1 2018-05-17 Steve Ellcey <sellcey@cavium.com>
3 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
4 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
6 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
7 Change logics_shift_reg to logics_shift_imm.
8 (thunderx2t99_fp_loadpair_basic): Delete.
9 (thunderx2t99_fp_storepair_basic): Delete.
10 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
11 (thunderx2t99_asimd_polynomial): Delete.
12 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
13 and neon_fp_mul_d_scalar_q.
14 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
15 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
16 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
17 (thunderx2t99_asimd_lut): Add missing tbl types.
18 (thunderx2t99_asimd_ext): Delete.
19 (thunderx2t99_asimd_load1_1_mult): Delete.
20 (thunderx2t99_asimd_load1_2_mult): Delete.
21 (thunderx2t99_asimd_load1_ldp): New.
22 (thunderx2t99_asimd_load1): New.
23 (thunderx2t99_asimd_load2): Add missing *load2* types.
24 (thunderx2t99_asimd_load3): New.
25 (thunderx2t99_asimd_load4): New.
26 (thunderx2t99_asimd_store1_1_mult): Delete.
27 (thunderx2t99_asimd_store1_2_mult): Delete.
28 (thunderx2t99_asimd_store2_mult): Delete.
29 (thunderx2t99_asimd_store2_onelane): Delete.
30 (thunderx2t99_asimd_store_stp): New.
31 (thunderx2t99_asimd_store1): New.
32 (thunderx2t99_asimd_store2): New.
33 (thunderx2t99_asimd_store3): New.
34 (thunderx2t99_asimd_store4): New.
36 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
38 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
39 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
41 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
42 Segher Boessenkool <segher@kernel.crashing.org>
45 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
48 2018-05-17 Richard Biener <rguenther@suse.de>
50 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
51 for pruning loop and prune defs feeding only already visited PHIs.
53 2018-05-17 Richard Biener <rguenther@suse.de>
55 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
57 2018-05-17 Bin Cheng <bin.cheng@arm.com>
58 Richard Biener <rguenther@suse.de>
60 PR tree-optimization/85793
61 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
64 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
66 * internal-fn.h (lookup_internal_fn): Declare
67 * internal-fn.c (lookup_internal_fn): New function.
68 * gimple.c (gimple_build_call_from_tree): Handle calls to
70 * gimple-pretty-print.c (dump_gimple_call): Print "." before
71 internal function names.
72 * tree-pretty-print.c (dump_generic_node): Likewise.
73 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
75 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
77 * gimple-fold.h (gimple_build): Make the function forms take
78 combined_fn rather than built_in_function.
79 (gimple_simplify): Likewise.
80 * gimple-match-head.c (gimple_simplify): Likewise.
81 * gimple-fold.c (gimple_build): Likewise.
82 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
83 rather than gimple_build_call_internal.
84 (get_initial_defs_for_reduction): Likewise.
85 (vect_create_epilog_for_reduction): Likewise.
86 (vectorizable_live_operation): Likewise.
88 2018-05-17 Martin Liska <mliska@suse.cz>
90 * gimple-ssa-sprintf.c (format_directive): Do not use
91 space in between 'G_' and '('.
93 2018-05-17 Jakub Jelinek <jakub@redhat.com>
96 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
97 even if the mask is not all ones.
100 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
102 (ix86_gimple_fold_builtin): Likewise.
105 * config/i386/i386.c: Include tree-vector-builder.h.
106 (ix86_vector_shift_count): New function.
107 (ix86_fold_builtin): Fold shift builtins by scalar count.
108 (ix86_gimple_fold_builtin): Likewise.
110 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
111 _mm512_setzero): New intrinsics.
113 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
114 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
116 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
117 code generation for cases where splatting a value is not useful.
118 * simplify-rtx.c (simplify_ternary_operation): Simplify
119 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
120 mode to a vec_concat.
122 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
124 * config.gcc: Support "goldmont-plus".
125 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
127 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
128 PROCESSOR_GOLDMONT_PLUS.
129 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
130 (processor_target_table): Add "goldmont-plus".
131 (PTA_GOLDMONT_PLUS): Define.
132 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
133 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
134 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
135 (fold_builtin_cpu): Add "goldmont-plus".
136 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
137 (ix86_option_override_internal): Add "goldmont-plus".
138 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
139 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
140 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
141 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
143 2018-05-17 Richard Biener <rguenther@suse.de>
145 PR tree-optimization/85757
146 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
147 remove defs that only feed that PHI from further processing.
149 2018-05-16 Jim Wilson <jimw@sifive.com>
151 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
153 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
155 2018-05-16 Mark Wielaard <mark@klomp.org>
157 * dwarf2out.c (count_index_strings): New function.
158 (output_indirect_strings): Call count_index_strings and generate
159 header for dwarf_version >= 5.
161 2018-05-16 Mark Wielaard <mark@klomp.org>
163 * dwarf2out.c (dwarf_FORM): New function.
164 (set_indirect_string): Use dwarf_FORM.
165 (reset_indirect_string): Likewise.
166 (size_of_die): Likewise.
167 (value_format): Likewise.
168 (output_die): Likewise.
169 (add_skeleton_AT_string): Likewise.
170 (output_macinfo_op): Likewise.
171 (index_string): Likewise.
172 (output_index_string_offset): Likewise.
173 (output_index_string): Likewise.
174 (count_index_strings): Likewise.
