1 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3 * internal-fn.h (lookup_internal_fn): Declare
4 * internal-fn.c (lookup_internal_fn): New function.
5 * gimple.c (gimple_build_call_from_tree): Handle calls to
7 * gimple-pretty-print.c (dump_gimple_call): Print "." before
8 internal function names.
9 * tree-pretty-print.c (dump_generic_node): Likewise.
10 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
12 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
14 * gimple-fold.h (gimple_build): Make the function forms take
15 combined_fn rather than built_in_function.
16 (gimple_simplify): Likewise.
17 * gimple-match-head.c (gimple_simplify): Likewise.
18 * gimple-fold.c (gimple_build): Likewise.
19 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
20 rather than gimple_build_call_internal.
21 (get_initial_defs_for_reduction): Likewise.
22 (vect_create_epilog_for_reduction): Likewise.
23 (vectorizable_live_operation): Likewise.
25 2018-05-17 Martin Liska <mliska@suse.cz>
27 * gimple-ssa-sprintf.c (format_directive): Do not use
28 space in between 'G_' and '('.
30 2018-05-17 Jakub Jelinek <jakub@redhat.com>
33 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
34 even if the mask is not all ones.
37 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
39 (ix86_gimple_fold_builtin): Likewise.
42 * config/i386/i386.c: Include tree-vector-builder.h.
43 (ix86_vector_shift_count): New function.
44 (ix86_fold_builtin): Fold shift builtins by scalar count.
45 (ix86_gimple_fold_builtin): Likewise.
47 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
48 _mm512_setzero): New intrinsics.
50 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
51 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
53 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
54 code generation for cases where splatting a value is not useful.
55 * simplify-rtx.c (simplify_ternary_operation): Simplify
56 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
59 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
61 * config.gcc: Support "goldmont-plus".
62 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
64 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
65 PROCESSOR_GOLDMONT_PLUS.
66 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
67 (processor_target_table): Add "goldmont-plus".
68 (PTA_GOLDMONT_PLUS): Define.
69 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
70 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
71 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
72 (fold_builtin_cpu): Add "goldmont-plus".
73 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
74 (ix86_option_override_internal): Add "goldmont-plus".
75 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
76 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
77 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
78 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
80 2018-05-17 Richard Biener <rguenther@suse.de>
82 PR tree-optimization/85757
83 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
84 remove defs that only feed that PHI from further processing.
86 2018-05-16 Jim Wilson <jimw@sifive.com>
88 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
90 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
92 2018-05-16 Mark Wielaard <mark@klomp.org>
94 * dwarf2out.c (count_index_strings): New function.
95 (output_indirect_strings): Call count_index_strings and generate
96 header for dwarf_version >= 5.
98 2018-05-16 Mark Wielaard <mark@klomp.org>
100 * dwarf2out.c (dwarf_FORM): New function.
101 (set_indirect_string): Use dwarf_FORM.
102 (reset_indirect_string): Likewise.
103 (size_of_die): Likewise.
104 (value_format): Likewise.
105 (output_die): Likewise.
106 (add_skeleton_AT_string): Likewise.
107 (output_macinfo_op): Likewise.
108 (index_string): Likewise.
109 (output_index_string_offset): Likewise.
110 (output_index_string): Likewise.
111 (count_index_strings): Likewise.
113 2018-05-16 Carl Love <cel@us.ibm.com>
115 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
116 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
118 2018-05-16 Martin Jambor <mjambor@suse.cz>
120 * ipa-prop.c (ipa_free_all_edge_args): Remove.
121 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
123 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
125 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
126 (fnma<mode>4): Likewise.
127 (fms<mode>4): Likewise.
128 (fnms<mode>4): Likewise.
129 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
130 (aarch64_fnma<mode>4): Likewise.
131 (aarch64_fms<mode>4): Likewise.
132 (aarch64_fnms<mode>4): Likewise.
133 (aarch64_fnmadd<mode>4): Likewise.
135 2018-05-16 Jason Merrill <jason@redhat.com>
137 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
139 2018-05-16 Richard Biener <rguenther@suse.de>
141 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
142 (dump_stmt_cost): Declare.
143 (add_stmt_cost): Dump cost we add.
144 (add_stmt_costs): New function.
145 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
147 (vect_analyze_stmt): Adjust prototype.
148 (vectorizable_condition): Likewise.
149 (vectorizable_live_operation): Likewise.
150 (vectorizable_reduction): Likewise.
151 (vectorizable_induction): Likewise.
152 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
153 cost vector to pass to vectorizable_ and record afterwards.
154 (vect_model_reduction_cost): Take cost vector argument and adjust.
155 (vect_model_induction_cost): Likewise.
156 (vectorizable_reduction): Likewise.
157 (vectorizable_induction): Likewise.
158 (vectorizable_live_operation): Likewise.
159 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
160 SLP_TREE_NUMBER_OF_VEC_STMTS.
161 (vect_analyze_slp_cost_1): Remove.
162 (vect_analyze_slp_cost): Likewise.
163 (vect_slp_analyze_node_operations): Take visited args and
164 a target cost vector. Avoid processing already visited stmt sets.
165 (vect_slp_analyze_operations): Use a local cost vector to gather
166 costs and register those of non-discarded instances.
167 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
168 (vect_schedule_slp_instance): Remove copying of
169 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
171 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
172 adding cost. Record cost entry location.
173 (vect_prologue_cost_for_slp_op): Function to compute cost of
174 a constant or invariant generated for SLP vect in the prologue,
175 split out from vect_analyze_slp_cost_1.
176 (vect_model_simple_cost): Make static. Adjust for SLP costing.
177 (vect_model_promotion_demotion_cost): Likewise.
178 (vect_model_store_cost): Likewise, make static.
179 (vect_model_load_cost): Likewise.
180 (vectorizable_bswap): Add cost vector arg and adjust.
181 (vectorizable_call): Likewise.
182 (vectorizable_simd_clone_call): Likewise.
183 (vectorizable_conversion): Likewise.
184 (vectorizable_assignment): Likewise.
185 (vectorizable_shift): Likewise.
186 (vectorizable_operation): Likewise.
187 (vectorizable_store): Likewise.
188 (vectorizable_load): Likewise.
189 (vectorizable_condition): Likewise.
190 (vectorizable_comparison): Likewise.
191 (can_vectorize_live_stmts): Likewise.
192 (vect_analyze_stmt): Likewise.
193 (vect_transform_stmt): Adjust calls to vectorizable_*.
194 * tree-vectorizer.c: Include gimple-pretty-print.h.
195 (dump_stmt_cost): New function.
197 2018-05-16 Richard Biener <rguenther@suse.de>
199 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
200 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
201 * tree-ssa-dse.c: Include tree-ssa-loop.h.
202 (check_name): New callback.
203 (dse_classify_store): Track cycles via a visited bitmap of PHI
204 defs and simplify handling of in-loop and across loop dead stores
205 and properly fail for loop-variant refs. Handle byte-tracking with
206 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
209 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
211 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
212 (vect_get_mask_type_for_stmt): Likewise.
213 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
215 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
216 to determine the statement's vector type and the vector type that
217 should be used for calculating nunits. Deal with cases in which
218 the type has to be deferred.
219 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
220 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
221 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
222 (vect_determine_vf_for_stmt): New functions, split out from...
223 (vect_determine_vectorization_factor): ...here.
224 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
225 (vect_get_mask_type_for_stmt): New functions, split out from
226 vect_determine_vectorization_factor.
228 2018-05-16 Richard Biener <rguenther@suse.de>
230 * tree-cfg.c (verify_gimple_assign_ternary): Properly
231 verify the [VEC_]COND_EXPR embedded comparison.
233 2018-05-15 Martin Sebor <msebor@redhat.com>
235 PR tree-optimization/85753
236 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
237 RECORD_TYPE in addition to ARRAY_TYPE.
239 2018-05-15 Martin Sebor <msebor@redhat.com>
242 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
244 2018-05-15 Richard Biener <rguenther@suse.de>
246 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
247 add by_clobber_p one. Change algorithm to collect all defs
248 representing uses we need to walk and try reducing them to
249 a single one before failing.
250 (dse_dom_walker::dse_optimize_stmt): Adjust.
252 2018-05-13 Mark Wielaard <mark@klomp.org>
254 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
255 (size_of_loc_descr): Likewise.
256 (output_loc_operands): Likewise.
257 (output_loc_operands_raw): Likewise.
258 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
259 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
260 (hash_loc_operands): Likewise.
261 (compare_loc_operands): Likewise.
263 2018-05-14 Mark Wielaard <mark@klomp.org>
265 * dwarf2out.c (count_index_addrs): New function.
266 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
268 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
270 PR tree-optimization/83648
271 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
272 return value as malloc candidate.
274 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
277 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
278 as true in call to suggest_attribute.
280 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
282 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
285 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
287 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
289 (aarch64_crypto_aesd_fused): Likewise.
291 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
293 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
294 (movsi_aarch64): Likewise.
295 (load_pairsi): Likewise.
296 (load_pairdi): Likewise.
297 (store_pairsi): Likewise.
298 (store_pairdi): Likewise.
299 (load_pairsf): Likewise.
300 (load_pairdf): Likewise.
301 (store_pairsf): Likewise.
302 (store_pairdf): Likewise.
303 (zero_extend): Likewise.
304 (trunc): Swap alternatives.
305 (fcvt_target): Add '?' to prefer w over r.
307 2018-05-14 Jakub Jelinek <jakub@redhat.com>
310 * config/i386/i386.md: Disallow non-commutative arithmetics in
311 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
312 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
313 in the peephole2 before it.
315 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
317 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
318 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
319 (ix86_handle_option): Handle -mcldemote.
320 * config.gcc: New header.
321 * config/i386/cldemoteintrin.h: New file.
322 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
323 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
325 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
326 OPTION_MASK_ISA_CLDEMOTE.
327 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
328 (ix86_valid_target_attribute_inner_p): Ditto.
329 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
330 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
331 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
332 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
333 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
335 * config/i386/i386.opt: Add -mcldemote.
336 * config/i386/x86intrin.h: New header.
337 * doc/invoke.texi: Add -mcldemote.
339 2018-05-14 Richard Biener <rguenther@suse.de>
341 * doc/match-and-simplify.texi: Adjust :s documentation.
343 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
345 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
346 intended memcpy size.
347 (REORDER_45): Likewise.
349 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
352 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
353 * vec.c (qsort_chk): Use gcc_qsort.
354 * Makefile.in (OBJS-libcommon): Add sort.o.
355 (build/sort.o): New target. Use it...
356 (BUILD_RTL): ... here, and...
357 (build/gencfn-macros): ... here, and...
358 (build/genmatch): ... here.
360 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
361 Chung-Ju Wu <jasonwucj@gmail.com>
363 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
364 * config/nds32/nds32-graywolf.md: New file.
365 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
366 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
368 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
369 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
370 * config/nds32/nds32.md (pipeline_model): Add graywolf.
371 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
372 * config/nds32/pipelines.md: Include n15 settings.
374 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
375 Chung-Ju Wu <jasonwucj@gmail.com>
377 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
378 * config/nds32/nds32-n13.md: New file.
379 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
380 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
382 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
383 * config/nds32/nds32.md (pipeline_model): Add n13.
384 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
385 * config/nds32/pipelines.md: Include n13 settings.
387 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
388 Chung-Ju Wu <jasonwucj@gmail.com>
390 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
391 * config/nds32/nds32-n10.md: New file.
392 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
393 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
395 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
396 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
397 * config/nds32/nds32.md (pipeline_model): Add n10.
398 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
399 * config/nds32/pipelines.md: Include n10 settings.
401 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
402 Kito Cheng <kito.cheng@gmail.com>
403 Chung-Ju Wu <jasonwucj@gmail.com>
405 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
406 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
407 Add enum values for DSP extension instructions.
408 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
410 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
411 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
413 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
414 * config/nds32/nds32-dspext.md: New file for DSP implementation.
415 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
416 * config/nds32/nds32-intrinsic.md: Likewise.
417 * config/nds32/nds32_intrinsic.h: Likewise.
418 * config/nds32/nds32-md-auxiliary.c: Likewise.
419 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
420 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
421 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
422 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
423 * config/nds32/nds32-protos.h: New declarations for DSP extension.
424 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
425 TYPE_DMAC in switch statement.
426 * config/nds32/nds32.c: New checking and implementation for DSP
427 extension instructions.
428 * config/nds32/nds32.h: Likewise.
429 * config/nds32/nds32.md: Likewise.
430 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
431 * config/nds32/predicates.md: Implement new predicates for DSP
434 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
436 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
437 Reformat alternatives and attributes so it is easier to identify
438 which constraints/attributes go with which instruction.
439 (mov<mode>_hardfloat32, FMOVE64): Likewise.
440 (mov<mode>_softfloat32, FMOVE64): Likewise.
441 (mov<mode>_hardfloat64, FMOVE64): Likewise.
442 (mov<mode>_softfloat64, FMOVE64): Likewise.
444 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
446 * doc/extend.texi (PowerPC Built-in Functions): Rename this
448 (Basic PowerPC Built-in Functions): The new name of the
449 subsection previously known as "PowerPC Built-in Functions".
450 (Basic PowerPC Built-in Functions Available on all Configurations):
452 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
453 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
454 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
455 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
457 2018-05-11 Martin Jambor <mjambor@suse.cz>
460 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
463 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
466 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
468 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
470 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
471 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
472 (ix86_handle_option): Handle -mwaitpkg.
473 * config.gcc: New header.
474 * config/i386/cpuid.h (bit_WAITPKG): New bit.
475 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
476 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
478 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
479 OPTION_MASK_ISA_WAITPKG.
480 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
481 (ix86_option_override_internal): Add PTA_WAITPKG.
482 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
483 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
485 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
486 __builtin_ia32_umwait and __builtin_ia32_tpause.
487 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
488 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
489 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
490 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
491 UNSPECV_TPAUSE): New.
492 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
493 * config/i386/i386.opt: Add -mwaitpkg.
494 * config/i386/waitpkgintrin.h: New file.
495 * config/i386/x86intrin.h: New header.
496 * doc/invoke.texi: Add -mwaitpkg.
498 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
501 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
503 (cortex-m0): Use armv6s-m isa.
504 (cortex-m0plus): Likewise.
505 (cortex-m1): Likewise.
506 (cortex-m0.small-multiply): Likewise.
507 (cortex-m0plus.small-multiply): Likewise.
508 (cortex-m1.small-multiply): Likewise.
510 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
511 Jakub Jelinek <jakub@redhat.com>
513 PR tree-optimization/85692
514 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
515 source permute as well.
517 2018-05-11 Martin Liska <mliska@suse.cz>
520 * doc/extend.texi: Document LLVM style format for no_sanitize
523 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
525 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
526 mode_supports_vsx_dform_quad to mode_supports_dq_form.
527 (mode_supports_vsx_dform_quad): Likewise.
528 (mode_supports_vmx_dform): Move these functions to be next to the
529 other mode_supports functions.
530 (mode_supports_dq_form): Likewise.
531 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
532 mode_supports_dq_form.
533 (reg_offset_addressing_ok_p): Likewise.
534 (offsettable_ok_by_alignment): Likewise.
535 (rs6000_legitimate_offset_address_p): Likewise.
536 (legitimate_lo_sum_address_p): Likewise.
537 (rs6000_legitimize_address): Likewise.
538 (rs6000_legitimize_reload_address): Likewise.
539 (rs6000_secondary_reload_inner): Likewise.
540 (rs6000_preferred_reload_class): Likewise.
541 (rs6000_output_move_128bit): Likewise.
543 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
545 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
546 Generate SImode target register for null target.
547 <case IX86_BUILTIN_XGETBV>: Ditto.
548 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
549 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
551 2018-05-10 Carl Love <cel@us.ibm.com>
553 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
554 dcbtt and dcbtstt if operands[2] is 0.
556 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
559 * config/i386/sse.md (usadv64qi): New expander.
561 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
563 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
564 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
565 -maltivec=be support.
566 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
567 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
568 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
569 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
570 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
571 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
572 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
573 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
574 altivec_vsumsws): Adjust.
575 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
576 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
577 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
579 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
580 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
581 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
582 (altivec_lve<VI_char>x): Delete expand.
583 (*altivec_lve<VI_char>x_internal): Rename to...
584 (altivec_lve<VI_char>x): ... this.
585 (altivec_lvxl_<mode>): Delete expand.
586 (*altivec_lvxl_<mode>_internal): Rename to ...
587 (altivec_lvxl_<mode>): ... this.
588 (altivec_stvxl_<mode>): Delete expand.
589 (*altivec_stvxl_<mode>_internal): Rename to ...
590 (altivec_stvxl_<mode>): ... this.
591 (altivec_stve<VI_char>x): Delete expand.
592 (*altivec_stve<VI_char>x_internal): Rename to ...
593 (altivec_stve<VI_char>x): ... this.
594 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
595 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
596 reduc_plus_scal_<mode>): Adjust.
597 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
599 (rs6000_cpu_cpp_builtins): Adjust.
600 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
601 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
602 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
603 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
604 -maltivec=be support.
605 (rs6000_split_vec_extract_var): Adjust.
606 (rs6000_split_v4si_init): Adjust.
607 (swap_selector_for_mode): Delete.
608 (altivec_expand_lvx_be, altivec_expand_stvx_be,
609 altivec_expand_stvex_be): Delete.
610 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
611 -maltivec=be support.
612 (rs6000_gimple_fold_builtin): Ditto.
613 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
615 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
616 (TARGET_DIRECT_MOVE_64BIT): Adjust.
617 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
618 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
619 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
620 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
621 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
622 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
623 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
624 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
625 anonymous split): Adjust.
626 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
627 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
629 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
631 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
632 when --with-gxx-include-dir is also specified.
633 * configure: Regenerate.
635 2018-05-09 Jim Wilson <jimw@sifive.com>
638 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
639 * config/riscv/t-withmultilib: New.
640 * config/riscv/withmultilib.h: New.
641 * doc/install.texi: Document RISC-V --with-multilib-list support.
643 2018-05-09 Richard Biener <rguenther@suse.de>
645 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
647 (vect_bb_vectorization_profitable_p): Adjust. Compute
648 actual scalar cost using the cost vector and the add_stmt_cost
651 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
653 PR rtl-optimization/85645
654 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
655 in the REG_CFA_REGISTER note for LR, don't leave it empty.
657 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
659 PR rtl-optimization/85645
660 * shrink-wrap.c (spread_components): Return a boolean saying if
661 anything was changed.
662 (try_shrink_wrapping_separate): Iterate spread_components until
663 nothing changes anymore.
665 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
667 PR rtl-optimization/85645
668 * regrename.c (build_def_use): Also kill the chains that include the
669 destination of a REG_CFA_REGISTER note.
671 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
673 PR rtl-optimization/85645
674 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
675 insn that has a REG_CFA_REGISTER note.
677 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
679 * cfgexpand.c (expand_clobber): New function.
680 (expand_gimple_stmt_1): Use it.
681 * tree-vect-stmts.c (vect_clobber_variable): New function,
683 (vectorizable_simd_clone_call): ...here.
684 (vectorizable_store): Emit a clobber either side of an
685 IFN_STORE_LANES sequence.
686 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
688 2018-05-09 Tom de Vries <tom@codesourcery.com>
691 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
692 (define_insn "trap_if_false"): Add exit after trap.
694 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
696 PR rtl-optimization/85638
697 * bb-reorder.c: Include common/common-target.h.
698 (create_forwarder_block): New function extracted from...
699 (fix_up_crossing_landing_pad): ...here. Rename into...
700 (dw2_fix_up_crossing_landing_pad): ...this.
701 (sjlj_fix_up_crossing_landing_pad): New function.
702 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
703 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
704 from both partitions and exit the loop after one iteration.
706 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
709 * doc/extend.texi (PowerPC Built-in Functions): Rename this
711 (Basic PowerPC Built-in Functions): The new name of the
712 subsection previously known as "PowerPC Built-in Functions".
713 (Basic PowerPC Built-in Functions Available on all Configurations):
715 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
717 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
719 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
721 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
724 2018-05-08 Jim Wilson <jimw@sifive.com>
726 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
727 (LD_EMUL_SUFFIX): New.
730 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
732 * doc/extend.texi (PowerPC Built-in Functions): Rename this
734 (Basic PowerPC Built-in Functions): The new name of the
735 subsection previously known as "PowerPC Built-in Functions".
736 (Basic PowerPC Built-in Functions Available on all Configurations):
738 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
740 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
742 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
744 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
747 2018-05-08 Jakub Jelinek <jakub@redhat.com>
750 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
751 after cmpelim optimization.
753 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
755 * config.gcc: Support "goldmont".
756 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
757 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
759 * config/i386/i386.c (m_GOLDMONT): Define.
760 (processor_target_table): Add "goldmont".
761 (PTA_GOLDMONT): Define.
762 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
763 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
764 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
765 (fold_builtin_cpu): Add "goldmont".
766 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
767 (ix86_option_override_internal): Add "goldmont".
768 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
769 (processor_type): Add PROCESSOR_GOLDMONT.
770 * config/i386/i386.md: Add CPU "glm".
771 * config/i386/glm.md: New file.
772 * config/i386/x86-tune.def: Add m_GOLDMONT.
773 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
775 2018-05-08 Jakub Jelinek <jakub@redhat.com>
778 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
780 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
781 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
782 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
785 * config/i386/i386.c (ix86_fold_builtin): Handle
786 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
789 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
790 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
792 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
795 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
796 (check_arch): Likewise.
797 (check_fpu): Return the result rather than printing it.
798 (end arch): Fix operator precedence.
800 (END): Print the result from check_fpu.
802 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
803 Alan Hayward <alan.hayward@arm.com>
804 David Sherwood <david.sherwood@arm.com>
806 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
807 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
808 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
809 (*fcmuo<mode>_and): New patterns.
811 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
813 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
814 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
815 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
816 (cmp_op, sve_imm_con): New code attributes.
817 (SVE_COND_INT_CMP, imm_con): Delete.
818 (cmp_op): Remove above unspecs from int attribute.
819 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
821 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
822 comparison-specific unspecs.
823 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
824 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
825 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
826 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
827 (*vec_fcm<cmp_op><mode>): Rename to...
828 (*fcm<cmp_op><mode>): ...this and adjust likewise.
829 (*vec_fcmuo<mode>): Rename to...
830 (*fcmuo<mode>): ...this and adjust likewise.
831 (*pred_fcm<cmp_op><mode>): New pattern.
832 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
833 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
835 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
837 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
838 (aarch64_emit_sve_predicated_cond): New function.
839 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
840 (aarch64_emit_unspec_cond_or): Replace with...
841 (aarch64_emit_sve_or_conds): ...this new function. Use
842 aarch64_emit_sve_ptrue_op for the individual comparisons and
843 aarch64_emit_binop to OR them together.
844 (aarch64_emit_inverted_unspec_cond): Replace with...
845 (aarch64_emit_sve_inverted_cond): ...this new function. Use
846 aarch64_emit_sve_ptrue_op for the comparison and
847 aarch64_emit_unop to invert the result.
848 (aarch64_expand_sve_vec_cmp_float): Update after the above
849 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
851 2018-05-07 Nathan Sidwell <nathan@acm.org>
853 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
854 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
855 (Backwards Compatibility): Likewise.
857 2018-05-07 Luis Machado <luis.machado@linaro.org>
861 2018-05-07 Luis Machado <luis.machado@linaro.org>
863 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
864 <prefetch_dynamic_strides>: New const bool field.
865 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
866 prefetch_dynamic_strides.
867 (exynosm1_prefetch_tune): Likewise.
868 (thunderxt88_prefetch_tune): Likewise.
869 (thunderx_prefetch_tune): Likewise.
870 (thunderx2t99_prefetch_tune): Likewise.
871 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
872 (aarch64_override_options_internal): Update to set
873 PARAM_PREFETCH_DYNAMIC_STRIDES.
874 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
875 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
876 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
877 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
878 prefetch-dynamic-strides setting.
880 2018-05-07 Luis Machado <luis.machado@linaro.org>
882 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
883 <minimum_stride>: New const int field.
884 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
885 minimum_stride field.
886 (exynosm1_prefetch_tune): Likewise.
887 (thunderxt88_prefetch_tune): Likewise.
888 (thunderx_prefetch_tune): Likewise.
889 (thunderx2t99_prefetch_tune): Likewise.
890 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
891 (aarch64_override_options_internal): Update to set
892 PARAM_PREFETCH_MINIMUM_STRIDE.
893 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
894 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
895 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
896 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
897 stride is constant and is below the minimum stride threshold.
899 2018-05-07 Luis Machado <luis.machado@linaro.org>
901 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
904 2018-05-07 Luis Machado <luis.machado@linaro.org>
906 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
907 <prefetch_dynamic_strides>: New const bool field.
908 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
909 prefetch_dynamic_strides.
910 (exynosm1_prefetch_tune): Likewise.
911 (thunderxt88_prefetch_tune): Likewise.
912 (thunderx_prefetch_tune): Likewise.
913 (thunderx2t99_prefetch_tune): Likewise.
914 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
915 (aarch64_override_options_internal): Update to set
916 PARAM_PREFETCH_DYNAMIC_STRIDES.
917 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
918 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
919 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
920 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
921 prefetch-dynamic-strides setting.
923 2018-05-07 Luis Machado <luis.machado@linaro.org>
925 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
926 <minimum_stride>: New const int field.
927 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
928 minimum_stride field.
929 (exynosm1_prefetch_tune): Likewise.
930 (thunderxt88_prefetch_tune): Likewise.
931 (thunderx_prefetch_tune): Likewise.
932 (thunderx2t99_prefetch_tune): Likewise.
933 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
934 (aarch64_override_options_internal): Update to set
935 PARAM_PREFETCH_MINIMUM_STRIDE.
936 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
937 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
938 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
939 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
940 stride is constant and is below the minimum stride threshold.
942 2018-05-06 Jakub Jelinek <jakub@redhat.com>
945 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
946 the type is addressable. Don't force op into register if it has
949 2018-05-05 Roland McGrath <mcgrathr@google.com>
952 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
953 any section for which we don't know a specific type it should have,
954 regardless of name. Previously this was done only for the exact
955 names ".init_array", ".fini_array", and ".preinit_array".
956 (default_elf_asm_named_section): Add comment about
957 relationship with default_section_type_flags and SECTION_NOTYPE.
958 (get_section): Don't consider it a type conflict if one side has
959 SECTION_NOTYPE and the other doesn't, as long as neither has the
960 SECTION_BSS et al used in the default_section_type_flags logic.
962 2018-05-05 Tom de Vries <tom@codesourcery.com>
965 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
966 (workaround_barsyncs): New function.
967 (nvptx_reorg): Use workaround_barsyncs.
968 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
969 (define_expand "nvptx_membar_cta"): New define_expand.
970 (define_insn "*nvptx_membar_cta"): New insn.
972 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
974 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
975 To improve optimization opportunities.
976 * builtin-types.def: The new needed builtin types for the above.
978 2018-05-04 Richard Biener <rguenther@suse.de>
980 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
981 * gimple-ssa-store-merging.c
982 (imm_store_chain_info::output_merged_store): Remove redundant create,
983 release split_store vector contents on failure.
984 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
985 scalar stmt vector on cache hit.
987 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
989 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
991 * config.gcc (powerpc-xilinx-eabi*): Remove.
992 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
994 (fusion_addis_mem_combo_load): Ditto.
995 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
997 (rs6000_cpu_cpp_builtins): Ditto.
998 * config/rs6000/rs6000-linux.c
999 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1000 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1001 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1003 (rs6000_setup_reg_addr_masks): Ditto.
1004 (rs6000_init_hard_regno_mode_ok): Ditto.
1005 (rs6000_option_override_internal): Ditto.
1006 (legitimate_lo_sum_address_p): Ditto.
1007 (rs6000_legitimize_address): Ditto.
1008 (rs6000_legitimize_reload_address): Ditto.
1009 (rs6000_legitimate_address_p): Ditto.
1010 (abi_v4_pass_in_fpr): Ditto.
1011 (setup_incoming_varargs): Ditto.
1012 (rs6000_gimplify_va_arg): Ditto.
1013 (rs6000_split_multireg_move): Ditto.
1014 (rs6000_savres_strategy): Ditto.
1015 (rs6000_emit_prologue_components): Ditto.
1016 (rs6000_emit_epilogue_components): Ditto.
1017 (rs6000_emit_prologue): Ditto.
1018 (rs6000_emit_epilogue): Ditto.
1019 (rs6000_elf_file_end): Ditto.
1020 (rs6000_function_value): Ditto.
1021 (rs6000_libcall_value): Ditto.
1022 * config/rs6000/rs6000.h: Ditto.
1023 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1024 (TARGET_MINMAX): ... this. New.
1025 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1026 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1027 (*movsi_internal1_single): Delete.
1028 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1029 mfpu=, mxilinx-fpu): Delete.
1030 * config/rs6000/singlefp.h: Delete.
1031 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1032 * config/rs6000/t-rs6000: Ditto.
1033 * config/rs6000/t-xilinx: Delete.
1034 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1035 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1036 (VStype_simple): Delete.
1037 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1038 * config/rs6000/xfpu.h: Delete.
1039 * config/rs6000/xfpu.md: Delete.
1040 * config/rs6000/xilinx.h: Delete.
1041 * config/rs6000/xilinx.opt: Delete.
1042 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1043 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1045 2018-05-04 Tom de Vries <tom@codesourcery.com>
1048 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1051 2018-05-04 Richard Biener <rguenther@suse.de>
1054 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1055 (expand_complex_div_wide): Likewise.
1056 (expand_complex_operations_1): Likewise.
1057 (expand_complex_libcall): Preserve EH info of the original stmt.
1058 (tree_lower_complex): Handle removed blocks.
1059 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1060 on complex multiplication and division libcall builtins.
1062 2018-05-04 Richard Biener <rguenther@suse.de>
1065 * fold-const.c (negate_expr_p): Restrict negation of operand
1066 zero of a division to when we know that can happen without
1068 (fold_negate_expr_1): Likewise.
1070 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1073 * real.h (real_nextafter): Declare.
1074 * real.c (real_nextafter): New function.
1075 * fold-const-call.c (fold_const_nextafter): New function.
1076 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1077 CASE_CFN_NEXTTOWARD.
1078 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1079 even when arg1_mode is different from arg0_mode.
1081 2018-05-03 Nathan Sidwell <nathan@acm.org>
1083 * doc/extend.texi (Deprecated Features): Remove
1085 (Backwards Compatibility): Likewise.
1086 * doc/invoke.texi (C++ Language Options): Likewise.
1087 (C++ Dialect Options): Likewise.
1089 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1092 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1093 _mm512_mask_mullox_epi64): New intrinsics.
1095 2018-05-03 Tom de Vries <tom@codesourcery.com>
1098 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1099 dump files): Add offload-tree.
1101 2018-05-03 Richard Biener <rguenther@suse.de>
1103 PR tree-optimization/85615
1104 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1105 to loops not nested in BBs loop father to avoid creating multi-entry
1108 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1110 PR tree-optimization/70291
1111 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1112 arguments. Change return type to tree. Emit libcall as a new
1113 statement rather than replacing existing one when inplace_p is true.
1114 (expand_complex_multiplication_components): New function.
1115 (expand_complex_multiplication): Expand floating-point complex
1116 multiplication using the above.
1117 (expand_complex_division): Rename inner_type parameter to type.
1118 Update expand_complex_libcall call-site.
1119 (expand_complex_operations_1): Update expand_complex_multiplication
1120 and expand_complex_division call-sites.
1122 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1125 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1126 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1127 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1128 the highest significant bit of the shift count mask is clear. In
1129 check whether and[sq]i3 is needed verify that all significant bits
1130 of the shift count other than the highest are set.
1132 2018-05-02 Tom de Vries <tom@codesourcery.com>
1135 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1136 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1137 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1138 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1139 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1140 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1141 __builtin_goacc_parlevel_size.
1143 2018-05-02 Richard Biener <rguenther@suse.de>
1145 PR tree-optimization/85597
1146 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1147 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1150 2018-05-02 Tom de Vries <tom@codesourcery.com>
1153 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1154 dump files): Add ltrans-tree.
1156 2018-05-02 Tom de Vries <tom@codesourcery.com>
1159 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1160 dump files): Add wpa-ipa.
1162 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1164 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1165 powerpc*-*-linux*paired* target.
1166 * config/rs6000/750cl.h: Delete.
1167 * config/rs6000/paired.h: Delete.
1168 * config/rs6000/paired.md: Delete.
1169 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1171 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1172 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1173 comment. Remove paired float support.
1174 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1175 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1177 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1178 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1180 * config/rs6000/rs6000.c: Remove paired float support.
1181 (paired_expand_vector_init, paired_expand_vector_move,
1182 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1183 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1184 paired_expand_builtin, paired_expand_predicate_builtin,
1185 paired_init_builtins): Delete.
1186 * config/rs6000/rs6000.h: Remove paired float support.
1187 * config/rs6000/rs6000.md: Remove paired float support.
1188 (move_from_CR_ov_bit): Delete.
1189 * config/rs6000/rs6000.opt (mpaired): Delete.
1190 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1191 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1193 2018-05-02 Richard Biener <rguenther@suse.de>
1196 * gimplify.c (gimplify_save_expr): When in SSA form allow
1197 SAVE_EXPRs to compute to SSA vars.
1199 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1202 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1203 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1204 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1205 clobber operands[2], instead use a new pseudo. Formatting fixes.
1207 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1209 PR tree-optimization/85586
1210 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1211 exit early for statements in the same group if the accesses are
1214 2018-05-02 Tom de Vries <tom@codesourcery.com>
1217 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1220 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1222 PR tree-optimization/85143
1223 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1225 2018-05-01 Tom de Vries <tom@codesourcery.com>
1228 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1229 not found" error message.
1231 2018-05-01 Tom de Vries <tom@codesourcery.com>
1234 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1235 * vec.c (test_ordered_remove_if): New function.
1236 (vec_c_tests): Call test_ordered_remove_if.
1237 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1238 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1239 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1240 VEC_ORDERED_REMOVE_IF.
1242 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1244 PR tree-optimization/82665
1245 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1246 pointer subtraction where arguments come from a memchr call.
1248 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1250 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1251 --push-state --as-needed and --pop-state instead of --as-needed and
1252 --no-as-needed if ld supports it.
1253 * configure: Regenerated.
1256 * doc/install.texi2html: Replace _002d with - and _002a with * in
1257 generated html files using sed.
1259 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1262 * gcc-rich-location.c (blank_line_before_p): New function.
1263 (use_new_line): New function.
1264 (gcc_rich_location::add_fixit_insert_formatted): New function.
1265 * gcc-rich-location.h
1266 (gcc_rich_location::add_fixit_insert_formatted): New function.
1268 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1270 * selftest.c (assert_streq): Rename "expected" and "actual" to
1271 "val1" and "val2". Extend NULL-handling to cover both inputs
1272 symmetrically, while still requiring both to be non-NULL for a pass.
1273 * selftest.h (assert_streq): Rename "expected" and "actual" to
1275 (ASSERT_EQ): Likewise.
1276 (ASSERT_EQ_AT): Likewise.
1277 (ASSERT_KNOWN_EQ): Likewise.
1278 (ASSERT_KNOWN_EQ_AT): Likewise.
1279 (ASSERT_NE): Likewise.
1280 (ASSERT_MAYBE_NE): Likewise.
1281 (ASSERT_MAYBE_NE_AT): Likewise.
1282 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1283 the assertion to pass.
1284 (ASSERT_STREQ_AT): Likewise.
1286 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1288 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1289 interaction with -pie.
1291 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1293 * selftest.h: Fix alphabetization of per-source-file selftest
1296 2018-04-30 Jason Merrill <jason@redhat.com>
1298 PR c++/61982 - dead stores to destroyed objects.
1299 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1302 2018-04-30 Jason Merrill <jason@redhat.com>
1304 * tree.c (build_clobber): New.
1305 * tree.h: Declare it.
1306 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1308 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1310 * diagnostic-show-locus.c (layout::layout): Update for
1311 location_get_source_line returning a char_span.
1312 (struct char_span): Move to input.h.
1313 (struct correction): Update for fields in char_span becoming
1315 (struct source_line): Update for location_get_source_line
1316 returning a char_span.
1317 (layout::print_line): Likewise.
1318 * edit-context.c (edited_file::print_content): Likewise.
1319 (edited_file::print_diff_hunk): Likewise.
1320 (edited_file::print_run_of_changed_lines): Likewise.
1321 (edited_file::get_num_lines): Likewise.
1322 (edited_line::edited_line): Likewise.
1323 * final.c (asm_show_source): Likewise.
1324 * input.c (location_get_source_line): Convert return type
1325 from const char * to char_span, losing the final "line_len"
1327 (dump_location_info): Update for the above.
1328 (get_substring_ranges_for_loc): Likewise. Use a char_span
1329 when handling the literal within the line.
1330 (test_reading_source_line): Update for location_get_source_line
1331 returning a char_span.
1332 * input.h (class char_span): Move here from
1333 diagnostic-show-locus.c, converting from a struct to a class.
1334 Make data members private.
1335 (char_span::operator bool): New.
1336 (char_span::length): New.
1337 (char_span::get_buffer): New.
1338 (char_span::operator[]): New.
1339 (char_span::subspan): Make const.
1340 (char_span::xstrdup): New.
1341 (location_get_source_line): Convert return type from const char *
1342 to char_span, losing the final "line_size" param.
1344 2018-04-30 Jan Hubicka <jh@suse.cz>
1346 * lto-wrapper.c (ltrans_priorities): New static var.
1347 (cmp_priority): New.
1348 (run_gcc): Read priorities and if doing parallel build order
1349 the Makefile by them.
1351 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1353 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1355 2018-04-30 Richard Biener <rguenther@suse.de>
1357 * tree-cfg.c (verify_address): Remove base argument, add
1358 flag whether to check TREE_ADDRESSABLE and do that.
1359 (verify_expr): Remove.
1360 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1361 (verify_gimple_assign_single): Likewise.
1362 (verify_gimple_switch): Likewise.
1363 (verify_expr_location_1): Dereference tp once. Add (disabled)
1364 piece from verify_expr.
1365 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1367 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1369 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1371 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1373 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1374 (small_data_pattern): Likewise.
1375 (arc_rewrite_small_data): Likewise.
1376 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1377 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1378 (get_symbol_alignment): New function.
1379 (legitimate_small_data_address_p): Likewise.
1380 (legitimate_scaled_address): Update, call
1381 legitimate_small_data_address_p.
1382 (output_sdata): New static variable.
1383 (arc_print_operand): Update how we handle small data operands.
1384 (arc_print_operand_address): Likewise.
1385 (arc_legitimate_address_p): Update, use
1386 legitimate_small_data_address_p.
1387 (arc_rewrite_small_data_p): Remove.
1388 (arc_rewrite_small_data_1): Likewise.
1389 (arc_rewrite_small_data): Likewise.
1390 (small_data_pattern): Likewise.
1391 (compact_sda_memory_operand): Update to use
1392 legitimate_small_data_address_p and get_symbol_alignment.
1393 (prepare_move_operands): Don't rewite sdata pattern.
1394 (prepare_extend_operands): Remove.
1395 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1397 (zero_extendqisi2): Likewise.
1398 (zero_extendhisi2): Likewise.
1399 (extendqihi2): Likewise.
1400 (extendqisi2): Likewise.
1401 (extendhisi2): Likewise.
1405 * config/arc/constraints.md (Usd): Change it to memory constraint.
1407 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1409 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1410 as source of std instructions.
1411 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1412 allow 6-bit constants as source for store instructions.
1413 (movdi_insn): Update instruction pattern to allow 6-bit constants
1414 as source for store instructions.
1416 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1418 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1420 2018-04-30 Nathan Sidwell <nathan@acm.org>
1421 Sandra Loosemore <sandra@codesourcery.com>
1423 * dumpfile.c (dump_open): Allow '-' for stdout.
1424 * doc/invoke.texi (Developer Options): Document dump filename
1425 determination early. Document stdin/stdout selection.
1427 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1429 Microblaze Target: PIC data text relative
1431 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1432 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1434 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1435 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1436 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1437 New addressing mode for data-text relative position indepenedent code.
1438 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1439 'ADDRESS_SYMBOLIC_TXT_REL'.
1440 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1441 (microblaze_legitimate_pic_operand): Exclude function calls from
1442 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1443 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1445 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1446 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1447 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1448 for 'address + offset'.
1449 (microblaze_expand_prologue): Add new function prologue call for
1451 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1452 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1453 table in case of TARGET_PIC_DATA_TEXT_REL.
1454 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1455 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1456 Add new macros 'UNSPEC_TEXT',
1457 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1458 + exclude function calls from 'UNSPEC_PLT' in case of data text
1460 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1461 new target hook for generating address diff vector tables in case of
1463 * doc/tm.texi : Regenerate.
1464 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1465 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1466 of addr diff vector generation.
1467 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1468 target hook definition.
1469 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1470 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1471 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1472 option for data text relative.
1474 2018-04-30 Richard Biener <rguenther@suse.de>
1476 * tree-chrec.h (evolution_function_is_constant_p): Remove
1478 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1480 2018-04-30 Richard Biener <rguenther@suse.de>
1483 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1485 2018-04-30 Richard Biener <rguenther@suse.de>
1487 PR tree-optimization/28364
1488 PR tree-optimization/85275
1489 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1490 copying first exit test.
1492 2018-04-28 Mark Wielaard <mark@klomp.org>
1494 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1496 (dwarf_AT): Handle DW_AT_addr_base.
1497 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1499 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1502 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1503 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1504 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1505 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1507 2018-04-28 Richard Biener <rguenther@suse.de>
1509 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1510 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1511 to reflect use. Only add interesting stmts.
1513 2018-04-27 Martin Jambor <mjambor@suse.cz>
1516 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1517 the jump function allows for passing through aggregate values.
1519 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1521 * input.h (in_system_header_at): Convert from macro to inline
1523 (from_macro_expansion_at): Likewise.
1524 (from_macro_definition_at): Likewise.
1526 2018-04-27 Jeff Law <law@redhat.com>
1528 * config.gcc: Mark tile* targets as deprecated/obsolete.
1530 2018-04-27 Richard Biener <rguenther@suse.de>
1532 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1535 2018-04-27 Richard Biener <rguenther@suse.de>
1537 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1539 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1541 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1542 with Yd constraint. Set "preferred_for_speed" attribute from
1543 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1545 (*movdi_internal): Ditto.
1546 (movti_interunit splitters): Remove
1547 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1548 (movdi_interunit splitters): Ditto.
1549 * config/i386/constraints.md (Ye): Remove.
1550 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1552 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1555 * config/aarch64/constraints.md (Usg): Limit to 31.
1558 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1560 PR tree-optimization/85529
1561 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1562 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1563 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1564 zero extension or masking of the MSB bit.
1565 (optimize_range_tests): Add FIRST_BB argument, pass it through
1566 to optimize_range_tests_var_bound.
1567 (maybe_optimize_range_tests, reassociate_bb): Adjust
1568 optimize_range_tests callers.
