1 2018-05-17 Jim Wilson <jimw@sifive.com>
3 * expr.c (do_tablejump): When converting index to Pmode, if we have a
4 sign extended promoted subreg, and the range does not have the sign bit
5 set, then do a sign extend.
7 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
8 test, check for sign extended subreg and/or constant operands, and
9 do a sign extend in that case.
11 2018-05-17 Steve Ellcey <sellcey@cavium.com>
13 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
14 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
16 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
17 Change logics_shift_reg to logics_shift_imm.
18 (thunderx2t99_fp_loadpair_basic): Delete.
19 (thunderx2t99_fp_storepair_basic): Delete.
20 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
21 (thunderx2t99_asimd_polynomial): Delete.
22 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
23 and neon_fp_mul_d_scalar_q.
24 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
25 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
26 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
27 (thunderx2t99_asimd_lut): Add missing tbl types.
28 (thunderx2t99_asimd_ext): Delete.
29 (thunderx2t99_asimd_load1_1_mult): Delete.
30 (thunderx2t99_asimd_load1_2_mult): Delete.
31 (thunderx2t99_asimd_load1_ldp): New.
32 (thunderx2t99_asimd_load1): New.
33 (thunderx2t99_asimd_load2): Add missing *load2* types.
34 (thunderx2t99_asimd_load3): New.
35 (thunderx2t99_asimd_load4): New.
36 (thunderx2t99_asimd_store1_1_mult): Delete.
37 (thunderx2t99_asimd_store1_2_mult): Delete.
38 (thunderx2t99_asimd_store2_mult): Delete.
39 (thunderx2t99_asimd_store2_onelane): Delete.
40 (thunderx2t99_asimd_store_stp): New.
41 (thunderx2t99_asimd_store1): New.
42 (thunderx2t99_asimd_store2): New.
43 (thunderx2t99_asimd_store3): New.
44 (thunderx2t99_asimd_store4): New.
46 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
48 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
49 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
51 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
52 Segher Boessenkool <segher@kernel.crashing.org>
55 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
58 2018-05-17 Richard Biener <rguenther@suse.de>
60 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
61 for pruning loop and prune defs feeding only already visited PHIs.
63 2018-05-17 Richard Biener <rguenther@suse.de>
65 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
67 2018-05-17 Bin Cheng <bin.cheng@arm.com>
68 Richard Biener <rguenther@suse.de>
70 PR tree-optimization/85793
71 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
74 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
76 * internal-fn.h (lookup_internal_fn): Declare
77 * internal-fn.c (lookup_internal_fn): New function.
78 * gimple.c (gimple_build_call_from_tree): Handle calls to
80 * gimple-pretty-print.c (dump_gimple_call): Print "." before
81 internal function names.
82 * tree-pretty-print.c (dump_generic_node): Likewise.
83 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
85 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
87 * gimple-fold.h (gimple_build): Make the function forms take
88 combined_fn rather than built_in_function.
89 (gimple_simplify): Likewise.
90 * gimple-match-head.c (gimple_simplify): Likewise.
91 * gimple-fold.c (gimple_build): Likewise.
92 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
93 rather than gimple_build_call_internal.
94 (get_initial_defs_for_reduction): Likewise.
95 (vect_create_epilog_for_reduction): Likewise.
96 (vectorizable_live_operation): Likewise.
98 2018-05-17 Martin Liska <mliska@suse.cz>
100 * gimple-ssa-sprintf.c (format_directive): Do not use
101 space in between 'G_' and '('.
103 2018-05-17 Jakub Jelinek <jakub@redhat.com>
106 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
107 even if the mask is not all ones.
110 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
112 (ix86_gimple_fold_builtin): Likewise.
115 * config/i386/i386.c: Include tree-vector-builder.h.
116 (ix86_vector_shift_count): New function.
117 (ix86_fold_builtin): Fold shift builtins by scalar count.
118 (ix86_gimple_fold_builtin): Likewise.
120 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
121 _mm512_setzero): New intrinsics.
123 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
124 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
126 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
127 code generation for cases where splatting a value is not useful.
128 * simplify-rtx.c (simplify_ternary_operation): Simplify
129 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
130 mode to a vec_concat.
132 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
134 * config.gcc: Support "goldmont-plus".
135 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
137 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
138 PROCESSOR_GOLDMONT_PLUS.
139 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
140 (processor_target_table): Add "goldmont-plus".
141 (PTA_GOLDMONT_PLUS): Define.
142 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
143 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
144 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
145 (fold_builtin_cpu): Add "goldmont-plus".
146 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
147 (ix86_option_override_internal): Add "goldmont-plus".
148 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
149 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
150 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
151 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
153 2018-05-17 Richard Biener <rguenther@suse.de>
155 PR tree-optimization/85757
156 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
157 remove defs that only feed that PHI from further processing.
159 2018-05-16 Jim Wilson <jimw@sifive.com>
161 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
163 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
165 2018-05-16 Mark Wielaard <mark@klomp.org>
167 * dwarf2out.c (count_index_strings): New function.
168 (output_indirect_strings): Call count_index_strings and generate
169 header for dwarf_version >= 5.
171 2018-05-16 Mark Wielaard <mark@klomp.org>
173 * dwarf2out.c (dwarf_FORM): New function.
174 (set_indirect_string): Use dwarf_FORM.
175 (reset_indirect_string): Likewise.
176 (size_of_die): Likewise.
177 (value_format): Likewise.
178 (output_die): Likewise.
179 (add_skeleton_AT_string): Likewise.
180 (output_macinfo_op): Likewise.
181 (index_string): Likewise.
182 (output_index_string_offset): Likewise.
183 (output_index_string): Likewise.
184 (count_index_strings): Likewise.
186 2018-05-16 Carl Love <cel@us.ibm.com>
188 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
189 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
191 2018-05-16 Martin Jambor <mjambor@suse.cz>
193 * ipa-prop.c (ipa_free_all_edge_args): Remove.
194 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
196 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
198 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
199 (fnma<mode>4): Likewise.
200 (fms<mode>4): Likewise.
201 (fnms<mode>4): Likewise.
202 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
203 (aarch64_fnma<mode>4): Likewise.
204 (aarch64_fms<mode>4): Likewise.
205 (aarch64_fnms<mode>4): Likewise.
206 (aarch64_fnmadd<mode>4): Likewise.
208 2018-05-16 Jason Merrill <jason@redhat.com>
210 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
212 2018-05-16 Richard Biener <rguenther@suse.de>
214 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
215 (dump_stmt_cost): Declare.
216 (add_stmt_cost): Dump cost we add.
217 (add_stmt_costs): New function.
218 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
220 (vect_analyze_stmt): Adjust prototype.
221 (vectorizable_condition): Likewise.
222 (vectorizable_live_operation): Likewise.
223 (vectorizable_reduction): Likewise.
224 (vectorizable_induction): Likewise.
225 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
226 cost vector to pass to vectorizable_ and record afterwards.
227 (vect_model_reduction_cost): Take cost vector argument and adjust.
228 (vect_model_induction_cost): Likewise.
229 (vectorizable_reduction): Likewise.
230 (vectorizable_induction): Likewise.
231 (vectorizable_live_operation): Likewise.
232 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
233 SLP_TREE_NUMBER_OF_VEC_STMTS.
234 (vect_analyze_slp_cost_1): Remove.
235 (vect_analyze_slp_cost): Likewise.
236 (vect_slp_analyze_node_operations): Take visited args and
237 a target cost vector. Avoid processing already visited stmt sets.
238 (vect_slp_analyze_operations): Use a local cost vector to gather
239 costs and register those of non-discarded instances.
240 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
241 (vect_schedule_slp_instance): Remove copying of
242 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
244 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
245 adding cost. Record cost entry location.
246 (vect_prologue_cost_for_slp_op): Function to compute cost of
247 a constant or invariant generated for SLP vect in the prologue,
248 split out from vect_analyze_slp_cost_1.
249 (vect_model_simple_cost): Make static. Adjust for SLP costing.
250 (vect_model_promotion_demotion_cost): Likewise.
251 (vect_model_store_cost): Likewise, make static.
252 (vect_model_load_cost): Likewise.
253 (vectorizable_bswap): Add cost vector arg and adjust.
254 (vectorizable_call): Likewise.
255 (vectorizable_simd_clone_call): Likewise.
256 (vectorizable_conversion): Likewise.
257 (vectorizable_assignment): Likewise.
258 (vectorizable_shift): Likewise.
259 (vectorizable_operation): Likewise.
260 (vectorizable_store): Likewise.
261 (vectorizable_load): Likewise.
262 (vectorizable_condition): Likewise.
263 (vectorizable_comparison): Likewise.
264 (can_vectorize_live_stmts): Likewise.
265 (vect_analyze_stmt): Likewise.
266 (vect_transform_stmt): Adjust calls to vectorizable_*.
267 * tree-vectorizer.c: Include gimple-pretty-print.h.
268 (dump_stmt_cost): New function.
270 2018-05-16 Richard Biener <rguenther@suse.de>
272 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
273 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
274 * tree-ssa-dse.c: Include tree-ssa-loop.h.
275 (check_name): New callback.
276 (dse_classify_store): Track cycles via a visited bitmap of PHI
277 defs and simplify handling of in-loop and across loop dead stores
278 and properly fail for loop-variant refs. Handle byte-tracking with
279 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
282 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
284 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
285 (vect_get_mask_type_for_stmt): Likewise.
286 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
288 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
289 to determine the statement's vector type and the vector type that
290 should be used for calculating nunits. Deal with cases in which
291 the type has to be deferred.
292 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
293 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
294 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
295 (vect_determine_vf_for_stmt): New functions, split out from...
296 (vect_determine_vectorization_factor): ...here.
297 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
298 (vect_get_mask_type_for_stmt): New functions, split out from
299 vect_determine_vectorization_factor.
301 2018-05-16 Richard Biener <rguenther@suse.de>
303 * tree-cfg.c (verify_gimple_assign_ternary): Properly
304 verify the [VEC_]COND_EXPR embedded comparison.
306 2018-05-15 Martin Sebor <msebor@redhat.com>
308 PR tree-optimization/85753
309 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
310 RECORD_TYPE in addition to ARRAY_TYPE.
312 2018-05-15 Martin Sebor <msebor@redhat.com>
315 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
317 2018-05-15 Richard Biener <rguenther@suse.de>
319 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
320 add by_clobber_p one. Change algorithm to collect all defs
321 representing uses we need to walk and try reducing them to
322 a single one before failing.
323 (dse_dom_walker::dse_optimize_stmt): Adjust.
325 2018-05-13 Mark Wielaard <mark@klomp.org>
327 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
328 (size_of_loc_descr): Likewise.
329 (output_loc_operands): Likewise.
330 (output_loc_operands_raw): Likewise.
331 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
332 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
333 (hash_loc_operands): Likewise.
334 (compare_loc_operands): Likewise.
336 2018-05-14 Mark Wielaard <mark@klomp.org>
338 * dwarf2out.c (count_index_addrs): New function.
339 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
341 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
343 PR tree-optimization/83648
344 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
345 return value as malloc candidate.
347 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
350 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
351 as true in call to suggest_attribute.
353 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
355 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
358 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
360 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
362 (aarch64_crypto_aesd_fused): Likewise.
364 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
366 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
367 (movsi_aarch64): Likewise.
368 (load_pairsi): Likewise.
369 (load_pairdi): Likewise.
370 (store_pairsi): Likewise.
371 (store_pairdi): Likewise.
372 (load_pairsf): Likewise.
373 (load_pairdf): Likewise.
374 (store_pairsf): Likewise.
375 (store_pairdf): Likewise.
376 (zero_extend): Likewise.
377 (trunc): Swap alternatives.
378 (fcvt_target): Add '?' to prefer w over r.
380 2018-05-14 Jakub Jelinek <jakub@redhat.com>
383 * config/i386/i386.md: Disallow non-commutative arithmetics in
384 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
385 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
386 in the peephole2 before it.
388 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
390 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
391 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
392 (ix86_handle_option): Handle -mcldemote.
393 * config.gcc: New header.
394 * config/i386/cldemoteintrin.h: New file.
395 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
396 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
398 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
399 OPTION_MASK_ISA_CLDEMOTE.
400 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
401 (ix86_valid_target_attribute_inner_p): Ditto.
402 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
403 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
404 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
405 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
406 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
408 * config/i386/i386.opt: Add -mcldemote.
409 * config/i386/x86intrin.h: New header.
410 * doc/invoke.texi: Add -mcldemote.
412 2018-05-14 Richard Biener <rguenther@suse.de>
414 * doc/match-and-simplify.texi: Adjust :s documentation.
416 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
418 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
419 intended memcpy size.
420 (REORDER_45): Likewise.
422 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
425 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
426 * vec.c (qsort_chk): Use gcc_qsort.
427 * Makefile.in (OBJS-libcommon): Add sort.o.
428 (build/sort.o): New target. Use it...
429 (BUILD_RTL): ... here, and...
430 (build/gencfn-macros): ... here, and...
431 (build/genmatch): ... here.
433 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
434 Chung-Ju Wu <jasonwucj@gmail.com>
436 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
437 * config/nds32/nds32-graywolf.md: New file.
438 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
439 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
441 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
442 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
443 * config/nds32/nds32.md (pipeline_model): Add graywolf.
444 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
445 * config/nds32/pipelines.md: Include n15 settings.
447 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
448 Chung-Ju Wu <jasonwucj@gmail.com>
450 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
451 * config/nds32/nds32-n13.md: New file.
452 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
453 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
455 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
456 * config/nds32/nds32.md (pipeline_model): Add n13.
457 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
458 * config/nds32/pipelines.md: Include n13 settings.
460 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
461 Chung-Ju Wu <jasonwucj@gmail.com>
463 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
464 * config/nds32/nds32-n10.md: New file.
465 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
466 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
468 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
469 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
470 * config/nds32/nds32.md (pipeline_model): Add n10.
471 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
472 * config/nds32/pipelines.md: Include n10 settings.
474 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
475 Kito Cheng <kito.cheng@gmail.com>
476 Chung-Ju Wu <jasonwucj@gmail.com>
478 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
479 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
480 Add enum values for DSP extension instructions.
481 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
483 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
484 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
486 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
487 * config/nds32/nds32-dspext.md: New file for DSP implementation.
488 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
489 * config/nds32/nds32-intrinsic.md: Likewise.
490 * config/nds32/nds32_intrinsic.h: Likewise.
491 * config/nds32/nds32-md-auxiliary.c: Likewise.
492 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
493 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
494 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
495 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
496 * config/nds32/nds32-protos.h: New declarations for DSP extension.
497 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
498 TYPE_DMAC in switch statement.
499 * config/nds32/nds32.c: New checking and implementation for DSP
500 extension instructions.
501 * config/nds32/nds32.h: Likewise.
502 * config/nds32/nds32.md: Likewise.
503 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
504 * config/nds32/predicates.md: Implement new predicates for DSP
507 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
509 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
510 Reformat alternatives and attributes so it is easier to identify
511 which constraints/attributes go with which instruction.
512 (mov<mode>_hardfloat32, FMOVE64): Likewise.
513 (mov<mode>_softfloat32, FMOVE64): Likewise.
514 (mov<mode>_hardfloat64, FMOVE64): Likewise.
515 (mov<mode>_softfloat64, FMOVE64): Likewise.
517 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
519 * doc/extend.texi (PowerPC Built-in Functions): Rename this
521 (Basic PowerPC Built-in Functions): The new name of the
522 subsection previously known as "PowerPC Built-in Functions".
523 (Basic PowerPC Built-in Functions Available on all Configurations):
525 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
526 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
527 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
528 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
530 2018-05-11 Martin Jambor <mjambor@suse.cz>
533 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
536 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
539 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
541 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
543 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
544 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
545 (ix86_handle_option): Handle -mwaitpkg.
546 * config.gcc: New header.
547 * config/i386/cpuid.h (bit_WAITPKG): New bit.
548 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
549 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
551 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
552 OPTION_MASK_ISA_WAITPKG.
553 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
554 (ix86_option_override_internal): Add PTA_WAITPKG.
555 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
556 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
558 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
559 __builtin_ia32_umwait and __builtin_ia32_tpause.
560 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
561 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
562 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
563 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
564 UNSPECV_TPAUSE): New.
565 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
566 * config/i386/i386.opt: Add -mwaitpkg.
567 * config/i386/waitpkgintrin.h: New file.
568 * config/i386/x86intrin.h: New header.
569 * doc/invoke.texi: Add -mwaitpkg.
571 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
574 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
576 (cortex-m0): Use armv6s-m isa.
577 (cortex-m0plus): Likewise.
578 (cortex-m1): Likewise.
579 (cortex-m0.small-multiply): Likewise.
580 (cortex-m0plus.small-multiply): Likewise.
581 (cortex-m1.small-multiply): Likewise.
583 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
584 Jakub Jelinek <jakub@redhat.com>
586 PR tree-optimization/85692
587 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
588 source permute as well.
590 2018-05-11 Martin Liska <mliska@suse.cz>
593 * doc/extend.texi: Document LLVM style format for no_sanitize
596 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
598 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
599 mode_supports_vsx_dform_quad to mode_supports_dq_form.
600 (mode_supports_vsx_dform_quad): Likewise.
601 (mode_supports_vmx_dform): Move these functions to be next to the
602 other mode_supports functions.
603 (mode_supports_dq_form): Likewise.
604 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
605 mode_supports_dq_form.
606 (reg_offset_addressing_ok_p): Likewise.
607 (offsettable_ok_by_alignment): Likewise.
608 (rs6000_legitimate_offset_address_p): Likewise.
609 (legitimate_lo_sum_address_p): Likewise.
610 (rs6000_legitimize_address): Likewise.
611 (rs6000_legitimize_reload_address): Likewise.
612 (rs6000_secondary_reload_inner): Likewise.
613 (rs6000_preferred_reload_class): Likewise.
614 (rs6000_output_move_128bit): Likewise.
616 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
618 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
619 Generate SImode target register for null target.
620 <case IX86_BUILTIN_XGETBV>: Ditto.
621 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
622 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
624 2018-05-10 Carl Love <cel@us.ibm.com>
626 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
627 dcbtt and dcbtstt if operands[2] is 0.
629 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
632 * config/i386/sse.md (usadv64qi): New expander.
634 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
636 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
637 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
638 -maltivec=be support.
639 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
640 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
641 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
642 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
643 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
644 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
645 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
646 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
647 altivec_vsumsws): Adjust.
648 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
649 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
650 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
652 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
653 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
654 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
655 (altivec_lve<VI_char>x): Delete expand.
656 (*altivec_lve<VI_char>x_internal): Rename to...
657 (altivec_lve<VI_char>x): ... this.
658 (altivec_lvxl_<mode>): Delete expand.
659 (*altivec_lvxl_<mode>_internal): Rename to ...
660 (altivec_lvxl_<mode>): ... this.
661 (altivec_stvxl_<mode>): Delete expand.
662 (*altivec_stvxl_<mode>_internal): Rename to ...
663 (altivec_stvxl_<mode>): ... this.
664 (altivec_stve<VI_char>x): Delete expand.
665 (*altivec_stve<VI_char>x_internal): Rename to ...
666 (altivec_stve<VI_char>x): ... this.
667 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
668 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
669 reduc_plus_scal_<mode>): Adjust.
670 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
672 (rs6000_cpu_cpp_builtins): Adjust.
673 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
674 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
675 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
676 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
677 -maltivec=be support.
678 (rs6000_split_vec_extract_var): Adjust.
679 (rs6000_split_v4si_init): Adjust.
680 (swap_selector_for_mode): Delete.
681 (altivec_expand_lvx_be, altivec_expand_stvx_be,
682 altivec_expand_stvex_be): Delete.
683 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
684 -maltivec=be support.
685 (rs6000_gimple_fold_builtin): Ditto.
686 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
688 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
689 (TARGET_DIRECT_MOVE_64BIT): Adjust.
690 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
691 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
692 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
693 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
694 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
695 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
696 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
697 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
698 anonymous split): Adjust.
699 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
700 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
702 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
704 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
705 when --with-gxx-include-dir is also specified.
706 * configure: Regenerate.
708 2018-05-09 Jim Wilson <jimw@sifive.com>
711 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
712 * config/riscv/t-withmultilib: New.
713 * config/riscv/withmultilib.h: New.
714 * doc/install.texi: Document RISC-V --with-multilib-list support.
716 2018-05-09 Richard Biener <rguenther@suse.de>
718 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
720 (vect_bb_vectorization_profitable_p): Adjust. Compute
721 actual scalar cost using the cost vector and the add_stmt_cost
724 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
726 PR rtl-optimization/85645
727 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
728 in the REG_CFA_REGISTER note for LR, don't leave it empty.
730 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
732 PR rtl-optimization/85645
733 * shrink-wrap.c (spread_components): Return a boolean saying if
734 anything was changed.
735 (try_shrink_wrapping_separate): Iterate spread_components until
736 nothing changes anymore.
738 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
740 PR rtl-optimization/85645
741 * regrename.c (build_def_use): Also kill the chains that include the
742 destination of a REG_CFA_REGISTER note.
744 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
746 PR rtl-optimization/85645
747 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
748 insn that has a REG_CFA_REGISTER note.
750 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
752 * cfgexpand.c (expand_clobber): New function.
753 (expand_gimple_stmt_1): Use it.
754 * tree-vect-stmts.c (vect_clobber_variable): New function,
756 (vectorizable_simd_clone_call): ...here.
757 (vectorizable_store): Emit a clobber either side of an
758 IFN_STORE_LANES sequence.
759 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
761 2018-05-09 Tom de Vries <tom@codesourcery.com>
764 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
765 (define_insn "trap_if_false"): Add exit after trap.
767 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
769 PR rtl-optimization/85638
770 * bb-reorder.c: Include common/common-target.h.
771 (create_forwarder_block): New function extracted from...
772 (fix_up_crossing_landing_pad): ...here. Rename into...
773 (dw2_fix_up_crossing_landing_pad): ...this.
774 (sjlj_fix_up_crossing_landing_pad): New function.
775 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
776 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
777 from both partitions and exit the loop after one iteration.
779 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
782 * doc/extend.texi (PowerPC Built-in Functions): Rename this
784 (Basic PowerPC Built-in Functions): The new name of the
785 subsection previously known as "PowerPC Built-in Functions".
786 (Basic PowerPC Built-in Functions Available on all Configurations):
788 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
790 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
792 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
794 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
797 2018-05-08 Jim Wilson <jimw@sifive.com>
799 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
800 (LD_EMUL_SUFFIX): New.
803 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
805 * doc/extend.texi (PowerPC Built-in Functions): Rename this
807 (Basic PowerPC Built-in Functions): The new name of the
808 subsection previously known as "PowerPC Built-in Functions".
809 (Basic PowerPC Built-in Functions Available on all Configurations):
811 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
813 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
815 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
817 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
820 2018-05-08 Jakub Jelinek <jakub@redhat.com>
823 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
824 after cmpelim optimization.
826 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
828 * config.gcc: Support "goldmont".
829 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
830 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
832 * config/i386/i386.c (m_GOLDMONT): Define.
833 (processor_target_table): Add "goldmont".
834 (PTA_GOLDMONT): Define.
835 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
836 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
837 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
838 (fold_builtin_cpu): Add "goldmont".
839 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
840 (ix86_option_override_internal): Add "goldmont".
841 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
842 (processor_type): Add PROCESSOR_GOLDMONT.
843 * config/i386/i386.md: Add CPU "glm".
844 * config/i386/glm.md: New file.
845 * config/i386/x86-tune.def: Add m_GOLDMONT.
846 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
848 2018-05-08 Jakub Jelinek <jakub@redhat.com>
851 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
853 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
854 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
855 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
858 * config/i386/i386.c (ix86_fold_builtin): Handle
859 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
862 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
863 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
865 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
868 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
869 (check_arch): Likewise.
870 (check_fpu): Return the result rather than printing it.
871 (end arch): Fix operator precedence.
873 (END): Print the result from check_fpu.
875 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
876 Alan Hayward <alan.hayward@arm.com>
877 David Sherwood <david.sherwood@arm.com>
879 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
880 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
881 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
882 (*fcmuo<mode>_and): New patterns.
884 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
886 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
887 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
888 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
889 (cmp_op, sve_imm_con): New code attributes.
890 (SVE_COND_INT_CMP, imm_con): Delete.
891 (cmp_op): Remove above unspecs from int attribute.
892 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
894 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
895 comparison-specific unspecs.
896 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
897 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
898 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
899 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
900 (*vec_fcm<cmp_op><mode>): Rename to...
901 (*fcm<cmp_op><mode>): ...this and adjust likewise.
902 (*vec_fcmuo<mode>): Rename to...
903 (*fcmuo<mode>): ...this and adjust likewise.
904 (*pred_fcm<cmp_op><mode>): New pattern.
905 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
906 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
908 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
910 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
911 (aarch64_emit_sve_predicated_cond): New function.
912 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
913 (aarch64_emit_unspec_cond_or): Replace with...
914 (aarch64_emit_sve_or_conds): ...this new function. Use
915 aarch64_emit_sve_ptrue_op for the individual comparisons and
916 aarch64_emit_binop to OR them together.
917 (aarch64_emit_inverted_unspec_cond): Replace with...
918 (aarch64_emit_sve_inverted_cond): ...this new function. Use
919 aarch64_emit_sve_ptrue_op for the comparison and
920 aarch64_emit_unop to invert the result.
921 (aarch64_expand_sve_vec_cmp_float): Update after the above
922 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
924 2018-05-07 Nathan Sidwell <nathan@acm.org>
926 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
927 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
928 (Backwards Compatibility): Likewise.
930 2018-05-07 Luis Machado <luis.machado@linaro.org>
934 2018-05-07 Luis Machado <luis.machado@linaro.org>
936 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
937 <prefetch_dynamic_strides>: New const bool field.
938 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
939 prefetch_dynamic_strides.
940 (exynosm1_prefetch_tune): Likewise.
941 (thunderxt88_prefetch_tune): Likewise.
942 (thunderx_prefetch_tune): Likewise.
943 (thunderx2t99_prefetch_tune): Likewise.
944 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
945 (aarch64_override_options_internal): Update to set
946 PARAM_PREFETCH_DYNAMIC_STRIDES.
947 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
948 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
949 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
950 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
951 prefetch-dynamic-strides setting.
953 2018-05-07 Luis Machado <luis.machado@linaro.org>
955 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
956 <minimum_stride>: New const int field.
957 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
958 minimum_stride field.
959 (exynosm1_prefetch_tune): Likewise.
960 (thunderxt88_prefetch_tune): Likewise.
961 (thunderx_prefetch_tune): Likewise.
962 (thunderx2t99_prefetch_tune): Likewise.
963 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
964 (aarch64_override_options_internal): Update to set
965 PARAM_PREFETCH_MINIMUM_STRIDE.
966 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
967 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
968 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
969 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
970 stride is constant and is below the minimum stride threshold.
972 2018-05-07 Luis Machado <luis.machado@linaro.org>
974 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
977 2018-05-07 Luis Machado <luis.machado@linaro.org>
979 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
980 <prefetch_dynamic_strides>: New const bool field.
981 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
982 prefetch_dynamic_strides.
983 (exynosm1_prefetch_tune): Likewise.
984 (thunderxt88_prefetch_tune): Likewise.
985 (thunderx_prefetch_tune): Likewise.
986 (thunderx2t99_prefetch_tune): Likewise.
987 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
988 (aarch64_override_options_internal): Update to set
989 PARAM_PREFETCH_DYNAMIC_STRIDES.
990 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
991 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
992 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
993 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
994 prefetch-dynamic-strides setting.
996 2018-05-07 Luis Machado <luis.machado@linaro.org>
998 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
999 <minimum_stride>: New const int field.
1000 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1001 minimum_stride field.
1002 (exynosm1_prefetch_tune): Likewise.
1003 (thunderxt88_prefetch_tune): Likewise.
1004 (thunderx_prefetch_tune): Likewise.
1005 (thunderx2t99_prefetch_tune): Likewise.
1006 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1007 (aarch64_override_options_internal): Update to set
1008 PARAM_PREFETCH_MINIMUM_STRIDE.
1009 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1010 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1011 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1012 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1013 stride is constant and is below the minimum stride threshold.
1015 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1018 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1019 the type is addressable. Don't force op into register if it has
1022 2018-05-05 Roland McGrath <mcgrathr@google.com>
1025 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1026 any section for which we don't know a specific type it should have,
1027 regardless of name. Previously this was done only for the exact
1028 names ".init_array", ".fini_array", and ".preinit_array".
1029 (default_elf_asm_named_section): Add comment about
1030 relationship with default_section_type_flags and SECTION_NOTYPE.
1031 (get_section): Don't consider it a type conflict if one side has
1032 SECTION_NOTYPE and the other doesn't, as long as neither has the
1033 SECTION_BSS et al used in the default_section_type_flags logic.
1035 2018-05-05 Tom de Vries <tom@codesourcery.com>
1038 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1039 (workaround_barsyncs): New function.
1040 (nvptx_reorg): Use workaround_barsyncs.
1041 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1042 (define_expand "nvptx_membar_cta"): New define_expand.
1043 (define_insn "*nvptx_membar_cta"): New insn.
1045 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1047 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1048 To improve optimization opportunities.
1049 * builtin-types.def: The new needed builtin types for the above.
1051 2018-05-04 Richard Biener <rguenther@suse.de>
1053 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1054 * gimple-ssa-store-merging.c
1055 (imm_store_chain_info::output_merged_store): Remove redundant create,
1056 release split_store vector contents on failure.
1057 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1058 scalar stmt vector on cache hit.
1060 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1062 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1064 * config.gcc (powerpc-xilinx-eabi*): Remove.
1065 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1067 (fusion_addis_mem_combo_load): Ditto.
1068 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1070 (rs6000_cpu_cpp_builtins): Ditto.
1071 * config/rs6000/rs6000-linux.c
1072 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1073 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1074 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1076 (rs6000_setup_reg_addr_masks): Ditto.
1077 (rs6000_init_hard_regno_mode_ok): Ditto.
1078 (rs6000_option_override_internal): Ditto.
1079 (legitimate_lo_sum_address_p): Ditto.
1080 (rs6000_legitimize_address): Ditto.
1081 (rs6000_legitimize_reload_address): Ditto.
1082 (rs6000_legitimate_address_p): Ditto.
1083 (abi_v4_pass_in_fpr): Ditto.
1084 (setup_incoming_varargs): Ditto.
1085 (rs6000_gimplify_va_arg): Ditto.
1086 (rs6000_split_multireg_move): Ditto.
1087 (rs6000_savres_strategy): Ditto.
1088 (rs6000_emit_prologue_components): Ditto.
1089 (rs6000_emit_epilogue_components): Ditto.
1090 (rs6000_emit_prologue): Ditto.
1091 (rs6000_emit_epilogue): Ditto.
1092 (rs6000_elf_file_end): Ditto.
1093 (rs6000_function_value): Ditto.
1094 (rs6000_libcall_value): Ditto.
1095 * config/rs6000/rs6000.h: Ditto.
1096 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1097 (TARGET_MINMAX): ... this. New.
1098 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1099 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1100 (*movsi_internal1_single): Delete.
1101 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1102 mfpu=, mxilinx-fpu): Delete.
1103 * config/rs6000/singlefp.h: Delete.
1104 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1105 * config/rs6000/t-rs6000: Ditto.
1106 * config/rs6000/t-xilinx: Delete.
1107 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1108 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1109 (VStype_simple): Delete.
1110 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1111 * config/rs6000/xfpu.h: Delete.
1112 * config/rs6000/xfpu.md: Delete.
1113 * config/rs6000/xilinx.h: Delete.
1114 * config/rs6000/xilinx.opt: Delete.
1115 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1116 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1118 2018-05-04 Tom de Vries <tom@codesourcery.com>
1121 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1124 2018-05-04 Richard Biener <rguenther@suse.de>
1127 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1128 (expand_complex_div_wide): Likewise.
1129 (expand_complex_operations_1): Likewise.
1130 (expand_complex_libcall): Preserve EH info of the original stmt.
1131 (tree_lower_complex): Handle removed blocks.
1132 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1133 on complex multiplication and division libcall builtins.
1135 2018-05-04 Richard Biener <rguenther@suse.de>
1138 * fold-const.c (negate_expr_p): Restrict negation of operand
1139 zero of a division to when we know that can happen without
1141 (fold_negate_expr_1): Likewise.
1143 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1146 * real.h (real_nextafter): Declare.
1147 * real.c (real_nextafter): New function.
1148 * fold-const-call.c (fold_const_nextafter): New function.
1149 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1150 CASE_CFN_NEXTTOWARD.
1151 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1152 even when arg1_mode is different from arg0_mode.
1154 2018-05-03 Nathan Sidwell <nathan@acm.org>
1156 * doc/extend.texi (Deprecated Features): Remove
1158 (Backwards Compatibility): Likewise.
1159 * doc/invoke.texi (C++ Language Options): Likewise.
1160 (C++ Dialect Options): Likewise.
1162 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1165 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1166 _mm512_mask_mullox_epi64): New intrinsics.
1168 2018-05-03 Tom de Vries <tom@codesourcery.com>
1171 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1172 dump files): Add offload-tree.
1174 2018-05-03 Richard Biener <rguenther@suse.de>
1176 PR tree-optimization/85615
1177 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1178 to loops not nested in BBs loop father to avoid creating multi-entry
1181 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1183 PR tree-optimization/70291
1184 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1185 arguments. Change return type to tree. Emit libcall as a new
1186 statement rather than replacing existing one when inplace_p is true.
1187 (expand_complex_multiplication_components): New function.
1188 (expand_complex_multiplication): Expand floating-point complex
1189 multiplication using the above.
1190 (expand_complex_division): Rename inner_type parameter to type.
1191 Update expand_complex_libcall call-site.
1192 (expand_complex_operations_1): Update expand_complex_multiplication
1193 and expand_complex_division call-sites.
1195 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1198 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1199 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1200 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1201 the highest significant bit of the shift count mask is clear. In
1202 check whether and[sq]i3 is needed verify that all significant bits
1203 of the shift count other than the highest are set.
1205 2018-05-02 Tom de Vries <tom@codesourcery.com>
1208 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1209 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1210 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1211 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1212 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1213 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1214 __builtin_goacc_parlevel_size.
1216 2018-05-02 Richard Biener <rguenther@suse.de>
1218 PR tree-optimization/85597
1219 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1220 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1223 2018-05-02 Tom de Vries <tom@codesourcery.com>
1226 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1227 dump files): Add ltrans-tree.
1229 2018-05-02 Tom de Vries <tom@codesourcery.com>
1232 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1233 dump files): Add wpa-ipa.
1235 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1237 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1238 powerpc*-*-linux*paired* target.
1239 * config/rs6000/750cl.h: Delete.
1240 * config/rs6000/paired.h: Delete.
1241 * config/rs6000/paired.md: Delete.
1242 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1244 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1245 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1246 comment. Remove paired float support.
1247 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1248 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1250 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1251 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1253 * config/rs6000/rs6000.c: Remove paired float support.
1254 (paired_expand_vector_init, paired_expand_vector_move,
1255 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1256 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1257 paired_expand_builtin, paired_expand_predicate_builtin,
1258 paired_init_builtins): Delete.
1259 * config/rs6000/rs6000.h: Remove paired float support.
1260 * config/rs6000/rs6000.md: Remove paired float support.
1261 (move_from_CR_ov_bit): Delete.
1262 * config/rs6000/rs6000.opt (mpaired): Delete.
1263 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1264 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1266 2018-05-02 Richard Biener <rguenther@suse.de>
1269 * gimplify.c (gimplify_save_expr): When in SSA form allow
1270 SAVE_EXPRs to compute to SSA vars.
1272 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1275 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1276 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1277 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1278 clobber operands[2], instead use a new pseudo. Formatting fixes.
1280 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1282 PR tree-optimization/85586
1283 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1284 exit early for statements in the same group if the accesses are
1287 2018-05-02 Tom de Vries <tom@codesourcery.com>
1290 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1293 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1295 PR tree-optimization/85143
1296 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1298 2018-05-01 Tom de Vries <tom@codesourcery.com>
1301 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1302 not found" error message.
1304 2018-05-01 Tom de Vries <tom@codesourcery.com>
1307 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1308 * vec.c (test_ordered_remove_if): New function.
1309 (vec_c_tests): Call test_ordered_remove_if.
1310 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1311 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1312 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1313 VEC_ORDERED_REMOVE_IF.
1315 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1317 PR tree-optimization/82665
1318 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1319 pointer subtraction where arguments come from a memchr call.
1321 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1323 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1324 --push-state --as-needed and --pop-state instead of --as-needed and
1325 --no-as-needed if ld supports it.
1326 * configure: Regenerated.
1329 * doc/install.texi2html: Replace _002d with - and _002a with * in
1330 generated html files using sed.
1332 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1335 * gcc-rich-location.c (blank_line_before_p): New function.
1336 (use_new_line): New function.
1337 (gcc_rich_location::add_fixit_insert_formatted): New function.
1338 * gcc-rich-location.h
1339 (gcc_rich_location::add_fixit_insert_formatted): New function.
1341 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1343 * selftest.c (assert_streq): Rename "expected" and "actual" to
1344 "val1" and "val2". Extend NULL-handling to cover both inputs
1345 symmetrically, while still requiring both to be non-NULL for a pass.
1346 * selftest.h (assert_streq): Rename "expected" and "actual" to
1348 (ASSERT_EQ): Likewise.
1349 (ASSERT_EQ_AT): Likewise.
1350 (ASSERT_KNOWN_EQ): Likewise.
1351 (ASSERT_KNOWN_EQ_AT): Likewise.
1352 (ASSERT_NE): Likewise.
1353 (ASSERT_MAYBE_NE): Likewise.
1354 (ASSERT_MAYBE_NE_AT): Likewise.
1355 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1356 the assertion to pass.
1357 (ASSERT_STREQ_AT): Likewise.
1359 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1361 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1362 interaction with -pie.
1364 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1366 * selftest.h: Fix alphabetization of per-source-file selftest
1369 2018-04-30 Jason Merrill <jason@redhat.com>
1371 PR c++/61982 - dead stores to destroyed objects.
1372 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1375 2018-04-30 Jason Merrill <jason@redhat.com>
1377 * tree.c (build_clobber): New.
1378 * tree.h: Declare it.
1379 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1381 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1383 * diagnostic-show-locus.c (layout::layout): Update for
1384 location_get_source_line returning a char_span.
1385 (struct char_span): Move to input.h.
1386 (struct correction): Update for fields in char_span becoming
1388 (struct source_line): Update for location_get_source_line
1389 returning a char_span.
1390 (layout::print_line): Likewise.
1391 * edit-context.c (edited_file::print_content): Likewise.
1392 (edited_file::print_diff_hunk): Likewise.
1393 (edited_file::print_run_of_changed_lines): Likewise.
1394 (edited_file::get_num_lines): Likewise.
1395 (edited_line::edited_line): Likewise.
1396 * final.c (asm_show_source): Likewise.
1397 * input.c (location_get_source_line): Convert return type
1398 from const char * to char_span, losing the final "line_len"
1400 (dump_location_info): Update for the above.
1401 (get_substring_ranges_for_loc): Likewise. Use a char_span
1402 when handling the literal within the line.
1403 (test_reading_source_line): Update for location_get_source_line
1404 returning a char_span.
1405 * input.h (class char_span): Move here from
1406 diagnostic-show-locus.c, converting from a struct to a class.
1407 Make data members private.
1408 (char_span::operator bool): New.
1409 (char_span::length): New.
1410 (char_span::get_buffer): New.
1411 (char_span::operator[]): New.
1412 (char_span::subspan): Make const.
1413 (char_span::xstrdup): New.
1414 (location_get_source_line): Convert return type from const char *
1415 to char_span, losing the final "line_size" param.
1417 2018-04-30 Jan Hubicka <jh@suse.cz>
1419 * lto-wrapper.c (ltrans_priorities): New static var.
1420 (cmp_priority): New.
1421 (run_gcc): Read priorities and if doing parallel build order
1422 the Makefile by them.
1424 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1426 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1428 2018-04-30 Richard Biener <rguenther@suse.de>
1430 * tree-cfg.c (verify_address): Remove base argument, add
1431 flag whether to check TREE_ADDRESSABLE and do that.
1432 (verify_expr): Remove.
1433 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1434 (verify_gimple_assign_single): Likewise.
1435 (verify_gimple_switch): Likewise.
1436 (verify_expr_location_1): Dereference tp once. Add (disabled)
1437 piece from verify_expr.
1438 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1440 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1442 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1444 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1446 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1447 (small_data_pattern): Likewise.
1448 (arc_rewrite_small_data): Likewise.
1449 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1450 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1451 (get_symbol_alignment): New function.
1452 (legitimate_small_data_address_p): Likewise.
1453 (legitimate_scaled_address): Update, call
1454 legitimate_small_data_address_p.
1455 (output_sdata): New static variable.
1456 (arc_print_operand): Update how we handle small data operands.
1457 (arc_print_operand_address): Likewise.
1458 (arc_legitimate_address_p): Update, use
1459 legitimate_small_data_address_p.
1460 (arc_rewrite_small_data_p): Remove.
1461 (arc_rewrite_small_data_1): Likewise.
1462 (arc_rewrite_small_data): Likewise.
1463 (small_data_pattern): Likewise.
1464 (compact_sda_memory_operand): Update to use
1465 legitimate_small_data_address_p and get_symbol_alignment.
1466 (prepare_move_operands): Don't rewite sdata pattern.
1467 (prepare_extend_operands): Remove.
1468 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1470 (zero_extendqisi2): Likewise.
1471 (zero_extendhisi2): Likewise.
1472 (extendqihi2): Likewise.
1473 (extendqisi2): Likewise.
1474 (extendhisi2): Likewise.
1478 * config/arc/constraints.md (Usd): Change it to memory constraint.
1480 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1482 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1483 as source of std instructions.
1484 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1485 allow 6-bit constants as source for store instructions.
1486 (movdi_insn): Update instruction pattern to allow 6-bit constants
1487 as source for store instructions.
1489 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1491 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1493 2018-04-30 Nathan Sidwell <nathan@acm.org>
1494 Sandra Loosemore <sandra@codesourcery.com>
1496 * dumpfile.c (dump_open): Allow '-' for stdout.
1497 * doc/invoke.texi (Developer Options): Document dump filename
1498 determination early. Document stdin/stdout selection.
1500 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1502 Microblaze Target: PIC data text relative
1504 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1505 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1507 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1508 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1509 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1510 New addressing mode for data-text relative position indepenedent code.
1511 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1512 'ADDRESS_SYMBOLIC_TXT_REL'.
1513 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1514 (microblaze_legitimate_pic_operand): Exclude function calls from
1515 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1516 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1518 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1519 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1520 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1521 for 'address + offset'.
1522 (microblaze_expand_prologue): Add new function prologue call for
1524 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1525 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1526 table in case of TARGET_PIC_DATA_TEXT_REL.
1527 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1528 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1529 Add new macros 'UNSPEC_TEXT',
1530 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1531 + exclude function calls from 'UNSPEC_PLT' in case of data text
1533 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1534 new target hook for generating address diff vector tables in case of
1536 * doc/tm.texi : Regenerate.
1537 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1538 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1539 of addr diff vector generation.
1540 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1541 target hook definition.