176 2018-05-16 Carl Love <cel@us.ibm.com>
178 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
179 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
181 2018-05-16 Martin Jambor <mjambor@suse.cz>
183 * ipa-prop.c (ipa_free_all_edge_args): Remove.
184 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
186 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
188 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
189 (fnma<mode>4): Likewise.
190 (fms<mode>4): Likewise.
191 (fnms<mode>4): Likewise.
192 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
193 (aarch64_fnma<mode>4): Likewise.
194 (aarch64_fms<mode>4): Likewise.
195 (aarch64_fnms<mode>4): Likewise.
196 (aarch64_fnmadd<mode>4): Likewise.
198 2018-05-16 Jason Merrill <jason@redhat.com>
200 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
202 2018-05-16 Richard Biener <rguenther@suse.de>
204 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
205 (dump_stmt_cost): Declare.
206 (add_stmt_cost): Dump cost we add.
207 (add_stmt_costs): New function.
208 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
210 (vect_analyze_stmt): Adjust prototype.
211 (vectorizable_condition): Likewise.
212 (vectorizable_live_operation): Likewise.
213 (vectorizable_reduction): Likewise.
214 (vectorizable_induction): Likewise.
215 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
216 cost vector to pass to vectorizable_ and record afterwards.
217 (vect_model_reduction_cost): Take cost vector argument and adjust.
218 (vect_model_induction_cost): Likewise.
219 (vectorizable_reduction): Likewise.
220 (vectorizable_induction): Likewise.
221 (vectorizable_live_operation): Likewise.
222 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
223 SLP_TREE_NUMBER_OF_VEC_STMTS.
224 (vect_analyze_slp_cost_1): Remove.
225 (vect_analyze_slp_cost): Likewise.
226 (vect_slp_analyze_node_operations): Take visited args and
227 a target cost vector. Avoid processing already visited stmt sets.
228 (vect_slp_analyze_operations): Use a local cost vector to gather
229 costs and register those of non-discarded instances.
230 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
231 (vect_schedule_slp_instance): Remove copying of
232 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
234 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
235 adding cost. Record cost entry location.
236 (vect_prologue_cost_for_slp_op): Function to compute cost of
237 a constant or invariant generated for SLP vect in the prologue,
238 split out from vect_analyze_slp_cost_1.
239 (vect_model_simple_cost): Make static. Adjust for SLP costing.
240 (vect_model_promotion_demotion_cost): Likewise.
241 (vect_model_store_cost): Likewise, make static.
242 (vect_model_load_cost): Likewise.
243 (vectorizable_bswap): Add cost vector arg and adjust.
244 (vectorizable_call): Likewise.
245 (vectorizable_simd_clone_call): Likewise.
246 (vectorizable_conversion): Likewise.
247 (vectorizable_assignment): Likewise.
248 (vectorizable_shift): Likewise.
249 (vectorizable_operation): Likewise.
250 (vectorizable_store): Likewise.
251 (vectorizable_load): Likewise.
252 (vectorizable_condition): Likewise.
253 (vectorizable_comparison): Likewise.
254 (can_vectorize_live_stmts): Likewise.
255 (vect_analyze_stmt): Likewise.
256 (vect_transform_stmt): Adjust calls to vectorizable_*.
257 * tree-vectorizer.c: Include gimple-pretty-print.h.
258 (dump_stmt_cost): New function.
260 2018-05-16 Richard Biener <rguenther@suse.de>
262 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
263 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
264 * tree-ssa-dse.c: Include tree-ssa-loop.h.
265 (check_name): New callback.
266 (dse_classify_store): Track cycles via a visited bitmap of PHI
267 defs and simplify handling of in-loop and across loop dead stores
268 and properly fail for loop-variant refs. Handle byte-tracking with
269 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
272 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
274 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
275 (vect_get_mask_type_for_stmt): Likewise.
276 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
278 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
279 to determine the statement's vector type and the vector type that
280 should be used for calculating nunits. Deal with cases in which
281 the type has to be deferred.
282 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
283 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
284 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
285 (vect_determine_vf_for_stmt): New functions, split out from...
286 (vect_determine_vectorization_factor): ...here.
287 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
288 (vect_get_mask_type_for_stmt): New functions, split out from
289 vect_determine_vectorization_factor.
291 2018-05-16 Richard Biener <rguenther@suse.de>
293 * tree-cfg.c (verify_gimple_assign_ternary): Properly
294 verify the [VEC_]COND_EXPR embedded comparison.
296 2018-05-15 Martin Sebor <msebor@redhat.com>
298 PR tree-optimization/85753
299 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
300 RECORD_TYPE in addition to ARRAY_TYPE.
302 2018-05-15 Martin Sebor <msebor@redhat.com>
305 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
307 2018-05-15 Richard Biener <rguenther@suse.de>
309 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
310 add by_clobber_p one. Change algorithm to collect all defs
311 representing uses we need to walk and try reducing them to
312 a single one before failing.
313 (dse_dom_walker::dse_optimize_stmt): Adjust.
315 2018-05-13 Mark Wielaard <mark@klomp.org>
317 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
318 (size_of_loc_descr): Likewise.
319 (output_loc_operands): Likewise.
320 (output_loc_operands_raw): Likewise.
321 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
322 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
323 (hash_loc_operands): Likewise.
324 (compare_loc_operands): Likewise.
326 2018-05-14 Mark Wielaard <mark@klomp.org>
328 * dwarf2out.c (count_index_addrs): New function.