1570 2018-04-26 Richard Biener <rguenther@suse.de>
1571 Jakub Jelinek <jakub@redhat.com>
1573 * cgraph.h (symbol_table): Just declare debug method here.
1574 * symtab.c (symbol_table::debug): Define.
1576 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1578 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1580 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1582 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1583 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1584 (*movdi_internal): Substitute Yi and Yj constraint with x
1585 and Ym and Yn constraint with y constraint. Update "isa"
1586 attribute and set "preferred_for_speed" attribute from
1587 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1588 (*movsi_internal): Ditto.
1589 (*movdf_internal): Ditto.
1590 (*movsf_internal): Ditto.
1591 (*zero_extendsidi2): Ditto.
1592 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1593 (sse2_loadld): Ditto.
1594 (*vec_extract<ssevecmodelower>_0): Ditto.
1595 (*vec_extractv4si_0_zext_sse4): Ditto.
1596 (vec_concatv2di): Ditto.
1597 (*vec_dup<mode>): Ditto.
1598 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1599 * config/i386/constraints.md (Yi): Remove.
1604 2018-04-26 Nathan Sidwell <nathan@acm.org>
1606 * dumpfile.c (dump_open): New.
1607 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1608 (dump_finish): Detect stdio/stderr by value not name.
1610 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1612 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1614 2018-04-26 Tom de Vries <tom@codesourcery.com>
1617 * config/nvptx/nvptx.c (verify_neutering_jumps)
1618 (verify_neutering_labels): New function
1619 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1621 2018-04-26 Tom de Vries <tom@codesourcery.com>
1624 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1625 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1628 2018-04-26 Richard Biener <rguenther@suse.de>
1629 Tom de Vries <tom@codesourcery.com>
1632 * lto-streamer-out.c (output_function): Fixup loops if required to match
1633 discovery done in the reader.
1635 2018-04-26 Richard Biener <rguenther@suse.de>
1637 PR tree-optimization/85116
1638 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1639 have a loop exit from the single latch predecessor. Remove
1640 case of header with just condition.
1641 (ch_base::copy_headers): Exclude infinite loops from any
1643 (pass_ch::execute): Record exits.
1645 2018-04-26 Richard Biener <rguenther@suse.de>
1647 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1648 prologue cost vector and pass it to vect_get_load_cost.
1649 (vect_get_peeling_costs_all_drs): Likewise.
1650 (vect_peeling_hash_get_lowest_cost): Likewise.
1651 (vect_enhance_data_refs_alignment): Likewise.
1653 2018-04-26 Richard Biener <rguenther@suse.de>
1656 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1657 checking of integer<->pointer conversions.
1658 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1659 sign-/zero-extending pointer types.
1660 (expand_omp_for_static_chunk): Likewise.
1662 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1663 Jean Lee <xiaoyur347@gmail.com>
1665 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1666 (TARGET_ASAN_SHADOW_OFFSET): Define.
1667 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1668 true for -fsanitize=address.
1670 2018-04-25 Mark Wielaard <mark@klomp.org>
1672 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1675 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1677 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1678 than "alu", remove explicit "memory" and "imm_disp" attributes.
1679 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1682 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1683 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1684 gen_lowpart_no_emit.
1686 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1689 * config/i386/i386.c (ix86_expand_builtin): Change memory
1690 operand to XI, extend p0 to Pmode.
1691 * config/i386/i386.md: Change unspec volatile and operand
1692 1 mode to XI, change operand 0 mode to P.
1694 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1696 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1697 GET_MODE_MASK before any checking.
1698 (nds32_can_use_bset_p): Likewise.
1699 (nds32_can_use_btgl_p): Likewise.
1701 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1703 * config/nds32/nds32-doubleword.md: New define_split pattern for
1704 illegal register number.
1706 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1708 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1710 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1712 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1714 2018-04-25 Richard Biener <rguenther@suse.de>
1716 * lto-streamer.h (LTO_major_version): Bump to 8.
1718 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1720 * BASE-VER: Set to 9.0.0.
1722 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1724 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1725 in __abskf2 and __powikf2.
1727 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1730 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1731 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1732 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1733 Use the above on operand 2. Reindent.
1734 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1736 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1739 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1741 * config/i386/i386.opt (mcet): Removed.
1742 * doc/install.texi: Remove -mcet documentation.
1743 * doc/invoke.texi: Likewise.
1745 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1748 * doc/install.texi: Remove -mcet from bootstrap-cet.
1750 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1753 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1754 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1758 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1759 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1760 containing a CONST_VECTOR.
1762 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1764 * doc/install.texi: Update newlib dependency for nvptx.
1766 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1769 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1770 instead of INTVAL when shifting x left.
1772 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1774 PR tree-optimization/85478
1775 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1776 vect_grouped_store_supported for single element vectors.
1778 2018-04-24 Richard Biener <rguenther@suse.de>
1781 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1782 load cost increase to the case of non-constant step.
1784 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1787 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1788 destination if any_malformed_asm.
1790 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1793 * expr.c (store_field): In the bitfield case, if the value comes from
1794 a function call and is returned in registers by means of a PARALLEL,
1795 do not change the mode of the temporary unless BLKmode and VOIDmode.
1797 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1799 PR rtl-optimization/85423
1800 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1801 dependencies to debug insns when the previous insn is non-debug.
1803 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1805 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1806 enums into a single definition.
1807 (fls): Fix predicates and printing.
1810 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1812 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1813 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1814 and short u6 immediate.
1815 (check_if_valid_sleep_operand): Remove.
1816 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1818 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1820 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1821 flag_always_save_lp condition.
1822 * config/nds32/nds32.opt (malways-save-lp): New option.
1824 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1826 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1827 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1828 * config/nds32/nds32.h
1829 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1830 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1832 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1834 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1835 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1837 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1838 Chung-Ju Wu <jasonwucj@gmail.com>
1840 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1842 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1843 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1845 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1847 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1849 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1851 * config/nds32/nds32-protos.h (nds32_data_alignment,
1852 nds32_local_alignment): Declare.
1853 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1854 nds32_local_alignment): New functions.
1855 (TARGET_CONSTANT_ALIGNMENT): Define.
1856 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1858 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1860 * config/nds32/nds32.c
1861 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1862 (TARGET_MODES_TIEABLE_P): Likewise.
1864 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1866 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1869 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1870 Chung-Ju Wu <jasonwucj@gmail.com>
1872 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1873 for unaligned access.
1874 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1876 * config/nds32/nds32-intrinsic.md: Likewise.
1877 * config/nds32/nds32_intrinsic.h: Likewise.
1878 * config/nds32/nds32.h (nds32_builtins): Likewise.
1879 * config/nds32/nds32.opt (munaligned-access): New option.
1880 * config/nds32/nds32.c (nds32_asm_file_start): Display
1881 flag_unaligned_access status.
1883 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1885 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1886 -mno-relax is present.
1887 * config/riscv/linux.h (LINK_SPEC): Ditto.
1889 2018-04-20 Martin Sebor <msebor@redhat.com>
1892 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1894 (gimple_fold_builtin_stxcpy_chk): Same.
1895 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1897 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1900 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1901 __powikf2 when long double is IEEE 128-bit.
1903 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1905 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1906 step to make sure stack always aligned.
1908 2018-04-20 Carl Love <cel@us.ibm.com>
1911 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1912 size check for arg0.
1914 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1915 Tom de Vries <tom@codesourcery.com>
1918 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1919 Emit insns for calls too.
1920 (nvptx_find_par): Always look for worker-level predecessor insn.
1921 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1923 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1924 (nvptx_process_pars): Propagate frames for calls.
1926 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1929 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1931 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1932 (ix86_handle_option): Don't handle OPT_mibt.
1933 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1935 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1937 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1938 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1939 (ix86_target_macros): Define __CET__ with flag_cf_protection
1940 for -fcf-protection.
1941 * config/i386/i386.c (isa2_opts): Remove -mibt.
1942 * config/i386/i386.h (TARGET_IBT): Removed.
1943 (TARGET_IBT_P): Likewise.
1944 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1945 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1946 * config/i386/i386.opt (mcet): Update help message.
1949 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1950 -mcet as an alias for -mshstk.
1952 2018-04-20 Richard Biener <rguenther@suse.de>
1955 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1956 complexity by forcing a single use of the multiply operand.
1958 2018-04-20 Martin Jambor <mjambor@suse.cz>
1961 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1962 recursion dependency to only apply to non-clones.
1964 2018-04-20 Martin Jambor <mjambor@suse.cz>
1967 * ipa-cp.c (create_specialized_node): Check that clones of
1968 self-recursive edges exist during IPA-CP.
1970 2018-04-19 Toon Moene <toon@moene.org>
1972 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1975 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1977 PR tree-optimization/85467
1978 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1979 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1980 VECTOR_CST element to type.
1982 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1985 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1986 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1987 (builtin_longjmp): Likewise.
1988 (save_stack_nonlocal): New pattern.
1989 (restore_stack_nonlocal): Likewise.
1991 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1994 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1995 Replace ASM_OUTPUT_LABEL with fprintf.
1997 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2000 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2001 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2002 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2003 define __IBT__ and __SHSTK__ for -fcf-protection.
2004 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2006 (ix86_trampoline_init): Likewise.
2007 (x86_output_mi_thunk): Likewise.
2008 (ix86_notrack_prefixed_insn_p): Likewise.
2009 (ix86_option_override_internal): Don't disallow -fcf-protection.
2010 * config/i386/i386.md (rdssp<mode>): Also enable for
2012 (incssp<mode>): Likewise.
2013 (nop_endbr): Likewise.
2014 * config/i386/i386.opt (mcet): Change help message to built-in
2018 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2019 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2020 enable CET built-in functions.
2022 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2024 * common/config/i386/i386-common.c
2025 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2026 OPTION_MASK_ISA_MOVDIRI_UNSET,
2027 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2028 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2029 * config.gcc (movdirintrin.h): New header.
2030 * config/i386/cpuid.h (bit_MOVDIRI,
2031 bit_MOVDIR64B): New bits.
2032 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2034 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2035 (VOID, PVOID, PCVOID)): New function types.
2036 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2037 __builtin_ia32_directstoreu_u64,
2038 __builtin_ia32_movdir64b): New builtins.
2039 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2040 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2042 (ix86_valid_target_attribute_inner_p): Ditto.
2043 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2044 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2045 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2046 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2047 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2048 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2049 (movdiri<mode>, movdir64b_<mode>): New.
2050 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2051 * config/i386/immintrin.h: Include movdirintrin.h.
2052 * config/i386/movdirintrin.h: New file.
2053 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2055 2018-04-19 Richard Biener <rguenther@suse.de>
2058 * cfg.c (clear_bb_flags): When loop state says we have
2059 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2061 2018-04-19 Richard Biener <rguenther@suse.de>
2063 PR tree-optimization/84737
2064 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2065 copying restrict info.
2066 (vect_setup_realignment): Use it.
2067 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2068 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2069 the first DR to all generated stores.
2070 (vectorizable_load): Likewise for loads.
2072 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2074 PR tree-optimization/85446
2075 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2076 the integral and pointer types to have the same precision.
2078 * doc/install.texi: Document --disable-cet being the default and
2081 2018-04-18 Martin Liska <mliska@suse.cz>
2083 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2086 2018-04-18 Martin Liska <mliska@suse.cz>
2089 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2092 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2093 arguments if they are comparable.
2095 2018-04-18 Martin Liska <mliska@suse.cz>
2098 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2101 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2104 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2107 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2108 ENDBR after calling __morestack.
2110 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2113 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2114 by using gcc_base_ver to generate a gcc_driver_version, and use
2115 it when generating GCC_DRIVER_NAME.
2116 * configure: Regenerate.
2118 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2121 * config.gcc: Obsolete powerpc*-*-*spe*.
2123 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2126 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2127 (stabstr_D): Change type of unum from unsigned int to
2128 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2131 2018-04-17 Jim Wilson <jimw@sifive.com>
2134 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2135 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2136 Set arg_pointer_offset after using pretend_args_size.
2138 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2140 PR rtl-optimization/85431
2141 * dse.c (record_store): Ignore zero width stores.
2144 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2145 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2146 __builtin_stack_restore rather than after it.
2147 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2148 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2149 argument instead of virtual_dynamic_stack_rtx.
2151 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2153 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2155 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2156 Add note to error message to explain internal mapping of overloaded
2157 built-in function name to non-overloaded built-in function name.
2158 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2161 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2164 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2165 where the inputs overlap with the output.
2167 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2169 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2170 (=v, v) alternative and explicit "memory" attribute.
2171 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2172 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2174 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2175 "sselog1" type instead of "sselog".
2176 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2177 "sselog". Remove explicit "memory" attribute.
2178 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2179 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2181 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2182 "sselog1" type instead of "sselog". Remove explicit "memory"
2184 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2185 use "sselog1" type instead of "sselog". Remove explicit "memory"
2187 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2188 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2190 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2191 "sselog1" type instead of "sselog". Remove explicit "memory"
2193 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2194 use "sselog1" type instead of "sselog". Remove explicit "memory"
2198 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2201 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2204 2018-04-17 Martin Jambor <mjambor@suse.cz>
2207 * ipa-cp.c (create_specialized_node): Call
2208 expand_all_artificial_thunks if necessary.
2210 2018-04-17 Martin Liska <mliska@suse.cz>
2213 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2214 in message, remote space in between '_G' and '('.
2216 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2219 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2220 avx512f_vmcmp<mode>3<round_saeonly_name>,
2221 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2222 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2223 avx512f_rndscale<mode><round_saeonly_name>,
2224 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2225 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2226 Use %<iptr>2 instead of %2 for -masm=intel.
2227 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2228 avx512f_vcvttss2usi<round_saeonly_name>,
2229 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2231 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2232 avx512f_vcvttsd2usi<round_saeonly_name>,
2233 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2234 Use %q1 instead of %1 for -masm=intel.
2235 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2236 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2237 of %3 for -masm=intel.
2238 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2240 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2242 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2244 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2245 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2247 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2249 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2250 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2251 %g1 and one with %0 and %1.
2252 (avx512er_vmrcp28<mode><round_saeonly_name>,
2253 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2255 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2256 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2257 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2258 of %0 and %{%4%} for -masm=intel.
2259 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2260 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2261 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2262 order of %0 and %{%5%}%{z%} for -masm=intel.
2264 2018-04-17 Jan Hubicka <jh@suse.cz>
2267 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2269 2018-04-17 Martin Liska <mliska@suse.cz>
2272 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2273 for target_clone error message. Make default implementation
2274 clone to be a local declaration.
2275 (separate_attrs): Add new argument and check for an empty
2277 (expand_target_clones): Handle it.
2278 (ipa_target_clone): Make redirection just for target_clones
2281 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2282 Tom de Vries <tom@codesourcery.com>
2285 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2286 tiled basic blocks without omp continue statements.
2288 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2291 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2292 vec_extract expression as having side effects to make sure it gets
2295 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2298 * config/i386/i386.c (get_builtin_code_for_version): Check
2301 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2304 * gcc/config.gcc: Support "skylake".
2305 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2307 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2308 (processor_target_table): Add "skylake".
2309 (ix86_option_override_internal): Add "skylake".
2310 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2311 PROCESSOR_CANNONLAKE.
2312 (get_builtin_code_for_version): Fix priority for
2313 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2314 PROCESSOR_SKYLAKE-AVX512.
2315 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2316 (processor_type): Add PROCESSOR_SKYLAKE.
2318 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2319 Jason Merrill <jason@redhat.com>
2322 * convert.c (convert_to_integer_1): Use direct recursion for
2323 enumeral types and types with a precision less than the number
2324 of bits in their mode.
2326 2018-04-16 Julia Koval <julia.koval@intel.com>
2329 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2330 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2332 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2335 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2336 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2337 and -mno-direct-move.
2339 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2342 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2343 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2344 Ensure negative shifts result in {0}.
2346 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2348 PR rtl-optimization/79916
2349 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2350 regs (if any) to define how to gnerate SD moves when LRA is in
2353 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2355 PR rtl-optimization/85393
2356 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2357 * except.c (expand_dw2_landing_pad_for_region): Make static.
2358 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2359 a label and unconditional jump to old_bb, rather than
2360 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2363 PR rtl-optimization/85376
2364 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2365 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2366 instead of a specific value.
2368 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2369 Bin Cheng <bin.cheng@arm.com>
2371 PR tree-optimization/82965
2372 PR tree-optimization/83991
2373 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2374 by_profile_only parameter.
2375 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2376 information if the loop was predicted to iterate too many times.
2377 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2379 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2382 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2385 2018-04-13 Martin Liska <mliska@suse.cz>
2386 Jakub Jelinek <jakub@redhat.com>
2389 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2390 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2391 * builtins.c (expand_builtin_memory_copy_args): Use
2392 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2393 handle dest_addr == pc_rtx.
2395 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2398 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2399 asked to not generate direct moves.
2400 (fix_trunc<mode>si2_stfiwx): Similar.
2401 (fix_trunc<mode>si2_internal): Similar.
2403 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2406 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2407 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2408 lookup if dest in some wider mode is known to be const0_rtx and
2409 if so, record permanent equivalence for it to be ZERO_EXTEND of
2410 the narrower mode destination.
2412 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2414 * lto-streamer-out.c (output_function): Revert 259346.
2415 * omp-expand.c (expand_oacc_for): Likewise.
2417 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2419 PR rtl-optimization/85354
2420 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2421 * sel-sched.c (sel_global_init): ... here.
2423 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2426 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2427 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2428 mode for PE-COFF targets.
2429 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2430 (i386_pe_asm_lto_end): Likewise.
2431 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2432 (TARGET_ASM_LTO_END): Likewise.
2433 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2434 (i386_pe_asm_lto_start): New function.
2435 (i386_pe_asm_lto_end): Likewise.
2437 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2438 Richard Biener <rguenther@suse.de>
2441 * lto-streamer-out.c (output_function): Fix CFG loop state before
2443 * omp-expand.c (expand_oacc_for): Handle calls to internal
2444 functions like regular functions.
2446 2018-04-12 Richard Biener <rguenther@suse.de>
2449 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2450 for the early LTO debug to properly generate references to it
2451 during DIE emission. Do not re-use that for the skeleton for
2453 (dwarf2out_early_finish): Likewise.
2455 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2458 * config/i386/sse.md
2459 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2460 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2461 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2462 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2463 and output is a reg, avoid creating invalid lowpart subreg, but
2464 instead split into a 512-bit move. Don't split if not AVX512VL,
2465 input is xmm16+ reg and output is a mem.
2466 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2467 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2468 xmm16+ reg and output is a mem.
2470 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2472 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2473 also for flag_dwarf2_cfi_asm.
2475 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2477 PR rtl-optimization/85342
2478 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2479 a bool scalar var inside of the loop instead. Don't try to update
2480 recog_data.operand after failed apply_change_group.
2482 2018-04-12 Tom de Vries <tom@codesourcery.com>
2485 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2486 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2487 array with flexible array member as array without given dimension.
2488 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2489 argument for undefined param to true.
2491 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2494 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2495 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2496 from PowerPC section.
2497 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2498 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2499 help text that is too long.
2500 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2501 help text that is too long.
2502 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2503 help text that is too long.
2505 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2507 * config/alpha/alpha.md (stack_probe_internal): Rename
2508 from "probe_stack". Update all callers.
2510 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2512 PR rtl-optimization/84566
2513 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2514 sched_macro_fuse_insns.
2516 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2519 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2520 (compute_block_dependences): ... from here.
2522 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2524 PR tree-optimization/85331
2525 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2526 from int to HOST_WIDE_INT.
2528 2018-04-11 Martin Jambor <mjambor@suse.cz>
2531 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2532 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2533 not the same as the source val.
2534 (cgraph_edge_brings_value_p): New parameter.
2535 (gather_edges_for_value): Pass destination value to
2536 cgraph_edge_brings_value_p.
2537 (perhaps_add_new_callers): Likewise.
2538 (get_info_about_necessary_edges): Likewise and exclude values brought
2539 only by self-recursive edges.
2540 (create_specialized_node): Redirect only clones of self-calling edges.
2541 (+self_recursive_pass_through_p): New function.
2542 (find_more_scalar_values_for_callers_subset): Use it.
2543 (find_aggregate_values_for_callers_subset): Likewise.
2544 (known_aggs_to_agg_replacement_list): Removed.
2545 (decide_whether_version_node): Re-calculate known constants for all
2546 remaining context clones.
2548 2018-04-11 Richard Biener <rguenther@suse.de>
2551 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2552 from early DWARF output.
2553 (dwarf2out_early_finish): Output line info unconditionally into
2554 early DWARF and add reference to it.
2556 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2559 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2560 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2561 other than V2DFmode using iptr mode attribute.
2562 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2564 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2566 PR rtl-optimization/84659
2567 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2569 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2572 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2574 (output_loc_list): Pass address of a dummy size variable even in the
2575 locview handling loop.
2576 (index_location_lists): Add comment on why skip_loc_list_entry can't
2579 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2582 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2585 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2588 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2589 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2590 and -mstring-compare-inline-limit.
2592 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2595 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2596 for stack clash protection in a register whenever we need it to be in
2599 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2601 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2602 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2604 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2607 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2609 (mlong-double-): Ditto.
2610 * config/rs6000/sysv4.opt (msdata=): Ditto.
2611 (mtls-size=): Ditto.
2613 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2615 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2616 erroneous entries for
2617 "vector int vec_ldl (int, long int *)", and
2618 "vector unsigned int vec_ldl (int, unsigned long int *)".
2619 Add comments and entries for
2620 "vector bool char vec_ldl (int, bool char *)",
2621 "vector bool short vec_ldl (int, bool short *)",
2622 "vector bool int vec_ldl (int, bool int *)",
2623 "vector bool long long vec_ldl (int, bool long long *)",
2624 "vector pixel vec_ldl (int, pixel *)",
2625 "vector long long vec_ldl (int, long long *)",
2626 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2627 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2628 type tree bool_long_long_type_node and correct definition of
2629 bool_V2DI_type_node to make reference to this new type tree.
2630 (rs6000_mangle_type): Replace erroneous reference to
2631 bool_long_type_node with bool_long_long_type_node.
2632 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2633 comments to emphasize sign distinctions for char and int types and
2634 replace RS6000_BTI_bool_long constant with
2635 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2636 use of RS6000_BTI_pixel.
2637 (bool_long_type_node): Remove this macro definition.
2638 (bool_long_long_type_node): New macro definition
2640 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2642 PR rtl-optimization/85300
2643 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2644 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2645 simplify_unary_operation fails.
2647 2018-04-10 Martin Liska <mliska@suse.cz>
2649 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2650 cgraph_edge and ipa_ref.
2652 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2656 * config/i386/sse.md
2657 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2658 computation of the VEC_MERGE selector from mask.
2659 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2660 Fix decoding of the VEC_MERGE selector into mask.
2662 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2664 PR tree-optimization/85286
2665 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2667 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2669 * final.c (final_1): Set insn_last_address as well as
2670 insn_current_address.
2672 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2675 * explow.c (emit_stack_probe): Call validize_mem on memory location
2676 before passing it to gen_probe_stack. Create address operand and
2677 legitimize it for the probe_stack_address case.
2679 2018-04-09 Jan Hubicka <jh@suse.cz>
2682 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2683 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2684 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2686 (free_lang_data): Rebuild type inheritance graph.
2688 2018-04-09 Martin Sebor <msebor@redhat.com>
2690 * invoke.texi (-finline-small-functions): Mention other optimization
2692 (-findirect-inlining, -fpartial-inlining): Same.
2693 (-finline-functions-called-once): Same.
2694 (-freorder-blocks-and-partition): Same.
2696 2018-04-09 Jan Hubicka <jh@suse.cz>
2699 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2700 jumps; choose last target that matches the criteria (i.e.
2701 no partition changes for non-crossing jumps).
2702 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2703 support for redirecting crossing jumps to non-crossing.
2705 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2707 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2708 also for naked functions.
2710 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2712 * config/arc/arc.md (add_shift): New pattern.
2713 (add_shift2): Likewise.
2714 (sub_shift): Likewise.
2715 (sub_shift_cmp0_noout): Likewise.
2716 (compare_si_ashiftsi): Likewise.
2717 (xbfu_cmp0_noout): New combine pattern.
2718 (xbfu_cmp0"): Likewise.
2719 (movsi_set_cc_insn): Place the predicable variant first.
2720 (commutative_binary_cmp0_noout): Remove clobber.
2721 (commutative_binary_cmp0): New pattern.
2722 (noncommutative_binary_cmp0): Likewise.
2723 (noncommutative_binary_cmp0_noout): Likewise.
2724 (noncommutative_binary_comparison_result_used): Removed.
2725 (rsub_cmp0): New pattern.
2726 (rsub_cmp0_noout): Likewise.
2727 (extzvsi): Changed, keep only meaningful variants.
2728 (SQH, SEZ): New iterators.
2729 (SQH_postfix): New mode attribute.
2730 (SEZ_prefix): New code attribute.
2731 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2732 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2733 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2735 (noncommutative_operator): Check the availability of barrel
2738 2018-04-09 Richard Biener <rguenther@suse.de>
2740 PR tree-optimization/85284
2741 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2742 Only use the niter constraining form of simple_iv when the exit
2745 2018-04-09 Tom de Vries <tom@codesourcery.com>
2748 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2749 (define_expand "*memory_barrier"): New define_expand.
2750 (define_insn "memory_barrier"): New insn.
2752 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2754 PR rtl-optimization/80463
2755 PR rtl-optimization/83972
2756 PR rtl-optimization/83480
2758 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2759 correct producer for the insn.
2760 (tidy_control_flow): Fixup seqnos in case of debug insns.
2762 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2764 PR rtl-optimization/83913
2766 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2767 different sched-times when merging exprs.
2769 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2771 PR rtl-optimization/83962
2773 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2774 tidy_fallthru_edge and tidy_control_flow.
2776 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2778 PR rtl-optimization/83530
2780 * sel-sched.c (force_next_insn): New global variable.
2781 (remove_insn_for_debug): When force_next_insn is true, also leave only
2782 next insn in the ready list.
2783 (sel_sched_region): When the region wasn't scheduled, make another pass
2784 over it with force_next_insn set to 1.
2786 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2788 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2790 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2791 for interrupt control.
2792 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2793 functions for interrupt control.
2794 * config/nds32/nds32-intrinsic.md: Likewise.
2795 * config/nds32/nds32_intrinsic.h: Likewise.
2796 * config/nds32/nds32.h (nds32_builtins): Likewise.
2798 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2800 * config/nds32/nds32.c (nds32_init_machine_status,
2801 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2802 strict_aligned_p field.
2803 (nds32_expand_to_rtl_hook): New function.
2804 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2805 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2807 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2808 Chung-Ju Wu <jasonwucj@gmail.com>
2810 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2811 * config/nds32/nds32-n7.md: New file.
2812 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2813 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2815 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2816 * config/nds32/nds32.md (pipeline_model): Add n7.
2817 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2818 * config/nds32/pipelines.md: Include n7 settings.
2820 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2821 Chung-Ju Wu <jasonwucj@gmail.com>
2823 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2824 * config/nds32/nds32-e8.md: New file.
2825 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2826 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2828 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2829 * config/nds32/nds32.md (pipeline_model): Add e8.
2830 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2831 * config/nds32/pipelines.md: Include e8 settings.
2833 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2834 Chung-Ju Wu <jasonwucj@gmail.com>
2836 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2837 * config/nds32/nds32-n8.md: New file.
2838 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2839 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2841 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2842 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2843 * config/nds32/nds32.md (pipeline_model): Add n8.
2844 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2845 * config/nds32/pipelines.md: Include n8 settings.
2847 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2848 Chung-Ju Wu <jasonwucj@gmail.com>
2850 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2851 * config/nds32/nds32-n9-2r1w.md: New file.
2852 * config/nds32/nds32-n9-3r2w.md: New file.
2853 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2854 nds32_register_ports): New or modify for cpu n9.
2855 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2857 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2858 * config/nds32/nds32-utils.c: New file.
2859 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2860 TARGET_MUL_SLOW): Define.
2861 * config/nds32/nds32.md (pipeline_model): New attribute.
2862 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2863 New options that support cpu n9.
2864 * config/nds32/pipelines.md: Include n9 settings.
2865 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2867 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2869 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2870 information if necessary.
2871 (output_cond_branch_compare_zero): Likewise.
2872 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2873 (nds32_target_alignment): Refine for alignment.
2874 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2875 (FUNCTION_BOUNDARY): Modify.
2876 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2878 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2880 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2882 * config/nds32/constants.md (unspec_volatile_element): Add values for
2883 TLB operation and data prefetch.
2884 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2885 functions for TLB operation and data prefetch.
2886 * config/nds32/nds32-intrinsic.md: Likewise.
2887 * config/nds32/nds32_intrinsic.h: Likewise.
2888 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2889 (nds32_print_operand): Likewise.
2890 * config/nds32/nds32.h (nds32_builtins): Likewise.
2892 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2893 Andrew Pinski <pinsika@gcc.gnu.org>
2896 * match.pd: Use constant_boolean_node of correct type instead of
2897 boolean_true_node or boolean_false_node for simplifying
2898 pointer comparisons to zero.
2900 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2902 PR tree-optimization/80021
2903 * tree.c (verify_type_variant): Make error call in verify_variant_match
2904 translatable and remove final full stop.
2906 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2908 * config/nds32/constants.md (unspec_volatile_element): Add
2909 UNSPEC_VOLATILE_EH_RETURN.
2910 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2911 nds32_output_stack_pop): Support dwarf exception handling process.
2912 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2913 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2914 exception handling process.
2915 (nds32_compute_stack_frame): Likewise.
2916 (nds32_return_addr_rtx): Likewise.
2917 (nds32_initial_elimination_offset): Likewise.
2918 (nds32_expand_prologue): Likewise.
2919 (nds32_expand_epilogue): Likewise.
2920 (nds32_dynamic_chain_address): New function.
2921 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2923 (DYNAMIC_CHAIN_ADDRESS): Define.
2924 (EH_RETURN_DATA_REGNO): Define.
2925 (EH_RETURN_STACKADJ_RTX): Define.
2926 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2927 patterns for dwarf exception handling.
2929 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2931 * config/nds32/nds32.h: Clean up obsolete macros.
2933 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2935 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2936 Add enum values for particular instructions.
2937 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2938 particular intrinsic functions.
2939 * config/nds32/nds32-intrinsic.md: Likewise.
2940 * config/nds32/nds32_intrinsic.h: Likewise.
2941 * config/nds32/nds32.h (nds32_builtins): Likewise.
2942 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2943 (btst, ave): New patterns for particular instructions.
2945 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2947 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2948 Add enum values for atomic load/store and memory sync.
2949 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2951 * config/nds32/nds32-intrinsic.md: Likewise.
2952 * config/nds32/nds32_intrinsic.h: Likewise.
2953 * config/nds32/nds32.h (nds32_builtins): Likewise.
2955 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2957 PR tree-optimization/85257
2958 * fold-const.c (native_encode_vector): If not all elts could fit
2959 and off is -1, return 0 rather than offset.
2960 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2961 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2962 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2963 adjust buffer in native_interpret_expr call.
2965 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2967 * config/nds32/constants.md (unspec_volatile_element): Add cache
2968 control enum values.
2969 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2970 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2971 * config/nds32/nds32.c (nds32_cctl_names): New.
2972 (nds32_print_operand): Handle cache control register names.
2973 * config/nds32/nds32.h (nds32_builtins): New enum values.
2974 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2976 * config/nds32/nds32.md (type): Add mmu.
2977 * config/nds32/pipelines.md (simple_insn): Add mmu.
2979 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2981 * config/nds32/nds32.md (type): Remove call.
2982 * config/nds32/pipelines.md (simple_insn): Likewise.
2984 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2986 * config/nds32/constants.md (unspec_volatile_element): Add
2987 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2988 UNSPEC_VOLATILE_FMFCFG.
2989 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2990 description for fmfcfg and fmfcsr.
2991 (bdesc_1arg): Add fmtcsr.
2992 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2993 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2994 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2995 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2996 unspec_fmfcfg): New patterns.
2997 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2998 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2999 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3000 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3001 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3002 __nds32__fmfcfg): Define.
3004 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3006 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3007 intrinsic register names.
3008 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3009 intrinsic register enum values and macros.
3011 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3013 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3014 for load/store addressing form.
3015 (nds32_print_operand_address): Likewise.
3017 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3020 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3021 based on LABEL_REF. Remove useless assertion.
3022 (pic_address_needs_scratch): Fix formatting.
3023 (sparc_legitimize_pic_address): Minor tweaks.
3024 (sparc_delegitimize_address): Adjust assertion accordingly.
3025 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3026 into symbolic_operand.
3027 (movsi_high_pic_label_ref): Likewise.
3028 (movsi_lo_sum_pic_label_ref): Likewise.
3029 (movdi_pic_label_ref): Likewise.
3030 (movdi_high_pic_label_ref): Likewise.
3031 (movdi_lo_sum_pic_label_ref): Likewise.
3033 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3035 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3036 custom LIB_SPEC setup.
3038 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3039 Kito Cheng <kito.cheng@gmail.com>
3041 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3042 * config/riscv/freebsd.h: New.
3044 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3046 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3047 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3050 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3051 Kito Cheng <kito.cheng@gmail.com>
3053 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3054 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3055 * config/nds32/nds32-protos.h (nds32_output_call,
3056 nds32_output_return): Declare.
3057 * config/nds32/nds32.md: Refine all the call and return patterns.
3059 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3062 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3063 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3065 PR rtl-optimization/84872
3066 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3067 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3070 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3072 * expr.c (copy_blkmode_to_reg): Revert 254862.
3073 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3075 2018-04-06 Richard Biener <rguenther@suse.de>
3078 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3079 after seeing a component reference with an adjacent field. Treat
3080 refs to arrays at struct end of external decls similar to
3081 refs to unconstrained commons.
3083 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3086 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3087 look through SAVE_EXPRs with non-side-effects argument. Adjust
3089 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3092 2018-04-06 Richard Biener <rguenther@suse.de>
3095 * alias.c (find_base_term): New wrapper around find_base_term
3096 unwinding CSELIB_VAL_PTR changes.
3097 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3100 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3102 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3104 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3105 constant definitions.
3106 ("nop"): lr 0,0 -> nopr r0
3107 ("nop_lr0", "nop_lr1"): New insn definitions.
3109 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3111 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3112 NDS32_V3PUSH_AVAILABLE_P macro.
3114 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3115 Chung-Ju Wu <jasonwucj@gmail.com>
3117 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3118 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3119 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3120 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3121 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3122 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3123 * config/nds32/constraints.md: New constraints and checking for hard
3124 float configuration.
3125 * config/nds32/iterators.md: New mode iterator and attribute for hard
3126 float configuration.
3127 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3129 * config/nds32/nds32-fpu.md: New file.
3130 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3131 deal with hard float code generation.
3132 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3134 (abi_type, float_reg_number): New enum type.
3135 * config/nds32/nds32-predicates.c: New predicates for hard float.
3136 * config/nds32/nds32-protos.h: Declare functions for hard float.
3137 * config/nds32/nds32.c: Implementation for hard float configuration.
3138 * config/nds32/nds32.h: Definitions for hard float configuration.
3139 * config/nds32/nds32.md: Include hard float machine description and
3140 modify patterns for hard float configuration.
3141 * config/nds32/nds32.opt: New options for hard float configuration.
3142 * config/nds32/predicates.md: New predicates for hard float
3145 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3147 * common/config/nds32/nds32-common.c
3148 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3150 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3153 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3154 CONSTRUCTOR_ELT (ctor, ...)->value.
3156 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3159 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3161 2018-04-05 Tom de Vries <tom@codesourcery.com>
3164 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3167 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3168 Kito Cheng <kito.cheng@gmail.com>
3170 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3171 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3172 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3173 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3175 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3176 Kito Cheng <kito.cheng@gmail.com>
3178 * config/nds32/constraints.md (Ufe): New memory constraint.
3179 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3180 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3181 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3183 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3184 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3186 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3188 * config/nds32/nds32.md: Use optimize_size in the condition for
3189 alu-shift instructions.
3191 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3193 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3195 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3197 * config/nds32/nds32.md (negsi2): Refine pattern.
3199 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3200 Chung-Ju Wu <jasonwucj@gmail.com>
3202 * config/nds32/iterators.md (shift_rotate): New code iterator.
3203 (shift): New code attribute.
3204 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3205 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3206 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3207 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3208 bit-wise operations.
3209 (andsi3, *andsi3): Ditto.
3210 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3211 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3212 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3213 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3214 nds32_ior_operand, nds32_xor_operand): New predicates.
3216 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3218 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3219 (addsi3, subsi3): ... this.
3221 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3223 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3225 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3227 * config/nds32/nds32.md: Adjust indention.
3229 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3231 * config/nds32/nds32.md (feature): New attribute.
3233 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3235 * config/nds32/nds32.md (subtype): New attribute.
3237 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3240 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3241 expansion to perform a bitwise AND of the argument followed by a
3242 boolean negation of the result.
3244 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3246 PR rtl-optimization/84878
3247 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3248 the basic block. Assert the use reference is not artificial and that
3249 it has an associated insn.
3251 2018-04-04 Michael Matz <matz@suse.de>
3253 * builtins.c (compute_objsize): Pass correct operand
3254 to array_at_struct_end_p.
3256 2018-04-04 Richard Biener <rguenther@suse.de>
3259 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3260 from contexts for DINFO_LEVEL_TERSE and below.
3262 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3264 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3265 resiter_operand condition.
3266 * config/nds32/nds32.md (*move<mode>): Ditto.
3268 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3269 Monk Chiang <sh.chiang04@gmail.com>
3271 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3273 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3275 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3277 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3278 Kito Cheng <kito.cheng@gmail.com>
3280 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3281 nds32_cond_code_str, output_cond_branch,
3282 output_cond_branch_compare_zero, nds32_expand_cbranch,
3283 nds32_expand_cstore, nds32_expand_movcc,
3284 nds32_output_cbranchsi4_equality_zero,
3285 nds32_output_cbranchsi4_equality_reg,
3286 nds32_output_cbranchsi4_equality_reg_or_const_int,
3287 nds32_output_cbranchsi4_greater_less_zero: New functions.
3288 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3289 nds32_expand_cstore, nds32_expand_movcc,
3290 nds32_output_cbranchsi4_equality_zero,
3291 nds32_output_cbranchsi4_equality_reg,
3292 nds32_output_cbranchsi4_equality_reg_or_const_int,
3293 nds32_output_cbranchsi4_greater_less_zero): Declare.
3294 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3295 nds32_rimm11s_operand): New predicates.
3296 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3297 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3300 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3302 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3303 * config/nds32/nds32.md: Ditto.
3304 * config/nds32/pipelines.md: Ditto.
3306 2018-04-04 Richard Biener <rguenther@suse.de>
3308 PR tree-optimization/85168
3309 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3310 propagating abnormals.
3312 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3314 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3316 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3317 Kito Cheng <kito.cheng@gmail.com>
3319 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3320 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3321 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3322 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3323 * config/nds32/nds32.md (sibcall_internal): New.
3324 (sibcall_register): Remove.
3325 (sibcall_immediate): Remove.
3326 (sibcall_value_internal): New.
3327 (sibcall_value_register): Remove.
3328 (sibcall_value_immediate): Remove.
3329 * config/nds32/predicates.md (nds32_general_register_operand): New.
3330 (nds32_call_address_operand): New.
3332 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3334 PR rtl-optimization/85167
3335 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3336 bb_defs if *split_p, instead preinitialize it to NULL.
3338 PR tree-optimization/85156
3339 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3340 evaluating the argument multiple times.
3342 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3344 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3346 (_mm_cvtpd_ps): Likewise.
3347 (_mm_cvttpd_epi32): Likewise.
3348 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3349 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3350 vector, pixel, and bool following altivec.h include.
3352 2018-04-03 Martin Sebor <msebor@redhat.com>
3354 * doc/extend.texi (Common Function Attributes): Clarify.
3355 (const attribute): Likewise.
3356 (pure attribute): Likewise.
3358 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3361 * config/i386/i386.c (ix86_expand_vector_set): Use
3362 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3364 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3366 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3367 instructions when changing rounding bits to preserve precision bits
3368 in the x87 control word.
3370 2018-04-03 Martin Liska <mliska@suse.cz>
3372 PR tree-optimization/82491
3373 * rtl.h (strip_offset_and_add): Replace += suboffset with
3374 poly_uint64 () + suboffset.
3376 2018-03-29 Martin Liska <mliska@suse.cz>
3377 Martin Jambor <mjambor@suse.cz>
3380 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3381 param_type is not an integral or pointer type.
3383 2018-04-03 Richard Biener <rguenther@suse.de>
3385 * sese.h (recompute_all_dominators): Remove.
3387 2018-04-02 Martin Sebor <msebor@redhat.com>
3389 * doc/invoke.texi (-Wrestrict): Fix typos.
3391 2018-04-02 Jim Wilson <jimw@sifive.com>
3393 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3394 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3395 (<optab>di3, <optab>si3_extend): Likewise.
3396 (<optab>si3_mask, <optab>si3_mask_1): New.
3397 (<optab>di3_mask, <optab>di3_mask_1): New.
3398 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3399 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3400 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3402 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3404 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3407 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3409 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3410 (nds32_canonicalize_comparison): New function.
3412 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3413 Kito Cheng <kito.cheng@gmail.com>
3414 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3416 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3417 * config/nds32/constants.md (unspec_volatile_element): Add
3418 UNSPEC_VOLATILE_RELAX_GROUP.
3419 * config/nds32/nds32-relax-opt.c: New file.
3420 * config/nds32/nds32-predicates.c
3421 (nds32_symbol_load_store_p): New function.
3422 * config/nds32/nds32-protos.h
3423 (nds32_symbol_load_store_p): Declare function.
3424 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3425 * config/nds32/nds32.c
3426 (nds32_register_pass): New function to register pass.
3427 (nds32_register_passes): New function to register passes.
3428 * config/nds32/nds32.md (relax_group): New pattern.
3429 * config/nds32/nds32.opt (mrelax-hint): New option.
3430 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3432 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3434 * config/nds32/t-nds32: Modify files dependency.
3436 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3438 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3439 (PROFILE_HOOK): Define its implementation.
3441 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3443 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3444 type and 32-bit size.
3446 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3449 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3450 (V_128_256): New mode iterator.
3451 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3452 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3453 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3455 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3456 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3458 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3461 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3462 NaN inputs correctly.
3464 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3467 * config/rs6000/vsx.md (??r): New mode attribute.
3468 (*vsx_mov<mode>_64bit): Use it.
3469 (*vsx_mov<mode>_32bit): Likewise.
3471 2018-03-30 Martin Sebor <msebor@redhat.com>
3473 PR tree-optimization/84818
3474 * builtins.c (check_access): Use warning_n.