1542 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1543 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1544 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1545 option for data text relative.
1547 2018-04-30 Richard Biener <rguenther@suse.de>
1549 * tree-chrec.h (evolution_function_is_constant_p): Remove
1551 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1553 2018-04-30 Richard Biener <rguenther@suse.de>
1556 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1558 2018-04-30 Richard Biener <rguenther@suse.de>
1560 PR tree-optimization/28364
1561 PR tree-optimization/85275
1562 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1563 copying first exit test.
1565 2018-04-28 Mark Wielaard <mark@klomp.org>
1567 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1569 (dwarf_AT): Handle DW_AT_addr_base.
1570 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1572 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1575 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1576 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1577 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1578 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1580 2018-04-28 Richard Biener <rguenther@suse.de>
1582 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1583 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1584 to reflect use. Only add interesting stmts.
1586 2018-04-27 Martin Jambor <mjambor@suse.cz>
1589 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1590 the jump function allows for passing through aggregate values.
1592 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1594 * input.h (in_system_header_at): Convert from macro to inline
1596 (from_macro_expansion_at): Likewise.
1597 (from_macro_definition_at): Likewise.
1599 2018-04-27 Jeff Law <law@redhat.com>
1601 * config.gcc: Mark tile* targets as deprecated/obsolete.
1603 2018-04-27 Richard Biener <rguenther@suse.de>
1605 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1608 2018-04-27 Richard Biener <rguenther@suse.de>
1610 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1612 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1614 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1615 with Yd constraint. Set "preferred_for_speed" attribute from
1616 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1618 (*movdi_internal): Ditto.
1619 (movti_interunit splitters): Remove
1620 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1621 (movdi_interunit splitters): Ditto.
1622 * config/i386/constraints.md (Ye): Remove.
1623 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1625 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1628 * config/aarch64/constraints.md (Usg): Limit to 31.
1631 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1633 PR tree-optimization/85529
1634 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1635 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1636 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1637 zero extension or masking of the MSB bit.
1638 (optimize_range_tests): Add FIRST_BB argument, pass it through
1639 to optimize_range_tests_var_bound.
1640 (maybe_optimize_range_tests, reassociate_bb): Adjust
1641 optimize_range_tests callers.
1643 2018-04-26 Richard Biener <rguenther@suse.de>
1644 Jakub Jelinek <jakub@redhat.com>
1646 * cgraph.h (symbol_table): Just declare debug method here.
1647 * symtab.c (symbol_table::debug): Define.
1649 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1651 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1653 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1655 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1656 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1657 (*movdi_internal): Substitute Yi and Yj constraint with x
1658 and Ym and Yn constraint with y constraint. Update "isa"
1659 attribute and set "preferred_for_speed" attribute from
1660 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1661 (*movsi_internal): Ditto.
1662 (*movdf_internal): Ditto.
1663 (*movsf_internal): Ditto.
1664 (*zero_extendsidi2): Ditto.
1665 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1666 (sse2_loadld): Ditto.
1667 (*vec_extract<ssevecmodelower>_0): Ditto.
1668 (*vec_extractv4si_0_zext_sse4): Ditto.
1669 (vec_concatv2di): Ditto.
1670 (*vec_dup<mode>): Ditto.
1671 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1672 * config/i386/constraints.md (Yi): Remove.
1677 2018-04-26 Nathan Sidwell <nathan@acm.org>
1679 * dumpfile.c (dump_open): New.
1680 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1681 (dump_finish): Detect stdio/stderr by value not name.
1683 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1685 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1687 2018-04-26 Tom de Vries <tom@codesourcery.com>
1690 * config/nvptx/nvptx.c (verify_neutering_jumps)
1691 (verify_neutering_labels): New function
1692 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1694 2018-04-26 Tom de Vries <tom@codesourcery.com>
1697 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1698 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1701 2018-04-26 Richard Biener <rguenther@suse.de>
1702 Tom de Vries <tom@codesourcery.com>
1705 * lto-streamer-out.c (output_function): Fixup loops if required to match
1706 discovery done in the reader.
1708 2018-04-26 Richard Biener <rguenther@suse.de>
1710 PR tree-optimization/85116
1711 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1712 have a loop exit from the single latch predecessor. Remove
1713 case of header with just condition.
1714 (ch_base::copy_headers): Exclude infinite loops from any
1716 (pass_ch::execute): Record exits.
1718 2018-04-26 Richard Biener <rguenther@suse.de>
1720 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1721 prologue cost vector and pass it to vect_get_load_cost.
1722 (vect_get_peeling_costs_all_drs): Likewise.
1723 (vect_peeling_hash_get_lowest_cost): Likewise.
1724 (vect_enhance_data_refs_alignment): Likewise.
1726 2018-04-26 Richard Biener <rguenther@suse.de>
1729 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1730 checking of integer<->pointer conversions.
1731 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1732 sign-/zero-extending pointer types.
1733 (expand_omp_for_static_chunk): Likewise.
1735 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1736 Jean Lee <xiaoyur347@gmail.com>
1738 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1739 (TARGET_ASAN_SHADOW_OFFSET): Define.
1740 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1741 true for -fsanitize=address.
1743 2018-04-25 Mark Wielaard <mark@klomp.org>
1745 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1748 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1750 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1751 than "alu", remove explicit "memory" and "imm_disp" attributes.
1752 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1755 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1756 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1757 gen_lowpart_no_emit.
1759 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1762 * config/i386/i386.c (ix86_expand_builtin): Change memory
1763 operand to XI, extend p0 to Pmode.
1764 * config/i386/i386.md: Change unspec volatile and operand
1765 1 mode to XI, change operand 0 mode to P.
1767 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1769 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1770 GET_MODE_MASK before any checking.
1771 (nds32_can_use_bset_p): Likewise.
1772 (nds32_can_use_btgl_p): Likewise.
1774 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1776 * config/nds32/nds32-doubleword.md: New define_split pattern for
1777 illegal register number.
1779 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1781 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1783 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1785 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1787 2018-04-25 Richard Biener <rguenther@suse.de>
1789 * lto-streamer.h (LTO_major_version): Bump to 8.
1791 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1793 * BASE-VER: Set to 9.0.0.
1795 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1797 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1798 in __abskf2 and __powikf2.
1800 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1803 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1804 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1805 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1806 Use the above on operand 2. Reindent.
1807 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1809 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1812 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1814 * config/i386/i386.opt (mcet): Removed.
1815 * doc/install.texi: Remove -mcet documentation.
1816 * doc/invoke.texi: Likewise.
1818 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1821 * doc/install.texi: Remove -mcet from bootstrap-cet.
1823 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1826 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1827 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1831 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1832 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1833 containing a CONST_VECTOR.
1835 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1837 * doc/install.texi: Update newlib dependency for nvptx.
1839 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1842 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1843 instead of INTVAL when shifting x left.
1845 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1847 PR tree-optimization/85478
1848 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1849 vect_grouped_store_supported for single element vectors.
1851 2018-04-24 Richard Biener <rguenther@suse.de>
1854 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1855 load cost increase to the case of non-constant step.
1857 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1860 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1861 destination if any_malformed_asm.
1863 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1866 * expr.c (store_field): In the bitfield case, if the value comes from
1867 a function call and is returned in registers by means of a PARALLEL,
1868 do not change the mode of the temporary unless BLKmode and VOIDmode.
1870 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1872 PR rtl-optimization/85423
1873 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1874 dependencies to debug insns when the previous insn is non-debug.
1876 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1878 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1879 enums into a single definition.
1880 (fls): Fix predicates and printing.
1883 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1885 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1886 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1887 and short u6 immediate.
1888 (check_if_valid_sleep_operand): Remove.
1889 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1891 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1893 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1894 flag_always_save_lp condition.
1895 * config/nds32/nds32.opt (malways-save-lp): New option.
1897 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1899 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1900 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1901 * config/nds32/nds32.h
1902 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1903 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1905 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1907 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1908 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1910 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1911 Chung-Ju Wu <jasonwucj@gmail.com>
1913 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1915 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1916 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1918 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1920 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1922 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1924 * config/nds32/nds32-protos.h (nds32_data_alignment,
1925 nds32_local_alignment): Declare.
1926 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1927 nds32_local_alignment): New functions.
1928 (TARGET_CONSTANT_ALIGNMENT): Define.
1929 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1931 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1933 * config/nds32/nds32.c
1934 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1935 (TARGET_MODES_TIEABLE_P): Likewise.
1937 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1939 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1942 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1943 Chung-Ju Wu <jasonwucj@gmail.com>
1945 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1946 for unaligned access.
1947 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1949 * config/nds32/nds32-intrinsic.md: Likewise.
1950 * config/nds32/nds32_intrinsic.h: Likewise.
1951 * config/nds32/nds32.h (nds32_builtins): Likewise.
1952 * config/nds32/nds32.opt (munaligned-access): New option.
1953 * config/nds32/nds32.c (nds32_asm_file_start): Display
1954 flag_unaligned_access status.
1956 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1958 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1959 -mno-relax is present.
1960 * config/riscv/linux.h (LINK_SPEC): Ditto.
1962 2018-04-20 Martin Sebor <msebor@redhat.com>
1965 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1967 (gimple_fold_builtin_stxcpy_chk): Same.
1968 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1970 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1973 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1974 __powikf2 when long double is IEEE 128-bit.
1976 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1978 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1979 step to make sure stack always aligned.
1981 2018-04-20 Carl Love <cel@us.ibm.com>
1984 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1985 size check for arg0.
1987 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1988 Tom de Vries <tom@codesourcery.com>
1991 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1992 Emit insns for calls too.
1993 (nvptx_find_par): Always look for worker-level predecessor insn.
1994 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1996 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1997 (nvptx_process_pars): Propagate frames for calls.
1999 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2002 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2004 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2005 (ix86_handle_option): Don't handle OPT_mibt.
2006 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2008 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2010 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2011 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2012 (ix86_target_macros): Define __CET__ with flag_cf_protection
2013 for -fcf-protection.
2014 * config/i386/i386.c (isa2_opts): Remove -mibt.
2015 * config/i386/i386.h (TARGET_IBT): Removed.
2016 (TARGET_IBT_P): Likewise.
2017 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2018 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2019 * config/i386/i386.opt (mcet): Update help message.
2022 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2023 -mcet as an alias for -mshstk.
2025 2018-04-20 Richard Biener <rguenther@suse.de>
2028 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2029 complexity by forcing a single use of the multiply operand.
2031 2018-04-20 Martin Jambor <mjambor@suse.cz>
2034 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2035 recursion dependency to only apply to non-clones.
2037 2018-04-20 Martin Jambor <mjambor@suse.cz>
2040 * ipa-cp.c (create_specialized_node): Check that clones of
2041 self-recursive edges exist during IPA-CP.
2043 2018-04-19 Toon Moene <toon@moene.org>
2045 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2048 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2050 PR tree-optimization/85467
2051 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2052 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2053 VECTOR_CST element to type.
2055 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2058 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2059 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2060 (builtin_longjmp): Likewise.
2061 (save_stack_nonlocal): New pattern.
2062 (restore_stack_nonlocal): Likewise.
2064 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2067 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2068 Replace ASM_OUTPUT_LABEL with fprintf.
2070 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2073 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2074 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2075 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2076 define __IBT__ and __SHSTK__ for -fcf-protection.
2077 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2079 (ix86_trampoline_init): Likewise.
2080 (x86_output_mi_thunk): Likewise.
2081 (ix86_notrack_prefixed_insn_p): Likewise.
2082 (ix86_option_override_internal): Don't disallow -fcf-protection.
2083 * config/i386/i386.md (rdssp<mode>): Also enable for
2085 (incssp<mode>): Likewise.
2086 (nop_endbr): Likewise.
2087 * config/i386/i386.opt (mcet): Change help message to built-in
2091 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2092 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2093 enable CET built-in functions.
2095 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2097 * common/config/i386/i386-common.c
2098 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2099 OPTION_MASK_ISA_MOVDIRI_UNSET,
2100 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2101 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2102 * config.gcc (movdirintrin.h): New header.
2103 * config/i386/cpuid.h (bit_MOVDIRI,
2104 bit_MOVDIR64B): New bits.
2105 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2107 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2108 (VOID, PVOID, PCVOID)): New function types.
2109 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2110 __builtin_ia32_directstoreu_u64,
2111 __builtin_ia32_movdir64b): New builtins.
2112 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2113 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2115 (ix86_valid_target_attribute_inner_p): Ditto.
2116 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2117 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2118 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2119 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2120 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2121 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2122 (movdiri<mode>, movdir64b_<mode>): New.
2123 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2124 * config/i386/immintrin.h: Include movdirintrin.h.
2125 * config/i386/movdirintrin.h: New file.
2126 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2128 2018-04-19 Richard Biener <rguenther@suse.de>
2131 * cfg.c (clear_bb_flags): When loop state says we have
2132 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2134 2018-04-19 Richard Biener <rguenther@suse.de>
2136 PR tree-optimization/84737
2137 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2138 copying restrict info.
2139 (vect_setup_realignment): Use it.
2140 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2141 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2142 the first DR to all generated stores.
2143 (vectorizable_load): Likewise for loads.
2145 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2147 PR tree-optimization/85446
2148 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2149 the integral and pointer types to have the same precision.
2151 * doc/install.texi: Document --disable-cet being the default and
2154 2018-04-18 Martin Liska <mliska@suse.cz>
2156 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2159 2018-04-18 Martin Liska <mliska@suse.cz>
2162 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2165 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2166 arguments if they are comparable.
2168 2018-04-18 Martin Liska <mliska@suse.cz>
2171 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2174 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2177 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2180 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2181 ENDBR after calling __morestack.
2183 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2186 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2187 by using gcc_base_ver to generate a gcc_driver_version, and use
2188 it when generating GCC_DRIVER_NAME.
2189 * configure: Regenerate.
2191 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2194 * config.gcc: Obsolete powerpc*-*-*spe*.
2196 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2199 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2200 (stabstr_D): Change type of unum from unsigned int to
2201 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2204 2018-04-17 Jim Wilson <jimw@sifive.com>
2207 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2208 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2209 Set arg_pointer_offset after using pretend_args_size.
2211 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2213 PR rtl-optimization/85431
2214 * dse.c (record_store): Ignore zero width stores.
2217 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2218 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2219 __builtin_stack_restore rather than after it.
2220 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2221 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2222 argument instead of virtual_dynamic_stack_rtx.
2224 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2226 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2228 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2229 Add note to error message to explain internal mapping of overloaded
2230 built-in function name to non-overloaded built-in function name.
2231 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2234 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2237 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2238 where the inputs overlap with the output.
2240 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2242 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2243 (=v, v) alternative and explicit "memory" attribute.
2244 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2245 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2247 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2248 "sselog1" type instead of "sselog".
2249 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2250 "sselog". Remove explicit "memory" attribute.
2251 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2252 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2254 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2255 "sselog1" type instead of "sselog". Remove explicit "memory"
2257 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2258 use "sselog1" type instead of "sselog". Remove explicit "memory"
2260 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2261 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2263 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2264 "sselog1" type instead of "sselog". Remove explicit "memory"
2266 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2267 use "sselog1" type instead of "sselog". Remove explicit "memory"
2271 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2274 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2277 2018-04-17 Martin Jambor <mjambor@suse.cz>
2280 * ipa-cp.c (create_specialized_node): Call
2281 expand_all_artificial_thunks if necessary.
2283 2018-04-17 Martin Liska <mliska@suse.cz>
2286 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2287 in message, remote space in between '_G' and '('.
2289 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2292 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2293 avx512f_vmcmp<mode>3<round_saeonly_name>,
2294 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2295 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2296 avx512f_rndscale<mode><round_saeonly_name>,
2297 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2298 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2299 Use %<iptr>2 instead of %2 for -masm=intel.
2300 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2301 avx512f_vcvttss2usi<round_saeonly_name>,
2302 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2304 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2305 avx512f_vcvttsd2usi<round_saeonly_name>,
2306 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2307 Use %q1 instead of %1 for -masm=intel.
2308 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2309 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2310 of %3 for -masm=intel.
2311 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2313 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2315 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2317 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2318 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2320 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2322 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2323 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2324 %g1 and one with %0 and %1.
2325 (avx512er_vmrcp28<mode><round_saeonly_name>,
2326 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2328 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2329 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2330 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2331 of %0 and %{%4%} for -masm=intel.
2332 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2333 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2334 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2335 order of %0 and %{%5%}%{z%} for -masm=intel.
2337 2018-04-17 Jan Hubicka <jh@suse.cz>
2340 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2342 2018-04-17 Martin Liska <mliska@suse.cz>
2345 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2346 for target_clone error message. Make default implementation
2347 clone to be a local declaration.
2348 (separate_attrs): Add new argument and check for an empty
2350 (expand_target_clones): Handle it.
2351 (ipa_target_clone): Make redirection just for target_clones
2354 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2355 Tom de Vries <tom@codesourcery.com>
2358 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2359 tiled basic blocks without omp continue statements.
2361 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2364 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2365 vec_extract expression as having side effects to make sure it gets
2368 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2371 * config/i386/i386.c (get_builtin_code_for_version): Check
2374 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2377 * gcc/config.gcc: Support "skylake".
2378 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2380 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2381 (processor_target_table): Add "skylake".
2382 (ix86_option_override_internal): Add "skylake".
2383 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2384 PROCESSOR_CANNONLAKE.
2385 (get_builtin_code_for_version): Fix priority for
2386 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2387 PROCESSOR_SKYLAKE-AVX512.
2388 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2389 (processor_type): Add PROCESSOR_SKYLAKE.
2391 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2392 Jason Merrill <jason@redhat.com>
2395 * convert.c (convert_to_integer_1): Use direct recursion for
2396 enumeral types and types with a precision less than the number
2397 of bits in their mode.
2399 2018-04-16 Julia Koval <julia.koval@intel.com>
2402 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2403 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2405 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2408 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2409 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2410 and -mno-direct-move.
2412 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2415 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2416 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2417 Ensure negative shifts result in {0}.
2419 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2421 PR rtl-optimization/79916
2422 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2423 regs (if any) to define how to gnerate SD moves when LRA is in
2426 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2428 PR rtl-optimization/85393
2429 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2430 * except.c (expand_dw2_landing_pad_for_region): Make static.
2431 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2432 a label and unconditional jump to old_bb, rather than
2433 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2436 PR rtl-optimization/85376
2437 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2438 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2439 instead of a specific value.
2441 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2442 Bin Cheng <bin.cheng@arm.com>
2444 PR tree-optimization/82965
2445 PR tree-optimization/83991
2446 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2447 by_profile_only parameter.
2448 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2449 information if the loop was predicted to iterate too many times.
2450 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2452 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2455 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2458 2018-04-13 Martin Liska <mliska@suse.cz>
2459 Jakub Jelinek <jakub@redhat.com>
2462 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2463 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2464 * builtins.c (expand_builtin_memory_copy_args): Use
2465 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2466 handle dest_addr == pc_rtx.
2468 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2471 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2472 asked to not generate direct moves.
2473 (fix_trunc<mode>si2_stfiwx): Similar.
2474 (fix_trunc<mode>si2_internal): Similar.
2476 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2479 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2480 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2481 lookup if dest in some wider mode is known to be const0_rtx and
2482 if so, record permanent equivalence for it to be ZERO_EXTEND of
2483 the narrower mode destination.
2485 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2487 * lto-streamer-out.c (output_function): Revert 259346.
2488 * omp-expand.c (expand_oacc_for): Likewise.
2490 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2492 PR rtl-optimization/85354
2493 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2494 * sel-sched.c (sel_global_init): ... here.
2496 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2499 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2500 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2501 mode for PE-COFF targets.
2502 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2503 (i386_pe_asm_lto_end): Likewise.
2504 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2505 (TARGET_ASM_LTO_END): Likewise.
2506 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2507 (i386_pe_asm_lto_start): New function.
2508 (i386_pe_asm_lto_end): Likewise.
2510 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2511 Richard Biener <rguenther@suse.de>
2514 * lto-streamer-out.c (output_function): Fix CFG loop state before
2516 * omp-expand.c (expand_oacc_for): Handle calls to internal
2517 functions like regular functions.
2519 2018-04-12 Richard Biener <rguenther@suse.de>
2522 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2523 for the early LTO debug to properly generate references to it
2524 during DIE emission. Do not re-use that for the skeleton for
2526 (dwarf2out_early_finish): Likewise.
2528 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2531 * config/i386/sse.md
2532 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2533 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2534 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2535 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2536 and output is a reg, avoid creating invalid lowpart subreg, but
2537 instead split into a 512-bit move. Don't split if not AVX512VL,
2538 input is xmm16+ reg and output is a mem.
2539 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2540 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2541 xmm16+ reg and output is a mem.
2543 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2545 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2546 also for flag_dwarf2_cfi_asm.
2548 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2550 PR rtl-optimization/85342
2551 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2552 a bool scalar var inside of the loop instead. Don't try to update
2553 recog_data.operand after failed apply_change_group.
2555 2018-04-12 Tom de Vries <tom@codesourcery.com>
2558 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2559 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2560 array with flexible array member as array without given dimension.
2561 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2562 argument for undefined param to true.
2564 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2567 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2568 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2569 from PowerPC section.
2570 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2571 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2572 help text that is too long.
2573 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2574 help text that is too long.
2575 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2576 help text that is too long.
2578 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2580 * config/alpha/alpha.md (stack_probe_internal): Rename
2581 from "probe_stack". Update all callers.
2583 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2585 PR rtl-optimization/84566
2586 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2587 sched_macro_fuse_insns.
2589 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2592 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2593 (compute_block_dependences): ... from here.
2595 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2597 PR tree-optimization/85331
2598 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2599 from int to HOST_WIDE_INT.
2601 2018-04-11 Martin Jambor <mjambor@suse.cz>
2604 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2605 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2606 not the same as the source val.
2607 (cgraph_edge_brings_value_p): New parameter.
2608 (gather_edges_for_value): Pass destination value to
2609 cgraph_edge_brings_value_p.
2610 (perhaps_add_new_callers): Likewise.
2611 (get_info_about_necessary_edges): Likewise and exclude values brought
2612 only by self-recursive edges.
2613 (create_specialized_node): Redirect only clones of self-calling edges.
2614 (+self_recursive_pass_through_p): New function.
2615 (find_more_scalar_values_for_callers_subset): Use it.
2616 (find_aggregate_values_for_callers_subset): Likewise.
2617 (known_aggs_to_agg_replacement_list): Removed.
2618 (decide_whether_version_node): Re-calculate known constants for all
2619 remaining context clones.
2621 2018-04-11 Richard Biener <rguenther@suse.de>
2624 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2625 from early DWARF output.
2626 (dwarf2out_early_finish): Output line info unconditionally into
2627 early DWARF and add reference to it.
2629 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2632 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2633 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2634 other than V2DFmode using iptr mode attribute.
2635 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2637 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2639 PR rtl-optimization/84659
2640 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2642 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2645 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2647 (output_loc_list): Pass address of a dummy size variable even in the
2648 locview handling loop.
2649 (index_location_lists): Add comment on why skip_loc_list_entry can't
2652 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2655 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2658 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2661 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2662 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2663 and -mstring-compare-inline-limit.
2665 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2668 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2669 for stack clash protection in a register whenever we need it to be in
2672 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2674 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2675 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2677 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2680 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2682 (mlong-double-): Ditto.
2683 * config/rs6000/sysv4.opt (msdata=): Ditto.
2684 (mtls-size=): Ditto.
2686 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2688 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2689 erroneous entries for
2690 "vector int vec_ldl (int, long int *)", and
2691 "vector unsigned int vec_ldl (int, unsigned long int *)".
2692 Add comments and entries for
2693 "vector bool char vec_ldl (int, bool char *)",
2694 "vector bool short vec_ldl (int, bool short *)",
2695 "vector bool int vec_ldl (int, bool int *)",
2696 "vector bool long long vec_ldl (int, bool long long *)",
2697 "vector pixel vec_ldl (int, pixel *)",
2698 "vector long long vec_ldl (int, long long *)",
2699 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2700 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2701 type tree bool_long_long_type_node and correct definition of
2702 bool_V2DI_type_node to make reference to this new type tree.
2703 (rs6000_mangle_type): Replace erroneous reference to
2704 bool_long_type_node with bool_long_long_type_node.
2705 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2706 comments to emphasize sign distinctions for char and int types and
2707 replace RS6000_BTI_bool_long constant with
2708 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2709 use of RS6000_BTI_pixel.
2710 (bool_long_type_node): Remove this macro definition.
2711 (bool_long_long_type_node): New macro definition
2713 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2715 PR rtl-optimization/85300
2716 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2717 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2718 simplify_unary_operation fails.
2720 2018-04-10 Martin Liska <mliska@suse.cz>
2722 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2723 cgraph_edge and ipa_ref.
2725 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2729 * config/i386/sse.md
2730 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2731 computation of the VEC_MERGE selector from mask.
2732 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2733 Fix decoding of the VEC_MERGE selector into mask.
2735 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2737 PR tree-optimization/85286
2738 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2740 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2742 * final.c (final_1): Set insn_last_address as well as
2743 insn_current_address.
2745 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2748 * explow.c (emit_stack_probe): Call validize_mem on memory location
2749 before passing it to gen_probe_stack. Create address operand and
2750 legitimize it for the probe_stack_address case.
2752 2018-04-09 Jan Hubicka <jh@suse.cz>
2755 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2756 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2757 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2759 (free_lang_data): Rebuild type inheritance graph.
2761 2018-04-09 Martin Sebor <msebor@redhat.com>
2763 * invoke.texi (-finline-small-functions): Mention other optimization
2765 (-findirect-inlining, -fpartial-inlining): Same.
2766 (-finline-functions-called-once): Same.
2767 (-freorder-blocks-and-partition): Same.
2769 2018-04-09 Jan Hubicka <jh@suse.cz>
2772 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2773 jumps; choose last target that matches the criteria (i.e.
2774 no partition changes for non-crossing jumps).
2775 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2776 support for redirecting crossing jumps to non-crossing.
2778 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2780 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2781 also for naked functions.
2783 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2785 * config/arc/arc.md (add_shift): New pattern.
2786 (add_shift2): Likewise.
2787 (sub_shift): Likewise.
2788 (sub_shift_cmp0_noout): Likewise.
2789 (compare_si_ashiftsi): Likewise.
2790 (xbfu_cmp0_noout): New combine pattern.
2791 (xbfu_cmp0"): Likewise.
2792 (movsi_set_cc_insn): Place the predicable variant first.
2793 (commutative_binary_cmp0_noout): Remove clobber.
2794 (commutative_binary_cmp0): New pattern.
2795 (noncommutative_binary_cmp0): Likewise.
2796 (noncommutative_binary_cmp0_noout): Likewise.
2797 (noncommutative_binary_comparison_result_used): Removed.
2798 (rsub_cmp0): New pattern.
2799 (rsub_cmp0_noout): Likewise.
2800 (extzvsi): Changed, keep only meaningful variants.
2801 (SQH, SEZ): New iterators.
2802 (SQH_postfix): New mode attribute.
2803 (SEZ_prefix): New code attribute.
2804 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2805 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2806 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2808 (noncommutative_operator): Check the availability of barrel
2811 2018-04-09 Richard Biener <rguenther@suse.de>
2813 PR tree-optimization/85284
2814 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2815 Only use the niter constraining form of simple_iv when the exit
2818 2018-04-09 Tom de Vries <tom@codesourcery.com>
2821 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2822 (define_expand "*memory_barrier"): New define_expand.
2823 (define_insn "memory_barrier"): New insn.
2825 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2827 PR rtl-optimization/80463
2828 PR rtl-optimization/83972
2829 PR rtl-optimization/83480
2831 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2832 correct producer for the insn.
2833 (tidy_control_flow): Fixup seqnos in case of debug insns.
2835 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2837 PR rtl-optimization/83913
2839 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2840 different sched-times when merging exprs.
2842 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2844 PR rtl-optimization/83962
2846 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2847 tidy_fallthru_edge and tidy_control_flow.
2849 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2851 PR rtl-optimization/83530
2853 * sel-sched.c (force_next_insn): New global variable.
2854 (remove_insn_for_debug): When force_next_insn is true, also leave only
2855 next insn in the ready list.
2856 (sel_sched_region): When the region wasn't scheduled, make another pass
2857 over it with force_next_insn set to 1.
2859 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2861 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2863 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2864 for interrupt control.
2865 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2866 functions for interrupt control.
2867 * config/nds32/nds32-intrinsic.md: Likewise.
2868 * config/nds32/nds32_intrinsic.h: Likewise.
2869 * config/nds32/nds32.h (nds32_builtins): Likewise.
2871 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2873 * config/nds32/nds32.c (nds32_init_machine_status,
2874 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2875 strict_aligned_p field.
2876 (nds32_expand_to_rtl_hook): New function.
2877 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2878 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2880 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2881 Chung-Ju Wu <jasonwucj@gmail.com>
2883 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2884 * config/nds32/nds32-n7.md: New file.
2885 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2886 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2888 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2889 * config/nds32/nds32.md (pipeline_model): Add n7.
2890 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2891 * config/nds32/pipelines.md: Include n7 settings.
2893 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2894 Chung-Ju Wu <jasonwucj@gmail.com>
2896 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2897 * config/nds32/nds32-e8.md: New file.
2898 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2899 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2901 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2902 * config/nds32/nds32.md (pipeline_model): Add e8.
2903 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2904 * config/nds32/pipelines.md: Include e8 settings.
2906 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2907 Chung-Ju Wu <jasonwucj@gmail.com>
2909 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2910 * config/nds32/nds32-n8.md: New file.
2911 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2912 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2914 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2915 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2916 * config/nds32/nds32.md (pipeline_model): Add n8.
2917 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2918 * config/nds32/pipelines.md: Include n8 settings.
2920 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2921 Chung-Ju Wu <jasonwucj@gmail.com>
2923 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2924 * config/nds32/nds32-n9-2r1w.md: New file.
2925 * config/nds32/nds32-n9-3r2w.md: New file.
2926 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2927 nds32_register_ports): New or modify for cpu n9.
2928 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2930 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2931 * config/nds32/nds32-utils.c: New file.
2932 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2933 TARGET_MUL_SLOW): Define.
2934 * config/nds32/nds32.md (pipeline_model): New attribute.
2935 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2936 New options that support cpu n9.
2937 * config/nds32/pipelines.md: Include n9 settings.
2938 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2940 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2942 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2943 information if necessary.
2944 (output_cond_branch_compare_zero): Likewise.
2945 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2946 (nds32_target_alignment): Refine for alignment.
2947 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2948 (FUNCTION_BOUNDARY): Modify.
2949 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2951 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2953 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2955 * config/nds32/constants.md (unspec_volatile_element): Add values for
2956 TLB operation and data prefetch.
2957 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2958 functions for TLB operation and data prefetch.
2959 * config/nds32/nds32-intrinsic.md: Likewise.
2960 * config/nds32/nds32_intrinsic.h: Likewise.
2961 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2962 (nds32_print_operand): Likewise.
2963 * config/nds32/nds32.h (nds32_builtins): Likewise.
2965 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2966 Andrew Pinski <pinsika@gcc.gnu.org>
2969 * match.pd: Use constant_boolean_node of correct type instead of
2970 boolean_true_node or boolean_false_node for simplifying
2971 pointer comparisons to zero.
2973 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2975 PR tree-optimization/80021
2976 * tree.c (verify_type_variant): Make error call in verify_variant_match
2977 translatable and remove final full stop.
2979 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2981 * config/nds32/constants.md (unspec_volatile_element): Add
2982 UNSPEC_VOLATILE_EH_RETURN.
2983 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2984 nds32_output_stack_pop): Support dwarf exception handling process.
2985 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2986 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2987 exception handling process.
2988 (nds32_compute_stack_frame): Likewise.
2989 (nds32_return_addr_rtx): Likewise.
2990 (nds32_initial_elimination_offset): Likewise.
2991 (nds32_expand_prologue): Likewise.
2992 (nds32_expand_epilogue): Likewise.
2993 (nds32_dynamic_chain_address): New function.
2994 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2996 (DYNAMIC_CHAIN_ADDRESS): Define.
2997 (EH_RETURN_DATA_REGNO): Define.
2998 (EH_RETURN_STACKADJ_RTX): Define.
2999 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3000 patterns for dwarf exception handling.
3002 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3004 * config/nds32/nds32.h: Clean up obsolete macros.
3006 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3008 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3009 Add enum values for particular instructions.
3010 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3011 particular intrinsic functions.
3012 * config/nds32/nds32-intrinsic.md: Likewise.
3013 * config/nds32/nds32_intrinsic.h: Likewise.
3014 * config/nds32/nds32.h (nds32_builtins): Likewise.
3015 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3016 (btst, ave): New patterns for particular instructions.
3018 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3020 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3021 Add enum values for atomic load/store and memory sync.
3022 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3024 * config/nds32/nds32-intrinsic.md: Likewise.
3025 * config/nds32/nds32_intrinsic.h: Likewise.
3026 * config/nds32/nds32.h (nds32_builtins): Likewise.
3028 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3030 PR tree-optimization/85257
3031 * fold-const.c (native_encode_vector): If not all elts could fit
3032 and off is -1, return 0 rather than offset.
3033 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3034 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3035 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3036 adjust buffer in native_interpret_expr call.
3038 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3040 * config/nds32/constants.md (unspec_volatile_element): Add cache
3041 control enum values.
3042 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3043 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3044 * config/nds32/nds32.c (nds32_cctl_names): New.
3045 (nds32_print_operand): Handle cache control register names.
3046 * config/nds32/nds32.h (nds32_builtins): New enum values.
3047 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3049 * config/nds32/nds32.md (type): Add mmu.
3050 * config/nds32/pipelines.md (simple_insn): Add mmu.
3052 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3054 * config/nds32/nds32.md (type): Remove call.
3055 * config/nds32/pipelines.md (simple_insn): Likewise.
3057 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3059 * config/nds32/constants.md (unspec_volatile_element): Add
3060 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3061 UNSPEC_VOLATILE_FMFCFG.
3062 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3063 description for fmfcfg and fmfcsr.
3064 (bdesc_1arg): Add fmtcsr.
3065 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3066 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3067 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3068 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3069 unspec_fmfcfg): New patterns.
3070 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3071 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3072 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3073 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3074 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3075 __nds32__fmfcfg): Define.
3077 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3079 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3080 intrinsic register names.
3081 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3082 intrinsic register enum values and macros.
3084 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3086 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3087 for load/store addressing form.
3088 (nds32_print_operand_address): Likewise.
3090 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3093 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3094 based on LABEL_REF. Remove useless assertion.
3095 (pic_address_needs_scratch): Fix formatting.
3096 (sparc_legitimize_pic_address): Minor tweaks.
3097 (sparc_delegitimize_address): Adjust assertion accordingly.
3098 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3099 into symbolic_operand.
3100 (movsi_high_pic_label_ref): Likewise.
3101 (movsi_lo_sum_pic_label_ref): Likewise.
3102 (movdi_pic_label_ref): Likewise.
3103 (movdi_high_pic_label_ref): Likewise.
3104 (movdi_lo_sum_pic_label_ref): Likewise.
3106 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3108 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3109 custom LIB_SPEC setup.
3111 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3112 Kito Cheng <kito.cheng@gmail.com>
3114 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3115 * config/riscv/freebsd.h: New.
3117 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3119 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3120 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3123 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3124 Kito Cheng <kito.cheng@gmail.com>
3126 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3127 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3128 * config/nds32/nds32-protos.h (nds32_output_call,
3129 nds32_output_return): Declare.
3130 * config/nds32/nds32.md: Refine all the call and return patterns.
3132 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3135 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3136 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3138 PR rtl-optimization/84872
3139 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3140 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3143 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3145 * expr.c (copy_blkmode_to_reg): Revert 254862.
3146 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3148 2018-04-06 Richard Biener <rguenther@suse.de>
3151 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3152 after seeing a component reference with an adjacent field. Treat
3153 refs to arrays at struct end of external decls similar to
3154 refs to unconstrained commons.
3156 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3159 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3160 look through SAVE_EXPRs with non-side-effects argument. Adjust
3162 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3165 2018-04-06 Richard Biener <rguenther@suse.de>
3168 * alias.c (find_base_term): New wrapper around find_base_term
3169 unwinding CSELIB_VAL_PTR changes.
3170 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3173 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3175 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3177 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3178 constant definitions.
3179 ("nop"): lr 0,0 -> nopr r0
3180 ("nop_lr0", "nop_lr1"): New insn definitions.
3182 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3184 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3185 NDS32_V3PUSH_AVAILABLE_P macro.
3187 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3188 Chung-Ju Wu <jasonwucj@gmail.com>
3190 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3191 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3192 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3193 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3194 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3195 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3196 * config/nds32/constraints.md: New constraints and checking for hard
3197 float configuration.
3198 * config/nds32/iterators.md: New mode iterator and attribute for hard
3199 float configuration.
3200 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3202 * config/nds32/nds32-fpu.md: New file.
3203 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3204 deal with hard float code generation.
3205 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3207 (abi_type, float_reg_number): New enum type.
3208 * config/nds32/nds32-predicates.c: New predicates for hard float.
3209 * config/nds32/nds32-protos.h: Declare functions for hard float.
3210 * config/nds32/nds32.c: Implementation for hard float configuration.
3211 * config/nds32/nds32.h: Definitions for hard float configuration.
3212 * config/nds32/nds32.md: Include hard float machine description and
3213 modify patterns for hard float configuration.
3214 * config/nds32/nds32.opt: New options for hard float configuration.
3215 * config/nds32/predicates.md: New predicates for hard float
3218 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3220 * common/config/nds32/nds32-common.c
3221 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3223 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3226 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3227 CONSTRUCTOR_ELT (ctor, ...)->value.
3229 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3232 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3234 2018-04-05 Tom de Vries <tom@codesourcery.com>
3237 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3240 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3241 Kito Cheng <kito.cheng@gmail.com>
3243 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3244 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3245 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3246 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3248 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3249 Kito Cheng <kito.cheng@gmail.com>
3251 * config/nds32/constraints.md (Ufe): New memory constraint.
3252 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3253 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3254 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3256 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3257 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3259 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3261 * config/nds32/nds32.md: Use optimize_size in the condition for
3262 alu-shift instructions.
3264 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3266 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3268 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3270 * config/nds32/nds32.md (negsi2): Refine pattern.
3272 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3273 Chung-Ju Wu <jasonwucj@gmail.com>
3275 * config/nds32/iterators.md (shift_rotate): New code iterator.
3276 (shift): New code attribute.
3277 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3278 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3279 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3280 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3281 bit-wise operations.
3282 (andsi3, *andsi3): Ditto.
3283 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3284 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3285 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3286 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3287 nds32_ior_operand, nds32_xor_operand): New predicates.
3289 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3291 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3292 (addsi3, subsi3): ... this.
3294 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3296 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3298 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3300 * config/nds32/nds32.md: Adjust indention.
3302 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3304 * config/nds32/nds32.md (feature): New attribute.
3306 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3308 * config/nds32/nds32.md (subtype): New attribute.
3310 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3313 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3314 expansion to perform a bitwise AND of the argument followed by a
3315 boolean negation of the result.
3317 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3319 PR rtl-optimization/84878
3320 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3321 the basic block. Assert the use reference is not artificial and that
3322 it has an associated insn.
3324 2018-04-04 Michael Matz <matz@suse.de>
3326 * builtins.c (compute_objsize): Pass correct operand
3327 to array_at_struct_end_p.
3329 2018-04-04 Richard Biener <rguenther@suse.de>
3332 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3333 from contexts for DINFO_LEVEL_TERSE and below.
3335 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3337 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3338 resiter_operand condition.
3339 * config/nds32/nds32.md (*move<mode>): Ditto.
3341 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3342 Monk Chiang <sh.chiang04@gmail.com>
3344 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3346 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3348 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3350 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3351 Kito Cheng <kito.cheng@gmail.com>
3353 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3354 nds32_cond_code_str, output_cond_branch,
3355 output_cond_branch_compare_zero, nds32_expand_cbranch,
3356 nds32_expand_cstore, nds32_expand_movcc,
3357 nds32_output_cbranchsi4_equality_zero,
3358 nds32_output_cbranchsi4_equality_reg,
3359 nds32_output_cbranchsi4_equality_reg_or_const_int,
3360 nds32_output_cbranchsi4_greater_less_zero: New functions.
3361 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3362 nds32_expand_cstore, nds32_expand_movcc,
3363 nds32_output_cbranchsi4_equality_zero,
3364 nds32_output_cbranchsi4_equality_reg,
3365 nds32_output_cbranchsi4_equality_reg_or_const_int,
3366 nds32_output_cbranchsi4_greater_less_zero): Declare.
3367 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3368 nds32_rimm11s_operand): New predicates.
3369 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3370 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3373 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3375 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3376 * config/nds32/nds32.md: Ditto.
3377 * config/nds32/pipelines.md: Ditto.
3379 2018-04-04 Richard Biener <rguenther@suse.de>
3381 PR tree-optimization/85168
3382 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3383 propagating abnormals.
3385 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3387 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3389 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3390 Kito Cheng <kito.cheng@gmail.com>
3392 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3393 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3394 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3395 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3396 * config/nds32/nds32.md (sibcall_internal): New.
3397 (sibcall_register): Remove.
3398 (sibcall_immediate): Remove.
3399 (sibcall_value_internal): New.
3400 (sibcall_value_register): Remove.
3401 (sibcall_value_immediate): Remove.
3402 * config/nds32/predicates.md (nds32_general_register_operand): New.
3403 (nds32_call_address_operand): New.
3405 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3407 PR rtl-optimization/85167
3408 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3409 bb_defs if *split_p, instead preinitialize it to NULL.
3411 PR tree-optimization/85156
3412 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3413 evaluating the argument multiple times.
3415 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3417 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3419 (_mm_cvtpd_ps): Likewise.
3420 (_mm_cvttpd_epi32): Likewise.
3421 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3422 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3423 vector, pixel, and bool following altivec.h include.
3425 2018-04-03 Martin Sebor <msebor@redhat.com>
3427 * doc/extend.texi (Common Function Attributes): Clarify.
3428 (const attribute): Likewise.
3429 (pure attribute): Likewise.
3431 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3434 * config/i386/i386.c (ix86_expand_vector_set): Use
3435 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3437 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3439 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3440 instructions when changing rounding bits to preserve precision bits
3441 in the x87 control word.
3443 2018-04-03 Martin Liska <mliska@suse.cz>
3445 PR tree-optimization/82491
3446 * rtl.h (strip_offset_and_add): Replace += suboffset with
3447 poly_uint64 () + suboffset.
3449 2018-03-29 Martin Liska <mliska@suse.cz>
3450 Martin Jambor <mjambor@suse.cz>
3453 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3454 param_type is not an integral or pointer type.
3456 2018-04-03 Richard Biener <rguenther@suse.de>
3458 * sese.h (recompute_all_dominators): Remove.
3460 2018-04-02 Martin Sebor <msebor@redhat.com>
3462 * doc/invoke.texi (-Wrestrict): Fix typos.
3464 2018-04-02 Jim Wilson <jimw@sifive.com>
3466 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3467 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3468 (<optab>di3, <optab>si3_extend): Likewise.
3469 (<optab>si3_mask, <optab>si3_mask_1): New.
3470 (<optab>di3_mask, <optab>di3_mask_1): New.