329 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
331 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
333 PR tree-optimization/83648
334 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
335 return value as malloc candidate.
337 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
340 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
341 as true in call to suggest_attribute.
343 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
345 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
348 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
350 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
352 (aarch64_crypto_aesd_fused): Likewise.
354 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
356 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
357 (movsi_aarch64): Likewise.
358 (load_pairsi): Likewise.
359 (load_pairdi): Likewise.
360 (store_pairsi): Likewise.
361 (store_pairdi): Likewise.
362 (load_pairsf): Likewise.
363 (load_pairdf): Likewise.
364 (store_pairsf): Likewise.
365 (store_pairdf): Likewise.
366 (zero_extend): Likewise.
367 (trunc): Swap alternatives.
368 (fcvt_target): Add '?' to prefer w over r.
370 2018-05-14 Jakub Jelinek <jakub@redhat.com>
373 * config/i386/i386.md: Disallow non-commutative arithmetics in
374 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
375 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
376 in the peephole2 before it.
378 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
380 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
381 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
382 (ix86_handle_option): Handle -mcldemote.
383 * config.gcc: New header.
384 * config/i386/cldemoteintrin.h: New file.
385 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
386 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
388 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
389 OPTION_MASK_ISA_CLDEMOTE.
390 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
391 (ix86_valid_target_attribute_inner_p): Ditto.
392 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
393 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
394 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
395 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
396 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
398 * config/i386/i386.opt: Add -mcldemote.
399 * config/i386/x86intrin.h: New header.
400 * doc/invoke.texi: Add -mcldemote.
402 2018-05-14 Richard Biener <rguenther@suse.de>
404 * doc/match-and-simplify.texi: Adjust :s documentation.
406 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
408 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
409 intended memcpy size.
410 (REORDER_45): Likewise.
412 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
415 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
416 * vec.c (qsort_chk): Use gcc_qsort.
417 * Makefile.in (OBJS-libcommon): Add sort.o.
418 (build/sort.o): New target. Use it...
419 (BUILD_RTL): ... here, and...
420 (build/gencfn-macros): ... here, and...
421 (build/genmatch): ... here.
423 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
424 Chung-Ju Wu <jasonwucj@gmail.com>
426 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
427 * config/nds32/nds32-graywolf.md: New file.
428 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
429 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
431 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
432 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
433 * config/nds32/nds32.md (pipeline_model): Add graywolf.
434 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
435 * config/nds32/pipelines.md: Include n15 settings.
437 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
438 Chung-Ju Wu <jasonwucj@gmail.com>
440 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
441 * config/nds32/nds32-n13.md: New file.
442 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
443 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
445 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
446 * config/nds32/nds32.md (pipeline_model): Add n13.
447 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
448 * config/nds32/pipelines.md: Include n13 settings.
450 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
451 Chung-Ju Wu <jasonwucj@gmail.com>
453 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
454 * config/nds32/nds32-n10.md: New file.
455 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
456 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
458 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
459 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
460 * config/nds32/nds32.md (pipeline_model): Add n10.
461 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
462 * config/nds32/pipelines.md: Include n10 settings.
464 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
465 Kito Cheng <kito.cheng@gmail.com>
466 Chung-Ju Wu <jasonwucj@gmail.com>
468 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
469 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
470 Add enum values for DSP extension instructions.
471 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
473 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
474 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
476 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
477 * config/nds32/nds32-dspext.md: New file for DSP implementation.
478 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
479 * config/nds32/nds32-intrinsic.md: Likewise.
480 * config/nds32/nds32_intrinsic.h: Likewise.
481 * config/nds32/nds32-md-auxiliary.c: Likewise.
482 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
483 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
484 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
485 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
486 * config/nds32/nds32-protos.h: New declarations for DSP extension.
487 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
488 TYPE_DMAC in switch statement.
489 * config/nds32/nds32.c: New checking and implementation for DSP
490 extension instructions.
491 * config/nds32/nds32.h: Likewise.
492 * config/nds32/nds32.md: Likewise.
493 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
494 * config/nds32/predicates.md: Implement new predicates for DSP
497 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
499 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
500 Reformat alternatives and attributes so it is easier to identify
501 which constraints/attributes go with which instruction.
502 (mov<mode>_hardfloat32, FMOVE64): Likewise.
503 (mov<mode>_softfloat32, FMOVE64): Likewise.
504 (mov<mode>_hardfloat64, FMOVE64): Likewise.
505 (mov<mode>_softfloat64, FMOVE64): Likewise.
507 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
509 * doc/extend.texi (PowerPC Built-in Functions): Rename this
511 (Basic PowerPC Built-in Functions): The new name of the
512 subsection previously known as "PowerPC Built-in Functions".
513 (Basic PowerPC Built-in Functions Available on all Configurations):
515 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
516 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
517 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
518 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
520 2018-05-11 Martin Jambor <mjambor@suse.cz>
523 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
526 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
529 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
531 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
533 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
534 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
535 (ix86_handle_option): Handle -mwaitpkg.
536 * config.gcc: New header.
537 * config/i386/cpuid.h (bit_WAITPKG): New bit.
538 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
539 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
541 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
542 OPTION_MASK_ISA_WAITPKG.
543 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
544 (ix86_option_override_internal): Add PTA_WAITPKG.
545 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
546 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
548 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
549 __builtin_ia32_umwait and __builtin_ia32_tpause.