3476 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3479 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3481 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3484 2018-03-30 Julia Koval <julia.koval@intel.com>
3487 * x86-tune.def (movx, partial_reg_dependency): Enable for
3490 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3493 * lra-constraints.c (process_alt_operands): Move setting
3494 this_alternative_matches below.
3496 2018-03-29 Martin Liska <mliska@suse.cz>
3499 * doc/invoke.texi: Document how LTO works with debug info.
3500 Describe auto-load support of binutils. Mention 'x86-64'
3501 as valid option value of -march option.
3503 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3505 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3508 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3509 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3510 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3513 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3516 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3517 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3518 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3519 for RS6000_BTM_POWERPC64.
3520 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3521 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3522 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3527 2018-03-28 Carl Love <cel@us.ibm.com>
3530 2017-09-27 Carl Love <cel@us.ibm.com>
3532 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3533 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3534 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3537 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3539 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3540 instead of __vector bool.
3541 (_mm_max_pu8): Likewise.
3542 (_mm_min_pi16): Likewise.
3544 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3547 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3548 (DIVWEUO): Likewise.
3550 (DIVDEUO): Likewise.
3551 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3552 DIVWEUO and DIVDEUO.
3553 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3554 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3555 (div_extend): Likewise.
3556 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3558 (__builtin_divweuo): Likewise.
3559 (__builtin_divdeo): Likewise.
3560 (__builtin_divdeuo): Likewise.
3562 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3565 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3566 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3568 PR tree-optimization/82004
3569 * gimple-match-head.c (optimize_pow_to_exp): New function.
3570 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3571 Don't fold to exp if optimize_pow_to_exp is false.
3573 2018-03-28 Martin Liska <mliska@suse.cz>
3576 * calls.c (initialize_argument_information): Fix trailing space.
3577 * common.opt: Fix typo and provide better explanation for
3578 -fsanitize-coverage option.
3579 * config/i386/i386.opt: Fix typo.
3581 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3582 Martin Liska <mliska@suse.cz>
3585 * gimplify.c (asan_poison_variable): Don't do the check for
3586 gimplify_omp_ctxp here.
3587 (gimplify_decl_expr): Do it here.
3588 (gimplify_target_expr): Likewise.
3590 2018-03-28 Martin Liska <mliska@suse.cz>
3593 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3594 chkp_type_bounds_count if MPX is not enabled.
3596 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3598 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3600 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3603 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3604 function to create the function decl for complex long double
3605 multiply and divide for -mabi=ieeelongdouble.
3606 (init_float128_ieee): Call it.
3608 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3611 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3612 -fcf-protection=branch -mibt.
3613 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3615 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3618 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3620 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3623 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3624 extern array declarations.
3626 2018-03-27 Richard Biener <rguenther@suse.de>
3629 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3630 explicit single_use checks.
3632 2018-03-27 Richard Biener <rguenther@suse.de>
3634 PR tree-optimization/85082
3635 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3638 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3640 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3641 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3642 Turn on fasynchronous-unwind-tables and funwind-tables.
3644 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3647 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3648 (*bmi_blsr_<mode>_ccz): Ditto.
3650 2018-03-26 Tom de Vries <tom@codesourcery.com>
3652 PR tree-optimization/85063
3653 * omp-general.c (offloading_function_p): New function. Factor out
3655 * omp-offload.c (pass_omp_target_link::gate): ... here.
3656 * omp-general.h (offloading_function_p): Declare.
3657 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3658 with attribute omp declare target for offloading functions.
3660 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3662 PR tree-optimization/84005
3663 * tree-data-ref.h (get_base_for_alignment): Declare.
3664 * tree-data-ref.c (get_base_for_alignment_1): New function.
3665 (get_base_for_alignment): Likewise.
3666 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3667 get_base_for_alignment to find a suitable base object, instead
3668 of always using drb->base_address.
3670 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3673 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3674 known size by default.
3676 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3679 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3680 and non BLKmode operands.
3682 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3685 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3686 Clean up attributes.
3688 2018-03-23 Richard Biener <rguenther@suse.de>
3691 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3692 we are going to emit early debug for LTO.
3694 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3697 * function.c (match_asm_constraints_1): Don't optimize if input
3698 doesn't satisfy general_operand predicate for output's mode.
3701 * alias.c (write_dependence_p): Don't require for x_canonicalized
3702 non-VOIDmode if x has VOIDmode.
3705 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3706 just don't try to optimize it rather than assert it never happens.
3708 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3710 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3711 macro expansions for definition of ST_INTERNAL_<mode> and
3712 LD_INTERNAL_<mode> builtins.
3713 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3715 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3717 (altivec_expand_st_builtin): Likewise.
3718 (altivec_expand_builtin): Remove calls to deleted functions.
3719 (rs6000_address_for_altivec): Delete this function.
3720 * config/rs6000/vector.md: Remove expands for
3721 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3723 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3726 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3727 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3728 re-computing once computed.
3729 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3730 (arm_init_machine_status): Initialize
3731 machine->static_chain_stack_bytes.
3733 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3736 * doc/extend.texi: Add four new prototypes for vec_ld.
3737 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3738 definitions for more logical presentation.
3739 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3740 entries for V1TI variants of __builtin_altivec_ld builtin.
3741 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3742 handling of V1TI variant of LVX icode pattern.
3743 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3744 (rs6000_gimple_fold_builtin): Likewise.
3745 (altivec_init_builtins): Add code to define
3746 __builtin_altivec_lvx_v1ti function.
3748 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3751 * function.c (match_asm_constraints_1): Don't do the optimization
3752 if input isn't a REG, SUBREG, MEM or constant.
3754 2018-03-22 Tom de Vries <tom@codesourcery.com>
3756 PR tree-optimization/84956
3757 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3758 bb_has_abnormal_pred.
3760 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3763 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3764 DECL_INITIAL (decl) to decl at the end.
3765 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3768 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3770 * doc/extend.texi (__builtin_tgmath): Document when complex
3771 integer types are treated as _Complex _Float64.
3773 2018-03-21 Tom de Vries <tom@codesourcery.com>
3775 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3777 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3779 PR tree-optimization/84960
3780 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3781 if it is ENTRY block, move them into single succ of ENTRY in that case.
3783 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3785 PR tree-optimization/84811
3786 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3787 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3788 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3789 (known_subrange_p): Update accordingly. Cast each value involved
3790 in the size comparison, rather than casting the result of the
3793 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3795 PR tree-optimization/84982
3796 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3797 by flipping the least significant bit rather than all bits from
3798 bitpos to bitpos + bitsize - 1.
3800 2018-03-21 Nathan Sidwell <nathan@acm.org>
3802 * doc/extend.texi (Deprecated Features): Remove mention of
3803 long-deleted deprecations.
3805 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3808 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3809 * configure: Regenerate.
3811 2018-03-21 Tom de Vries <tom@codesourcery.com>
3813 PR tree-optimization/83126
3814 * tree-parloops.c (num_phis): New function.
3815 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3817 2018-03-21 Nathan Sidwell <nathan@acm.org>
3819 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3820 mention anon-struct/union members and trailing attributes.
3822 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3824 PR tree-optimization/84969
3825 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3826 builtin memset partitions if they set different rhs values.
3828 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3830 PR rtl-optimization/84989
3831 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3832 VEC_DUPLICATE with scalar result mode.
3834 2018-03-21 Martin Liska <mliska@suse.cz>
3837 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3838 not intended return statement.
3840 2018-03-21 Martin Liska <mliska@suse.cz>
3843 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3844 (chkp_find_bound_slots_1): Limit number of iterations.
3846 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3849 * Minor grammar fixes for x86 options.
3851 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3854 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3855 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3857 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3860 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3861 (altivec_lvx_<mode>_1op): Likewise.
3862 (altivec_stvx_<mode>_2op): Likewise.
3863 (altivec_stvx_<mode>_1op): Likewise.
3864 (altivec_lvx_<VM2:mode>): New define_expand.
3865 (altivec_stvx_<VM2:mode>): Likewise.
3866 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3867 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3868 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3869 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3870 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3871 (rs6000_gen_lvx): Likewise.
3872 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3873 (altivec_expand_stv_builtin): Likewise.
3874 (altivec_expand_builtin): Likewise.
3875 * config/rs6000/vector.md: Likewise.
3877 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3880 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3883 2018-03-20 Richard Biener <rguenther@suse.de>
3886 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3887 sign-conversions as zero, fall back to standard scalar_stmt
3890 2018-03-20 Martin Liska <mliska@suse.cz>
3893 * predict.c (rebuild_frequencies): Handle case when we have
3894 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3896 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3899 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3900 flag_section_anchors.
3901 * varasm.c (use_blocks_for_decl_p): Remove hack for
3902 dw2_force_const_mem.
3905 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3907 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3908 be created, use lowpart_subreg of operands[0] rather than operands[0]
3910 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3911 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3912 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3913 and n constraint instead of aarch64_shift_imm_di and Usd.
3914 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3915 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3917 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3920 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3921 to favor GPR over NEON registers.
3922 (<shift>di3_neon): Likewise.
3924 2018-03-20 Tom de Vries <tom@codesourcery.com>
3927 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3928 (nvptx_process_pars): Emit bar.sync asap and alap.
3930 2018-03-20 Tom de Vries <tom@codesourcery.com>
3933 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3934 seen_label if seen_label is already set.
3936 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3939 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3940 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3941 Use 1U instead of 1. Formatting fixes.
3944 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3945 instead of TREE_TYPE (s1) for the return value.
3947 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3949 PR tree-optimization/84946
3950 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3951 bitsize + bitsize in poly_uint64 rather than poly_int64.
3954 * dwarf2asm.c: Include fold-const.c.
3955 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3956 of decl rather than decl itself.
3958 PR rtl-optimization/84643
3959 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3961 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3964 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3965 calling assemble_variable.
3967 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3970 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3971 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3973 2018-03-19 Jim Wilson <jimw@sifive.com>
3976 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3977 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3978 (riscv_first_stack_step): Likewise.
3979 (riscv_option_override): Use STACK_BOUNDARY instead of
3981 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3983 (BIGGEST_ALIGNMENT): Set to 128.
3984 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3985 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3988 2018-03-19 Richard Biener <rguenther@suse.de>
3990 PR tree-optimization/84933
3991 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3992 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3994 2018-03-19 Richard Biener <rguenther@suse.de>
3996 PR tree-optimization/84859
3997 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3998 (cond_if_else_store_replacement): Perform sinking operation on
3999 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4000 Generalize what a BB with a single eligible store is.
4002 2018-03-19 Richard Biener <rguenther@suse.de>
4004 PR tree-optimization/84929
4005 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4006 chrec_is_positive against non-chrec arg.
4008 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4011 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4013 2018-03-18 Martin Liska <mliska@suse.cz>
4015 PR rtl-optimization/84635
4016 * regrename.c (build_def_use): Use matches_mode only when
4019 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4021 PR tree-optimization/84913
4022 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4023 vectorize chains of COND_EXPRs.
4025 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4027 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4029 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4031 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4033 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4035 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4037 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4038 Kito Cheng <kito.cheng@gmail.com>
4040 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4041 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4042 (nds32_adjust_reg_alloc_order): New function.
4043 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4045 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4047 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4048 nds32_print_operand, nds32_print_operand_address): Use
4049 HOST_WIDE_INT_PRINT_DEC instead.
4051 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4053 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4055 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4058 * config/i386/i386.c (initial_ix86_tune_features,
4059 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4061 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4062 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4063 rather than 1u << ix86_tune. Formatting fix.
4064 (ix86_option_override_internal): Change ix86_arch_mask from
4065 unsigned int to unsigned HOST_WIDE_INT, initialize to
4066 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4067 (ix86_function_specific_restore): Likewise.
4069 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4072 * postreload.c (reload_combine_recognize_pattern): Perform
4073 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4074 truncate_int_for_mode the result for the destination's mode.
4077 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4078 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4080 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4083 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4086 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4087 regno_allocno_class_array and sorted_pseudos.
4088 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4089 insns where regno is used.
4091 2018-03-16 Martin Liska <mliska@suse.cz>
4094 * multiple_target.c (create_dispatcher_calls): Redirect
4095 reference in the symbol table.
4097 2018-03-16 Martin Liska <mliska@suse.cz>
4100 * multiple_target.c (create_dispatcher_calls): Redirect also
4103 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4107 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4108 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4109 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4111 2018-03-16 Julia Koval <julia.koval@intel.com>
4113 * doc/invoke.texi (Skylake Server): Add CLWB.
4114 Cannonlake): Remove CLWB.
4116 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4118 PR tree-optimization/84841
4119 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4121 (FLOAT_ONE_CONST_TYPE): Define.
4122 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4123 (sort_by_operand_rank): Put entries with higher constant_type last
4124 rather than first to match comments.
4126 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4128 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4131 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4134 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4135 check and use address of target always.
4137 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4140 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4141 (indirect_thunk_bnd_needed): Likewise.
4142 (indirect_thunks_used): Likewise.
4143 (indirect_thunks_bnd_used): Likewise.
4144 (indirect_return_needed): New.
4145 (indirect_return_bnd_needed): Likewise.
4146 (output_indirect_thunk_function): Add a bool argument for
4148 (output_indirect_thunk_function): Don't generate alias for
4149 function return thunk.
4150 (ix86_code_end): Call output_indirect_thunk_function to generate
4151 function return thunks.
4152 (ix86_output_function_return): Set indirect_return_bnd_needed
4153 and indirect_return_needed instead of indirect_thunk_bnd_needed
4154 and indirect_thunk_needed.
4156 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4158 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4159 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4160 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4162 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4163 Paul Hua <paul.hua.gm@gmail.com>
4166 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4168 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4170 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4171 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4174 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4177 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4178 instead of GET_MODE_SIZE when comparing Units.
4180 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4183 * varasm.c (hash_section): Return an unchangeble hash value
4184 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4185 Return !aarch64_can_use_per_function_literal_pools_p ().
4187 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4190 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4191 rather than address of cmode as last argument to prepare_cmp_insn.
4193 2018-03-15 Julia Koval <julia.koval@intel.com>
4195 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4196 F_AVX512VNNI, F_AVX512BITALG): New.
4198 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4201 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4202 insn for floating-point loads and stores.
4204 2018-03-14 Carl Love <cel@us.ibm.com>
4206 * config/rs6000/rs6000-c.c: Add macro definitions for
4207 ALTIVEC_BUILTIN_VEC_PERMXOR.
4208 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4209 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4210 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4211 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4213 * config/doc/extend.texi: Add prototypes for vec_permxor.
4215 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4218 * diagnostic-show-locus.c (class layout_point): Convert m_line
4219 from int to linenum_type.
4220 (line_span::comparator): Use linenum "compare" function when
4221 comparing line numbers.
4222 (test_line_span): New function.
4223 (layout_range::contains_point): Convert param "row" from int to
4225 (layout_range::intersects_line_p): Likewise.
4226 (layout::will_show_line_p): Likewise.
4227 (layout::print_source_line): Likewise.
4228 (layout::should_print_annotation_line_p): Likewise.
4229 (layout::print_annotation_line): Likewise.
4230 (layout::print_leading_fixits): Likewise.
4231 (layout::annotation_line_showed_range_p): Likewise.
4232 (struct line_corrections): Likewise for field m_row.
4233 (line_corrections::line_corrections): Likewise for param "row".
4234 (layout::print_trailing_fixits): Likewise.
4235 (layout::get_state_at_point): Likewise.
4236 (layout::get_x_bound_for_row): Likewise.
4237 (layout::print_line): Likewise.
4238 (diagnostic_show_locus): Likewise for locals "last_line" and
4240 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4241 * input.c (selftest::test_linenum_comparisons): New function.
4242 (selftest::input_c_tests): Call it.
4243 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4244 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4245 * selftest.h (ASSERT_GT): New macro.
4246 (ASSERT_GT_AT): New macro.
4247 (ASSERT_LT): New macro.
4248 (ASSERT_LT_AT): New macro.
4250 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4252 PR rtl-optimization/84780
4253 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4255 2018-03-14 Martin Liska <mliska@suse.cz>
4257 * tree.c (record_node_allocation_statistics): Use
4258 get_stats_node_kind.
4259 (get_stats_node_kind): New function extracted from
4260 record_node_allocation_statistics.
4261 (free_node): Use get_stats_node_kind.
4263 2018-03-14 Richard Biener <rguenther@suse.de>
4265 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4266 that the value-set of ANTIC_IN doesn't grow.
4269 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4271 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4272 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4274 2018-03-14 Julia Koval <julia.koval@intel.com>
4276 * config.gcc (icelake-client, icelake-server): New.
4278 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4279 (initial_ix86_arch_features): Ditto.
4280 (PTA_SKYLAKE): Add SGX.
4281 (PTA_ICELAKE): Remove.
4282 (PTA_ICELAKE_CLIENT): New.
4283 (PTA_ICELAKE_SERVER): New.
4284 (ix86_option_override_internal): Split up icelake on icelake client and
4286 (get_builtin_code_for_version): Ditto.
4287 (fold_builtin_cpu): Ditto.
4288 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4289 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4290 * config/i386/i386.h (processor_type): Ditto.
4291 * doc/invoke.texi: Ditto.
4293 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4296 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4297 INTEGER_CST offset, add it together with bitpos / 8 and
4298 sign extend based on POINTER_SIZE.
4302 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4305 * config/i386/constraints.md (Yc): New register constraint.
4306 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4307 Use Yc constraint for alternative 2 of operand 0. Remove
4308 preferred_for_speed attribute.
4310 2018-03-14 Richard Biener <rguenther@suse.de>
4312 PR tree-optimization/84830
4313 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4314 with the old one to avoid oscillations.
4316 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4319 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4321 (assign_by_spills): Return a flag of reload assignment failure.
4322 Do not process the reload assignment failures. Do not spill other
4323 reload pseudos if they has the same reg class. Update n if
4325 (lra_assign): Add a return arg. Set up from the result of
4326 assign_by_spills call.
4327 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4328 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4329 usage_insns if it is not NULL.
4330 (spill_hard_reg_in_range): New function.
4331 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4332 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4333 function prototypes.
4334 (lra_assign): Change prototype.
4335 * lra.c (lra): Add code to deal with fails by splitting hard reg
4338 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4340 * config/riscv/riscv.opt (mrelax): New option.
4341 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4342 "norelax" when riscv_mrelax is disabled.
4343 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4345 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4348 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4349 reassociation for int modes.
4351 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4353 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4354 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4356 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4357 * config/aarch64/aarch64-sve.md
4358 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4359 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4360 (*extend<mode><Vwide>2): Rename to...
4361 (aarch64_sve_extend<mode><Vwide>2): ...this.
4362 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4363 renaming the old pattern to...
4364 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4366 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4367 define_expand, renaming the old pattern to...
4368 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4369 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4370 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4371 account when deciding which SVE instruction the optab should use.
4372 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4374 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4376 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4377 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4378 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4379 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4380 (tlsdesc_small_<mode>): Turn a define_expand and use
4381 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4382 (tlsdesc_small_advsimd_<mode>): ...this.
4383 (tlsdesc_small_sve_<mode>): New pattern.
4385 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4387 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4388 (UNSPEC_UMUL_HIGHPART): New constants.
4389 (MUL_HIGHPART): New int iteraor.
4390 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4391 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4393 (*<su>mul<mode>3_highpart): New define_insn.
4395 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4398 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4401 2018-03-13 Martin Liska <mliska@suse.cz>
4404 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4406 (sem_item_optimizer::~sem_item_optimizer): Release it.
4407 (sem_item_optimizer::merge_classes): Register variable aliases.
4408 (sem_item_optimizer::fixup_pt_set): New function.
4409 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4410 * ipa-icf.h: Declare new variables and functions.
4412 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4415 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4416 integer_pow2p@2 and test integer_pow2p in condition.
4417 (A < 0 ? C : 0): Similarly for @1.
4420 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4421 characters starting at p contain '\0' character, don't look beyond
4425 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4426 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4429 * reg-stack.c (change_stack): Change update_end var from int to
4430 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4431 also call set_block_for_insn on the newly added insns and rescan.
4434 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4435 on the last operand.
4438 * tree.c (stabilize_reference_1): Return save_expr (e) for
4439 STATEMENT_LIST even if it doesn't have side-effects.
4441 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4443 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4445 2018-03-12 Renlin Li <renlin.li@arm.com>
4447 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4448 aarch64_output_scalar_simd_mov_immediate.
4450 2018-03-12 Martin Sebor <msebor@redhat.com>
4452 PR tree-optimization/83456
4453 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4454 for perfectly overlapping calls to memcpy.
4455 (gimple_fold_builtin_memory_chk): Same.
4456 (gimple_fold_builtin_strcpy): Handle no-warning.
4457 (gimple_fold_builtin_stxcpy_chk): Same.
4458 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4460 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4462 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4463 parameter. Use it for SFmode.
4464 (rs6000_function_arg_advance_1): Adjust.
4465 (rs6000_function_arg): Adjust.
4466 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4468 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4470 PR rtl-optimization/84169
4471 PR rtl-optimization/84780
4472 * combine.c (can_combine_p): Check for a 2-insn combination whether
4473 the destination register is used between the two insns, too.
4475 2018-03-12 Richard Biener <rguenther@suse.de>
4477 PR tree-optimization/84803
4478 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4479 for refs DR analysis didn't process.
4481 2018-03-12 Richard Biener <rguenther@suse.de>
4483 PR tree-optimization/84777
4484 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4485 force-vectorize loops ignore whether we are optimizing for size.
4487 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4489 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4490 (TARGET_MD_ASM_ADJUST): Define.
4492 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4493 Kito Cheng <kito.cheng@gmail.com>
4494 Chung-Ju Wu <jasonwucj@gmail.com>
4496 * config/nds32/nds32.c (nds32_compute_stack_frame,
4497 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4498 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4499 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4500 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4501 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4502 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4503 * config/nds32/nds32.md (prologue, epilogue): Use macro
4504 NDS32_V3PUSH_AVAILABLE_P to do checking.
4506 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4509 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4510 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4511 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4512 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4513 addition of most attributes on !orig_type_die or the attribute not
4514 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4516 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4517 Chung-Ju Wu <jasonwucj@gmail.com>
4519 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4521 * config/nds32/nds32.opt (mvh): New option.
4523 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4524 Chung-Ju Wu <jasonwucj@gmail.com>
4526 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4528 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4529 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4532 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4533 Chung-Ju Wu <jasonwucj@gmail.com>
4535 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4537 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4538 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4540 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4541 Kito Cheng <kito.cheng@gmail.com>
4542 Chung-Ju Wu <jasonwucj@gmail.com>
4544 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4545 UNSPEC_FFMISM and UNSPEC_FLMISM.
4546 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4547 for ffb, ffmism and flmism.
4548 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4549 (unspec_ffmism): Ditto.
4550 (unspec_flmism): Ditto.
4551 (nds32_expand_builtin_impl): Check if string extension is available.
4552 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4553 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4555 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4558 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4561 * lra-assigns.c (assign_by_spills): Return a flag of reload
4562 assignment failure. Do not process the reload assignment
4563 failures. Do not spill other reload pseudos if they has the same
4565 (lra_assign): Add a return arg. Set up from the result of
4566 assign_by_spills call.
4567 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4568 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4569 usage_insns if it is not NULL.
4570 (spill_hard_reg_in_range): New function.
4571 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4572 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4573 function prototypes.
4574 (lra_assign): Change prototype.
4575 * lra.c (lra): Add code to deal with fails by splitting hard reg
4578 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4581 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4583 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4586 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4587 (dw_val_node): Add val_symbolic_view.
4588 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4589 (symview_upper_bound): New.
4590 (new_line_info_table): Initialize symviews_since_reset.
4591 (dwarf2out_source_line): Count symviews_since_reset and set
4592 symview_upper_bound.
4593 (dw_val_equal_p): Handle symview.
4594 (add_AT_symview): New.
4595 (print_dw_val): Handle symview.
4596 (attr_checksum, attr_checksum_ordered): Likewise.
4597 (same_dw_val_p, size_of_die): Likewise.
4598 (value_format, output_die): Likewise.
4599 (add_high_low_attributes): Use add_AT_symview for entry_view.
4600 (dwarf2out_finish): Reset symview_upper_bound, clear
4603 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4606 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4607 Add strict argument and use it.
4608 (rs6000_split_multireg_move): Update for new strict argument.
4609 (mem_operand_gpr): Disallow all non-offsettable addresses.
4610 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4612 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4615 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4616 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4617 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4620 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4621 decl, use remap_type if we want to use the type.
4623 2018-03-09 Martin Sebor <msebor@redhat.com>
4625 PR tree-optimization/84526
4626 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4628 (builtin_access::generic_overlap): Be prepared to handle non-array
4631 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4633 PR rtl-optimization/84682
4634 * lra-constraints.c (process_address_1): Check is_address flag
4635 for address constraints.
4636 (process_alt_operands): Likewise.
4637 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4638 preprocess_constraints.
4639 * recog.h (preprocess_constraints): Add oploc parameter.
4641 * recog.c (preprocess_constraints): Test address_operand for
4642 CT_ADDRESS constraints.
4644 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4647 * lra-assigns.c (assign_by_spills): Return a flag of reload
4648 assignment failure. Do not process the reload assignment
4649 failures. Do not spill other reload pseudos if they has the same
4651 (lra_assign): Add a return arg. Set up from the result of
4652 assign_by_spills call.
4653 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4654 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4655 usage_insns if it is not NULL.
4656 (spill_hard_reg_in_range): New function.
4657 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4658 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4659 function prototypes.
4660 (lra_assign): Change prototype.
4661 * lra.c (lra): Add code to deal with fails by splitting hard reg
4664 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4667 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4668 Accept complain bool parameter. Only emit errors if it is true.
4669 (arm_parse_cpu_option_name): Likewise.
4670 (arm_target_thumb_only): Adjust callers of the above.
4671 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4672 prototype to take a default true bool parameter.
4673 (arm_parse_arch_option_name): Likewise.
4675 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4676 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4680 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4681 * configure: Regenerate.
4682 * configure.ac ("linker --version-script option"): New.
4683 ("linker soname option"): New.
4685 2018-03-09 Richard Biener <rguenther@suse.de>
4687 PR tree-optimization/84775
4688 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4689 immediate uses of predicate stmts and mark them modified.
4692 PR tree-optimization/84178
4693 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4695 (version_loop_for_if_conversion): Delay update_ssa call.
4696 (tree_if_conversion): Delay update_ssa until after predicate
4699 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4702 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4703 when the function accesses prior frames.
4705 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4708 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4709 gen_llsym, otherwise call maybe_gen_llsym.
4712 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4713 has ',' character inside of it.
4715 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4718 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4719 as clobbering CC_REGNUM.
4721 2018-03-08 Richard Biener <rguenther@suse.de>
4724 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4725 (follow_copies_to_constant): Do not follow SSA names registered
4728 2018-03-08 Richard Biener <rguenther@suse.de>
4730 PR tree-optimization/84178
4731 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4733 (version_loop_for_if_conversion): Delay update_ssa call.
4734 (tree_if_conversion): Delay update_ssa until after predicate
4737 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4739 PR tree-optimization/84178
4740 * tree-if-conv.c (release_bb_predicate): Remove the
4741 the assertion that the stmts have NULL use_ops.
4742 Discard the statements, asserting that they haven't
4743 yet been added to a BB.
4745 2018-03-08 Richard Biener <rguenther@suse.de>
4747 PR tree-optimization/84746
4748 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4749 (phi_translate): Pass in destination ANTIC_OUT set.
4750 (phi_translate_1): Likewise. For a simplified result lookup
4751 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4752 (phi_translate_set): Adjust.
4753 (do_pre_regular_insertion): Likewise.
4754 (do_pre_partial_partial_insertion): Likewise.
4756 2018-03-08 Martin Liska <mliska@suse.cz>
4758 PR gcov-profile/84735
4759 * doc/gcov.texi: Document usage of profile files.
4760 * gcov-io.h: Document changes in the format.
4762 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4766 * dwarf2out.c (struct dw_line_info_table): Update comments for
4768 (FORCE_RESET_NEXT_VIEW): New.
4769 (FORCE_RESETTING_VIEW_P): New.
4770 (RESETTING_VIEW_P): Check for -1 too.
4771 (ZERO_VIEW_P): Likewise.
4772 (new_line_info_table): Force-reset next view.
4773 (dwarf2out_begin_function): Likewise.
4774 (dwarf2out_source_line): Simplify zero_view_p initialization.
4775 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4776 view directly. Omit view when omitting .loc at line 0.
4778 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4780 PR tree-optimization/84740
4781 * tree-switch-conversion.c (process_switch): Call build_constructors
4782 only if info.phi_count is non-zero.
4784 PR tree-optimization/84739
4785 * tree-tailcall.c (find_tail_calls): Check call arguments against
4786 DECL_ARGUMENTS (current_function_decl) rather than
4787 DECL_ARGUMENTS (func) when checking for tail recursion.
4789 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4791 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4792 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4793 Volker Reichelt's entry and add entries for people that perform
4794 GCC fuzzy testing and report numerous bugs.
4796 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4799 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4800 readonly data in sdata, if that is disabled.
4801 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4802 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4803 -mreadonly-in-sdata option.
4805 2018-03-07 Martin Sebor <msebor@redhat.com>
4807 PR tree-optimization/84468
4808 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4809 basic block when looking for nul assignment.
4811 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4814 * except.h (output_function_exception_table): Adjust prototype.
4815 * except.c (output_function_exception_table): Remove FNNAME parameter
4816 and add SECTION parameter. Ouput one part of the table at a time.
4817 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4818 the first part of the exception table and emit unwind directives.
4819 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4820 (i386_pe_seh_cold_init): Likewise.
4821 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4822 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4823 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4824 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4825 * config/i386/winnt.c: Include except.h.
4826 (struct seh_frame_state): Add reg_offset, after_prologue and
4827 in_cold_section fields.
4828 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4829 (i386_pe_seh_cold_init): New function.
4830 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4831 to seh->in_cold_section.
4832 (seh_emit_push): Record the offset of the push.
4833 (seh_emit_save): Record the offet of the save.
4834 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4835 Test seh->after_prologue to disregard the epilogue.
4836 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4837 (i386_pe_end_cold_function): New function.
4839 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4842 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4843 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4846 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4848 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4849 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4852 * multiple_target.c: Include tree-inline.h and intl.h.
4853 (expand_target_clones): Diagnose and fail if node->definition and
4854 !tree_versionable_function_p (node->decl).
4856 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4858 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4860 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4861 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4862 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4864 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4867 * combine.c (try_combine): Use reg_or_subregno instead of handling
4868 just paradoxical SUBREGs and REGs.
4870 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4872 * config/arc/arc.c (arc_finalize_pic): Remove function.
4873 (arc_must_save_register): We use single base PIC register, remove
4874 checks to save/restore the PIC register.
4875 (arc_expand_prologue): Likewise.
4876 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4878 (arc_verify_short): Likewise.
4879 (arc_attr_type): Likewise.
4880 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4881 (walk_stores): Likewise.
4882 (arc_address_cost): Make it static.
4883 (arc_verify_short): Likewise.
4884 (branch_dest): Likewise.
4885 (arc_attr_type): Likewise.
4886 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4887 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4888 (arc_final_prescan_insn): Remove inserting the nops due to
4889 hardware hazards. It is done in reorg step.
4890 (insn_length_variant_t): Remove.
4891 (insn_length_parameters_t): Likewise.
4892 (arc_insn_length_parameters): Likewise.
4893 (arc_get_insn_variants): Likewise.
4894 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4896 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4899 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4902 PR tree-optimization/84687
4903 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4905 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4907 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4909 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4910 Rename to ppc_speculation_barrier.
4911 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4912 __builtin_ppc_speculation_barrier.
4914 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4917 * combine.c (combine_simplify_rtx): Don't try to simplify if
4918 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4921 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4923 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4924 to 32 bytes when compiling for POWER9.
4926 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4929 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4930 regparm >= 3 with no arg reg available also for calls with
4931 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4934 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4936 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4938 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4941 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4943 2018-03-05 Richard Biener <rguenther@suse.de>
4945 PR tree-optimization/84486
4946 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4947 When inserting a __builtin_assume_aligned call set the LHS
4948 SSA name alignment info accordingly.
4950 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4952 PR tree-optimization/84114
4953 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4954 Avoid reassociation of FLOAT_MODE addition.
4956 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4958 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4959 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4960 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4961 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4962 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4963 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4964 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4966 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4967 __builtin_ia32_wbinvd): New builtins.
4968 (SPECIAL_ARGS2): New.
4969 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4970 (SPECIAL_ARGS2): New.
4971 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4972 (ix86_valid_target_attribute_inner_p): Ditto.
4973 (ix86_init_mmx_sse_builtins): Add special_args2.
4974 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4975 TARGET_WBNOINVD_P): New.
4976 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4977 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4978 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4979 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4980 * config/i386/pconfigintrin.h: New file.
4981 * config/i386/wbnoinvdintrin.h: Ditto.
4982 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4983 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4985 2018-03-05 Richard Biener <rguenther@suse.de>
4987 PR tree-optimization/84670
4988 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4990 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4991 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4992 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4993 doesn't grow if all successors (recursively) were visited at least
4996 2018-03-05 Richard Biener <rguenther@suse.de>
4998 PR tree-optimization/84650
4999 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5000 if executed in the loop pipeline.
5002 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5004 * doc/configfiles.texi (Configuration Files): Move info about
5005 conditionalizing $target-protos.h to...
5006 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5007 differs from $target-protos.h.
5009 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5010 Chung-Ju Wu <jasonwucj@gmail.com>
5012 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5013 * config/nds32/nds32-multiple.md (setmemsi): Define.
5014 * config/nds32/nds32-memory-manipulation.c
5015 (nds32_gen_dup_4_byte_to_word_value): New.
5016 (emit_setmem_word_loop): New.
5017 (emit_setmem_byte_loop): New.
5018 (nds32_expand_setmem_loop): New.
5019 (nds32_expand_setmem_loop_v3m): New.
5020 (nds32_expand_setmem_unroll): New.
5021 (nds32_expand_setmem): New.
5023 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5024 Chung-Ju Wu <jasonwucj@gmail.com>
5026 * config/nds32/nds32-memory-manipulation.c
5027 (nds32_emit_load_store): New.
5028 (nds32_emit_post_inc_load_store): New.
5029 (nds32_emit_mem_move): New.
5030 (nds32_emit_mem_move_block): New.
5031 (nds32_expand_movmemsi_loop_unknown_size): New.
5032 (nds32_expand_movmemsi_loop_known_size): New.
5033 (nds32_expand_movmemsi_loop): New.
5034 (nds32_expand_movmemsi_unroll): New.
5035 (nds32_expand_movmemqi): Rename ...
5036 (nds32_expand_movmemsi): ... to this.
5037 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5038 (movmemsi): ... to this.
5039 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5040 (nds32_expand_movmemsi): ... to this.
5042 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5043 Monk Chiang <sh.chiang04@gmail.com>
5044 Chung-Ju Wu <jasonwucj@gmail.com>
5046 * config/nds32/nds32-protos.h
5047 (nds32_expand_load_multiple): New arguments.
5048 (nds32_expand_store_multiple): Ditto.
5049 (nds32_valid_multiple_load_store): Rename ...
5050 (nds32_valid_multiple_load_store_p): ... to this.
5051 * config/nds32/nds32-memory-manipulation.c
5052 (nds32_expand_load_multiple): Refine implementation.
5053 (nds32_expand_store_multiple): Ditto.
5054 * config/nds32/nds32-multiple.md
5055 (load_multiple): Update nds32_expand_load_multiple interface.
5056 (store_multiple): Update nds32_expand_store_multiple interface.
5057 * config/nds32/nds32-predicates.c
5058 (nds32_valid_multiple_load_store): Rename ...
5059 (nds32_valid_multiple_load_store_p): ... to this and refine
5061 * config/nds32/predicates.md
5062 (nds32_load_multiple_and_update_address_operation): New predicate.
5063 (nds32_store_multiple_and_update_address_operation): New predicate.
5065 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5066 Chung-Ju Wu <jasonwucj@gmail.com>
5068 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5069 (combo): New attribute.
5070 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5072 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5074 * config/nds32/nds32.opt: Change -mcmodel= default value.
5076 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5077 Monk Chiang <sh.chiang04@gmail.com>
5078 Chung-Ju Wu <jasonwucj@gmail.com>
5080 * config/nds32/constants.md (unspec_element): New enum.
5081 * config/nds32/constraints.md (Umw): New constraint.
5082 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5083 * config/nds32/nds32-intrinsic.md: Likewise.
5084 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5085 (nds32_valid_smw_lwm_base_p): New.
5086 (nds32_output_smw_single_word): New.
5087 (nds32_output_lmw_single_word): New.
5088 (nds32_expand_unaligned_load): New.
5089 (nds32_expand_unaligned_store): New.
5090 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5091 (nds32_output_smw_single_word): Declare.
5092 (nds32_output_lmw_single_word): Declare.
5093 (nds32_expand_unaligned_load): Declare.
5094 (nds32_expand_unaligned_store): Declare.
5095 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5096 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5097 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5098 NDS32_BUILTIN_UASTORE_DW.
5099 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5102 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5103 Kito Cheng <kito.cheng@gmail.com>
5104 Chung-Ju Wu <jasonwucj@gmail.com>
5106 * config/nds32/nds32-intrinsic.c
5107 (nds32_expand_builtin_null_ftype_reg): Delete.
5108 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5109 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5110 (nds32_read_argument): New.
5111 (nds32_legitimize_target): Ditto.
5112 (nds32_legitimize_argument): Ditto.
5113 (nds32_check_constant_argument): Ditto.
5114 (nds32_expand_unop_builtin): Ditto.
5115 (nds32_expand_unopimm_builtin): Ditto.
5116 (nds32_expand_binop_builtin): Ditto.
5117 (nds32_builtin_decl_impl): Ditto.
5118 (builtin_description): Ditto.
5119 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5120 (nds32_init_builtins_impl): Ditto.
5121 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5122 (nds32_builtin_decl): New.
5123 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5124 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5126 2018-03-02 Jeff Law <law@redhat.com>
5128 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5129 (redundant_insn, fill_simple_delay_slots): Likewise.
5130 (fill_slots_from_thread): Likewise.
5131 * resource.c (mark_referenced_resources): Likewise.
5132 (mark_set_resources, find_dead_or_set_registers): Likewise.
5134 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5136 * substring-locations.h (format_warning_va): Formatting fix for
5138 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5140 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5141 * substring-locations.c: Include intl.h.
5142 (format_warning_va): Turned into small wrapper around
5143 format_warning_n_va, renamed to ...
5144 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5145 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5147 (format_warning_at_substring_n): New function.
5148 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5149 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5150 format_warning_at_substring with just a shorter name instead of
5151 const function pointer.
5152 (fmtwarn_n): New function.
5153 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5154 appropriate, get rid of all the fmtstr temporaries, move conditionals
5155 with G_() wrapped string literals directly into fmtwarn arguments,
5156 cast dir.len to (int), formatting fixes.
5158 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5160 * doc/invoke.texi: Remove "Cilk Plus" references.
5162 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5163 Richard Biener <rguenther@suse.de>
5166 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5167 for error or warning attributes if CALL_FROM_THUNK_P is set.
5170 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5173 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5174 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5177 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5178 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5180 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5181 instead of -1U in last predictors element's probability member.
5183 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5186 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5187 arguments if they are comparable.
5189 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5191 PR tree-optimization/84634
5192 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5193 masks and masked_loop_p with a single loop_masks, making sure it's
5194 null for bb vectorization.
5196 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5198 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5199 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5200 loop->force_vectorize to check whether there is no alias.
5202 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5205 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5207 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5209 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5210 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5211 instead of a loop around prev_real_insn.
5212 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5216 * config/i386/i386.c (ix86_print_operand): Use conditional
5217 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5220 2018-03-02 Richard Biener <rguenther@suse.de>
5222 PR tree-optimization/84427
5223 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5224 (bitmap_set_subtract_values): Rewrite to handle multiple
5227 (prune_clobbered_mems): Likewise.
5228 (phi_translate): Take edge instead of pred/phiblock.
5229 (phi_translate_1): Likewise.
5230 (phi_translate_set): Likewise. Insert all translated
5231 exprs for a value into the set, keeping possibly multiple
5232 expressions per value.
5233 (compute_antic_aux): Adjust for phi_translate changes.
5234 When intersecting union the expressions and prune those
5235 not in the final value set, keeping possibly multiple
5236 expressions per value. Do not use value-insertion
5237 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5238 all expressions. Add verification that the value-sets
5239 only shrink during iteration.
5240 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5241 (do_pre_regular_insertion): Likewise.
5242 (do_pre_partial_partial_insertion): Likewise.
5244 2018-03-02 Richard Biener <rguenther@suse.de>
5247 * config/darwin.c (saved_debug_info_level): New static global.
5248 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5249 (darwin_asm_lto_end): Restore debug info generation settings.
5251 2018-03-01 Martin Liska <mliska@suse.cz>
5254 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5257 2018-03-01 Richard Biener <rguenther@suse.de>
5260 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5261 type annotation with LTO when debug was disabled at compile-time.
5263 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5265 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5267 (mips_final_postscan_insn): Likewise.
5269 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5271 PR rtl-optimization/84528
5272 * alias.c (init_alias_target): Add commentary.
5273 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5274 a unique base value if the frame pointer is not eliminated
5275 to the stack pointer.
5277 2018-03-01 Tom de Vries <tom@codesourcery.com>
5279 PR rtl-optimization/83327
5280 * lra-int.h (hard_regs_spilled_into): Declare.
5281 * lra.c (hard_regs_spilled_into): Define.
5282 (init_reg_info): Init hard_regs_spilled_into.
5283 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5284 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5285 (process_bb_lives): Handle hard_regs_spilled_into.
5286 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5287 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5289 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5291 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5292 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5293 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5294 * config/rs6000/aix72.h: New file.
5296 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5298 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5299 instead of warning_at with conditional singular and plural messages
5303 * stor-layout.c (update_alignment_for_field): For
5304 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5305 && !DECL_PACKED (field), do the alignment update, just use
5306 only desired_align instead of MAX (type_align, desired_align)
5308 (place_field): Don't do known_align < desired_align handling
5309 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5310 is non-NULL, instead do it after rli->prev_field handling and
5311 only if not within a bitfield word. For DECL_PACKED (field)
5312 use type_align of BITS_PER_UNIT.
5314 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5316 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5317 superfluous parentheses and trailing spaces.
5319 2018-02-28 Richard Biener <rguenther@suse.de>
5321 PR tree-optimization/84584
5322 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5323 SCoPs with fake exit edge.
5325 2018-02-28 Martin Liska <mliska@suse.cz>
5328 * timevar.c (timer::print): Fix format to properly print 100%
5331 2018-02-28 Richard Biener <rguenther@suse.de>
5334 * genmatch.c (capture_info::walk_match): Do not mark
5335 captured expressions without operands as expr_p given
5336 they act more like predicates and should be subject to
5337 "lost tail" side-effect preserving.