3471 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3472 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3473 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3475 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3477 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3480 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3482 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3483 (nds32_canonicalize_comparison): New function.
3485 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3486 Kito Cheng <kito.cheng@gmail.com>
3487 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3489 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3490 * config/nds32/constants.md (unspec_volatile_element): Add
3491 UNSPEC_VOLATILE_RELAX_GROUP.
3492 * config/nds32/nds32-relax-opt.c: New file.
3493 * config/nds32/nds32-predicates.c
3494 (nds32_symbol_load_store_p): New function.
3495 * config/nds32/nds32-protos.h
3496 (nds32_symbol_load_store_p): Declare function.
3497 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3498 * config/nds32/nds32.c
3499 (nds32_register_pass): New function to register pass.
3500 (nds32_register_passes): New function to register passes.
3501 * config/nds32/nds32.md (relax_group): New pattern.
3502 * config/nds32/nds32.opt (mrelax-hint): New option.
3503 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3505 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3507 * config/nds32/t-nds32: Modify files dependency.
3509 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3511 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3512 (PROFILE_HOOK): Define its implementation.
3514 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3516 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3517 type and 32-bit size.
3519 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3522 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3523 (V_128_256): New mode iterator.
3524 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3525 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3526 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3528 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3529 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3531 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3534 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3535 NaN inputs correctly.
3537 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3540 * config/rs6000/vsx.md (??r): New mode attribute.
3541 (*vsx_mov<mode>_64bit): Use it.
3542 (*vsx_mov<mode>_32bit): Likewise.
3544 2018-03-30 Martin Sebor <msebor@redhat.com>
3546 PR tree-optimization/84818
3547 * builtins.c (check_access): Use warning_n.
3549 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3552 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3554 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3557 2018-03-30 Julia Koval <julia.koval@intel.com>
3560 * x86-tune.def (movx, partial_reg_dependency): Enable for
3563 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3566 * lra-constraints.c (process_alt_operands): Move setting
3567 this_alternative_matches below.
3569 2018-03-29 Martin Liska <mliska@suse.cz>
3572 * doc/invoke.texi: Document how LTO works with debug info.
3573 Describe auto-load support of binutils. Mention 'x86-64'
3574 as valid option value of -march option.
3576 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3578 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3581 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3582 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3583 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3586 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3589 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3590 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3591 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3592 for RS6000_BTM_POWERPC64.
3593 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3594 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3595 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3600 2018-03-28 Carl Love <cel@us.ibm.com>
3603 2017-09-27 Carl Love <cel@us.ibm.com>
3605 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3606 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3607 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3610 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3612 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3613 instead of __vector bool.
3614 (_mm_max_pu8): Likewise.
3615 (_mm_min_pi16): Likewise.
3617 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3620 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3621 (DIVWEUO): Likewise.
3623 (DIVDEUO): Likewise.
3624 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3625 DIVWEUO and DIVDEUO.
3626 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3627 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3628 (div_extend): Likewise.
3629 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3631 (__builtin_divweuo): Likewise.
3632 (__builtin_divdeo): Likewise.
3633 (__builtin_divdeuo): Likewise.
3635 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3638 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3639 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3641 PR tree-optimization/82004
3642 * gimple-match-head.c (optimize_pow_to_exp): New function.
3643 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3644 Don't fold to exp if optimize_pow_to_exp is false.
3646 2018-03-28 Martin Liska <mliska@suse.cz>
3649 * calls.c (initialize_argument_information): Fix trailing space.
3650 * common.opt: Fix typo and provide better explanation for
3651 -fsanitize-coverage option.
3652 * config/i386/i386.opt: Fix typo.
3654 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3655 Martin Liska <mliska@suse.cz>
3658 * gimplify.c (asan_poison_variable): Don't do the check for
3659 gimplify_omp_ctxp here.
3660 (gimplify_decl_expr): Do it here.
3661 (gimplify_target_expr): Likewise.
3663 2018-03-28 Martin Liska <mliska@suse.cz>
3666 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3667 chkp_type_bounds_count if MPX is not enabled.
3669 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3671 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3673 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3676 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3677 function to create the function decl for complex long double
3678 multiply and divide for -mabi=ieeelongdouble.
3679 (init_float128_ieee): Call it.
3681 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3684 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3685 -fcf-protection=branch -mibt.
3686 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3688 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3691 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3693 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3696 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3697 extern array declarations.
3699 2018-03-27 Richard Biener <rguenther@suse.de>
3702 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3703 explicit single_use checks.
3705 2018-03-27 Richard Biener <rguenther@suse.de>
3707 PR tree-optimization/85082
3708 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3711 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3713 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3714 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3715 Turn on fasynchronous-unwind-tables and funwind-tables.
3717 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3720 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3721 (*bmi_blsr_<mode>_ccz): Ditto.
3723 2018-03-26 Tom de Vries <tom@codesourcery.com>
3725 PR tree-optimization/85063
3726 * omp-general.c (offloading_function_p): New function. Factor out
3728 * omp-offload.c (pass_omp_target_link::gate): ... here.
3729 * omp-general.h (offloading_function_p): Declare.
3730 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3731 with attribute omp declare target for offloading functions.
3733 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3735 PR tree-optimization/84005
3736 * tree-data-ref.h (get_base_for_alignment): Declare.
3737 * tree-data-ref.c (get_base_for_alignment_1): New function.
3738 (get_base_for_alignment): Likewise.
3739 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3740 get_base_for_alignment to find a suitable base object, instead
3741 of always using drb->base_address.
3743 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3746 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3747 known size by default.
3749 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3752 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3753 and non BLKmode operands.
3755 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3758 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3759 Clean up attributes.
3761 2018-03-23 Richard Biener <rguenther@suse.de>
3764 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3765 we are going to emit early debug for LTO.
3767 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3770 * function.c (match_asm_constraints_1): Don't optimize if input
3771 doesn't satisfy general_operand predicate for output's mode.
3774 * alias.c (write_dependence_p): Don't require for x_canonicalized
3775 non-VOIDmode if x has VOIDmode.
3778 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3779 just don't try to optimize it rather than assert it never happens.
3781 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3783 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3784 macro expansions for definition of ST_INTERNAL_<mode> and
3785 LD_INTERNAL_<mode> builtins.
3786 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3788 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3790 (altivec_expand_st_builtin): Likewise.
3791 (altivec_expand_builtin): Remove calls to deleted functions.
3792 (rs6000_address_for_altivec): Delete this function.
3793 * config/rs6000/vector.md: Remove expands for
3794 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3796 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3799 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3800 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3801 re-computing once computed.
3802 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3803 (arm_init_machine_status): Initialize
3804 machine->static_chain_stack_bytes.
3806 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3809 * doc/extend.texi: Add four new prototypes for vec_ld.
3810 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3811 definitions for more logical presentation.
3812 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3813 entries for V1TI variants of __builtin_altivec_ld builtin.
3814 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3815 handling of V1TI variant of LVX icode pattern.
3816 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3817 (rs6000_gimple_fold_builtin): Likewise.
3818 (altivec_init_builtins): Add code to define
3819 __builtin_altivec_lvx_v1ti function.
3821 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3824 * function.c (match_asm_constraints_1): Don't do the optimization
3825 if input isn't a REG, SUBREG, MEM or constant.
3827 2018-03-22 Tom de Vries <tom@codesourcery.com>
3829 PR tree-optimization/84956
3830 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3831 bb_has_abnormal_pred.
3833 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3836 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3837 DECL_INITIAL (decl) to decl at the end.
3838 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3841 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3843 * doc/extend.texi (__builtin_tgmath): Document when complex
3844 integer types are treated as _Complex _Float64.
3846 2018-03-21 Tom de Vries <tom@codesourcery.com>
3848 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3850 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3852 PR tree-optimization/84960
3853 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3854 if it is ENTRY block, move them into single succ of ENTRY in that case.
3856 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3858 PR tree-optimization/84811
3859 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3860 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3861 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3862 (known_subrange_p): Update accordingly. Cast each value involved
3863 in the size comparison, rather than casting the result of the
3866 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3868 PR tree-optimization/84982
3869 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3870 by flipping the least significant bit rather than all bits from
3871 bitpos to bitpos + bitsize - 1.
3873 2018-03-21 Nathan Sidwell <nathan@acm.org>
3875 * doc/extend.texi (Deprecated Features): Remove mention of
3876 long-deleted deprecations.
3878 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3881 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3882 * configure: Regenerate.
3884 2018-03-21 Tom de Vries <tom@codesourcery.com>
3886 PR tree-optimization/83126
3887 * tree-parloops.c (num_phis): New function.
3888 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3890 2018-03-21 Nathan Sidwell <nathan@acm.org>
3892 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3893 mention anon-struct/union members and trailing attributes.
3895 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3897 PR tree-optimization/84969
3898 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3899 builtin memset partitions if they set different rhs values.
3901 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3903 PR rtl-optimization/84989
3904 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3905 VEC_DUPLICATE with scalar result mode.
3907 2018-03-21 Martin Liska <mliska@suse.cz>
3910 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3911 not intended return statement.
3913 2018-03-21 Martin Liska <mliska@suse.cz>
3916 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3917 (chkp_find_bound_slots_1): Limit number of iterations.
3919 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3922 * Minor grammar fixes for x86 options.
3924 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3927 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3928 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3930 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3933 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3934 (altivec_lvx_<mode>_1op): Likewise.
3935 (altivec_stvx_<mode>_2op): Likewise.
3936 (altivec_stvx_<mode>_1op): Likewise.
3937 (altivec_lvx_<VM2:mode>): New define_expand.
3938 (altivec_stvx_<VM2:mode>): Likewise.
3939 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3940 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3941 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3942 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3943 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3944 (rs6000_gen_lvx): Likewise.
3945 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3946 (altivec_expand_stv_builtin): Likewise.
3947 (altivec_expand_builtin): Likewise.
3948 * config/rs6000/vector.md: Likewise.
3950 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3953 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3956 2018-03-20 Richard Biener <rguenther@suse.de>
3959 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3960 sign-conversions as zero, fall back to standard scalar_stmt
3963 2018-03-20 Martin Liska <mliska@suse.cz>
3966 * predict.c (rebuild_frequencies): Handle case when we have
3967 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3969 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3972 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3973 flag_section_anchors.
3974 * varasm.c (use_blocks_for_decl_p): Remove hack for
3975 dw2_force_const_mem.
3978 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3980 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3981 be created, use lowpart_subreg of operands[0] rather than operands[0]
3983 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3984 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3985 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3986 and n constraint instead of aarch64_shift_imm_di and Usd.
3987 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3988 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3990 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3993 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3994 to favor GPR over NEON registers.
3995 (<shift>di3_neon): Likewise.
3997 2018-03-20 Tom de Vries <tom@codesourcery.com>
4000 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4001 (nvptx_process_pars): Emit bar.sync asap and alap.
4003 2018-03-20 Tom de Vries <tom@codesourcery.com>
4006 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4007 seen_label if seen_label is already set.
4009 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4012 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4013 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4014 Use 1U instead of 1. Formatting fixes.
4017 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4018 instead of TREE_TYPE (s1) for the return value.
4020 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4022 PR tree-optimization/84946
4023 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4024 bitsize + bitsize in poly_uint64 rather than poly_int64.
4027 * dwarf2asm.c: Include fold-const.c.
4028 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4029 of decl rather than decl itself.
4031 PR rtl-optimization/84643
4032 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4034 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4037 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4038 calling assemble_variable.
4040 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4043 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4044 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4046 2018-03-19 Jim Wilson <jimw@sifive.com>
4049 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4050 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4051 (riscv_first_stack_step): Likewise.
4052 (riscv_option_override): Use STACK_BOUNDARY instead of
4054 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4056 (BIGGEST_ALIGNMENT): Set to 128.
4057 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4058 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4061 2018-03-19 Richard Biener <rguenther@suse.de>
4063 PR tree-optimization/84933
4064 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4065 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4067 2018-03-19 Richard Biener <rguenther@suse.de>
4069 PR tree-optimization/84859
4070 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4071 (cond_if_else_store_replacement): Perform sinking operation on
4072 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4073 Generalize what a BB with a single eligible store is.
4075 2018-03-19 Richard Biener <rguenther@suse.de>
4077 PR tree-optimization/84929
4078 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4079 chrec_is_positive against non-chrec arg.
4081 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4084 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4086 2018-03-18 Martin Liska <mliska@suse.cz>
4088 PR rtl-optimization/84635
4089 * regrename.c (build_def_use): Use matches_mode only when
4092 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4094 PR tree-optimization/84913
4095 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4096 vectorize chains of COND_EXPRs.
4098 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4100 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4102 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4104 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4106 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4108 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4110 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4111 Kito Cheng <kito.cheng@gmail.com>
4113 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4114 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4115 (nds32_adjust_reg_alloc_order): New function.
4116 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4118 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4120 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4121 nds32_print_operand, nds32_print_operand_address): Use
4122 HOST_WIDE_INT_PRINT_DEC instead.
4124 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4126 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4128 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4131 * config/i386/i386.c (initial_ix86_tune_features,
4132 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4134 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4135 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4136 rather than 1u << ix86_tune. Formatting fix.
4137 (ix86_option_override_internal): Change ix86_arch_mask from
4138 unsigned int to unsigned HOST_WIDE_INT, initialize to
4139 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4140 (ix86_function_specific_restore): Likewise.
4142 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4145 * postreload.c (reload_combine_recognize_pattern): Perform
4146 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4147 truncate_int_for_mode the result for the destination's mode.
4150 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4151 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4153 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4156 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4159 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4160 regno_allocno_class_array and sorted_pseudos.
4161 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4162 insns where regno is used.
4164 2018-03-16 Martin Liska <mliska@suse.cz>
4167 * multiple_target.c (create_dispatcher_calls): Redirect
4168 reference in the symbol table.
4170 2018-03-16 Martin Liska <mliska@suse.cz>
4173 * multiple_target.c (create_dispatcher_calls): Redirect also
4176 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4180 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4181 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4182 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4184 2018-03-16 Julia Koval <julia.koval@intel.com>
4186 * doc/invoke.texi (Skylake Server): Add CLWB.
4187 Cannonlake): Remove CLWB.
4189 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4191 PR tree-optimization/84841
4192 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4194 (FLOAT_ONE_CONST_TYPE): Define.
4195 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4196 (sort_by_operand_rank): Put entries with higher constant_type last
4197 rather than first to match comments.
4199 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4201 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4204 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4207 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4208 check and use address of target always.
4210 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4213 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4214 (indirect_thunk_bnd_needed): Likewise.
4215 (indirect_thunks_used): Likewise.
4216 (indirect_thunks_bnd_used): Likewise.
4217 (indirect_return_needed): New.
4218 (indirect_return_bnd_needed): Likewise.
4219 (output_indirect_thunk_function): Add a bool argument for
4221 (output_indirect_thunk_function): Don't generate alias for
4222 function return thunk.
4223 (ix86_code_end): Call output_indirect_thunk_function to generate
4224 function return thunks.
4225 (ix86_output_function_return): Set indirect_return_bnd_needed
4226 and indirect_return_needed instead of indirect_thunk_bnd_needed
4227 and indirect_thunk_needed.
4229 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4231 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4232 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4233 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4235 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4236 Paul Hua <paul.hua.gm@gmail.com>
4239 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4241 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4243 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4244 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4247 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4250 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4251 instead of GET_MODE_SIZE when comparing Units.
4253 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4256 * varasm.c (hash_section): Return an unchangeble hash value
4257 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4258 Return !aarch64_can_use_per_function_literal_pools_p ().
4260 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4263 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4264 rather than address of cmode as last argument to prepare_cmp_insn.
4266 2018-03-15 Julia Koval <julia.koval@intel.com>
4268 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4269 F_AVX512VNNI, F_AVX512BITALG): New.
4271 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4274 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4275 insn for floating-point loads and stores.
4277 2018-03-14 Carl Love <cel@us.ibm.com>
4279 * config/rs6000/rs6000-c.c: Add macro definitions for
4280 ALTIVEC_BUILTIN_VEC_PERMXOR.
4281 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4282 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4283 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4284 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4286 * config/doc/extend.texi: Add prototypes for vec_permxor.
4288 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4291 * diagnostic-show-locus.c (class layout_point): Convert m_line
4292 from int to linenum_type.
4293 (line_span::comparator): Use linenum "compare" function when
4294 comparing line numbers.
4295 (test_line_span): New function.
4296 (layout_range::contains_point): Convert param "row" from int to
4298 (layout_range::intersects_line_p): Likewise.
4299 (layout::will_show_line_p): Likewise.
4300 (layout::print_source_line): Likewise.
4301 (layout::should_print_annotation_line_p): Likewise.
4302 (layout::print_annotation_line): Likewise.
4303 (layout::print_leading_fixits): Likewise.
4304 (layout::annotation_line_showed_range_p): Likewise.
4305 (struct line_corrections): Likewise for field m_row.
4306 (line_corrections::line_corrections): Likewise for param "row".
4307 (layout::print_trailing_fixits): Likewise.
4308 (layout::get_state_at_point): Likewise.
4309 (layout::get_x_bound_for_row): Likewise.
4310 (layout::print_line): Likewise.
4311 (diagnostic_show_locus): Likewise for locals "last_line" and
4313 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4314 * input.c (selftest::test_linenum_comparisons): New function.
4315 (selftest::input_c_tests): Call it.
4316 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4317 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4318 * selftest.h (ASSERT_GT): New macro.
4319 (ASSERT_GT_AT): New macro.
4320 (ASSERT_LT): New macro.
4321 (ASSERT_LT_AT): New macro.
4323 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4325 PR rtl-optimization/84780
4326 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4328 2018-03-14 Martin Liska <mliska@suse.cz>
4330 * tree.c (record_node_allocation_statistics): Use
4331 get_stats_node_kind.
4332 (get_stats_node_kind): New function extracted from
4333 record_node_allocation_statistics.
4334 (free_node): Use get_stats_node_kind.
4336 2018-03-14 Richard Biener <rguenther@suse.de>
4338 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4339 that the value-set of ANTIC_IN doesn't grow.
4342 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4344 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4345 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4347 2018-03-14 Julia Koval <julia.koval@intel.com>
4349 * config.gcc (icelake-client, icelake-server): New.
4351 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4352 (initial_ix86_arch_features): Ditto.
4353 (PTA_SKYLAKE): Add SGX.
4354 (PTA_ICELAKE): Remove.
4355 (PTA_ICELAKE_CLIENT): New.
4356 (PTA_ICELAKE_SERVER): New.
4357 (ix86_option_override_internal): Split up icelake on icelake client and
4359 (get_builtin_code_for_version): Ditto.
4360 (fold_builtin_cpu): Ditto.
4361 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4362 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4363 * config/i386/i386.h (processor_type): Ditto.
4364 * doc/invoke.texi: Ditto.
4366 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4369 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4370 INTEGER_CST offset, add it together with bitpos / 8 and
4371 sign extend based on POINTER_SIZE.
4375 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4378 * config/i386/constraints.md (Yc): New register constraint.
4379 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4380 Use Yc constraint for alternative 2 of operand 0. Remove
4381 preferred_for_speed attribute.
4383 2018-03-14 Richard Biener <rguenther@suse.de>
4385 PR tree-optimization/84830
4386 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4387 with the old one to avoid oscillations.
4389 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4392 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4394 (assign_by_spills): Return a flag of reload assignment failure.
4395 Do not process the reload assignment failures. Do not spill other
4396 reload pseudos if they has the same reg class. Update n if
4398 (lra_assign): Add a return arg. Set up from the result of
4399 assign_by_spills call.
4400 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4401 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4402 usage_insns if it is not NULL.
4403 (spill_hard_reg_in_range): New function.
4404 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4405 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4406 function prototypes.
4407 (lra_assign): Change prototype.
4408 * lra.c (lra): Add code to deal with fails by splitting hard reg
4411 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4413 * config/riscv/riscv.opt (mrelax): New option.
4414 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4415 "norelax" when riscv_mrelax is disabled.
4416 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4418 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4421 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4422 reassociation for int modes.
4424 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4426 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4427 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4429 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4430 * config/aarch64/aarch64-sve.md
4431 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4432 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4433 (*extend<mode><Vwide>2): Rename to...
4434 (aarch64_sve_extend<mode><Vwide>2): ...this.
4435 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4436 renaming the old pattern to...
4437 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4439 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4440 define_expand, renaming the old pattern to...
4441 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4442 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4443 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4444 account when deciding which SVE instruction the optab should use.
4445 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4447 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4449 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4450 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4451 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4452 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4453 (tlsdesc_small_<mode>): Turn a define_expand and use
4454 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4455 (tlsdesc_small_advsimd_<mode>): ...this.
4456 (tlsdesc_small_sve_<mode>): New pattern.
4458 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4460 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4461 (UNSPEC_UMUL_HIGHPART): New constants.
4462 (MUL_HIGHPART): New int iteraor.
4463 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4464 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4466 (*<su>mul<mode>3_highpart): New define_insn.
4468 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4471 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4474 2018-03-13 Martin Liska <mliska@suse.cz>
4477 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4479 (sem_item_optimizer::~sem_item_optimizer): Release it.
4480 (sem_item_optimizer::merge_classes): Register variable aliases.
4481 (sem_item_optimizer::fixup_pt_set): New function.
4482 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4483 * ipa-icf.h: Declare new variables and functions.
4485 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4488 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4489 integer_pow2p@2 and test integer_pow2p in condition.
4490 (A < 0 ? C : 0): Similarly for @1.
4493 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4494 characters starting at p contain '\0' character, don't look beyond
4498 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4499 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4502 * reg-stack.c (change_stack): Change update_end var from int to
4503 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4504 also call set_block_for_insn on the newly added insns and rescan.
4507 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4508 on the last operand.
4511 * tree.c (stabilize_reference_1): Return save_expr (e) for
4512 STATEMENT_LIST even if it doesn't have side-effects.
4514 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4516 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4518 2018-03-12 Renlin Li <renlin.li@arm.com>
4520 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4521 aarch64_output_scalar_simd_mov_immediate.
4523 2018-03-12 Martin Sebor <msebor@redhat.com>
4525 PR tree-optimization/83456
4526 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4527 for perfectly overlapping calls to memcpy.
4528 (gimple_fold_builtin_memory_chk): Same.
4529 (gimple_fold_builtin_strcpy): Handle no-warning.
4530 (gimple_fold_builtin_stxcpy_chk): Same.
4531 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4533 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4535 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4536 parameter. Use it for SFmode.
4537 (rs6000_function_arg_advance_1): Adjust.
4538 (rs6000_function_arg): Adjust.
4539 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4541 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4543 PR rtl-optimization/84169
4544 PR rtl-optimization/84780
4545 * combine.c (can_combine_p): Check for a 2-insn combination whether
4546 the destination register is used between the two insns, too.
4548 2018-03-12 Richard Biener <rguenther@suse.de>
4550 PR tree-optimization/84803
4551 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4552 for refs DR analysis didn't process.
4554 2018-03-12 Richard Biener <rguenther@suse.de>
4556 PR tree-optimization/84777
4557 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4558 force-vectorize loops ignore whether we are optimizing for size.
4560 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4562 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4563 (TARGET_MD_ASM_ADJUST): Define.
4565 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4566 Kito Cheng <kito.cheng@gmail.com>
4567 Chung-Ju Wu <jasonwucj@gmail.com>
4569 * config/nds32/nds32.c (nds32_compute_stack_frame,
4570 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4571 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4572 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4573 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4574 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4575 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4576 * config/nds32/nds32.md (prologue, epilogue): Use macro
4577 NDS32_V3PUSH_AVAILABLE_P to do checking.
4579 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4582 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4583 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4584 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4585 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4586 addition of most attributes on !orig_type_die or the attribute not
4587 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4589 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4590 Chung-Ju Wu <jasonwucj@gmail.com>
4592 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4594 * config/nds32/nds32.opt (mvh): New option.
4596 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4597 Chung-Ju Wu <jasonwucj@gmail.com>
4599 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4601 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4602 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4605 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4606 Chung-Ju Wu <jasonwucj@gmail.com>
4608 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4610 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4611 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4613 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4614 Kito Cheng <kito.cheng@gmail.com>
4615 Chung-Ju Wu <jasonwucj@gmail.com>
4617 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4618 UNSPEC_FFMISM and UNSPEC_FLMISM.
4619 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4620 for ffb, ffmism and flmism.
4621 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4622 (unspec_ffmism): Ditto.
4623 (unspec_flmism): Ditto.
4624 (nds32_expand_builtin_impl): Check if string extension is available.
4625 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4626 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4628 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4631 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4634 * lra-assigns.c (assign_by_spills): Return a flag of reload
4635 assignment failure. Do not process the reload assignment
4636 failures. Do not spill other reload pseudos if they has the same
4638 (lra_assign): Add a return arg. Set up from the result of
4639 assign_by_spills call.
4640 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4641 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4642 usage_insns if it is not NULL.
4643 (spill_hard_reg_in_range): New function.
4644 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4645 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4646 function prototypes.
4647 (lra_assign): Change prototype.
4648 * lra.c (lra): Add code to deal with fails by splitting hard reg
4651 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4654 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4656 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4659 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4660 (dw_val_node): Add val_symbolic_view.
4661 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4662 (symview_upper_bound): New.
4663 (new_line_info_table): Initialize symviews_since_reset.
4664 (dwarf2out_source_line): Count symviews_since_reset and set
4665 symview_upper_bound.
4666 (dw_val_equal_p): Handle symview.
4667 (add_AT_symview): New.
4668 (print_dw_val): Handle symview.
4669 (attr_checksum, attr_checksum_ordered): Likewise.
4670 (same_dw_val_p, size_of_die): Likewise.
4671 (value_format, output_die): Likewise.
4672 (add_high_low_attributes): Use add_AT_symview for entry_view.
4673 (dwarf2out_finish): Reset symview_upper_bound, clear
4676 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4679 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4680 Add strict argument and use it.
4681 (rs6000_split_multireg_move): Update for new strict argument.
4682 (mem_operand_gpr): Disallow all non-offsettable addresses.
4683 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4685 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4688 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4689 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4690 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4693 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4694 decl, use remap_type if we want to use the type.
4696 2018-03-09 Martin Sebor <msebor@redhat.com>
4698 PR tree-optimization/84526
4699 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4701 (builtin_access::generic_overlap): Be prepared to handle non-array
4704 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4706 PR rtl-optimization/84682
4707 * lra-constraints.c (process_address_1): Check is_address flag
4708 for address constraints.
4709 (process_alt_operands): Likewise.
4710 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4711 preprocess_constraints.
4712 * recog.h (preprocess_constraints): Add oploc parameter.
4714 * recog.c (preprocess_constraints): Test address_operand for
4715 CT_ADDRESS constraints.
4717 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4720 * lra-assigns.c (assign_by_spills): Return a flag of reload
4721 assignment failure. Do not process the reload assignment
4722 failures. Do not spill other reload pseudos if they has the same
4724 (lra_assign): Add a return arg. Set up from the result of
4725 assign_by_spills call.
4726 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4727 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4728 usage_insns if it is not NULL.
4729 (spill_hard_reg_in_range): New function.
4730 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4731 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4732 function prototypes.
4733 (lra_assign): Change prototype.
4734 * lra.c (lra): Add code to deal with fails by splitting hard reg
4737 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4740 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4741 Accept complain bool parameter. Only emit errors if it is true.
4742 (arm_parse_cpu_option_name): Likewise.
4743 (arm_target_thumb_only): Adjust callers of the above.
4744 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4745 prototype to take a default true bool parameter.
4746 (arm_parse_arch_option_name): Likewise.
4748 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4749 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4753 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4754 * configure: Regenerate.
4755 * configure.ac ("linker --version-script option"): New.
4756 ("linker soname option"): New.
4758 2018-03-09 Richard Biener <rguenther@suse.de>
4760 PR tree-optimization/84775
4761 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4762 immediate uses of predicate stmts and mark them modified.
4765 PR tree-optimization/84178
4766 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4768 (version_loop_for_if_conversion): Delay update_ssa call.
4769 (tree_if_conversion): Delay update_ssa until after predicate
4772 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4775 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4776 when the function accesses prior frames.
4778 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4781 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4782 gen_llsym, otherwise call maybe_gen_llsym.
4785 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4786 has ',' character inside of it.
4788 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4791 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4792 as clobbering CC_REGNUM.
4794 2018-03-08 Richard Biener <rguenther@suse.de>
4797 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4798 (follow_copies_to_constant): Do not follow SSA names registered
4801 2018-03-08 Richard Biener <rguenther@suse.de>
4803 PR tree-optimization/84178
4804 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4806 (version_loop_for_if_conversion): Delay update_ssa call.
4807 (tree_if_conversion): Delay update_ssa until after predicate
4810 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4812 PR tree-optimization/84178
4813 * tree-if-conv.c (release_bb_predicate): Remove the
4814 the assertion that the stmts have NULL use_ops.
4815 Discard the statements, asserting that they haven't
4816 yet been added to a BB.
4818 2018-03-08 Richard Biener <rguenther@suse.de>
4820 PR tree-optimization/84746
4821 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4822 (phi_translate): Pass in destination ANTIC_OUT set.
4823 (phi_translate_1): Likewise. For a simplified result lookup
4824 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4825 (phi_translate_set): Adjust.
4826 (do_pre_regular_insertion): Likewise.
4827 (do_pre_partial_partial_insertion): Likewise.
4829 2018-03-08 Martin Liska <mliska@suse.cz>
4831 PR gcov-profile/84735
4832 * doc/gcov.texi: Document usage of profile files.
4833 * gcov-io.h: Document changes in the format.
4835 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4839 * dwarf2out.c (struct dw_line_info_table): Update comments for
4841 (FORCE_RESET_NEXT_VIEW): New.
4842 (FORCE_RESETTING_VIEW_P): New.
4843 (RESETTING_VIEW_P): Check for -1 too.
4844 (ZERO_VIEW_P): Likewise.
4845 (new_line_info_table): Force-reset next view.
4846 (dwarf2out_begin_function): Likewise.
4847 (dwarf2out_source_line): Simplify zero_view_p initialization.
4848 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4849 view directly. Omit view when omitting .loc at line 0.
4851 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4853 PR tree-optimization/84740
4854 * tree-switch-conversion.c (process_switch): Call build_constructors
4855 only if info.phi_count is non-zero.
4857 PR tree-optimization/84739
4858 * tree-tailcall.c (find_tail_calls): Check call arguments against
4859 DECL_ARGUMENTS (current_function_decl) rather than
4860 DECL_ARGUMENTS (func) when checking for tail recursion.
4862 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4864 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4865 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4866 Volker Reichelt's entry and add entries for people that perform
4867 GCC fuzzy testing and report numerous bugs.
4869 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4872 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4873 readonly data in sdata, if that is disabled.
4874 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4875 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4876 -mreadonly-in-sdata option.
4878 2018-03-07 Martin Sebor <msebor@redhat.com>
4880 PR tree-optimization/84468
4881 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4882 basic block when looking for nul assignment.
4884 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4887 * except.h (output_function_exception_table): Adjust prototype.
4888 * except.c (output_function_exception_table): Remove FNNAME parameter
4889 and add SECTION parameter. Ouput one part of the table at a time.
4890 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4891 the first part of the exception table and emit unwind directives.
4892 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4893 (i386_pe_seh_cold_init): Likewise.
4894 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4895 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4896 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4897 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4898 * config/i386/winnt.c: Include except.h.
4899 (struct seh_frame_state): Add reg_offset, after_prologue and
4900 in_cold_section fields.
4901 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4902 (i386_pe_seh_cold_init): New function.
4903 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4904 to seh->in_cold_section.
4905 (seh_emit_push): Record the offset of the push.
4906 (seh_emit_save): Record the offet of the save.
4907 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4908 Test seh->after_prologue to disregard the epilogue.
4909 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4910 (i386_pe_end_cold_function): New function.
4912 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4915 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4916 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4919 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4921 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4922 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4925 * multiple_target.c: Include tree-inline.h and intl.h.
4926 (expand_target_clones): Diagnose and fail if node->definition and
4927 !tree_versionable_function_p (node->decl).
4929 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4931 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4933 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4934 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4935 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4937 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4940 * combine.c (try_combine): Use reg_or_subregno instead of handling
4941 just paradoxical SUBREGs and REGs.
4943 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4945 * config/arc/arc.c (arc_finalize_pic): Remove function.
4946 (arc_must_save_register): We use single base PIC register, remove
4947 checks to save/restore the PIC register.
4948 (arc_expand_prologue): Likewise.
4949 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4951 (arc_verify_short): Likewise.
4952 (arc_attr_type): Likewise.
4953 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4954 (walk_stores): Likewise.
4955 (arc_address_cost): Make it static.
4956 (arc_verify_short): Likewise.
4957 (branch_dest): Likewise.
4958 (arc_attr_type): Likewise.
4959 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4960 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4961 (arc_final_prescan_insn): Remove inserting the nops due to
4962 hardware hazards. It is done in reorg step.
4963 (insn_length_variant_t): Remove.
4964 (insn_length_parameters_t): Likewise.
4965 (arc_insn_length_parameters): Likewise.
4966 (arc_get_insn_variants): Likewise.
4967 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4969 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4972 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4975 PR tree-optimization/84687
4976 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4978 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4980 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4982 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4983 Rename to ppc_speculation_barrier.
4984 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4985 __builtin_ppc_speculation_barrier.
4987 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4990 * combine.c (combine_simplify_rtx): Don't try to simplify if
4991 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4994 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4996 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4997 to 32 bytes when compiling for POWER9.
4999 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5002 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5003 regparm >= 3 with no arg reg available also for calls with
5004 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5007 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5009 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5011 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5014 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5016 2018-03-05 Richard Biener <rguenther@suse.de>
5018 PR tree-optimization/84486
5019 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5020 When inserting a __builtin_assume_aligned call set the LHS
5021 SSA name alignment info accordingly.
5023 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5025 PR tree-optimization/84114
5026 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5027 Avoid reassociation of FLOAT_MODE addition.
5029 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5031 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5032 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5033 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5034 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5035 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5036 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5037 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5039 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5040 __builtin_ia32_wbinvd): New builtins.
5041 (SPECIAL_ARGS2): New.
5042 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5043 (SPECIAL_ARGS2): New.
5044 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5045 (ix86_valid_target_attribute_inner_p): Ditto.
5046 (ix86_init_mmx_sse_builtins): Add special_args2.
5047 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5048 TARGET_WBNOINVD_P): New.
5049 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5050 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5051 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5052 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5053 * config/i386/pconfigintrin.h: New file.
5054 * config/i386/wbnoinvdintrin.h: Ditto.
5055 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5056 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5058 2018-03-05 Richard Biener <rguenther@suse.de>
5060 PR tree-optimization/84670
5061 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5063 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5064 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5065 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5066 doesn't grow if all successors (recursively) were visited at least
5069 2018-03-05 Richard Biener <rguenther@suse.de>
5071 PR tree-optimization/84650
5072 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5073 if executed in the loop pipeline.
5075 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5077 * doc/configfiles.texi (Configuration Files): Move info about
5078 conditionalizing $target-protos.h to...
5079 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5080 differs from $target-protos.h.
5082 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5083 Chung-Ju Wu <jasonwucj@gmail.com>
5085 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5086 * config/nds32/nds32-multiple.md (setmemsi): Define.
5087 * config/nds32/nds32-memory-manipulation.c
5088 (nds32_gen_dup_4_byte_to_word_value): New.
5089 (emit_setmem_word_loop): New.
5090 (emit_setmem_byte_loop): New.
5091 (nds32_expand_setmem_loop): New.
5092 (nds32_expand_setmem_loop_v3m): New.
5093 (nds32_expand_setmem_unroll): New.
5094 (nds32_expand_setmem): New.
5096 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5097 Chung-Ju Wu <jasonwucj@gmail.com>
5099 * config/nds32/nds32-memory-manipulation.c
5100 (nds32_emit_load_store): New.
5101 (nds32_emit_post_inc_load_store): New.
5102 (nds32_emit_mem_move): New.
5103 (nds32_emit_mem_move_block): New.
5104 (nds32_expand_movmemsi_loop_unknown_size): New.
5105 (nds32_expand_movmemsi_loop_known_size): New.
5106 (nds32_expand_movmemsi_loop): New.
5107 (nds32_expand_movmemsi_unroll): New.
5108 (nds32_expand_movmemqi): Rename ...
5109 (nds32_expand_movmemsi): ... to this.
5110 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5111 (movmemsi): ... to this.
5112 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5113 (nds32_expand_movmemsi): ... to this.
5115 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5116 Monk Chiang <sh.chiang04@gmail.com>
5117 Chung-Ju Wu <jasonwucj@gmail.com>
5119 * config/nds32/nds32-protos.h
5120 (nds32_expand_load_multiple): New arguments.
5121 (nds32_expand_store_multiple): Ditto.
5122 (nds32_valid_multiple_load_store): Rename ...
5123 (nds32_valid_multiple_load_store_p): ... to this.
5124 * config/nds32/nds32-memory-manipulation.c
5125 (nds32_expand_load_multiple): Refine implementation.
5126 (nds32_expand_store_multiple): Ditto.
5127 * config/nds32/nds32-multiple.md
5128 (load_multiple): Update nds32_expand_load_multiple interface.
5129 (store_multiple): Update nds32_expand_store_multiple interface.
5130 * config/nds32/nds32-predicates.c
5131 (nds32_valid_multiple_load_store): Rename ...
5132 (nds32_valid_multiple_load_store_p): ... to this and refine
5134 * config/nds32/predicates.md
5135 (nds32_load_multiple_and_update_address_operation): New predicate.
5136 (nds32_store_multiple_and_update_address_operation): New predicate.
5138 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5139 Chung-Ju Wu <jasonwucj@gmail.com>
5141 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5142 (combo): New attribute.
5143 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5145 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5147 * config/nds32/nds32.opt: Change -mcmodel= default value.
5149 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5150 Monk Chiang <sh.chiang04@gmail.com>
5151 Chung-Ju Wu <jasonwucj@gmail.com>
5153 * config/nds32/constants.md (unspec_element): New enum.
5154 * config/nds32/constraints.md (Umw): New constraint.
5155 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5156 * config/nds32/nds32-intrinsic.md: Likewise.
5157 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5158 (nds32_valid_smw_lwm_base_p): New.
5159 (nds32_output_smw_single_word): New.
5160 (nds32_output_lmw_single_word): New.
5161 (nds32_expand_unaligned_load): New.
5162 (nds32_expand_unaligned_store): New.
5163 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5164 (nds32_output_smw_single_word): Declare.
5165 (nds32_output_lmw_single_word): Declare.
5166 (nds32_expand_unaligned_load): Declare.
5167 (nds32_expand_unaligned_store): Declare.
5168 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5169 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5170 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5171 NDS32_BUILTIN_UASTORE_DW.
5172 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5175 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5176 Kito Cheng <kito.cheng@gmail.com>
5177 Chung-Ju Wu <jasonwucj@gmail.com>
5179 * config/nds32/nds32-intrinsic.c
5180 (nds32_expand_builtin_null_ftype_reg): Delete.
5181 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5182 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5183 (nds32_read_argument): New.
5184 (nds32_legitimize_target): Ditto.
5185 (nds32_legitimize_argument): Ditto.
5186 (nds32_check_constant_argument): Ditto.
5187 (nds32_expand_unop_builtin): Ditto.
5188 (nds32_expand_unopimm_builtin): Ditto.
5189 (nds32_expand_binop_builtin): Ditto.
5190 (nds32_builtin_decl_impl): Ditto.
5191 (builtin_description): Ditto.
5192 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5193 (nds32_init_builtins_impl): Ditto.
5194 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5195 (nds32_builtin_decl): New.
5196 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5197 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5199 2018-03-02 Jeff Law <law@redhat.com>
5201 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5202 (redundant_insn, fill_simple_delay_slots): Likewise.
5203 (fill_slots_from_thread): Likewise.
5204 * resource.c (mark_referenced_resources): Likewise.
5205 (mark_set_resources, find_dead_or_set_registers): Likewise.
5207 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5209 * substring-locations.h (format_warning_va): Formatting fix for
5211 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5213 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5214 * substring-locations.c: Include intl.h.
5215 (format_warning_va): Turned into small wrapper around
5216 format_warning_n_va, renamed to ...
5217 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5218 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5220 (format_warning_at_substring_n): New function.
5221 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5222 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5223 format_warning_at_substring with just a shorter name instead of
5224 const function pointer.
5225 (fmtwarn_n): New function.
5226 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5227 appropriate, get rid of all the fmtstr temporaries, move conditionals
5228 with G_() wrapped string literals directly into fmtwarn arguments,
5229 cast dir.len to (int), formatting fixes.
5231 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5233 * doc/invoke.texi: Remove "Cilk Plus" references.
5235 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5236 Richard Biener <rguenther@suse.de>
5239 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5240 for error or warning attributes if CALL_FROM_THUNK_P is set.
5243 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5246 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5247 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5250 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5251 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5253 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5254 instead of -1U in last predictors element's probability member.
5256 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5259 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5260 arguments if they are comparable.
5262 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5264 PR tree-optimization/84634
5265 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5266 masks and masked_loop_p with a single loop_masks, making sure it's
5267 null for bb vectorization.
5269 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5271 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5272 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5273 loop->force_vectorize to check whether there is no alias.
5275 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5278 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5280 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5282 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5283 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5284 instead of a loop around prev_real_insn.
5285 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5289 * config/i386/i386.c (ix86_print_operand): Use conditional
5290 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5293 2018-03-02 Richard Biener <rguenther@suse.de>
5295 PR tree-optimization/84427
5296 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5297 (bitmap_set_subtract_values): Rewrite to handle multiple
5300 (prune_clobbered_mems): Likewise.
5301 (phi_translate): Take edge instead of pred/phiblock.
5302 (phi_translate_1): Likewise.
5303 (phi_translate_set): Likewise. Insert all translated
5304 exprs for a value into the set, keeping possibly multiple
5305 expressions per value.
5306 (compute_antic_aux): Adjust for phi_translate changes.
5307 When intersecting union the expressions and prune those
5308 not in the final value set, keeping possibly multiple
5309 expressions per value. Do not use value-insertion
5310 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5311 all expressions. Add verification that the value-sets
5312 only shrink during iteration.
5313 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5314 (do_pre_regular_insertion): Likewise.
5315 (do_pre_partial_partial_insertion): Likewise.
5317 2018-03-02 Richard Biener <rguenther@suse.de>
5320 * config/darwin.c (saved_debug_info_level): New static global.
5321 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5322 (darwin_asm_lto_end): Restore debug info generation settings.
5324 2018-03-01 Martin Liska <mliska@suse.cz>
5327 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5330 2018-03-01 Richard Biener <rguenther@suse.de>
5333 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5334 type annotation with LTO when debug was disabled at compile-time.
5336 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5338 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5340 (mips_final_postscan_insn): Likewise.
5342 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5344 PR rtl-optimization/84528
5345 * alias.c (init_alias_target): Add commentary.
5346 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5347 a unique base value if the frame pointer is not eliminated
5348 to the stack pointer.
5350 2018-03-01 Tom de Vries <tom@codesourcery.com>
5352 PR rtl-optimization/83327
5353 * lra-int.h (hard_regs_spilled_into): Declare.
5354 * lra.c (hard_regs_spilled_into): Define.
5355 (init_reg_info): Init hard_regs_spilled_into.