550 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
551 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
552 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
553 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
554 UNSPECV_TPAUSE): New.
555 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
556 * config/i386/i386.opt: Add -mwaitpkg.
557 * config/i386/waitpkgintrin.h: New file.
558 * config/i386/x86intrin.h: New header.
559 * doc/invoke.texi: Add -mwaitpkg.
561 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
564 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
566 (cortex-m0): Use armv6s-m isa.
567 (cortex-m0plus): Likewise.
568 (cortex-m1): Likewise.
569 (cortex-m0.small-multiply): Likewise.
570 (cortex-m0plus.small-multiply): Likewise.
571 (cortex-m1.small-multiply): Likewise.
573 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
574 Jakub Jelinek <jakub@redhat.com>
576 PR tree-optimization/85692
577 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
578 source permute as well.
580 2018-05-11 Martin Liska <mliska@suse.cz>
583 * doc/extend.texi: Document LLVM style format for no_sanitize
586 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
588 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
589 mode_supports_vsx_dform_quad to mode_supports_dq_form.
590 (mode_supports_vsx_dform_quad): Likewise.
591 (mode_supports_vmx_dform): Move these functions to be next to the
592 other mode_supports functions.
593 (mode_supports_dq_form): Likewise.
594 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
595 mode_supports_dq_form.
596 (reg_offset_addressing_ok_p): Likewise.
597 (offsettable_ok_by_alignment): Likewise.
598 (rs6000_legitimate_offset_address_p): Likewise.
599 (legitimate_lo_sum_address_p): Likewise.
600 (rs6000_legitimize_address): Likewise.
601 (rs6000_legitimize_reload_address): Likewise.
602 (rs6000_secondary_reload_inner): Likewise.
603 (rs6000_preferred_reload_class): Likewise.
604 (rs6000_output_move_128bit): Likewise.
606 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
608 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
609 Generate SImode target register for null target.
610 <case IX86_BUILTIN_XGETBV>: Ditto.
611 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
612 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
614 2018-05-10 Carl Love <cel@us.ibm.com>
616 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
617 dcbtt and dcbtstt if operands[2] is 0.
619 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
622 * config/i386/sse.md (usadv64qi): New expander.
624 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
626 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
627 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
628 -maltivec=be support.
629 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
630 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
631 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
632 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
633 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
634 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
635 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
636 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
637 altivec_vsumsws): Adjust.
638 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
639 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
640 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
642 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
643 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
644 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
645 (altivec_lve<VI_char>x): Delete expand.
646 (*altivec_lve<VI_char>x_internal): Rename to...
647 (altivec_lve<VI_char>x): ... this.
648 (altivec_lvxl_<mode>): Delete expand.
649 (*altivec_lvxl_<mode>_internal): Rename to ...
650 (altivec_lvxl_<mode>): ... this.
651 (altivec_stvxl_<mode>): Delete expand.
652 (*altivec_stvxl_<mode>_internal): Rename to ...
653 (altivec_stvxl_<mode>): ... this.
654 (altivec_stve<VI_char>x): Delete expand.
655 (*altivec_stve<VI_char>x_internal): Rename to ...
656 (altivec_stve<VI_char>x): ... this.
657 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
658 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
659 reduc_plus_scal_<mode>): Adjust.
660 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
662 (rs6000_cpu_cpp_builtins): Adjust.
663 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
664 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
665 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
666 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
667 -maltivec=be support.
668 (rs6000_split_vec_extract_var): Adjust.
669 (rs6000_split_v4si_init): Adjust.
670 (swap_selector_for_mode): Delete.
671 (altivec_expand_lvx_be, altivec_expand_stvx_be,
672 altivec_expand_stvex_be): Delete.
673 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
674 -maltivec=be support.
675 (rs6000_gimple_fold_builtin): Ditto.
676 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
678 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
679 (TARGET_DIRECT_MOVE_64BIT): Adjust.
680 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
681 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
682 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
683 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
684 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
685 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
686 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
687 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
688 anonymous split): Adjust.
689 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
690 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
692 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
694 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
695 when --with-gxx-include-dir is also specified.
696 * configure: Regenerate.
698 2018-05-09 Jim Wilson <jimw@sifive.com>
701 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
702 * config/riscv/t-withmultilib: New.
703 * config/riscv/withmultilib.h: New.
704 * doc/install.texi: Document RISC-V --with-multilib-list support.
706 2018-05-09 Richard Biener <rguenther@suse.de>
708 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
710 (vect_bb_vectorization_profitable_p): Adjust. Compute
711 actual scalar cost using the cost vector and the add_stmt_cost
714 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
716 PR rtl-optimization/85645
717 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
718 in the REG_CFA_REGISTER note for LR, don't leave it empty.
720 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
722 PR rtl-optimization/85645
723 * shrink-wrap.c (spread_components): Return a boolean saying if
724 anything was changed.
725 (try_shrink_wrapping_separate): Iterate spread_components until
726 nothing changes anymore.
728 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
730 PR rtl-optimization/85645
731 * regrename.c (build_def_use): Also kill the chains that include the
732 destination of a REG_CFA_REGISTER note.
734 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
736 PR rtl-optimization/85645
737 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
738 insn that has a REG_CFA_REGISTER note.
740 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
742 * cfgexpand.c (expand_clobber): New function.
743 (expand_gimple_stmt_1): Use it.