5339 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5341 PR rtl-optimization/81611
5342 * auto-inc-dec.c (attempt_change): Move dead note from
5343 mem_insn if it's the next use of regno
5344 (find_address): Take address use of reg holding
5345 non-incremented value. Add parm to limit search to the named
5347 (merge_in_block): Attempt to use a mem insn that is the next
5348 use of the original regno.
5350 2018-02-27 Martin Sebor <msebor@redhat.com>
5353 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5354 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5356 2018-02-27 Martin Sebor <msebor@redhat.com>
5358 PR translation/84207
5359 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5360 n argument to unsigned HOST_WIDE_INT.
5361 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5362 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5363 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5364 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5366 2018-02-27 Richard Biener <rguenther@suse.de>
5368 PR tree-optimization/84512
5369 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5370 Do not use the estimate returned from record_stmt_cost for
5371 the scalar iteration cost but sum properly using add_stmt_cost.
5373 2018-02-27 Richard Biener <rguenther@suse.de>
5375 PR tree-optimization/84466
5376 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5377 Adjust last change to less strictly validate use operands.
5379 2018-02-27 Martin Liska <mliska@suse.cz>
5381 PR gcov-profile/84548
5382 * gcov.c (process_file): Allow partial overlap and consider it
5383 also as group functions.
5384 (output_lines): Properly calculate range of lines for a group.
5386 2018-02-27 Martin Liska <mliska@suse.cz>
5388 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5389 'ggc' suffixes. Change first column width.
5390 (timer::print): Fix formatting of the column.
5392 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5394 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5395 preserve inline entry blocks for the sake of debug inline
5396 entry point markers alone.
5397 (remove_unused_locals): Suggest in comments a better place to
5398 force the preservation of inline entry blocks that are
5399 otherwise unused, but do not preserve them.
5401 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5403 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5405 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5408 * config/i386/constraints.md (Bs): Replace
5409 ix86_indirect_branch_register with
5410 TARGET_INDIRECT_BRANCH_REGISTER.
5412 * config/i386/i386.md (indirect_jump): Likewise.
5413 (tablejump): Likewise.
5414 (*sibcall_memory): Likewise.
5415 (*sibcall_value_memory): Likewise.
5416 Peepholes of indirect call and jump via memory: Likewise.
5417 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5418 (*sibcall_value_GOT_32): Likewise.
5419 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5420 (GOT_memory_operand): Likewise.
5421 (call_insn_operand): Likewise.
5422 (sibcall_insn_operand): Likewise.
5423 (GOT32_symbol_operand): Likewise.
5424 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5426 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5428 PR rtl-optimization/83496
5429 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5430 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5431 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5432 redundant insn, if any.
5433 (relax_delay_slots): Likewise.
5434 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5436 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5438 PR tree-optimization/83965
5439 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5440 that grouped statements are part of a reduction chain. Return
5441 true if the statement is not marked as a reduction itself but
5443 (vect_recog_dot_prod_pattern): Don't check whether the statement
5444 is part of a group here.
5445 (vect_recog_sad_pattern): Likewise.
5446 (vect_recog_widen_sum_pattern): Likewise.
5448 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5451 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5453 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5456 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5458 (ix86_output_indirect_function_return): New prototype.
5459 (ix86_split_simple_return_pop_internal): Likewise.
5460 * config/i386/i386.c (indirect_return_via_cx): New.
5461 (indirect_return_via_cx_bnd): Likewise.
5462 (indirect_thunk_name): Handle return va CX_REG.
5463 (output_indirect_thunk_function): Create alias for
5464 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5465 (ix86_output_indirect_jmp): Remove the bool argument.
5466 (ix86_output_indirect_function_return): New function.
5467 (ix86_split_simple_return_pop_internal): Likewise.
5468 * config/i386/i386.md (*indirect_jump): Don't pass false
5469 to ix86_output_indirect_jmp.
5470 (*tablejump_1): Likewise.
5471 (simple_return_pop_internal): Change it to define_insn_and_split.
5472 Call ix86_split_simple_return_pop_internal to split it for
5474 (simple_return_indirect_internal): Call
5475 ix86_output_indirect_function_return instead of
5476 ix86_output_indirect_jmp.
5478 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5481 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5482 memset and value initialization afterwards.
5484 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5486 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5488 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5491 * common/config/aarch64/aarch64-common.c
5492 (aarch_option_optimization_table[]): Switch
5493 off fomit-frame-pointer
5495 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5496 Chung-Ju Wu <jasonwucj@gmail.com>
5498 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5500 (store_multiple): Ditto.
5502 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5504 * config.gcc: Add --with-cpu support for nds32 target.
5505 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5506 * config/nds32/nds32.opt: Add -mcpu= option.
5508 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5510 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5511 isel=yes): Warn for these deprecated options.
5513 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5515 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5516 ISA_2_5_MASKS_EMBEDDED.
5518 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5520 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5521 p->max as pointers rather than using iterative_hash_expr.
5523 2018-02-23 Carl Love <cel@us.ibm.com>
5525 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5526 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5528 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5529 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5530 P8V_BUILTIN_VEC_VUNSIGNED2.
5532 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5535 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5536 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5538 * lra-constraints.c (curr_insn_transform): Set up
5539 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5541 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5542 * lra-eliminations.c (spill_pseudos): Ditto.
5543 (process_insn_for_elimination): Ditto.
5544 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5545 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5546 LRA_NON_CLOBBERED_ALT.
5548 2018-02-22 Martin Sebor <msebor@redhat.com>
5550 PR tree-optimization/84480
5551 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5552 to maybe_diag_stxncpy_trunc. Call it.
5553 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5554 from gimple_fold_builtin_strcpy. Print inlining stack.
5555 (handle_builtin_stxncpy): Print inlining stack.
5556 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5558 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5561 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5562 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5563 and -fcheck-pointer-bounds are used together.
5564 (indirect_thunk_prefix): New enum.
5565 (indirect_thunk_need_prefix): New function.
5566 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5567 "_nt" instead of "_bnd" for NOTRACK prefix.
5568 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5569 (output_indirect_thunk_function): Likewise.
5571 (ix86_code_end): Update output_indirect_thunk_function calls.
5572 (ix86_output_indirect_branch_via_reg): Replace
5573 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5574 (ix86_output_indirect_branch_via_push): Likewise.
5575 (ix86_output_function_return): Likewise.
5576 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5577 incompatible with -fcf-protection=branch and
5578 -fcheck-pointer-bounds.
5580 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5583 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5584 Change gcc_assert call to output_operand_lossage.
5586 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5588 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5590 2018-02-22 DJ Delorie <dj@redhat.com>
5591 Sebastian Perta <sebastian.perta@renesas.com>
5592 Oleg Endo <olegendo@gcc.gnu.org>
5594 * config/rx/rx.c (rx_rtx_costs): New function.
5595 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5597 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5599 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5601 2018-02-22 Martin Liska <mliska@suse.cz>
5604 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5605 Add "native" as a possible value.
5607 2018-02-22 Martin Liska <mliska@suse.cz>
5610 * config/i386/i386.c (ix86_option_override_internal):
5611 Add "native" as a possible value for -march and -mtune.
5613 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5616 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5617 to all type variants.
5619 PR tree-optimization/84503
5620 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5621 width as info->bitpos + info->bitsize - start.
5622 (merged_store_group::merge_overlapping): Simplify width computation.
5623 (check_no_overlap): New function.
5624 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5625 start + width and last_order of the group, fail if check_no_overlap
5627 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5628 to group if check_no_overlap fails.
5630 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5632 * config/rs6000/altivec.md: Delete contraint arguments to
5633 define_expand, define_split, and define_peephole2, and in
5634 define_insn_and_split if always unused.
5635 * config/rs6000/darwin.md: Ditto.
5636 * config/rs6000/dfp.md: Ditto.
5637 * config/rs6000/rs6000.md: Ditto.
5638 * config/rs6000/sync.md: Ditto.
5639 * config/rs6000/vector.md: Ditto.
5640 * config/rs6000/vsx.md: Ditto.
5642 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5644 * config/rs6000/altivec.md: Write output control strings as braced
5645 blocks instead of double-quoted strings.
5646 * config/rs6000/darwin.md: Ditto.
5647 * config/rs6000/rs6000.md: Ditto.
5648 * config/rs6000/vector.md: Ditto.
5649 * config/rs6000/vsx.md: Ditto.
5651 2018-02-21 Jason Merrill <jason@redhat.com>
5653 PR c++/84314 - ICE with templates and fastcall attribute.
5654 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5656 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5658 * ipa-cp.c (determine_versionability): Fix comment typos.
5660 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5663 * ipa-cp.c (determine_versionability): Do not version functions caling
5666 2018-02-21 Martin Liska <mliska@suse.cz>
5669 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5670 Add "native" as a possible value.
5671 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5672 the macro when native cpu detection is available.
5674 2018-02-21 Martin Liska <mliska@suse.cz>
5677 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5678 Add "native" as a possible value.
5679 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5680 when native cpu detection is available.
5682 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5683 Martin Sebor <msebor@redhat.com>
5685 PR tree-optimization/84478
5686 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5688 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5689 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5690 support which is conservatively correct, for 2 only stay conservative
5691 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5692 argument to the 2 argument get_range_strlen, adjust 6 arg
5693 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5695 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5696 (gimple_fold_builtin_strlen): Pass true as last argument to
5699 2018-02-20 Martin Sebor <msebor@redhat.com>
5702 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5703 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5704 (builtin_memref::builtin_memref): Factor out parts into
5705 set_base_and_offset and call it.
5707 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5710 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5711 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5712 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5713 search at the associated MODE_INT.
5715 2018-02-20 Jeff Law <law@redhat.com>
5718 PR tree-optimization/81592
5720 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5721 for range data rather than using global data.
5722 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5723 range data rather than using global data.
5724 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5725 pass it to children as needed.
5726 (struct directive::fmtresult): Similarly.
5727 (struct directive::set_width): Similarly.
5728 (struct directive::set_precision): Similarly.
5729 (format_integer, format_directive, parse_directive): Similarly.
5730 (format_none): Accept unnamed vr_values parameter.
5731 (format_percent, format_floating, format_character): Similarly.
5732 (format_string, format_plain): Similarly.
5733 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5734 the EVRP range analyzer for range data rather than using global data.
5735 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5736 gimple-ssa-evrp-analyze.h
5737 (class sprintf_dom_walker): Add after_dom_children member function.
5738 Add evrp_range_analyzer member.
5739 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5740 range analyzer as needed.
5741 (sprintf_dom_walker::after_dom_children): New member function.
5742 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5744 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5745 (evrp_range_analyzer::pop_to_marker): Likewise.
5747 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5749 PR tree-optimization/84419
5750 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5751 with the required type if its current type is compatible but
5754 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5757 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5758 after vectorization.
5760 2018-02-20 Martin Liska <mliska@suse.cz>
5763 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5764 possible values if we don't have a hint.
5766 2018-02-20 Martin Liska <mliska@suse.cz>
5770 * final.c (shorten_branches): Build align_tab array with one
5772 * opts.c (finish_options): Add alignment option limit check.
5773 (MAX_CODE_ALIGN): Likewise.
5774 (MAX_CODE_ALIGN_VALUE): Likewise.
5775 * doc/invoke.texi: Document maximum allowed option value for
5776 all -falign-* options.
5778 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5781 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5782 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5783 * var-tracking.c (emit_note_insn_var_location): Remove all references
5784 to NOTE_INSN_CALL_ARG_LOCATION.
5785 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5786 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5787 Use copy_rtx_if_shared.
5788 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5789 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5790 (dwarf2out_var_location): Remove handling of
5791 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5793 * final.c (final_scan_insn): Remove all references to
5794 NOTE_INSN_CALL_ARG_LOCATION.
5795 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5796 before dumping final insns.
5797 * except.c (emit_note_eh_region_end): Remove all references to
5798 NOTE_INSN_CALL_ARG_LOCATION.
5799 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5800 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5801 * config/arc/arc.c (hwloop_optimize): Likewise.
5802 * config/arm/arm.c (create_fix_barrier): Likewise.
5803 * config/s390/s390.c (s390_chunkify_start): Likewise.
5804 * config/sh/sh.c (find_barrier): Likewise.
5805 * config/i386/i386.c (rest_of_insert_endbranch,
5806 ix86_seh_fixup_eh_fallthru): Likewise.
5807 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5808 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5809 * config/frv/frv.c (frv_function_prologue): Likewise.
5810 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5812 (note_outside_basic_block_p): Remove all references to
5813 NOTE_INSN_CALL_ARG_LOCATION.
5814 * gengtype.c (adjust_field_rtx_def): Likewise.
5815 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5817 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5818 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5821 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5824 PR tree-optimization/84452
5825 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5826 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5829 2018-02-19 Martin Liska <mliska@suse.cz>
5832 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5834 2018-02-19 Martin Liska <mliska@suse.cz>
5835 Richard Sandiford <richard.sandiford@linaro.org>
5837 PR tree-optimization/82491
5838 * gimple-fold.c (get_base_constructor): Make earlier bail out
5841 2018-02-19 Carl Love <cel@us.ibm.com>
5843 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5844 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5846 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5847 P8V_BUILTIN_VEC_NEG.
5849 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5851 * config/rl78/rl78.md (movdf): New define expand.
5853 2018-02-19 Martin Liska <mliska@suse.cz>
5856 * doc/invoke.texi: Fix typo.
5857 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5859 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5861 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5862 handle rs6000_single_float and rs6000_double_float specially for
5865 2018-02-16 Jeff Law <law@redhat.com>
5867 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5870 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5873 * ipa-inline.c (inline_small_functions): Fix a typo.
5875 2018-02-16 Nathan Sidwell <nathan@acm.org>
5877 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5879 2018-02-16 Carl Love <cel@us.ibm.com>
5881 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5882 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5883 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5884 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5885 expansion to P8V_BUILTIN_VEC_FLOAT2.
5887 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5889 PR rtl-optimization/70023
5890 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5891 src_regno into account.
5893 2018-02-16 Carl Love <cel@us.ibm.com>
5895 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5896 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5897 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5898 * config/rs6000/rs6000.c: Remove case statements for
5899 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5900 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5901 and P9V_BUILTIN_VEC_VINSERT4B.
5902 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5903 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5904 * config/rs6000/vsx.md:
5905 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5908 2018-02-16 Carl Love <cel@us.ibm.com>
5910 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5912 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5914 * config/rs6000/rs6000-c.c: Add the definitions for
5915 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5916 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5917 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5918 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5919 definition for insert4b and define insn *insert3b_internal.
5920 * doc/extend.texi: Add documentation for vec_extract4b.
5922 2018-02-16 Nathan Sidwell <nathan@acm.org>
5924 * doc/extend.texi (Backwards Compatibility): Mention friend
5925 injection. Note for-scope is deprecated.
5926 * doc/invoke.texi (-ffriend-injection): Deprecate.
5928 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5930 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5931 that moved to I2, also allow destinations that are a paradoxical
5932 subreg (instead of a normal reg).
5934 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5937 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5940 2018-02-16 Richard Biener <rguenther@suse.de>
5942 PR tree-optimization/84037
5943 PR tree-optimization/84016
5945 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5946 Adjust vec_construct for the fact we need additional higher latency
5947 128bit inserts for AVX256 and AVX512 vector builds.
5948 (ix86_add_stmt_cost): Scale vector construction cost for
5951 2018-02-16 Richard Biener <rguenther@suse.de>
5953 PR tree-optimization/84417
5954 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5955 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5956 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5958 2018-02-16 Martin Liska <mliska@suse.cz>
5961 * internal-fn.def (ASAN_CHECK): Set proper flags.
5962 (ASAN_MARK): Likewise.
5964 2018-02-16 Julia Koval <julia.koval@intel.com>
5966 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5967 from PTA_CANNONLAKE.
5969 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5972 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5973 Use ++iter rather than iter++ for std::list iterators.
5974 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5975 defer deleting them until all nodes in the forest are processed. Do
5976 free even leaf nodes. Change to_process into auto_vec.
5979 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5980 * vec.h (vec_default_construct): Use memset instead of placement new
5981 if BROKEN_VALUE_INITIALIZATION is defined.
5982 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5983 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5986 PR rtl-optimization/83723
5987 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5988 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5989 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5991 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5993 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5995 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5997 PR rtl-optimization/81443
5998 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5999 from inner REGs to paradoxical SUBREGs.
6001 2018-02-16 Richard Biener <rguenther@suse.de>
6003 PR tree-optimization/84399
6004 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6005 For operands we can analyze at their definition make sure we can
6006 analyze them at each use as well.
6008 2018-02-16 Richard Biener <rguenther@suse.de>
6010 PR tree-optimization/84190
6011 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6012 volatile accesses if the decl isn't volatile.
6014 2018-02-15 Jason Merrill <jason@redhat.com>
6016 PR c++/84314 - ICE with templates and fastcall attribute.
6017 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6018 TYPE_CANONICAL on an existing type.
6020 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6022 PR tree-optimization/84383
6023 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6024 dstoff nor call operand_equal_p if dstbase is NULL.
6026 PR tree-optimization/84334
6027 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6028 also a CONSTANT_CLASS_P, punt.
6030 2018-02-14 Jim Wilson <jimw@sifive.com>
6032 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6033 first SMALL_OPERAND check. New local min_second_step. Move assert
6034 to where locals are set. Add TARGET_RVC support.
6035 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6037 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6039 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6041 2018-02-14 Martin Sebor <msebor@redhat.com>
6043 PR tree-optimization/83698
6044 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6045 arrays constrain the offset range to their bounds.
6046 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6047 (builtin_access::overlap): Avoid setting the size of overlap if it's
6049 (maybe_diag_overlap): Also consider arrays when deciding what values
6050 of offsets to include in diagnostics.
6052 2018-02-14 Martin Sebor <msebor@redhat.com>
6055 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6056 that correspond to the kind of a declaration.
6058 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6061 * config/pa/pa.md: Load address of PIC label using the linkage table
6062 if the label is nonlocal.
6064 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6066 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6067 warning message if user requests -maltivec=be.
6068 * doc/invoke.texi: Document deprecation of -maltivec=be.
6070 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6073 * config/rs6000/rs6000-c.c: Update definitions for
6074 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6075 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6077 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6080 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6081 add _get_ssp intrinsics. Remove argument from
6082 __builtin_ia32_rdssp[d|q].
6083 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6084 * config/i386/i386-builtin.def: Remove argument from
6085 __builtin_ia32_rdssp[d|q].
6086 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6087 ix86_expand_special_args_builtin for _rdssp[d|q].
6088 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6089 Clear register before usage.
6090 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6091 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6093 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6095 PR tree-optimization/84357
6096 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6097 operand 1 of an ARRAY_REF too.
6099 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6102 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6103 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6105 (set_of_reg): New struct.
6106 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6107 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6108 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6110 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6111 Split into bitclr, bitset, bitinvert patterns if appropriate.
6112 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6113 use rx_fuse_in_memory_bitop.
6114 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6115 to named insn, correct maximum insn length.
6117 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6120 * machmode.def: Define a complex mode for PARTIAL_INT.
6121 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6123 * doc/rtl.texi: Document CSPImode.
6124 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6126 (msp430_hard_regno_nregs_with_padding): Likewise.
6128 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6131 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6133 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6135 PR rtl-optimization/84169
6136 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6137 we generated a parallel as new i3 and we split that to new i2 and i3
6138 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6139 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6140 those to i2, not i1. Partially rewrite this scan code.
6142 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6145 * stor-layout.c (place_field): For variable length fields, adjust
6146 offset_align afterwards not just based on the field's alignment,
6147 but also on the size.
6150 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6151 of exps and logs in the use_exp2 case.
6153 2018-02-13 Jeff Law <law@redhat.com>
6155 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6158 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6161 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6165 * common.opt (gas-loc-support, gas-locview-support): New.
6166 (ginline-points, ginternal-reset-location-views): New.
6167 * doc/invoke.texi: Document them. Use @itemx where intended.
6168 (gvariable-location-views): Adjust.
6169 * target.def (reset_location_view): New.
6170 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6171 (TARGET_RESET_LOCATION_VIEW): New.
6172 * doc/tm.texi: Rebuilt.
6173 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6174 (dwarf2out_default_as_locview_support): New.
6175 (output_asm_line_debug_info): Use option variables.
6176 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6177 (output_loc_list): Likewise.
6178 (add_high_low_attributes): Check option variables.
6179 Don't output entry view attribute in strict mode.
6180 (gen_inlined_subroutine_die): Check option variables.
6181 (dwarf2out_inline_entry): Likewise.
6182 (init_sections_and_labels): Likewise.
6183 (dwarf2out_early_finish): Likewise.
6184 (maybe_reset_location_view): New, from...
6185 (dwarf2out_var_location): ... here. Call it.
6186 * debug.h (dwarf2out_default_as_loc_support): Declare.
6187 (dwarf2out_default_as_locview_support): Declare.
6188 * hooks.c (hook_int_rtx_insn_0): New.
6189 * hooks.h (hook_int_rtx_insn_0): Declare.
6190 * toplev.c (process_options): Take -gas-loc-support and
6191 -gas-locview-support from dwarf2out. Enable
6192 -gvariable-location-views by default only with locview
6193 assembler support. Enable -ginternal-reset-location-views by
6194 default only if the target defines the corresponding hook.
6195 Enable -ginline-points by default if location views are
6196 enabled; force it disabled if statement frontiers are
6198 * tree-inline.c (expand_call_inline): Check option variables.
6199 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6201 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6203 PR tree-optimization/84321
6204 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6205 handling. Also check whether the anti-range contains any values
6206 that satisfy the mask; switch to a VR_RANGE if not.
6208 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6211 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6213 2018-02-13 Martin Jambor <mjambor@suse.cz>
6216 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6217 of call statements, also set location of a load to a temporary.
6219 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6221 * config/rl78/rl78.c (add_vector_labels): New function.
6222 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6223 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6224 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6225 which checks that no arguments are passed.
6226 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6227 * doc/extend.texi: Documentation for the new attribute.
6229 2018-02-13 Andreas Schwab <schwab@suse.de>
6231 * config/riscv/linux.h (CPP_SPEC): Define.
6233 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6236 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6237 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6238 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6239 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6240 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6241 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6242 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6243 temporarily for AES and PCLMUL builtins.
6245 PR tree-optimization/84339
6246 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6247 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6251 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6252 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6253 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6255 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6257 * omp-simd-clone.h: New file.
6258 * omp-simd-clone.c: Include omp-simd-clone.h.
6259 (expand_simd_clones): No longer static.
6260 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6261 cgraph.h and omp-simd-clone.h.
6262 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6263 (vect_recog_widen_shift_pattern): Formatting fix.
6264 (vect_pattern_recog_1): Don't check optab for calls.
6267 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6268 operands[2] into a REG before using gen_lowpart on it.
6270 2018-02-12 Jeff Law <law@redhat.com>
6273 * config/sh/sh.c (find_barrier): Consider a sibling call
6276 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6277 successfully back substituting a reg.
6279 2018-02-12 Richard Biener <rguenther@suse.de>
6281 PR tree-optimization/84037
6282 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6283 parameter, move visited init to caller.
6284 (vect_slp_analyze_operations): Separate cost from validity
6285 check, initialize visited once for all instances.
6286 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6288 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6290 (vect_model_promotion_demotion_cost): Likewise.
6291 (vectorizable_bswap): Guard cost modeling with !slp_node
6292 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6294 (vectorizable_call): Likewise.
6295 (vectorizable_conversion): Likewise.
6296 (vectorizable_assignment): Likewise.
6297 (vectorizable_shift): Likewise.
6298 (vectorizable_operation): Likewise.
6299 (vectorizable_store): Likewise.
6300 (vectorizable_load): Likewise.
6301 (vectorizable_condition): Likewise.
6302 (vectorizable_comparison): Likewise.
6304 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6307 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6308 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6309 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6311 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6314 * params.def (inline-min-speedup): Increase from 8 to 15.
6315 (max-inline-insns-auto): Decrease from 40 to 30.
6316 * ipa-split.c (consider_split): Add some buffer for function to
6317 be considered inlining candidate.
6318 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6321 2018-02-12 Richard Biener <rguenther@suse.de>
6323 PR tree-optimization/84037
6324 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6325 matched stmts if we cannot swap the non-matched ones.
6327 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6329 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6330 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6331 _mm_maskz_scalef_round_ss): New intrinsics.
6332 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6333 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6334 __builtin_ia32_scalefss_round): Remove.
6335 (__builtin_ia32_scalefsd_mask_round,
6336 __builtin_ia32_scalefss_mask_round): New intrinsics.
6337 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6338 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6339 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6340 "<round_constraint>")): Changed to ...
6341 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6342 "<round_scalar_constraint>")): ... this.
6343 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6344 %0, %1, %2<round_op3>}"): Changed to ...
6345 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6346 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6347 %2<round_scalar_mask_op3>}"): ... this.
6348 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6350 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6352 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6353 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6354 (_mm_maskz_sqrt_round_ss): New intrinsics.
6355 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6356 (__builtin_ia32_sqrtsd_mask_round)
6357 (__builtin_ia32_sqrtss_mask_round): New builtins.
6358 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6359 (__builtin_ia32_sqrtss_round): Remove.
6360 (__builtin_ia32_sqrtsd_mask_round)
6361 (__builtin_ia32_sqrtss_mask_round): New builtins.
6362 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6363 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6364 ((match_operand:VF_128 1 "vector_operand"
6365 "xBm,<round_constraint>")): Changed to ...
6366 ((match_operand:VF_128 1 "vector_operand"
6367 "xBm,<round_scalar_constraint>")): ... this.
6368 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6369 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6370 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6371 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6372 %<iptr>1<round_scalar_mask_op3>}): ... this.
6373 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6374 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6376 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6379 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6380 Cast vec_cmpeq result to correct type.
6381 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6382 Cast vec_cmpgt result to correct type.
6384 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6386 * final.c (final_scan_insn_1): Renamed from...
6387 (final_scan_insn): ... this. New wrapper, to recover
6388 seen from the outermost call in recursive ones.
6389 * config/sparc/sparc.c (output_return): Drop seen from call.
6390 (output_sibcall): Likewise.
6391 * config/visium/visium.c (output_branch): Likewise.
6393 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6395 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6398 2018-02-10 Alan Modra <amodra@gmail.com>
6401 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6402 Specify LR as an input.
6404 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6407 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6408 remove_member_access_dummy_vars): New functions.
6409 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6410 lower_omp_1, execute_lower_omp): Use them.
6412 PR rtl-optimization/84308
6413 * shrink-wrap.c (spread_components): Release todo vector.
6415 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6417 PR rtl-optimization/57193
6418 * ira-color.c (struct allocno_color_data): Add member
6419 conflict_allocno_hard_prefs.
6420 (update_conflict_allocno_hard_prefs): New.
6421 (bucket_allocno_compare_func): Add a preference based on
6422 conflict_allocno_hard_prefs.
6423 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6424 (color_allocnos): Remove a dead code. Initiate
6425 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6427 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6430 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6431 constraint from =wa to wa. Avoid a subreg on the output operand,
6432 instead use a pseudo and subreg it in a move.
6433 (p9_xxbrd_<mode>): Changed to ...
6434 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6435 (p9_xxbrd_v2df): New expander.
6436 (p9_xxbrw_<mode>): Changed to ...
6437 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6438 (p9_xxbrw_v4sf): New expander.
6440 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6442 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6444 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6447 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6448 multiply in 32-bit mode.
6449 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6450 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6453 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6455 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6456 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6457 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6458 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6460 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6463 * dwarf2out.c (is_trivial_indirect_ref): New function.
6464 (dwarf2out_late_global_decl): Do not generate a location
6465 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6466 and that are not defined in the current unit.
6468 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6470 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6471 instead of a libcall for UNORDERED.
6473 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6476 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6477 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6479 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6482 * config/s390/s390.c (s390_set_current_function): Invoke
6483 s390_indirect_branch_settings also if fndecl didn't change.
6485 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6487 * config/rs6000/rs6000.md (blockage): Set length to zero.
6489 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6491 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6493 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6496 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6497 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6501 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6502 PARALLEL incoming that failed vt_get_decl_and_offset check.
6505 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6506 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6508 (get_variable_section): For decls in named .bss* sections pass true as
6509 second argument to bss_initializer_p.
6511 2018-02-09 Marek Polacek <polacek@redhat.com>
6512 Jakub Jelinek <jakub@redhat.com>
6515 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6516 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6517 Sync some changes from cxx_fold_indirect_ref.
6519 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6521 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6523 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6524 (BLOCK_INLINE_ENTRY_LABEL): New.
6525 (dwarf2out_var_location): Disregard inline entry markers.
6526 (inline_entry_data): New struct.
6527 (inline_entry_data_hasher): New hashtable type.
6528 (inline_entry_data_hasher::hash): New.
6529 (inline_entry_data_hasher::equal): New.
6530 (inline_entry_data_table): New variable.
6531 (add_high_low_attributes): Add DW_AT_entry_pc and
6532 DW_AT_GNU_entry_view attributes if a pending entry is found
6533 in inline_entry_data_table. Add old entry_pc attribute only
6534 if debug nonbinding markers are disabled.
6535 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6536 markers are enabled.
6537 (block_within_block_p, dwarf2out_inline_entry): New.
6538 (dwarf2out_finish): Check that no entries remained in
6539 inline_entry_data_table.
6540 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6541 (final_scan_insn, notice_source_line): Likewise.
6542 (rest_of_clean_state): Skip inline entry markers.
6543 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6545 * gimple.c (gimple_build_debug_inline_entry): New.
6546 * gimple.h (enum gimple_debug_subcode): Add
6547 GIMPLE_DEBUG_INLINE_ENTRY.
6548 (gimple_build_debug_inline_entry): Declare.
6549 (gimple_debug_inline_entry_p): New.
6550 (gimple_debug_nonbind_marker_p): Adjust.
6551 * insn-notes.def (INLINE_ENTRY): New.
6552 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6553 inline entry marker notes.
6554 (print_insn): Likewise.
6555 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6556 (INSN_DEBUG_MARKER_KIND): Likewise.
6557 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6558 * tree-inline.c (expand_call_inline): Build and insert
6559 debug_inline_entry stmt.
6560 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6561 inline entry blocks early, if nonbind markers are enabled.
6562 (dump_scope_block): Dump fragment info.
6563 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6564 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6565 (gimple_build_debug_inline_entry): New.
6566 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6567 Enable/disable inline entry points too.
6568 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6569 (DEBUG_INSN): Describe inline entry markers.
6571 * common.opt (gvariable-location-views): New.
6572 (gvariable-location-views=incompat5): New.
6573 * config.in: Rebuilt.
6574 * configure: Rebuilt.
6575 * configure.ac: Test assembler for view support.
6576 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6577 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6578 * dwarf2out.c (var_loc_view): New typedef.
6579 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6580 (dwarf2out_locviews_in_attribute): New.
6581 (dwarf2out_locviews_in_loclist): New.
6582 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6583 (enum dw_line_info_opcode): Add LI_adv_address.
6584 (struct dw_line_info_table): Add view.
6585 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6586 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6587 (zero_view_p): New variable.
6588 (ZERO_VIEW_P): New macro.
6589 (output_asm_line_debug_info): New.
6590 (struct var_loc_node): Add view.
6591 (add_AT_view_list, AT_loc_list): New.
6592 (add_var_loc_to_decl): Add view param. Test it against last.
6593 (new_loc_list): Add view params. Record them.
6594 (AT_loc_list_ptr): Handle loc and view lists.
6595 (view_list_to_loc_list_val_node): New.
6596 (print_dw_val): Handle dw_val_class_view_list.
6597 (size_of_die): Likewise.
6598 (value_format): Likewise.
6599 (loc_list_has_views): New.
6600 (gen_llsym): Set vl_symbol too.
6601 (maybe_gen_llsym, skip_loc_list_entry): New.
6602 (dwarf2out_maybe_output_loclist_view_pair): New.
6603 (output_loc_list): Output view list or entries too.
6604 (output_view_list_offset): New.
6605 (output_die): Handle dw_val_class_view_list.
6606 (output_dwarf_version): New.
6607 (output_compilation_unit_header): Use it.
6608 (output_skeleton_debug_sections): Likewise.
6609 (output_rnglists, output_line_info): Likewise.
6610 (output_pubnames, output_aranges): Update version comments.
6611 (output_one_line_info_table): Output view numbers in asm comments.
6612 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6613 Call maybe_gen_llsym.
6614 (loc_list_from_tree_1): Adjust.
6615 (add_AT_location_description): Create view list attribute if
6616 needed, check it's absent otherwise.
6617 (convert_cfa_to_fb_loc_list): Adjust.
6618 (maybe_emit_file): Call output_asm_line_debug_info for test.
6619 (dwarf2out_var_location): Reset views as needed. Precompute
6620 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6621 attribute. Set view.
6622 (new_line_info_table): Reset next view.
6623 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6624 (dwarf2out_source_line): Likewise. Output view resets and labels to
6625 the assembler, or select appropriate line info opcodes.
6626 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6627 (optimize_string_length): Catch it. Adjust.
6628 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6629 dw_val_class_view_list, and remove it if no longer needed.
6630 (hash_loc_list): Hash view numbers.
6631 (loc_list_hasher::equal): Compare them.
6632 (optimize_location_lists): Check whether a view list symbol is
6633 needed, and whether the locview attribute is present, and
6634 whether they match. Remove the locview attribute if no longer
6636 (index_location_lists): Call skip_loc_list_entry for test.
6637 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6638 Use output_dwarf_version.
6639 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6640 (struct dw_val_node): Add val_view_list.
6641 * final.c (SEEN_NEXT_VIEW): New.
6642 (set_next_view_needed): New.
6643 (clear_next_view_needed): New.
6644 (maybe_output_next_view): New.
6645 (final_start_function): Rename to...
6646 (final_start_function_1): ... this. Take pointer to FIRST,
6647 add SEEN parameter. Emit param bindings in the initial view.
6648 (final_start_function): Reintroduce SEEN-less interface.
6649 (final): Rename to...
6650 (final_1): ... this. Take SEEN parameter. Output final pending
6651 next view at the end.
6652 (final): Reintroduce seen-less interface.
6653 (final_scan_insn): Output pending next view before switching
6654 sections or ending a block. Mark the next view as needed when
6655 outputting variable locations. Notify debug backend of section
6656 changes, and of location view changes.
6657 (rest_of_handle_final): Adjust.
6658 * toplev.c (process_options): Autodetect value for debug variable
6659 location views option. Warn on incompat5 without -gdwarf-5.
6660 * doc/invoke.texi (gvariable-location-views): New.
6661 (gvariable-location-views=incompat5): New.
6662 (gno-variable-location-views): New.
6664 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6666 PR tree-optimization/84136
6667 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6668 that the result of find_edge is non-NULL.
6670 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6673 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6674 storing integer register in SImode. Fix cost of 256 and 512
6675 byte aligned SSE register store.
6677 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6679 * config/i386/i386.c (ix86_multiplication_cost): Fix
6680 multiplication cost for TARGET_AVX512DQ.
6682 2018-02-08 Marek Polacek <polacek@redhat.com>
6684 PR tree-optimization/84238
6685 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6688 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6690 PR tree-optimization/84265
6691 * tree-vect-stmts.c (vectorizable_store): Don't treat
6692 VMAT_CONTIGUOUS accesses as grouped.
6693 (vectorizable_load): Likewise.
6695 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6697 PR tree-optimization/81635
6698 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6699 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6700 (test_round_for_mask): New functions.
6701 (wide_int_cc_tests): Call test_round_for_mask.
6702 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6703 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6704 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6705 range returned by get_range_info.
6707 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6710 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6711 * symtab.c: Include builtins.h
6712 (symtab_node::output_to_lto_symbol_table_p): Move here
6713 from lto-streamer-out.c:output_symbol_p.
6714 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6715 (output_symbol_p): Move all logic to symtab.c
6716 (produce_symtab): Update.
6718 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6720 * config/s390/s390-opts.h (enum indirect_branch): Define.
6721 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6722 (s390_indirect_branch_via_thunk)
6723 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6724 (enum s390_indirect_branch_type): Define.
6725 * config/s390/s390.c (struct s390_frame_layout, struct
6726 machine_function): Remove.
6727 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6728 (indirect_branch_table_label_no, indirect_branch_table_name):
6730 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6731 (enum s390_indirect_branch_option): Define.
6732 (s390_return_addr_from_memory): New function.
6733 (s390_handle_string_attribute): New function.
6734 (s390_attribute_table): Add new attribute handler.
6735 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6736 (s390_indirect_branch_via_thunk): New function.
6737 (s390_indirect_branch_via_inline_thunk): New function.
6738 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6739 sibling call optimization for non z10 compiles.
6740 (s390_emit_call): Force indirect branch target to be a single
6741 register. Add r1 clobber for non-z10 compiles.
6742 (s390_emit_epilogue): Emit return jump via return_use expander.
6743 (s390_reorg): Handle JUMP_INSNs as execute targets.
6744 (s390_option_override_internal): Perform validity checks for the
6745 new command line options.
6746 (s390_indirect_branch_attrvalue): New function.
6747 (s390_indirect_branch_settings): New function.
6748 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6749 (s390_output_indirect_thunk_function): New function.
6750 (s390_code_end): Implement target hook.
6751 (s390_case_values_threshold): Implement target hook.
6752 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6754 * config/s390/s390.h (struct s390_frame_layout)
6755 (struct machine_function): Move here from s390.c.
6756 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6757 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6758 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6759 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6760 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6761 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6762 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6763 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6764 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6765 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6766 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6767 (mnemonic attribute): Add values which aren't recognized
6769 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6770 pattern for branch conversion. Fix mnemonic attribute.
6771 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6772 indirect branch via thunk if requested.
6773 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6774 ("*indirect_jump"): Disable for branch conversion using out of
6776 ("indirect_jump_via_thunk<mode>_z10")
6777 ("indirect_jump_via_thunk<mode>")
6778 ("indirect_jump_via_inlinethunk<mode>_z10")
6779 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6780 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6781 ("casesi_jump_via_inlinethunk<mode>_z10")
6782 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6783 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6784 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6785 ("*indirect2_jump"): Disable for branch conversion.
6786 ("casesi_jump"): Turn into expander and expand patterns for branch
6788 ("return_use"): New expander.
6789 ("*return"): Emit return via thunk and rename it to ...
6790 ("*return<mode>"): ... this one.
6791 * config/s390/s390.opt: Add new options and and enum for the
6794 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6796 * lra-constraints.c (match_reload): Unconditionally use
6797 gen_lowpart_SUBREG, rather than selecting between that
6798 and equivalent gen_rtx_SUBREG code.
6800 2018-02-08 Richard Biener <rguenther@suse.de>
6802 PR tree-optimization/84233
6803 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6804 changed flag instead of boguously re-using phi_inserted.
6806 2018-02-08 Martin Jambor <mjambor@suse.cz>
6808 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6809 static local variables.
6811 2018-02-08 Richard Biener <rguenther@suse.de>
6813 PR tree-optimization/84278
6814 * tree-vect-stmts.c (vectorizable_store): When looking for
6815 smaller vector types to perform grouped strided loads/stores
6816 make sure the mode is supported by the target.
6817 (vectorizable_load): Likewise.
6819 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6821 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6822 Increase LDP/STP opportunities by adding adjacent callee-saves.
6824 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6826 PR rtl-optimization/84068
6827 PR rtl-optimization/83459
6828 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6830 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6832 PR tree-optimization/84224
6833 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6834 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6837 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6840 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6841 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6842 count, remove one USE.
6844 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6846 * doc/install.texi (Configuration): Document the
6847 --with-long-double-format={ibm,ieee} PowerPC configuration
6851 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6852 Convert from define_expand to be define_insn_and_split. Rework
6853 float/double/_Float128 conversions to QI/HI/SImode to work with
6854 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6855 conversions to QI/HImode types did a store and then a load to
6856 truncate the value. For conversions to VSX registers, don't split
6857 the insn, instead emit the code directly. Use the code iterator
6858 any_fix to combine signed and unsigned conversions.
6859 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6860 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6861 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6862 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6863 (fix_<mode>di2_hw): Likewise.
6864 (fixuns_<mode>di2_hw): Likewise.
6865 (fix_<mode>si2_hw): Likewise.
6866 (fixuns_<mode>si2_hw): Likewise.
6867 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6868 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6869 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6870 fix<uns>_trunc<SFDF:mode>si2_p8.
6871 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6873 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6874 (fix<uns>_<mode>_mem): Likewise.
6875 (fctiw<u>z_<mode>_mem): Likewise.
6876 (fix<uns>_<mode>_mem): Likewise.
6877 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6878 the register allocator from doing a direct move to the GPRs to do
6879 a store, and instead use the ISA 3.0 store byte/half-word from
6880 vector register instruction. For IEEE 128-bit floating point,
6881 also optimize stores of 32-bit ints.
6882 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6884 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6886 * genextract.c (push_pathstr_operand): New function to support
6888 (walk_rtx): Call push_pathstr_operand.
6889 (print_path): Support [a-zA-Z].
6891 2018-02-07 Richard Biener <rguenther@suse.de>
6893 PR tree-optimization/84037
6894 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6895 (cse_and_gimplify_to_preheader): Declare.
6896 (vect_get_place_in_interleaving_chain): Likewise.
6897 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6899 (_loop_vec_info::~_loop_vec_info): Delete it.
6900 (cse_and_gimplify_to_preheader): New function.
6901 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6902 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6903 (vectorizable_load): Likewise. For grouped stores always base
6904 the IV on the first element.
6905 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6906 condition before gimplifying.
6908 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6910 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6911 *DIV_EXPR and *MOD_EXPR.
6913 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6916 * config/i386/i386.c (ix86_option_override_internal): Mask out
6917 the CF_SET bit when checking -fcf-protection.
6919 2018-02-07 Tom de Vries <tom@codesourcery.com>
6922 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6925 2018-02-07 Richard Biener <rguenther@suse.de>
6927 PR tree-optimization/84204
6928 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6931 PR tree-optimization/84205
6932 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6933 special-case isl_ast_op_zdiv_r.
6935 PR tree-optimization/84223
6936 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6937 Only add conditions from within the region.
6938 (gather_bbs::after_dom_children): Adjust.
6940 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6943 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6944 * config/avr/avr.md: Only post-reload split REG-REG moves if
6945 either register is GENERAL_REG_P.
6947 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6949 PR tree-optimization/84235
6950 * tree-ssa-scopedtables.c
6951 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6952 if the subtraction is performed in floating point type where NaNs are
6953 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6954 build 1. Formatting fix.
6956 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6959 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6960 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6961 and skip it regardless of bb boundaries. Use CALL_P macro,
6962 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6963 unnecessarily, formatting fix.
6965 2018-02-06 Michael Collison <michael.collison@arm.com>
6967 * config/arm/thumb2.md:
6968 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6969 (*thumb_mov_notscc): Ditto.
6971 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6974 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6975 unsigned_fix, not "s".
6977 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6979 * configure.ac (gcc_fn_eh_frame_ro): New function.
6980 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6981 correct .eh_frame permissions.
6982 * configure: Regenerate.