5356 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5357 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5358 (process_bb_lives): Handle hard_regs_spilled_into.
5359 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5360 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5362 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5364 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5365 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5366 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5367 * config/rs6000/aix72.h: New file.
5369 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5371 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5372 instead of warning_at with conditional singular and plural messages
5376 * stor-layout.c (update_alignment_for_field): For
5377 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5378 && !DECL_PACKED (field), do the alignment update, just use
5379 only desired_align instead of MAX (type_align, desired_align)
5381 (place_field): Don't do known_align < desired_align handling
5382 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5383 is non-NULL, instead do it after rli->prev_field handling and
5384 only if not within a bitfield word. For DECL_PACKED (field)
5385 use type_align of BITS_PER_UNIT.
5387 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5389 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5390 superfluous parentheses and trailing spaces.
5392 2018-02-28 Richard Biener <rguenther@suse.de>
5394 PR tree-optimization/84584
5395 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5396 SCoPs with fake exit edge.
5398 2018-02-28 Martin Liska <mliska@suse.cz>
5401 * timevar.c (timer::print): Fix format to properly print 100%
5404 2018-02-28 Richard Biener <rguenther@suse.de>
5407 * genmatch.c (capture_info::walk_match): Do not mark
5408 captured expressions without operands as expr_p given
5409 they act more like predicates and should be subject to
5410 "lost tail" side-effect preserving.
5412 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5414 PR rtl-optimization/81611
5415 * auto-inc-dec.c (attempt_change): Move dead note from
5416 mem_insn if it's the next use of regno
5417 (find_address): Take address use of reg holding
5418 non-incremented value. Add parm to limit search to the named
5420 (merge_in_block): Attempt to use a mem insn that is the next
5421 use of the original regno.
5423 2018-02-27 Martin Sebor <msebor@redhat.com>
5426 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5427 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5429 2018-02-27 Martin Sebor <msebor@redhat.com>
5431 PR translation/84207
5432 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5433 n argument to unsigned HOST_WIDE_INT.
5434 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5435 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5436 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5437 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5439 2018-02-27 Richard Biener <rguenther@suse.de>
5441 PR tree-optimization/84512
5442 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5443 Do not use the estimate returned from record_stmt_cost for
5444 the scalar iteration cost but sum properly using add_stmt_cost.
5446 2018-02-27 Richard Biener <rguenther@suse.de>
5448 PR tree-optimization/84466
5449 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5450 Adjust last change to less strictly validate use operands.
5452 2018-02-27 Martin Liska <mliska@suse.cz>
5454 PR gcov-profile/84548
5455 * gcov.c (process_file): Allow partial overlap and consider it
5456 also as group functions.
5457 (output_lines): Properly calculate range of lines for a group.
5459 2018-02-27 Martin Liska <mliska@suse.cz>
5461 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5462 'ggc' suffixes. Change first column width.
5463 (timer::print): Fix formatting of the column.
5465 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5467 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5468 preserve inline entry blocks for the sake of debug inline
5469 entry point markers alone.
5470 (remove_unused_locals): Suggest in comments a better place to
5471 force the preservation of inline entry blocks that are
5472 otherwise unused, but do not preserve them.
5474 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5476 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5478 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5481 * config/i386/constraints.md (Bs): Replace
5482 ix86_indirect_branch_register with
5483 TARGET_INDIRECT_BRANCH_REGISTER.
5485 * config/i386/i386.md (indirect_jump): Likewise.
5486 (tablejump): Likewise.
5487 (*sibcall_memory): Likewise.
5488 (*sibcall_value_memory): Likewise.
5489 Peepholes of indirect call and jump via memory: Likewise.
5490 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5491 (*sibcall_value_GOT_32): Likewise.
5492 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5493 (GOT_memory_operand): Likewise.
5494 (call_insn_operand): Likewise.
5495 (sibcall_insn_operand): Likewise.
5496 (GOT32_symbol_operand): Likewise.
5497 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5499 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5501 PR rtl-optimization/83496
5502 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5503 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5504 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5505 redundant insn, if any.
5506 (relax_delay_slots): Likewise.
5507 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5509 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5511 PR tree-optimization/83965
5512 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5513 that grouped statements are part of a reduction chain. Return
5514 true if the statement is not marked as a reduction itself but
5516 (vect_recog_dot_prod_pattern): Don't check whether the statement
5517 is part of a group here.
5518 (vect_recog_sad_pattern): Likewise.
5519 (vect_recog_widen_sum_pattern): Likewise.
5521 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5524 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5526 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5529 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5531 (ix86_output_indirect_function_return): New prototype.
5532 (ix86_split_simple_return_pop_internal): Likewise.
5533 * config/i386/i386.c (indirect_return_via_cx): New.
5534 (indirect_return_via_cx_bnd): Likewise.
5535 (indirect_thunk_name): Handle return va CX_REG.
5536 (output_indirect_thunk_function): Create alias for
5537 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5538 (ix86_output_indirect_jmp): Remove the bool argument.
5539 (ix86_output_indirect_function_return): New function.
5540 (ix86_split_simple_return_pop_internal): Likewise.
5541 * config/i386/i386.md (*indirect_jump): Don't pass false
5542 to ix86_output_indirect_jmp.
5543 (*tablejump_1): Likewise.
5544 (simple_return_pop_internal): Change it to define_insn_and_split.
5545 Call ix86_split_simple_return_pop_internal to split it for
5547 (simple_return_indirect_internal): Call
5548 ix86_output_indirect_function_return instead of
5549 ix86_output_indirect_jmp.
5551 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5554 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5555 memset and value initialization afterwards.
5557 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5559 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5561 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5564 * common/config/aarch64/aarch64-common.c
5565 (aarch_option_optimization_table[]): Switch
5566 off fomit-frame-pointer
5568 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5569 Chung-Ju Wu <jasonwucj@gmail.com>
5571 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5573 (store_multiple): Ditto.
5575 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5577 * config.gcc: Add --with-cpu support for nds32 target.
5578 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5579 * config/nds32/nds32.opt: Add -mcpu= option.
5581 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5583 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5584 isel=yes): Warn for these deprecated options.
5586 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5588 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5589 ISA_2_5_MASKS_EMBEDDED.
5591 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5593 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5594 p->max as pointers rather than using iterative_hash_expr.
5596 2018-02-23 Carl Love <cel@us.ibm.com>
5598 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5599 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5601 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5602 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5603 P8V_BUILTIN_VEC_VUNSIGNED2.
5605 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5608 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5609 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5611 * lra-constraints.c (curr_insn_transform): Set up
5612 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5614 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5615 * lra-eliminations.c (spill_pseudos): Ditto.
5616 (process_insn_for_elimination): Ditto.
5617 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5618 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5619 LRA_NON_CLOBBERED_ALT.
5621 2018-02-22 Martin Sebor <msebor@redhat.com>
5623 PR tree-optimization/84480
5624 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5625 to maybe_diag_stxncpy_trunc. Call it.
5626 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5627 from gimple_fold_builtin_strcpy. Print inlining stack.
5628 (handle_builtin_stxncpy): Print inlining stack.
5629 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5631 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5634 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5635 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5636 and -fcheck-pointer-bounds are used together.
5637 (indirect_thunk_prefix): New enum.
5638 (indirect_thunk_need_prefix): New function.
5639 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5640 "_nt" instead of "_bnd" for NOTRACK prefix.
5641 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5642 (output_indirect_thunk_function): Likewise.
5644 (ix86_code_end): Update output_indirect_thunk_function calls.
5645 (ix86_output_indirect_branch_via_reg): Replace
5646 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5647 (ix86_output_indirect_branch_via_push): Likewise.
5648 (ix86_output_function_return): Likewise.
5649 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5650 incompatible with -fcf-protection=branch and
5651 -fcheck-pointer-bounds.
5653 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5656 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5657 Change gcc_assert call to output_operand_lossage.
5659 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5661 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5663 2018-02-22 DJ Delorie <dj@redhat.com>
5664 Sebastian Perta <sebastian.perta@renesas.com>
5665 Oleg Endo <olegendo@gcc.gnu.org>
5667 * config/rx/rx.c (rx_rtx_costs): New function.
5668 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5670 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5672 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5674 2018-02-22 Martin Liska <mliska@suse.cz>
5677 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5678 Add "native" as a possible value.
5680 2018-02-22 Martin Liska <mliska@suse.cz>
5683 * config/i386/i386.c (ix86_option_override_internal):
5684 Add "native" as a possible value for -march and -mtune.
5686 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5689 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5690 to all type variants.
5692 PR tree-optimization/84503
5693 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5694 width as info->bitpos + info->bitsize - start.
5695 (merged_store_group::merge_overlapping): Simplify width computation.
5696 (check_no_overlap): New function.
5697 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5698 start + width and last_order of the group, fail if check_no_overlap
5700 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5701 to group if check_no_overlap fails.
5703 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5705 * config/rs6000/altivec.md: Delete contraint arguments to
5706 define_expand, define_split, and define_peephole2, and in
5707 define_insn_and_split if always unused.
5708 * config/rs6000/darwin.md: Ditto.
5709 * config/rs6000/dfp.md: Ditto.
5710 * config/rs6000/rs6000.md: Ditto.
5711 * config/rs6000/sync.md: Ditto.
5712 * config/rs6000/vector.md: Ditto.
5713 * config/rs6000/vsx.md: Ditto.
5715 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5717 * config/rs6000/altivec.md: Write output control strings as braced
5718 blocks instead of double-quoted strings.
5719 * config/rs6000/darwin.md: Ditto.
5720 * config/rs6000/rs6000.md: Ditto.
5721 * config/rs6000/vector.md: Ditto.
5722 * config/rs6000/vsx.md: Ditto.
5724 2018-02-21 Jason Merrill <jason@redhat.com>
5726 PR c++/84314 - ICE with templates and fastcall attribute.
5727 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5729 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5731 * ipa-cp.c (determine_versionability): Fix comment typos.
5733 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5736 * ipa-cp.c (determine_versionability): Do not version functions caling
5739 2018-02-21 Martin Liska <mliska@suse.cz>
5742 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5743 Add "native" as a possible value.
5744 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5745 the macro when native cpu detection is available.
5747 2018-02-21 Martin Liska <mliska@suse.cz>
5750 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5751 Add "native" as a possible value.
5752 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5753 when native cpu detection is available.
5755 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5756 Martin Sebor <msebor@redhat.com>
5758 PR tree-optimization/84478
5759 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5761 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5762 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5763 support which is conservatively correct, for 2 only stay conservative
5764 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5765 argument to the 2 argument get_range_strlen, adjust 6 arg
5766 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5768 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5769 (gimple_fold_builtin_strlen): Pass true as last argument to
5772 2018-02-20 Martin Sebor <msebor@redhat.com>
5775 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5776 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5777 (builtin_memref::builtin_memref): Factor out parts into
5778 set_base_and_offset and call it.
5780 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5783 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5784 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5785 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5786 search at the associated MODE_INT.
5788 2018-02-20 Jeff Law <law@redhat.com>
5791 PR tree-optimization/81592
5793 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5794 for range data rather than using global data.
5795 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5796 range data rather than using global data.
5797 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5798 pass it to children as needed.
5799 (struct directive::fmtresult): Similarly.
5800 (struct directive::set_width): Similarly.
5801 (struct directive::set_precision): Similarly.
5802 (format_integer, format_directive, parse_directive): Similarly.
5803 (format_none): Accept unnamed vr_values parameter.
5804 (format_percent, format_floating, format_character): Similarly.
5805 (format_string, format_plain): Similarly.
5806 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5807 the EVRP range analyzer for range data rather than using global data.
5808 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5809 gimple-ssa-evrp-analyze.h
5810 (class sprintf_dom_walker): Add after_dom_children member function.
5811 Add evrp_range_analyzer member.
5812 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5813 range analyzer as needed.
5814 (sprintf_dom_walker::after_dom_children): New member function.
5815 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5817 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5818 (evrp_range_analyzer::pop_to_marker): Likewise.
5820 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5822 PR tree-optimization/84419
5823 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5824 with the required type if its current type is compatible but
5827 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5830 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5831 after vectorization.
5833 2018-02-20 Martin Liska <mliska@suse.cz>
5836 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5837 possible values if we don't have a hint.
5839 2018-02-20 Martin Liska <mliska@suse.cz>
5843 * final.c (shorten_branches): Build align_tab array with one
5845 * opts.c (finish_options): Add alignment option limit check.
5846 (MAX_CODE_ALIGN): Likewise.
5847 (MAX_CODE_ALIGN_VALUE): Likewise.
5848 * doc/invoke.texi: Document maximum allowed option value for
5849 all -falign-* options.
5851 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5854 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5855 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5856 * var-tracking.c (emit_note_insn_var_location): Remove all references
5857 to NOTE_INSN_CALL_ARG_LOCATION.
5858 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5859 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5860 Use copy_rtx_if_shared.
5861 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5862 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5863 (dwarf2out_var_location): Remove handling of
5864 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5866 * final.c (final_scan_insn): Remove all references to
5867 NOTE_INSN_CALL_ARG_LOCATION.
5868 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5869 before dumping final insns.
5870 * except.c (emit_note_eh_region_end): Remove all references to
5871 NOTE_INSN_CALL_ARG_LOCATION.
5872 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5873 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5874 * config/arc/arc.c (hwloop_optimize): Likewise.
5875 * config/arm/arm.c (create_fix_barrier): Likewise.
5876 * config/s390/s390.c (s390_chunkify_start): Likewise.
5877 * config/sh/sh.c (find_barrier): Likewise.
5878 * config/i386/i386.c (rest_of_insert_endbranch,
5879 ix86_seh_fixup_eh_fallthru): Likewise.
5880 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5881 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5882 * config/frv/frv.c (frv_function_prologue): Likewise.
5883 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5885 (note_outside_basic_block_p): Remove all references to
5886 NOTE_INSN_CALL_ARG_LOCATION.
5887 * gengtype.c (adjust_field_rtx_def): Likewise.
5888 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5890 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5891 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5894 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5897 PR tree-optimization/84452
5898 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5899 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5902 2018-02-19 Martin Liska <mliska@suse.cz>
5905 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5907 2018-02-19 Martin Liska <mliska@suse.cz>
5908 Richard Sandiford <richard.sandiford@linaro.org>
5910 PR tree-optimization/82491
5911 * gimple-fold.c (get_base_constructor): Make earlier bail out
5914 2018-02-19 Carl Love <cel@us.ibm.com>
5916 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5917 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5919 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5920 P8V_BUILTIN_VEC_NEG.
5922 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5924 * config/rl78/rl78.md (movdf): New define expand.
5926 2018-02-19 Martin Liska <mliska@suse.cz>
5929 * doc/invoke.texi: Fix typo.
5930 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5932 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5934 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5935 handle rs6000_single_float and rs6000_double_float specially for
5938 2018-02-16 Jeff Law <law@redhat.com>
5940 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5943 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5946 * ipa-inline.c (inline_small_functions): Fix a typo.
5948 2018-02-16 Nathan Sidwell <nathan@acm.org>
5950 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5952 2018-02-16 Carl Love <cel@us.ibm.com>
5954 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5955 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5956 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5957 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5958 expansion to P8V_BUILTIN_VEC_FLOAT2.
5960 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5962 PR rtl-optimization/70023
5963 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5964 src_regno into account.
5966 2018-02-16 Carl Love <cel@us.ibm.com>
5968 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5969 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5970 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5971 * config/rs6000/rs6000.c: Remove case statements for
5972 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5973 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5974 and P9V_BUILTIN_VEC_VINSERT4B.
5975 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5976 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5977 * config/rs6000/vsx.md:
5978 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5981 2018-02-16 Carl Love <cel@us.ibm.com>
5983 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5985 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5987 * config/rs6000/rs6000-c.c: Add the definitions for
5988 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5989 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5990 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5991 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5992 definition for insert4b and define insn *insert3b_internal.
5993 * doc/extend.texi: Add documentation for vec_extract4b.
5995 2018-02-16 Nathan Sidwell <nathan@acm.org>
5997 * doc/extend.texi (Backwards Compatibility): Mention friend
5998 injection. Note for-scope is deprecated.
5999 * doc/invoke.texi (-ffriend-injection): Deprecate.
6001 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6003 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6004 that moved to I2, also allow destinations that are a paradoxical
6005 subreg (instead of a normal reg).
6007 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6010 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6013 2018-02-16 Richard Biener <rguenther@suse.de>
6015 PR tree-optimization/84037
6016 PR tree-optimization/84016
6018 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6019 Adjust vec_construct for the fact we need additional higher latency
6020 128bit inserts for AVX256 and AVX512 vector builds.
6021 (ix86_add_stmt_cost): Scale vector construction cost for
6024 2018-02-16 Richard Biener <rguenther@suse.de>
6026 PR tree-optimization/84417
6027 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6028 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6029 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6031 2018-02-16 Martin Liska <mliska@suse.cz>
6034 * internal-fn.def (ASAN_CHECK): Set proper flags.
6035 (ASAN_MARK): Likewise.
6037 2018-02-16 Julia Koval <julia.koval@intel.com>
6039 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6040 from PTA_CANNONLAKE.
6042 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6045 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6046 Use ++iter rather than iter++ for std::list iterators.
6047 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6048 defer deleting them until all nodes in the forest are processed. Do
6049 free even leaf nodes. Change to_process into auto_vec.
6052 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6053 * vec.h (vec_default_construct): Use memset instead of placement new
6054 if BROKEN_VALUE_INITIALIZATION is defined.
6055 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6056 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6059 PR rtl-optimization/83723
6060 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6061 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6062 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6064 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6066 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6068 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6070 PR rtl-optimization/81443
6071 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6072 from inner REGs to paradoxical SUBREGs.
6074 2018-02-16 Richard Biener <rguenther@suse.de>
6076 PR tree-optimization/84399
6077 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6078 For operands we can analyze at their definition make sure we can
6079 analyze them at each use as well.
6081 2018-02-16 Richard Biener <rguenther@suse.de>
6083 PR tree-optimization/84190
6084 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6085 volatile accesses if the decl isn't volatile.
6087 2018-02-15 Jason Merrill <jason@redhat.com>
6089 PR c++/84314 - ICE with templates and fastcall attribute.
6090 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6091 TYPE_CANONICAL on an existing type.
6093 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6095 PR tree-optimization/84383
6096 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6097 dstoff nor call operand_equal_p if dstbase is NULL.
6099 PR tree-optimization/84334
6100 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6101 also a CONSTANT_CLASS_P, punt.
6103 2018-02-14 Jim Wilson <jimw@sifive.com>
6105 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6106 first SMALL_OPERAND check. New local min_second_step. Move assert
6107 to where locals are set. Add TARGET_RVC support.
6108 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6110 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6112 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6114 2018-02-14 Martin Sebor <msebor@redhat.com>
6116 PR tree-optimization/83698
6117 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6118 arrays constrain the offset range to their bounds.
6119 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6120 (builtin_access::overlap): Avoid setting the size of overlap if it's
6122 (maybe_diag_overlap): Also consider arrays when deciding what values
6123 of offsets to include in diagnostics.
6125 2018-02-14 Martin Sebor <msebor@redhat.com>
6128 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6129 that correspond to the kind of a declaration.
6131 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6134 * config/pa/pa.md: Load address of PIC label using the linkage table
6135 if the label is nonlocal.
6137 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6139 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6140 warning message if user requests -maltivec=be.
6141 * doc/invoke.texi: Document deprecation of -maltivec=be.
6143 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6146 * config/rs6000/rs6000-c.c: Update definitions for
6147 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6148 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6150 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6153 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6154 add _get_ssp intrinsics. Remove argument from
6155 __builtin_ia32_rdssp[d|q].
6156 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6157 * config/i386/i386-builtin.def: Remove argument from
6158 __builtin_ia32_rdssp[d|q].
6159 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6160 ix86_expand_special_args_builtin for _rdssp[d|q].
6161 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6162 Clear register before usage.
6163 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6164 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6166 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6168 PR tree-optimization/84357
6169 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6170 operand 1 of an ARRAY_REF too.
6172 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6175 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6176 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6178 (set_of_reg): New struct.
6179 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6180 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6181 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6183 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6184 Split into bitclr, bitset, bitinvert patterns if appropriate.
6185 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6186 use rx_fuse_in_memory_bitop.
6187 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6188 to named insn, correct maximum insn length.
6190 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6193 * machmode.def: Define a complex mode for PARTIAL_INT.
6194 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6196 * doc/rtl.texi: Document CSPImode.
6197 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6199 (msp430_hard_regno_nregs_with_padding): Likewise.
6201 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6204 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6206 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6208 PR rtl-optimization/84169
6209 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6210 we generated a parallel as new i3 and we split that to new i2 and i3
6211 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6212 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6213 those to i2, not i1. Partially rewrite this scan code.
6215 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6218 * stor-layout.c (place_field): For variable length fields, adjust
6219 offset_align afterwards not just based on the field's alignment,
6220 but also on the size.
6223 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6224 of exps and logs in the use_exp2 case.
6226 2018-02-13 Jeff Law <law@redhat.com>
6228 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6231 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6234 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6238 * common.opt (gas-loc-support, gas-locview-support): New.
6239 (ginline-points, ginternal-reset-location-views): New.
6240 * doc/invoke.texi: Document them. Use @itemx where intended.
6241 (gvariable-location-views): Adjust.
6242 * target.def (reset_location_view): New.
6243 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6244 (TARGET_RESET_LOCATION_VIEW): New.
6245 * doc/tm.texi: Rebuilt.
6246 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6247 (dwarf2out_default_as_locview_support): New.
6248 (output_asm_line_debug_info): Use option variables.
6249 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6250 (output_loc_list): Likewise.
6251 (add_high_low_attributes): Check option variables.
6252 Don't output entry view attribute in strict mode.
6253 (gen_inlined_subroutine_die): Check option variables.
6254 (dwarf2out_inline_entry): Likewise.
6255 (init_sections_and_labels): Likewise.
6256 (dwarf2out_early_finish): Likewise.
6257 (maybe_reset_location_view): New, from...
6258 (dwarf2out_var_location): ... here. Call it.
6259 * debug.h (dwarf2out_default_as_loc_support): Declare.
6260 (dwarf2out_default_as_locview_support): Declare.
6261 * hooks.c (hook_int_rtx_insn_0): New.
6262 * hooks.h (hook_int_rtx_insn_0): Declare.
6263 * toplev.c (process_options): Take -gas-loc-support and
6264 -gas-locview-support from dwarf2out. Enable
6265 -gvariable-location-views by default only with locview
6266 assembler support. Enable -ginternal-reset-location-views by
6267 default only if the target defines the corresponding hook.
6268 Enable -ginline-points by default if location views are
6269 enabled; force it disabled if statement frontiers are
6271 * tree-inline.c (expand_call_inline): Check option variables.
6272 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6274 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6276 PR tree-optimization/84321
6277 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6278 handling. Also check whether the anti-range contains any values
6279 that satisfy the mask; switch to a VR_RANGE if not.
6281 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6284 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6286 2018-02-13 Martin Jambor <mjambor@suse.cz>
6289 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6290 of call statements, also set location of a load to a temporary.
6292 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6294 * config/rl78/rl78.c (add_vector_labels): New function.
6295 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6296 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6297 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6298 which checks that no arguments are passed.
6299 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6300 * doc/extend.texi: Documentation for the new attribute.
6302 2018-02-13 Andreas Schwab <schwab@suse.de>
6304 * config/riscv/linux.h (CPP_SPEC): Define.
6306 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6309 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6310 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6311 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6312 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6313 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6314 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6315 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6316 temporarily for AES and PCLMUL builtins.
6318 PR tree-optimization/84339
6319 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6320 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6324 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6325 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6326 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6328 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6330 * omp-simd-clone.h: New file.
6331 * omp-simd-clone.c: Include omp-simd-clone.h.
6332 (expand_simd_clones): No longer static.
6333 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6334 cgraph.h and omp-simd-clone.h.
6335 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6336 (vect_recog_widen_shift_pattern): Formatting fix.
6337 (vect_pattern_recog_1): Don't check optab for calls.
6340 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6341 operands[2] into a REG before using gen_lowpart on it.
6343 2018-02-12 Jeff Law <law@redhat.com>
6346 * config/sh/sh.c (find_barrier): Consider a sibling call
6349 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6350 successfully back substituting a reg.
6352 2018-02-12 Richard Biener <rguenther@suse.de>
6354 PR tree-optimization/84037
6355 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6356 parameter, move visited init to caller.
6357 (vect_slp_analyze_operations): Separate cost from validity
6358 check, initialize visited once for all instances.
6359 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6361 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6363 (vect_model_promotion_demotion_cost): Likewise.
6364 (vectorizable_bswap): Guard cost modeling with !slp_node
6365 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6367 (vectorizable_call): Likewise.
6368 (vectorizable_conversion): Likewise.
6369 (vectorizable_assignment): Likewise.
6370 (vectorizable_shift): Likewise.
6371 (vectorizable_operation): Likewise.
6372 (vectorizable_store): Likewise.
6373 (vectorizable_load): Likewise.
6374 (vectorizable_condition): Likewise.
6375 (vectorizable_comparison): Likewise.
6377 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6380 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6381 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6382 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6384 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6387 * params.def (inline-min-speedup): Increase from 8 to 15.
6388 (max-inline-insns-auto): Decrease from 40 to 30.
6389 * ipa-split.c (consider_split): Add some buffer for function to
6390 be considered inlining candidate.
6391 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6394 2018-02-12 Richard Biener <rguenther@suse.de>
6396 PR tree-optimization/84037
6397 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6398 matched stmts if we cannot swap the non-matched ones.
6400 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6402 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6403 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6404 _mm_maskz_scalef_round_ss): New intrinsics.
6405 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6406 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6407 __builtin_ia32_scalefss_round): Remove.
6408 (__builtin_ia32_scalefsd_mask_round,
6409 __builtin_ia32_scalefss_mask_round): New intrinsics.
6410 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6411 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6412 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6413 "<round_constraint>")): Changed to ...
6414 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6415 "<round_scalar_constraint>")): ... this.
6416 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6417 %0, %1, %2<round_op3>}"): Changed to ...
6418 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6419 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6420 %2<round_scalar_mask_op3>}"): ... this.
6421 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6423 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6425 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6426 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6427 (_mm_maskz_sqrt_round_ss): New intrinsics.
6428 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6429 (__builtin_ia32_sqrtsd_mask_round)
6430 (__builtin_ia32_sqrtss_mask_round): New builtins.
6431 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6432 (__builtin_ia32_sqrtss_round): Remove.
6433 (__builtin_ia32_sqrtsd_mask_round)
6434 (__builtin_ia32_sqrtss_mask_round): New builtins.
6435 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6436 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6437 ((match_operand:VF_128 1 "vector_operand"
6438 "xBm,<round_constraint>")): Changed to ...
6439 ((match_operand:VF_128 1 "vector_operand"
6440 "xBm,<round_scalar_constraint>")): ... this.
6441 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6442 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6443 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6444 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6445 %<iptr>1<round_scalar_mask_op3>}): ... this.
6446 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6447 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6449 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6452 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6453 Cast vec_cmpeq result to correct type.
6454 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6455 Cast vec_cmpgt result to correct type.
6457 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6459 * final.c (final_scan_insn_1): Renamed from...
6460 (final_scan_insn): ... this. New wrapper, to recover
6461 seen from the outermost call in recursive ones.
6462 * config/sparc/sparc.c (output_return): Drop seen from call.
6463 (output_sibcall): Likewise.
6464 * config/visium/visium.c (output_branch): Likewise.
6466 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6468 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6471 2018-02-10 Alan Modra <amodra@gmail.com>
6474 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6475 Specify LR as an input.
6477 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6480 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6481 remove_member_access_dummy_vars): New functions.
6482 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6483 lower_omp_1, execute_lower_omp): Use them.
6485 PR rtl-optimization/84308
6486 * shrink-wrap.c (spread_components): Release todo vector.
6488 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6490 PR rtl-optimization/57193
6491 * ira-color.c (struct allocno_color_data): Add member
6492 conflict_allocno_hard_prefs.
6493 (update_conflict_allocno_hard_prefs): New.
6494 (bucket_allocno_compare_func): Add a preference based on
6495 conflict_allocno_hard_prefs.
6496 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6497 (color_allocnos): Remove a dead code. Initiate
6498 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6500 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6503 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6504 constraint from =wa to wa. Avoid a subreg on the output operand,
6505 instead use a pseudo and subreg it in a move.
6506 (p9_xxbrd_<mode>): Changed to ...
6507 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6508 (p9_xxbrd_v2df): New expander.
6509 (p9_xxbrw_<mode>): Changed to ...
6510 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6511 (p9_xxbrw_v4sf): New expander.
6513 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6515 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6517 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6520 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6521 multiply in 32-bit mode.
6522 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6523 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6526 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6528 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6529 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6530 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6531 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6533 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6536 * dwarf2out.c (is_trivial_indirect_ref): New function.
6537 (dwarf2out_late_global_decl): Do not generate a location
6538 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6539 and that are not defined in the current unit.
6541 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6543 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6544 instead of a libcall for UNORDERED.
6546 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6549 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6550 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6552 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6555 * config/s390/s390.c (s390_set_current_function): Invoke
6556 s390_indirect_branch_settings also if fndecl didn't change.
6558 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6560 * config/rs6000/rs6000.md (blockage): Set length to zero.
6562 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6564 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6566 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6569 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6570 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6574 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6575 PARALLEL incoming that failed vt_get_decl_and_offset check.
6578 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6579 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6581 (get_variable_section): For decls in named .bss* sections pass true as
6582 second argument to bss_initializer_p.
6584 2018-02-09 Marek Polacek <polacek@redhat.com>
6585 Jakub Jelinek <jakub@redhat.com>
6588 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6589 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6590 Sync some changes from cxx_fold_indirect_ref.
6592 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6594 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6596 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6597 (BLOCK_INLINE_ENTRY_LABEL): New.
6598 (dwarf2out_var_location): Disregard inline entry markers.
6599 (inline_entry_data): New struct.
6600 (inline_entry_data_hasher): New hashtable type.
6601 (inline_entry_data_hasher::hash): New.
6602 (inline_entry_data_hasher::equal): New.
6603 (inline_entry_data_table): New variable.
6604 (add_high_low_attributes): Add DW_AT_entry_pc and
6605 DW_AT_GNU_entry_view attributes if a pending entry is found
6606 in inline_entry_data_table. Add old entry_pc attribute only
6607 if debug nonbinding markers are disabled.
6608 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6609 markers are enabled.
6610 (block_within_block_p, dwarf2out_inline_entry): New.
6611 (dwarf2out_finish): Check that no entries remained in
6612 inline_entry_data_table.
6613 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6614 (final_scan_insn, notice_source_line): Likewise.
6615 (rest_of_clean_state): Skip inline entry markers.
6616 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6618 * gimple.c (gimple_build_debug_inline_entry): New.
6619 * gimple.h (enum gimple_debug_subcode): Add
6620 GIMPLE_DEBUG_INLINE_ENTRY.
6621 (gimple_build_debug_inline_entry): Declare.
6622 (gimple_debug_inline_entry_p): New.
6623 (gimple_debug_nonbind_marker_p): Adjust.
6624 * insn-notes.def (INLINE_ENTRY): New.
6625 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6626 inline entry marker notes.
6627 (print_insn): Likewise.
6628 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6629 (INSN_DEBUG_MARKER_KIND): Likewise.
6630 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6631 * tree-inline.c (expand_call_inline): Build and insert
6632 debug_inline_entry stmt.
6633 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6634 inline entry blocks early, if nonbind markers are enabled.
6635 (dump_scope_block): Dump fragment info.
6636 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6637 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6638 (gimple_build_debug_inline_entry): New.
6639 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6640 Enable/disable inline entry points too.
6641 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6642 (DEBUG_INSN): Describe inline entry markers.
6644 * common.opt (gvariable-location-views): New.
6645 (gvariable-location-views=incompat5): New.
6646 * config.in: Rebuilt.
6647 * configure: Rebuilt.
6648 * configure.ac: Test assembler for view support.
6649 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6650 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6651 * dwarf2out.c (var_loc_view): New typedef.
6652 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6653 (dwarf2out_locviews_in_attribute): New.
6654 (dwarf2out_locviews_in_loclist): New.
6655 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6656 (enum dw_line_info_opcode): Add LI_adv_address.
6657 (struct dw_line_info_table): Add view.
6658 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6659 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6660 (zero_view_p): New variable.
6661 (ZERO_VIEW_P): New macro.
6662 (output_asm_line_debug_info): New.
6663 (struct var_loc_node): Add view.
6664 (add_AT_view_list, AT_loc_list): New.
6665 (add_var_loc_to_decl): Add view param. Test it against last.
6666 (new_loc_list): Add view params. Record them.
6667 (AT_loc_list_ptr): Handle loc and view lists.
6668 (view_list_to_loc_list_val_node): New.
6669 (print_dw_val): Handle dw_val_class_view_list.
6670 (size_of_die): Likewise.
6671 (value_format): Likewise.
6672 (loc_list_has_views): New.
6673 (gen_llsym): Set vl_symbol too.
6674 (maybe_gen_llsym, skip_loc_list_entry): New.
6675 (dwarf2out_maybe_output_loclist_view_pair): New.
6676 (output_loc_list): Output view list or entries too.
6677 (output_view_list_offset): New.
6678 (output_die): Handle dw_val_class_view_list.
6679 (output_dwarf_version): New.
6680 (output_compilation_unit_header): Use it.
6681 (output_skeleton_debug_sections): Likewise.
6682 (output_rnglists, output_line_info): Likewise.
6683 (output_pubnames, output_aranges): Update version comments.
6684 (output_one_line_info_table): Output view numbers in asm comments.
6685 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6686 Call maybe_gen_llsym.
6687 (loc_list_from_tree_1): Adjust.
6688 (add_AT_location_description): Create view list attribute if
6689 needed, check it's absent otherwise.
6690 (convert_cfa_to_fb_loc_list): Adjust.
6691 (maybe_emit_file): Call output_asm_line_debug_info for test.
6692 (dwarf2out_var_location): Reset views as needed. Precompute
6693 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6694 attribute. Set view.
6695 (new_line_info_table): Reset next view.
6696 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6697 (dwarf2out_source_line): Likewise. Output view resets and labels to
6698 the assembler, or select appropriate line info opcodes.
6699 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6700 (optimize_string_length): Catch it. Adjust.
6701 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6702 dw_val_class_view_list, and remove it if no longer needed.
6703 (hash_loc_list): Hash view numbers.
6704 (loc_list_hasher::equal): Compare them.
6705 (optimize_location_lists): Check whether a view list symbol is
6706 needed, and whether the locview attribute is present, and
6707 whether they match. Remove the locview attribute if no longer
6709 (index_location_lists): Call skip_loc_list_entry for test.
6710 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6711 Use output_dwarf_version.
6712 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6713 (struct dw_val_node): Add val_view_list.
6714 * final.c (SEEN_NEXT_VIEW): New.
6715 (set_next_view_needed): New.
6716 (clear_next_view_needed): New.
6717 (maybe_output_next_view): New.
6718 (final_start_function): Rename to...
6719 (final_start_function_1): ... this. Take pointer to FIRST,
6720 add SEEN parameter. Emit param bindings in the initial view.
6721 (final_start_function): Reintroduce SEEN-less interface.
6722 (final): Rename to...
6723 (final_1): ... this. Take SEEN parameter. Output final pending
6724 next view at the end.
6725 (final): Reintroduce seen-less interface.
6726 (final_scan_insn): Output pending next view before switching
6727 sections or ending a block. Mark the next view as needed when
6728 outputting variable locations. Notify debug backend of section
6729 changes, and of location view changes.
6730 (rest_of_handle_final): Adjust.
6731 * toplev.c (process_options): Autodetect value for debug variable
6732 location views option. Warn on incompat5 without -gdwarf-5.
6733 * doc/invoke.texi (gvariable-location-views): New.
6734 (gvariable-location-views=incompat5): New.
6735 (gno-variable-location-views): New.
6737 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6739 PR tree-optimization/84136
6740 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6741 that the result of find_edge is non-NULL.
6743 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6746 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6747 storing integer register in SImode. Fix cost of 256 and 512
6748 byte aligned SSE register store.
6750 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6752 * config/i386/i386.c (ix86_multiplication_cost): Fix
6753 multiplication cost for TARGET_AVX512DQ.
6755 2018-02-08 Marek Polacek <polacek@redhat.com>
6757 PR tree-optimization/84238
6758 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6761 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6763 PR tree-optimization/84265
6764 * tree-vect-stmts.c (vectorizable_store): Don't treat
6765 VMAT_CONTIGUOUS accesses as grouped.
6766 (vectorizable_load): Likewise.
6768 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6770 PR tree-optimization/81635
6771 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6772 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6773 (test_round_for_mask): New functions.
6774 (wide_int_cc_tests): Call test_round_for_mask.
6775 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6776 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6777 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6778 range returned by get_range_info.
6780 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6783 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6784 * symtab.c: Include builtins.h
6785 (symtab_node::output_to_lto_symbol_table_p): Move here
6786 from lto-streamer-out.c:output_symbol_p.
6787 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6788 (output_symbol_p): Move all logic to symtab.c
6789 (produce_symtab): Update.
6791 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6793 * config/s390/s390-opts.h (enum indirect_branch): Define.
6794 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6795 (s390_indirect_branch_via_thunk)
6796 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6797 (enum s390_indirect_branch_type): Define.
6798 * config/s390/s390.c (struct s390_frame_layout, struct
6799 machine_function): Remove.
6800 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6801 (indirect_branch_table_label_no, indirect_branch_table_name):
6803 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6804 (enum s390_indirect_branch_option): Define.
6805 (s390_return_addr_from_memory): New function.
6806 (s390_handle_string_attribute): New function.
6807 (s390_attribute_table): Add new attribute handler.
6808 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6809 (s390_indirect_branch_via_thunk): New function.
6810 (s390_indirect_branch_via_inline_thunk): New function.
6811 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6812 sibling call optimization for non z10 compiles.
6813 (s390_emit_call): Force indirect branch target to be a single
6814 register. Add r1 clobber for non-z10 compiles.
6815 (s390_emit_epilogue): Emit return jump via return_use expander.
6816 (s390_reorg): Handle JUMP_INSNs as execute targets.
6817 (s390_option_override_internal): Perform validity checks for the
6818 new command line options.
6819 (s390_indirect_branch_attrvalue): New function.
6820 (s390_indirect_branch_settings): New function.
6821 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6822 (s390_output_indirect_thunk_function): New function.
6823 (s390_code_end): Implement target hook.
6824 (s390_case_values_threshold): Implement target hook.
6825 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6827 * config/s390/s390.h (struct s390_frame_layout)
6828 (struct machine_function): Move here from s390.c.
6829 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6830 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6831 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6832 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6833 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6834 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6835 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6836 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6837 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6838 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6839 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6840 (mnemonic attribute): Add values which aren't recognized
6842 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6843 pattern for branch conversion. Fix mnemonic attribute.
6844 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6845 indirect branch via thunk if requested.
6846 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6847 ("*indirect_jump"): Disable for branch conversion using out of
6849 ("indirect_jump_via_thunk<mode>_z10")
6850 ("indirect_jump_via_thunk<mode>")
6851 ("indirect_jump_via_inlinethunk<mode>_z10")
6852 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6853 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6854 ("casesi_jump_via_inlinethunk<mode>_z10")
6855 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6856 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6857 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6858 ("*indirect2_jump"): Disable for branch conversion.
6859 ("casesi_jump"): Turn into expander and expand patterns for branch
6861 ("return_use"): New expander.
6862 ("*return"): Emit return via thunk and rename it to ...
6863 ("*return<mode>"): ... this one.
6864 * config/s390/s390.opt: Add new options and and enum for the
6867 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6869 * lra-constraints.c (match_reload): Unconditionally use
6870 gen_lowpart_SUBREG, rather than selecting between that
6871 and equivalent gen_rtx_SUBREG code.
6873 2018-02-08 Richard Biener <rguenther@suse.de>
6875 PR tree-optimization/84233
6876 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6877 changed flag instead of boguously re-using phi_inserted.
6879 2018-02-08 Martin Jambor <mjambor@suse.cz>
6881 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6882 static local variables.
6884 2018-02-08 Richard Biener <rguenther@suse.de>
6886 PR tree-optimization/84278
6887 * tree-vect-stmts.c (vectorizable_store): When looking for
6888 smaller vector types to perform grouped strided loads/stores
6889 make sure the mode is supported by the target.
6890 (vectorizable_load): Likewise.
6892 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6894 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6895 Increase LDP/STP opportunities by adding adjacent callee-saves.
6897 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6899 PR rtl-optimization/84068
6900 PR rtl-optimization/83459
6901 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6903 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6905 PR tree-optimization/84224
6906 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6907 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6910 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6913 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6914 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6915 count, remove one USE.
6917 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6919 * doc/install.texi (Configuration): Document the
6920 --with-long-double-format={ibm,ieee} PowerPC configuration
6924 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6925 Convert from define_expand to be define_insn_and_split. Rework
6926 float/double/_Float128 conversions to QI/HI/SImode to work with
6927 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6928 conversions to QI/HImode types did a store and then a load to
6929 truncate the value. For conversions to VSX registers, don't split
6930 the insn, instead emit the code directly. Use the code iterator
6931 any_fix to combine signed and unsigned conversions.
6932 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6933 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6934 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6935 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6936 (fix_<mode>di2_hw): Likewise.
6937 (fixuns_<mode>di2_hw): Likewise.
6938 (fix_<mode>si2_hw): Likewise.
6939 (fixuns_<mode>si2_hw): Likewise.
6940 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6941 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6942 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6943 fix<uns>_trunc<SFDF:mode>si2_p8.
6944 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6946 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6947 (fix<uns>_<mode>_mem): Likewise.
6948 (fctiw<u>z_<mode>_mem): Likewise.
6949 (fix<uns>_<mode>_mem): Likewise.
6950 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6951 the register allocator from doing a direct move to the GPRs to do
6952 a store, and instead use the ISA 3.0 store byte/half-word from
6953 vector register instruction. For IEEE 128-bit floating point,
6954 also optimize stores of 32-bit ints.
6955 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6957 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6959 * genextract.c (push_pathstr_operand): New function to support
6961 (walk_rtx): Call push_pathstr_operand.
6962 (print_path): Support [a-zA-Z].
6964 2018-02-07 Richard Biener <rguenther@suse.de>
6966 PR tree-optimization/84037
6967 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6968 (cse_and_gimplify_to_preheader): Declare.
6969 (vect_get_place_in_interleaving_chain): Likewise.
6970 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6972 (_loop_vec_info::~_loop_vec_info): Delete it.
6973 (cse_and_gimplify_to_preheader): New function.
6974 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6975 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6976 (vectorizable_load): Likewise. For grouped stores always base
6977 the IV on the first element.
6978 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6979 condition before gimplifying.
6981 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6983 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6984 *DIV_EXPR and *MOD_EXPR.
6986 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6989 * config/i386/i386.c (ix86_option_override_internal): Mask out
6990 the CF_SET bit when checking -fcf-protection.
6992 2018-02-07 Tom de Vries <tom@codesourcery.com>
6995 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6998 2018-02-07 Richard Biener <rguenther@suse.de>
7000 PR tree-optimization/84204
7001 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7004 PR tree-optimization/84205
7005 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7006 special-case isl_ast_op_zdiv_r.