744 * tree-vect-stmts.c (vect_clobber_variable): New function,
746 (vectorizable_simd_clone_call): ...here.
747 (vectorizable_store): Emit a clobber either side of an
748 IFN_STORE_LANES sequence.
749 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
751 2018-05-09 Tom de Vries <tom@codesourcery.com>
754 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
755 (define_insn "trap_if_false"): Add exit after trap.
757 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
759 PR rtl-optimization/85638
760 * bb-reorder.c: Include common/common-target.h.
761 (create_forwarder_block): New function extracted from...
762 (fix_up_crossing_landing_pad): ...here. Rename into...
763 (dw2_fix_up_crossing_landing_pad): ...this.
764 (sjlj_fix_up_crossing_landing_pad): New function.
765 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
766 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
767 from both partitions and exit the loop after one iteration.
769 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
772 * doc/extend.texi (PowerPC Built-in Functions): Rename this
774 (Basic PowerPC Built-in Functions): The new name of the
775 subsection previously known as "PowerPC Built-in Functions".
776 (Basic PowerPC Built-in Functions Available on all Configurations):
778 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
780 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
782 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
784 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
787 2018-05-08 Jim Wilson <jimw@sifive.com>
789 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
790 (LD_EMUL_SUFFIX): New.
793 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
795 * doc/extend.texi (PowerPC Built-in Functions): Rename this
797 (Basic PowerPC Built-in Functions): The new name of the
798 subsection previously known as "PowerPC Built-in Functions".
799 (Basic PowerPC Built-in Functions Available on all Configurations):
801 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
803 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
805 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
807 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
810 2018-05-08 Jakub Jelinek <jakub@redhat.com>
813 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
814 after cmpelim optimization.
816 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
818 * config.gcc: Support "goldmont".
819 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
820 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
822 * config/i386/i386.c (m_GOLDMONT): Define.
823 (processor_target_table): Add "goldmont".
824 (PTA_GOLDMONT): Define.
825 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
826 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
827 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
828 (fold_builtin_cpu): Add "goldmont".
829 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
830 (ix86_option_override_internal): Add "goldmont".
831 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
832 (processor_type): Add PROCESSOR_GOLDMONT.
833 * config/i386/i386.md: Add CPU "glm".
834 * config/i386/glm.md: New file.
835 * config/i386/x86-tune.def: Add m_GOLDMONT.
836 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
838 2018-05-08 Jakub Jelinek <jakub@redhat.com>
841 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
843 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
844 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
845 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
848 * config/i386/i386.c (ix86_fold_builtin): Handle
849 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
852 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
853 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
855 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
858 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
859 (check_arch): Likewise.
860 (check_fpu): Return the result rather than printing it.
861 (end arch): Fix operator precedence.
863 (END): Print the result from check_fpu.
865 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
866 Alan Hayward <alan.hayward@arm.com>
867 David Sherwood <david.sherwood@arm.com>
869 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
870 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
871 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
872 (*fcmuo<mode>_and): New patterns.
874 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
876 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
877 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
878 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
879 (cmp_op, sve_imm_con): New code attributes.
880 (SVE_COND_INT_CMP, imm_con): Delete.
881 (cmp_op): Remove above unspecs from int attribute.
882 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
884 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
885 comparison-specific unspecs.
886 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
887 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
888 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
889 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
890 (*vec_fcm<cmp_op><mode>): Rename to...
891 (*fcm<cmp_op><mode>): ...this and adjust likewise.
892 (*vec_fcmuo<mode>): Rename to...
893 (*fcmuo<mode>): ...this and adjust likewise.
894 (*pred_fcm<cmp_op><mode>): New pattern.
895 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
896 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
898 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
900 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
901 (aarch64_emit_sve_predicated_cond): New function.
902 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
903 (aarch64_emit_unspec_cond_or): Replace with...
904 (aarch64_emit_sve_or_conds): ...this new function. Use
905 aarch64_emit_sve_ptrue_op for the individual comparisons and
906 aarch64_emit_binop to OR them together.
907 (aarch64_emit_inverted_unspec_cond): Replace with...
908 (aarch64_emit_sve_inverted_cond): ...this new function. Use
909 aarch64_emit_sve_ptrue_op for the comparison and
910 aarch64_emit_unop to invert the result.
911 (aarch64_expand_sve_vec_cmp_float): Update after the above
912 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
914 2018-05-07 Nathan Sidwell <nathan@acm.org>
916 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
917 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
918 (Backwards Compatibility): Likewise.
920 2018-05-07 Luis Machado <luis.machado@linaro.org>
924 2018-05-07 Luis Machado <luis.machado@linaro.org>
926 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
927 <prefetch_dynamic_strides>: New const bool field.
928 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
929 prefetch_dynamic_strides.
930 (exynosm1_prefetch_tune): Likewise.
931 (thunderxt88_prefetch_tune): Likewise.
932 (thunderx_prefetch_tune): Likewise.
933 (thunderx2t99_prefetch_tune): Likewise.
934 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
935 (aarch64_override_options_internal): Update to set
936 PARAM_PREFETCH_DYNAMIC_STRIDES.
937 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
938 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
939 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
940 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
941 prefetch-dynamic-strides setting.
943 2018-05-07 Luis Machado <luis.machado@linaro.org>
945 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
946 <minimum_stride>: New const int field.
947 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
948 minimum_stride field.
949 (exynosm1_prefetch_tune): Likewise.