6984 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6986 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6989 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6991 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6992 Display warning message for -mno-speculate-indirect-jumps.
6994 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6996 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6998 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7000 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7002 PR tree-optimization/84225
7003 * tree-eh.c (find_trapping_overflow): Only call
7004 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7006 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7009 * config/i386/i386.c: Reimplement the check of possible options
7010 -mibt/-mshstk conbination. Change error messages.
7011 * doc/invoke.texi: Fix a typo: remove extra '='.
7013 2018-02-06 Marek Polacek <polacek@redhat.com>
7015 PR tree-optimization/84228
7016 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7018 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7021 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7022 emitted arch directives.
7023 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7024 __ARM_FEATURE_COPROC before changing architectures.
7026 2018-02-06 Richard Biener <rguenther@suse.de>
7028 * config/i386/i386.c (print_reg): Fix typo.
7029 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7031 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7033 * configure: Regenerate.
7035 2018-02-05 Martin Sebor <msebor@redhat.com>
7037 PR tree-optimization/83369
7038 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7041 2018-02-05 Martin Liska <mliska@suse.cz>
7043 * doc/invoke.texi: Cherry-pick upstream r323995.
7045 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7047 * ira.c (ira_init_register_move_cost): Adjust comment.
7049 2018-02-05 Martin Liska <mliska@suse.cz>
7051 PR gcov-profile/84137
7052 * doc/gcov.texi: Fix typo in documentation.
7054 2018-02-05 Martin Liska <mliska@suse.cz>
7056 PR gcov-profile/83879
7057 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7058 using dlopen functionality.
7060 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7062 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7063 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7064 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7065 _mm_maskz_range_round_ss): New intrinsics.
7066 (__builtin_ia32_rangesd128_round)
7067 (__builtin_ia32_rangess128_round): Remove.
7068 (__builtin_ia32_rangesd128_mask_round,
7069 __builtin_ia32_rangess128_mask_round): New builtins.
7070 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7071 __builtin_ia32_rangess128_round): Remove.
7072 (__builtin_ia32_rangesd128_mask_round,
7073 __builtin_ia32_rangess128_mask_round): New builtins.
7074 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7075 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7076 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7077 "<round_saeonly_constraint>")): Changed to ...
7078 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7079 "<round_saeonly_scalar_constraint>")): ... this.
7080 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7081 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7082 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7083 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7084 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7086 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7088 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7090 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7091 Remove all values except native, 8540 and 8548.
7093 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7095 * config/i386/i386.c (ix86_output_function_return): Pass
7096 INVALID_REGNUM, instead of -1, as invalid register number to
7097 indirect_thunk_name and output_indirect_thunk.
7099 2018-02-02 Julia Koval <julia.koval@intel.com>
7101 * config.gcc: Add -march=icelake.
7102 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7103 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7104 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7105 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7106 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7107 (processor_target_table): Add icelake.
7108 (ix86_option_override_internal): Handle new PTAs.
7109 (get_builtin_code_for_version): Handle icelake.
7110 (M_INTEL_COREI7_ICELAKE): New.
7111 (fold_builtin_cpu): Handle icelake.
7112 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7113 * doc/invoke.texi: Add -march=icelake.
7115 2018-02-02 Julia Koval <julia.koval@intel.com>
7117 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7118 to wide_int_bitmask.
7119 * wide-int-bitmask.h: New.
7121 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7124 * config/i386/i386.md: Replace Pmode with word_mode in
7125 builtin_setjmp_setup and builtin_longjmp to support x32.
7127 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7131 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7133 (rs6000_supported_cpu_names): New static variable.
7134 (linux_cpu_translation_table): Likewise.
7135 (elf_platform) <cpu>: Define new static variable and use it.
7136 Translate kernel AT_PLATFORM name to canonical name if needed.
7137 Error if platform name is unknown.
7139 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7142 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7144 2018-02-01 Jeff Law <law@redhat.com>
7147 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7148 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7149 the scratch if RELEASE_VIA_POP is false.
7150 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7151 If we have to save a temporary register, decrement SIZE appropriately.
7152 Pass new arguments to release_scratch_register_on_entry.
7153 (ix86_adjust_stack_and_probe): Likewise.
7154 (ix86_emit_probe_stack_range): Pass new arguments to
7155 release_scratch_register_on_entry.
7157 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7159 PR rtl-optimization/84157
7160 * combine.c (change_zero_ext): Use REG_P predicate in
7161 front of HARD_REGISTER_P predicate.
7163 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7165 * config/avr/avr.c (avr_option_override): Move disabling of
7166 -fdelete-null-pointer-checks to...
7167 * common/config/avr/avr-common.c (avr_option_optimization_table):
7170 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7172 PR tree-optimization/81635
7173 * tree-data-ref.c (split_constant_offset_1): For types that
7174 wrap on overflow, try to use range info to prove that wrapping
7177 2018-02-01 Renlin Li <renlin.li@arm.com>
7180 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7182 (aarch64_register_move_cost): Likewise.
7183 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7185 (REG_CLASS_NAMES): Likewise.
7186 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7187 TAILCALL_ADDR_REGS. Remove IP registers.
7188 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7190 2018-02-01 Richard Biener <rguenther@suse.de>
7192 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7193 for specifying RPO order and allow NULL for that.
7194 * domwalk.c (dom_walker::dom_walker): Likewise.
7195 (dom_walker::walk): Handle NULL RPO order.
7196 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7198 (rewrite_update_dom_walker): Likewise.
7199 (mark_def_dom_walker): Likewise.
7201 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7203 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7204 (aarch64_maybe_expand_sve_subreg_move): Declare.
7205 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7206 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7208 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7209 that are semantically a reverse operation.
7210 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7211 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7212 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7214 (aarch64_can_change_mode_class): For big-endian, forbid changes
7215 between two SVE modes if they have different element sizes.
7217 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7219 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7220 the TImode handling for big-endian targets.
7222 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7224 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7225 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7227 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7228 Remove BSWAP handing for big-endian targets and use the form of
7229 LD1RQ appropariate for the mode.
7231 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7233 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7234 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7237 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7240 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7241 check for operands that need to go through aarch64_sve_reload_be.
7243 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7245 PR tree-optimization/81661
7246 PR tree-optimization/84117
7247 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7248 * tree-eh.c: Include gimplify.h.
7249 (find_trapping_overflow, replace_trapping_overflow,
7250 rewrite_to_non_trapping_overflow): New functions.
7251 * tree-vect-loop.c: Include tree-eh.h.
7252 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7253 * tree-data-ref.c: Include tree-eh.h.
7254 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7256 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7258 PR rtl-optimization/84123
7259 * combine.c (change_zero_ext): Check if hard register satisfies
7260 can_change_dest_mode before calling gen_lowpart_SUBREG.
7262 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7265 * ira.c (ira_init_register_move_cost): Remove assert.
7267 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7269 PR rtl-optimization/84071
7270 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7271 * doc/tm.texi: Regenerate.
7273 2018-01-31 Richard Biener <rguenther@suse.de>
7275 PR tree-optimization/84132
7276 * tree-data-ref.c (analyze_miv_subscript): Properly
7277 check whether evolution_function_is_affine_multivariate_p
7278 before calling gcd_of_steps_may_divide_p.
7280 2018-01-31 Julia Koval <julia.koval@intel.com>
7283 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7284 * config/i386/i386.md (rdpid_rex64) New.
7285 (rdpid): Make 32bit only.
7287 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7290 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7291 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7293 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7296 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7298 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7300 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7302 PR rtl-optimization/84071
7303 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7304 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7306 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7308 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7309 (arc_attribute_table): Add 'aux' attribute.
7310 (arc_in_small_data_p): Consider aux like variables.
7311 (arc_is_aux_reg_p): New function.
7312 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7313 (arc_get_aux_arg): New function.
7314 (prepare_move_operands): Handle aux-register access.
7315 (arc_handle_aux_attribute): New function.
7316 * doc/extend.texi (ARC Variable attributes): Add subsection.
7318 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7320 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7321 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7322 (arc_attribute_table): Add 'uncached' attribute.
7323 (arc_print_operand): Print '.di' flag for uncached memory
7325 (arc_in_small_data_p): Do not consider for small data the uncached
7327 (arc_is_uncached_mem_p): New function.
7328 * config/arc/predicates.md (compact_store_memory_operand): Check
7329 for uncached memory accesses.
7330 (nonvol_nonimm_operand): Likewise.
7331 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7333 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7336 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7337 falign-loops=): Add Optimization flag.
7339 2018-01-30 Jeff Law <law@redhat.com>
7342 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7343 INT_REGISTERS_SAVED. Check it prior to calling
7344 get_scratch_register_on_entry.
7345 (ix86_adjust_stack_and_probe): Similarly.
7346 (ix86_emit_probe_stack_range): Similarly.
7347 (ix86_expand_prologue): Corresponding changes.
7349 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7352 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7353 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7355 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7358 * lra-constraints.c (curr_insn_transform): Process AND in the
7361 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7363 PR rtl-optimization/83986
7364 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7365 dependence against last_pending_memory_flush in addition to
7368 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7370 PR tree-optimization/81611
7371 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7374 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7377 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7380 2018-01-30 Richard Biener <rguenther@suse.de>
7381 Jakub Jelinek <jakub@redhat.com>
7383 PR tree-optimization/84111
7384 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7385 inner loops added during recursion, as they don't have up-to-date
7388 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7391 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7392 (can_inline_edge_by_limits_p): ... here.
7393 (can_early_inline_edge_p, check_callers,
7394 update_caller_keys, update_callee_keys, recursive_inlining,
7395 add_new_edges_to_heap, speculation_useful_p,
7396 inline_small_functions,
7397 inline_small_functions, flatten_function,
7398 inline_to_all_callers_1): Update.
7400 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7402 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7405 2018-01-30 Richard Biener <rguenther@suse.de>
7407 PR tree-optimization/83008
7408 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7409 invariant and constant vector uses in stmts when they need
7412 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7415 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7416 * configure: Regenerate.
7418 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7420 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7422 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7423 Use gen_rtx_REG rather than gen_lowpart.
7425 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7427 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7428 rather than 0 when creating partial subregs.
7430 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7432 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7435 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7438 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7439 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7440 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7441 flags. This restores the settings used before the 2017-07-24.
7442 Turning off pre increment/decrement/modify allows IVOPTS to
7443 optimize DF/SF loops where the index is an int.
7445 2018-01-29 Richard Biener <rguenther@suse.de>
7446 Kelvin Nilsen <kelvin@gcc.gnu.org>
7449 * tree-vect-stmts.c (vectorizable_call): Don't call
7450 targetm.vectorize_builtin_md_vectorized_function if callee is
7453 2018-01-22 Carl Love <cel@us.ibm.com>
7455 * doc/extend.tex: Fix typo in second arg in
7456 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7458 2018-01-29 Richard Biener <rguenther@suse.de>
7460 PR tree-optimization/84086
7461 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7462 (flush_ssaname_freelist): When SSA names were released reset
7463 the SCEV hash table.
7465 2018-01-29 Richard Biener <rguenther@suse.de>
7467 PR tree-optimization/84057
7468 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7469 removed paths when removing edges.
7471 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7473 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7474 -mfunction-return=@var{choice}.
7476 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7479 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7481 (layout::print_source_line): Likewise.
7482 (test_get_line_width_without_trailing_whitespace): Add test cases.
7484 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7487 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7490 2018-01-26 Jim Wilson <jimw@sifive.com>
7492 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7494 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7497 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7499 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7500 and CMP + SUB-immediate -> SUBS.
7502 2018-01-26 Martin Sebor <msebor@redhat.com>
7504 PR tree-optimization/83896
7505 * tree-ssa-strlen.c (get_string_len): Rename...
7506 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7507 Avoid assuming length is constant.
7508 (handle_char_store): Use HOST_WIDE_INT for string length.
7510 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7513 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7514 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7516 2018-01-26 Richard Biener <rguenther@suse.de>
7518 PR rtl-optimization/84003
7519 * dse.c (record_store): Only record redundant stores when
7520 the earlier store aliases at least all accesses the later one does.
7522 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7524 PR rtl-optimization/83985
7525 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7526 REG_CFA_RESTORE insns.
7527 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7528 REG_CFA_RESTORE insns here.
7531 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7532 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7534 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7536 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7537 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7538 (arc_init): Likewise.
7539 (arc_override_options): Likewise.
7540 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7542 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7544 * config/arc/arc.h (TARGET_DBNZ): Define.
7545 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7546 properly set the tune attribute.
7547 (dbnz): Use TARGET_DBNZ guard.
7548 * config/arc/arc.opt (mtune): Add core3 option.
7550 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7552 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7553 recognize new pic like addresses.
7554 (arc_delegitimize_address): Clean up.
7556 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7558 * config/arc/arc-arches.def: Option mrf16 valid for all
7560 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7561 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7562 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7563 * config/arc/arc-tables.opt: Regenerate.
7564 * config/arc/arc.c (arc_conditional_register_usage): Handle
7565 reduced register file case.
7566 (arc_file_start): Set must have build attributes.
7567 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7569 * config/arc/arc.opt (mrf16): Add new option.
7570 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7571 * config/arc/genmultilib.awk: Handle new mrf16 option.
7572 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7573 * config/arc/t-multilib: Regenerate.
7574 * doc/invoke.texi (ARC Options): Document mrf16 option.
7576 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7578 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7579 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7580 (arc_attribute_table): Add 'secure_call' attribute.
7581 (arc_print_operand): Print secure call operand.
7582 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7584 (arc_is_secure_call_p): New function. * config/arc/arc.md
7585 (call_i): Add support for sjli instruction.
7586 (call_value_i): Likewise.
7587 * config/arc/constraints.md (Csc): New constraint.
7589 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7590 John Eric Martin <John.Martin@emmicro-us.com>
7592 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7593 * config/arc/arc.c (_arc_jli_section): New struct.
7594 (arc_jli_section): New type.
7595 (rc_jli_sections): New static variable.
7596 (arc_handle_jli_attribute): New function.
7597 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7598 (arc_file_end): New function.
7599 (TARGET_ASM_FILE_END): Define.
7600 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7601 (arc_add_jli_section): New function.
7602 (jli_call_scan): Likewise.
7603 (arc_reorg): Call jli_call_scan.
7604 (arc_output_addsi): Remove 'S' from printing asm operand.
7605 (arc_is_jli_call_p): New function.
7606 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7608 (movhi_insn): Likewise.
7609 (movsi_insn): Likewise.
7610 (movsi_set_cc_insn): Likewise.
7611 (loadqi_update): Likewise.
7612 (load_zeroextendqisi_update): Likewise.
7613 (load_signextendqisi_update): Likewise.
7614 (loadhi_update): Likewise.
7615 (load_zeroextendhisi_update): Likewise.
7616 (load_signextendhisi_update): Likewise.
7617 (loadsi_update): Likewise.
7618 (loadsf_update): Likewise.
7619 (movsicc_insn): Likewise.
7620 (bset_insn): Likewise.
7621 (bxor_insn): Likewise.
7622 (bclr_insn): Likewise.
7623 (bmsk_insn): Likewise.
7624 (bicsi3_insn): Likewise.
7625 (cmpsi_cc_c_insn): Likewise.
7626 (movsi_ne): Likewise.
7627 (movsi_cond_exec): Likewise.
7628 (clrsbsi2): Likewise.
7639 (call_i): Remove 'S' asm letter, add jli instruction.
7640 (call_value_i): Likewise.
7641 * config/arc/arc.op (mjli-always): New option.
7642 * config/arc/constraints.md (Cji): New constraint.
7643 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7645 (subsf3_fpx): Likewise.
7646 (mulsf3_fpx): Likewise.
7647 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7649 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7651 * doc/invoke.texi (ARC): Document mjli-always option.
7653 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7655 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7656 avoid addition with 0 and use incw and decw where possible.
7658 2018-01-26 Richard Biener <rguenther@suse.de>
7660 PR tree-optimization/81082
7661 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7662 association if it requires casting to unsigned.
7663 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7664 from fold_plusminus_mult_expr to catch important cases late when
7665 range info is available.
7667 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7669 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7670 * configure.ac (hidden_linkonce): New test.
7671 * configure: Regenerate.
7672 * config.in: Regenerate.
7674 2018-01-26 Julia Koval <julia.koval@intel.com>
7676 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7677 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7678 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7679 _mm_mask_bitshuffle_epi64_mask): Fix type.
7680 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7681 USI_FTYPE_V4DI_V4DI_USI): Remove.
7682 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7683 __builtin_ia32_vpshufbitqmb256_mask,
7684 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7685 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7686 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7688 2018-01-26 Alan Modra <amodra@gmail.com>
7691 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7692 UNSPEC_VBPERMQ. Sort other unspecs.
7694 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7696 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7698 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7701 * predict.c (drop_profile): Do not push/pop cfun; update also
7703 (handle_missing_profiles): Fix logic looking for zero profiles.
7705 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7708 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7709 on functions with #pragma omp declare simd or functions with simd
7711 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7712 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7713 Remove trailing \n from warning_at calls.
7715 2018-01-25 Tom de Vries <tom@codesourcery.com>
7718 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7719 for neutered workers.
7721 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7724 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7725 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7727 2018-01-24 Jeff Law <law@redhat.com>
7730 * i386.c (get_probe_interval): Move to earlier point.
7731 (ix86_compute_frame_layout): If -fstack-clash-protection and
7732 the frame is larger than the probe interval, then use pushes
7733 to save registers rather than reg->mem moves.
7734 (ix86_expand_prologue): Remove conditional for int_registers_saved
7737 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7740 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7741 min/max for never referenced object.
7743 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7746 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7748 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7749 attributes from DECL_ATTRIBUTES (decl) without affecting
7750 DECL_ATTRIBUTES (current_function_decl).
7751 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7752 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7754 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7756 PR tree-optimization/83979
7757 * fold-const.c (fold_comparison): Use constant_boolean_node
7758 instead of boolean_{true,false}_node.
7760 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7762 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7765 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7767 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7768 Simplify the clause that sets the length attribute.
7769 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7770 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7771 clause that sets the length attribute.
7772 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7774 2018-01-24 Tom de Vries <tom@codesourcery.com>
7777 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7778 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7779 Add strict parameter.
7780 (prevent_branch_around_nothing): Insert dummy insn between branch to
7781 label and label with no ptx insn inbetween.
7782 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7784 2018-01-24 Tom de Vries <tom@codesourcery.com>
7787 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7788 for neutered threads in warp.
7789 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7791 2018-01-24 Richard Biener <rguenther@suse.de>
7793 PR tree-optimization/83176
7794 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7797 2018-01-24 Richard Biener <rguenther@suse.de>
7799 PR tree-optimization/82819
7800 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7801 code generating pluses that are no-ops in the target precision.
7803 2018-01-24 Richard Biener <rguenther@suse.de>
7806 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7808 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7810 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7811 to merge probabilities.
7812 * predict.c (probably_never_executed): Also mark as cold functions
7813 with global 0 profile and guessed local profile.
7814 * profile-count.c (profile_probability::combine_with_count): New
7816 * profile-count.h (profile_probability::operator*,
7817 profile_probability::operator*=, profile_probability::operator/,
7818 profile_probability::operator/=): Reduce precision to adjusted
7819 and set value to guessed on contradictory divisions.
7820 (profile_probability::combine_with_freq): Remove.
7821 (profile_probability::combine_wiht_count): Declare.
7822 (profile_count::force_nonzero):: Set to adjusted.
7823 (profile_count::probability_in):: Set quality to adjusted.
7824 * tree-ssa-tail-merge.c (replace_block_by): Use
7827 2018-01-23 Andrew Waterman <andrew@sifive.com>
7828 Jim Wilson <jimw@sifive.com>
7830 * config/riscv/riscv.c (riscv_stack_boundary): New.
7831 (riscv_option_override): Set riscv_stack_boundary. Handle
7832 riscv_preferred_stack_boundary_arg.
7833 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7834 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7835 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7836 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7837 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7838 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7840 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7843 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7844 of struct ix86_frame.
7845 (ix86_expand_epilogue): Likewise. Add a local variable for
7846 the reg_save_offset field in struct ix86_frame.
7848 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7850 PR tree-optimization/82604
7851 * tree-loop-distribution.c (enum partition_kind): New enum item
7852 PKIND_PARTIAL_MEMSET.
7853 (partition_builtin_p): Support above new enum item.
7854 (generate_code_for_partition): Ditto.
7855 (compute_access_range): Differentiate cases that equality can be
7856 proven at all loops, the innermost loops or no loops.
7857 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7858 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7859 (finalize_partitions, distribute_loop): Don't fuse partition of
7860 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7861 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7864 2018-01-23 Martin Liska <mliska@suse.cz>
7866 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7867 order to ignore the predictor.
7868 (PRED_POLYMORPHIC_CALL): Likewise.
7869 (PRED_RECURSIVE_CALL): Likewise.
7871 2018-01-23 Martin Liska <mliska@suse.cz>
7873 * tree-profile.c (tree_profiling): Print function header to
7874 aware reader which function we are working on.
7875 * value-prof.c (gimple_find_values_to_profile): Do not print
7876 not interesting value histograms.
7878 2018-01-23 Martin Liska <mliska@suse.cz>
7880 * profile-count.h (enum profile_quality): Add
7881 profile_uninitialized as the first value. Do not number values
7882 as they are zero based.
7883 (profile_count::verify): Update sanity check.
7884 (profile_probability::verify): Likewise.
7886 2018-01-23 Nathan Sidwell <nathan@acm.org>
7888 * doc/invoke.texi (ffor-scope): Deprecate.
7890 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7892 PR tree-optimization/83510
7893 * domwalk.c (set_all_edges_as_executable): New function.
7894 (dom_walker::dom_walker): Convert bool param
7895 "skip_unreachable_blocks" to enum reachability. Move setup of
7896 edge flags to set_all_edges_as_executable and only do it when
7897 reachability is REACHABLE_BLOCKS.
7898 * domwalk.h (enum dom_walker::reachability): New enum.
7899 (dom_walker::dom_walker): Convert bool param
7900 "skip_unreachable_blocks" to enum reachability.
7901 (set_all_edges_as_executable): New decl.
7902 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7903 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7905 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7906 but converting true to REACHABLE_BLOCKS.
7907 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7909 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7910 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7911 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7913 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7914 if check_all_array_refs will be called.
7916 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7918 * tree.c (selftest::test_location_wrappers): Add more test
7921 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7923 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7924 (selftest::test_bit_in_range): Likewise.
7926 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7929 * doc/sourcebuild.texi (vect_float): Say that the selector
7930 only describes the situation when -funsafe-math-optimizations is on.
7931 (vect_float_strict): Document.
7933 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7935 PR tree-optimization/83965
7936 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7937 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7938 instead of checking only for a reduction.
7939 (vect_recog_widen_sum_pattern): Likewise.
7941 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7943 * predict.c (probably_never_executed): Only use precise profile info.
7944 (compute_function_frequency): Skip after inlining hack since we now
7945 have quality checking.
7947 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7949 * profile-count.h (profile_probability::very_unlikely,
7950 profile_probability::unlikely, profile_probability::even): Set
7951 precision to guessed.
7953 2018-01-23 Richard Biener <rguenther@suse.de>
7955 PR tree-optimization/83963
7956 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7957 Properly terminate dominator walk when crossing the exit edge not
7958 when visiting its source block.
7960 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7963 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7964 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7966 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7968 PR tree-optimization/83957
7969 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7970 semicolon after for body surrounded by braces.
7972 PR tree-optimization/83081
7973 * profile-count.h (profile_probability::split): New method.
7974 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7975 Use profile_probability::split.
7976 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7977 when splitting a single conditional jump into 2.
7979 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7981 PR tree-optimization/69452
7982 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7985 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7987 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7988 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7989 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7991 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7993 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7994 * config/rl78/rl78.md (movdi): New define_expand.
7995 * config/rl78/rl78.c (rl78_split_movdi): New function.
7997 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8000 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8002 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8003 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8004 128-bit to produce an UNSPEC move to get the double word with the
8005 signbit and then a shift directly to do signbit.
8006 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8007 implementation with a new version that just does either a direct
8008 move or a regular move. Move memory interface to separate insns.
8009 Move insns so they are next to the expander.
8010 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8011 with signbit move. Split big and little endian case.
8012 (signbit<mode>2_dm_mem_le): Likewise.
8013 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8014 (signbit<mode>2_dm2): Likewise.
8016 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8018 * config/rl78/rl78.md (anddi3): New define_expand.
8020 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8022 * config/rl78/rl78.md (umindi3): New define_expand.
8024 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8026 * config/rl78/rl78.md (smindi3): New define_expand.
8028 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8030 * config/rl78/rl78.md (smaxdi3): New define_expand.
8032 2018-01-22 Carl Love <cel@us.ibm.com>
8034 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8035 LVX_V1TI): Add macro expansion.
8036 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8037 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8038 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8039 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8040 Change check to determine if the instruction is a byte reversing
8041 entry. Fix typo in comment.
8042 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8043 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8044 Add def_builtin calls for new builtins.
8045 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8046 Add define_insn expansion.
8048 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8050 * config/rl78/rl78.md (umaxdi3): New define_expand.
8052 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8054 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8055 for non-QImode registers.
8057 2018-01-22 Richard Biener <rguenther@suse.de>
8059 PR tree-optimization/83963
8060 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8061 including the loop exit block.
8062 (scop_detection::merge_sese): Likewise.
8063 (scop_detection::add_scop): Do it here instead.
8065 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8067 * doc/sourcebuild.texi (arm_softfloat): Document.
8069 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8072 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8073 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8074 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8076 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8077 David Edelsohn <dje.gcc@gmail.com>
8080 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8081 Change "crset eq" to "crset 2".
8082 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8083 (*call_indirect_aix<mode>_nospec): Likewise.
8084 (*call_value_indirect_aix<mode>_nospec): Likewise.
8085 (*call_indirect_elfv2<mode>_nospec): Likewise.
8086 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8087 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8088 change assembly output from . to $.
8089 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8090 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8091 (*tablejump<mode>_internal1_nospec): Likewise.
8093 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8096 * config/sh/sh_optimize_sett_clrt.cc:
8097 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8099 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8101 PR tree-optimization/83940
8102 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8103 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8104 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8105 use it to pass back the definition type.
8106 (vect_check_store_rhs): Likewise rhs_dt_out.
8107 (vect_build_gather_load_calls): Add a mask_dt argument and use
8108 it instead of a call to vect_is_simple_use.
8109 (vectorizable_store): Update calls to vect_check_load_store_mask
8110 and vect_check_store_rhs. Use the dt returned by the latter instead
8111 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8112 instead of calls to vect_is_simple_use. Pass the scalar rather
8113 than the vector operand to vect_is_simple_use when handling
8114 second and subsequent copies of an rhs value.
8115 (vectorizable_load): Update calls to vect_check_load_store_mask
8116 and vect_build_gather_load_calls. Use the cached mask_dt and
8117 gs_info.offset_dt instead of calls to vect_is_simple_use.
8119 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8122 * tree-emutls.c: Include gimplify.h.
8123 (lower_emutls_2): New function.
8124 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8125 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8126 it before further processing.
8129 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8130 UINTVAL (trueop1) instead of INTVAL (op1).
8132 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8136 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8137 INCOMING_FRAME_SP_OFFSET if not defined.
8138 (scan_trace): Add ENTRY argument. If true and
8139 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8140 emit a note to adjust the CFA offset.
8141 (create_cfi_notes): Adjust scan_trace callers.
8142 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8143 INCOMING_FRAME_SP_OFFSET in the CIE.
8144 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8145 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8147 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8148 * doc/tm.texi: Regenerated.
8150 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8152 PR rtl-optimization/83147
8153 * lra-constraints.c (remove_inheritance_pseudos): Use
8154 lra_substitute_pseudo_within_insn.
8156 2018-01-19 Tom de Vries <tom@codesourcery.com>
8157 Cesar Philippidis <cesar@codesourcery.com>
8160 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8162 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8165 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8166 spaces for function labels.
8168 2018-01-19 Martin Liska <mliska@suse.cz>
8170 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8171 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8172 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8173 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8174 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8175 (PRED_CONST_RETURN): Change from 69 to 65.
8176 (PRED_NULL_RETURN): Change from 91 to 71.
8177 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8178 (PRED_LOOP_GUARD): Change from 66 to 73.
8180 2018-01-19 Martin Liska <mliska@suse.cz>
8182 * predict.c (predict_insn_def): Add new assert.
8183 (struct branch_predictor): Change type to signed integer.
8184 (test_prediction_value_range): Amend test to cover
8186 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8187 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8188 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8189 (PRED_LOOP_IV_COMPARE): Likewise.
8190 * predict.h (PROB_UNINITIALIZED): Define new constant.
8192 2018-01-19 Martin Liska <mliska@suse.cz>
8194 * predict.c (dump_prediction): Add new format for
8195 analyze_brprob.py script which is enabled with -details
8197 * profile-count.h (precise_p): New function.
8199 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8201 PR tree-optimization/83922
8202 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8203 there are no statements that need masking.
8204 (vect_active_double_reduction_p): New function.
8205 (vect_analyze_loop_operations): Use it when handling phis that
8206 are not in the loop header.
8208 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8210 PR tree-optimization/83914
8211 * tree-vect-loop.c (vectorizable_induction): Don't convert
8212 init_expr or apply the peeling adjustment for inductions
8213 that are nested within the vectorized loop.
8215 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8217 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8220 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8224 * function.h (gimplify_parameters): Add gimple_seq * argument.
8225 * function.c: Include gimple.h and options.h.
8226 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8227 for the added local temporaries if needed.
8228 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8229 if there are any parameter cleanups, wrap whole body into a
8230 try/finally with the cleanups.
8232 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8235 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8236 Use GET_MODE_CLASS for scalar floating point.
8238 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8242 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8243 Fix call of call_cgraph_insertion_hooks.
8245 2018-01-18 Martin Sebor <msebor@redhat.com>
8247 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8249 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8252 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8255 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8258 * common.opt: (-ffile-prefix-map): New option.
8259 * opts.c (common_handle_option): Defer it.
8260 * opts-global.c (handle_common_deferred_options): Handle it.
8261 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8262 * file-prefix-map.h: New file.
8263 (remap_debug_filename, add_debug_prefix_map): ...here.
8264 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8265 * final.c (debug_prefix_map, add_debug_prefix_map
8266 remap_debug_filename): Move to...
8267 * file-prefix-map.c: New file.
8268 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8269 generalize, get rid of alloca(), use strrchr() instead of strchr().
8270 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8271 Implement in terms of add_prefix_map().
8272 (remap_macro_filename, remap_debug_filename): Implement in term of
8274 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8275 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8276 * dbxout.c: Include file-prefix-map.h.
8277 * varasm.c: Likewise.
8278 * vmsdbgout.c: Likewise.
8279 * xcoffout.c: Likewise.
8280 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8281 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8282 * doc/invoke.texi (-ffile-prefix-map): Document.
8283 (-fdebug-prefix-map): Update description.
8285 2018-01-18 Martin Liska <mliska@suse.cz>
8287 * config/i386/i386.c (indirect_thunk_name): Document that also
8289 (output_indirect_thunk): Document why both instructions
8290 (pause and lfence) are generated.
8292 2018-01-18 Richard Biener <rguenther@suse.de>
8294 PR tree-optimization/83887
8295 * graphite-scop-detection.c
8296 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8297 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8298 (scop_detection::merge_sese): Re-implement with a flood-fill
8299 algorithm that properly finds a SESE region if it exists.
8301 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8304 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8305 pointer_diff optimizations use view_convert instead of convert.
8307 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8309 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8310 Generate different code for -mno-speculate-indirect-jumps.
8311 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8312 (*call_indirect_aix<mode>): Disable for
8313 -mno-speculate-indirect-jumps.
8314 (*call_indirect_aix<mode>_nospec): New define_insn.
8315 (*call_value_indirect_aix<mode>): Disable for
8316 -mno-speculate-indirect-jumps.
8317 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8318 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8319 -mno-speculate-indirect-jumps.
8320 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8322 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8324 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8325 long double type, set the flags for noting the default long double
8326 type, even if we don't pass or return a long double type.
8328 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8331 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8334 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8336 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8337 support for merge[hl].
8338 (fold_mergehl_helper): New helper function.
8339 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8340 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8341 (altivec_vmrglw_direct): Add xxmrglw insn.
8343 2018-01-17 Andrew Waterman <andrew@sifive.com>
8345 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8346 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8348 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8351 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8352 call the lto_location_cache before reading the
8353 DECL_SOURCE_LOCATION of the types.
8355 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8356 Richard Sandiford <richard.sandiford@linaro.org>
8358 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8359 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8360 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8361 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8362 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8364 * config/aarch64/constraints.md (aarch64_movti_operand):
8366 * config/aarch64/predicates.md (Uti): Add new constraint.
8368 2018-01-17 Carl Love <cel@us.ibm.com>
8370 * config/rs6000/vsx.md (define_expand xl_len_r,
8371 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8372 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8374 (define_expand, define_insn): Move the shift left from the
8375 define_insn to the define_expand for lxvl and stxvl instructions.
8376 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8377 and XL_LEN_R definitions to PURE.
8379 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8381 * config/i386/i386.c (indirect_thunk_name): Declare regno
8382 as unsigned int. Compare regno with INVALID_REGNUM.
8383 (output_indirect_thunk): Ditto.
8384 (output_indirect_thunk_function): Ditto.
8385 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8386 in the call to output_indirect_thunk_function.
8388 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8391 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8392 rather than the size of inner_type to determine the stack slot size
8393 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8395 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8398 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8401 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8403 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8404 endian Linux systems to optionally enable multilibs for selecting
8405 the long double type if the user configured an explicit type.
8406 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8407 have no long double multilibs if not defined.
8408 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8409 warn if the user used -mabi={ieee,ibm}longdouble and we built
8410 multilibs for long double.
8411 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8412 appropriate multilib option.
8413 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8415 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8416 for building long double multilibs.
8417 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8419 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8421 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8424 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8426 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8429 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8432 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8435 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8437 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8438 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8440 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8442 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8443 different rtl trees depending on TARGET_64BIT.
8444 (rs6000_gen_lvx): Likewise.
8446 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8448 * config/visium/visium.md (nop): Tweak comment.
8449 (hazard_nop): Likewise.
8451 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8453 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8454 -mspeculate-indirect-jumps.
8455 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8456 for -mno-speculate-indirect-jumps.
8457 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8458 (*call_value_indirect_elfv2<mode>): Disable for
8459 -mno-speculate-indirect-jumps.
8460 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8461 (indirect_jump): Emit different RTL for
8462 -mno-speculate-indirect-jumps.
8463 (*indirect_jump<mode>): Disable for
8464 -mno-speculate-indirect-jumps.
8465 (*indirect_jump<mode>_nospec): New define_insn.
8466 (tablejump): Emit different RTL for
8467 -mno-speculate-indirect-jumps.
8468 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8469 (tablejumpsi_nospec): New define_expand.
8470 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8471 (tablejumpdi_nospec): New define_expand.
8472 (*tablejump<mode>_internal1): Disable for
8473 -mno-speculate-indirect-jumps.
8474 (*tablejump<mode>_internal1_nospec): New define_insn.
8475 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8478 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8480 * caller-save.c (insert_save): Drop unnecessary parameter. All
8483 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8484 Richard Biener <rguenth@suse.de>
8487 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8488 return early, inline manually is_gimple_sizepos. Make sure if we
8489 call gimplify_expr we don't end up with a gimple constant.
8490 * tree.c (variably_modified_type_p): Don't return true for
8491 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8492 * gimplify.h (is_gimple_sizepos): Remove.
8494 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8496 PR tree-optimization/83857
8497 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8498 vectorizable_live_operation for pure SLP statements.
8499 (vectorizable_live_operation): Handle PHIs.
8501 2018-01-16 Richard Biener <rguenther@suse.de>
8503 PR tree-optimization/83867
8504 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8505 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8507 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8510 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8511 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8512 If off is not INTEGER_CST, issue a may not be aligned warning
8513 rather than isn't aligned. Use isn%'t rather than isn't.
8514 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8516 <case MULT_EXPR>: Improve the case when bottom and one of the
8517 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8518 operand, in that case check if the other operand is multiple of
8519 bottom divided by the INTEGER_CST operand.
8521 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8524 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8525 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8526 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8527 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8528 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8529 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8530 (pa_function_arg_size): New function.
8532 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8534 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8535 in a separate statement.
8537 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8539 PR tree-optimization/83847
8540 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8541 group gathers and scatters.
8543 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8545 PR rtl-optimization/86620
8546 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8548 PR rtl-optimization/83213
8549 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8550 to last if both are JUMP_INSNs.
8552 PR tree-optimization/83843
8553 * gimple-ssa-store-merging.c
8554 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8555 store_immediate_info for bswap/nop orig_stores.
8557 2018-01-15 Andrew Waterman <andrew@sifive.com>
8559 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8561 <UDIV>: Increase cost if !TARGET_DIV.
8563 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8565 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8566 (define_attr "cr_logical_3op"): New.
8567 (cceq_ior_compare): Adjust.
8568 (cceq_ior_compare_complement): Adjust.
8569 (*cceq_rev_compare): Adjust.
8570 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8571 (is_cracked_insn): Adjust.
8572 (insn_must_be_first_in_group): Adjust.
8573 * config/rs6000/40x.md: Adjust.
8574 * config/rs6000/440.md: Adjust.
8575 * config/rs6000/476.md: Adjust.
8576 * config/rs6000/601.md: Adjust.
8577 * config/rs6000/603.md: Adjust.
8578 * config/rs6000/6xx.md: Adjust.
8579 * config/rs6000/7450.md: Adjust.
8580 * config/rs6000/7xx.md: Adjust.
8581 * config/rs6000/8540.md: Adjust.
8582 * config/rs6000/cell.md: Adjust.
8583 * config/rs6000/e300c2c3.md: Adjust.
8584 * config/rs6000/e500mc.md: Adjust.
8585 * config/rs6000/e500mc64.md: Adjust.
8586 * config/rs6000/e5500.md: Adjust.
8587 * config/rs6000/e6500.md: Adjust.
8588 * config/rs6000/mpc.md: Adjust.
8589 * config/rs6000/power4.md: Adjust.
8590 * config/rs6000/power5.md: Adjust.
8591 * config/rs6000/power6.md: Adjust.
8592 * config/rs6000/power7.md: Adjust.
8593 * config/rs6000/power8.md: Adjust.
8594 * config/rs6000/power9.md: Adjust.
8595 * config/rs6000/rs64.md: Adjust.
8596 * config/rs6000/titan.md: Adjust.
8598 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8600 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8601 ix86_indirect_branch_register logic.
8603 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8605 * config/i386/constraints.md (Bs): Update
8606 ix86_indirect_branch_register check. Don't check
8607 ix86_indirect_branch_register with GOT_memory_operand.
8609 * config/i386/predicates.md (GOT_memory_operand): Don't check
8610 ix86_indirect_branch_register here.
8611 (GOT32_symbol_operand): Likewise.
8613 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8615 * config/i386/predicates.md (constant_call_address_operand):
8616 Rewrite ix86_indirect_branch_register logic.
8617 (sibcall_insn_operand): Likewise.
8619 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8621 * config/i386/constraints.md (Bs): Replace
8622 ix86_indirect_branch_thunk_register with
8623 ix86_indirect_branch_register.
8625 * config/i386/i386.md (indirect_jump): Likewise.
8626 (tablejump): Likewise.
8627 (*sibcall_memory): Likewise.
8628 (*sibcall_value_memory): Likewise.
8629 Peepholes of indirect call and jump via memory: Likewise.
8630 * config/i386/i386.opt: Likewise.
8631 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8632 (GOT_memory_operand): Likewise.
8633 (call_insn_operand): Likewise.
8634 (sibcall_insn_operand): Likewise.
8635 (GOT32_symbol_operand): Likewise.
8637 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8640 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8641 type rather than type addr's type points to.
8642 (expand_omp_atomic_mutex): Likewise.
8643 (expand_omp_atomic): Likewise.
8645 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8648 * config/i386/i386.c (output_indirect_thunk_function): Use
8649 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8650 for __x86_return_thunk.
8652 2018-01-15 Richard Biener <rguenther@suse.de>
8655 * expmed.c (extract_bit_field_1): Fix typo.
8657 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8660 * config/arm/iterators.md (VF): New mode iterator.
8661 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8662 Remove integer-related logic from pattern.
8663 (neon_vabd<mode>_3): Likewise.
8665 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8668 * common.opt (fstrict-overflow): No longer an alias.
8669 (fwrapv-pointer): New option.
8670 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8671 also for pointer types based on flag_wrapv_pointer.
8672 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8673 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8674 opts->x_flag_wrapv got set.
8675 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8676 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8677 POINTER_TYPE_OVERFLOW_UNDEFINED.
8678 * match.pd: Likewise in address comparison pattern.
8679 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8681 2018-01-15 Richard Biener <rguenther@suse.de>
8684 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8685 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8686 Reset type names to their identifier if their TYPE_DECL doesn't
8687 have linkage (and thus is used for ODR and devirt).
8688 (save_debug_info_for_decl): Remove.
8689 (save_debug_info_for_type): Likewise.
8690 (add_tree_to_fld_list): Adjust.
8691 * tree-pretty-print.c (dump_generic_node): Make dumping of
8692 type names more robust.
8694 2018-01-15 Richard Biener <rguenther@suse.de>
8696 * BASE-VER: Bump to 8.0.1.
8698 2018-01-14 Martin Sebor <msebor@redhat.com>
8701 * builtins.c (check_access): Avoid warning when the no-warning bit
8704 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8706 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8707 * ira-color (allocno_hard_regs_compare): Likewise.
8709 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8712 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8713 Use .pushsection/.popsection.
8715 2018-01-14 Martin Sebor <msebor@redhat.com>
8718 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8720 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8722 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8723 entry from extra_headers.
8724 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8725 extra_headers, make the list bitwise identical to the i?86-*-* one.
8727 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8729 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8730 -mcmodel=large with -mindirect-branch=thunk,
8731 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8732 -mfunction-return=thunk-extern.
8733 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8734 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8735 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8737 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8739 * config/i386/i386.c (print_reg): Print the name of the full
8740 integer register without '%'.
8741 (ix86_print_operand): Handle 'V'.
8742 * doc/extend.texi: Document 'V' modifier.
8744 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8746 * config/i386/constraints.md (Bs): Disallow memory operand for
8747 -mindirect-branch-register.
8749 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8750 (GOT_memory_operand): Likewise.
8751 (call_insn_operand): Likewise.
8752 (sibcall_insn_operand): Likewise.
8753 (GOT32_symbol_operand): Likewise.
8754 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8755 for -mindirect-branch-register.
8756 (tablejump): Likewise.
8757 (*sibcall_memory): Likewise.
8758 (*sibcall_value_memory): Likewise.
8759 Disallow peepholes of indirect call and jump via memory for
8760 -mindirect-branch-register.
8761 (*call_pop): Replace m with Bw.
8762 (*call_value_pop): Likewise.
8763 (*sibcall_pop_memory): Replace m with Bs.