7008 PR tree-optimization/84223
7009 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7010 Only add conditions from within the region.
7011 (gather_bbs::after_dom_children): Adjust.
7013 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7016 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7017 * config/avr/avr.md: Only post-reload split REG-REG moves if
7018 either register is GENERAL_REG_P.
7020 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7022 PR tree-optimization/84235
7023 * tree-ssa-scopedtables.c
7024 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7025 if the subtraction is performed in floating point type where NaNs are
7026 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7027 build 1. Formatting fix.
7029 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7032 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7033 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7034 and skip it regardless of bb boundaries. Use CALL_P macro,
7035 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7036 unnecessarily, formatting fix.
7038 2018-02-06 Michael Collison <michael.collison@arm.com>
7040 * config/arm/thumb2.md:
7041 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7042 (*thumb_mov_notscc): Ditto.
7044 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7047 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7048 unsigned_fix, not "s".
7050 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7052 * configure.ac (gcc_fn_eh_frame_ro): New function.
7053 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7054 correct .eh_frame permissions.
7055 * configure: Regenerate.
7057 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7059 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7062 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7064 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7065 Display warning message for -mno-speculate-indirect-jumps.
7067 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7069 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7071 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7073 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7075 PR tree-optimization/84225
7076 * tree-eh.c (find_trapping_overflow): Only call
7077 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7079 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7082 * config/i386/i386.c: Reimplement the check of possible options
7083 -mibt/-mshstk conbination. Change error messages.
7084 * doc/invoke.texi: Fix a typo: remove extra '='.
7086 2018-02-06 Marek Polacek <polacek@redhat.com>
7088 PR tree-optimization/84228
7089 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7091 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7094 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7095 emitted arch directives.
7096 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7097 __ARM_FEATURE_COPROC before changing architectures.
7099 2018-02-06 Richard Biener <rguenther@suse.de>
7101 * config/i386/i386.c (print_reg): Fix typo.
7102 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7104 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7106 * configure: Regenerate.
7108 2018-02-05 Martin Sebor <msebor@redhat.com>
7110 PR tree-optimization/83369
7111 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7114 2018-02-05 Martin Liska <mliska@suse.cz>
7116 * doc/invoke.texi: Cherry-pick upstream r323995.
7118 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7120 * ira.c (ira_init_register_move_cost): Adjust comment.
7122 2018-02-05 Martin Liska <mliska@suse.cz>
7124 PR gcov-profile/84137
7125 * doc/gcov.texi: Fix typo in documentation.
7127 2018-02-05 Martin Liska <mliska@suse.cz>
7129 PR gcov-profile/83879
7130 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7131 using dlopen functionality.
7133 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7135 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7136 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7137 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7138 _mm_maskz_range_round_ss): New intrinsics.
7139 (__builtin_ia32_rangesd128_round)
7140 (__builtin_ia32_rangess128_round): Remove.
7141 (__builtin_ia32_rangesd128_mask_round,
7142 __builtin_ia32_rangess128_mask_round): New builtins.
7143 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7144 __builtin_ia32_rangess128_round): Remove.
7145 (__builtin_ia32_rangesd128_mask_round,
7146 __builtin_ia32_rangess128_mask_round): New builtins.
7147 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7148 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7149 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7150 "<round_saeonly_constraint>")): Changed to ...
7151 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7152 "<round_saeonly_scalar_constraint>")): ... this.
7153 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7154 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7155 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7156 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7157 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7159 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7161 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7163 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7164 Remove all values except native, 8540 and 8548.
7166 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7168 * config/i386/i386.c (ix86_output_function_return): Pass
7169 INVALID_REGNUM, instead of -1, as invalid register number to
7170 indirect_thunk_name and output_indirect_thunk.
7172 2018-02-02 Julia Koval <julia.koval@intel.com>
7174 * config.gcc: Add -march=icelake.
7175 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7176 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7177 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7178 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7179 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7180 (processor_target_table): Add icelake.
7181 (ix86_option_override_internal): Handle new PTAs.
7182 (get_builtin_code_for_version): Handle icelake.
7183 (M_INTEL_COREI7_ICELAKE): New.
7184 (fold_builtin_cpu): Handle icelake.
7185 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7186 * doc/invoke.texi: Add -march=icelake.
7188 2018-02-02 Julia Koval <julia.koval@intel.com>
7190 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7191 to wide_int_bitmask.
7192 * wide-int-bitmask.h: New.
7194 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7197 * config/i386/i386.md: Replace Pmode with word_mode in
7198 builtin_setjmp_setup and builtin_longjmp to support x32.
7200 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7204 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7206 (rs6000_supported_cpu_names): New static variable.
7207 (linux_cpu_translation_table): Likewise.
7208 (elf_platform) <cpu>: Define new static variable and use it.
7209 Translate kernel AT_PLATFORM name to canonical name if needed.
7210 Error if platform name is unknown.
7212 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7215 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7217 2018-02-01 Jeff Law <law@redhat.com>
7220 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7221 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7222 the scratch if RELEASE_VIA_POP is false.
7223 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7224 If we have to save a temporary register, decrement SIZE appropriately.
7225 Pass new arguments to release_scratch_register_on_entry.
7226 (ix86_adjust_stack_and_probe): Likewise.
7227 (ix86_emit_probe_stack_range): Pass new arguments to
7228 release_scratch_register_on_entry.
7230 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7232 PR rtl-optimization/84157
7233 * combine.c (change_zero_ext): Use REG_P predicate in
7234 front of HARD_REGISTER_P predicate.
7236 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7238 * config/avr/avr.c (avr_option_override): Move disabling of
7239 -fdelete-null-pointer-checks to...
7240 * common/config/avr/avr-common.c (avr_option_optimization_table):
7243 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7245 PR tree-optimization/81635
7246 * tree-data-ref.c (split_constant_offset_1): For types that
7247 wrap on overflow, try to use range info to prove that wrapping
7250 2018-02-01 Renlin Li <renlin.li@arm.com>
7253 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7255 (aarch64_register_move_cost): Likewise.
7256 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7258 (REG_CLASS_NAMES): Likewise.
7259 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7260 TAILCALL_ADDR_REGS. Remove IP registers.
7261 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7263 2018-02-01 Richard Biener <rguenther@suse.de>
7265 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7266 for specifying RPO order and allow NULL for that.
7267 * domwalk.c (dom_walker::dom_walker): Likewise.
7268 (dom_walker::walk): Handle NULL RPO order.
7269 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7271 (rewrite_update_dom_walker): Likewise.
7272 (mark_def_dom_walker): Likewise.
7274 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7276 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7277 (aarch64_maybe_expand_sve_subreg_move): Declare.
7278 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7279 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7281 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7282 that are semantically a reverse operation.
7283 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7284 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7285 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7287 (aarch64_can_change_mode_class): For big-endian, forbid changes
7288 between two SVE modes if they have different element sizes.
7290 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7292 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7293 the TImode handling for big-endian targets.
7295 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7297 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7298 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7300 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7301 Remove BSWAP handing for big-endian targets and use the form of
7302 LD1RQ appropariate for the mode.
7304 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7306 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7307 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7310 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7313 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7314 check for operands that need to go through aarch64_sve_reload_be.
7316 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7318 PR tree-optimization/81661
7319 PR tree-optimization/84117
7320 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7321 * tree-eh.c: Include gimplify.h.
7322 (find_trapping_overflow, replace_trapping_overflow,
7323 rewrite_to_non_trapping_overflow): New functions.
7324 * tree-vect-loop.c: Include tree-eh.h.
7325 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7326 * tree-data-ref.c: Include tree-eh.h.
7327 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7329 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7331 PR rtl-optimization/84123
7332 * combine.c (change_zero_ext): Check if hard register satisfies
7333 can_change_dest_mode before calling gen_lowpart_SUBREG.
7335 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7338 * ira.c (ira_init_register_move_cost): Remove assert.
7340 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7342 PR rtl-optimization/84071
7343 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7344 * doc/tm.texi: Regenerate.
7346 2018-01-31 Richard Biener <rguenther@suse.de>
7348 PR tree-optimization/84132
7349 * tree-data-ref.c (analyze_miv_subscript): Properly
7350 check whether evolution_function_is_affine_multivariate_p
7351 before calling gcd_of_steps_may_divide_p.
7353 2018-01-31 Julia Koval <julia.koval@intel.com>
7356 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7357 * config/i386/i386.md (rdpid_rex64) New.
7358 (rdpid): Make 32bit only.
7360 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7363 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7364 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7366 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7369 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7371 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7373 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7375 PR rtl-optimization/84071
7376 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7377 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7379 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7381 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7382 (arc_attribute_table): Add 'aux' attribute.
7383 (arc_in_small_data_p): Consider aux like variables.
7384 (arc_is_aux_reg_p): New function.
7385 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7386 (arc_get_aux_arg): New function.
7387 (prepare_move_operands): Handle aux-register access.
7388 (arc_handle_aux_attribute): New function.
7389 * doc/extend.texi (ARC Variable attributes): Add subsection.
7391 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7393 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7394 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7395 (arc_attribute_table): Add 'uncached' attribute.
7396 (arc_print_operand): Print '.di' flag for uncached memory
7398 (arc_in_small_data_p): Do not consider for small data the uncached
7400 (arc_is_uncached_mem_p): New function.
7401 * config/arc/predicates.md (compact_store_memory_operand): Check
7402 for uncached memory accesses.
7403 (nonvol_nonimm_operand): Likewise.
7404 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7406 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7409 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7410 falign-loops=): Add Optimization flag.
7412 2018-01-30 Jeff Law <law@redhat.com>
7415 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7416 INT_REGISTERS_SAVED. Check it prior to calling
7417 get_scratch_register_on_entry.
7418 (ix86_adjust_stack_and_probe): Similarly.
7419 (ix86_emit_probe_stack_range): Similarly.
7420 (ix86_expand_prologue): Corresponding changes.
7422 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7425 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7426 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7428 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7431 * lra-constraints.c (curr_insn_transform): Process AND in the
7434 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7436 PR rtl-optimization/83986
7437 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7438 dependence against last_pending_memory_flush in addition to
7441 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7443 PR tree-optimization/81611
7444 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7447 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7450 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7453 2018-01-30 Richard Biener <rguenther@suse.de>
7454 Jakub Jelinek <jakub@redhat.com>
7456 PR tree-optimization/84111
7457 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7458 inner loops added during recursion, as they don't have up-to-date
7461 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7464 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7465 (can_inline_edge_by_limits_p): ... here.
7466 (can_early_inline_edge_p, check_callers,
7467 update_caller_keys, update_callee_keys, recursive_inlining,
7468 add_new_edges_to_heap, speculation_useful_p,
7469 inline_small_functions,
7470 inline_small_functions, flatten_function,
7471 inline_to_all_callers_1): Update.
7473 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7475 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7478 2018-01-30 Richard Biener <rguenther@suse.de>
7480 PR tree-optimization/83008
7481 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7482 invariant and constant vector uses in stmts when they need
7485 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7488 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7489 * configure: Regenerate.
7491 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7493 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7495 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7496 Use gen_rtx_REG rather than gen_lowpart.
7498 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7500 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7501 rather than 0 when creating partial subregs.
7503 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7505 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7508 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7511 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7512 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7513 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7514 flags. This restores the settings used before the 2017-07-24.
7515 Turning off pre increment/decrement/modify allows IVOPTS to
7516 optimize DF/SF loops where the index is an int.
7518 2018-01-29 Richard Biener <rguenther@suse.de>
7519 Kelvin Nilsen <kelvin@gcc.gnu.org>
7522 * tree-vect-stmts.c (vectorizable_call): Don't call
7523 targetm.vectorize_builtin_md_vectorized_function if callee is
7526 2018-01-22 Carl Love <cel@us.ibm.com>
7528 * doc/extend.tex: Fix typo in second arg in
7529 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7531 2018-01-29 Richard Biener <rguenther@suse.de>
7533 PR tree-optimization/84086
7534 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7535 (flush_ssaname_freelist): When SSA names were released reset
7536 the SCEV hash table.
7538 2018-01-29 Richard Biener <rguenther@suse.de>
7540 PR tree-optimization/84057
7541 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7542 removed paths when removing edges.
7544 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7546 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7547 -mfunction-return=@var{choice}.
7549 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7552 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7554 (layout::print_source_line): Likewise.
7555 (test_get_line_width_without_trailing_whitespace): Add test cases.
7557 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7560 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7563 2018-01-26 Jim Wilson <jimw@sifive.com>
7565 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7567 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7570 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7572 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7573 and CMP + SUB-immediate -> SUBS.
7575 2018-01-26 Martin Sebor <msebor@redhat.com>
7577 PR tree-optimization/83896
7578 * tree-ssa-strlen.c (get_string_len): Rename...
7579 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7580 Avoid assuming length is constant.
7581 (handle_char_store): Use HOST_WIDE_INT for string length.
7583 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7586 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7587 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7589 2018-01-26 Richard Biener <rguenther@suse.de>
7591 PR rtl-optimization/84003
7592 * dse.c (record_store): Only record redundant stores when
7593 the earlier store aliases at least all accesses the later one does.
7595 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7597 PR rtl-optimization/83985
7598 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7599 REG_CFA_RESTORE insns.
7600 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7601 REG_CFA_RESTORE insns here.
7604 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7605 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7607 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7609 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7610 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7611 (arc_init): Likewise.
7612 (arc_override_options): Likewise.
7613 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7615 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7617 * config/arc/arc.h (TARGET_DBNZ): Define.
7618 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7619 properly set the tune attribute.
7620 (dbnz): Use TARGET_DBNZ guard.
7621 * config/arc/arc.opt (mtune): Add core3 option.
7623 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7625 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7626 recognize new pic like addresses.
7627 (arc_delegitimize_address): Clean up.
7629 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7631 * config/arc/arc-arches.def: Option mrf16 valid for all
7633 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7634 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7635 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7636 * config/arc/arc-tables.opt: Regenerate.
7637 * config/arc/arc.c (arc_conditional_register_usage): Handle
7638 reduced register file case.
7639 (arc_file_start): Set must have build attributes.
7640 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7642 * config/arc/arc.opt (mrf16): Add new option.
7643 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7644 * config/arc/genmultilib.awk: Handle new mrf16 option.
7645 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7646 * config/arc/t-multilib: Regenerate.
7647 * doc/invoke.texi (ARC Options): Document mrf16 option.
7649 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7651 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7652 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7653 (arc_attribute_table): Add 'secure_call' attribute.
7654 (arc_print_operand): Print secure call operand.
7655 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7657 (arc_is_secure_call_p): New function. * config/arc/arc.md
7658 (call_i): Add support for sjli instruction.
7659 (call_value_i): Likewise.
7660 * config/arc/constraints.md (Csc): New constraint.
7662 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7663 John Eric Martin <John.Martin@emmicro-us.com>
7665 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7666 * config/arc/arc.c (_arc_jli_section): New struct.
7667 (arc_jli_section): New type.
7668 (rc_jli_sections): New static variable.
7669 (arc_handle_jli_attribute): New function.
7670 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7671 (arc_file_end): New function.
7672 (TARGET_ASM_FILE_END): Define.
7673 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7674 (arc_add_jli_section): New function.
7675 (jli_call_scan): Likewise.
7676 (arc_reorg): Call jli_call_scan.
7677 (arc_output_addsi): Remove 'S' from printing asm operand.
7678 (arc_is_jli_call_p): New function.
7679 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7681 (movhi_insn): Likewise.
7682 (movsi_insn): Likewise.
7683 (movsi_set_cc_insn): Likewise.
7684 (loadqi_update): Likewise.
7685 (load_zeroextendqisi_update): Likewise.
7686 (load_signextendqisi_update): Likewise.
7687 (loadhi_update): Likewise.
7688 (load_zeroextendhisi_update): Likewise.
7689 (load_signextendhisi_update): Likewise.
7690 (loadsi_update): Likewise.
7691 (loadsf_update): Likewise.
7692 (movsicc_insn): Likewise.
7693 (bset_insn): Likewise.
7694 (bxor_insn): Likewise.
7695 (bclr_insn): Likewise.
7696 (bmsk_insn): Likewise.
7697 (bicsi3_insn): Likewise.
7698 (cmpsi_cc_c_insn): Likewise.
7699 (movsi_ne): Likewise.
7700 (movsi_cond_exec): Likewise.
7701 (clrsbsi2): Likewise.
7712 (call_i): Remove 'S' asm letter, add jli instruction.
7713 (call_value_i): Likewise.
7714 * config/arc/arc.op (mjli-always): New option.
7715 * config/arc/constraints.md (Cji): New constraint.
7716 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7718 (subsf3_fpx): Likewise.
7719 (mulsf3_fpx): Likewise.
7720 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7722 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7724 * doc/invoke.texi (ARC): Document mjli-always option.
7726 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7728 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7729 avoid addition with 0 and use incw and decw where possible.
7731 2018-01-26 Richard Biener <rguenther@suse.de>
7733 PR tree-optimization/81082
7734 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7735 association if it requires casting to unsigned.
7736 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7737 from fold_plusminus_mult_expr to catch important cases late when
7738 range info is available.
7740 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7742 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7743 * configure.ac (hidden_linkonce): New test.
7744 * configure: Regenerate.
7745 * config.in: Regenerate.
7747 2018-01-26 Julia Koval <julia.koval@intel.com>
7749 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7750 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7751 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7752 _mm_mask_bitshuffle_epi64_mask): Fix type.
7753 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7754 USI_FTYPE_V4DI_V4DI_USI): Remove.
7755 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7756 __builtin_ia32_vpshufbitqmb256_mask,
7757 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7758 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7759 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7761 2018-01-26 Alan Modra <amodra@gmail.com>
7764 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7765 UNSPEC_VBPERMQ. Sort other unspecs.
7767 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7769 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7771 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7774 * predict.c (drop_profile): Do not push/pop cfun; update also
7776 (handle_missing_profiles): Fix logic looking for zero profiles.
7778 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7781 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7782 on functions with #pragma omp declare simd or functions with simd
7784 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7785 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7786 Remove trailing \n from warning_at calls.
7788 2018-01-25 Tom de Vries <tom@codesourcery.com>
7791 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7792 for neutered workers.
7794 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7797 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7798 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7800 2018-01-24 Jeff Law <law@redhat.com>
7803 * i386.c (get_probe_interval): Move to earlier point.
7804 (ix86_compute_frame_layout): If -fstack-clash-protection and
7805 the frame is larger than the probe interval, then use pushes
7806 to save registers rather than reg->mem moves.
7807 (ix86_expand_prologue): Remove conditional for int_registers_saved
7810 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7813 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7814 min/max for never referenced object.
7816 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7819 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7821 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7822 attributes from DECL_ATTRIBUTES (decl) without affecting
7823 DECL_ATTRIBUTES (current_function_decl).
7824 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7825 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7827 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7829 PR tree-optimization/83979
7830 * fold-const.c (fold_comparison): Use constant_boolean_node
7831 instead of boolean_{true,false}_node.
7833 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7835 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7838 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7840 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7841 Simplify the clause that sets the length attribute.
7842 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7843 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7844 clause that sets the length attribute.
7845 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7847 2018-01-24 Tom de Vries <tom@codesourcery.com>
7850 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7851 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7852 Add strict parameter.
7853 (prevent_branch_around_nothing): Insert dummy insn between branch to
7854 label and label with no ptx insn inbetween.
7855 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7857 2018-01-24 Tom de Vries <tom@codesourcery.com>
7860 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7861 for neutered threads in warp.
7862 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7864 2018-01-24 Richard Biener <rguenther@suse.de>
7866 PR tree-optimization/83176
7867 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7870 2018-01-24 Richard Biener <rguenther@suse.de>
7872 PR tree-optimization/82819
7873 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7874 code generating pluses that are no-ops in the target precision.
7876 2018-01-24 Richard Biener <rguenther@suse.de>
7879 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7881 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7883 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7884 to merge probabilities.
7885 * predict.c (probably_never_executed): Also mark as cold functions
7886 with global 0 profile and guessed local profile.
7887 * profile-count.c (profile_probability::combine_with_count): New
7889 * profile-count.h (profile_probability::operator*,
7890 profile_probability::operator*=, profile_probability::operator/,
7891 profile_probability::operator/=): Reduce precision to adjusted
7892 and set value to guessed on contradictory divisions.
7893 (profile_probability::combine_with_freq): Remove.
7894 (profile_probability::combine_wiht_count): Declare.
7895 (profile_count::force_nonzero):: Set to adjusted.
7896 (profile_count::probability_in):: Set quality to adjusted.
7897 * tree-ssa-tail-merge.c (replace_block_by): Use
7900 2018-01-23 Andrew Waterman <andrew@sifive.com>
7901 Jim Wilson <jimw@sifive.com>
7903 * config/riscv/riscv.c (riscv_stack_boundary): New.
7904 (riscv_option_override): Set riscv_stack_boundary. Handle
7905 riscv_preferred_stack_boundary_arg.
7906 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7907 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7908 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7909 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7910 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7911 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7913 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7916 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7917 of struct ix86_frame.
7918 (ix86_expand_epilogue): Likewise. Add a local variable for
7919 the reg_save_offset field in struct ix86_frame.
7921 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7923 PR tree-optimization/82604
7924 * tree-loop-distribution.c (enum partition_kind): New enum item
7925 PKIND_PARTIAL_MEMSET.
7926 (partition_builtin_p): Support above new enum item.
7927 (generate_code_for_partition): Ditto.
7928 (compute_access_range): Differentiate cases that equality can be
7929 proven at all loops, the innermost loops or no loops.
7930 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7931 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7932 (finalize_partitions, distribute_loop): Don't fuse partition of
7933 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7934 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7937 2018-01-23 Martin Liska <mliska@suse.cz>
7939 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7940 order to ignore the predictor.
7941 (PRED_POLYMORPHIC_CALL): Likewise.
7942 (PRED_RECURSIVE_CALL): Likewise.
7944 2018-01-23 Martin Liska <mliska@suse.cz>
7946 * tree-profile.c (tree_profiling): Print function header to
7947 aware reader which function we are working on.
7948 * value-prof.c (gimple_find_values_to_profile): Do not print
7949 not interesting value histograms.
7951 2018-01-23 Martin Liska <mliska@suse.cz>
7953 * profile-count.h (enum profile_quality): Add
7954 profile_uninitialized as the first value. Do not number values
7955 as they are zero based.
7956 (profile_count::verify): Update sanity check.
7957 (profile_probability::verify): Likewise.
7959 2018-01-23 Nathan Sidwell <nathan@acm.org>
7961 * doc/invoke.texi (ffor-scope): Deprecate.
7963 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7965 PR tree-optimization/83510
7966 * domwalk.c (set_all_edges_as_executable): New function.
7967 (dom_walker::dom_walker): Convert bool param
7968 "skip_unreachable_blocks" to enum reachability. Move setup of
7969 edge flags to set_all_edges_as_executable and only do it when
7970 reachability is REACHABLE_BLOCKS.
7971 * domwalk.h (enum dom_walker::reachability): New enum.
7972 (dom_walker::dom_walker): Convert bool param
7973 "skip_unreachable_blocks" to enum reachability.
7974 (set_all_edges_as_executable): New decl.
7975 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7976 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7978 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7979 but converting true to REACHABLE_BLOCKS.
7980 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7982 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7983 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7984 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7986 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7987 if check_all_array_refs will be called.
7989 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7991 * tree.c (selftest::test_location_wrappers): Add more test
7994 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7996 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7997 (selftest::test_bit_in_range): Likewise.
7999 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8002 * doc/sourcebuild.texi (vect_float): Say that the selector
8003 only describes the situation when -funsafe-math-optimizations is on.
8004 (vect_float_strict): Document.
8006 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8008 PR tree-optimization/83965
8009 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8010 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8011 instead of checking only for a reduction.
8012 (vect_recog_widen_sum_pattern): Likewise.
8014 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8016 * predict.c (probably_never_executed): Only use precise profile info.
8017 (compute_function_frequency): Skip after inlining hack since we now
8018 have quality checking.
8020 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8022 * profile-count.h (profile_probability::very_unlikely,
8023 profile_probability::unlikely, profile_probability::even): Set
8024 precision to guessed.
8026 2018-01-23 Richard Biener <rguenther@suse.de>
8028 PR tree-optimization/83963
8029 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8030 Properly terminate dominator walk when crossing the exit edge not
8031 when visiting its source block.
8033 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8036 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8037 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8039 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8041 PR tree-optimization/83957
8042 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8043 semicolon after for body surrounded by braces.
8045 PR tree-optimization/83081
8046 * profile-count.h (profile_probability::split): New method.
8047 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8048 Use profile_probability::split.
8049 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8050 when splitting a single conditional jump into 2.
8052 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8054 PR tree-optimization/69452
8055 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8058 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8060 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8061 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8062 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8064 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8066 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8067 * config/rl78/rl78.md (movdi): New define_expand.
8068 * config/rl78/rl78.c (rl78_split_movdi): New function.
8070 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8073 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8075 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8076 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8077 128-bit to produce an UNSPEC move to get the double word with the
8078 signbit and then a shift directly to do signbit.
8079 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8080 implementation with a new version that just does either a direct
8081 move or a regular move. Move memory interface to separate insns.
8082 Move insns so they are next to the expander.
8083 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8084 with signbit move. Split big and little endian case.
8085 (signbit<mode>2_dm_mem_le): Likewise.
8086 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8087 (signbit<mode>2_dm2): Likewise.
8089 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8091 * config/rl78/rl78.md (anddi3): New define_expand.
8093 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8095 * config/rl78/rl78.md (umindi3): New define_expand.
8097 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8099 * config/rl78/rl78.md (smindi3): New define_expand.
8101 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8103 * config/rl78/rl78.md (smaxdi3): New define_expand.
8105 2018-01-22 Carl Love <cel@us.ibm.com>
8107 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8108 LVX_V1TI): Add macro expansion.
8109 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8110 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8111 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8112 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8113 Change check to determine if the instruction is a byte reversing
8114 entry. Fix typo in comment.
8115 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8116 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8117 Add def_builtin calls for new builtins.
8118 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8119 Add define_insn expansion.
8121 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8123 * config/rl78/rl78.md (umaxdi3): New define_expand.
8125 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8127 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8128 for non-QImode registers.
8130 2018-01-22 Richard Biener <rguenther@suse.de>
8132 PR tree-optimization/83963
8133 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8134 including the loop exit block.
8135 (scop_detection::merge_sese): Likewise.
8136 (scop_detection::add_scop): Do it here instead.
8138 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8140 * doc/sourcebuild.texi (arm_softfloat): Document.
8142 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8145 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8146 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8147 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8149 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8150 David Edelsohn <dje.gcc@gmail.com>
8153 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8154 Change "crset eq" to "crset 2".
8155 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8156 (*call_indirect_aix<mode>_nospec): Likewise.
8157 (*call_value_indirect_aix<mode>_nospec): Likewise.
8158 (*call_indirect_elfv2<mode>_nospec): Likewise.
8159 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8160 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8161 change assembly output from . to $.
8162 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8163 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8164 (*tablejump<mode>_internal1_nospec): Likewise.
8166 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8169 * config/sh/sh_optimize_sett_clrt.cc:
8170 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8172 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8174 PR tree-optimization/83940
8175 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8176 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8177 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8178 use it to pass back the definition type.
8179 (vect_check_store_rhs): Likewise rhs_dt_out.
8180 (vect_build_gather_load_calls): Add a mask_dt argument and use
8181 it instead of a call to vect_is_simple_use.
8182 (vectorizable_store): Update calls to vect_check_load_store_mask
8183 and vect_check_store_rhs. Use the dt returned by the latter instead
8184 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8185 instead of calls to vect_is_simple_use. Pass the scalar rather
8186 than the vector operand to vect_is_simple_use when handling
8187 second and subsequent copies of an rhs value.
8188 (vectorizable_load): Update calls to vect_check_load_store_mask
8189 and vect_build_gather_load_calls. Use the cached mask_dt and
8190 gs_info.offset_dt instead of calls to vect_is_simple_use.
8192 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8195 * tree-emutls.c: Include gimplify.h.
8196 (lower_emutls_2): New function.
8197 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8198 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8199 it before further processing.
8202 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8203 UINTVAL (trueop1) instead of INTVAL (op1).
8205 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8209 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8210 INCOMING_FRAME_SP_OFFSET if not defined.
8211 (scan_trace): Add ENTRY argument. If true and
8212 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8213 emit a note to adjust the CFA offset.
8214 (create_cfi_notes): Adjust scan_trace callers.
8215 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8216 INCOMING_FRAME_SP_OFFSET in the CIE.
8217 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8218 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8220 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8221 * doc/tm.texi: Regenerated.
8223 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8225 PR rtl-optimization/83147
8226 * lra-constraints.c (remove_inheritance_pseudos): Use
8227 lra_substitute_pseudo_within_insn.
8229 2018-01-19 Tom de Vries <tom@codesourcery.com>
8230 Cesar Philippidis <cesar@codesourcery.com>
8233 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8235 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8238 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8239 spaces for function labels.
8241 2018-01-19 Martin Liska <mliska@suse.cz>
8243 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8244 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8245 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8246 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8247 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8248 (PRED_CONST_RETURN): Change from 69 to 65.
8249 (PRED_NULL_RETURN): Change from 91 to 71.
8250 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8251 (PRED_LOOP_GUARD): Change from 66 to 73.
8253 2018-01-19 Martin Liska <mliska@suse.cz>
8255 * predict.c (predict_insn_def): Add new assert.
8256 (struct branch_predictor): Change type to signed integer.
8257 (test_prediction_value_range): Amend test to cover
8259 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8260 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8261 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8262 (PRED_LOOP_IV_COMPARE): Likewise.
8263 * predict.h (PROB_UNINITIALIZED): Define new constant.
8265 2018-01-19 Martin Liska <mliska@suse.cz>
8267 * predict.c (dump_prediction): Add new format for
8268 analyze_brprob.py script which is enabled with -details
8270 * profile-count.h (precise_p): New function.
8272 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8274 PR tree-optimization/83922
8275 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8276 there are no statements that need masking.
8277 (vect_active_double_reduction_p): New function.
8278 (vect_analyze_loop_operations): Use it when handling phis that
8279 are not in the loop header.
8281 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8283 PR tree-optimization/83914
8284 * tree-vect-loop.c (vectorizable_induction): Don't convert
8285 init_expr or apply the peeling adjustment for inductions
8286 that are nested within the vectorized loop.
8288 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8290 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8293 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8297 * function.h (gimplify_parameters): Add gimple_seq * argument.
8298 * function.c: Include gimple.h and options.h.
8299 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8300 for the added local temporaries if needed.
8301 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8302 if there are any parameter cleanups, wrap whole body into a
8303 try/finally with the cleanups.
8305 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8308 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8309 Use GET_MODE_CLASS for scalar floating point.
8311 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8315 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8316 Fix call of call_cgraph_insertion_hooks.
8318 2018-01-18 Martin Sebor <msebor@redhat.com>
8320 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8322 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8325 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8328 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8331 * common.opt: (-ffile-prefix-map): New option.
8332 * opts.c (common_handle_option): Defer it.
8333 * opts-global.c (handle_common_deferred_options): Handle it.
8334 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8335 * file-prefix-map.h: New file.
8336 (remap_debug_filename, add_debug_prefix_map): ...here.
8337 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8338 * final.c (debug_prefix_map, add_debug_prefix_map
8339 remap_debug_filename): Move to...
8340 * file-prefix-map.c: New file.
8341 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8342 generalize, get rid of alloca(), use strrchr() instead of strchr().
8343 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8344 Implement in terms of add_prefix_map().
8345 (remap_macro_filename, remap_debug_filename): Implement in term of
8347 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8348 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8349 * dbxout.c: Include file-prefix-map.h.
8350 * varasm.c: Likewise.
8351 * vmsdbgout.c: Likewise.
8352 * xcoffout.c: Likewise.
8353 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8354 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8355 * doc/invoke.texi (-ffile-prefix-map): Document.
8356 (-fdebug-prefix-map): Update description.
8358 2018-01-18 Martin Liska <mliska@suse.cz>
8360 * config/i386/i386.c (indirect_thunk_name): Document that also
8362 (output_indirect_thunk): Document why both instructions
8363 (pause and lfence) are generated.
8365 2018-01-18 Richard Biener <rguenther@suse.de>
8367 PR tree-optimization/83887
8368 * graphite-scop-detection.c
8369 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8370 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8371 (scop_detection::merge_sese): Re-implement with a flood-fill
8372 algorithm that properly finds a SESE region if it exists.
8374 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8377 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8378 pointer_diff optimizations use view_convert instead of convert.
8380 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8382 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8383 Generate different code for -mno-speculate-indirect-jumps.
8384 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8385 (*call_indirect_aix<mode>): Disable for
8386 -mno-speculate-indirect-jumps.
8387 (*call_indirect_aix<mode>_nospec): New define_insn.
8388 (*call_value_indirect_aix<mode>): Disable for
8389 -mno-speculate-indirect-jumps.
8390 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8391 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8392 -mno-speculate-indirect-jumps.
8393 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8395 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8397 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8398 long double type, set the flags for noting the default long double
8399 type, even if we don't pass or return a long double type.
8401 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8404 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8407 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8409 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8410 support for merge[hl].
8411 (fold_mergehl_helper): New helper function.
8412 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8413 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8414 (altivec_vmrglw_direct): Add xxmrglw insn.
8416 2018-01-17 Andrew Waterman <andrew@sifive.com>
8418 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8419 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8421 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8424 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8425 call the lto_location_cache before reading the
8426 DECL_SOURCE_LOCATION of the types.
8428 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8429 Richard Sandiford <richard.sandiford@linaro.org>
8431 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8432 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8433 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8434 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8435 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8437 * config/aarch64/constraints.md (aarch64_movti_operand):
8439 * config/aarch64/predicates.md (Uti): Add new constraint.
8441 2018-01-17 Carl Love <cel@us.ibm.com>
8443 * config/rs6000/vsx.md (define_expand xl_len_r,
8444 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8445 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8447 (define_expand, define_insn): Move the shift left from the
8448 define_insn to the define_expand for lxvl and stxvl instructions.
8449 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8450 and XL_LEN_R definitions to PURE.
8452 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8454 * config/i386/i386.c (indirect_thunk_name): Declare regno
8455 as unsigned int. Compare regno with INVALID_REGNUM.
8456 (output_indirect_thunk): Ditto.
8457 (output_indirect_thunk_function): Ditto.
8458 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8459 in the call to output_indirect_thunk_function.
8461 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8464 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8465 rather than the size of inner_type to determine the stack slot size
8466 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8468 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8471 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8474 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8476 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8477 endian Linux systems to optionally enable multilibs for selecting
8478 the long double type if the user configured an explicit type.
8479 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8480 have no long double multilibs if not defined.
8481 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8482 warn if the user used -mabi={ieee,ibm}longdouble and we built
8483 multilibs for long double.
8484 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8485 appropriate multilib option.
8486 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8488 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8489 for building long double multilibs.
8490 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8492 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8494 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8497 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8499 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8502 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8505 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8508 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8510 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8511 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8513 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8515 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8516 different rtl trees depending on TARGET_64BIT.
8517 (rs6000_gen_lvx): Likewise.
8519 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8521 * config/visium/visium.md (nop): Tweak comment.
8522 (hazard_nop): Likewise.
8524 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8526 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8527 -mspeculate-indirect-jumps.
8528 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8529 for -mno-speculate-indirect-jumps.
8530 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8531 (*call_value_indirect_elfv2<mode>): Disable for
8532 -mno-speculate-indirect-jumps.
8533 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8534 (indirect_jump): Emit different RTL for
8535 -mno-speculate-indirect-jumps.
8536 (*indirect_jump<mode>): Disable for
8537 -mno-speculate-indirect-jumps.
8538 (*indirect_jump<mode>_nospec): New define_insn.
8539 (tablejump): Emit different RTL for
8540 -mno-speculate-indirect-jumps.
8541 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8542 (tablejumpsi_nospec): New define_expand.
8543 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8544 (tablejumpdi_nospec): New define_expand.
8545 (*tablejump<mode>_internal1): Disable for
8546 -mno-speculate-indirect-jumps.
8547 (*tablejump<mode>_internal1_nospec): New define_insn.
8548 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8551 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8553 * caller-save.c (insert_save): Drop unnecessary parameter. All
8556 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8557 Richard Biener <rguenth@suse.de>
8560 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8561 return early, inline manually is_gimple_sizepos. Make sure if we
8562 call gimplify_expr we don't end up with a gimple constant.
8563 * tree.c (variably_modified_type_p): Don't return true for
8564 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8565 * gimplify.h (is_gimple_sizepos): Remove.
8567 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8569 PR tree-optimization/83857
8570 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8571 vectorizable_live_operation for pure SLP statements.
8572 (vectorizable_live_operation): Handle PHIs.
8574 2018-01-16 Richard Biener <rguenther@suse.de>
8576 PR tree-optimization/83867
8577 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8578 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8580 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8583 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8584 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8585 If off is not INTEGER_CST, issue a may not be aligned warning
8586 rather than isn't aligned. Use isn%'t rather than isn't.
8587 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8589 <case MULT_EXPR>: Improve the case when bottom and one of the
8590 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8591 operand, in that case check if the other operand is multiple of
8592 bottom divided by the INTEGER_CST operand.
8594 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8597 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8598 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8599 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8600 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8601 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8602 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8603 (pa_function_arg_size): New function.
8605 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8607 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8608 in a separate statement.
8610 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8612 PR tree-optimization/83847
8613 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8614 group gathers and scatters.
8616 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8618 PR rtl-optimization/86620
8619 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8621 PR rtl-optimization/83213
8622 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8623 to last if both are JUMP_INSNs.
8625 PR tree-optimization/83843
8626 * gimple-ssa-store-merging.c
8627 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8628 store_immediate_info for bswap/nop orig_stores.
8630 2018-01-15 Andrew Waterman <andrew@sifive.com>
8632 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8634 <UDIV>: Increase cost if !TARGET_DIV.
8636 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8638 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8639 (define_attr "cr_logical_3op"): New.
8640 (cceq_ior_compare): Adjust.
8641 (cceq_ior_compare_complement): Adjust.
8642 (*cceq_rev_compare): Adjust.
8643 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8644 (is_cracked_insn): Adjust.
8645 (insn_must_be_first_in_group): Adjust.
8646 * config/rs6000/40x.md: Adjust.
8647 * config/rs6000/440.md: Adjust.
8648 * config/rs6000/476.md: Adjust.
8649 * config/rs6000/601.md: Adjust.
8650 * config/rs6000/603.md: Adjust.
8651 * config/rs6000/6xx.md: Adjust.
8652 * config/rs6000/7450.md: Adjust.
8653 * config/rs6000/7xx.md: Adjust.
8654 * config/rs6000/8540.md: Adjust.
8655 * config/rs6000/cell.md: Adjust.
8656 * config/rs6000/e300c2c3.md: Adjust.
8657 * config/rs6000/e500mc.md: Adjust.
8658 * config/rs6000/e500mc64.md: Adjust.
8659 * config/rs6000/e5500.md: Adjust.
8660 * config/rs6000/e6500.md: Adjust.
8661 * config/rs6000/mpc.md: Adjust.
8662 * config/rs6000/power4.md: Adjust.
8663 * config/rs6000/power5.md: Adjust.
8664 * config/rs6000/power6.md: Adjust.
8665 * config/rs6000/power7.md: Adjust.
8666 * config/rs6000/power8.md: Adjust.
8667 * config/rs6000/power9.md: Adjust.
8668 * config/rs6000/rs64.md: Adjust.
8669 * config/rs6000/titan.md: Adjust.
8671 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8673 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8674 ix86_indirect_branch_register logic.
8676 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8678 * config/i386/constraints.md (Bs): Update
8679 ix86_indirect_branch_register check. Don't check
8680 ix86_indirect_branch_register with GOT_memory_operand.
8682 * config/i386/predicates.md (GOT_memory_operand): Don't check
8683 ix86_indirect_branch_register here.
8684 (GOT32_symbol_operand): Likewise.
8686 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8688 * config/i386/predicates.md (constant_call_address_operand):
8689 Rewrite ix86_indirect_branch_register logic.
8690 (sibcall_insn_operand): Likewise.
8692 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8694 * config/i386/constraints.md (Bs): Replace
8695 ix86_indirect_branch_thunk_register with
8696 ix86_indirect_branch_register.
8698 * config/i386/i386.md (indirect_jump): Likewise.
8699 (tablejump): Likewise.
8700 (*sibcall_memory): Likewise.
8701 (*sibcall_value_memory): Likewise.
8702 Peepholes of indirect call and jump via memory: Likewise.
8703 * config/i386/i386.opt: Likewise.
8704 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8705 (GOT_memory_operand): Likewise.
8706 (call_insn_operand): Likewise.
8707 (sibcall_insn_operand): Likewise.
8708 (GOT32_symbol_operand): Likewise.
8710 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8713 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8714 type rather than type addr's type points to.
8715 (expand_omp_atomic_mutex): Likewise.
8716 (expand_omp_atomic): Likewise.
8718 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8721 * config/i386/i386.c (output_indirect_thunk_function): Use
8722 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8723 for __x86_return_thunk.
8725 2018-01-15 Richard Biener <rguenther@suse.de>
8728 * expmed.c (extract_bit_field_1): Fix typo.
8730 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8733 * config/arm/iterators.md (VF): New mode iterator.
8734 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8735 Remove integer-related logic from pattern.
8736 (neon_vabd<mode>_3): Likewise.
8738 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8741 * common.opt (fstrict-overflow): No longer an alias.
8742 (fwrapv-pointer): New option.
8743 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8744 also for pointer types based on flag_wrapv_pointer.
8745 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8746 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8747 opts->x_flag_wrapv got set.
8748 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8749 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8750 POINTER_TYPE_OVERFLOW_UNDEFINED.
8751 * match.pd: Likewise in address comparison pattern.
8752 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8754 2018-01-15 Richard Biener <rguenther@suse.de>
8757 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8758 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8759 Reset type names to their identifier if their TYPE_DECL doesn't
8760 have linkage (and thus is used for ODR and devirt).
8761 (save_debug_info_for_decl): Remove.
8762 (save_debug_info_for_type): Likewise.
8763 (add_tree_to_fld_list): Adjust.
8764 * tree-pretty-print.c (dump_generic_node): Make dumping of
8765 type names more robust.
8767 2018-01-15 Richard Biener <rguenther@suse.de>
8769 * BASE-VER: Bump to 8.0.1.
8771 2018-01-14 Martin Sebor <msebor@redhat.com>
8774 * builtins.c (check_access): Avoid warning when the no-warning bit
8777 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8779 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8780 * ira-color (allocno_hard_regs_compare): Likewise.
8782 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8785 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8786 Use .pushsection/.popsection.
8788 2018-01-14 Martin Sebor <msebor@redhat.com>
8791 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8793 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8795 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8796 entry from extra_headers.
8797 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8798 extra_headers, make the list bitwise identical to the i?86-*-* one.
8800 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8802 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8803 -mcmodel=large with -mindirect-branch=thunk,
8804 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8805 -mfunction-return=thunk-extern.
8806 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8807 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8808 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8810 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8812 * config/i386/i386.c (print_reg): Print the name of the full
8813 integer register without '%'.