950 (thunderxt88_prefetch_tune): Likewise.
951 (thunderx_prefetch_tune): Likewise.
952 (thunderx2t99_prefetch_tune): Likewise.
953 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
954 (aarch64_override_options_internal): Update to set
955 PARAM_PREFETCH_MINIMUM_STRIDE.
956 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
957 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
958 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
959 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
960 stride is constant and is below the minimum stride threshold.
962 2018-05-07 Luis Machado <luis.machado@linaro.org>
964 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
967 2018-05-07 Luis Machado <luis.machado@linaro.org>
969 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
970 <prefetch_dynamic_strides>: New const bool field.
971 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
972 prefetch_dynamic_strides.
973 (exynosm1_prefetch_tune): Likewise.
974 (thunderxt88_prefetch_tune): Likewise.
975 (thunderx_prefetch_tune): Likewise.
976 (thunderx2t99_prefetch_tune): Likewise.
977 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
978 (aarch64_override_options_internal): Update to set
979 PARAM_PREFETCH_DYNAMIC_STRIDES.
980 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
981 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
982 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
983 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
984 prefetch-dynamic-strides setting.
986 2018-05-07 Luis Machado <luis.machado@linaro.org>
988 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
989 <minimum_stride>: New const int field.
990 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
991 minimum_stride field.
992 (exynosm1_prefetch_tune): Likewise.
993 (thunderxt88_prefetch_tune): Likewise.
994 (thunderx_prefetch_tune): Likewise.
995 (thunderx2t99_prefetch_tune): Likewise.
996 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
997 (aarch64_override_options_internal): Update to set
998 PARAM_PREFETCH_MINIMUM_STRIDE.
999 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1000 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1001 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1002 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1003 stride is constant and is below the minimum stride threshold.
1005 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1008 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1009 the type is addressable. Don't force op into register if it has
1012 2018-05-05 Roland McGrath <mcgrathr@google.com>
1015 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1016 any section for which we don't know a specific type it should have,
1017 regardless of name. Previously this was done only for the exact
1018 names ".init_array", ".fini_array", and ".preinit_array".
1019 (default_elf_asm_named_section): Add comment about
1020 relationship with default_section_type_flags and SECTION_NOTYPE.
1021 (get_section): Don't consider it a type conflict if one side has
1022 SECTION_NOTYPE and the other doesn't, as long as neither has the
1023 SECTION_BSS et al used in the default_section_type_flags logic.
1025 2018-05-05 Tom de Vries <tom@codesourcery.com>
1028 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1029 (workaround_barsyncs): New function.
1030 (nvptx_reorg): Use workaround_barsyncs.
1031 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1032 (define_expand "nvptx_membar_cta"): New define_expand.
1033 (define_insn "*nvptx_membar_cta"): New insn.
1035 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1037 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1038 To improve optimization opportunities.
1039 * builtin-types.def: The new needed builtin types for the above.
1041 2018-05-04 Richard Biener <rguenther@suse.de>
1043 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1044 * gimple-ssa-store-merging.c
1045 (imm_store_chain_info::output_merged_store): Remove redundant create,
1046 release split_store vector contents on failure.
1047 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1048 scalar stmt vector on cache hit.
1050 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1052 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1054 * config.gcc (powerpc-xilinx-eabi*): Remove.
1055 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1057 (fusion_addis_mem_combo_load): Ditto.
1058 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1060 (rs6000_cpu_cpp_builtins): Ditto.
1061 * config/rs6000/rs6000-linux.c
1062 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1063 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1064 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1066 (rs6000_setup_reg_addr_masks): Ditto.
1067 (rs6000_init_hard_regno_mode_ok): Ditto.
1068 (rs6000_option_override_internal): Ditto.
1069 (legitimate_lo_sum_address_p): Ditto.
1070 (rs6000_legitimize_address): Ditto.
1071 (rs6000_legitimize_reload_address): Ditto.
1072 (rs6000_legitimate_address_p): Ditto.
1073 (abi_v4_pass_in_fpr): Ditto.
1074 (setup_incoming_varargs): Ditto.
1075 (rs6000_gimplify_va_arg): Ditto.
1076 (rs6000_split_multireg_move): Ditto.
1077 (rs6000_savres_strategy): Ditto.
1078 (rs6000_emit_prologue_components): Ditto.
1079 (rs6000_emit_epilogue_components): Ditto.
1080 (rs6000_emit_prologue): Ditto.
1081 (rs6000_emit_epilogue): Ditto.
1082 (rs6000_elf_file_end): Ditto.
1083 (rs6000_function_value): Ditto.
1084 (rs6000_libcall_value): Ditto.
1085 * config/rs6000/rs6000.h: Ditto.
1086 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1087 (TARGET_MINMAX): ... this. New.
1088 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1089 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1090 (*movsi_internal1_single): Delete.
1091 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1092 mfpu=, mxilinx-fpu): Delete.
1093 * config/rs6000/singlefp.h: Delete.
1094 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1095 * config/rs6000/t-rs6000: Ditto.
1096 * config/rs6000/t-xilinx: Delete.
1097 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1098 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1099 (VStype_simple): Delete.
1100 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1101 * config/rs6000/xfpu.h: Delete.
1102 * config/rs6000/xfpu.md: Delete.
1103 * config/rs6000/xilinx.h: Delete.
1104 * config/rs6000/xilinx.opt: Delete.