8764 * config/i386/i386.opt (mindirect-branch-register): New option.
8765 * doc/invoke.texi: Document -mindirect-branch-register option.
8767 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8769 * config/i386/i386-protos.h (ix86_output_function_return): New.
8770 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8771 set function_return_type.
8772 (indirect_thunk_name): Add ret_p to indicate thunk for function
8774 (output_indirect_thunk_function): Pass false to
8775 indirect_thunk_name.
8776 (ix86_output_indirect_branch_via_reg): Likewise.
8777 (ix86_output_indirect_branch_via_push): Likewise.
8778 (output_indirect_thunk_function): Create alias for function
8779 return thunk if regno < 0.
8780 (ix86_output_function_return): New function.
8781 (ix86_handle_fndecl_attribute): Handle function_return.
8782 (ix86_attribute_table): Add function_return.
8783 * config/i386/i386.h (machine_function): Add
8784 function_return_type.
8785 * config/i386/i386.md (simple_return_internal): Use
8786 ix86_output_function_return.
8787 (simple_return_internal_long): Likewise.
8788 * config/i386/i386.opt (mfunction-return=): New option.
8789 (indirect_branch): Mention -mfunction-return=.
8790 * doc/extend.texi: Document function_return function attribute.
8791 * doc/invoke.texi: Document -mfunction-return= option.
8793 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8795 * config/i386/i386-opts.h (indirect_branch): New.
8796 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8797 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8798 with local indirect jump when converting indirect call and jump.
8799 (ix86_set_indirect_branch_type): New.
8800 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8801 (indirectlabelno): New.
8802 (indirect_thunk_needed): Likewise.
8803 (indirect_thunk_bnd_needed): Likewise.
8804 (indirect_thunks_used): Likewise.
8805 (indirect_thunks_bnd_used): Likewise.
8806 (INDIRECT_LABEL): Likewise.
8807 (indirect_thunk_name): Likewise.
8808 (output_indirect_thunk): Likewise.
8809 (output_indirect_thunk_function): Likewise.
8810 (ix86_output_indirect_branch_via_reg): Likewise.
8811 (ix86_output_indirect_branch_via_push): Likewise.
8812 (ix86_output_indirect_branch): Likewise.
8813 (ix86_output_indirect_jmp): Likewise.
8814 (ix86_code_end): Call output_indirect_thunk_function if needed.
8815 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8817 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8818 (ix86_attribute_table): Add indirect_branch.
8819 * config/i386/i386.h (machine_function): Add indirect_branch_type
8820 and has_local_indirect_jump.
8821 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8823 (tablejump): Likewise.
8824 (*indirect_jump): Use ix86_output_indirect_jmp.
8825 (*tablejump_1): Likewise.
8826 (simple_return_indirect_internal): Likewise.
8827 * config/i386/i386.opt (mindirect-branch=): New option.
8828 (indirect_branch): New.
8831 (thunk-inline): Likewise.
8832 (thunk-extern): Likewise.
8833 * doc/extend.texi: Document indirect_branch function attribute.
8834 * doc/invoke.texi: Document -mindirect-branch= option.
8836 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8839 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8841 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8843 * ipa-inline.c (want_inline_small_function_p): Return false if
8844 inlining has already failed with CIF_FINAL_ERROR.
8845 (update_caller_keys): Call want_inline_small_function_p before
8847 (update_callee_keys): Likewise.
8849 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8851 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8853 (rs6000_quadword_masked_address_p): Likewise.
8854 (quad_aligned_load_p): Likewise.
8855 (quad_aligned_store_p): Likewise.
8856 (const_load_sequence_p): Add comment to describe the outer-most loop.
8857 (mimic_memory_attributes_and_flags): New function.
8858 (rs6000_gen_stvx): Likewise.
8859 (replace_swapped_aligned_store): Likewise.
8860 (rs6000_gen_lvx): Likewise.
8861 (replace_swapped_aligned_load): Likewise.
8862 (replace_swapped_load_constant): Capitalize argument name in
8863 comment describing this function.
8864 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8865 and stores that access quad-word aligned addresses and replace
8866 with stvx or lvx instructions when appropriate.
8867 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8868 New function prototype.
8869 (rs6000_quadword_masked_address_p): Likewise.
8870 (rs6000_gen_lvx): Likewise.
8871 (rs6000_gen_stvx): Likewise.
8872 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8873 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8874 when memory address is aligned.
8875 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8876 this split to select lvx instruction when memory address is aligned.
8877 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8878 instruction when memory address is aligned.
8879 (*vsx_le_perm_load_v16qi): Likewise.
8880 (four unnamed splitters): Modify to select the stvx instruction
8881 when memory is aligned.
8883 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8885 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8886 which appears in the queue multiple times.
8888 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8889 Alan Hayward <alan.hayward@arm.com>
8890 David Sherwood <david.sherwood@arm.com>
8892 * tree-vectorizer.h (vec_lower_bound): New structure.
8893 (_loop_vec_info): Add check_nonzero and lower_bounds.
8894 (LOOP_VINFO_CHECK_NONZERO): New macro.
8895 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8896 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8897 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8898 fields. Make seg_len the distance travelled, not including the
8900 (dr_direction_indicator): Declare.
8901 (dr_zero_step_indicator): Likewise.
8902 (dr_known_forward_stride_p): Likewise.
8903 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8905 (runtime_alias_check_p): Allow runtime alias checks with
8907 (operator ==): Compare access_size and align.
8908 (prune_runtime_alias_test_list): Rework for new distinction between
8909 the access_size and seg_len.
8910 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8912 (get_segment_min_max): New function.
8913 (create_intersect_range_checks): Use it.
8914 (dr_step_indicator): New function.
8915 (dr_direction_indicator): Likewise.
8916 (dr_zero_step_indicator): Likewise.
8917 (dr_known_forward_stride_p): Likewise.
8918 * tree-loop-distribution.c (data_ref_segment_size): Return
8919 DR_STEP * (niters - 1).
8920 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8922 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8923 (vect_preserves_scalar_order_p): New function, split out from...
8924 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8925 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8926 (vect_vfa_access_size): New function.
8927 (vect_vfa_align): Likewise.
8928 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8929 (dump_lower_bound): New function.
8930 (vect_check_lower_bound): Likewise.
8931 (vect_small_gap_p): Likewise.
8932 (vectorizable_with_step_bound_p): Likewise.
8933 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8934 depencies if the vectorization factor is 1. Convert the checks
8935 for nonzero steps into checks on the bounds of DR_STEP. Try using
8936 a bunds check for variable steps if the minimum required step is
8937 relatively small. Update calls to the dr_with_seg_len
8938 constructor and to vect_compile_time_alias.
8939 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8941 (vect_loop_versioning): Call it.
8942 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8944 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8946 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8947 Alan Hayward <alan.hayward@arm.com>
8948 David Sherwood <david.sherwood@arm.com>
8950 * doc/sourcebuild.texi (vect_scatter_store): Document.
8951 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8953 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8955 * genopinit.c (main): Add supports_vec_scatter_store and
8956 supports_vec_scatter_store_cached to target_optabs.
8957 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8958 IFN_MASK_SCATTER_STORE.
8959 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8961 * internal-fn.h (internal_store_fn_p): Declare.
8962 (internal_fn_stored_value_index): Likewise.
8963 * internal-fn.c (scatter_store_direct): New macro.
8964 (expand_scatter_store_optab_fn): New function.
8965 (direct_scatter_store_optab_supported_p): New macro.
8966 (internal_store_fn_p): New function.
8967 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8968 IFN_MASK_SCATTER_STORE.
8969 (internal_fn_mask_index): Likewise.
8970 (internal_fn_stored_value_index): New function.
8971 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8973 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8974 * optabs-query.c (supports_vec_scatter_store_p): New function.
8975 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8976 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8977 true for scatter stores.
8978 (vect_gather_scatter_fn_p): Handle scatter stores too.
8979 (vect_check_gather_scatter): Consider using scatter stores if
8980 supports_vec_scatter_store_p.
8981 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8983 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8984 internal_fn_stored_value_index.
8985 (check_load_store_masking): Handle scatter stores too.
8986 (vect_get_store_rhs): Make public.
8987 (vectorizable_call): Use internal_store_fn_p.
8988 (vectorizable_store): Handle scatter store internal functions.
8989 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8990 when deciding whether the end of the group has been reached.
8991 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8992 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8993 (mask_scatter_store<mode>): New insns.
8995 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8996 Alan Hayward <alan.hayward@arm.com>
8997 David Sherwood <david.sherwood@arm.com>
8999 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9000 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9001 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9003 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9004 Use vect_truncate_gather_scatter_offset if we can't treat the
9005 operation as a normal gather load or scatter store.
9006 (get_group_load_store_type): Take the gather_scatter_info
9007 as argument. Try using a gather load or scatter store for
9008 single-element groups.
9009 (get_load_store_type): Update calls to get_group_load_store_type
9010 and vect_use_strided_gather_scatters_p.
9012 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9013 Alan Hayward <alan.hayward@arm.com>
9014 David Sherwood <david.sherwood@arm.com>
9016 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9017 optional tree argument.
9018 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9020 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9021 but continue to use the current value as a fallback.
9022 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9023 to compare the updates.
9024 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9025 (get_load_store_type): Use it when handling a strided access.
9026 (vect_get_strided_load_store_ops): New function.
9027 (vect_get_data_ptr_increment): Likewise.
9028 (vectorizable_load): Handle strided gather loads. Always pass
9029 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9031 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9032 Alan Hayward <alan.hayward@arm.com>
9033 David Sherwood <david.sherwood@arm.com>
9035 * doc/md.texi (gather_load@var{m}): Document.
9036 (mask_gather_load@var{m}): Likewise.
9037 * genopinit.c (main): Add supports_vec_gather_load and
9038 supports_vec_gather_load_cached to target_optabs.
9039 * optabs-tree.c (init_tree_optimization_optabs): Use
9040 ggc_cleared_alloc to allocate target_optabs.
9041 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9042 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9044 * internal-fn.h (internal_load_fn_p): Declare.
9045 (internal_gather_scatter_fn_p): Likewise.
9046 (internal_fn_mask_index): Likewise.
9047 (internal_gather_scatter_fn_supported_p): Likewise.
9048 * internal-fn.c (gather_load_direct): New macro.
9049 (expand_gather_load_optab_fn): New function.
9050 (direct_gather_load_optab_supported_p): New macro.
9051 (direct_internal_fn_optab): New function.
9052 (internal_load_fn_p): Likewise.
9053 (internal_gather_scatter_fn_p): Likewise.
9054 (internal_fn_mask_index): Likewise.
9055 (internal_gather_scatter_fn_supported_p): Likewise.
9056 * optabs-query.c (supports_at_least_one_mode_p): New function.
9057 (supports_vec_gather_load_p): Likewise.
9058 * optabs-query.h (supports_vec_gather_load_p): Declare.
9059 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9060 and memory_type field.
9061 (NUM_PATTERNS): Bump to 15.
9062 * tree-vect-data-refs.c: Include internal-fn.h.
9063 (vect_gather_scatter_fn_p): New function.
9064 (vect_describe_gather_scatter_call): Likewise.
9065 (vect_check_gather_scatter): Try using internal functions for
9066 gather loads. Recognize existing calls to a gather load function.
9067 (vect_analyze_data_refs): Consider using gather loads if
9068 supports_vec_gather_load_p.
9069 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9070 (vect_get_gather_scatter_offset_type): Likewise.
9071 (vect_convert_mask_for_vectype): Likewise.
9072 (vect_add_conversion_to_patterm): Likewise.
9073 (vect_try_gather_scatter_pattern): Likewise.
9074 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9075 (vect_vect_recog_func_ptrs): Add it.
9076 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9077 internal_fn_mask_index and internal_gather_scatter_fn_p.
9078 (check_load_store_masking): Take the gather_scatter_info as an
9079 argument and handle gather loads.
9080 (vect_get_gather_scatter_ops): New function.
9081 (vectorizable_call): Check internal_load_fn_p.
9082 (vectorizable_load): Likewise. Handle gather load internal
9084 (vectorizable_store): Update call to check_load_store_masking.
9085 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9086 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9087 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9088 (aarch64_gather_scale_operand_d): New predicates.
9089 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9090 (mask_gather_load<mode>): New insns.
9092 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9093 Alan Hayward <alan.hayward@arm.com>
9094 David Sherwood <david.sherwood@arm.com>
9096 * optabs.def (fold_left_plus_optab): New optab.
9097 * doc/md.texi (fold_left_plus_@var{m}): Document.
9098 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9099 * internal-fn.c (fold_left_direct): Define.
9100 (expand_fold_left_optab_fn): Likewise.
9101 (direct_fold_left_optab_supported_p): Likewise.
9102 * fold-const-call.c (fold_const_fold_left): New function.
9103 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9104 * tree-parloops.c (valid_reduction_p): New function.
9105 (gather_scalar_reductions): Use it.
9106 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9107 (vect_finish_replace_stmt): Declare.
9108 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9109 (needs_fold_left_reduction_p): New function, split out from...
9110 (vect_is_simple_reduction): ...here. Accept reductions that
9111 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9112 (vect_force_simple_reduction): Also store the reduction type in
9113 the assignment's STMT_VINFO_REDUC_TYPE.
9114 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9115 (merge_with_identity): New function.
9116 (vect_expand_fold_left): Likewise.
9117 (vectorize_fold_left_reduction): Likewise.
9118 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9119 scalar phi in place for it. Check for target support and reject
9120 cases that would reassociate the operation. Defer the transform
9121 phase to vectorize_fold_left_reduction.
9122 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9123 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9124 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9126 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9128 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9129 call to ifc_temp_var.
9131 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9132 Alan Hayward <alan.hayward@arm.com>
9133 David Sherwood <david.sherwood@arm.com>
9135 * target.def (legitimize_address_displacement): Take the original
9136 offset as a poly_int.
9137 * targhooks.h (default_legitimize_address_displacement): Update
9139 * targhooks.c (default_legitimize_address_displacement): Likewise.
9140 * doc/tm.texi: Regenerate.
9141 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9142 as an argument, moving assert of ad->disp == ad->disp_term to...
9143 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9144 Try calling targetm.legitimize_address_displacement before expanding
9145 the address rather than afterwards, and adjust for the new interface.
9146 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9147 Match the new hook interface. Handle SVE addresses.
9148 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9151 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9153 * Makefile.in (OBJS): Add early-remat.o.
9154 * target.def (select_early_remat_modes): New hook.
9155 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9156 * doc/tm.texi: Regenerate.
9157 * targhooks.h (default_select_early_remat_modes): Declare.
9158 * targhooks.c (default_select_early_remat_modes): New function.
9159 * timevar.def (TV_EARLY_REMAT): New timevar.
9160 * passes.def (pass_early_remat): New pass.
9161 * tree-pass.h (make_pass_early_remat): Declare.
9162 * early-remat.c: New file.
9163 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9165 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9167 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9168 Alan Hayward <alan.hayward@arm.com>
9169 David Sherwood <david.sherwood@arm.com>
9171 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9172 vfm1 with a bound_epilog parameter.
9173 (vect_do_peeling): Update calls accordingly, and move the prologue
9174 call earlier in the function. Treat the base bound_epilog as 0 for
9175 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9176 this base when peeling for gaps.
9177 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9178 with fully-masked loops.
9179 (vect_estimate_min_profitable_iters): Handle the single peeled
9180 iteration in that case.
9182 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9183 Alan Hayward <alan.hayward@arm.com>
9184 David Sherwood <david.sherwood@arm.com>
9186 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9187 single-element interleaving even if the size is not a power of 2.
9188 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9189 accesses for single-element interleaving if the group size is
9192 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9193 Alan Hayward <alan.hayward@arm.com>
9194 David Sherwood <david.sherwood@arm.com>
9196 * doc/md.texi (fold_extract_last_@var{m}): Document.
9197 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9198 * optabs.def (fold_extract_last_optab): New optab.
9199 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9200 * internal-fn.c (fold_extract_direct): New macro.
9201 (expand_fold_extract_optab_fn): Likewise.
9202 (direct_fold_extract_optab_supported_p): Likewise.
9203 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9204 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9205 EXTRACT_LAST_REDUCTION.
9206 (get_initial_def_for_reduction): Do not create an initial vector
9207 for EXTRACT_LAST_REDUCTION reductions.
9208 (vectorizable_reduction): Leave the scalar phi in place for
9209 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9210 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9211 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9212 transform phase to vectorizable_condition.
9213 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9215 (vect_finish_stmt_generation): ...here.
9216 (vect_finish_replace_stmt): New function.
9217 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9218 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9220 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9222 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9223 Alan Hayward <alan.hayward@arm.com>
9224 David Sherwood <david.sherwood@arm.com>
9226 * doc/md.texi (extract_last_@var{m}): Document.
9227 * optabs.def (extract_last_optab): New optab.
9228 * internal-fn.def (EXTRACT_LAST): New internal function.
9229 * internal-fn.c (cond_unary_direct): New macro.
9230 (expand_cond_unary_optab_fn): Likewise.
9231 (direct_cond_unary_optab_supported_p): Likewise.
9232 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9233 loops using EXTRACT_LAST.
9234 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9235 (extract_last_<mode>): ...this optab.
9236 (vec_extract<mode><Vel>): Update accordingly.
9238 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9239 Alan Hayward <alan.hayward@arm.com>
9240 David Sherwood <david.sherwood@arm.com>
9242 * target.def (empty_mask_is_expensive): New hook.
9243 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9244 * doc/tm.texi: Regenerate.
9245 * targhooks.h (default_empty_mask_is_expensive): Declare.
9246 * targhooks.c (default_empty_mask_is_expensive): New function.
9247 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9248 if the target says that empty masks are expensive.
9249 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9251 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9253 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9254 Alan Hayward <alan.hayward@arm.com>
9255 David Sherwood <david.sherwood@arm.com>
9257 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9258 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9259 (vect_use_loop_mask_for_alignment_p): New function.
9260 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9261 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9262 niters_skip argument. Make sure that the first niters_skip elements
9263 of the first iteration are inactive.
9264 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9265 Update call to vect_set_loop_masks_directly.
9266 (get_misalign_in_elems): New function, split out from...
9267 (vect_gen_prolog_loop_niters): ...here.
9268 (vect_update_init_of_dr): Take a code argument that specifies whether
9269 the adjustment should be added or subtracted.
9270 (vect_update_init_of_drs): Likewise.
9271 (vect_prepare_for_masked_peels): New function.
9272 (vect_do_peeling): Skip prologue peeling if we're using a mask
9273 instead. Update call to vect_update_inits_of_drs.
9274 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9276 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9277 alignment. Do not include the number of peeled iterations in
9278 the minimum threshold in that case.
9279 (vectorizable_induction): Adjust the start value down by
9280 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9281 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9282 Take the number of skipped iterations into account when calculating
9284 * tree-vect-stmts.c (vect_gen_while_not): New function.
9286 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9287 Alan Hayward <alan.hayward@arm.com>
9288 David Sherwood <david.sherwood@arm.com>
9290 * doc/sourcebuild.texi (vect_fully_masked): Document.
9291 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9293 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9295 (vect_analyze_loop_2): ...here. Don't check the vectorization
9296 factor against the number of loop iterations if the loop is
9299 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9300 Alan Hayward <alan.hayward@arm.com>
9301 David Sherwood <david.sherwood@arm.com>
9303 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9304 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9305 (dump_groups): Update accordingly.
9306 (iv_use::mem_type): New member variable.
9307 (address_p): New function.
9308 (record_use): Add a mem_type argument and initialize the new
9310 (record_group_use): Add a mem_type argument. Use address_p.
9311 Remove obsolete null checks of base_object. Update call to record_use.
9312 (find_interesting_uses_op): Update call to record_group_use.
9313 (find_interesting_uses_cond): Likewise.
9314 (find_interesting_uses_address): Likewise.
9315 (get_mem_type_for_internal_fn): New function.
9316 (find_address_like_use): Likewise.
9317 (find_interesting_uses_stmt): Try find_address_like_use before
9318 calling find_interesting_uses_op.
9319 (addr_offset_valid_p): Use the iv mem_type field as the type
9320 of the addressed memory.
9321 (add_autoinc_candidates): Likewise.
9322 (get_address_cost): Likewise.
9323 (split_small_address_groups_p): Use address_p.
9324 (split_address_groups): Likewise.
9325 (add_iv_candidate_for_use): Likewise.
9326 (autoinc_possible_for_pair): Likewise.
9327 (rewrite_groups): Likewise.
9328 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9329 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9330 (get_alias_ptr_type_for_ptr_address): New function.
9331 (rewrite_use_address): Rewrite address uses in calls that were
9332 identified by find_address_like_use.
9334 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9335 Alan Hayward <alan.hayward@arm.com>
9336 David Sherwood <david.sherwood@arm.com>
9338 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9340 * gimple-expr.h (is_gimple_addressable: Likewise.
9341 * gimple-expr.c (is_gimple_address): Likewise.
9342 * internal-fn.c (expand_call_mem_ref): New function.
9343 (expand_mask_load_optab_fn): Use it.
9344 (expand_mask_store_optab_fn): Likewise.
9346 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9347 Alan Hayward <alan.hayward@arm.com>
9348 David Sherwood <david.sherwood@arm.com>
9350 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9351 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9352 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9353 (cond_umax@var{mode}): Document.
9354 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9355 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9356 (cond_umin_optab, cond_umax_optab): New optabs.
9357 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9358 (COND_IOR, COND_XOR): New internal functions.
9359 * internal-fn.h (get_conditional_internal_fn): Declare.
9360 * internal-fn.c (cond_binary_direct): New macro.
9361 (expand_cond_binary_optab_fn): Likewise.
9362 (direct_cond_binary_optab_supported_p): Likewise.
9363 (get_conditional_internal_fn): New function.
9364 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9365 Cope with reduction statements that are vectorized as calls rather
9367 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9368 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9369 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9370 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9371 (UNSPEC_COND_EOR): New unspecs.
9372 (optab): Add mappings for them.
9373 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9374 (sve_int_op, sve_fp_op): New int attributes.
9376 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9377 Alan Hayward <alan.hayward@arm.com>
9378 David Sherwood <david.sherwood@arm.com>
9380 * optabs.def (while_ult_optab): New optab.
9381 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9382 * internal-fn.def (WHILE_ULT): New internal function.
9383 * internal-fn.h (direct_internal_fn_supported_p): New override
9384 that takes two types as argument.
9385 * internal-fn.c (while_direct): New macro.
9386 (expand_while_optab_fn): New function.
9387 (convert_optab_supported_p): Likewise.
9388 (direct_while_optab_supported_p): New macro.
9389 * wide-int.h (wi::udiv_ceil): New function.
9390 * tree-vectorizer.h (rgroup_masks): New structure.
9391 (vec_loop_masks): New typedef.
9392 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9394 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9395 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9396 (vect_max_vf): New function.
9397 (slpeel_make_loop_iterate_ntimes): Delete.
9398 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9399 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9400 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9401 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9402 internal-fn.h, stor-layout.h and optabs-query.h.
9403 (vect_set_loop_mask): New function.
9404 (add_preheader_seq): Likewise.
9405 (add_header_seq): Likewise.
9406 (interleave_supported_p): Likewise.
9407 (vect_maybe_permute_loop_masks): Likewise.
9408 (vect_set_loop_masks_directly): Likewise.
9409 (vect_set_loop_condition_masked): Likewise.
9410 (vect_set_loop_condition_unmasked): New function, split out from
9411 slpeel_make_loop_iterate_ntimes.
9412 (slpeel_make_loop_iterate_ntimes): Rename to..
9413 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9414 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9415 (vect_do_peeling): Update call accordingly.
9416 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9418 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9419 mask_compare_type, can_fully_mask_p and fully_masked_p.
9420 (release_vec_loop_masks): New function.
9421 (_loop_vec_info): Use it to free the loop masks.
9422 (can_produce_all_loop_masks_p): New function.
9423 (vect_get_max_nscalars_per_iter): Likewise.
9424 (vect_verify_full_masking): Likewise.
9425 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9426 retries, and free the mask rgroups before retrying. Check loop-wide
9427 reasons for disallowing fully-masked loops. Make the final decision
9428 about whether use a fully-masked loop or not.
9429 (vect_estimate_min_profitable_iters): Do not assume that peeling
9430 for the number of iterations will be needed for fully-masked loops.
9431 (vectorizable_reduction): Disable fully-masked loops.
9432 (vectorizable_live_operation): Likewise.
9433 (vect_halve_mask_nunits): New function.
9434 (vect_double_mask_nunits): Likewise.
9435 (vect_record_loop_mask): Likewise.
9436 (vect_get_loop_mask): Likewise.
9437 (vect_transform_loop): Handle the case in which the final loop
9438 iteration might handle a partial vector. Call vect_set_loop_condition
9439 instead of slpeel_make_loop_iterate_ntimes.
9440 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9441 (check_load_store_masking): New function.
9442 (prepare_load_store_mask): Likewise.
9443 (vectorizable_store): Handle fully-masked loops.
9444 (vectorizable_load): Likewise.
9445 (supportable_widening_operation): Use vect_halve_mask_nunits for
9447 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9448 (vect_gen_while): New function.
9449 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9450 (aarch64_uqdec<mode>): New insn.
9452 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9453 Alan Hayward <alan.hayward@arm.com>
9454 David Sherwood <david.sherwood@arm.com>
9456 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9457 (reduc_xor_scal_optab): New optabs.
9458 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9459 (reduc_xor_scal_@var{m}): Document.
9460 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9461 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9463 * fold-const-call.c (fold_const_call): Handle them.
9464 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9465 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9466 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9467 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9468 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9469 (UNSPEC_XORV): New unspecs.
9470 (optab): Add entries for them.
9471 (BITWISEV): New int iterator.
9472 (bit_reduc_op): New int attributes.
9474 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9475 Alan Hayward <alan.hayward@arm.com>
9476 David Sherwood <david.sherwood@arm.com>
9478 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9479 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9480 * optabs.def (vec_shl_insert_optab): New optab.
9481 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9482 (duplicate_and_interleave): Likewise.
9483 * tree-vect-loop.c: Include internal-fn.h.
9484 (neutral_op_for_slp_reduction): New function, split out from
9485 get_initial_defs_for_reduction.
9486 (get_initial_def_for_reduction): Handle option 2 for variable-length
9487 vectors by loading the neutral value into a vector and then shifting
9488 the initial value into element 0.
9489 (get_initial_defs_for_reduction): Replace the code argument with
9490 the neutral value calculated by neutral_op_for_slp_reduction.
9491 Use gimple_build_vector for constant-length vectors.
9492 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9493 but the first group_size elements have a neutral value.
9494 Use duplicate_and_interleave otherwise.
9495 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9496 Update call to get_initial_defs_for_reduction. Handle SLP
9497 reductions for variable-length vectors by creating one vector
9498 result for each scalar result, with the elements associated
9499 with other scalar results stubbed out with the neutral value.
9500 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9501 Require IFN_VEC_SHL_INSERT for double reductions on
9502 variable-length vectors, or SLP reductions that have
9503 a neutral value. Require can_duplicate_and_interleave_p
9504 support for variable-length unchained SLP reductions if there
9505 is no neutral value, such as for MIN/MAX reductions. Also require
9506 the number of vector elements to be a multiple of the number of
9507 SLP statements when doing variable-length unchained SLP reductions.
9508 Update call to vect_create_epilog_for_reduction.
9509 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9510 and remove initial values.
9511 (duplicate_and_interleave): Make public.
9512 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9513 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9515 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9516 Alan Hayward <alan.hayward@arm.com>
9517 David Sherwood <david.sherwood@arm.com>
9519 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9520 (can_duplicate_and_interleave_p): New function.
9521 (vect_get_and_check_slp_defs): Take the vector of statements
9522 rather than just the current one. Remove excess parentheses.
9523 Restriction rejectinon of vect_constant_def and vect_external_def
9524 for variable-length vectors to boolean types, or types for which
9525 can_duplicate_and_interleave_p is false.
9526 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9527 (duplicate_and_interleave): New function.
9528 (vect_get_constant_vectors): Use gimple_build_vector for
9529 constant-length vectors and suitable variable-length constant
9530 vectors. Use duplicate_and_interleave for other variable-length
9531 vectors. Don't defer the update when inserting new statements.
9533 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9534 Alan Hayward <alan.hayward@arm.com>
9535 David Sherwood <david.sherwood@arm.com>
9537 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9538 min_profitable_iters doesn't go negative.
9540 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9541 Alan Hayward <alan.hayward@arm.com>
9542 David Sherwood <david.sherwood@arm.com>
9544 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9545 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9546 * optabs.def (vec_mask_load_lanes_optab): New optab.
9547 (vec_mask_store_lanes_optab): Likewise.
9548 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9549 (MASK_STORE_LANES): Likewise.
9550 * internal-fn.c (mask_load_lanes_direct): New macro.
9551 (mask_store_lanes_direct): Likewise.
9552 (expand_mask_load_optab_fn): Handle masked operations.
9553 (expand_mask_load_lanes_optab_fn): New macro.
9554 (expand_mask_store_optab_fn): Handle masked operations.
9555 (expand_mask_store_lanes_optab_fn): New macro.
9556 (direct_mask_load_lanes_optab_supported_p): Likewise.
9557 (direct_mask_store_lanes_optab_supported_p): Likewise.
9558 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9560 (vect_load_lanes_supported): Likewise.
9561 * tree-vect-data-refs.c (strip_conversion): New function.
9562 (can_group_stmts_p): Likewise.
9563 (vect_analyze_data_ref_accesses): Use it instead of checking
9564 for a pair of assignments.
9565 (vect_store_lanes_supported): Take a masked_p parameter.
9566 (vect_load_lanes_supported): Likewise.
9567 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9568 vect_store_lanes_supported and vect_load_lanes_supported.
9569 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9570 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9571 parameter. Don't allow gaps for masked accesses.
9572 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9573 and vect_load_lanes_supported.
9574 (get_load_store_type): Take a masked_p parameter and update
9575 call to get_group_load_store_type.
9576 (vectorizable_store): Update call to get_load_store_type.
9577 Handle IFN_MASK_STORE_LANES.
9578 (vectorizable_load): Update call to get_load_store_type.
9579 Handle IFN_MASK_LOAD_LANES.
9581 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9582 Alan Hayward <alan.hayward@arm.com>
9583 David Sherwood <david.sherwood@arm.com>
9585 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9587 * config/aarch64/aarch64-protos.h
9588 (aarch64_sve_struct_memory_operand_p): Declare.
9589 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9590 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9591 (VPRED, vpred): Handle SVE structure modes.
9592 * config/aarch64/constraints.md (Utx): New constraint.
9593 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9594 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9595 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9596 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9597 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9598 structure modes. Split into pieces after RA.
9599 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9600 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9602 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9603 SVE structure modes.
9604 (aarch64_classify_address): Likewise.
9605 (sizetochar): Move earlier in file.
9606 (aarch64_print_operand): Handle SVE register lists.
9607 (aarch64_array_mode): New function.
9608 (aarch64_sve_struct_memory_operand_p): Likewise.
9609 (TARGET_ARRAY_MODE): Redefine.
9611 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9612 Alan Hayward <alan.hayward@arm.com>
9613 David Sherwood <david.sherwood@arm.com>
9615 * target.def (array_mode): New target hook.
9616 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9617 * doc/tm.texi: Regenerate.
9618 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9619 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9620 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9622 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9625 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9626 Alan Hayward <alan.hayward@arm.com>
9627 David Sherwood <david.sherwood@arm.com>
9629 * fold-const.c (fold_binary_loc): Check the argument types
9630 rather than the result type when testing for a vector operation.
9632 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9634 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9635 * doc/tm.texi: Regenerate.
9637 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9638 Alan Hayward <alan.hayward@arm.com>
9639 David Sherwood <david.sherwood@arm.com>
9641 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9642 (sve): Document new AArch64 extension.
9643 * doc/md.texi (w): Extend the description of the AArch64
9644 constraint to include SVE vectors.
9645 (Upl, Upa): Document new AArch64 predicate constraints.
9646 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9648 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9649 (msve-vector-bits=): New option.
9650 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9651 SVE when these are disabled.
9652 (sve): New extension.
9653 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9654 modes. Adjust their number of units based on aarch64_sve_vg.
9655 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9656 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9657 aarch64_addr_query_type.
9658 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9659 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9660 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9661 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9662 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9663 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9664 (aarch64_simd_imm_zero_p): Delete.
9665 (aarch64_check_zero_based_sve_index_immediate): Declare.
9666 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9667 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9668 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9669 (aarch64_sve_float_mul_immediate_p): Likewise.
9670 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9672 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9673 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9674 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9675 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9676 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9677 (aarch64_regmode_natural_size): Likewise.
9678 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9679 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9681 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9682 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9683 for VG and the SVE predicate registers.
9684 (V_ALIASES): Add a "z"-prefixed alias.
9685 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9686 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9687 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9688 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9689 (REG_CLASS_NAMES): Add entries for them.
9690 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9691 and the predicate registers.
9692 (aarch64_sve_vg): Declare.
9693 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9694 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9695 (REGMODE_NATURAL_SIZE): Define.
9696 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9698 * config/aarch64/aarch64.c: Include cfgrtl.h.
9699 (simd_immediate_info): Add a constructor for series vectors,
9700 and an associated step field.
9701 (aarch64_sve_vg): New variable.
9702 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9703 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9704 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9705 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9706 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9707 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9708 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9709 (aarch64_get_mask_mode): New functions.
9710 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9711 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9712 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9713 predicate modes and predicate registers. Explicitly restrict
9714 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9715 to store a vector mode if it is recognized by
9716 aarch64_classify_vector_mode.
9717 (aarch64_regmode_natural_size): New function.
9718 (aarch64_hard_regno_caller_save_mode): Return the original mode
9720 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9721 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9722 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9723 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9725 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9726 does not overlap dest if the function is frame-related. Handle
9728 (aarch64_split_add_offset): New function.
9729 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9730 them aarch64_add_offset.
9731 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9732 and update call to aarch64_sub_sp.
9733 (aarch64_add_cfa_expression): New function.
9734 (aarch64_expand_prologue): Pass extra temporary registers to the
9735 functions above. Handle the case in which we need to emit new
9736 DW_CFA_expressions for registers that were originally saved
9737 relative to the stack pointer, but now have to be expressed
9738 relative to the frame pointer.
9739 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9741 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9742 IP0 and IP1 values for SVE frames.
9743 (aarch64_expand_vec_series): New function.
9744 (aarch64_expand_sve_widened_duplicate): Likewise.
9745 (aarch64_expand_sve_const_vector): Likewise.
9746 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9747 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9748 into the register, rather than emitting a SET directly.
9749 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9750 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9751 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9752 (offset_9bit_signed_scaled_p): New functions.
9753 (aarch64_replicate_bitmask_imm): New function.
9754 (aarch64_bitmask_imm): Use it.
9755 (aarch64_cannot_force_const_mem): Reject expressions involving
9756 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9757 (aarch64_classify_index): Handle SVE indices, by requiring
9758 a plain register index with a scale that matches the element size.
9759 (aarch64_classify_address): Handle SVE addresses. Assert that
9760 the mode of the address is VOIDmode or an integer mode.
9761 Update call to aarch64_classify_symbol.
9762 (aarch64_classify_symbolic_expression): Update call to
9763 aarch64_classify_symbol.
9764 (aarch64_const_vec_all_in_range_p): New function.
9765 (aarch64_print_vector_float_operand): Likewise.
9766 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9767 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9768 and the FP immediates 1.0 and 0.5.
9769 (aarch64_print_address_internal): Handle SVE addresses.
9770 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9771 (aarch64_regno_regclass): Handle predicate registers.
9772 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9774 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9775 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9776 (aarch64_convert_sve_vector_bits): New function.
9777 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9778 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9780 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9781 Handle SVE vector and predicate modes. Accept VL-based constants
9782 that need only one temporary register, and VL offsets that require
9783 no temporary registers.
9784 (aarch64_conditional_register_usage): Mark the predicate registers
9785 as fixed if SVE isn't available.
9786 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9787 Return true for SVE vector and predicate modes.
9788 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9789 rather than an unsigned int. Handle SVE modes.
9790 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9792 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9794 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9795 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9796 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9797 (aarch64_sve_float_mul_immediate_p): New functions.
9798 (aarch64_sve_valid_immediate): New function.
9799 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9800 Explicitly reject structure modes. Check for INDEX constants.
9801 Handle PTRUE and PFALSE constants.
9802 (aarch64_check_zero_based_sve_index_immediate): New function.
9803 (aarch64_simd_imm_zero_p): Delete.
9804 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9805 vector modes. Accept constants in the range of CNT[BHWD].
9806 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9807 ask for an Advanced SIMD mode.
9808 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9809 (aarch64_simd_vector_alignment): Handle SVE predicates.
9810 (aarch64_vectorize_preferred_vector_alignment): New function.
9811 (aarch64_simd_vector_alignment_reachable): Use it instead of
9813 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9814 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9816 (MAX_VECT_LEN): Delete.
9817 (expand_vec_perm_d): Add a vec_flags field.
9818 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9819 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9820 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9822 (aarch64_evpc_rev): Rename to...
9823 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9824 (aarch64_evpc_rev_global): New function.
9825 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9826 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9828 (aarch64_evpc_sve_tbl): New function.
9829 (aarch64_expand_vec_perm_const_1): Update after rename of
9830 aarch64_evpc_rev. Handle SVE permutes too, trying
9831 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9832 than aarch64_evpc_tbl.
9833 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9834 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9835 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9836 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9837 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9838 (aarch64_expand_sve_vcond): New functions.
9839 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9840 of aarch64_vector_mode_p.
9841 (aarch64_dwarf_poly_indeterminate_value): New function.
9842 (aarch64_compute_pressure_classes): Likewise.
9843 (aarch64_can_change_mode_class): Likewise.
9844 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9845 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9846 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9847 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9848 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9849 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9850 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9851 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9853 (Dn, Dl, Dr): Accept const as well as const_vector.
9854 (Dz): Likewise. Compare against CONST0_RTX.
9855 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9856 of "vector" where appropriate.
9857 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9858 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9859 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9860 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9861 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9862 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9863 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9864 (v_int_equiv): Extend to SVE modes.
9865 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9867 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9868 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9869 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9870 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9871 (SVE_COND_FP_CMP): New int iterators.
9872 (perm_hilo): Handle the new unpack unspecs.
9873 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9875 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9876 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9877 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9878 (aarch64_equality_operator, aarch64_constant_vector_operand)
9879 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9880 (aarch64_sve_nonimmediate_operand): Likewise.
9881 (aarch64_sve_general_operand): Likewise.
9882 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9883 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9884 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9885 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9886 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9887 (aarch64_sve_float_arith_immediate): Likewise.
9888 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9889 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9890 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9891 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9892 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9893 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9894 (aarch64_sve_float_arith_operand): Likewise.
9895 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9896 (aarch64_sve_float_mul_operand): Likewise.
9897 (aarch64_sve_vec_perm_operand): Likewise.
9898 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9899 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9900 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9901 as well as const_vector.
9902 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9903 in file. Use CONST0_RTX and CONSTM1_RTX.
9904 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9905 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9906 Use aarch64_simd_imm_zero.
9907 * config/aarch64/aarch64-sve.md: New file.
9908 * config/aarch64/aarch64.md: Include it.
9909 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9910 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9911 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9912 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9913 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9914 (sve): New attribute.
9915 (enabled): Disable instructions with the sve attribute unless
9917 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9918 aarch64_expand_mov_immediate.
9919 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9920 CNT[BHSD] immediates.
9921 (movti): Split CONST_POLY_INT moves into two halves.
9922 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9923 Split additions that need a temporary here if the destination
9924 is the stack pointer.
9925 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9926 (*add<mode>3_poly_1): New instruction.
9927 (set_clobber_cc): New expander.
9929 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9931 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9932 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9933 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9934 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9935 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9936 Change innermode from fixed_mode_size to machine_mode.
9937 (simplify_subreg): Update call accordingly. Handle a constant-sized
9938 subreg of a variable-length CONST_VECTOR.
9940 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9941 Alan Hayward <alan.hayward@arm.com>
9942 David Sherwood <david.sherwood@arm.com>
9944 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9945 (add_offset_to_base): New function, split out from...
9946 (create_mem_ref): ...here. When handling a scale other than 1,
9947 check first whether the address is valid without the offset.
9948 Add it into the base if so, leaving the index and scale as-is.
9950 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9953 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9954 fold_for_warn before checking if arg2 is INTEGER_CST.
9956 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9958 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9959 (store_multiple_operation): Delete.
9960 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9961 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9962 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9963 guarded by TARGET_STRING.
9964 (rs6000_output_load_multiple): Delete.
9965 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9966 OPTION_MASK_STRING / TARGET_STRING handling.
9967 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9968 (const rs6000_opt_masks) <"string">: Change mask to 0.
9969 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9970 (MASK_STRING): Delete.
9971 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9973 (load_multiple): Delete.
9980 (store_multiple): Delete.
9987 (movmemsi_8reg): Delete.
9988 (corresponding unnamed define_insn): Delete.
9989 (movmemsi_6reg): Delete.
9990 (corresponding unnamed define_insn): Delete.
9991 (movmemsi_4reg): Delete.
9992 (corresponding unnamed define_insn): Delete.
9993 (movmemsi_2reg): Delete.
9994 (corresponding unnamed define_insn): Delete.
9995 (movmemsi_1reg): Delete.
9996 (corresponding unnamed define_insn): Delete.
9997 * config/rs6000/rs6000.opt (mno-string): New.
9998 (mstring): Replace by deprecation warning stub.
9999 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10001 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10003 * regrename.c (regrename_do_replace): If replacing the same
10004 reg multiple times, try to reuse last created gen_raw_REG.
10007 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10008 main to workaround a bug in GDB.
10010 2018-01-12 Tom de Vries <tom@codesourcery.com>
10013 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10015 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10017 PR rtl-optimization/80481
10018 * ira-color.c (get_cap_member): New function.
10019 (allocnos_conflict_by_live_ranges_p): Use it.
10020 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10021 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10023 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10026 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10027 (*saddl_se_1): Ditto.
10028 (*ssubsi_1): Ditto.
10029 (*ssubl_se_1): Ditto.
10031 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10033 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10034 rather than wi::to_widest for DR_INITs.
10035 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10036 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10037 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10039 (vect_analyze_group_access_1): Note that here.
10041 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10043 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10044 polynomial type sizes.
10046 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10048 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10049 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10050 (gimple_add_tmp_var): Likewise.
10052 2018-01-12 Martin Liska <mliska@suse.cz>
10054 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10055 (gimple_alloc_sizes): Likewise.
10056 (dump_gimple_statistics): Use PRIu64 in printf format.
10057 * gimple.h: Change uint64_t to int.
10059 2018-01-12 Martin Liska <mliska@suse.cz>
10061 * tree-core.h: Use uint64_t instead of int.
10062 * tree.c (tree_node_counts): Likewise.
10063 (tree_node_sizes): Likewise.
10064 (dump_tree_statistics): Use PRIu64 in printf format.