8814 (ix86_print_operand): Handle 'V'.
8815 * doc/extend.texi: Document 'V' modifier.
8817 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8819 * config/i386/constraints.md (Bs): Disallow memory operand for
8820 -mindirect-branch-register.
8822 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8823 (GOT_memory_operand): Likewise.
8824 (call_insn_operand): Likewise.
8825 (sibcall_insn_operand): Likewise.
8826 (GOT32_symbol_operand): Likewise.
8827 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8828 for -mindirect-branch-register.
8829 (tablejump): Likewise.
8830 (*sibcall_memory): Likewise.
8831 (*sibcall_value_memory): Likewise.
8832 Disallow peepholes of indirect call and jump via memory for
8833 -mindirect-branch-register.
8834 (*call_pop): Replace m with Bw.
8835 (*call_value_pop): Likewise.
8836 (*sibcall_pop_memory): Replace m with Bs.
8837 * config/i386/i386.opt (mindirect-branch-register): New option.
8838 * doc/invoke.texi: Document -mindirect-branch-register option.
8840 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8842 * config/i386/i386-protos.h (ix86_output_function_return): New.
8843 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8844 set function_return_type.
8845 (indirect_thunk_name): Add ret_p to indicate thunk for function
8847 (output_indirect_thunk_function): Pass false to
8848 indirect_thunk_name.
8849 (ix86_output_indirect_branch_via_reg): Likewise.
8850 (ix86_output_indirect_branch_via_push): Likewise.
8851 (output_indirect_thunk_function): Create alias for function
8852 return thunk if regno < 0.
8853 (ix86_output_function_return): New function.
8854 (ix86_handle_fndecl_attribute): Handle function_return.
8855 (ix86_attribute_table): Add function_return.
8856 * config/i386/i386.h (machine_function): Add
8857 function_return_type.
8858 * config/i386/i386.md (simple_return_internal): Use
8859 ix86_output_function_return.
8860 (simple_return_internal_long): Likewise.
8861 * config/i386/i386.opt (mfunction-return=): New option.
8862 (indirect_branch): Mention -mfunction-return=.
8863 * doc/extend.texi: Document function_return function attribute.
8864 * doc/invoke.texi: Document -mfunction-return= option.
8866 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8868 * config/i386/i386-opts.h (indirect_branch): New.
8869 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8870 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8871 with local indirect jump when converting indirect call and jump.
8872 (ix86_set_indirect_branch_type): New.
8873 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8874 (indirectlabelno): New.
8875 (indirect_thunk_needed): Likewise.
8876 (indirect_thunk_bnd_needed): Likewise.
8877 (indirect_thunks_used): Likewise.
8878 (indirect_thunks_bnd_used): Likewise.
8879 (INDIRECT_LABEL): Likewise.
8880 (indirect_thunk_name): Likewise.
8881 (output_indirect_thunk): Likewise.
8882 (output_indirect_thunk_function): Likewise.
8883 (ix86_output_indirect_branch_via_reg): Likewise.
8884 (ix86_output_indirect_branch_via_push): Likewise.
8885 (ix86_output_indirect_branch): Likewise.
8886 (ix86_output_indirect_jmp): Likewise.
8887 (ix86_code_end): Call output_indirect_thunk_function if needed.
8888 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8890 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8891 (ix86_attribute_table): Add indirect_branch.
8892 * config/i386/i386.h (machine_function): Add indirect_branch_type
8893 and has_local_indirect_jump.
8894 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8896 (tablejump): Likewise.
8897 (*indirect_jump): Use ix86_output_indirect_jmp.
8898 (*tablejump_1): Likewise.
8899 (simple_return_indirect_internal): Likewise.
8900 * config/i386/i386.opt (mindirect-branch=): New option.
8901 (indirect_branch): New.
8904 (thunk-inline): Likewise.
8905 (thunk-extern): Likewise.
8906 * doc/extend.texi: Document indirect_branch function attribute.
8907 * doc/invoke.texi: Document -mindirect-branch= option.
8909 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8912 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8914 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8916 * ipa-inline.c (want_inline_small_function_p): Return false if
8917 inlining has already failed with CIF_FINAL_ERROR.
8918 (update_caller_keys): Call want_inline_small_function_p before
8920 (update_callee_keys): Likewise.
8922 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8924 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8926 (rs6000_quadword_masked_address_p): Likewise.
8927 (quad_aligned_load_p): Likewise.
8928 (quad_aligned_store_p): Likewise.
8929 (const_load_sequence_p): Add comment to describe the outer-most loop.
8930 (mimic_memory_attributes_and_flags): New function.
8931 (rs6000_gen_stvx): Likewise.
8932 (replace_swapped_aligned_store): Likewise.
8933 (rs6000_gen_lvx): Likewise.
8934 (replace_swapped_aligned_load): Likewise.
8935 (replace_swapped_load_constant): Capitalize argument name in
8936 comment describing this function.
8937 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8938 and stores that access quad-word aligned addresses and replace
8939 with stvx or lvx instructions when appropriate.
8940 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8941 New function prototype.
8942 (rs6000_quadword_masked_address_p): Likewise.
8943 (rs6000_gen_lvx): Likewise.
8944 (rs6000_gen_stvx): Likewise.
8945 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8946 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8947 when memory address is aligned.
8948 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8949 this split to select lvx instruction when memory address is aligned.
8950 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8951 instruction when memory address is aligned.
8952 (*vsx_le_perm_load_v16qi): Likewise.
8953 (four unnamed splitters): Modify to select the stvx instruction
8954 when memory is aligned.
8956 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8958 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8959 which appears in the queue multiple times.
8961 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8962 Alan Hayward <alan.hayward@arm.com>
8963 David Sherwood <david.sherwood@arm.com>
8965 * tree-vectorizer.h (vec_lower_bound): New structure.
8966 (_loop_vec_info): Add check_nonzero and lower_bounds.
8967 (LOOP_VINFO_CHECK_NONZERO): New macro.
8968 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8969 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8970 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8971 fields. Make seg_len the distance travelled, not including the
8973 (dr_direction_indicator): Declare.
8974 (dr_zero_step_indicator): Likewise.
8975 (dr_known_forward_stride_p): Likewise.
8976 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8978 (runtime_alias_check_p): Allow runtime alias checks with
8980 (operator ==): Compare access_size and align.
8981 (prune_runtime_alias_test_list): Rework for new distinction between
8982 the access_size and seg_len.
8983 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8985 (get_segment_min_max): New function.
8986 (create_intersect_range_checks): Use it.
8987 (dr_step_indicator): New function.
8988 (dr_direction_indicator): Likewise.
8989 (dr_zero_step_indicator): Likewise.
8990 (dr_known_forward_stride_p): Likewise.
8991 * tree-loop-distribution.c (data_ref_segment_size): Return
8992 DR_STEP * (niters - 1).
8993 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8995 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8996 (vect_preserves_scalar_order_p): New function, split out from...
8997 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8998 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8999 (vect_vfa_access_size): New function.
9000 (vect_vfa_align): Likewise.
9001 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9002 (dump_lower_bound): New function.
9003 (vect_check_lower_bound): Likewise.
9004 (vect_small_gap_p): Likewise.
9005 (vectorizable_with_step_bound_p): Likewise.
9006 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9007 depencies if the vectorization factor is 1. Convert the checks
9008 for nonzero steps into checks on the bounds of DR_STEP. Try using
9009 a bunds check for variable steps if the minimum required step is
9010 relatively small. Update calls to the dr_with_seg_len
9011 constructor and to vect_compile_time_alias.
9012 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9014 (vect_loop_versioning): Call it.
9015 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9017 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9020 Alan Hayward <alan.hayward@arm.com>
9021 David Sherwood <david.sherwood@arm.com>
9023 * doc/sourcebuild.texi (vect_scatter_store): Document.
9024 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9026 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9028 * genopinit.c (main): Add supports_vec_scatter_store and
9029 supports_vec_scatter_store_cached to target_optabs.
9030 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9031 IFN_MASK_SCATTER_STORE.
9032 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9034 * internal-fn.h (internal_store_fn_p): Declare.
9035 (internal_fn_stored_value_index): Likewise.
9036 * internal-fn.c (scatter_store_direct): New macro.
9037 (expand_scatter_store_optab_fn): New function.
9038 (direct_scatter_store_optab_supported_p): New macro.
9039 (internal_store_fn_p): New function.
9040 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9041 IFN_MASK_SCATTER_STORE.
9042 (internal_fn_mask_index): Likewise.
9043 (internal_fn_stored_value_index): New function.
9044 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9046 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9047 * optabs-query.c (supports_vec_scatter_store_p): New function.
9048 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9049 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9050 true for scatter stores.
9051 (vect_gather_scatter_fn_p): Handle scatter stores too.
9052 (vect_check_gather_scatter): Consider using scatter stores if
9053 supports_vec_scatter_store_p.
9054 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9056 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9057 internal_fn_stored_value_index.
9058 (check_load_store_masking): Handle scatter stores too.
9059 (vect_get_store_rhs): Make public.
9060 (vectorizable_call): Use internal_store_fn_p.
9061 (vectorizable_store): Handle scatter store internal functions.
9062 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9063 when deciding whether the end of the group has been reached.
9064 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9065 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9066 (mask_scatter_store<mode>): New insns.
9068 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9069 Alan Hayward <alan.hayward@arm.com>
9070 David Sherwood <david.sherwood@arm.com>
9072 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9073 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9074 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9076 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9077 Use vect_truncate_gather_scatter_offset if we can't treat the
9078 operation as a normal gather load or scatter store.
9079 (get_group_load_store_type): Take the gather_scatter_info
9080 as argument. Try using a gather load or scatter store for
9081 single-element groups.
9082 (get_load_store_type): Update calls to get_group_load_store_type
9083 and vect_use_strided_gather_scatters_p.
9085 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9086 Alan Hayward <alan.hayward@arm.com>
9087 David Sherwood <david.sherwood@arm.com>
9089 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9090 optional tree argument.
9091 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9093 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9094 but continue to use the current value as a fallback.
9095 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9096 to compare the updates.
9097 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9098 (get_load_store_type): Use it when handling a strided access.
9099 (vect_get_strided_load_store_ops): New function.
9100 (vect_get_data_ptr_increment): Likewise.
9101 (vectorizable_load): Handle strided gather loads. Always pass
9102 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9104 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9105 Alan Hayward <alan.hayward@arm.com>
9106 David Sherwood <david.sherwood@arm.com>
9108 * doc/md.texi (gather_load@var{m}): Document.
9109 (mask_gather_load@var{m}): Likewise.
9110 * genopinit.c (main): Add supports_vec_gather_load and
9111 supports_vec_gather_load_cached to target_optabs.
9112 * optabs-tree.c (init_tree_optimization_optabs): Use
9113 ggc_cleared_alloc to allocate target_optabs.
9114 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9115 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9117 * internal-fn.h (internal_load_fn_p): Declare.
9118 (internal_gather_scatter_fn_p): Likewise.
9119 (internal_fn_mask_index): Likewise.
9120 (internal_gather_scatter_fn_supported_p): Likewise.
9121 * internal-fn.c (gather_load_direct): New macro.
9122 (expand_gather_load_optab_fn): New function.
9123 (direct_gather_load_optab_supported_p): New macro.
9124 (direct_internal_fn_optab): New function.
9125 (internal_load_fn_p): Likewise.
9126 (internal_gather_scatter_fn_p): Likewise.
9127 (internal_fn_mask_index): Likewise.
9128 (internal_gather_scatter_fn_supported_p): Likewise.
9129 * optabs-query.c (supports_at_least_one_mode_p): New function.
9130 (supports_vec_gather_load_p): Likewise.
9131 * optabs-query.h (supports_vec_gather_load_p): Declare.
9132 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9133 and memory_type field.
9134 (NUM_PATTERNS): Bump to 15.
9135 * tree-vect-data-refs.c: Include internal-fn.h.
9136 (vect_gather_scatter_fn_p): New function.
9137 (vect_describe_gather_scatter_call): Likewise.
9138 (vect_check_gather_scatter): Try using internal functions for
9139 gather loads. Recognize existing calls to a gather load function.
9140 (vect_analyze_data_refs): Consider using gather loads if
9141 supports_vec_gather_load_p.
9142 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9143 (vect_get_gather_scatter_offset_type): Likewise.
9144 (vect_convert_mask_for_vectype): Likewise.
9145 (vect_add_conversion_to_patterm): Likewise.
9146 (vect_try_gather_scatter_pattern): Likewise.
9147 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9148 (vect_vect_recog_func_ptrs): Add it.
9149 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9150 internal_fn_mask_index and internal_gather_scatter_fn_p.
9151 (check_load_store_masking): Take the gather_scatter_info as an
9152 argument and handle gather loads.
9153 (vect_get_gather_scatter_ops): New function.
9154 (vectorizable_call): Check internal_load_fn_p.
9155 (vectorizable_load): Likewise. Handle gather load internal
9157 (vectorizable_store): Update call to check_load_store_masking.
9158 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9159 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9160 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9161 (aarch64_gather_scale_operand_d): New predicates.
9162 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9163 (mask_gather_load<mode>): New insns.
9165 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9166 Alan Hayward <alan.hayward@arm.com>
9167 David Sherwood <david.sherwood@arm.com>
9169 * optabs.def (fold_left_plus_optab): New optab.
9170 * doc/md.texi (fold_left_plus_@var{m}): Document.
9171 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9172 * internal-fn.c (fold_left_direct): Define.
9173 (expand_fold_left_optab_fn): Likewise.
9174 (direct_fold_left_optab_supported_p): Likewise.
9175 * fold-const-call.c (fold_const_fold_left): New function.
9176 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9177 * tree-parloops.c (valid_reduction_p): New function.
9178 (gather_scalar_reductions): Use it.
9179 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9180 (vect_finish_replace_stmt): Declare.
9181 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9182 (needs_fold_left_reduction_p): New function, split out from...
9183 (vect_is_simple_reduction): ...here. Accept reductions that
9184 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9185 (vect_force_simple_reduction): Also store the reduction type in
9186 the assignment's STMT_VINFO_REDUC_TYPE.
9187 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9188 (merge_with_identity): New function.
9189 (vect_expand_fold_left): Likewise.
9190 (vectorize_fold_left_reduction): Likewise.
9191 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9192 scalar phi in place for it. Check for target support and reject
9193 cases that would reassociate the operation. Defer the transform
9194 phase to vectorize_fold_left_reduction.
9195 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9196 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9197 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9199 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9201 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9202 call to ifc_temp_var.
9204 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9205 Alan Hayward <alan.hayward@arm.com>
9206 David Sherwood <david.sherwood@arm.com>
9208 * target.def (legitimize_address_displacement): Take the original
9209 offset as a poly_int.
9210 * targhooks.h (default_legitimize_address_displacement): Update
9212 * targhooks.c (default_legitimize_address_displacement): Likewise.
9213 * doc/tm.texi: Regenerate.
9214 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9215 as an argument, moving assert of ad->disp == ad->disp_term to...
9216 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9217 Try calling targetm.legitimize_address_displacement before expanding
9218 the address rather than afterwards, and adjust for the new interface.
9219 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9220 Match the new hook interface. Handle SVE addresses.
9221 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9224 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9226 * Makefile.in (OBJS): Add early-remat.o.
9227 * target.def (select_early_remat_modes): New hook.
9228 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9229 * doc/tm.texi: Regenerate.
9230 * targhooks.h (default_select_early_remat_modes): Declare.
9231 * targhooks.c (default_select_early_remat_modes): New function.
9232 * timevar.def (TV_EARLY_REMAT): New timevar.
9233 * passes.def (pass_early_remat): New pass.
9234 * tree-pass.h (make_pass_early_remat): Declare.
9235 * early-remat.c: New file.
9236 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9238 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9240 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9241 Alan Hayward <alan.hayward@arm.com>
9242 David Sherwood <david.sherwood@arm.com>
9244 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9245 vfm1 with a bound_epilog parameter.
9246 (vect_do_peeling): Update calls accordingly, and move the prologue
9247 call earlier in the function. Treat the base bound_epilog as 0 for
9248 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9249 this base when peeling for gaps.
9250 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9251 with fully-masked loops.
9252 (vect_estimate_min_profitable_iters): Handle the single peeled
9253 iteration in that case.
9255 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9256 Alan Hayward <alan.hayward@arm.com>
9257 David Sherwood <david.sherwood@arm.com>
9259 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9260 single-element interleaving even if the size is not a power of 2.
9261 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9262 accesses for single-element interleaving if the group size is
9265 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9266 Alan Hayward <alan.hayward@arm.com>
9267 David Sherwood <david.sherwood@arm.com>
9269 * doc/md.texi (fold_extract_last_@var{m}): Document.
9270 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9271 * optabs.def (fold_extract_last_optab): New optab.
9272 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9273 * internal-fn.c (fold_extract_direct): New macro.
9274 (expand_fold_extract_optab_fn): Likewise.
9275 (direct_fold_extract_optab_supported_p): Likewise.
9276 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9277 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9278 EXTRACT_LAST_REDUCTION.
9279 (get_initial_def_for_reduction): Do not create an initial vector
9280 for EXTRACT_LAST_REDUCTION reductions.
9281 (vectorizable_reduction): Leave the scalar phi in place for
9282 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9283 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9284 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9285 transform phase to vectorizable_condition.
9286 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9288 (vect_finish_stmt_generation): ...here.
9289 (vect_finish_replace_stmt): New function.
9290 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9291 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9293 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9295 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9296 Alan Hayward <alan.hayward@arm.com>
9297 David Sherwood <david.sherwood@arm.com>
9299 * doc/md.texi (extract_last_@var{m}): Document.
9300 * optabs.def (extract_last_optab): New optab.
9301 * internal-fn.def (EXTRACT_LAST): New internal function.
9302 * internal-fn.c (cond_unary_direct): New macro.
9303 (expand_cond_unary_optab_fn): Likewise.
9304 (direct_cond_unary_optab_supported_p): Likewise.
9305 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9306 loops using EXTRACT_LAST.
9307 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9308 (extract_last_<mode>): ...this optab.
9309 (vec_extract<mode><Vel>): Update accordingly.
9311 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9312 Alan Hayward <alan.hayward@arm.com>
9313 David Sherwood <david.sherwood@arm.com>
9315 * target.def (empty_mask_is_expensive): New hook.
9316 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9317 * doc/tm.texi: Regenerate.
9318 * targhooks.h (default_empty_mask_is_expensive): Declare.
9319 * targhooks.c (default_empty_mask_is_expensive): New function.
9320 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9321 if the target says that empty masks are expensive.
9322 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9324 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9326 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9327 Alan Hayward <alan.hayward@arm.com>
9328 David Sherwood <david.sherwood@arm.com>
9330 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9331 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9332 (vect_use_loop_mask_for_alignment_p): New function.
9333 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9334 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9335 niters_skip argument. Make sure that the first niters_skip elements
9336 of the first iteration are inactive.
9337 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9338 Update call to vect_set_loop_masks_directly.
9339 (get_misalign_in_elems): New function, split out from...
9340 (vect_gen_prolog_loop_niters): ...here.
9341 (vect_update_init_of_dr): Take a code argument that specifies whether
9342 the adjustment should be added or subtracted.
9343 (vect_update_init_of_drs): Likewise.
9344 (vect_prepare_for_masked_peels): New function.
9345 (vect_do_peeling): Skip prologue peeling if we're using a mask
9346 instead. Update call to vect_update_inits_of_drs.
9347 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9349 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9350 alignment. Do not include the number of peeled iterations in
9351 the minimum threshold in that case.
9352 (vectorizable_induction): Adjust the start value down by
9353 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9354 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9355 Take the number of skipped iterations into account when calculating
9357 * tree-vect-stmts.c (vect_gen_while_not): New function.
9359 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9360 Alan Hayward <alan.hayward@arm.com>
9361 David Sherwood <david.sherwood@arm.com>
9363 * doc/sourcebuild.texi (vect_fully_masked): Document.
9364 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9366 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9368 (vect_analyze_loop_2): ...here. Don't check the vectorization
9369 factor against the number of loop iterations if the loop is
9372 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9373 Alan Hayward <alan.hayward@arm.com>
9374 David Sherwood <david.sherwood@arm.com>
9376 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9377 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9378 (dump_groups): Update accordingly.
9379 (iv_use::mem_type): New member variable.
9380 (address_p): New function.
9381 (record_use): Add a mem_type argument and initialize the new
9383 (record_group_use): Add a mem_type argument. Use address_p.
9384 Remove obsolete null checks of base_object. Update call to record_use.
9385 (find_interesting_uses_op): Update call to record_group_use.
9386 (find_interesting_uses_cond): Likewise.
9387 (find_interesting_uses_address): Likewise.
9388 (get_mem_type_for_internal_fn): New function.
9389 (find_address_like_use): Likewise.
9390 (find_interesting_uses_stmt): Try find_address_like_use before
9391 calling find_interesting_uses_op.
9392 (addr_offset_valid_p): Use the iv mem_type field as the type
9393 of the addressed memory.
9394 (add_autoinc_candidates): Likewise.
9395 (get_address_cost): Likewise.
9396 (split_small_address_groups_p): Use address_p.
9397 (split_address_groups): Likewise.
9398 (add_iv_candidate_for_use): Likewise.
9399 (autoinc_possible_for_pair): Likewise.
9400 (rewrite_groups): Likewise.
9401 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9402 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9403 (get_alias_ptr_type_for_ptr_address): New function.
9404 (rewrite_use_address): Rewrite address uses in calls that were
9405 identified by find_address_like_use.
9407 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9408 Alan Hayward <alan.hayward@arm.com>
9409 David Sherwood <david.sherwood@arm.com>
9411 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9413 * gimple-expr.h (is_gimple_addressable: Likewise.
9414 * gimple-expr.c (is_gimple_address): Likewise.
9415 * internal-fn.c (expand_call_mem_ref): New function.
9416 (expand_mask_load_optab_fn): Use it.
9417 (expand_mask_store_optab_fn): Likewise.
9419 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9420 Alan Hayward <alan.hayward@arm.com>
9421 David Sherwood <david.sherwood@arm.com>
9423 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9424 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9425 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9426 (cond_umax@var{mode}): Document.
9427 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9428 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9429 (cond_umin_optab, cond_umax_optab): New optabs.
9430 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9431 (COND_IOR, COND_XOR): New internal functions.
9432 * internal-fn.h (get_conditional_internal_fn): Declare.
9433 * internal-fn.c (cond_binary_direct): New macro.
9434 (expand_cond_binary_optab_fn): Likewise.
9435 (direct_cond_binary_optab_supported_p): Likewise.
9436 (get_conditional_internal_fn): New function.
9437 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9438 Cope with reduction statements that are vectorized as calls rather
9440 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9441 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9442 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9443 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9444 (UNSPEC_COND_EOR): New unspecs.
9445 (optab): Add mappings for them.
9446 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9447 (sve_int_op, sve_fp_op): New int attributes.
9449 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9450 Alan Hayward <alan.hayward@arm.com>
9451 David Sherwood <david.sherwood@arm.com>
9453 * optabs.def (while_ult_optab): New optab.
9454 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9455 * internal-fn.def (WHILE_ULT): New internal function.
9456 * internal-fn.h (direct_internal_fn_supported_p): New override
9457 that takes two types as argument.
9458 * internal-fn.c (while_direct): New macro.
9459 (expand_while_optab_fn): New function.
9460 (convert_optab_supported_p): Likewise.
9461 (direct_while_optab_supported_p): New macro.
9462 * wide-int.h (wi::udiv_ceil): New function.
9463 * tree-vectorizer.h (rgroup_masks): New structure.
9464 (vec_loop_masks): New typedef.
9465 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9467 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9468 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9469 (vect_max_vf): New function.
9470 (slpeel_make_loop_iterate_ntimes): Delete.
9471 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9472 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9473 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9474 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9475 internal-fn.h, stor-layout.h and optabs-query.h.
9476 (vect_set_loop_mask): New function.
9477 (add_preheader_seq): Likewise.
9478 (add_header_seq): Likewise.
9479 (interleave_supported_p): Likewise.
9480 (vect_maybe_permute_loop_masks): Likewise.
9481 (vect_set_loop_masks_directly): Likewise.
9482 (vect_set_loop_condition_masked): Likewise.
9483 (vect_set_loop_condition_unmasked): New function, split out from
9484 slpeel_make_loop_iterate_ntimes.
9485 (slpeel_make_loop_iterate_ntimes): Rename to..
9486 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9487 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9488 (vect_do_peeling): Update call accordingly.
9489 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9491 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9492 mask_compare_type, can_fully_mask_p and fully_masked_p.
9493 (release_vec_loop_masks): New function.
9494 (_loop_vec_info): Use it to free the loop masks.
9495 (can_produce_all_loop_masks_p): New function.
9496 (vect_get_max_nscalars_per_iter): Likewise.
9497 (vect_verify_full_masking): Likewise.
9498 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9499 retries, and free the mask rgroups before retrying. Check loop-wide
9500 reasons for disallowing fully-masked loops. Make the final decision
9501 about whether use a fully-masked loop or not.
9502 (vect_estimate_min_profitable_iters): Do not assume that peeling
9503 for the number of iterations will be needed for fully-masked loops.
9504 (vectorizable_reduction): Disable fully-masked loops.
9505 (vectorizable_live_operation): Likewise.
9506 (vect_halve_mask_nunits): New function.
9507 (vect_double_mask_nunits): Likewise.
9508 (vect_record_loop_mask): Likewise.
9509 (vect_get_loop_mask): Likewise.
9510 (vect_transform_loop): Handle the case in which the final loop
9511 iteration might handle a partial vector. Call vect_set_loop_condition
9512 instead of slpeel_make_loop_iterate_ntimes.
9513 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9514 (check_load_store_masking): New function.
9515 (prepare_load_store_mask): Likewise.
9516 (vectorizable_store): Handle fully-masked loops.
9517 (vectorizable_load): Likewise.
9518 (supportable_widening_operation): Use vect_halve_mask_nunits for
9520 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9521 (vect_gen_while): New function.
9522 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9523 (aarch64_uqdec<mode>): New insn.
9525 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9526 Alan Hayward <alan.hayward@arm.com>
9527 David Sherwood <david.sherwood@arm.com>
9529 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9530 (reduc_xor_scal_optab): New optabs.
9531 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9532 (reduc_xor_scal_@var{m}): Document.
9533 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9534 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9536 * fold-const-call.c (fold_const_call): Handle them.
9537 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9538 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9539 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9540 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9541 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9542 (UNSPEC_XORV): New unspecs.
9543 (optab): Add entries for them.
9544 (BITWISEV): New int iterator.
9545 (bit_reduc_op): New int attributes.
9547 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9548 Alan Hayward <alan.hayward@arm.com>
9549 David Sherwood <david.sherwood@arm.com>
9551 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9552 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9553 * optabs.def (vec_shl_insert_optab): New optab.
9554 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9555 (duplicate_and_interleave): Likewise.
9556 * tree-vect-loop.c: Include internal-fn.h.
9557 (neutral_op_for_slp_reduction): New function, split out from
9558 get_initial_defs_for_reduction.
9559 (get_initial_def_for_reduction): Handle option 2 for variable-length
9560 vectors by loading the neutral value into a vector and then shifting
9561 the initial value into element 0.
9562 (get_initial_defs_for_reduction): Replace the code argument with
9563 the neutral value calculated by neutral_op_for_slp_reduction.
9564 Use gimple_build_vector for constant-length vectors.
9565 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9566 but the first group_size elements have a neutral value.
9567 Use duplicate_and_interleave otherwise.
9568 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9569 Update call to get_initial_defs_for_reduction. Handle SLP
9570 reductions for variable-length vectors by creating one vector
9571 result for each scalar result, with the elements associated
9572 with other scalar results stubbed out with the neutral value.
9573 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9574 Require IFN_VEC_SHL_INSERT for double reductions on
9575 variable-length vectors, or SLP reductions that have
9576 a neutral value. Require can_duplicate_and_interleave_p
9577 support for variable-length unchained SLP reductions if there
9578 is no neutral value, such as for MIN/MAX reductions. Also require
9579 the number of vector elements to be a multiple of the number of
9580 SLP statements when doing variable-length unchained SLP reductions.
9581 Update call to vect_create_epilog_for_reduction.
9582 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9583 and remove initial values.
9584 (duplicate_and_interleave): Make public.
9585 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9586 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9588 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9589 Alan Hayward <alan.hayward@arm.com>
9590 David Sherwood <david.sherwood@arm.com>
9592 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9593 (can_duplicate_and_interleave_p): New function.
9594 (vect_get_and_check_slp_defs): Take the vector of statements
9595 rather than just the current one. Remove excess parentheses.
9596 Restriction rejectinon of vect_constant_def and vect_external_def
9597 for variable-length vectors to boolean types, or types for which
9598 can_duplicate_and_interleave_p is false.
9599 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9600 (duplicate_and_interleave): New function.
9601 (vect_get_constant_vectors): Use gimple_build_vector for
9602 constant-length vectors and suitable variable-length constant
9603 vectors. Use duplicate_and_interleave for other variable-length
9604 vectors. Don't defer the update when inserting new statements.
9606 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9607 Alan Hayward <alan.hayward@arm.com>
9608 David Sherwood <david.sherwood@arm.com>
9610 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9611 min_profitable_iters doesn't go negative.
9613 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9614 Alan Hayward <alan.hayward@arm.com>
9615 David Sherwood <david.sherwood@arm.com>
9617 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9618 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9619 * optabs.def (vec_mask_load_lanes_optab): New optab.
9620 (vec_mask_store_lanes_optab): Likewise.
9621 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9622 (MASK_STORE_LANES): Likewise.
9623 * internal-fn.c (mask_load_lanes_direct): New macro.
9624 (mask_store_lanes_direct): Likewise.
9625 (expand_mask_load_optab_fn): Handle masked operations.
9626 (expand_mask_load_lanes_optab_fn): New macro.
9627 (expand_mask_store_optab_fn): Handle masked operations.
9628 (expand_mask_store_lanes_optab_fn): New macro.
9629 (direct_mask_load_lanes_optab_supported_p): Likewise.
9630 (direct_mask_store_lanes_optab_supported_p): Likewise.
9631 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9633 (vect_load_lanes_supported): Likewise.
9634 * tree-vect-data-refs.c (strip_conversion): New function.
9635 (can_group_stmts_p): Likewise.
9636 (vect_analyze_data_ref_accesses): Use it instead of checking
9637 for a pair of assignments.
9638 (vect_store_lanes_supported): Take a masked_p parameter.
9639 (vect_load_lanes_supported): Likewise.
9640 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9641 vect_store_lanes_supported and vect_load_lanes_supported.
9642 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9643 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9644 parameter. Don't allow gaps for masked accesses.
9645 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9646 and vect_load_lanes_supported.
9647 (get_load_store_type): Take a masked_p parameter and update
9648 call to get_group_load_store_type.
9649 (vectorizable_store): Update call to get_load_store_type.
9650 Handle IFN_MASK_STORE_LANES.
9651 (vectorizable_load): Update call to get_load_store_type.
9652 Handle IFN_MASK_LOAD_LANES.
9654 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9655 Alan Hayward <alan.hayward@arm.com>
9656 David Sherwood <david.sherwood@arm.com>
9658 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9660 * config/aarch64/aarch64-protos.h
9661 (aarch64_sve_struct_memory_operand_p): Declare.
9662 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9663 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9664 (VPRED, vpred): Handle SVE structure modes.
9665 * config/aarch64/constraints.md (Utx): New constraint.
9666 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9667 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9668 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9669 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9670 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9671 structure modes. Split into pieces after RA.
9672 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9673 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9675 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9676 SVE structure modes.
9677 (aarch64_classify_address): Likewise.
9678 (sizetochar): Move earlier in file.
9679 (aarch64_print_operand): Handle SVE register lists.
9680 (aarch64_array_mode): New function.
9681 (aarch64_sve_struct_memory_operand_p): Likewise.
9682 (TARGET_ARRAY_MODE): Redefine.
9684 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9685 Alan Hayward <alan.hayward@arm.com>
9686 David Sherwood <david.sherwood@arm.com>
9688 * target.def (array_mode): New target hook.
9689 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9690 * doc/tm.texi: Regenerate.
9691 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9692 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9693 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9695 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9698 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9699 Alan Hayward <alan.hayward@arm.com>
9700 David Sherwood <david.sherwood@arm.com>
9702 * fold-const.c (fold_binary_loc): Check the argument types
9703 rather than the result type when testing for a vector operation.
9705 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9707 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9708 * doc/tm.texi: Regenerate.
9710 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9711 Alan Hayward <alan.hayward@arm.com>
9712 David Sherwood <david.sherwood@arm.com>
9714 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9715 (sve): Document new AArch64 extension.
9716 * doc/md.texi (w): Extend the description of the AArch64
9717 constraint to include SVE vectors.
9718 (Upl, Upa): Document new AArch64 predicate constraints.
9719 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9721 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9722 (msve-vector-bits=): New option.
9723 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9724 SVE when these are disabled.
9725 (sve): New extension.
9726 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9727 modes. Adjust their number of units based on aarch64_sve_vg.
9728 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9729 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9730 aarch64_addr_query_type.
9731 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9732 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9733 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9734 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9735 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9736 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9737 (aarch64_simd_imm_zero_p): Delete.
9738 (aarch64_check_zero_based_sve_index_immediate): Declare.
9739 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9740 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9741 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9742 (aarch64_sve_float_mul_immediate_p): Likewise.
9743 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9745 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9746 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9747 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9748 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9749 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9750 (aarch64_regmode_natural_size): Likewise.
9751 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9752 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9754 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9755 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9756 for VG and the SVE predicate registers.
9757 (V_ALIASES): Add a "z"-prefixed alias.
9758 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9759 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9760 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9761 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9762 (REG_CLASS_NAMES): Add entries for them.
9763 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9764 and the predicate registers.
9765 (aarch64_sve_vg): Declare.
9766 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9767 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9768 (REGMODE_NATURAL_SIZE): Define.
9769 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9771 * config/aarch64/aarch64.c: Include cfgrtl.h.
9772 (simd_immediate_info): Add a constructor for series vectors,
9773 and an associated step field.
9774 (aarch64_sve_vg): New variable.
9775 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9776 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9777 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9778 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9779 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9780 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9781 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9782 (aarch64_get_mask_mode): New functions.
9783 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9784 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9785 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9786 predicate modes and predicate registers. Explicitly restrict
9787 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9788 to store a vector mode if it is recognized by
9789 aarch64_classify_vector_mode.
9790 (aarch64_regmode_natural_size): New function.
9791 (aarch64_hard_regno_caller_save_mode): Return the original mode
9793 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9794 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9795 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9796 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9798 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9799 does not overlap dest if the function is frame-related. Handle
9801 (aarch64_split_add_offset): New function.
9802 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9803 them aarch64_add_offset.
9804 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9805 and update call to aarch64_sub_sp.
9806 (aarch64_add_cfa_expression): New function.
9807 (aarch64_expand_prologue): Pass extra temporary registers to the
9808 functions above. Handle the case in which we need to emit new
9809 DW_CFA_expressions for registers that were originally saved
9810 relative to the stack pointer, but now have to be expressed
9811 relative to the frame pointer.
9812 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9814 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9815 IP0 and IP1 values for SVE frames.
9816 (aarch64_expand_vec_series): New function.
9817 (aarch64_expand_sve_widened_duplicate): Likewise.
9818 (aarch64_expand_sve_const_vector): Likewise.
9819 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9820 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9821 into the register, rather than emitting a SET directly.
9822 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9823 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9824 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9825 (offset_9bit_signed_scaled_p): New functions.
9826 (aarch64_replicate_bitmask_imm): New function.
9827 (aarch64_bitmask_imm): Use it.
9828 (aarch64_cannot_force_const_mem): Reject expressions involving
9829 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9830 (aarch64_classify_index): Handle SVE indices, by requiring
9831 a plain register index with a scale that matches the element size.
9832 (aarch64_classify_address): Handle SVE addresses. Assert that
9833 the mode of the address is VOIDmode or an integer mode.
9834 Update call to aarch64_classify_symbol.
9835 (aarch64_classify_symbolic_expression): Update call to
9836 aarch64_classify_symbol.
9837 (aarch64_const_vec_all_in_range_p): New function.
9838 (aarch64_print_vector_float_operand): Likewise.
9839 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9840 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9841 and the FP immediates 1.0 and 0.5.
9842 (aarch64_print_address_internal): Handle SVE addresses.
9843 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9844 (aarch64_regno_regclass): Handle predicate registers.
9845 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9847 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9848 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9849 (aarch64_convert_sve_vector_bits): New function.
9850 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9851 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9853 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9854 Handle SVE vector and predicate modes. Accept VL-based constants
9855 that need only one temporary register, and VL offsets that require
9856 no temporary registers.
9857 (aarch64_conditional_register_usage): Mark the predicate registers
9858 as fixed if SVE isn't available.
9859 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9860 Return true for SVE vector and predicate modes.
9861 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9862 rather than an unsigned int. Handle SVE modes.
9863 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9865 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9867 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9868 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9869 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9870 (aarch64_sve_float_mul_immediate_p): New functions.
9871 (aarch64_sve_valid_immediate): New function.
9872 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9873 Explicitly reject structure modes. Check for INDEX constants.
9874 Handle PTRUE and PFALSE constants.
9875 (aarch64_check_zero_based_sve_index_immediate): New function.
9876 (aarch64_simd_imm_zero_p): Delete.
9877 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9878 vector modes. Accept constants in the range of CNT[BHWD].
9879 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9880 ask for an Advanced SIMD mode.
9881 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9882 (aarch64_simd_vector_alignment): Handle SVE predicates.
9883 (aarch64_vectorize_preferred_vector_alignment): New function.
9884 (aarch64_simd_vector_alignment_reachable): Use it instead of
9886 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9887 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9889 (MAX_VECT_LEN): Delete.
9890 (expand_vec_perm_d): Add a vec_flags field.
9891 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9892 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9893 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9895 (aarch64_evpc_rev): Rename to...
9896 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9897 (aarch64_evpc_rev_global): New function.
9898 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9899 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9901 (aarch64_evpc_sve_tbl): New function.
9902 (aarch64_expand_vec_perm_const_1): Update after rename of
9903 aarch64_evpc_rev. Handle SVE permutes too, trying
9904 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9905 than aarch64_evpc_tbl.
9906 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9907 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9908 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9909 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9910 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9911 (aarch64_expand_sve_vcond): New functions.
9912 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9913 of aarch64_vector_mode_p.
9914 (aarch64_dwarf_poly_indeterminate_value): New function.
9915 (aarch64_compute_pressure_classes): Likewise.
9916 (aarch64_can_change_mode_class): Likewise.
9917 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9918 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9919 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9920 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9921 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9922 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9923 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9924 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9926 (Dn, Dl, Dr): Accept const as well as const_vector.
9927 (Dz): Likewise. Compare against CONST0_RTX.
9928 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9929 of "vector" where appropriate.
9930 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9931 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9932 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9933 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9934 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9935 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9936 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9937 (v_int_equiv): Extend to SVE modes.
9938 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9940 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9941 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9942 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9943 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9944 (SVE_COND_FP_CMP): New int iterators.
9945 (perm_hilo): Handle the new unpack unspecs.
9946 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9948 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9949 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9950 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9951 (aarch64_equality_operator, aarch64_constant_vector_operand)
9952 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9953 (aarch64_sve_nonimmediate_operand): Likewise.
9954 (aarch64_sve_general_operand): Likewise.
9955 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9956 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9957 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9958 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9959 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9960 (aarch64_sve_float_arith_immediate): Likewise.
9961 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9962 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9963 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9964 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9965 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9966 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9967 (aarch64_sve_float_arith_operand): Likewise.
9968 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9969 (aarch64_sve_float_mul_operand): Likewise.
9970 (aarch64_sve_vec_perm_operand): Likewise.
9971 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9972 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9973 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9974 as well as const_vector.
9975 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9976 in file. Use CONST0_RTX and CONSTM1_RTX.
9977 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9978 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9979 Use aarch64_simd_imm_zero.
9980 * config/aarch64/aarch64-sve.md: New file.
9981 * config/aarch64/aarch64.md: Include it.
9982 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9983 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9984 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9985 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9986 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9987 (sve): New attribute.
9988 (enabled): Disable instructions with the sve attribute unless
9990 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9991 aarch64_expand_mov_immediate.
9992 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9993 CNT[BHSD] immediates.
9994 (movti): Split CONST_POLY_INT moves into two halves.
9995 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9996 Split additions that need a temporary here if the destination
9997 is the stack pointer.
9998 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9999 (*add<mode>3_poly_1): New instruction.
10000 (set_clobber_cc): New expander.
10002 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10004 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10005 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10006 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10007 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10008 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10009 Change innermode from fixed_mode_size to machine_mode.
10010 (simplify_subreg): Update call accordingly. Handle a constant-sized
10011 subreg of a variable-length CONST_VECTOR.
10013 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10014 Alan Hayward <alan.hayward@arm.com>
10015 David Sherwood <david.sherwood@arm.com>
10017 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10018 (add_offset_to_base): New function, split out from...
10019 (create_mem_ref): ...here. When handling a scale other than 1,
10020 check first whether the address is valid without the offset.
10021 Add it into the base if so, leaving the index and scale as-is.
10023 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10026 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10027 fold_for_warn before checking if arg2 is INTEGER_CST.
10029 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10031 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10032 (store_multiple_operation): Delete.
10033 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10034 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10035 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10036 guarded by TARGET_STRING.
10037 (rs6000_output_load_multiple): Delete.
10038 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10039 OPTION_MASK_STRING / TARGET_STRING handling.
10040 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10041 (const rs6000_opt_masks) <"string">: Change mask to 0.
10042 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10043 (MASK_STRING): Delete.
10044 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10046 (load_multiple): Delete.
10053 (store_multiple): Delete.
10060 (movmemsi_8reg): Delete.
10061 (corresponding unnamed define_insn): Delete.
10062 (movmemsi_6reg): Delete.
10063 (corresponding unnamed define_insn): Delete.
10064 (movmemsi_4reg): Delete.
10065 (corresponding unnamed define_insn): Delete.
10066 (movmemsi_2reg): Delete.
10067 (corresponding unnamed define_insn): Delete.
10068 (movmemsi_1reg): Delete.
10069 (corresponding unnamed define_insn): Delete.
10070 * config/rs6000/rs6000.opt (mno-string): New.
10071 (mstring): Replace by deprecation warning stub.
10072 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10074 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10076 * regrename.c (regrename_do_replace): If replacing the same
10077 reg multiple times, try to reuse last created gen_raw_REG.
10080 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10081 main to workaround a bug in GDB.
10083 2018-01-12 Tom de Vries <tom@codesourcery.com>
10086 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10088 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10090 PR rtl-optimization/80481
10091 * ira-color.c (get_cap_member): New function.
10092 (allocnos_conflict_by_live_ranges_p): Use it.
10093 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10094 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10096 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10099 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10100 (*saddl_se_1): Ditto.
10101 (*ssubsi_1): Ditto.
10102 (*ssubl_se_1): Ditto.
10104 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10106 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10107 rather than wi::to_widest for DR_INITs.
10108 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10109 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10110 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10112 (vect_analyze_group_access_1): Note that here.
10114 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10116 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10117 polynomial type sizes.
10119 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10121 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10122 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10123 (gimple_add_tmp_var): Likewise.