1105 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1106 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1108 2018-05-04 Tom de Vries <tom@codesourcery.com>
1111 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1114 2018-05-04 Richard Biener <rguenther@suse.de>
1117 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1118 (expand_complex_div_wide): Likewise.
1119 (expand_complex_operations_1): Likewise.
1120 (expand_complex_libcall): Preserve EH info of the original stmt.
1121 (tree_lower_complex): Handle removed blocks.
1122 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1123 on complex multiplication and division libcall builtins.
1125 2018-05-04 Richard Biener <rguenther@suse.de>
1128 * fold-const.c (negate_expr_p): Restrict negation of operand
1129 zero of a division to when we know that can happen without
1131 (fold_negate_expr_1): Likewise.
1133 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1136 * real.h (real_nextafter): Declare.
1137 * real.c (real_nextafter): New function.
1138 * fold-const-call.c (fold_const_nextafter): New function.
1139 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1140 CASE_CFN_NEXTTOWARD.
1141 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1142 even when arg1_mode is different from arg0_mode.
1144 2018-05-03 Nathan Sidwell <nathan@acm.org>
1146 * doc/extend.texi (Deprecated Features): Remove
1148 (Backwards Compatibility): Likewise.
1149 * doc/invoke.texi (C++ Language Options): Likewise.
1150 (C++ Dialect Options): Likewise.
1152 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1155 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1156 _mm512_mask_mullox_epi64): New intrinsics.
1158 2018-05-03 Tom de Vries <tom@codesourcery.com>
1161 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1162 dump files): Add offload-tree.
1164 2018-05-03 Richard Biener <rguenther@suse.de>
1166 PR tree-optimization/85615
1167 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1168 to loops not nested in BBs loop father to avoid creating multi-entry
1171 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1173 PR tree-optimization/70291
1174 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1175 arguments. Change return type to tree. Emit libcall as a new
1176 statement rather than replacing existing one when inplace_p is true.
1177 (expand_complex_multiplication_components): New function.
1178 (expand_complex_multiplication): Expand floating-point complex
1179 multiplication using the above.
1180 (expand_complex_division): Rename inner_type parameter to type.
1181 Update expand_complex_libcall call-site.
1182 (expand_complex_operations_1): Update expand_complex_multiplication
1183 and expand_complex_division call-sites.
1185 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1188 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1189 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1190 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1191 the highest significant bit of the shift count mask is clear. In
1192 check whether and[sq]i3 is needed verify that all significant bits
1193 of the shift count other than the highest are set.
1195 2018-05-02 Tom de Vries <tom@codesourcery.com>
1198 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1199 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1200 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1201 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1202 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1203 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1204 __builtin_goacc_parlevel_size.
1206 2018-05-02 Richard Biener <rguenther@suse.de>
1208 PR tree-optimization/85597
1209 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1210 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1213 2018-05-02 Tom de Vries <tom@codesourcery.com>
1216 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1217 dump files): Add ltrans-tree.
1219 2018-05-02 Tom de Vries <tom@codesourcery.com>
1222 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1223 dump files): Add wpa-ipa.
1225 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1227 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1228 powerpc*-*-linux*paired* target.
1229 * config/rs6000/750cl.h: Delete.
1230 * config/rs6000/paired.h: Delete.
1231 * config/rs6000/paired.md: Delete.
1232 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1234 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1235 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1236 comment. Remove paired float support.
1237 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1238 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1240 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1241 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1243 * config/rs6000/rs6000.c: Remove paired float support.
1244 (paired_expand_vector_init, paired_expand_vector_move,
1245 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1246 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1247 paired_expand_builtin, paired_expand_predicate_builtin,
1248 paired_init_builtins): Delete.
1249 * config/rs6000/rs6000.h: Remove paired float support.
1250 * config/rs6000/rs6000.md: Remove paired float support.
1251 (move_from_CR_ov_bit): Delete.
1252 * config/rs6000/rs6000.opt (mpaired): Delete.
1253 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1254 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1256 2018-05-02 Richard Biener <rguenther@suse.de>
1259 * gimplify.c (gimplify_save_expr): When in SSA form allow
1260 SAVE_EXPRs to compute to SSA vars.
1262 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1265 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1266 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1267 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1268 clobber operands[2], instead use a new pseudo. Formatting fixes.
1270 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1272 PR tree-optimization/85586
1273 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1274 exit early for statements in the same group if the accesses are
1277 2018-05-02 Tom de Vries <tom@codesourcery.com>
1280 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1283 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1285 PR tree-optimization/85143
1286 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1288 2018-05-01 Tom de Vries <tom@codesourcery.com>
1291 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1292 not found" error message.
1294 2018-05-01 Tom de Vries <tom@codesourcery.com>
1297 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1298 * vec.c (test_ordered_remove_if): New function.
1299 (vec_c_tests): Call test_ordered_remove_if.
1300 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1301 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1302 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1303 VEC_ORDERED_REMOVE_IF.
1305 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1307 PR tree-optimization/82665
1308 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1309 pointer subtraction where arguments come from a memchr call.
1311 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1313 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1314 --push-state --as-needed and --pop-state instead of --as-needed and
1315 --no-as-needed if ld supports it.
1316 * configure: Regenerated.
1319 * doc/install.texi2html: Replace _002d with - and _002a with * in
1320 generated html files using sed.
1322 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1325 * gcc-rich-location.c (blank_line_before_p): New function.