10066 2018-01-12 Martin Liska <mliska@suse.cz>
10068 * Makefile.in: As qsort_chk is implemented in vec.c, add
10069 vec.o to linkage of gencfn-macros.
10070 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10071 passing the info to record_node_allocation_statistics.
10072 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10074 * ggc-common.c (struct ggc_usage): Add operator== and use
10075 it in operator< and compare function.
10076 * mem-stats.h (struct mem_usage): Likewise.
10077 * vec.c (struct vec_usage): Remove operator< and compare
10078 function. Can be simply inherited.
10080 2018-01-12 Martin Jambor <mjambor@suse.cz>
10083 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10084 * tree-ssa-math-opts.c: Include domwalk.h.
10085 (convert_mult_to_fma_1): New function.
10086 (fma_transformation_info): New type.
10087 (fma_deferring_state): Likewise.
10088 (cancel_fma_deferring): New function.
10089 (result_of_phi): Likewise.
10090 (last_fma_candidate_feeds_initial_phi): Likewise.
10091 (convert_mult_to_fma): Added deferring logic, split actual
10092 transformation to convert_mult_to_fma_1.
10093 (math_opts_dom_walker): New type.
10094 (math_opts_dom_walker::after_dom_children): New method, body moved
10095 here from pass_optimize_widening_mul::execute, added deferring logic
10097 (pass_optimize_widening_mul::execute): Moved most of code to
10098 math_opts_dom_walker::after_dom_children.
10099 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10100 * config/i386/i386.c (ix86_option_override_internal): Added
10101 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10103 2018-01-12 Richard Biener <rguenther@suse.de>
10106 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10107 inline instance vars.
10109 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10112 * config/rx/rx.c (rx_is_restricted_memory_address):
10113 Handle SUBREG case.
10115 2018-01-12 Richard Biener <rguenther@suse.de>
10117 PR tree-optimization/80846
10118 * target.def (split_reduction): New target hook.
10119 * targhooks.c (default_split_reduction): New function.
10120 * targhooks.h (default_split_reduction): Declare.
10121 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10122 target requests first reduce vectors by combining low and high
10124 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10125 (get_vectype_for_scalar_type_and_size): Export.
10126 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10127 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10128 * doc/tm.texi: Regenerate.
10129 * config/i386/i386.c (ix86_split_reduction): Implement
10130 TARGET_VECTORIZE_SPLIT_REDUCTION.
10132 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10135 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10136 in PIC mode except for TARGET_VXWORKS_RTP.
10137 * config/sparc/sparc.c: Include cfgrtl.h.
10138 (TARGET_INIT_PIC_REG): Define.
10139 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10140 (sparc_pic_register_p): New predicate.
10141 (sparc_legitimate_address_p): Use it.
10142 (sparc_legitimize_pic_address): Likewise.
10143 (sparc_delegitimize_address): Likewise.
10144 (sparc_mode_dependent_address_p): Likewise.
10145 (gen_load_pcrel_sym): Remove 4th parameter.
10146 (load_got_register): Adjust call to above. Remove obsolete stuff.
10147 (sparc_expand_prologue): Do not call load_got_register here.
10148 (sparc_flat_expand_prologue): Likewise.
10149 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10150 (sparc_use_pseudo_pic_reg): New function.
10151 (sparc_init_pic_reg): Likewise.
10152 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10153 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10155 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10157 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10158 Add item for branch_cost.
10160 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10162 PR rtl-optimization/83565
10163 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10164 not extend the result to a larger mode for rotate operations.
10165 (num_sign_bit_copies1): Likewise.
10167 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10170 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10172 Use values-Xc.o for -pedantic.
10173 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10175 2018-01-12 Martin Liska <mliska@suse.cz>
10178 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10180 (possible_polymorphic_call_targets): Use it.
10181 (ipa_devirt): Likewise.
10183 2018-01-12 Martin Liska <mliska@suse.cz>
10185 * profile-count.h (enum profile_quality): Use 0 as invalid
10186 enum value of profile_quality.
10188 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10190 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10191 -mext-string options.
10193 2018-01-12 Richard Biener <rguenther@suse.de>
10195 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10196 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10197 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10199 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10201 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10203 * configure.ac (--with-long-double-format): Add support for the
10204 configuration option to change the default long double format on
10206 * config.gcc (powerpc*-linux*-*): Likewise.
10207 * configure: Regenerate.
10208 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10209 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10210 used without modification.
10212 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10214 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10215 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10216 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10217 MISC_BUILTIN_SPEC_BARRIER.
10218 (rs6000_init_builtins): Likewise.
10219 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10221 (speculation_barrier): New define_insn.
10222 * doc/extend.texi: Document __builtin_speculation_barrier.
10224 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10227 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10228 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10229 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10231 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10232 integral modes instead of "ss" and "sd".
10233 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10234 vectors with 32-bit and 64-bit elements.
10235 (vecdupssescalarmodesuffix): New mode attribute.
10236 (vec_dup<mode>): Use it.
10238 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10241 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10242 frame if argument is passed on stack.
10244 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10247 * ree.c (combine_reaching_defs): Optimize also
10248 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10249 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10251 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10253 PR middle-end/83189
10254 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10256 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10258 PR middle-end/83718
10259 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10260 after they are computed.
10262 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10264 PR tree-optimization/83695
10265 * gimple-loop-linterchange.cc
10266 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10267 reset cached scev information after interchange.
10268 (pass_linterchange::execute): Remove call to scev_reset_htab.
10270 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10272 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10273 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10274 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10275 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10276 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10277 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10278 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10279 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10280 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10281 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10282 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10283 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10284 (V_lane_reg): Likewise.
10285 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10287 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10288 (vfmal_lane_low<mode>_intrinsic,
10289 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10290 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10291 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10292 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10293 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10294 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10296 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10298 * config/arm/arm-cpus.in (fp16fml): New feature.
10299 (ALL_SIMD): Add fp16fml.
10300 (armv8.2-a): Add fp16fml as an option.
10301 (armv8.3-a): Likewise.
10302 (armv8.4-a): Add fp16fml as part of fp16.
10303 * config/arm/arm.h (TARGET_FP16FML): Define.
10304 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10306 * config/arm/arm-modes.def (V2HF): Define.
10307 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10308 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10309 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10310 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10311 vfmsl_low, vfmsl_high): New set of builtins.
10312 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10313 (vfml_op): New code attribute.
10314 (VFMLHALVES): New int iterator.
10315 (VFML, VFMLSEL): New mode attributes.
10316 (V_reg): Define mapping for V2HF.
10317 (V_hi, V_lo): New mode attributes.
10318 (VF_constraint): Likewise.
10319 (vfml_half, vfml_half_selector): New int attributes.
10320 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10322 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10323 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10325 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10326 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10327 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10328 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10330 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10331 Document new effective target and option set.
10333 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10335 * config/arm/arm-cpus.in (armv8_4): New feature.
10336 (ARMv8_4a): New fgroup.
10337 (armv8.4-a): New arch.
10338 * config/arm/arm-tables.opt: Regenerate.
10339 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10340 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10341 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10342 Add matching rules for -march=armv8.4-a and extensions.
10343 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10345 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10348 * config/rx/rx.md (BW): New mode attribute.
10349 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10351 2018-01-11 Richard Biener <rguenther@suse.de>
10353 PR tree-optimization/83435
10354 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10355 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10356 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10358 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10359 Alan Hayward <alan.hayward@arm.com>
10360 David Sherwood <david.sherwood@arm.com>
10362 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10364 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10365 (aarch64_print_address_internal): Use it to check for a zero offset.
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-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10372 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10373 Return a poly_int64 rather than a HOST_WIDE_INT.
10374 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10375 rather than a HOST_WIDE_INT.
10376 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10377 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10378 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10379 final_offset from HOST_WIDE_INT to poly_int64.
10380 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10381 to_constant when getting the number of units in an Advanced SIMD
10383 (aarch64_builtin_vectorized_function): Check for a constant number
10385 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10387 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10388 attribute instead of GET_MODE_NUNITS.
10389 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10390 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10391 GET_MODE_SIZE for fixed-size registers.
10392 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10393 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10394 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10395 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10396 (aarch64_print_operand, aarch64_print_address_internal)
10397 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10398 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10399 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10400 Handle polynomial GET_MODE_SIZE.
10401 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10402 wider than SImode without modification.
10403 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10404 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10405 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10406 passing and returning SVE modes.
10407 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10408 rather than GEN_INT.
10409 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10410 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10411 (aarch64_allocate_and_probe_stack_space): Likewise.
10412 (aarch64_layout_frame): Cope with polynomial offsets.
10413 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10414 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10415 polynomial offsets.
10416 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10417 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10418 poly_int64 rather than a HOST_WIDE_INT.
10419 (aarch64_get_separate_components, aarch64_process_components)
10420 (aarch64_expand_prologue, aarch64_expand_epilogue)
10421 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10422 (aarch64_anchor_offset): New function, split out from...
10423 (aarch64_legitimize_address): ...here.
10424 (aarch64_builtin_vectorization_cost): Handle polynomial
10425 TYPE_VECTOR_SUBPARTS.
10426 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10428 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10429 number of elements from the PARALLEL rather than the mode.
10430 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10431 rather than GET_MODE_BITSIZE.
10432 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10433 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10434 (aarch64_expand_vec_perm_const_1): Handle polynomial
10435 d->perm.length () and d->perm elements.
10436 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10437 Apply to_constant to d->perm elements.
10438 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10439 polynomial CONST_VECTOR_NUNITS.
10440 (aarch64_move_pointer): Take amount as a poly_int64 rather
10442 (aarch64_progress_pointer): Avoid temporary variable.
10443 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10444 the mode attribute instead of GET_MODE.
10446 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10447 Alan Hayward <alan.hayward@arm.com>
10448 David Sherwood <david.sherwood@arm.com>
10450 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10451 x exists before using it.
10452 (aarch64_add_constant_internal): Rename to...
10453 (aarch64_add_offset_1): ...this. Replace regnum with separate
10454 src and dest rtxes. Handle the case in which they're different,
10455 including when the offset is zero. Replace scratchreg with an rtx.
10456 Use 2 additions if there is no spare register into which we can
10457 move a 16-bit constant.
10458 (aarch64_add_constant): Delete.
10459 (aarch64_add_offset): Replace reg with separate src and dest
10460 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10461 Use aarch64_add_offset_1.
10462 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10463 an rtx rather than an int. Take the delta as a poly_int64
10464 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10465 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10466 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10467 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10468 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10469 and aarch64_add_sp.
10470 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10471 aarch64_add_constant.
10473 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10475 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10476 Use scalar_float_mode.
10478 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10480 * config/aarch64/aarch64-simd.md
10481 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10482 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10483 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10484 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10485 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10486 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10487 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10488 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10489 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10490 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10492 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10495 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10496 targ_options->x_arm_arch_string is non NULL.
10498 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10500 * config/aarch64/aarch64.h
10501 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10503 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10506 * expmed.c (emit_store_flag_force): Swap if const op0
10507 and change VOIDmode to mode of op0.
10509 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10511 PR rtl-optimization/83761
10512 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10513 than bytes to mode_for_size.
10515 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10517 PR middle-end/83189
10518 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10519 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10522 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10524 PR middle-end/83575
10525 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10526 when in layout mode.
10527 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10528 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10531 2018-01-10 Michael Collison <michael.collison@arm.com>
10533 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10534 * config/aarch64/aarch64-option-extension.def: Add
10535 AARCH64_OPT_EXTENSION of 'fp16fml'.
10536 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10537 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10538 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10539 * config/aarch64/constraints.md (Ui7): New constraint.
10540 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10541 (VFMLA_SEL_W): Ditto.
10544 (VFMLA16_LOW): New int iterator.
10545 (VFMLA16_HIGH): Ditto.
10546 (UNSPEC_FMLAL): New unspec.
10547 (UNSPEC_FMLSL): Ditto.
10548 (UNSPEC_FMLAL2): Ditto.
10549 (UNSPEC_FMLSL2): Ditto.
10550 (f16mac): New code attribute.
10551 * config/aarch64/aarch64-simd-builtins.def
10552 (aarch64_fmlal_lowv2sf): Ditto.
10553 (aarch64_fmlsl_lowv2sf): Ditto.
10554 (aarch64_fmlalq_lowv4sf): Ditto.
10555 (aarch64_fmlslq_lowv4sf): Ditto.
10556 (aarch64_fmlal_highv2sf): Ditto.
10557 (aarch64_fmlsl_highv2sf): Ditto.
10558 (aarch64_fmlalq_highv4sf): Ditto.
10559 (aarch64_fmlslq_highv4sf): Ditto.
10560 (aarch64_fmlal_lane_lowv2sf): Ditto.
10561 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10562 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10563 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10564 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10565 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10566 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10567 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10568 (aarch64_fmlal_lane_highv2sf): Ditto.
10569 (aarch64_fmlsl_lane_highv2sf): Ditto.
10570 (aarch64_fmlal_laneq_highv2sf): Ditto.
10571 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10572 (aarch64_fmlalq_lane_highv4sf): Ditto.
10573 (aarch64_fmlsl_lane_highv4sf): Ditto.
10574 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10575 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10576 * config/aarch64/aarch64-simd.md:
10577 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10578 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10579 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10580 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10581 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10582 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10583 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10584 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10585 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10586 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10587 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10588 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10589 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10590 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10591 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10592 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10593 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10594 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10595 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10596 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10597 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10598 (vfmlsl_low_u32): Ditto.
10599 (vfmlalq_low_u32): Ditto.
10600 (vfmlslq_low_u32): Ditto.
10601 (vfmlal_high_u32): Ditto.
10602 (vfmlsl_high_u32): Ditto.
10603 (vfmlalq_high_u32): Ditto.
10604 (vfmlslq_high_u32): Ditto.
10605 (vfmlal_lane_low_u32): Ditto.
10606 (vfmlsl_lane_low_u32): Ditto.
10607 (vfmlal_laneq_low_u32): Ditto.
10608 (vfmlsl_laneq_low_u32): Ditto.
10609 (vfmlalq_lane_low_u32): Ditto.
10610 (vfmlslq_lane_low_u32): Ditto.
10611 (vfmlalq_laneq_low_u32): Ditto.
10612 (vfmlslq_laneq_low_u32): Ditto.
10613 (vfmlal_lane_high_u32): Ditto.
10614 (vfmlsl_lane_high_u32): Ditto.
10615 (vfmlal_laneq_high_u32): Ditto.
10616 (vfmlsl_laneq_high_u32): Ditto.
10617 (vfmlalq_lane_high_u32): Ditto.
10618 (vfmlslq_lane_high_u32): Ditto.
10619 (vfmlalq_laneq_high_u32): Ditto.
10620 (vfmlslq_laneq_high_u32): Ditto.
10621 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10622 (AARCH64_FL_FOR_ARCH8_4): New.
10623 (AARCH64_ISA_F16FML): New ISA flag.
10624 (TARGET_F16FML): New feature flag for fp16fml.
10625 (doc/invoke.texi): Document new fp16fml option.
10627 2018-01-10 Michael Collison <michael.collison@arm.com>
10629 * config/aarch64/aarch64-builtins.c:
10630 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10631 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10632 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10633 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10634 (AARCH64_ISA_SHA3): New ISA flag.
10635 (TARGET_SHA3): New feature flag for sha3.
10636 * config/aarch64/iterators.md (sha512_op): New int attribute.
10637 (CRYPTO_SHA512): New int iterator.
10638 (UNSPEC_SHA512H): New unspec.
10639 (UNSPEC_SHA512H2): Ditto.
10640 (UNSPEC_SHA512SU0): Ditto.
10641 (UNSPEC_SHA512SU1): Ditto.
10642 * config/aarch64/aarch64-simd-builtins.def
10643 (aarch64_crypto_sha512hqv2di): New builtin.
10644 (aarch64_crypto_sha512h2qv2di): Ditto.
10645 (aarch64_crypto_sha512su0qv2di): Ditto.
10646 (aarch64_crypto_sha512su1qv2di): Ditto.
10647 (aarch64_eor3qv8hi): Ditto.
10648 (aarch64_rax1qv2di): Ditto.
10649 (aarch64_xarqv2di): Ditto.
10650 (aarch64_bcaxqv8hi): Ditto.
10651 * config/aarch64/aarch64-simd.md:
10652 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10653 (aarch64_crypto_sha512su0qv2di): Ditto.
10654 (aarch64_crypto_sha512su1qv2di): Ditto.
10655 (aarch64_eor3qv8hi): Ditto.
10656 (aarch64_rax1qv2di): Ditto.
10657 (aarch64_xarqv2di): Ditto.
10658 (aarch64_bcaxqv8hi): Ditto.
10659 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10660 (vsha512h2q_u64): Ditto.
10661 (vsha512su0q_u64): Ditto.
10662 (vsha512su1q_u64): Ditto.
10663 (veor3q_u16): Ditto.
10664 (vrax1q_u64): Ditto.
10665 (vxarq_u64): Ditto.
10666 (vbcaxq_u16): Ditto.
10667 * config/arm/types.md (crypto_sha512): New type attribute.
10668 (crypto_sha3): Ditto.
10669 (doc/invoke.texi): Document new sha3 option.
10671 2018-01-10 Michael Collison <michael.collison@arm.com>
10673 * config/aarch64/aarch64-builtins.c:
10674 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10675 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10676 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10677 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10678 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10679 (AARCH64_ISA_SM4): New ISA flag.
10680 (TARGET_SM4): New feature flag for sm4.
10681 * config/aarch64/aarch64-simd-builtins.def
10682 (aarch64_sm3ss1qv4si): Ditto.
10683 (aarch64_sm3tt1aq4si): Ditto.
10684 (aarch64_sm3tt1bq4si): Ditto.
10685 (aarch64_sm3tt2aq4si): Ditto.
10686 (aarch64_sm3tt2bq4si): Ditto.
10687 (aarch64_sm3partw1qv4si): Ditto.
10688 (aarch64_sm3partw2qv4si): Ditto.
10689 (aarch64_sm4eqv4si): Ditto.
10690 (aarch64_sm4ekeyqv4si): Ditto.
10691 * config/aarch64/aarch64-simd.md:
10692 (aarch64_sm3ss1qv4si): Ditto.
10693 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10694 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10695 (aarch64_sm4eqv4si): Ditto.
10696 (aarch64_sm4ekeyqv4si): Ditto.
10697 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10698 (sm3part_op): Ditto.
10699 (CRYPTO_SM3TT): Ditto.
10700 (CRYPTO_SM3PART): Ditto.
10701 (UNSPEC_SM3SS1): New unspec.
10702 (UNSPEC_SM3TT1A): Ditto.
10703 (UNSPEC_SM3TT1B): Ditto.
10704 (UNSPEC_SM3TT2A): Ditto.
10705 (UNSPEC_SM3TT2B): Ditto.
10706 (UNSPEC_SM3PARTW1): Ditto.
10707 (UNSPEC_SM3PARTW2): Ditto.
10708 (UNSPEC_SM4E): Ditto.
10709 (UNSPEC_SM4EKEY): Ditto.
10710 * config/aarch64/constraints.md (Ui2): New constraint.
10711 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10712 * config/arm/types.md (crypto_sm3): New type attribute.
10713 (crypto_sm4): Ditto.
10714 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10715 (vsm3tt1aq_u32): Ditto.
10716 (vsm3tt1bq_u32): Ditto.
10717 (vsm3tt2aq_u32): Ditto.
10718 (vsm3tt2bq_u32): Ditto.
10719 (vsm3partw1q_u32): Ditto.
10720 (vsm3partw2q_u32): Ditto.
10721 (vsm4eq_u32): Ditto.
10722 (vsm4ekeyq_u32): Ditto.
10723 (doc/invoke.texi): Document new sm4 option.
10725 2018-01-10 Michael Collison <michael.collison@arm.com>
10727 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10728 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10729 (AARCH64_FL_FOR_ARCH8_4): New.
10730 (AARCH64_FL_V8_4): New flag.
10731 (doc/invoke.texi): Document new armv8.4-a option.
10733 2018-01-10 Michael Collison <michael.collison@arm.com>
10735 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10736 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10737 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10738 * config/aarch64/aarch64-option-extension.def: Add
10739 AARCH64_OPT_EXTENSION of 'sha2'.
10740 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10741 (crypto): Disable sha2 and aes if crypto disabled.
10742 (crypto): Enable aes and sha2 if enabled.
10743 (simd): Disable sha2 and aes if simd disabled.
10744 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10746 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10747 (TARGET_SHA2): New feature flag for sha2.
10748 (TARGET_AES): New feature flag for aes.
10749 * config/aarch64/aarch64-simd.md:
10750 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10751 conditional on TARGET_AES.
10752 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10753 (aarch64_crypto_sha1hsi): Make pattern conditional
10755 (aarch64_crypto_sha1hv4si): Ditto.
10756 (aarch64_be_crypto_sha1hv4si): Ditto.
10757 (aarch64_crypto_sha1su1v4si): Ditto.
10758 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10759 (aarch64_crypto_sha1su0v4si): Ditto.
10760 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10761 (aarch64_crypto_sha256su0v4si): Ditto.
10762 (aarch64_crypto_sha256su1v4si): Ditto.
10763 (doc/invoke.texi): Document new aes and sha2 options.
10765 2018-01-10 Martin Sebor <msebor@redhat.com>
10767 PR tree-optimization/83781
10768 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10771 2018-01-11 Martin Sebor <msebor@gmail.com>
10772 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10774 PR tree-optimization/83501
10775 PR tree-optimization/81703
10777 * tree-ssa-strlen.c (get_string_cst): Rename...
10778 (get_string_len): ...to this. Handle global constants.
10779 (handle_char_store): Adjust.
10781 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10782 Jim Wilson <jimw@sifive.com>
10784 * config/riscv/riscv-protos.h (riscv_output_return): New.
10785 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10786 (riscv_attribute_table, riscv_output_return),
10787 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10788 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10789 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10791 (riscv_expand_prologue): Add early return for naked function.
10792 (riscv_expand_epilogue): Likewise.
10793 (riscv_function_ok_for_sibcall): Return false for naked function.
10794 (riscv_set_current_function): New.
10795 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10796 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10797 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10798 * doc/extend.texi (RISC-V Function Attributes): New.
10800 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10802 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10803 check for 128-bit long double before checking TCmode.
10804 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10805 128-bit long doubles before checking TFmode or TCmode.
10806 (FLOAT128_IBM_P): Likewise.
10808 2018-01-10 Martin Sebor <msebor@redhat.com>
10810 PR tree-optimization/83671
10811 * builtins.c (c_strlen): Unconditionally return zero for the empty
10813 Use -Warray-bounds for warnings.
10814 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10815 for non-constant array indices with COMPONENT_REF, arrays of
10816 arrays, and pointers to arrays.
10817 (gimple_fold_builtin_strlen): Determine and set length range for
10818 non-constant character arrays.
10820 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10822 PR middle-end/81897
10823 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10826 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10828 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10830 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10833 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10834 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10835 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10836 indexed_or_indirect_operand predicate.
10837 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10838 (*vsx_le_perm_load_v8hi): Likewise.
10839 (*vsx_le_perm_load_v16qi): Likewise.
10840 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10841 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10842 (*vsx_le_perm_store_v8hi): Likewise.
10843 (*vsx_le_perm_store_v16qi): Likewise.
10844 (eight unnamed splitters): Likewise.
10846 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10848 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10849 * config/rs6000/emmintrin.h: Likewise.
10850 * config/rs6000/mmintrin.h: Likewise.
10851 * config/rs6000/xmmintrin.h: Likewise.
10853 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10856 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10858 * tree.c (tree_nop_conversion): Return true for location wrapper
10860 (maybe_wrap_with_location): New function.
10861 (selftest::check_strip_nops): New function.
10862 (selftest::test_location_wrappers): New function.
10863 (selftest::tree_c_tests): Call it.
10864 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10865 (maybe_wrap_with_location): New decl.
10866 (EXPR_LOCATION_WRAPPER_P): New macro.
10867 (location_wrapper_p): New inline function.
10868 (tree_strip_any_location_wrapper): New inline function.
10870 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10873 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10874 stack_realign_offset for the largest alignment of stack slot
10876 (ix86_find_max_used_stack_alignment): New function.
10877 (ix86_finalize_stack_frame_flags): Use it. Set
10878 max_used_stack_alignment if we don't realign stack.
10879 * config/i386/i386.h (machine_function): Add
10880 max_used_stack_alignment.
10882 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10884 * config/arm/arm.opt (-mbranch-cost): New option.
10885 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10888 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10891 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10892 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10894 2018-01-10 Richard Biener <rguenther@suse.de>
10897 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10898 early out so it also covers the case where we have a non-NULL
10901 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10903 PR tree-optimization/83753
10904 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10905 for non-strided grouped accesses if the number of elements is 1.
10907 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10910 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10911 * i386.h (TARGET_USE_GATHER): Define.
10912 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10914 2018-01-10 Martin Liska <mliska@suse.cz>
10917 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10918 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10920 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10921 CLEANUP_NO_PARTITIONING is not set.
10923 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10925 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10926 for vectors, as a partial revert of r254296.
10927 * rtl.h (const_vec_p): Delete.
10928 (const_vec_duplicate_p): Don't test for vector CONSTs.
10929 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10930 * expmed.c (make_tree): Likewise.
10933 * common.md (E, F): Use CONSTANT_P instead of checking for
10935 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10936 checking for CONST_VECTOR.
10938 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10940 PR middle-end/83575
10941 * predict.c (force_edge_cold): Handle in more sane way edges
10942 with no prediction.
10944 2018-01-09 Carl Love <cel@us.ibm.com>
10946 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10948 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10949 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10950 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10951 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10952 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10953 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10954 * config/rs6000/rs6000-protos.h: Add extern defition for
10955 rs6000_generate_float2_double_code.
10956 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10958 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10959 (float2_v2df): Add define_expand.
10961 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10964 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10965 op_mode in the force_to_mode call.
10967 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10969 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10970 instead of checking each element individually.
10971 (aarch64_evpc_uzp): Likewise.
10972 (aarch64_evpc_zip): Likewise.
10973 (aarch64_evpc_ext): Likewise.
10974 (aarch64_evpc_rev): Likewise.
10975 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10976 instead of checking each element individually. Return true without
10978 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10979 whether all selected elements come from the same input, instead of
10980 checking each element individually. Remove calls to gen_rtx_REG,
10981 start_sequence and end_sequence and instead assert that no rtl is
10984 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10986 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10987 order of HIGH and CONST checks.
10989 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10991 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10992 if the destination isn't an SSA_NAME.
10994 2018-01-09 Richard Biener <rguenther@suse.de>
10996 PR tree-optimization/83668
10997 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10999 (canonicalize_loop_form): ... here, renamed from ...
11000 (canonicalize_loop_closed_ssa_form): ... this and amended to
11001 swap successor edges for loop exit blocks to make us use
11002 the RPO order we need for initial schedule generation.
11004 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11006 PR tree-optimization/64811
11007 * match.pd: When optimizing comparisons with Inf, avoid
11008 introducing or losing exceptions from comparisons with NaN.
11010 2018-01-09 Martin Liska <mliska@suse.cz>
11013 * asan.c (shadow_mem_size): Add gcc_assert.
11015 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11017 Don't save registers in main().
11020 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11021 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11022 * config/avr/avr.c (avr_set_current_function): Don't error if
11023 naked, OS_task or OS_main are specified at the same time.
11024 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11026 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11028 * common/config/avr/avr-common.c (avr_option_optimization_table):
11029 Switch on -mmain-is-OS_task for optimizing compilations.
11031 2018-01-09 Richard Biener <rguenther@suse.de>
11033 PR tree-optimization/83572
11034 * graphite.c: Include cfganal.h.
11035 (graphite_transform_loops): Connect infinite loops to exit
11036 and remove fake edges at the end.
11038 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11040 * ipa-inline.c (edge_badness): Revert accidental checkin.
11042 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11045 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11046 symbols; not inline clones.
11048 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11051 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11052 hard registers. Formatting fixes.
11054 PR preprocessor/83722
11055 * gcc.c (try_generate_repro): Pass
11056 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11057 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11060 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11061 Kito Cheng <kito.cheng@gmail.com>
11063 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11064 (riscv_leaf_function_p): Delete.
11065 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11067 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11069 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11071 (do_ifelse): New function.
11072 (do_isel): New function.
11073 (do_sub3): New function.
11074 (do_add3): New function.
11075 (do_load_mask_compare): New function.
11076 (do_overlap_load_compare): New function.
11077 (expand_compare_loop): New function.
11078 (expand_block_compare): Call expand_compare_loop() when appropriate.
11079 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11080 option description.
11081 (-mblock-compare-inline-loop-limit): New option.
11083 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11086 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11087 Reverse order of second and third operands in first alternative.
11088 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11089 of first and second elements in UNSPEC_VPERMR vector.
11090 (altivec_expand_vec_perm_le): Likewise.
11092 2018-01-08 Jeff Law <law@redhat.com>
11094 PR rtl-optimizatin/81308
11095 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11096 (process_switch): If group_case_labels makes a change, then set
11098 (pass_convert_switch::execute): If a switch is converted, then
11099 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11101 PR rtl-optimization/81308
11102 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11105 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11107 PR target/83663 - Revert r255946
11108 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11109 generation for cases where splatting a value is not useful.
11110 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11111 across a vec_duplicate and a paradoxical subreg forming a vector
11112 mode to a vec_concat.
11114 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11116 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11117 -march=armv8.3-a variants.
11118 * config/arm/t-multilib: Likewise.
11119 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11121 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11123 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11125 (cceq_ior_compare_complement): Give it a name so I can use it, and
11126 change boolean_or_operator predicate to boolean_operator so it can
11127 be used to generate a crand.
11128 (eqne): New code iterator.
11129 (bd/bd_neg): New code_attrs.
11130 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11131 a single define_insn.
11132 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11133 decrement (bdnzt/bdnzf/bdzt/bdzf).
11134 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11135 with the new names of the branch decrement patterns, and added the
11136 names of the branch decrement conditional patterns.
11138 2018-01-08 Richard Biener <rguenther@suse.de>
11140 PR tree-optimization/83563
11141 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11144 2018-01-08 Richard Biener <rguenther@suse.de>
11146 PR middle-end/83713
11147 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11149 2018-01-08 Richard Biener <rguenther@suse.de>
11151 PR tree-optimization/83685
11152 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11153 references to abnormals.
11155 2018-01-08 Richard Biener <rguenther@suse.de>
11158 * dwarf2out.c (output_indirect_strings): Handle empty
11159 skeleton_debug_str_hash.
11160 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11162 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11164 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11165 (emit_store_direct): Likewise.
11166 (arc_trampoline_adjust_address): Likewise.
11167 (arc_asm_trampoline_template): New function.
11168 (arc_initialize_trampoline): Use asm_trampoline_template.
11169 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11170 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11171 * config/arc/arc.md (flush_icache): Delete pattern.
11173 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11175 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11176 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11179 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11182 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11183 by not USED_FOR_TARGET.
11184 (make_pass_resolve_sw_modes): Likewise.
11186 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11188 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11191 2018-01-08 Richard Biener <rguenther@suse.de>
11193 PR middle-end/83580
11194 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11196 2018-01-08 Richard Biener <rguenther@suse.de>
11198 PR middle-end/83517
11199 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11201 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11203 PR middle-end/81897
11204 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11205 basic blocks with a small number of successors.
11206 (convert_control_dep_chain_into_preds): Improve handling of
11208 (dump_predicates): Split apart into...
11209 (dump_pred_chain): ...here...
11210 (dump_pred_info): ...and here.
11211 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11212 (can_chain_union_be_invalidated_p): Improve check for invalidation
11214 (uninit_uses_cannot_happen): Avoid unnecessary if
11215 convert_control_dep_chain_into_preds yielded nothing.
11217 2018-01-06 Martin Sebor <msebor@redhat.com>
11219 PR tree-optimization/83640
11220 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11221 subtracting negative offset from size.
11222 (builtin_access::overlap): Adjust offset bounds of the access to fall
11223 within the size of the object if possible.
11225 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11227 PR rtl-optimization/83699
11228 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11229 extract_bit_field_as_subreg to cases in which the extracted
11230 value is also a vector.
11232 * lra-constraints.c (process_alt_operands): Test for the equivalence
11233 substitutions when detecting a possible reload cycle.
11235 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11238 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11239 by default if flag_selective_schedling{,2}. Formatting fixes.
11241 PR rtl-optimization/83682
11242 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11243 if it has non-VECTOR_MODE element mode.
11244 (vec_duplicate_p): Likewise.
11246 PR middle-end/83694
11247 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11248 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11250 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11253 * config/i386/i386-builtin.def
11254 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11255 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11256 Require also OPTION_MASK_ISA_AVX512F in addition to
11257 OPTION_MASK_ISA_GFNI.
11258 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11259 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11260 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11261 to OPTION_MASK_ISA_GFNI.
11262 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11263 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11264 OPTION_MASK_ISA_AVX512BW.
11265 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11266 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11267 addition to OPTION_MASK_ISA_GFNI.
11268 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11269 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11270 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11271 to OPTION_MASK_ISA_GFNI.
11272 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11273 a requirement for all ISAs rather than any of them with a few
11275 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11277 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11278 bitmasks to be enabled with 3 exceptions, instead of requiring any
11279 enabled ISA with lots of exceptions.
11280 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11281 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11282 Change avx512bw in isa attribute to avx512f.
11283 * config/i386/sgxintrin.h: Add license boilerplate.
11284 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11285 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11286 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11287 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11289 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11290 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11291 temporarily sse2 rather than sse if not enabled already.
11294 * config/i386/sse.md (VI248_VLBW): Rename to ...
11295 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11296 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11297 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11298 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11299 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11300 mode iterator instead of VI248_VLBW.
11302 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11304 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11305 (record_modified): Skip clobbers; add debug output.
11306 (param_change_prob): Use sreal frequencies.
11308 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11310 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11311 punt for user-aligned variables.
11313 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11315 * tree-chrec.c (chrec_contains_symbols): Return true for
11318 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11321 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11322 of (x|y) == x for BICS pattern.
11324 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11326 PR tree-optimization/83605
11327 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11328 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11331 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11333 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11334 * config/epiphany/rtems.h: New file.
11336 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11337 Uros Bizjak <ubizjak@gmail.com>
11340 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11341 QIreg_operand instead of register_operand predicate.
11342 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11343 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11344 comments instead of -fmitigate[-_]rop.
11346 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11349 * cgraphunit.c (symbol_table::compile): Switch to text_section
11350 before calling assembly_start debug hook.
11351 * run-rtl-passes.c (run_rtl_passes): Likewise.
11354 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11356 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11357 range_int_cst_p rather than !symbolic_range_p before calling
11358 extract_range_from_multiplicative_op_1.
11360 2018-01-04 Jeff Law <law@redhat.com>
11362 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11363 redundant test in assertion.
11365 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11367 * doc/rtl.texi: Document machine_mode wrapper classes.
11369 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11371 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11372 using tree_to_uhwi.
11374 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11376 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11377 the VEC_PERM_EXPR fold to fail.
11379 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11382 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11383 to switched_sections.
11385 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11388 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11389 test for d.testing.
11391 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11394 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11395 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11397 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11400 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11401 is BLKmode and bitpos not zero or mode change is needed.
11403 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11406 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11409 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11412 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11413 instead of MULT rtx. Update all corresponding splitters.
11414 (*saddl_se): Ditto.
11415 (*ssub<modesuffix>): Ditto.
11416 (*ssubl_se): Ditto.
11417 (*cmp_sadd_di): Update split patterns.
11418 (*cmp_sadd_si): Ditto.
11419 (*cmp_sadd_sidi): Ditto.
11420 (*cmp_ssub_di): Ditto.
11421 (*cmp_ssub_si): Ditto.
11422 (*cmp_ssub_sidi): Ditto.
11423 * config/alpha/predicates.md (const23_operand): New predicate.
11424 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11425 Look for ASHIFT, not MULT inner operand.
11426 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11428 2018-01-04 Martin Liska <mliska@suse.cz>
11430 PR gcov-profile/83669
11431 * gcov.c (output_intermediate_file): Add version to intermediate
11433 * doc/gcov.texi: Document new field 'version' in intermediate
11434 file format. Fix location of '-k' option of gcov command.
11436 2018-01-04 Martin Liska <mliska@suse.cz>
11439 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11441 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11443 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11445 2018-01-03 Martin Sebor <msebor@redhat.com>
11447 PR tree-optimization/83655
11448 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11449 checking calls with invalid arguments.
11451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11453 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11454 (vectorizable_mask_load_store): Delete.
11455 (vectorizable_call): Return false for masked loads and stores.
11456 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11457 instead of gimple_assign_rhs1.
11458 (vectorizable_load): Handle IFN_MASK_LOAD.
11459 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11463 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11465 (vectorizable_mask_load_store): ...here.
11466 (vectorizable_load): ...and here.
11468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11470 * tree-vect-stmts.c (vect_build_all_ones_mask)
11471 (vect_build_zero_merge_argument): New functions, split out from...
11472 (vectorizable_load): ...here.
11474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11476 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11478 (vectorizable_mask_load_store): ...here.
11479 (vectorizable_store): ...and here.
11481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11483 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11485 (vectorizable_mask_load_store): ...here.
11487 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11489 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11490 (vect_model_store_cost): Take a vec_load_store_type instead of a
11492 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11493 (vect_model_store_cost): Take a vec_load_store_type instead of a
11495 (vectorizable_mask_load_store): Update accordingly.
11496 (vectorizable_store): Likewise.
11497 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11499 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11501 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11502 IFN_MASK_LOAD calls here rather than...
11503 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11506 Alan Hayward <alan.hayward@arm.com>
11507 David Sherwood <david.sherwood@arm.com>
11509 * expmed.c (extract_bit_field_1): For vector extracts,
11510 fall back to extract_bit_field_as_subreg if vec_extract
11513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11514 Alan Hayward <alan.hayward@arm.com>
11515 David Sherwood <david.sherwood@arm.com>
11517 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11518 they are variable or constant sized.
11519 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11520 slots for constant-sized data.
11522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11523 Alan Hayward <alan.hayward@arm.com>
11524 David Sherwood <david.sherwood@arm.com>
11526 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11527 handling COND_EXPRs with boolean comparisons, try to find a better
11528 basis for the mask type than the boolean itself.
11530 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11532 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11533 is calculated and how it can be overridden.
11534 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11535 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11537 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11541 Alan Hayward <alan.hayward@arm.com>
11542 David Sherwood <david.sherwood@arm.com>
11544 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11545 Remove the mode argument.
11546 (aarch64_simd_valid_immediate): Remove the mode and inverse
11548 * config/aarch64/iterators.md (bitsize): New iterator.
11549 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11550 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11551 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11552 aarch64_simd_valid_immediate.
11553 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11554 (aarch64_reg_or_bic_imm): Likewise.
11555 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11556 with an insn_type enum and msl with a modifier_type enum.
11557 Replace element_width with a scalar_mode. Change the shift
11558 to unsigned int. Add constructors for scalar_float_mode and
11559 scalar_int_mode elements.
11560 (aarch64_vect_float_const_representable_p): Delete.
11561 (aarch64_can_const_movi_rtx_p)
11562 (aarch64_simd_scalar_immediate_valid_for_move)
11563 (aarch64_simd_make_constant): Update call to
11564 aarch64_simd_valid_immediate.
11565 (aarch64_advsimd_valid_immediate_hs): New function.
11566 (aarch64_advsimd_valid_immediate): Likewise.
11567 (aarch64_simd_valid_immediate): Remove mode and inverse
11568 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11569 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11570 and aarch64_float_const_representable_p on the result.
11571 (aarch64_output_simd_mov_immediate): Remove mode argument.
11572 Update call to aarch64_simd_valid_immediate and use of
11573 simd_immediate_info.
11574 (aarch64_output_scalar_simd_mov_immediate): Update call
11577 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11578 Alan Hayward <alan.hayward@arm.com>
11579 David Sherwood <david.sherwood@arm.com>
11581 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11582 (mode_nunits): Likewise CONST_MODE_NUNITS.
11583 * machmode.def (ADJUST_NUNITS): Document.
11584 * genmodes.c (mode_data::need_nunits_adj): New field.
11585 (blank_mode): Update accordingly.
11586 (adj_nunits): New variable.
11587 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11589 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11590 listed in adj_nunits.
11591 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11592 listed in adj_nunits. Don't emit case statements for such modes.
11593 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11594 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11595 nothing if adj_nunits is nonnull.
11596 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11597 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11598 (emit_mode_fbit): Update use of print_maybe_const_decl.
11599 (emit_move_size): Likewise. Treat the array as non-const
11601 (emit_mode_adjustments): Handle adj_nunits.
11603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11605 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11606 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11607 (VECTOR_MODES): Use it.
11608 (make_vector_modes): Take the prefix as an argument.
11610 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11611 Alan Hayward <alan.hayward@arm.com>
11612 David Sherwood <david.sherwood@arm.com>
11614 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11615 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11616 for MODE_VECTOR_BOOL.
11617 * machmode.def (VECTOR_BOOL_MODE): Document.
11618 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11619 (make_vector_bool_mode): New function.
11620 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11622 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11623 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11625 * stor-layout.c (int_mode_for_mode): Likewise.
11626 * tree.c (build_vector_type_for_mode): Likewise.
11627 * varasm.c (output_constant_pool_2): Likewise.
11628 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11629 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11630 for MODE_VECTOR_BOOL.
11631 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11632 of mode class checks.
11633 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11634 instead of a list of mode class checks.
11635 (expand_vector_scalar_condition): Likewise.
11636 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11639 Alan Hayward <alan.hayward@arm.com>
11640 David Sherwood <david.sherwood@arm.com>
11642 * machmode.h (mode_size): Change from unsigned short to
11644 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11645 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11646 or if measurement_type is not polynomial.
11647 (fixed_size_mode::includes_p): Check for constant-sized modes.
11648 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11649 return a poly_uint16 rather than an unsigned short.
11650 (emit_mode_size): Change the type of mode_size from unsigned short
11651 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11652 (emit_mode_adjustments): Cope with polynomial vector sizes.
11653 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11655 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11657 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11658 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11659 * caller-save.c (setup_save_areas): Likewise.
11660 (replace_reg_with_saved_mem): Likewise.
11661 * calls.c (emit_library_call_value_1): Likewise.
11662 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11663 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11664 (gen_lowpart_for_combine): Likewise.
11665 * convert.c (convert_to_integer_1): Likewise.
11666 * cse.c (equiv_constant, cse_insn): Likewise.
11667 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11668 (cselib_subst_to_values): Likewise.
11669 * dce.c (word_dce_process_block): Likewise.
11670 * df-problems.c (df_word_lr_mark_ref): Likewise.
11671 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11672 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11673 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11674 (rtl_for_decl_location): Likewise.
11675 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11676 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11677 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11678 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11679 (expand_expr_real_1): Likewise.
11680 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11681 (pad_below): Likewise.
11682 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11683 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11684 * ira.c (get_subreg_tracking_sizes): Likewise.