10125 2018-01-12 Martin Liska <mliska@suse.cz>
10127 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10128 (gimple_alloc_sizes): Likewise.
10129 (dump_gimple_statistics): Use PRIu64 in printf format.
10130 * gimple.h: Change uint64_t to int.
10132 2018-01-12 Martin Liska <mliska@suse.cz>
10134 * tree-core.h: Use uint64_t instead of int.
10135 * tree.c (tree_node_counts): Likewise.
10136 (tree_node_sizes): Likewise.
10137 (dump_tree_statistics): Use PRIu64 in printf format.
10139 2018-01-12 Martin Liska <mliska@suse.cz>
10141 * Makefile.in: As qsort_chk is implemented in vec.c, add
10142 vec.o to linkage of gencfn-macros.
10143 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10144 passing the info to record_node_allocation_statistics.
10145 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10147 * ggc-common.c (struct ggc_usage): Add operator== and use
10148 it in operator< and compare function.
10149 * mem-stats.h (struct mem_usage): Likewise.
10150 * vec.c (struct vec_usage): Remove operator< and compare
10151 function. Can be simply inherited.
10153 2018-01-12 Martin Jambor <mjambor@suse.cz>
10156 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10157 * tree-ssa-math-opts.c: Include domwalk.h.
10158 (convert_mult_to_fma_1): New function.
10159 (fma_transformation_info): New type.
10160 (fma_deferring_state): Likewise.
10161 (cancel_fma_deferring): New function.
10162 (result_of_phi): Likewise.
10163 (last_fma_candidate_feeds_initial_phi): Likewise.
10164 (convert_mult_to_fma): Added deferring logic, split actual
10165 transformation to convert_mult_to_fma_1.
10166 (math_opts_dom_walker): New type.
10167 (math_opts_dom_walker::after_dom_children): New method, body moved
10168 here from pass_optimize_widening_mul::execute, added deferring logic
10170 (pass_optimize_widening_mul::execute): Moved most of code to
10171 math_opts_dom_walker::after_dom_children.
10172 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10173 * config/i386/i386.c (ix86_option_override_internal): Added
10174 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10176 2018-01-12 Richard Biener <rguenther@suse.de>
10179 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10180 inline instance vars.
10182 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10185 * config/rx/rx.c (rx_is_restricted_memory_address):
10186 Handle SUBREG case.
10188 2018-01-12 Richard Biener <rguenther@suse.de>
10190 PR tree-optimization/80846
10191 * target.def (split_reduction): New target hook.
10192 * targhooks.c (default_split_reduction): New function.
10193 * targhooks.h (default_split_reduction): Declare.
10194 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10195 target requests first reduce vectors by combining low and high
10197 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10198 (get_vectype_for_scalar_type_and_size): Export.
10199 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10200 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10201 * doc/tm.texi: Regenerate.
10202 * config/i386/i386.c (ix86_split_reduction): Implement
10203 TARGET_VECTORIZE_SPLIT_REDUCTION.
10205 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10208 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10209 in PIC mode except for TARGET_VXWORKS_RTP.
10210 * config/sparc/sparc.c: Include cfgrtl.h.
10211 (TARGET_INIT_PIC_REG): Define.
10212 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10213 (sparc_pic_register_p): New predicate.
10214 (sparc_legitimate_address_p): Use it.
10215 (sparc_legitimize_pic_address): Likewise.
10216 (sparc_delegitimize_address): Likewise.
10217 (sparc_mode_dependent_address_p): Likewise.
10218 (gen_load_pcrel_sym): Remove 4th parameter.
10219 (load_got_register): Adjust call to above. Remove obsolete stuff.
10220 (sparc_expand_prologue): Do not call load_got_register here.
10221 (sparc_flat_expand_prologue): Likewise.
10222 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10223 (sparc_use_pseudo_pic_reg): New function.
10224 (sparc_init_pic_reg): Likewise.
10225 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10226 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10228 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10230 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10231 Add item for branch_cost.
10233 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10235 PR rtl-optimization/83565
10236 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10237 not extend the result to a larger mode for rotate operations.
10238 (num_sign_bit_copies1): Likewise.
10240 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10243 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10245 Use values-Xc.o for -pedantic.
10246 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10248 2018-01-12 Martin Liska <mliska@suse.cz>
10251 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10253 (possible_polymorphic_call_targets): Use it.
10254 (ipa_devirt): Likewise.
10256 2018-01-12 Martin Liska <mliska@suse.cz>
10258 * profile-count.h (enum profile_quality): Use 0 as invalid
10259 enum value of profile_quality.
10261 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10263 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10264 -mext-string options.
10266 2018-01-12 Richard Biener <rguenther@suse.de>
10268 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10269 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10270 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10272 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10274 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10276 * configure.ac (--with-long-double-format): Add support for the
10277 configuration option to change the default long double format on
10279 * config.gcc (powerpc*-linux*-*): Likewise.
10280 * configure: Regenerate.
10281 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10282 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10283 used without modification.
10285 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10287 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10288 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10289 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10290 MISC_BUILTIN_SPEC_BARRIER.
10291 (rs6000_init_builtins): Likewise.
10292 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10294 (speculation_barrier): New define_insn.
10295 * doc/extend.texi: Document __builtin_speculation_barrier.
10297 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10300 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10301 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10302 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10304 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10305 integral modes instead of "ss" and "sd".
10306 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10307 vectors with 32-bit and 64-bit elements.
10308 (vecdupssescalarmodesuffix): New mode attribute.
10309 (vec_dup<mode>): Use it.
10311 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10314 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10315 frame if argument is passed on stack.
10317 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10320 * ree.c (combine_reaching_defs): Optimize also
10321 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10322 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10324 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10326 PR middle-end/83189
10327 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10329 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10331 PR middle-end/83718
10332 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10333 after they are computed.
10335 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10337 PR tree-optimization/83695
10338 * gimple-loop-linterchange.cc
10339 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10340 reset cached scev information after interchange.
10341 (pass_linterchange::execute): Remove call to scev_reset_htab.
10343 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10345 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10346 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10347 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10348 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10349 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10350 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10351 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10352 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10353 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10354 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10355 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10356 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10357 (V_lane_reg): Likewise.
10358 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10360 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10361 (vfmal_lane_low<mode>_intrinsic,
10362 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10363 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10364 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10365 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10366 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10367 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10369 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10371 * config/arm/arm-cpus.in (fp16fml): New feature.
10372 (ALL_SIMD): Add fp16fml.
10373 (armv8.2-a): Add fp16fml as an option.
10374 (armv8.3-a): Likewise.
10375 (armv8.4-a): Add fp16fml as part of fp16.
10376 * config/arm/arm.h (TARGET_FP16FML): Define.
10377 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10379 * config/arm/arm-modes.def (V2HF): Define.
10380 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10381 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10382 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10383 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10384 vfmsl_low, vfmsl_high): New set of builtins.
10385 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10386 (vfml_op): New code attribute.
10387 (VFMLHALVES): New int iterator.
10388 (VFML, VFMLSEL): New mode attributes.
10389 (V_reg): Define mapping for V2HF.
10390 (V_hi, V_lo): New mode attributes.
10391 (VF_constraint): Likewise.
10392 (vfml_half, vfml_half_selector): New int attributes.
10393 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10395 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10396 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10398 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10399 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10400 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10401 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10403 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10404 Document new effective target and option set.
10406 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10408 * config/arm/arm-cpus.in (armv8_4): New feature.
10409 (ARMv8_4a): New fgroup.
10410 (armv8.4-a): New arch.
10411 * config/arm/arm-tables.opt: Regenerate.
10412 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10413 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10414 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10415 Add matching rules for -march=armv8.4-a and extensions.
10416 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10418 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10421 * config/rx/rx.md (BW): New mode attribute.
10422 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10424 2018-01-11 Richard Biener <rguenther@suse.de>
10426 PR tree-optimization/83435
10427 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10428 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10429 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10431 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10432 Alan Hayward <alan.hayward@arm.com>
10433 David Sherwood <david.sherwood@arm.com>
10435 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10437 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10438 (aarch64_print_address_internal): Use it to check for a zero offset.
10440 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10441 Alan Hayward <alan.hayward@arm.com>
10442 David Sherwood <david.sherwood@arm.com>
10444 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10445 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10446 Return a poly_int64 rather than a HOST_WIDE_INT.
10447 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10448 rather than a HOST_WIDE_INT.
10449 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10450 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10451 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10452 final_offset from HOST_WIDE_INT to poly_int64.
10453 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10454 to_constant when getting the number of units in an Advanced SIMD
10456 (aarch64_builtin_vectorized_function): Check for a constant number
10458 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10460 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10461 attribute instead of GET_MODE_NUNITS.
10462 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10463 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10464 GET_MODE_SIZE for fixed-size registers.
10465 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10466 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10467 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10468 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10469 (aarch64_print_operand, aarch64_print_address_internal)
10470 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10471 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10472 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10473 Handle polynomial GET_MODE_SIZE.
10474 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10475 wider than SImode without modification.
10476 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10477 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10478 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10479 passing and returning SVE modes.
10480 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10481 rather than GEN_INT.
10482 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10483 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10484 (aarch64_allocate_and_probe_stack_space): Likewise.
10485 (aarch64_layout_frame): Cope with polynomial offsets.
10486 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10487 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10488 polynomial offsets.
10489 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10490 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10491 poly_int64 rather than a HOST_WIDE_INT.
10492 (aarch64_get_separate_components, aarch64_process_components)
10493 (aarch64_expand_prologue, aarch64_expand_epilogue)
10494 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10495 (aarch64_anchor_offset): New function, split out from...
10496 (aarch64_legitimize_address): ...here.
10497 (aarch64_builtin_vectorization_cost): Handle polynomial
10498 TYPE_VECTOR_SUBPARTS.
10499 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10501 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10502 number of elements from the PARALLEL rather than the mode.
10503 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10504 rather than GET_MODE_BITSIZE.
10505 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10506 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10507 (aarch64_expand_vec_perm_const_1): Handle polynomial
10508 d->perm.length () and d->perm elements.
10509 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10510 Apply to_constant to d->perm elements.
10511 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10512 polynomial CONST_VECTOR_NUNITS.
10513 (aarch64_move_pointer): Take amount as a poly_int64 rather
10515 (aarch64_progress_pointer): Avoid temporary variable.
10516 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10517 the mode attribute instead of GET_MODE.
10519 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10520 Alan Hayward <alan.hayward@arm.com>
10521 David Sherwood <david.sherwood@arm.com>
10523 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10524 x exists before using it.
10525 (aarch64_add_constant_internal): Rename to...
10526 (aarch64_add_offset_1): ...this. Replace regnum with separate
10527 src and dest rtxes. Handle the case in which they're different,
10528 including when the offset is zero. Replace scratchreg with an rtx.
10529 Use 2 additions if there is no spare register into which we can
10530 move a 16-bit constant.
10531 (aarch64_add_constant): Delete.
10532 (aarch64_add_offset): Replace reg with separate src and dest
10533 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10534 Use aarch64_add_offset_1.
10535 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10536 an rtx rather than an int. Take the delta as a poly_int64
10537 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10538 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10539 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10540 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10541 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10542 and aarch64_add_sp.
10543 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10544 aarch64_add_constant.
10546 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10548 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10549 Use scalar_float_mode.
10551 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10553 * config/aarch64/aarch64-simd.md
10554 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10555 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10556 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10557 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10558 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10559 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10560 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10561 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10562 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10563 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10565 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10568 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10569 targ_options->x_arm_arch_string is non NULL.
10571 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10573 * config/aarch64/aarch64.h
10574 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10576 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10579 * expmed.c (emit_store_flag_force): Swap if const op0
10580 and change VOIDmode to mode of op0.
10582 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10584 PR rtl-optimization/83761
10585 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10586 than bytes to mode_for_size.
10588 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10590 PR middle-end/83189
10591 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10592 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10595 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10597 PR middle-end/83575
10598 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10599 when in layout mode.
10600 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10601 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10604 2018-01-10 Michael Collison <michael.collison@arm.com>
10606 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10607 * config/aarch64/aarch64-option-extension.def: Add
10608 AARCH64_OPT_EXTENSION of 'fp16fml'.
10609 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10610 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10611 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10612 * config/aarch64/constraints.md (Ui7): New constraint.
10613 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10614 (VFMLA_SEL_W): Ditto.
10617 (VFMLA16_LOW): New int iterator.
10618 (VFMLA16_HIGH): Ditto.
10619 (UNSPEC_FMLAL): New unspec.
10620 (UNSPEC_FMLSL): Ditto.
10621 (UNSPEC_FMLAL2): Ditto.
10622 (UNSPEC_FMLSL2): Ditto.
10623 (f16mac): New code attribute.
10624 * config/aarch64/aarch64-simd-builtins.def
10625 (aarch64_fmlal_lowv2sf): Ditto.
10626 (aarch64_fmlsl_lowv2sf): Ditto.
10627 (aarch64_fmlalq_lowv4sf): Ditto.
10628 (aarch64_fmlslq_lowv4sf): Ditto.
10629 (aarch64_fmlal_highv2sf): Ditto.
10630 (aarch64_fmlsl_highv2sf): Ditto.
10631 (aarch64_fmlalq_highv4sf): Ditto.
10632 (aarch64_fmlslq_highv4sf): Ditto.
10633 (aarch64_fmlal_lane_lowv2sf): Ditto.
10634 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10635 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10636 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10637 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10638 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10639 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10640 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10641 (aarch64_fmlal_lane_highv2sf): Ditto.
10642 (aarch64_fmlsl_lane_highv2sf): Ditto.
10643 (aarch64_fmlal_laneq_highv2sf): Ditto.
10644 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10645 (aarch64_fmlalq_lane_highv4sf): Ditto.
10646 (aarch64_fmlsl_lane_highv4sf): Ditto.
10647 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10648 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10649 * config/aarch64/aarch64-simd.md:
10650 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10651 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10652 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10653 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10654 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10655 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10656 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10657 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10658 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10659 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10660 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10661 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10662 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10663 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10664 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10665 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10666 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10667 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10668 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10669 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10670 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10671 (vfmlsl_low_u32): Ditto.
10672 (vfmlalq_low_u32): Ditto.
10673 (vfmlslq_low_u32): Ditto.
10674 (vfmlal_high_u32): Ditto.
10675 (vfmlsl_high_u32): Ditto.
10676 (vfmlalq_high_u32): Ditto.
10677 (vfmlslq_high_u32): Ditto.
10678 (vfmlal_lane_low_u32): Ditto.
10679 (vfmlsl_lane_low_u32): Ditto.
10680 (vfmlal_laneq_low_u32): Ditto.
10681 (vfmlsl_laneq_low_u32): Ditto.
10682 (vfmlalq_lane_low_u32): Ditto.
10683 (vfmlslq_lane_low_u32): Ditto.
10684 (vfmlalq_laneq_low_u32): Ditto.
10685 (vfmlslq_laneq_low_u32): Ditto.
10686 (vfmlal_lane_high_u32): Ditto.
10687 (vfmlsl_lane_high_u32): Ditto.
10688 (vfmlal_laneq_high_u32): Ditto.
10689 (vfmlsl_laneq_high_u32): Ditto.
10690 (vfmlalq_lane_high_u32): Ditto.
10691 (vfmlslq_lane_high_u32): Ditto.
10692 (vfmlalq_laneq_high_u32): Ditto.
10693 (vfmlslq_laneq_high_u32): Ditto.
10694 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10695 (AARCH64_FL_FOR_ARCH8_4): New.
10696 (AARCH64_ISA_F16FML): New ISA flag.
10697 (TARGET_F16FML): New feature flag for fp16fml.
10698 (doc/invoke.texi): Document new fp16fml option.
10700 2018-01-10 Michael Collison <michael.collison@arm.com>
10702 * config/aarch64/aarch64-builtins.c:
10703 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10704 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10705 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10706 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10707 (AARCH64_ISA_SHA3): New ISA flag.
10708 (TARGET_SHA3): New feature flag for sha3.
10709 * config/aarch64/iterators.md (sha512_op): New int attribute.
10710 (CRYPTO_SHA512): New int iterator.
10711 (UNSPEC_SHA512H): New unspec.
10712 (UNSPEC_SHA512H2): Ditto.
10713 (UNSPEC_SHA512SU0): Ditto.
10714 (UNSPEC_SHA512SU1): Ditto.
10715 * config/aarch64/aarch64-simd-builtins.def
10716 (aarch64_crypto_sha512hqv2di): New builtin.
10717 (aarch64_crypto_sha512h2qv2di): Ditto.
10718 (aarch64_crypto_sha512su0qv2di): Ditto.
10719 (aarch64_crypto_sha512su1qv2di): Ditto.
10720 (aarch64_eor3qv8hi): Ditto.
10721 (aarch64_rax1qv2di): Ditto.
10722 (aarch64_xarqv2di): Ditto.
10723 (aarch64_bcaxqv8hi): Ditto.
10724 * config/aarch64/aarch64-simd.md:
10725 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10726 (aarch64_crypto_sha512su0qv2di): Ditto.
10727 (aarch64_crypto_sha512su1qv2di): Ditto.
10728 (aarch64_eor3qv8hi): Ditto.
10729 (aarch64_rax1qv2di): Ditto.
10730 (aarch64_xarqv2di): Ditto.
10731 (aarch64_bcaxqv8hi): Ditto.
10732 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10733 (vsha512h2q_u64): Ditto.
10734 (vsha512su0q_u64): Ditto.
10735 (vsha512su1q_u64): Ditto.
10736 (veor3q_u16): Ditto.
10737 (vrax1q_u64): Ditto.
10738 (vxarq_u64): Ditto.
10739 (vbcaxq_u16): Ditto.
10740 * config/arm/types.md (crypto_sha512): New type attribute.
10741 (crypto_sha3): Ditto.
10742 (doc/invoke.texi): Document new sha3 option.
10744 2018-01-10 Michael Collison <michael.collison@arm.com>
10746 * config/aarch64/aarch64-builtins.c:
10747 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10748 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10749 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10750 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10751 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10752 (AARCH64_ISA_SM4): New ISA flag.
10753 (TARGET_SM4): New feature flag for sm4.
10754 * config/aarch64/aarch64-simd-builtins.def
10755 (aarch64_sm3ss1qv4si): Ditto.
10756 (aarch64_sm3tt1aq4si): Ditto.
10757 (aarch64_sm3tt1bq4si): Ditto.
10758 (aarch64_sm3tt2aq4si): Ditto.
10759 (aarch64_sm3tt2bq4si): Ditto.
10760 (aarch64_sm3partw1qv4si): Ditto.
10761 (aarch64_sm3partw2qv4si): Ditto.
10762 (aarch64_sm4eqv4si): Ditto.
10763 (aarch64_sm4ekeyqv4si): Ditto.
10764 * config/aarch64/aarch64-simd.md:
10765 (aarch64_sm3ss1qv4si): Ditto.
10766 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10767 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10768 (aarch64_sm4eqv4si): Ditto.
10769 (aarch64_sm4ekeyqv4si): Ditto.
10770 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10771 (sm3part_op): Ditto.
10772 (CRYPTO_SM3TT): Ditto.
10773 (CRYPTO_SM3PART): Ditto.
10774 (UNSPEC_SM3SS1): New unspec.
10775 (UNSPEC_SM3TT1A): Ditto.
10776 (UNSPEC_SM3TT1B): Ditto.
10777 (UNSPEC_SM3TT2A): Ditto.
10778 (UNSPEC_SM3TT2B): Ditto.
10779 (UNSPEC_SM3PARTW1): Ditto.
10780 (UNSPEC_SM3PARTW2): Ditto.
10781 (UNSPEC_SM4E): Ditto.
10782 (UNSPEC_SM4EKEY): Ditto.
10783 * config/aarch64/constraints.md (Ui2): New constraint.
10784 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10785 * config/arm/types.md (crypto_sm3): New type attribute.
10786 (crypto_sm4): Ditto.
10787 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10788 (vsm3tt1aq_u32): Ditto.
10789 (vsm3tt1bq_u32): Ditto.
10790 (vsm3tt2aq_u32): Ditto.
10791 (vsm3tt2bq_u32): Ditto.
10792 (vsm3partw1q_u32): Ditto.
10793 (vsm3partw2q_u32): Ditto.
10794 (vsm4eq_u32): Ditto.
10795 (vsm4ekeyq_u32): Ditto.
10796 (doc/invoke.texi): Document new sm4 option.
10798 2018-01-10 Michael Collison <michael.collison@arm.com>
10800 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10801 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10802 (AARCH64_FL_FOR_ARCH8_4): New.
10803 (AARCH64_FL_V8_4): New flag.
10804 (doc/invoke.texi): Document new armv8.4-a option.
10806 2018-01-10 Michael Collison <michael.collison@arm.com>
10808 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10809 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10810 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10811 * config/aarch64/aarch64-option-extension.def: Add
10812 AARCH64_OPT_EXTENSION of 'sha2'.
10813 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10814 (crypto): Disable sha2 and aes if crypto disabled.
10815 (crypto): Enable aes and sha2 if enabled.
10816 (simd): Disable sha2 and aes if simd disabled.
10817 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10819 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10820 (TARGET_SHA2): New feature flag for sha2.
10821 (TARGET_AES): New feature flag for aes.
10822 * config/aarch64/aarch64-simd.md:
10823 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10824 conditional on TARGET_AES.
10825 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10826 (aarch64_crypto_sha1hsi): Make pattern conditional
10828 (aarch64_crypto_sha1hv4si): Ditto.
10829 (aarch64_be_crypto_sha1hv4si): Ditto.
10830 (aarch64_crypto_sha1su1v4si): Ditto.
10831 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10832 (aarch64_crypto_sha1su0v4si): Ditto.
10833 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10834 (aarch64_crypto_sha256su0v4si): Ditto.
10835 (aarch64_crypto_sha256su1v4si): Ditto.
10836 (doc/invoke.texi): Document new aes and sha2 options.
10838 2018-01-10 Martin Sebor <msebor@redhat.com>
10840 PR tree-optimization/83781
10841 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10844 2018-01-11 Martin Sebor <msebor@gmail.com>
10845 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10847 PR tree-optimization/83501
10848 PR tree-optimization/81703
10850 * tree-ssa-strlen.c (get_string_cst): Rename...
10851 (get_string_len): ...to this. Handle global constants.
10852 (handle_char_store): Adjust.
10854 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10855 Jim Wilson <jimw@sifive.com>
10857 * config/riscv/riscv-protos.h (riscv_output_return): New.
10858 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10859 (riscv_attribute_table, riscv_output_return),
10860 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10861 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10862 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10864 (riscv_expand_prologue): Add early return for naked function.
10865 (riscv_expand_epilogue): Likewise.
10866 (riscv_function_ok_for_sibcall): Return false for naked function.
10867 (riscv_set_current_function): New.
10868 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10869 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10870 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10871 * doc/extend.texi (RISC-V Function Attributes): New.
10873 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10875 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10876 check for 128-bit long double before checking TCmode.
10877 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10878 128-bit long doubles before checking TFmode or TCmode.
10879 (FLOAT128_IBM_P): Likewise.
10881 2018-01-10 Martin Sebor <msebor@redhat.com>
10883 PR tree-optimization/83671
10884 * builtins.c (c_strlen): Unconditionally return zero for the empty
10886 Use -Warray-bounds for warnings.
10887 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10888 for non-constant array indices with COMPONENT_REF, arrays of
10889 arrays, and pointers to arrays.
10890 (gimple_fold_builtin_strlen): Determine and set length range for
10891 non-constant character arrays.
10893 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10895 PR middle-end/81897
10896 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10899 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10901 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10903 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10906 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10907 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10908 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10909 indexed_or_indirect_operand predicate.
10910 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10911 (*vsx_le_perm_load_v8hi): Likewise.
10912 (*vsx_le_perm_load_v16qi): Likewise.
10913 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10914 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10915 (*vsx_le_perm_store_v8hi): Likewise.
10916 (*vsx_le_perm_store_v16qi): Likewise.
10917 (eight unnamed splitters): Likewise.
10919 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10921 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10922 * config/rs6000/emmintrin.h: Likewise.
10923 * config/rs6000/mmintrin.h: Likewise.
10924 * config/rs6000/xmmintrin.h: Likewise.
10926 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10929 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10931 * tree.c (tree_nop_conversion): Return true for location wrapper
10933 (maybe_wrap_with_location): New function.
10934 (selftest::check_strip_nops): New function.
10935 (selftest::test_location_wrappers): New function.
10936 (selftest::tree_c_tests): Call it.
10937 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10938 (maybe_wrap_with_location): New decl.
10939 (EXPR_LOCATION_WRAPPER_P): New macro.
10940 (location_wrapper_p): New inline function.
10941 (tree_strip_any_location_wrapper): New inline function.
10943 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10946 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10947 stack_realign_offset for the largest alignment of stack slot
10949 (ix86_find_max_used_stack_alignment): New function.
10950 (ix86_finalize_stack_frame_flags): Use it. Set
10951 max_used_stack_alignment if we don't realign stack.
10952 * config/i386/i386.h (machine_function): Add
10953 max_used_stack_alignment.
10955 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10957 * config/arm/arm.opt (-mbranch-cost): New option.
10958 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10961 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10964 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10965 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10967 2018-01-10 Richard Biener <rguenther@suse.de>
10970 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10971 early out so it also covers the case where we have a non-NULL
10974 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10976 PR tree-optimization/83753
10977 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10978 for non-strided grouped accesses if the number of elements is 1.
10980 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10983 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10984 * i386.h (TARGET_USE_GATHER): Define.
10985 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10987 2018-01-10 Martin Liska <mliska@suse.cz>
10990 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10991 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10993 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10994 CLEANUP_NO_PARTITIONING is not set.
10996 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10998 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10999 for vectors, as a partial revert of r254296.
11000 * rtl.h (const_vec_p): Delete.
11001 (const_vec_duplicate_p): Don't test for vector CONSTs.
11002 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11003 * expmed.c (make_tree): Likewise.
11006 * common.md (E, F): Use CONSTANT_P instead of checking for
11008 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11009 checking for CONST_VECTOR.
11011 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11013 PR middle-end/83575
11014 * predict.c (force_edge_cold): Handle in more sane way edges
11015 with no prediction.
11017 2018-01-09 Carl Love <cel@us.ibm.com>
11019 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11021 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11022 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11023 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11024 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11025 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11026 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11027 * config/rs6000/rs6000-protos.h: Add extern defition for
11028 rs6000_generate_float2_double_code.
11029 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11031 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11032 (float2_v2df): Add define_expand.
11034 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11037 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11038 op_mode in the force_to_mode call.
11040 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11042 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11043 instead of checking each element individually.
11044 (aarch64_evpc_uzp): Likewise.
11045 (aarch64_evpc_zip): Likewise.
11046 (aarch64_evpc_ext): Likewise.
11047 (aarch64_evpc_rev): Likewise.
11048 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11049 instead of checking each element individually. Return true without
11051 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11052 whether all selected elements come from the same input, instead of
11053 checking each element individually. Remove calls to gen_rtx_REG,
11054 start_sequence and end_sequence and instead assert that no rtl is
11057 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11059 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11060 order of HIGH and CONST checks.
11062 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11064 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11065 if the destination isn't an SSA_NAME.
11067 2018-01-09 Richard Biener <rguenther@suse.de>
11069 PR tree-optimization/83668
11070 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11072 (canonicalize_loop_form): ... here, renamed from ...
11073 (canonicalize_loop_closed_ssa_form): ... this and amended to
11074 swap successor edges for loop exit blocks to make us use
11075 the RPO order we need for initial schedule generation.
11077 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11079 PR tree-optimization/64811
11080 * match.pd: When optimizing comparisons with Inf, avoid
11081 introducing or losing exceptions from comparisons with NaN.
11083 2018-01-09 Martin Liska <mliska@suse.cz>
11086 * asan.c (shadow_mem_size): Add gcc_assert.
11088 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11090 Don't save registers in main().
11093 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11094 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11095 * config/avr/avr.c (avr_set_current_function): Don't error if
11096 naked, OS_task or OS_main are specified at the same time.
11097 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11099 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11101 * common/config/avr/avr-common.c (avr_option_optimization_table):
11102 Switch on -mmain-is-OS_task for optimizing compilations.
11104 2018-01-09 Richard Biener <rguenther@suse.de>
11106 PR tree-optimization/83572
11107 * graphite.c: Include cfganal.h.
11108 (graphite_transform_loops): Connect infinite loops to exit
11109 and remove fake edges at the end.
11111 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11113 * ipa-inline.c (edge_badness): Revert accidental checkin.
11115 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11118 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11119 symbols; not inline clones.
11121 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11124 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11125 hard registers. Formatting fixes.
11127 PR preprocessor/83722
11128 * gcc.c (try_generate_repro): Pass
11129 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11130 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11133 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11134 Kito Cheng <kito.cheng@gmail.com>
11136 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11137 (riscv_leaf_function_p): Delete.
11138 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11140 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11142 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11144 (do_ifelse): New function.
11145 (do_isel): New function.
11146 (do_sub3): New function.
11147 (do_add3): New function.
11148 (do_load_mask_compare): New function.
11149 (do_overlap_load_compare): New function.
11150 (expand_compare_loop): New function.
11151 (expand_block_compare): Call expand_compare_loop() when appropriate.
11152 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11153 option description.
11154 (-mblock-compare-inline-loop-limit): New option.
11156 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11159 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11160 Reverse order of second and third operands in first alternative.
11161 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11162 of first and second elements in UNSPEC_VPERMR vector.
11163 (altivec_expand_vec_perm_le): Likewise.
11165 2018-01-08 Jeff Law <law@redhat.com>
11167 PR rtl-optimizatin/81308
11168 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11169 (process_switch): If group_case_labels makes a change, then set
11171 (pass_convert_switch::execute): If a switch is converted, then
11172 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11174 PR rtl-optimization/81308
11175 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11178 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11180 PR target/83663 - Revert r255946
11181 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11182 generation for cases where splatting a value is not useful.
11183 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11184 across a vec_duplicate and a paradoxical subreg forming a vector
11185 mode to a vec_concat.
11187 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11189 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11190 -march=armv8.3-a variants.
11191 * config/arm/t-multilib: Likewise.
11192 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11194 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11196 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11198 (cceq_ior_compare_complement): Give it a name so I can use it, and
11199 change boolean_or_operator predicate to boolean_operator so it can
11200 be used to generate a crand.
11201 (eqne): New code iterator.
11202 (bd/bd_neg): New code_attrs.
11203 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11204 a single define_insn.
11205 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11206 decrement (bdnzt/bdnzf/bdzt/bdzf).
11207 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11208 with the new names of the branch decrement patterns, and added the
11209 names of the branch decrement conditional patterns.
11211 2018-01-08 Richard Biener <rguenther@suse.de>
11213 PR tree-optimization/83563
11214 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11217 2018-01-08 Richard Biener <rguenther@suse.de>
11219 PR middle-end/83713
11220 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11222 2018-01-08 Richard Biener <rguenther@suse.de>
11224 PR tree-optimization/83685
11225 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11226 references to abnormals.
11228 2018-01-08 Richard Biener <rguenther@suse.de>
11231 * dwarf2out.c (output_indirect_strings): Handle empty
11232 skeleton_debug_str_hash.
11233 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11235 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11237 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11238 (emit_store_direct): Likewise.
11239 (arc_trampoline_adjust_address): Likewise.
11240 (arc_asm_trampoline_template): New function.
11241 (arc_initialize_trampoline): Use asm_trampoline_template.
11242 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11243 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11244 * config/arc/arc.md (flush_icache): Delete pattern.
11246 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11248 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11249 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11252 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11255 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11256 by not USED_FOR_TARGET.
11257 (make_pass_resolve_sw_modes): Likewise.
11259 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11261 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11264 2018-01-08 Richard Biener <rguenther@suse.de>
11266 PR middle-end/83580
11267 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11269 2018-01-08 Richard Biener <rguenther@suse.de>
11271 PR middle-end/83517
11272 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11274 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11276 PR middle-end/81897
11277 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11278 basic blocks with a small number of successors.
11279 (convert_control_dep_chain_into_preds): Improve handling of
11281 (dump_predicates): Split apart into...
11282 (dump_pred_chain): ...here...
11283 (dump_pred_info): ...and here.
11284 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11285 (can_chain_union_be_invalidated_p): Improve check for invalidation
11287 (uninit_uses_cannot_happen): Avoid unnecessary if
11288 convert_control_dep_chain_into_preds yielded nothing.
11290 2018-01-06 Martin Sebor <msebor@redhat.com>
11292 PR tree-optimization/83640
11293 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11294 subtracting negative offset from size.
11295 (builtin_access::overlap): Adjust offset bounds of the access to fall
11296 within the size of the object if possible.
11298 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11300 PR rtl-optimization/83699
11301 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11302 extract_bit_field_as_subreg to cases in which the extracted
11303 value is also a vector.
11305 * lra-constraints.c (process_alt_operands): Test for the equivalence
11306 substitutions when detecting a possible reload cycle.
11308 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11311 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11312 by default if flag_selective_schedling{,2}. Formatting fixes.
11314 PR rtl-optimization/83682
11315 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11316 if it has non-VECTOR_MODE element mode.
11317 (vec_duplicate_p): Likewise.
11319 PR middle-end/83694
11320 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11321 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11323 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11326 * config/i386/i386-builtin.def
11327 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11328 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11329 Require also OPTION_MASK_ISA_AVX512F in addition to
11330 OPTION_MASK_ISA_GFNI.
11331 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11332 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11333 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11334 to OPTION_MASK_ISA_GFNI.
11335 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11336 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11337 OPTION_MASK_ISA_AVX512BW.
11338 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11339 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11340 addition to OPTION_MASK_ISA_GFNI.
11341 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11342 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11343 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11344 to OPTION_MASK_ISA_GFNI.
11345 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11346 a requirement for all ISAs rather than any of them with a few
11348 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11350 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11351 bitmasks to be enabled with 3 exceptions, instead of requiring any
11352 enabled ISA with lots of exceptions.
11353 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11354 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11355 Change avx512bw in isa attribute to avx512f.
11356 * config/i386/sgxintrin.h: Add license boilerplate.
11357 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11358 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11359 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11360 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11362 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11363 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11364 temporarily sse2 rather than sse if not enabled already.
11367 * config/i386/sse.md (VI248_VLBW): Rename to ...
11368 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11369 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11370 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11371 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11372 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11373 mode iterator instead of VI248_VLBW.
11375 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11377 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11378 (record_modified): Skip clobbers; add debug output.
11379 (param_change_prob): Use sreal frequencies.
11381 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11383 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11384 punt for user-aligned variables.
11386 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11388 * tree-chrec.c (chrec_contains_symbols): Return true for
11391 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11394 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11395 of (x|y) == x for BICS pattern.
11397 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11399 PR tree-optimization/83605
11400 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11401 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11404 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11406 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11407 * config/epiphany/rtems.h: New file.
11409 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11410 Uros Bizjak <ubizjak@gmail.com>
11413 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11414 QIreg_operand instead of register_operand predicate.
11415 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11416 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11417 comments instead of -fmitigate[-_]rop.
11419 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11422 * cgraphunit.c (symbol_table::compile): Switch to text_section
11423 before calling assembly_start debug hook.
11424 * run-rtl-passes.c (run_rtl_passes): Likewise.
11427 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11429 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11430 range_int_cst_p rather than !symbolic_range_p before calling
11431 extract_range_from_multiplicative_op_1.
11433 2018-01-04 Jeff Law <law@redhat.com>
11435 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11436 redundant test in assertion.
11438 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11440 * doc/rtl.texi: Document machine_mode wrapper classes.
11442 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11444 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11445 using tree_to_uhwi.
11447 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11449 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11450 the VEC_PERM_EXPR fold to fail.
11452 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11455 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11456 to switched_sections.
11458 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11461 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11462 test for d.testing.
11464 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11467 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11468 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11470 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11473 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11474 is BLKmode and bitpos not zero or mode change is needed.
11476 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11479 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11482 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11485 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11486 instead of MULT rtx. Update all corresponding splitters.
11487 (*saddl_se): Ditto.
11488 (*ssub<modesuffix>): Ditto.
11489 (*ssubl_se): Ditto.
11490 (*cmp_sadd_di): Update split patterns.
11491 (*cmp_sadd_si): Ditto.
11492 (*cmp_sadd_sidi): Ditto.
11493 (*cmp_ssub_di): Ditto.
11494 (*cmp_ssub_si): Ditto.
11495 (*cmp_ssub_sidi): Ditto.
11496 * config/alpha/predicates.md (const23_operand): New predicate.
11497 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11498 Look for ASHIFT, not MULT inner operand.
11499 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11501 2018-01-04 Martin Liska <mliska@suse.cz>
11503 PR gcov-profile/83669
11504 * gcov.c (output_intermediate_file): Add version to intermediate
11506 * doc/gcov.texi: Document new field 'version' in intermediate
11507 file format. Fix location of '-k' option of gcov command.
11509 2018-01-04 Martin Liska <mliska@suse.cz>
11512 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11514 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11516 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11518 2018-01-03 Martin Sebor <msebor@redhat.com>
11520 PR tree-optimization/83655
11521 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11522 checking calls with invalid arguments.
11524 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11526 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11527 (vectorizable_mask_load_store): Delete.
11528 (vectorizable_call): Return false for masked loads and stores.
11529 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11530 instead of gimple_assign_rhs1.
11531 (vectorizable_load): Handle IFN_MASK_LOAD.
11532 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11534 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11536 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11538 (vectorizable_mask_load_store): ...here.
11539 (vectorizable_load): ...and here.
11541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11543 * tree-vect-stmts.c (vect_build_all_ones_mask)
11544 (vect_build_zero_merge_argument): New functions, split out from...
11545 (vectorizable_load): ...here.
11547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11549 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11551 (vectorizable_mask_load_store): ...here.
11552 (vectorizable_store): ...and here.
11554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11556 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11558 (vectorizable_mask_load_store): ...here.
11560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11562 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11563 (vect_model_store_cost): Take a vec_load_store_type instead of a
11565 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11566 (vect_model_store_cost): Take a vec_load_store_type instead of a
11568 (vectorizable_mask_load_store): Update accordingly.
11569 (vectorizable_store): Likewise.
11570 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11574 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11575 IFN_MASK_LOAD calls here rather than...
11576 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11578 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11579 Alan Hayward <alan.hayward@arm.com>
11580 David Sherwood <david.sherwood@arm.com>
11582 * expmed.c (extract_bit_field_1): For vector extracts,
11583 fall back to extract_bit_field_as_subreg if vec_extract
11586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11587 Alan Hayward <alan.hayward@arm.com>
11588 David Sherwood <david.sherwood@arm.com>
11590 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11591 they are variable or constant sized.
11592 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11593 slots for constant-sized data.
11595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11596 Alan Hayward <alan.hayward@arm.com>
11597 David Sherwood <david.sherwood@arm.com>
11599 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11600 handling COND_EXPRs with boolean comparisons, try to find a better
11601 basis for the mask type than the boolean itself.
11603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11605 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11606 is calculated and how it can be overridden.
11607 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11608 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11610 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11614 Alan Hayward <alan.hayward@arm.com>
11615 David Sherwood <david.sherwood@arm.com>
11617 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11618 Remove the mode argument.
11619 (aarch64_simd_valid_immediate): Remove the mode and inverse
11621 * config/aarch64/iterators.md (bitsize): New iterator.
11622 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11623 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11624 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11625 aarch64_simd_valid_immediate.
11626 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11627 (aarch64_reg_or_bic_imm): Likewise.
11628 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11629 with an insn_type enum and msl with a modifier_type enum.
11630 Replace element_width with a scalar_mode. Change the shift
11631 to unsigned int. Add constructors for scalar_float_mode and
11632 scalar_int_mode elements.
11633 (aarch64_vect_float_const_representable_p): Delete.
11634 (aarch64_can_const_movi_rtx_p)
11635 (aarch64_simd_scalar_immediate_valid_for_move)
11636 (aarch64_simd_make_constant): Update call to
11637 aarch64_simd_valid_immediate.
11638 (aarch64_advsimd_valid_immediate_hs): New function.
11639 (aarch64_advsimd_valid_immediate): Likewise.
11640 (aarch64_simd_valid_immediate): Remove mode and inverse
11641 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11642 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11643 and aarch64_float_const_representable_p on the result.
11644 (aarch64_output_simd_mov_immediate): Remove mode argument.
11645 Update call to aarch64_simd_valid_immediate and use of
11646 simd_immediate_info.
11647 (aarch64_output_scalar_simd_mov_immediate): Update call
11650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11651 Alan Hayward <alan.hayward@arm.com>
11652 David Sherwood <david.sherwood@arm.com>
11654 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11655 (mode_nunits): Likewise CONST_MODE_NUNITS.
11656 * machmode.def (ADJUST_NUNITS): Document.
11657 * genmodes.c (mode_data::need_nunits_adj): New field.
11658 (blank_mode): Update accordingly.
11659 (adj_nunits): New variable.
11660 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11662 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11663 listed in adj_nunits.
11664 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11665 listed in adj_nunits. Don't emit case statements for such modes.
11666 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11667 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11668 nothing if adj_nunits is nonnull.
11669 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11670 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11671 (emit_mode_fbit): Update use of print_maybe_const_decl.
11672 (emit_move_size): Likewise. Treat the array as non-const
11674 (emit_mode_adjustments): Handle adj_nunits.
11676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11678 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11679 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11680 (VECTOR_MODES): Use it.
11681 (make_vector_modes): Take the prefix as an argument.
11683 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11684 Alan Hayward <alan.hayward@arm.com>
11685 David Sherwood <david.sherwood@arm.com>
11687 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11688 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11689 for MODE_VECTOR_BOOL.
11690 * machmode.def (VECTOR_BOOL_MODE): Document.
11691 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11692 (make_vector_bool_mode): New function.
11693 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11695 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11696 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11698 * stor-layout.c (int_mode_for_mode): Likewise.
11699 * tree.c (build_vector_type_for_mode): Likewise.
11700 * varasm.c (output_constant_pool_2): Likewise.
11701 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11702 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11703 for MODE_VECTOR_BOOL.
11704 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11705 of mode class checks.
11706 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11707 instead of a list of mode class checks.
11708 (expand_vector_scalar_condition): Likewise.
11709 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11712 Alan Hayward <alan.hayward@arm.com>
11713 David Sherwood <david.sherwood@arm.com>
11715 * machmode.h (mode_size): Change from unsigned short to
11717 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11718 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11719 or if measurement_type is not polynomial.
11720 (fixed_size_mode::includes_p): Check for constant-sized modes.
11721 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11722 return a poly_uint16 rather than an unsigned short.
11723 (emit_mode_size): Change the type of mode_size from unsigned short
11724 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11725 (emit_mode_adjustments): Cope with polynomial vector sizes.
11726 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11728 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11730 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11731 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11732 * caller-save.c (setup_save_areas): Likewise.
11733 (replace_reg_with_saved_mem): Likewise.
11734 * calls.c (emit_library_call_value_1): Likewise.
11735 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11736 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11737 (gen_lowpart_for_combine): Likewise.
11738 * convert.c (convert_to_integer_1): Likewise.
11739 * cse.c (equiv_constant, cse_insn): Likewise.
11740 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11741 (cselib_subst_to_values): Likewise.
11742 * dce.c (word_dce_process_block): Likewise.
11743 * df-problems.c (df_word_lr_mark_ref): Likewise.
11744 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11745 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11746 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11747 (rtl_for_decl_location): Likewise.
11748 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11749 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11750 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11751 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11752 (expand_expr_real_1): Likewise.
11753 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11754 (pad_below): Likewise.
11755 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11756 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11757 * ira.c (get_subreg_tracking_sizes): Likewise.
11758 * ira-build.c (ira_create_allocno_objects): Likewise.
11759 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11760 (ira_sort_regnos_for_alter_reg): Likewise.
11761 * ira-costs.c (record_operand_costs): Likewise.
11762 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11763 (resolve_simple_move): Likewise.
11764 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11765 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11766 (lra_constraints): Likewise.
11767 (CONST_POOL_OK_P): Reject variable-sized modes.
11768 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11769 (add_pseudo_to_slot, lra_spill): Likewise.
11770 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11771 * optabs-query.c (get_best_extraction_insn): Likewise.
11772 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11773 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11774 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11775 * recog.c (offsettable_address_addr_space_p): Likewise.
11776 * regcprop.c (maybe_mode_change): Likewise.
11777 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11778 * regrename.c (build_def_use): Likewise.
11779 * regstat.c (dump_reg_info): Likewise.
11780 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11781 (find_reloads, find_reloads_subreg_address): Likewise.
11782 * reload1.c (eliminate_regs_1): Likewise.
11783 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11784 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11785 (simplify_binary_operation_1, simplify_subreg): Likewise.
11786 * targhooks.c (default_function_arg_padding): Likewise.
11787 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11788 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11789 (verify_gimple_assign_ternary): Likewise.
11790 * tree-inline.c (estimate_move_cost): Likewise.
11791 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11792 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11793 (get_address_cost_ainc): Likewise.
11794 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11795 (vect_supportable_dr_alignment): Likewise.
11796 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11797 (vectorizable_reduction): Likewise.
11798 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11799 (vectorizable_operation, vectorizable_load): Likewise.
11800 * tree.c (build_same_sized_truth_vector_type): Likewise.
11801 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11802 * var-tracking.c (emit_note_insn_var_location): Likewise.
11803 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11804 (ADDR_VEC_ALIGN): Likewise.
11806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11807 Alan Hayward <alan.hayward@arm.com>
11808 David Sherwood <david.sherwood@arm.com>
11810 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11812 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11813 or if measurement_type is polynomial.
11814 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11815 * combine.c (make_extraction): Likewise.
11816 * dse.c (find_shift_sequence): Likewise.
11817 * dwarf2out.c (mem_loc_descriptor): Likewise.
11818 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11819 (extract_bit_field, extract_low_bits): Likewise.
11820 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11821 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11822 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11823 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11824 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11825 * reload.c (find_reloads): Likewise.
11826 * reload1.c (alter_reg): Likewise.
11827 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11828 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11829 * tree-if-conv.c (predicate_mem_writes): Likewise.
11830 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11831 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11832 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11833 * valtrack.c (dead_debug_insert_temp): Likewise.
11834 * varasm.c (mergeable_constant_section): Likewise.
11835 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11837 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11838 Alan Hayward <alan.hayward@arm.com>
11839 David Sherwood <david.sherwood@arm.com>
11841 * expr.c (expand_assignment): Cope with polynomial mode sizes
11842 when assigning to a CONCAT.
11844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11845 Alan Hayward <alan.hayward@arm.com>
11846 David Sherwood <david.sherwood@arm.com>
11848 * machmode.h (mode_precision): Change from unsigned short to
11850 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11852 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11853 or if measurement_type is not polynomial.
11854 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11855 in which the mode is already known to be a scalar_int_mode.
11856 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11857 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11859 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11860 for GET_MODE_PRECISION.
11861 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11862 for GET_MODE_PRECISION.
11863 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11865 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11866 (expand_field_assignment, make_extraction): Likewise.
11867 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11868 (get_last_value): Likewise.
11869 * convert.c (convert_to_integer_1): Likewise.
11870 * cse.c (cse_insn): Likewise.
11871 * expr.c (expand_expr_real_1): Likewise.
11872 * lra-constraints.c (simplify_operand_subreg): Likewise.
11873 * optabs-query.c (can_atomic_load_p): Likewise.
11874 * optabs.c (expand_atomic_load): Likewise.
11875 (expand_atomic_store): Likewise.
11876 * ree.c (combine_reaching_defs): Likewise.
11877 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11878 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11879 * tree.h (type_has_mode_precision_p): Likewise.
11880 * ubsan.c (instrument_si_overflow): Likewise.
11882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11883 Alan Hayward <alan.hayward@arm.com>
11884 David Sherwood <david.sherwood@arm.com>
11886 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11887 polynomial numbers of units.
11888 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11889 (valid_vector_subparts_p): New function.
11890 (build_vector_type): Remove temporary shim and take the number
11891 of units as a poly_uint64 rather than an int.
11892 (build_opaque_vector_type): Take the number of units as a
11893 poly_uint64 rather than an int.
11894 * tree.c (build_vector_from_ctor): Handle polynomial
11895 TYPE_VECTOR_SUBPARTS.
11896 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11897 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11898 (build_vector_from_val): If the number of units is variable,
11899 use build_vec_duplicate_cst for constant operands and
11900 VEC_DUPLICATE_EXPR otherwise.
11901 (make_vector_type): Remove temporary is_constant ().
11902 (build_vector_type, build_opaque_vector_type): Take the number of
11903 units as a poly_uint64 rather than an int.
11904 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11906 * cfgexpand.c (expand_debug_expr): Likewise.
11907 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11908 (store_constructor, expand_expr_real_1): Likewise.
11909 (const_scalar_mask_from_tree): Likewise.
11910 * fold-const-call.c (fold_const_reduction): Likewise.
11911 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11912 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11913 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11914 (fold_relational_const): Likewise.
11915 (native_interpret_vector): Likewise. Change the size from an
11916 int to an unsigned int.
11917 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11918 TYPE_VECTOR_SUBPARTS.
11919 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11920 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11921 duplicating a non-constant operand into a variable-length vector.
11922 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11923 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11924 * ipa-icf.c (sem_variable::equals): Likewise.
11925 * match.pd: Likewise.
11926 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11927 * print-tree.c (print_node): Likewise.
11928 * stor-layout.c (layout_type): Likewise.
11929 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11930 * tree-cfg.c (verify_gimple_comparison): Likewise.
11931 (verify_gimple_assign_binary): Likewise.
11932 (verify_gimple_assign_ternary): Likewise.
11933 (verify_gimple_assign_single): Likewise.
11934 * tree-pretty-print.c (dump_generic_node): Likewise.
11935 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11936 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11937 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11938 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11939 (vect_shift_permute_load_chain): Likewise.
11940 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11941 (expand_vector_condition, optimize_vector_constructor): Likewise.
11942 (lower_vec_perm, get_compute_type): Likewise.
11943 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11944 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11945 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11946 (vect_recog_mask_conversion_pattern): Likewise.
11947 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11948 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11949 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11950 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11951 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11952 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11953 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11954 (supportable_widening_operation): Likewise.
11955 (supportable_narrowing_operation): Likewise.
11956 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11958 * varasm.c (output_constant): Likewise.
11960 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11961 Alan Hayward <alan.hayward@arm.com>
11962 David Sherwood <david.sherwood@arm.com>
11964 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11965 so that both the length == 3 and length != 3 cases set up their
11966 own permute vectors. Add comments explaining why we know the
11967 number of elements is constant.
11968 (vect_permute_load_chain): Likewise.
11970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11971 Alan Hayward <alan.hayward@arm.com>
11972 David Sherwood <david.sherwood@arm.com>
11974 * machmode.h (mode_nunits): Change from unsigned char to
11976 (ONLY_FIXED_SIZE_MODES): New macro.
11977 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11978 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11979 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11981 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11982 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11983 or if measurement_type is not polynomial.
11984 * genmodes.c (ZERO_COEFFS): New macro.
11985 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11987 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11988 Use ZERO_COEFFS when emitting initializers.
11989 * data-streamer.h (bp_pack_poly_value): New function.
11990 (bp_unpack_poly_value): Likewise.
11991 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11992 for GET_MODE_NUNITS.
11993 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11994 for GET_MODE_NUNITS.
11995 * tree.c (make_vector_type): Remove temporary shim and make
11996 the real function take the number of units as a poly_uint64
11997 rather than an int.
11998 (build_vector_type_for_mode): Handle polynomial nunits.
11999 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12000 * emit-rtl.c (const_vec_series_p_1): Likewise.
12001 (gen_rtx_CONST_VECTOR): Likewise.
12002 * fold-const.c (test_vec_duplicate_folding): Likewise.
12003 * genrecog.c (validate_pattern): Likewise.
12004 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12005 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12006 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12007 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12008 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12009 * rtlanal.c (subreg_get_info): Likewise.
12010 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12011 (vect_grouped_load_supported): Likewise.
12012 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12013 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12014 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12015 (simplify_const_unary_operation, simplify_binary_operation_1)
12016 (simplify_const_binary_operation, simplify_ternary_operation)
12017 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12018 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12019 instead of CONST_VECTOR_NUNITS.
12020 * varasm.c (output_constant_pool_2): Likewise.
12021 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12022 explicit-encoded elements in the XVEC for variable-length vectors.
12024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12026 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12029 Alan Hayward <alan.hayward@arm.com>
12030 David Sherwood <david.sherwood@arm.com>
12032 * coretypes.h (fixed_size_mode): Declare.
12033 (fixed_size_mode_pod): New typedef.
12034 * builtins.h (target_builtins::x_apply_args_mode)
12035 (target_builtins::x_apply_result_mode): Change type to
12036 fixed_size_mode_pod.
12037 * builtins.c (apply_args_size, apply_result_size, result_vector)
12038 (expand_builtin_apply_args_1, expand_builtin_apply)
12039 (expand_builtin_return): Update accordingly.
12041 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12043 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12044 * cselib.c (cselib_hash_rtx): Likewise.
12045 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12046 CONST_VECTOR encoding.
12048 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12049 Jeff Law <law@redhat.com>
12052 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12053 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12054 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12055 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12058 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12059 explicitly probe *sp in a noreturn function if there were any callee
12060 register saves or frame pointer is needed.
12062 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12065 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12066 BLKmode for ternary, binary or unary expressions.
12069 * var-tracking.c (delete_vta_debug_insn): New inline function.
12070 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12071 insns from get_insns () to NULL instead of each bb separately.
12072 Use delete_vta_debug_insn. No longer static.
12073 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12074 delete_vta_debug_insns callers.
12075 * rtl.h (delete_vta_debug_insns): Declare.
12076 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12077 instead of variable_tracking_main.
12079 2018-01-03 Martin Sebor <msebor@redhat.com>
12081 PR tree-optimization/83603
12082 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12083 arguments past the endof the argument list in functions declared
12084 without a prototype.
12085 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12086 Avoid checking when arguments are null.
12088 2018-01-03 Martin Sebor <msebor@redhat.com>
12091 * doc/extend.texi (attribute const): Fix a typo.
12092 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12093 issuing -Wsuggest-attribute for void functions.
12095 2018-01-03 Martin Sebor <msebor@redhat.com>
12097 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12098 offset_int::from instead of wide_int::to_shwi.
12099 (maybe_diag_overlap): Remove assertion.
12100 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12101 * gimple-ssa-sprintf.c (format_directive): Same.
12102 (parse_directive): Same.
12103 (sprintf_dom_walker::compute_format_length): Same.
12104 (try_substitute_return_value): Same.
12106 2018-01-03 Jeff Law <law@redhat.com>
12108 PR middle-end/83654
12109 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12110 non-constant residual for zero at runtime and avoid probing in
12111 that case. Reorganize code for trailing problem to mirror handling
12114 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12116 PR tree-optimization/83501
12117 * tree-ssa-strlen.c (get_string_cst): New.
12118 (handle_char_store): Call get_string_cst.
12120 2018-01-03 Martin Liska <mliska@suse.cz>
12122 PR tree-optimization/83593
12123 * tree-ssa-strlen.c: Include tree-cfg.h.
12124 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12125 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12126 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12128 (strlen_dom_walker::before_dom_children): Call
12129 gimple_purge_dead_eh_edges. Dump tranformation with details
12131 (strlen_dom_walker::before_dom_children): Update call by adding
12132 new argument cleanup_eh.
12133 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12135 2018-01-03 Martin Liska <mliska@suse.cz>
12138 * cif-code.def (VARIADIC_THUNK): New enum value.
12139 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12142 2018-01-03 Jan Beulich <jbeulich@suse.com>
12144 * sse.md (mov<mode>_internal): Tighten condition for when to use
12145 vmovdqu<ssescalarsize> for TI and OI modes.
12147 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12149 Update copyright years.
12151 2018-01-03 Martin Liska <mliska@suse.cz>
12154 * ipa-visibility.c (function_and_variable_visibility): Skip
12155 functions with noipa attribure.
12157 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12159 * gcc.c (process_command): Update copyright notice dates.
12160 * gcov-dump.c (print_version): Ditto.
12161 * gcov.c (print_version): Ditto.
12162 * gcov-tool.c (print_version): Ditto.
12163 * gengtype.c (create_file): Ditto.
12164 * doc/cpp.texi: Bump @copying's copyright year.
12165 * doc/cppinternals.texi: Ditto.
12166 * doc/gcc.texi: Ditto.
12167 * doc/gccint.texi: Ditto.
12168 * doc/gcov.texi: Ditto.
12169 * doc/install.texi: Ditto.
12170 * doc/invoke.texi: Ditto.
12172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12174 * vector-builder.h (vector_builder::m_full_nelts): Change from
12175 unsigned int to poly_uint64.
12176 (vector_builder::full_nelts): Update prototype accordingly.
12177 (vector_builder::new_vector): Likewise.
12178 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12179 (vector_builder::operator ==): Likewise.
12180 (vector_builder::finalize): Likewise.
12181 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12182 Take the number of elements as a poly_uint64 rather than an
12184 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12185 from unsigned int to poly_uint64.
12186 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12187 (vec_perm_indices::new_vector): Likewise.
12188 (vec_perm_indices::length): Likewise.
12189 (vec_perm_indices::nelts_per_input): Likewise.
12190 (vec_perm_indices::input_nelts): Likewise.
12191 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12192 number of elements per input as a poly_uint64 rather than an
12193 unsigned int. Use the original encoding for variable-length
12194 vectors, rather than clamping each individual element.
12195 For the second and subsequent elements in each pattern,
12196 clamp the step and base before clamping their sum.
12197 (vec_perm_indices::series_p): Handle polynomial element counts.
12198 (vec_perm_indices::all_in_range_p): Likewise.
12199 (vec_perm_indices_to_tree): Likewise.
12200 (vec_perm_indices_to_rtx): Likewise.
12201 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12202 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12203 (tree_vector_builder::new_binary_operation): Handle polynomial
12204 element counts. Return false if we need to know the number
12205 of elements at compile time.
12206 * fold-const.c (fold_vec_perm): Punt if the number of elements
12207 isn't known at compile time.
12209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12211 * vec-perm-indices.h (vec_perm_builder): Change element type
12212 from HOST_WIDE_INT to poly_int64.
12213 (vec_perm_indices::element_type): Update accordingly.
12214 (vec_perm_indices::clamp): Handle polynomial element_types.
12215 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12216 (vec_perm_indices::all_in_range_p): Likewise.
12217 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12219 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12220 polynomial vec_perm_indices element types.
12221 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12222 * fold-const.c (fold_vec_perm): Likewise.
12223 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12224 * tree-vect-generic.c (lower_vec_perm): Likewise.
12225 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12226 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12227 element type to HOST_WIDE_INT.
12229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12230 Alan Hayward <alan.hayward@arm.com>
12231 David Sherwood <david.sherwood@arm.com>
12233 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12234 rather than an int. Use plus_constant.
12235 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12236 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12239 Alan Hayward <alan.hayward@arm.com>
12240 David Sherwood <david.sherwood@arm.com>
12242 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12243 a HOST_WIDE_INT to a poly_int64.
12245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12246 Alan Hayward <alan.hayward@arm.com>
12247 David Sherwood <david.sherwood@arm.com>
12249 * calls.c (load_register_parameters): Cope with polynomial
12250 mode sizes. Require a constant size for BLKmode parameters
12251 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12252 forces a parameter to be padded at the lsb end in order to
12253 fill a complete number of words, require the parameter size
12254 to be ordered wrt UNITS_PER_WORD.
12256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12257 Alan Hayward <alan.hayward@arm.com>
12258 David Sherwood <david.sherwood@arm.com>
12260 * reload1.c (spill_stack_slot_width): Change element type
12261 from unsigned int to poly_uint64_pod.
12262 (alter_reg): Treat mode sizes as polynomial.
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 * reload.c (complex_word_subreg_p): New function.
12269 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
12276 operands whose sizes aren't ordered.
12277 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12284 that the mode size is in the set {1, 2, 4, 8, 16}.
12286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12287 Alan Hayward <alan.hayward@arm.com>
12288 David Sherwood <david.sherwood@arm.com>
12290 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12291 Use plus_constant instead of gen_rtx_PLUS.
12293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12294 Alan Hayward <alan.hayward@arm.com>
12295 David Sherwood <david.sherwood@arm.com>
12297 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12298 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12299 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12300 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12301 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12302 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12303 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12304 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12305 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12306 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12308 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12309 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12310 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12311 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12312 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12313 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12314 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12315 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12316 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12317 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12319 * expr.c (emit_move_resolve_push): Treat the input and result
12320 of PUSH_ROUNDING as a poly_int64.
12321 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12322 (emit_push_insn): Likewise.
12323 * lra-eliminations.c (mark_not_eliminable): Likewise.
12324 * recog.c (push_operand): Likewise.
12325 * reload1.c (elimination_effects): Likewise.
12326 * rtlanal.c (nonzero_bits1): Likewise.
12327 * calls.c (store_one_arg): Likewise. Require the padding to be
12328 known at compile time.
12330 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12331 Alan Hayward <alan.hayward@arm.com>
12332 David Sherwood <david.sherwood@arm.com>
12334 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12335 Use plus_constant instead of gen_rtx_PLUS.
12337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12338 Alan Hayward <alan.hayward@arm.com>
12339 David Sherwood <david.sherwood@arm.com>
12341 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12342 rather than an int.
12344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12345 Alan Hayward <alan.hayward@arm.com>
12346 David Sherwood <david.sherwood@arm.com>
12348 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12349 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12350 via stack temporaries. Treat the mode size as polynomial too.
12352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12353 Alan Hayward <alan.hayward@arm.com>
12354 David Sherwood <david.sherwood@arm.com>
12356 * expr.c (expand_expr_real_2): When handling conversions involving
12357 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12358 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12359 as a poly_uint64 too.
12361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12362 Alan Hayward <alan.hayward@arm.com>
12363 David Sherwood <david.sherwood@arm.com>
12365 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
12372 REGMODE_NATURAL_SIZE.
12373 * expmed.c (store_bit_field_1): Likewise.
12374 * expr.c (store_constructor): Likewise.
12375 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12376 and polynomial REGMODE_NATURAL_SIZE.
12377 (gen_lowpart_common): Likewise.
12378 * reginfo.c (record_subregs_of_mode): Likewise.
12379 * rtlanal.c (read_modify_subreg_p): Likewise.
12381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12382 Alan Hayward <alan.hayward@arm.com>
12383 David Sherwood <david.sherwood@arm.com>
12385 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12386 numbers of elements.
12388 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12389 Alan Hayward <alan.hayward@arm.com>
12390 David Sherwood <david.sherwood@arm.com>
12392 * match.pd: Cope with polynomial numbers of vector elements.
12394 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12395 Alan Hayward <alan.hayward@arm.com>
12396 David Sherwood <david.sherwood@arm.com>
12398 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12399 in a POINTER_PLUS_EXPR.
12401 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12402 Alan Hayward <alan.hayward@arm.com>
12403 David Sherwood <david.sherwood@arm.com>
12405 * omp-simd-clone.c (simd_clone_subparts): New function.
12406 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12407 (ipa_simd_modify_function_body): Likewise.
12409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12410 Alan Hayward <alan.hayward@arm.com>
12411 David Sherwood <david.sherwood@arm.com>
12413 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12414 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12415 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12416 (expand_vector_condition, vector_element): Likewise.
12417 (subparts_gt): New function.
12418 (get_compute_type): Use subparts_gt.
12419 (count_type_subparts): Delete.
12420 (expand_vector_operations_1): Use subparts_gt instead of
12421 count_type_subparts.
12423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12424 Alan Hayward <alan.hayward@arm.com>
12425 David Sherwood <david.sherwood@arm.com>
12427 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12428 (vect_compile_time_alias): ...this new function. Do the calculation
12429 on poly_ints rather than trees.
12430 (vect_prune_runtime_alias_test_list): Update call accordingly.
12432 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12433 Alan Hayward <alan.hayward@arm.com>
12434 David Sherwood <david.sherwood@arm.com>
12436 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12438 (vect_schedule_slp_instance): Likewise.
12440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12441 Alan Hayward <alan.hayward@arm.com>
12442 David Sherwood <david.sherwood@arm.com>
12444 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12445 constant and extern definitions for variable-length vectors.
12446 (vect_get_constant_vectors): Note that the number of units
12447 is known to be constant.
12449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12450 Alan Hayward <alan.hayward@arm.com>
12451 David Sherwood <david.sherwood@arm.com>
12453 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12454 of units as polynomial. Choose between WIDE and NARROW based
12457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12458 Alan Hayward <alan.hayward@arm.com>
12459 David Sherwood <david.sherwood@arm.com>
12461 * tree-vect-stmts.c (simd_clone_subparts): New function.
12462 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12465 Alan Hayward <alan.hayward@arm.com>
12466 David Sherwood <david.sherwood@arm.com>
12468 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12469 vectors as polynomial. Use build_index_vector for
12470 IFN_GOMP_SIMD_LANE.
12472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12473 Alan Hayward <alan.hayward@arm.com>
12474 David Sherwood <david.sherwood@arm.com>
12476 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12477 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12478 for variable-length vectors.
12479 (vectorizable_mask_load_store): Treat the number of units as
12480 polynomial, asserting that it is constant if the condition has
12481 already been enforced.
12482 (vectorizable_store, vectorizable_load): Likewise.
12484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12485 Alan Hayward <alan.hayward@arm.com>
12486 David Sherwood <david.sherwood@arm.com>
12488 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12489 of units as polynomial. Punt if we can't tell at compile time
12490 which vector contains the final result.
12492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12493 Alan Hayward <alan.hayward@arm.com>
12494 David Sherwood <david.sherwood@arm.com>
12496 * tree-vect-loop.c (vectorizable_induction): Treat the number
12497 of units as polynomial. Punt on SLP inductions. Use an integer
12498 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12499 cast of such a series for variable-length floating-point
12502 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12503 Alan Hayward <alan.hayward@arm.com>
12504 David Sherwood <david.sherwood@arm.com>
12506 * tree.h (build_index_vector): Declare.
12507 * tree.c (build_index_vector): New function.
12508 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12509 of units as polynomial, forcibly converting it to a constant if
12510 vectorizable_reduction has already enforced the condition.
12511 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12512 to create a {1,2,3,...} vector.
12513 (vectorizable_reduction): Treat the number of units as polynomial.
12514 Choose vectype_in based on the largest scalar element size rather
12515 than the smallest number of units. Enforce the restrictions
12518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12519 Alan Hayward <alan.hayward@arm.com>
12520 David Sherwood <david.sherwood@arm.com>
12522 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12523 number of units as polynomial.
12525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12526 Alan Hayward <alan.hayward@arm.com>
12527 David Sherwood <david.sherwood@arm.com>
12529 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12530 * target.def (autovectorize_vector_sizes): Return the vector sizes
12531 by pointer, using vector_sizes rather than a bitmask.
12532 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12533 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12534 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12536 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12537 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12538 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12539 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12540 * omp-general.c (omp_max_vf): Likewise.
12541 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12542 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12543 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12544 * tree-vect-slp.c (vect_slp_bb): Likewise.
12545 * doc/tm.texi: Regenerate.
12546 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12548 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12549 the vector size as a poly_uint64 rather than an unsigned int.
12550 (current_vector_size): Change from an unsigned int to a poly_uint64.
12551 (get_vectype_for_scalar_type): Update accordingly.
12552 * tree.h (build_truth_vector_type): Take the size and number of
12553 units as a poly_uint64 rather than an unsigned int.
12554 (build_vector_type): Add a temporary overload that takes
12555 the number of units as a poly_uint64 rather than an unsigned int.
12556 * tree.c (make_vector_type): Likewise.
12557 (build_truth_vector_type): Take the number of units as a poly_uint64
12558 rather than an unsigned int.
12560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12561 Alan Hayward <alan.hayward@arm.com>
12562 David Sherwood <david.sherwood@arm.com>
12564 * target.def (get_mask_mode): Take the number of units and length
12565 as poly_uint64s rather than unsigned ints.
12566 * targhooks.h (default_get_mask_mode): Update accordingly.
12567 * targhooks.c (default_get_mask_mode): Likewise.
12568 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12569 * doc/tm.texi: Regenerate.
12571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12572 Alan Hayward <alan.hayward@arm.com>
12573 David Sherwood <david.sherwood@arm.com>
12575 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12576 * omp-general.c (omp_max_vf): Likewise.
12577 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12578 (expand_omp_simd): Handle polynomial safelen.
12579 * omp-low.c (omplow_simd_context): Add a default constructor.
12580 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12581 (lower_rec_simd_input_clauses): Update accordingly.
12582 (lower_rec_input_clauses): Likewise.
12584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12585 Alan Hayward <alan.hayward@arm.com>
12586 David Sherwood <david.sherwood@arm.com>
12588 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12589 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12590 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12591 (vect_analyze_slp_cost): Likewise.
12592 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12593 (vect_model_load_cost): Likewise.
12595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12596 Alan Hayward <alan.hayward@arm.com>
12597 David Sherwood <david.sherwood@arm.com>
12599 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12600 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12601 from an unsigned int * to a poly_uint64_pod *.
12602 (calculate_unrolling_factor): New function.
12603 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12606 Alan Hayward <alan.hayward@arm.com>
12607 David Sherwood <david.sherwood@arm.com>
12609 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12610 from an unsigned int to a poly_uint64.
12611 (_loop_vec_info::slp_unrolling_factor): Likewise.
12612 (_loop_vec_info::vectorization_factor): Change from an int
12614 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12615 (vect_get_num_vectors): New function.
12616 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12617 (vect_get_num_copies): Use vect_get_num_vectors.
12618 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12619 to an unsigned int *.
12620 (vect_analyze_data_refs): Change min_vf from an int * to a
12622 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12623 than an unsigned HOST_WIDE_INT.
12624 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12625 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12626 to an unsigned int *.
12627 (vect_analyze_data_ref_dependences): Likewise.
12628 (vect_compute_data_ref_alignment): Handle polynomial vf.
12629 (vect_enhance_data_refs_alignment): Likewise.
12630 (vect_prune_runtime_alias_test_list): Likewise.
12631 (vect_shift_permute_load_chain): Likewise.
12632 (vect_supportable_dr_alignment): Likewise.
12633 (dependence_distance_ge_vf): Take the vectorization factor as a
12634 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12635 (vect_analyze_data_refs): Change min_vf from an int * to a
12637 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12638 vfm1 as a poly_uint64 rather than an int. Make the same change
12639 for the returned bound_scalar.
12640 (vect_gen_vector_loop_niters): Handle polynomial vf.
12641 (vect_do_peeling): Likewise. Update call to
12642 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12643 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12645 * tree-vect-loop.c (vect_determine_vectorization_factor)
12646 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12647 (vect_get_known_peeling_cost): Likewise.
12648 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12649 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12650 (vect_transform_loop): Likewise. Use the lowest possible VF when
12651 updating the upper bounds of the loop.
12652 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12653 rather than an int.
12654 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12655 polynomial unroll factors.
12656 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12657 (vect_make_slp_decision): Likewise.
12658 (vect_supported_load_permutation_p): Likewise, and polynomial
12660 (vect_analyze_slp_cost): Handle polynomial vf.
12661 (vect_slp_analyze_node_operations): Likewise.
12662 (vect_slp_analyze_bb_1): Likewise.
12663 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12664 than an unsigned HOST_WIDE_INT.
12665 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12666 (vectorizable_load): Handle polynomial vf.
12667 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12669 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12672 Alan Hayward <alan.hayward@arm.com>
12673 David Sherwood <david.sherwood@arm.com>
12675 * match.pd: Handle bit operations involving three constants
12676 and try to fold one pair.
12678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12680 * tree-vect-loop-manip.c: Include gimple-fold.h.
12681 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12682 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12683 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12684 Add a path that uses a step of VF instead of 1, but disable it
12686 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12687 and niters_no_overflow parameters. Update calls to
12688 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12689 Create a new SSA name if the latter choses to use a ste other
12690 than zero, and return it via niters_vector_mult_vf_var.
12691 * tree-vect-loop.c (vect_transform_loop): Update calls to
12692 vect_do_peeling, vect_gen_vector_loop_niters and
12693 slpeel_make_loop_iterate_ntimes.
12694 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12695 (vect_gen_vector_loop_niters): Update declarations after above changes.
12697 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12699 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12700 128-bit round to integer instructions.
12701 (ceil<mode>2): Likewise.
12702 (btrunc<mode>2): Likewise.
12703 (round<mode>2): Likewise.
12705 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12707 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12708 unaligned VSX load/store on P8/P9.
12709 (expand_block_clear): Allow the use of unaligned VSX
12710 load/store on P8/P9.
12712 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12714 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12716 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12717 swap associated with both a load and a store.
12719 2018-01-02 Andrew Waterman <andrew@sifive.com>
12721 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12722 * config/riscv/riscv.md (clear_cache): Use it.
12724 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12726 * web.c: Remove out-of-date comment.
12728 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12730 * expr.c (fixup_args_size_notes): Check that any existing
12731 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12732 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12733 (emit_single_push_insn): ...here.
12735 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12737 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12738 (const_vector_encoded_nelts): New function.
12739 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12740 (const_vector_int_elt, const_vector_elt): Declare.
12741 * emit-rtl.c (const_vector_int_elt_1): New function.
12742 (const_vector_elt): Likewise.
12743 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12744 of CONST_VECTOR_ELT.
12746 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12748 * expr.c: Include rtx-vector-builder.h.
12749 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12750 directly on the tree encoding.
12751 (const_vector_from_tree): Likewise.
12752 * optabs.c: Include rtx-vector-builder.h.
12753 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12754 sequence of "u" values.
12755 * vec-perm-indices.c: Include rtx-vector-builder.h.
12756 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12757 directly on the vec_perm_indices encoding.
12759 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12761 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12762 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12763 * rtx-vector-builder.h: New file.
12764 * rtx-vector-builder.c: Likewise.
12765 * rtl.h (rtx_def::u2): Add a const_vector field.
12766 (CONST_VECTOR_NPATTERNS): New macro.
12767 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12768 (CONST_VECTOR_DUPLICATE_P): Likewise.
12769 (CONST_VECTOR_STEPPED_P): Likewise.
12770 (CONST_VECTOR_ENCODED_ELT): Likewise.
12771 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12772 (unwrap_const_vec_duplicate): Likewise.
12773 (const_vec_series_p): Check for a non-duplicated vector encoding.
12774 Say that the function only returns true for integer vectors.
12775 * emit-rtl.c: Include rtx-vector-builder.h.
12776 (gen_const_vec_duplicate_1): Delete.
12777 (gen_const_vector): Call gen_const_vec_duplicate instead of
12778 gen_const_vec_duplicate_1.
12779 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12780 (gen_const_vec_duplicate): Use rtx_vector_builder.
12781 (gen_const_vec_series): Likewise.
12782 (gen_rtx_CONST_VECTOR): Likewise.
12783 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12784 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12785 Build a new vector rather than modifying a CONST_VECTOR in-place.
12786 (handle_special_swappables): Update call accordingly.
12787 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12788 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12789 Build a new vector rather than modifying a CONST_VECTOR in-place.
12790 (handle_special_swappables): Update call accordingly.
12792 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12794 * simplify-rtx.c (simplify_const_binary_operation): Use
12795 CONST_VECTOR_ELT instead of XVECEXP.
12797 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12799 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12800 the selector elements to be different from the data elements
12801 if the selector is a VECTOR_CST.
12802 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12803 ssizetype for the selector.
12805 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12807 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12808 before testing each element individually.
12809 * tree-vect-generic.c (lower_vec_perm): Likewise.
12811 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12813 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12814 * selftest-run-tests.c (selftest::run_tests): Call it.
12815 * vector-builder.h (vector_builder::operator ==): New function.
12816 (vector_builder::operator !=): Likewise.
12817 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12818 (vec_perm_indices::all_from_input_p): New function.
12819 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12820 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12821 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12822 instead of reading the VECTOR_CST directly. Detect whether both
12823 vector inputs are the same before constructing the vec_perm_indices,
12824 and update the number of inputs argument accordingly. Use the
12825 utility functions added above. Only construct sel2 if we need to.
12827 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12829 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12830 the broadcast of the low byte.
12831 (expand_mult_highpart): Use an explicit encoding for the permutes.
12832 * optabs-query.c (can_mult_highpart_p): Likewise.
12833 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12834 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12835 (vectorizable_bswap): Likewise.
12836 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12837 explicit encoding for the power-of-2 permutes.
12838 (vect_permute_store_chain): Likewise.
12839 (vect_grouped_load_supported): Likewise.
12840 (vect_permute_load_chain): Likewise.
12842 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12844 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12845 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12846 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12847 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12848 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12849 (vect_gen_perm_mask_any): Likewise.
12851 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12853 * int-vector-builder.h: New file.
12854 * vec-perm-indices.h: Include int-vector-builder.h.
12855 (vec_perm_indices): Redefine as an int_vector_builder.
12856 (auto_vec_perm_indices): Delete.
12857 (vec_perm_builder): Redefine as a stand-alone class.
12858 (vec_perm_indices::vec_perm_indices): New function.
12859 (vec_perm_indices::clamp): Likewise.
12860 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12861 (vec_perm_indices::new_vector): New function.
12862 (vec_perm_indices::new_expanded_vector): Update for new
12863 vec_perm_indices class.
12864 (vec_perm_indices::rotate_inputs): New function.
12865 (vec_perm_indices::all_in_range_p): Operate directly on the
12866 encoded form, without computing elided elements.
12867 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12868 encoding. Update for new vec_perm_indices class.
12869 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12870 the given vec_perm_builder.
12871 (expand_vec_perm_var): Update vec_perm_builder constructor.
12872 (expand_mult_highpart): Use vec_perm_builder instead of
12873 auto_vec_perm_indices.
12874 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12875 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12876 or double series encoding as appropriate.
12877 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12878 vec_perm_indices instead of auto_vec_perm_indices.
12879 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12880 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12881 (vect_permute_store_chain): Likewise.
12882 (vect_grouped_load_supported): Likewise.
12883 (vect_permute_load_chain): Likewise.
12884 (vect_shift_permute_load_chain): Likewise.
12885 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12886 (vect_transform_slp_perm_load): Likewise.
12887 (vect_schedule_slp_instance): Likewise.
12888 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12889 (vectorizable_mask_load_store): Likewise.
12890 (vectorizable_bswap): Likewise.
12891 (vectorizable_store): Likewise.
12892 (vectorizable_load): Likewise.
12893 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12894 vec_perm_indices instead of auto_vec_perm_indices. Use
12895 tree_to_vec_perm_builder to read the vector from a tree.
12896 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12897 vec_perm_builder instead of a vec_perm_indices.
12898 (have_whole_vector_shift): Use vec_perm_builder and
12899 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12900 truncation to calc_vec_perm_mask_for_shift.
12901 (vect_create_epilog_for_reduction): Likewise.
12902 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12903 from auto_vec_perm_indices to vec_perm_indices.
12904 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12905 instead of changing individual elements.
12906 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12907 the vector in d.perm.
12908 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12909 from auto_vec_perm_indices to vec_perm_indices.
12910 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12911 instead of changing individual elements.
12912 (arm_vectorize_vec_perm_const): Use new_vector to install
12913 the vector in d.perm.
12914 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12915 Update vec_perm_builder constructor.
12916 (rs6000_expand_interleave): Likewise.
12917 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12918 (rs6000_expand_interleave): Likewise.
12920 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12922 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12923 to qimode could truncate the indices.
12924 * optabs.c (expand_vec_perm_var): Likewise.
12926 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12928 * Makefile.in (OBJS): Add vec-perm-indices.o.
12929 * vec-perm-indices.h: New file.
12930 * vec-perm-indices.c: Likewise.
12931 * target.h (vec_perm_indices): Replace with a forward class
12933 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12934 * optabs.h: Include vec-perm-indices.h.
12935 (expand_vec_perm): Delete.
12936 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12937 (expand_vec_perm_const): Declare.
12938 * target.def (vec_perm_const_ok): Replace with...
12939 (vec_perm_const): ...this new hook.
12940 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12941 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12942 * doc/tm.texi: Regenerate.
12943 * optabs.def (vec_perm_const): Delete.
12944 * doc/md.texi (vec_perm_const): Likewise.
12945 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12946 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12947 expand_vec_perm for constant permutation vectors. Assert that
12948 the mode of variable permutation vectors is the integer equivalent
12949 of the mode that is being permuted.
12950 * optabs-query.h (selector_fits_mode_p): Declare.
12951 * optabs-query.c: Include vec-perm-indices.h.
12952 (selector_fits_mode_p): New function.
12953 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12954 is defined, instead of checking whether the vec_perm_const_optab
12955 exists. Use targetm.vectorize.vec_perm_const instead of
12956 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12957 fit in the vector mode before using a variable permute.
12958 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12959 vec_perm_indices instead of an rtx.
12960 (expand_vec_perm): Replace with...
12961 (expand_vec_perm_const): ...this new function. Take the selector
12962 as a vec_perm_indices rather than an rtx. Also take the mode of
12963 the selector. Update call to shift_amt_for_vec_perm_mask.
12964 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12965 Use vec_perm_indices::new_expanded_vector to expand the original
12966 selector into bytes. Check whether the indices fit in the vector
12967 mode before using a variable permute.
12968 (expand_vec_perm_var): Make global.
12969 (expand_mult_highpart): Use expand_vec_perm_const.
12970 * fold-const.c: Includes vec-perm-indices.h.
12971 * tree-ssa-forwprop.c: Likewise.
12972 * tree-vect-data-refs.c: Likewise.
12973 * tree-vect-generic.c: Likewise.
12974 * tree-vect-loop.c: Likewise.
12975 * tree-vect-slp.c: Likewise.
12976 * tree-vect-stmts.c: Likewise.
12977 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12979 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12980 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12981 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12982 (aarch64_vectorize_vec_perm_const): ...this new function.
12983 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12984 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12985 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12986 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12987 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12988 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12989 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12991 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12992 check for NEON modes.
12993 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12994 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12995 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12996 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12998 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12999 the old VEC_PERM_CONST conditions.
13000 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13001 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13002 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13003 (ia64_vectorize_vec_perm_const_ok): Merge into...
13004 (ia64_vectorize_vec_perm_const): ...this new function.
13005 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13006 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13007 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13008 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13009 * config/mips/mips.c (mips_expand_vec_perm_const)
13010 (mips_vectorize_vec_perm_const_ok): Merge into...
13011 (mips_vectorize_vec_perm_const): ...this new function.
13012 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13013 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13014 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13015 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13016 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13017 (rs6000_expand_vec_perm_const): Delete.
13018 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13020 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13021 (altivec_expand_vec_perm_const_le): Take each operand individually.
13022 Operate on constant selectors rather than rtxes.
13023 (altivec_expand_vec_perm_const): Likewise. Update call to
13024 altivec_expand_vec_perm_const_le.
13025 (rs6000_expand_vec_perm_const): Delete.
13026 (rs6000_vectorize_vec_perm_const_ok): Delete.
13027 (rs6000_vectorize_vec_perm_const): New function.
13028 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13029 an element count and rtx array.
13030 (rs6000_expand_extract_even): Update call accordingly.
13031 (rs6000_expand_interleave): Likewise.
13032 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13033 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13034 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13035 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13036 (rs6000_expand_vec_perm_const): Delete.
13037 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13038 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13039 (altivec_expand_vec_perm_const_le): Take each operand individually.
13040 Operate on constant selectors rather than rtxes.
13041 (altivec_expand_vec_perm_const): Likewise. Update call to
13042 altivec_expand_vec_perm_const_le.
13043 (rs6000_expand_vec_perm_const): Delete.
13044 (rs6000_vectorize_vec_perm_const_ok): Delete.
13045 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13046 reference to the SPE evmerge intructions.
13047 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13048 an element count and rtx array.
13049 (rs6000_expand_extract_even): Update call accordingly.
13050 (rs6000_expand_interleave): Likewise.
13051 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13052 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13054 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13056 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13058 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13059 vector mode and that that mode matches the mode of the data
13061 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13062 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13063 directly using expand_vec_perm_1 when forcing selectors into
13065 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13067 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13069 * optabs-query.h (can_vec_perm_p): Delete.
13070 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13071 * optabs-query.c (can_vec_perm_p): Split into...
13072 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13073 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13074 particular selector is valid.
13075 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13076 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13077 (vect_grouped_load_supported): Likewise.
13078 (vect_shift_permute_load_chain): Likewise.
13079 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13080 (vect_transform_slp_perm_load): Likewise.
13081 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13082 (vectorizable_bswap): Likewise.
13083 (vect_gen_perm_mask_checked): Likewise.
13084 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13085 implementations of variable permutation vectors into account
13086 when deciding which selector to use.
13087 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13088 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13089 with a false third argument.
13090 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13091 to test whether the constant selector is valid and can_vec_perm_var_p
13092 to test whether a variable selector is valid.
13094 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13096 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13097 * optabs-query.c (can_vec_perm_p): Likewise.
13098 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13099 instead of vec_perm_indices.
13100 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13101 (vect_gen_perm_mask_checked): Likewise,
13102 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13103 (vect_gen_perm_mask_checked): Likewise,
13105 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13107 * optabs-query.h (qimode_for_vec_perm): Declare.
13108 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13109 (qimode_for_vec_perm): ...this new function.
13110 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13112 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13114 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13115 does not have a conditional at the top.
13117 2018-01-02 Richard Biener <rguenther@suse.de>
13119 * ipa-inline.c (big_speedup_p): Fix expression.
13121 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13124 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13127 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13131 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13132 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13133 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13134 cond_taken_branch_cost 3->4.
13136 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13138 PR tree-optimization/83581
13139 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13140 TODO_cleanup_cfg if any changes have been made.
13142 PR middle-end/83608
13143 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13144 convert_modes if target mode has the right side, but different mode
13147 PR middle-end/83609
13148 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13149 last argument when extracting from CONCAT. If either from_real or
13150 from_imag is NULL, use expansion through memory. If result is not
13151 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13152 the parts directly to inner mode, if even that fails, use expansion
13155 PR middle-end/83623
13156 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13157 check for bswap in mode rather than HImode and use that in expand_unop
13160 Copyright (C) 2018 Free Software Foundation, Inc.
13162 Copying and distribution of this file, with or without modification,
13163 are permitted in any medium without royalty provided the copyright
13164 notice and this notice are preserved.