11685 * ira-build.c (ira_create_allocno_objects): Likewise.
11686 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11687 (ira_sort_regnos_for_alter_reg): Likewise.
11688 * ira-costs.c (record_operand_costs): Likewise.
11689 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11690 (resolve_simple_move): Likewise.
11691 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11692 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11693 (lra_constraints): Likewise.
11694 (CONST_POOL_OK_P): Reject variable-sized modes.
11695 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11696 (add_pseudo_to_slot, lra_spill): Likewise.
11697 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11698 * optabs-query.c (get_best_extraction_insn): Likewise.
11699 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11700 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11701 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11702 * recog.c (offsettable_address_addr_space_p): Likewise.
11703 * regcprop.c (maybe_mode_change): Likewise.
11704 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11705 * regrename.c (build_def_use): Likewise.
11706 * regstat.c (dump_reg_info): Likewise.
11707 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11708 (find_reloads, find_reloads_subreg_address): Likewise.
11709 * reload1.c (eliminate_regs_1): Likewise.
11710 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11711 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11712 (simplify_binary_operation_1, simplify_subreg): Likewise.
11713 * targhooks.c (default_function_arg_padding): Likewise.
11714 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11715 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11716 (verify_gimple_assign_ternary): Likewise.
11717 * tree-inline.c (estimate_move_cost): Likewise.
11718 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11719 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11720 (get_address_cost_ainc): Likewise.
11721 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11722 (vect_supportable_dr_alignment): Likewise.
11723 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11724 (vectorizable_reduction): Likewise.
11725 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11726 (vectorizable_operation, vectorizable_load): Likewise.
11727 * tree.c (build_same_sized_truth_vector_type): Likewise.
11728 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11729 * var-tracking.c (emit_note_insn_var_location): Likewise.
11730 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11731 (ADDR_VEC_ALIGN): Likewise.
11733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11734 Alan Hayward <alan.hayward@arm.com>
11735 David Sherwood <david.sherwood@arm.com>
11737 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11739 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11740 or if measurement_type is polynomial.
11741 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11742 * combine.c (make_extraction): Likewise.
11743 * dse.c (find_shift_sequence): Likewise.
11744 * dwarf2out.c (mem_loc_descriptor): Likewise.
11745 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11746 (extract_bit_field, extract_low_bits): Likewise.
11747 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11748 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11749 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11750 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11751 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11752 * reload.c (find_reloads): Likewise.
11753 * reload1.c (alter_reg): Likewise.
11754 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11755 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11756 * tree-if-conv.c (predicate_mem_writes): Likewise.
11757 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11758 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11759 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11760 * valtrack.c (dead_debug_insert_temp): Likewise.
11761 * varasm.c (mergeable_constant_section): Likewise.
11762 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11765 Alan Hayward <alan.hayward@arm.com>
11766 David Sherwood <david.sherwood@arm.com>
11768 * expr.c (expand_assignment): Cope with polynomial mode sizes
11769 when assigning to a CONCAT.
11771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11772 Alan Hayward <alan.hayward@arm.com>
11773 David Sherwood <david.sherwood@arm.com>
11775 * machmode.h (mode_precision): Change from unsigned short to
11777 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11779 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11780 or if measurement_type is not polynomial.
11781 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11782 in which the mode is already known to be a scalar_int_mode.
11783 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11784 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11786 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11787 for GET_MODE_PRECISION.
11788 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11789 for GET_MODE_PRECISION.
11790 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11792 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11793 (expand_field_assignment, make_extraction): Likewise.
11794 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11795 (get_last_value): Likewise.
11796 * convert.c (convert_to_integer_1): Likewise.
11797 * cse.c (cse_insn): Likewise.
11798 * expr.c (expand_expr_real_1): Likewise.
11799 * lra-constraints.c (simplify_operand_subreg): Likewise.
11800 * optabs-query.c (can_atomic_load_p): Likewise.
11801 * optabs.c (expand_atomic_load): Likewise.
11802 (expand_atomic_store): Likewise.
11803 * ree.c (combine_reaching_defs): Likewise.
11804 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11805 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11806 * tree.h (type_has_mode_precision_p): Likewise.
11807 * ubsan.c (instrument_si_overflow): Likewise.
11809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11810 Alan Hayward <alan.hayward@arm.com>
11811 David Sherwood <david.sherwood@arm.com>
11813 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11814 polynomial numbers of units.
11815 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11816 (valid_vector_subparts_p): New function.
11817 (build_vector_type): Remove temporary shim and take the number
11818 of units as a poly_uint64 rather than an int.
11819 (build_opaque_vector_type): Take the number of units as a
11820 poly_uint64 rather than an int.
11821 * tree.c (build_vector_from_ctor): Handle polynomial
11822 TYPE_VECTOR_SUBPARTS.
11823 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11824 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11825 (build_vector_from_val): If the number of units is variable,
11826 use build_vec_duplicate_cst for constant operands and
11827 VEC_DUPLICATE_EXPR otherwise.
11828 (make_vector_type): Remove temporary is_constant ().
11829 (build_vector_type, build_opaque_vector_type): Take the number of
11830 units as a poly_uint64 rather than an int.
11831 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11833 * cfgexpand.c (expand_debug_expr): Likewise.
11834 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11835 (store_constructor, expand_expr_real_1): Likewise.
11836 (const_scalar_mask_from_tree): Likewise.
11837 * fold-const-call.c (fold_const_reduction): Likewise.
11838 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11839 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11840 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11841 (fold_relational_const): Likewise.
11842 (native_interpret_vector): Likewise. Change the size from an
11843 int to an unsigned int.
11844 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11845 TYPE_VECTOR_SUBPARTS.
11846 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11847 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11848 duplicating a non-constant operand into a variable-length vector.
11849 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11850 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11851 * ipa-icf.c (sem_variable::equals): Likewise.
11852 * match.pd: Likewise.
11853 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11854 * print-tree.c (print_node): Likewise.
11855 * stor-layout.c (layout_type): Likewise.
11856 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11857 * tree-cfg.c (verify_gimple_comparison): Likewise.
11858 (verify_gimple_assign_binary): Likewise.
11859 (verify_gimple_assign_ternary): Likewise.
11860 (verify_gimple_assign_single): Likewise.
11861 * tree-pretty-print.c (dump_generic_node): Likewise.
11862 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11863 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11864 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11865 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11866 (vect_shift_permute_load_chain): Likewise.
11867 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11868 (expand_vector_condition, optimize_vector_constructor): Likewise.
11869 (lower_vec_perm, get_compute_type): Likewise.
11870 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11871 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11872 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11873 (vect_recog_mask_conversion_pattern): Likewise.
11874 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11875 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11876 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11877 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11878 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11879 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11880 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11881 (supportable_widening_operation): Likewise.
11882 (supportable_narrowing_operation): Likewise.
11883 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11885 * varasm.c (output_constant): Likewise.
11887 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11888 Alan Hayward <alan.hayward@arm.com>
11889 David Sherwood <david.sherwood@arm.com>
11891 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11892 so that both the length == 3 and length != 3 cases set up their
11893 own permute vectors. Add comments explaining why we know the
11894 number of elements is constant.
11895 (vect_permute_load_chain): Likewise.
11897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11898 Alan Hayward <alan.hayward@arm.com>
11899 David Sherwood <david.sherwood@arm.com>
11901 * machmode.h (mode_nunits): Change from unsigned char to
11903 (ONLY_FIXED_SIZE_MODES): New macro.
11904 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11905 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11906 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11908 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11909 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11910 or if measurement_type is not polynomial.
11911 * genmodes.c (ZERO_COEFFS): New macro.
11912 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11914 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11915 Use ZERO_COEFFS when emitting initializers.
11916 * data-streamer.h (bp_pack_poly_value): New function.
11917 (bp_unpack_poly_value): Likewise.
11918 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11919 for GET_MODE_NUNITS.
11920 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11921 for GET_MODE_NUNITS.
11922 * tree.c (make_vector_type): Remove temporary shim and make
11923 the real function take the number of units as a poly_uint64
11924 rather than an int.
11925 (build_vector_type_for_mode): Handle polynomial nunits.
11926 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11927 * emit-rtl.c (const_vec_series_p_1): Likewise.
11928 (gen_rtx_CONST_VECTOR): Likewise.
11929 * fold-const.c (test_vec_duplicate_folding): Likewise.
11930 * genrecog.c (validate_pattern): Likewise.
11931 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11932 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11933 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11934 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11935 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11936 * rtlanal.c (subreg_get_info): Likewise.
11937 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11938 (vect_grouped_load_supported): Likewise.
11939 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11940 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11941 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11942 (simplify_const_unary_operation, simplify_binary_operation_1)
11943 (simplify_const_binary_operation, simplify_ternary_operation)
11944 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11945 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11946 instead of CONST_VECTOR_NUNITS.
11947 * varasm.c (output_constant_pool_2): Likewise.
11948 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11949 explicit-encoded elements in the XVEC for variable-length vectors.
11951 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11953 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11956 Alan Hayward <alan.hayward@arm.com>
11957 David Sherwood <david.sherwood@arm.com>
11959 * coretypes.h (fixed_size_mode): Declare.
11960 (fixed_size_mode_pod): New typedef.
11961 * builtins.h (target_builtins::x_apply_args_mode)
11962 (target_builtins::x_apply_result_mode): Change type to
11963 fixed_size_mode_pod.
11964 * builtins.c (apply_args_size, apply_result_size, result_vector)
11965 (expand_builtin_apply_args_1, expand_builtin_apply)
11966 (expand_builtin_return): Update accordingly.
11968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11970 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11971 * cselib.c (cselib_hash_rtx): Likewise.
11972 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11973 CONST_VECTOR encoding.
11975 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11976 Jeff Law <law@redhat.com>
11979 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11980 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11981 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11982 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11985 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11986 explicitly probe *sp in a noreturn function if there were any callee
11987 register saves or frame pointer is needed.
11989 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11992 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11993 BLKmode for ternary, binary or unary expressions.
11996 * var-tracking.c (delete_vta_debug_insn): New inline function.
11997 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11998 insns from get_insns () to NULL instead of each bb separately.
11999 Use delete_vta_debug_insn. No longer static.
12000 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12001 delete_vta_debug_insns callers.
12002 * rtl.h (delete_vta_debug_insns): Declare.
12003 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12004 instead of variable_tracking_main.
12006 2018-01-03 Martin Sebor <msebor@redhat.com>
12008 PR tree-optimization/83603
12009 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12010 arguments past the endof the argument list in functions declared
12011 without a prototype.
12012 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12013 Avoid checking when arguments are null.
12015 2018-01-03 Martin Sebor <msebor@redhat.com>
12018 * doc/extend.texi (attribute const): Fix a typo.
12019 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12020 issuing -Wsuggest-attribute for void functions.
12022 2018-01-03 Martin Sebor <msebor@redhat.com>
12024 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12025 offset_int::from instead of wide_int::to_shwi.
12026 (maybe_diag_overlap): Remove assertion.
12027 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12028 * gimple-ssa-sprintf.c (format_directive): Same.
12029 (parse_directive): Same.
12030 (sprintf_dom_walker::compute_format_length): Same.
12031 (try_substitute_return_value): Same.
12033 2018-01-03 Jeff Law <law@redhat.com>
12035 PR middle-end/83654
12036 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12037 non-constant residual for zero at runtime and avoid probing in
12038 that case. Reorganize code for trailing problem to mirror handling
12041 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12043 PR tree-optimization/83501
12044 * tree-ssa-strlen.c (get_string_cst): New.
12045 (handle_char_store): Call get_string_cst.
12047 2018-01-03 Martin Liska <mliska@suse.cz>
12049 PR tree-optimization/83593
12050 * tree-ssa-strlen.c: Include tree-cfg.h.
12051 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12052 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12053 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12055 (strlen_dom_walker::before_dom_children): Call
12056 gimple_purge_dead_eh_edges. Dump tranformation with details
12058 (strlen_dom_walker::before_dom_children): Update call by adding
12059 new argument cleanup_eh.
12060 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12062 2018-01-03 Martin Liska <mliska@suse.cz>
12065 * cif-code.def (VARIADIC_THUNK): New enum value.
12066 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12069 2018-01-03 Jan Beulich <jbeulich@suse.com>
12071 * sse.md (mov<mode>_internal): Tighten condition for when to use
12072 vmovdqu<ssescalarsize> for TI and OI modes.
12074 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12076 Update copyright years.
12078 2018-01-03 Martin Liska <mliska@suse.cz>
12081 * ipa-visibility.c (function_and_variable_visibility): Skip
12082 functions with noipa attribure.
12084 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12086 * gcc.c (process_command): Update copyright notice dates.
12087 * gcov-dump.c (print_version): Ditto.
12088 * gcov.c (print_version): Ditto.
12089 * gcov-tool.c (print_version): Ditto.
12090 * gengtype.c (create_file): Ditto.
12091 * doc/cpp.texi: Bump @copying's copyright year.
12092 * doc/cppinternals.texi: Ditto.
12093 * doc/gcc.texi: Ditto.
12094 * doc/gccint.texi: Ditto.
12095 * doc/gcov.texi: Ditto.
12096 * doc/install.texi: Ditto.
12097 * doc/invoke.texi: Ditto.
12099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12101 * vector-builder.h (vector_builder::m_full_nelts): Change from
12102 unsigned int to poly_uint64.
12103 (vector_builder::full_nelts): Update prototype accordingly.
12104 (vector_builder::new_vector): Likewise.
12105 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12106 (vector_builder::operator ==): Likewise.
12107 (vector_builder::finalize): Likewise.
12108 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12109 Take the number of elements as a poly_uint64 rather than an
12111 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12112 from unsigned int to poly_uint64.
12113 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12114 (vec_perm_indices::new_vector): Likewise.
12115 (vec_perm_indices::length): Likewise.
12116 (vec_perm_indices::nelts_per_input): Likewise.
12117 (vec_perm_indices::input_nelts): Likewise.
12118 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12119 number of elements per input as a poly_uint64 rather than an
12120 unsigned int. Use the original encoding for variable-length
12121 vectors, rather than clamping each individual element.
12122 For the second and subsequent elements in each pattern,
12123 clamp the step and base before clamping their sum.
12124 (vec_perm_indices::series_p): Handle polynomial element counts.
12125 (vec_perm_indices::all_in_range_p): Likewise.
12126 (vec_perm_indices_to_tree): Likewise.
12127 (vec_perm_indices_to_rtx): Likewise.
12128 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12129 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12130 (tree_vector_builder::new_binary_operation): Handle polynomial
12131 element counts. Return false if we need to know the number
12132 of elements at compile time.
12133 * fold-const.c (fold_vec_perm): Punt if the number of elements
12134 isn't known at compile time.
12136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12138 * vec-perm-indices.h (vec_perm_builder): Change element type
12139 from HOST_WIDE_INT to poly_int64.
12140 (vec_perm_indices::element_type): Update accordingly.
12141 (vec_perm_indices::clamp): Handle polynomial element_types.
12142 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12143 (vec_perm_indices::all_in_range_p): Likewise.
12144 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12146 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12147 polynomial vec_perm_indices element types.
12148 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12149 * fold-const.c (fold_vec_perm): Likewise.
12150 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12151 * tree-vect-generic.c (lower_vec_perm): Likewise.
12152 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12153 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12154 element type to HOST_WIDE_INT.
12156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12157 Alan Hayward <alan.hayward@arm.com>
12158 David Sherwood <david.sherwood@arm.com>
12160 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12161 rather than an int. Use plus_constant.
12162 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12163 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12166 Alan Hayward <alan.hayward@arm.com>
12167 David Sherwood <david.sherwood@arm.com>
12169 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12170 a HOST_WIDE_INT to a poly_int64.
12172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12173 Alan Hayward <alan.hayward@arm.com>
12174 David Sherwood <david.sherwood@arm.com>
12176 * calls.c (load_register_parameters): Cope with polynomial
12177 mode sizes. Require a constant size for BLKmode parameters
12178 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12179 forces a parameter to be padded at the lsb end in order to
12180 fill a complete number of words, require the parameter size
12181 to be ordered wrt UNITS_PER_WORD.
12183 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12184 Alan Hayward <alan.hayward@arm.com>
12185 David Sherwood <david.sherwood@arm.com>
12187 * reload1.c (spill_stack_slot_width): Change element type
12188 from unsigned int to poly_uint64_pod.
12189 (alter_reg): Treat mode sizes as polynomial.
12191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12192 Alan Hayward <alan.hayward@arm.com>
12193 David Sherwood <david.sherwood@arm.com>
12195 * reload.c (complex_word_subreg_p): New function.
12196 (reload_inner_reg_of_subreg, push_reload): Use it.
12198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12199 Alan Hayward <alan.hayward@arm.com>
12200 David Sherwood <david.sherwood@arm.com>
12202 * lra-constraints.c (process_alt_operands): Reject matched
12203 operands whose sizes aren't ordered.
12204 (match_reload): Refer to this check here.
12206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12207 Alan Hayward <alan.hayward@arm.com>
12208 David Sherwood <david.sherwood@arm.com>
12210 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12211 that the mode size is in the set {1, 2, 4, 8, 16}.
12213 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12214 Alan Hayward <alan.hayward@arm.com>
12215 David Sherwood <david.sherwood@arm.com>
12217 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12218 Use plus_constant instead of gen_rtx_PLUS.
12220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12221 Alan Hayward <alan.hayward@arm.com>
12222 David Sherwood <david.sherwood@arm.com>
12224 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12225 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12226 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12227 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12228 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12229 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12230 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12231 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12232 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12233 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12235 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12236 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12237 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12238 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12239 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12240 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12241 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12242 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12243 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12244 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12246 * expr.c (emit_move_resolve_push): Treat the input and result
12247 of PUSH_ROUNDING as a poly_int64.
12248 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12249 (emit_push_insn): Likewise.
12250 * lra-eliminations.c (mark_not_eliminable): Likewise.
12251 * recog.c (push_operand): Likewise.
12252 * reload1.c (elimination_effects): Likewise.
12253 * rtlanal.c (nonzero_bits1): Likewise.
12254 * calls.c (store_one_arg): Likewise. Require the padding to be
12255 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12262 Use plus_constant instead of gen_rtx_PLUS.
12264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12265 Alan Hayward <alan.hayward@arm.com>
12266 David Sherwood <david.sherwood@arm.com>
12268 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12269 rather than an int.
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 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12276 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12277 via stack temporaries. Treat the mode size as polynomial too.
12279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12280 Alan Hayward <alan.hayward@arm.com>
12281 David Sherwood <david.sherwood@arm.com>
12283 * expr.c (expand_expr_real_2): When handling conversions involving
12284 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12285 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12286 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12295 Alan Hayward <alan.hayward@arm.com>
12296 David Sherwood <david.sherwood@arm.com>
12298 * combine.c (can_change_dest_mode): Handle polynomial
12299 REGMODE_NATURAL_SIZE.
12300 * expmed.c (store_bit_field_1): Likewise.
12301 * expr.c (store_constructor): Likewise.
12302 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12303 and polynomial REGMODE_NATURAL_SIZE.
12304 (gen_lowpart_common): Likewise.
12305 * reginfo.c (record_subregs_of_mode): Likewise.
12306 * rtlanal.c (read_modify_subreg_p): Likewise.
12308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12309 Alan Hayward <alan.hayward@arm.com>
12310 David Sherwood <david.sherwood@arm.com>
12312 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12313 numbers of elements.
12315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12316 Alan Hayward <alan.hayward@arm.com>
12317 David Sherwood <david.sherwood@arm.com>
12319 * match.pd: Cope with polynomial numbers of vector elements.
12321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12322 Alan Hayward <alan.hayward@arm.com>
12323 David Sherwood <david.sherwood@arm.com>
12325 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12326 in a POINTER_PLUS_EXPR.
12328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12329 Alan Hayward <alan.hayward@arm.com>
12330 David Sherwood <david.sherwood@arm.com>
12332 * omp-simd-clone.c (simd_clone_subparts): New function.
12333 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12334 (ipa_simd_modify_function_body): Likewise.
12336 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12337 Alan Hayward <alan.hayward@arm.com>
12338 David Sherwood <david.sherwood@arm.com>
12340 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12341 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12342 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12343 (expand_vector_condition, vector_element): Likewise.
12344 (subparts_gt): New function.
12345 (get_compute_type): Use subparts_gt.
12346 (count_type_subparts): Delete.
12347 (expand_vector_operations_1): Use subparts_gt instead of
12348 count_type_subparts.
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-vect-data-refs.c (vect_no_alias_p): Replace with...
12355 (vect_compile_time_alias): ...this new function. Do the calculation
12356 on poly_ints rather than trees.
12357 (vect_prune_runtime_alias_test_list): Update call accordingly.
12359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12360 Alan Hayward <alan.hayward@arm.com>
12361 David Sherwood <david.sherwood@arm.com>
12363 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12365 (vect_schedule_slp_instance): Likewise.
12367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12368 Alan Hayward <alan.hayward@arm.com>
12369 David Sherwood <david.sherwood@arm.com>
12371 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12372 constant and extern definitions for variable-length vectors.
12373 (vect_get_constant_vectors): Note that the number of units
12374 is known to be constant.
12376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12377 Alan Hayward <alan.hayward@arm.com>
12378 David Sherwood <david.sherwood@arm.com>
12380 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12381 of units as polynomial. Choose between WIDE and NARROW based
12384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12385 Alan Hayward <alan.hayward@arm.com>
12386 David Sherwood <david.sherwood@arm.com>
12388 * tree-vect-stmts.c (simd_clone_subparts): New function.
12389 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12392 Alan Hayward <alan.hayward@arm.com>
12393 David Sherwood <david.sherwood@arm.com>
12395 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12396 vectors as polynomial. Use build_index_vector for
12397 IFN_GOMP_SIMD_LANE.
12399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12400 Alan Hayward <alan.hayward@arm.com>
12401 David Sherwood <david.sherwood@arm.com>
12403 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12404 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12405 for variable-length vectors.
12406 (vectorizable_mask_load_store): Treat the number of units as
12407 polynomial, asserting that it is constant if the condition has
12408 already been enforced.
12409 (vectorizable_store, vectorizable_load): Likewise.
12411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12412 Alan Hayward <alan.hayward@arm.com>
12413 David Sherwood <david.sherwood@arm.com>
12415 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12416 of units as polynomial. Punt if we can't tell at compile time
12417 which vector contains the final result.
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 * tree-vect-loop.c (vectorizable_induction): Treat the number
12424 of units as polynomial. Punt on SLP inductions. Use an integer
12425 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12426 cast of such a series for variable-length floating-point
12429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12430 Alan Hayward <alan.hayward@arm.com>
12431 David Sherwood <david.sherwood@arm.com>
12433 * tree.h (build_index_vector): Declare.
12434 * tree.c (build_index_vector): New function.
12435 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12436 of units as polynomial, forcibly converting it to a constant if
12437 vectorizable_reduction has already enforced the condition.
12438 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12439 to create a {1,2,3,...} vector.
12440 (vectorizable_reduction): Treat the number of units as polynomial.
12441 Choose vectype_in based on the largest scalar element size rather
12442 than the smallest number of units. Enforce the restrictions
12445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12446 Alan Hayward <alan.hayward@arm.com>
12447 David Sherwood <david.sherwood@arm.com>
12449 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12450 number of units as polynomial.
12452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12453 Alan Hayward <alan.hayward@arm.com>
12454 David Sherwood <david.sherwood@arm.com>
12456 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12457 * target.def (autovectorize_vector_sizes): Return the vector sizes
12458 by pointer, using vector_sizes rather than a bitmask.
12459 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12460 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12461 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12463 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12464 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12465 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12466 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12467 * omp-general.c (omp_max_vf): Likewise.
12468 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12469 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12470 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12471 * tree-vect-slp.c (vect_slp_bb): Likewise.
12472 * doc/tm.texi: Regenerate.
12473 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12475 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12476 the vector size as a poly_uint64 rather than an unsigned int.
12477 (current_vector_size): Change from an unsigned int to a poly_uint64.
12478 (get_vectype_for_scalar_type): Update accordingly.
12479 * tree.h (build_truth_vector_type): Take the size and number of
12480 units as a poly_uint64 rather than an unsigned int.
12481 (build_vector_type): Add a temporary overload that takes
12482 the number of units as a poly_uint64 rather than an unsigned int.
12483 * tree.c (make_vector_type): Likewise.
12484 (build_truth_vector_type): Take the number of units as a poly_uint64
12485 rather than an unsigned int.
12487 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12488 Alan Hayward <alan.hayward@arm.com>
12489 David Sherwood <david.sherwood@arm.com>
12491 * target.def (get_mask_mode): Take the number of units and length
12492 as poly_uint64s rather than unsigned ints.
12493 * targhooks.h (default_get_mask_mode): Update accordingly.
12494 * targhooks.c (default_get_mask_mode): Likewise.
12495 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12496 * doc/tm.texi: Regenerate.
12498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12499 Alan Hayward <alan.hayward@arm.com>
12500 David Sherwood <david.sherwood@arm.com>
12502 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12503 * omp-general.c (omp_max_vf): Likewise.
12504 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12505 (expand_omp_simd): Handle polynomial safelen.
12506 * omp-low.c (omplow_simd_context): Add a default constructor.
12507 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12508 (lower_rec_simd_input_clauses): Update accordingly.
12509 (lower_rec_input_clauses): Likewise.
12511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12512 Alan Hayward <alan.hayward@arm.com>
12513 David Sherwood <david.sherwood@arm.com>
12515 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12516 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12517 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12518 (vect_analyze_slp_cost): Likewise.
12519 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12520 (vect_model_load_cost): Likewise.
12522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12523 Alan Hayward <alan.hayward@arm.com>
12524 David Sherwood <david.sherwood@arm.com>
12526 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12527 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12528 from an unsigned int * to a poly_uint64_pod *.
12529 (calculate_unrolling_factor): New function.
12530 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12533 Alan Hayward <alan.hayward@arm.com>
12534 David Sherwood <david.sherwood@arm.com>
12536 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12537 from an unsigned int to a poly_uint64.
12538 (_loop_vec_info::slp_unrolling_factor): Likewise.
12539 (_loop_vec_info::vectorization_factor): Change from an int
12541 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12542 (vect_get_num_vectors): New function.
12543 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12544 (vect_get_num_copies): Use vect_get_num_vectors.
12545 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12546 to an unsigned int *.
12547 (vect_analyze_data_refs): Change min_vf from an int * to a
12549 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12550 than an unsigned HOST_WIDE_INT.
12551 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12552 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12553 to an unsigned int *.
12554 (vect_analyze_data_ref_dependences): Likewise.
12555 (vect_compute_data_ref_alignment): Handle polynomial vf.
12556 (vect_enhance_data_refs_alignment): Likewise.
12557 (vect_prune_runtime_alias_test_list): Likewise.
12558 (vect_shift_permute_load_chain): Likewise.
12559 (vect_supportable_dr_alignment): Likewise.
12560 (dependence_distance_ge_vf): Take the vectorization factor as a
12561 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12562 (vect_analyze_data_refs): Change min_vf from an int * to a
12564 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12565 vfm1 as a poly_uint64 rather than an int. Make the same change
12566 for the returned bound_scalar.
12567 (vect_gen_vector_loop_niters): Handle polynomial vf.
12568 (vect_do_peeling): Likewise. Update call to
12569 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12570 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12572 * tree-vect-loop.c (vect_determine_vectorization_factor)
12573 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12574 (vect_get_known_peeling_cost): Likewise.
12575 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12576 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12577 (vect_transform_loop): Likewise. Use the lowest possible VF when
12578 updating the upper bounds of the loop.
12579 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12580 rather than an int.
12581 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12582 polynomial unroll factors.
12583 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12584 (vect_make_slp_decision): Likewise.
12585 (vect_supported_load_permutation_p): Likewise, and polynomial
12587 (vect_analyze_slp_cost): Handle polynomial vf.
12588 (vect_slp_analyze_node_operations): Likewise.
12589 (vect_slp_analyze_bb_1): Likewise.
12590 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12591 than an unsigned HOST_WIDE_INT.
12592 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12593 (vectorizable_load): Handle polynomial vf.
12594 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12596 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12599 Alan Hayward <alan.hayward@arm.com>
12600 David Sherwood <david.sherwood@arm.com>
12602 * match.pd: Handle bit operations involving three constants
12603 and try to fold one pair.
12605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12607 * tree-vect-loop-manip.c: Include gimple-fold.h.
12608 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12609 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12610 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12611 Add a path that uses a step of VF instead of 1, but disable it
12613 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12614 and niters_no_overflow parameters. Update calls to
12615 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12616 Create a new SSA name if the latter choses to use a ste other
12617 than zero, and return it via niters_vector_mult_vf_var.
12618 * tree-vect-loop.c (vect_transform_loop): Update calls to
12619 vect_do_peeling, vect_gen_vector_loop_niters and
12620 slpeel_make_loop_iterate_ntimes.
12621 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12622 (vect_gen_vector_loop_niters): Update declarations after above changes.
12624 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12626 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12627 128-bit round to integer instructions.
12628 (ceil<mode>2): Likewise.
12629 (btrunc<mode>2): Likewise.
12630 (round<mode>2): Likewise.
12632 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12634 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12635 unaligned VSX load/store on P8/P9.
12636 (expand_block_clear): Allow the use of unaligned VSX
12637 load/store on P8/P9.
12639 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12641 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12643 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12644 swap associated with both a load and a store.
12646 2018-01-02 Andrew Waterman <andrew@sifive.com>
12648 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12649 * config/riscv/riscv.md (clear_cache): Use it.
12651 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12653 * web.c: Remove out-of-date comment.
12655 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12657 * expr.c (fixup_args_size_notes): Check that any existing
12658 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12659 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12660 (emit_single_push_insn): ...here.
12662 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12664 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12665 (const_vector_encoded_nelts): New function.
12666 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12667 (const_vector_int_elt, const_vector_elt): Declare.
12668 * emit-rtl.c (const_vector_int_elt_1): New function.
12669 (const_vector_elt): Likewise.
12670 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12671 of CONST_VECTOR_ELT.
12673 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12675 * expr.c: Include rtx-vector-builder.h.
12676 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12677 directly on the tree encoding.
12678 (const_vector_from_tree): Likewise.
12679 * optabs.c: Include rtx-vector-builder.h.
12680 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12681 sequence of "u" values.
12682 * vec-perm-indices.c: Include rtx-vector-builder.h.
12683 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12684 directly on the vec_perm_indices encoding.
12686 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12688 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12689 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12690 * rtx-vector-builder.h: New file.
12691 * rtx-vector-builder.c: Likewise.
12692 * rtl.h (rtx_def::u2): Add a const_vector field.
12693 (CONST_VECTOR_NPATTERNS): New macro.
12694 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12695 (CONST_VECTOR_DUPLICATE_P): Likewise.
12696 (CONST_VECTOR_STEPPED_P): Likewise.
12697 (CONST_VECTOR_ENCODED_ELT): Likewise.
12698 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12699 (unwrap_const_vec_duplicate): Likewise.
12700 (const_vec_series_p): Check for a non-duplicated vector encoding.
12701 Say that the function only returns true for integer vectors.
12702 * emit-rtl.c: Include rtx-vector-builder.h.
12703 (gen_const_vec_duplicate_1): Delete.
12704 (gen_const_vector): Call gen_const_vec_duplicate instead of
12705 gen_const_vec_duplicate_1.
12706 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12707 (gen_const_vec_duplicate): Use rtx_vector_builder.
12708 (gen_const_vec_series): Likewise.
12709 (gen_rtx_CONST_VECTOR): Likewise.
12710 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12711 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12712 Build a new vector rather than modifying a CONST_VECTOR in-place.
12713 (handle_special_swappables): Update call accordingly.
12714 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12715 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12716 Build a new vector rather than modifying a CONST_VECTOR in-place.
12717 (handle_special_swappables): Update call accordingly.
12719 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12721 * simplify-rtx.c (simplify_const_binary_operation): Use
12722 CONST_VECTOR_ELT instead of XVECEXP.
12724 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12726 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12727 the selector elements to be different from the data elements
12728 if the selector is a VECTOR_CST.
12729 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12730 ssizetype for the selector.
12732 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12734 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12735 before testing each element individually.
12736 * tree-vect-generic.c (lower_vec_perm): Likewise.
12738 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12740 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12741 * selftest-run-tests.c (selftest::run_tests): Call it.
12742 * vector-builder.h (vector_builder::operator ==): New function.
12743 (vector_builder::operator !=): Likewise.
12744 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12745 (vec_perm_indices::all_from_input_p): New function.
12746 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12747 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12748 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12749 instead of reading the VECTOR_CST directly. Detect whether both
12750 vector inputs are the same before constructing the vec_perm_indices,
12751 and update the number of inputs argument accordingly. Use the
12752 utility functions added above. Only construct sel2 if we need to.
12754 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12756 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12757 the broadcast of the low byte.
12758 (expand_mult_highpart): Use an explicit encoding for the permutes.
12759 * optabs-query.c (can_mult_highpart_p): Likewise.
12760 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12761 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12762 (vectorizable_bswap): Likewise.
12763 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12764 explicit encoding for the power-of-2 permutes.
12765 (vect_permute_store_chain): Likewise.
12766 (vect_grouped_load_supported): Likewise.
12767 (vect_permute_load_chain): Likewise.
12769 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12771 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12772 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12773 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12774 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12775 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12776 (vect_gen_perm_mask_any): Likewise.
12778 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12780 * int-vector-builder.h: New file.
12781 * vec-perm-indices.h: Include int-vector-builder.h.
12782 (vec_perm_indices): Redefine as an int_vector_builder.
12783 (auto_vec_perm_indices): Delete.
12784 (vec_perm_builder): Redefine as a stand-alone class.
12785 (vec_perm_indices::vec_perm_indices): New function.
12786 (vec_perm_indices::clamp): Likewise.
12787 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12788 (vec_perm_indices::new_vector): New function.
12789 (vec_perm_indices::new_expanded_vector): Update for new
12790 vec_perm_indices class.
12791 (vec_perm_indices::rotate_inputs): New function.
12792 (vec_perm_indices::all_in_range_p): Operate directly on the
12793 encoded form, without computing elided elements.
12794 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12795 encoding. Update for new vec_perm_indices class.
12796 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12797 the given vec_perm_builder.
12798 (expand_vec_perm_var): Update vec_perm_builder constructor.
12799 (expand_mult_highpart): Use vec_perm_builder instead of
12800 auto_vec_perm_indices.
12801 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12802 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12803 or double series encoding as appropriate.
12804 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12805 vec_perm_indices instead of auto_vec_perm_indices.
12806 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12807 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12808 (vect_permute_store_chain): Likewise.
12809 (vect_grouped_load_supported): Likewise.
12810 (vect_permute_load_chain): Likewise.
12811 (vect_shift_permute_load_chain): Likewise.
12812 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12813 (vect_transform_slp_perm_load): Likewise.
12814 (vect_schedule_slp_instance): Likewise.
12815 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12816 (vectorizable_mask_load_store): Likewise.
12817 (vectorizable_bswap): Likewise.
12818 (vectorizable_store): Likewise.
12819 (vectorizable_load): Likewise.
12820 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12821 vec_perm_indices instead of auto_vec_perm_indices. Use
12822 tree_to_vec_perm_builder to read the vector from a tree.
12823 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12824 vec_perm_builder instead of a vec_perm_indices.
12825 (have_whole_vector_shift): Use vec_perm_builder and
12826 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12827 truncation to calc_vec_perm_mask_for_shift.
12828 (vect_create_epilog_for_reduction): Likewise.
12829 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12830 from auto_vec_perm_indices to vec_perm_indices.
12831 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12832 instead of changing individual elements.
12833 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12834 the vector in d.perm.
12835 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12836 from auto_vec_perm_indices to vec_perm_indices.
12837 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12838 instead of changing individual elements.
12839 (arm_vectorize_vec_perm_const): Use new_vector to install
12840 the vector in d.perm.
12841 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12842 Update vec_perm_builder constructor.
12843 (rs6000_expand_interleave): Likewise.
12844 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12845 (rs6000_expand_interleave): Likewise.
12847 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12849 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12850 to qimode could truncate the indices.
12851 * optabs.c (expand_vec_perm_var): Likewise.
12853 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12855 * Makefile.in (OBJS): Add vec-perm-indices.o.
12856 * vec-perm-indices.h: New file.
12857 * vec-perm-indices.c: Likewise.
12858 * target.h (vec_perm_indices): Replace with a forward class
12860 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12861 * optabs.h: Include vec-perm-indices.h.
12862 (expand_vec_perm): Delete.
12863 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12864 (expand_vec_perm_const): Declare.
12865 * target.def (vec_perm_const_ok): Replace with...
12866 (vec_perm_const): ...this new hook.
12867 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12868 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12869 * doc/tm.texi: Regenerate.
12870 * optabs.def (vec_perm_const): Delete.
12871 * doc/md.texi (vec_perm_const): Likewise.
12872 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12873 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12874 expand_vec_perm for constant permutation vectors. Assert that
12875 the mode of variable permutation vectors is the integer equivalent
12876 of the mode that is being permuted.
12877 * optabs-query.h (selector_fits_mode_p): Declare.
12878 * optabs-query.c: Include vec-perm-indices.h.
12879 (selector_fits_mode_p): New function.
12880 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12881 is defined, instead of checking whether the vec_perm_const_optab
12882 exists. Use targetm.vectorize.vec_perm_const instead of
12883 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12884 fit in the vector mode before using a variable permute.
12885 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12886 vec_perm_indices instead of an rtx.
12887 (expand_vec_perm): Replace with...
12888 (expand_vec_perm_const): ...this new function. Take the selector
12889 as a vec_perm_indices rather than an rtx. Also take the mode of
12890 the selector. Update call to shift_amt_for_vec_perm_mask.
12891 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12892 Use vec_perm_indices::new_expanded_vector to expand the original
12893 selector into bytes. Check whether the indices fit in the vector
12894 mode before using a variable permute.
12895 (expand_vec_perm_var): Make global.
12896 (expand_mult_highpart): Use expand_vec_perm_const.
12897 * fold-const.c: Includes vec-perm-indices.h.
12898 * tree-ssa-forwprop.c: Likewise.
12899 * tree-vect-data-refs.c: Likewise.
12900 * tree-vect-generic.c: Likewise.
12901 * tree-vect-loop.c: Likewise.
12902 * tree-vect-slp.c: Likewise.
12903 * tree-vect-stmts.c: Likewise.
12904 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12906 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12907 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12908 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12909 (aarch64_vectorize_vec_perm_const): ...this new function.
12910 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12911 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12912 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12913 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12914 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12915 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12916 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12918 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12919 check for NEON modes.
12920 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12921 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12922 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12923 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12925 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12926 the old VEC_PERM_CONST conditions.
12927 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12928 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12929 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12930 (ia64_vectorize_vec_perm_const_ok): Merge into...
12931 (ia64_vectorize_vec_perm_const): ...this new function.
12932 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12933 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12934 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12935 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12936 * config/mips/mips.c (mips_expand_vec_perm_const)
12937 (mips_vectorize_vec_perm_const_ok): Merge into...
12938 (mips_vectorize_vec_perm_const): ...this new function.
12939 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12940 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12941 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12942 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12943 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12944 (rs6000_expand_vec_perm_const): Delete.
12945 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12947 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12948 (altivec_expand_vec_perm_const_le): Take each operand individually.
12949 Operate on constant selectors rather than rtxes.
12950 (altivec_expand_vec_perm_const): Likewise. Update call to
12951 altivec_expand_vec_perm_const_le.
12952 (rs6000_expand_vec_perm_const): Delete.
12953 (rs6000_vectorize_vec_perm_const_ok): Delete.
12954 (rs6000_vectorize_vec_perm_const): New function.
12955 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12956 an element count and rtx array.
12957 (rs6000_expand_extract_even): Update call accordingly.
12958 (rs6000_expand_interleave): Likewise.
12959 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12960 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12961 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12962 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12963 (rs6000_expand_vec_perm_const): Delete.
12964 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12965 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12966 (altivec_expand_vec_perm_const_le): Take each operand individually.
12967 Operate on constant selectors rather than rtxes.
12968 (altivec_expand_vec_perm_const): Likewise. Update call to
12969 altivec_expand_vec_perm_const_le.
12970 (rs6000_expand_vec_perm_const): Delete.
12971 (rs6000_vectorize_vec_perm_const_ok): Delete.
12972 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12973 reference to the SPE evmerge intructions.
12974 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12975 an element count and rtx array.
12976 (rs6000_expand_extract_even): Update call accordingly.
12977 (rs6000_expand_interleave): Likewise.
12978 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12979 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12981 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12983 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12985 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12986 vector mode and that that mode matches the mode of the data
12988 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12989 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12990 directly using expand_vec_perm_1 when forcing selectors into
12992 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12994 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12996 * optabs-query.h (can_vec_perm_p): Delete.
12997 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12998 * optabs-query.c (can_vec_perm_p): Split into...
12999 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13000 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13001 particular selector is valid.
13002 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13003 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13004 (vect_grouped_load_supported): Likewise.
13005 (vect_shift_permute_load_chain): Likewise.
13006 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13007 (vect_transform_slp_perm_load): Likewise.
13008 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13009 (vectorizable_bswap): Likewise.
13010 (vect_gen_perm_mask_checked): Likewise.
13011 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13012 implementations of variable permutation vectors into account
13013 when deciding which selector to use.
13014 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13015 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13016 with a false third argument.
13017 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13018 to test whether the constant selector is valid and can_vec_perm_var_p
13019 to test whether a variable selector is valid.
13021 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13023 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13024 * optabs-query.c (can_vec_perm_p): Likewise.
13025 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13026 instead of vec_perm_indices.
13027 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13028 (vect_gen_perm_mask_checked): Likewise,
13029 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13030 (vect_gen_perm_mask_checked): Likewise,
13032 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13034 * optabs-query.h (qimode_for_vec_perm): Declare.
13035 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13036 (qimode_for_vec_perm): ...this new function.
13037 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13039 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13041 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13042 does not have a conditional at the top.
13044 2018-01-02 Richard Biener <rguenther@suse.de>
13046 * ipa-inline.c (big_speedup_p): Fix expression.
13048 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13051 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13054 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13058 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13059 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13060 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13061 cond_taken_branch_cost 3->4.
13063 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13065 PR tree-optimization/83581
13066 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13067 TODO_cleanup_cfg if any changes have been made.
13069 PR middle-end/83608
13070 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13071 convert_modes if target mode has the right side, but different mode
13074 PR middle-end/83609
13075 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13076 last argument when extracting from CONCAT. If either from_real or
13077 from_imag is NULL, use expansion through memory. If result is not
13078 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13079 the parts directly to inner mode, if even that fails, use expansion
13082 PR middle-end/83623
13083 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13084 check for bswap in mode rather than HImode and use that in expand_unop
13087 Copyright (C) 2018 Free Software Foundation, Inc.
13089 Copying and distribution of this file, with or without modification,
13090 are permitted in any medium without royalty provided the copyright
13091 notice and this notice are preserved.