1326 (use_new_line): New function.
1327 (gcc_rich_location::add_fixit_insert_formatted): New function.
1328 * gcc-rich-location.h
1329 (gcc_rich_location::add_fixit_insert_formatted): New function.
1331 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1333 * selftest.c (assert_streq): Rename "expected" and "actual" to
1334 "val1" and "val2". Extend NULL-handling to cover both inputs
1335 symmetrically, while still requiring both to be non-NULL for a pass.
1336 * selftest.h (assert_streq): Rename "expected" and "actual" to
1338 (ASSERT_EQ): Likewise.
1339 (ASSERT_EQ_AT): Likewise.
1340 (ASSERT_KNOWN_EQ): Likewise.
1341 (ASSERT_KNOWN_EQ_AT): Likewise.
1342 (ASSERT_NE): Likewise.
1343 (ASSERT_MAYBE_NE): Likewise.
1344 (ASSERT_MAYBE_NE_AT): Likewise.
1345 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1346 the assertion to pass.
1347 (ASSERT_STREQ_AT): Likewise.
1349 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1351 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1352 interaction with -pie.
1354 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1356 * selftest.h: Fix alphabetization of per-source-file selftest
1359 2018-04-30 Jason Merrill <jason@redhat.com>
1361 PR c++/61982 - dead stores to destroyed objects.
1362 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1365 2018-04-30 Jason Merrill <jason@redhat.com>
1367 * tree.c (build_clobber): New.
1368 * tree.h: Declare it.
1369 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1371 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1373 * diagnostic-show-locus.c (layout::layout): Update for
1374 location_get_source_line returning a char_span.
1375 (struct char_span): Move to input.h.
1376 (struct correction): Update for fields in char_span becoming
1378 (struct source_line): Update for location_get_source_line
1379 returning a char_span.
1380 (layout::print_line): Likewise.
1381 * edit-context.c (edited_file::print_content): Likewise.
1382 (edited_file::print_diff_hunk): Likewise.
1383 (edited_file::print_run_of_changed_lines): Likewise.
1384 (edited_file::get_num_lines): Likewise.
1385 (edited_line::edited_line): Likewise.
1386 * final.c (asm_show_source): Likewise.
1387 * input.c (location_get_source_line): Convert return type
1388 from const char * to char_span, losing the final "line_len"
1390 (dump_location_info): Update for the above.
1391 (get_substring_ranges_for_loc): Likewise. Use a char_span
1392 when handling the literal within the line.
1393 (test_reading_source_line): Update for location_get_source_line
1394 returning a char_span.
1395 * input.h (class char_span): Move here from
1396 diagnostic-show-locus.c, converting from a struct to a class.
1397 Make data members private.
1398 (char_span::operator bool): New.
1399 (char_span::length): New.
1400 (char_span::get_buffer): New.
1401 (char_span::operator[]): New.
1402 (char_span::subspan): Make const.
1403 (char_span::xstrdup): New.
1404 (location_get_source_line): Convert return type from const char *
1405 to char_span, losing the final "line_size" param.
1407 2018-04-30 Jan Hubicka <jh@suse.cz>
1409 * lto-wrapper.c (ltrans_priorities): New static var.
1410 (cmp_priority): New.
1411 (run_gcc): Read priorities and if doing parallel build order
1412 the Makefile by them.
1414 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1416 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1418 2018-04-30 Richard Biener <rguenther@suse.de>
1420 * tree-cfg.c (verify_address): Remove base argument, add
1421 flag whether to check TREE_ADDRESSABLE and do that.
1422 (verify_expr): Remove.
1423 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1424 (verify_gimple_assign_single): Likewise.
1425 (verify_gimple_switch): Likewise.
1426 (verify_expr_location_1): Dereference tp once. Add (disabled)
1427 piece from verify_expr.
1428 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1430 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1432 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1434 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1436 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1437 (small_data_pattern): Likewise.
1438 (arc_rewrite_small_data): Likewise.
1439 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1440 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1441 (get_symbol_alignment): New function.
1442 (legitimate_small_data_address_p): Likewise.
1443 (legitimate_scaled_address): Update, call
1444 legitimate_small_data_address_p.
1445 (output_sdata): New static variable.
1446 (arc_print_operand): Update how we handle small data operands.
1447 (arc_print_operand_address): Likewise.
1448 (arc_legitimate_address_p): Update, use
1449 legitimate_small_data_address_p.
1450 (arc_rewrite_small_data_p): Remove.
1451 (arc_rewrite_small_data_1): Likewise.
1452 (arc_rewrite_small_data): Likewise.
1453 (small_data_pattern): Likewise.
1454 (compact_sda_memory_operand): Update to use
1455 legitimate_small_data_address_p and get_symbol_alignment.
1456 (prepare_move_operands): Don't rewite sdata pattern.
1457 (prepare_extend_operands): Remove.
1458 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1460 (zero_extendqisi2): Likewise.
1461 (zero_extendhisi2): Likewise.
1462 (extendqihi2): Likewise.
1463 (extendqisi2): Likewise.
1464 (extendhisi2): Likewise.
1468 * config/arc/constraints.md (Usd): Change it to memory constraint.
1470 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1472 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1473 as source of std instructions.
1474 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1475 allow 6-bit constants as source for store instructions.
1476 (movdi_insn): Update instruction pattern to allow 6-bit constants
1477 as source for store instructions.
1479 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1481 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1483 2018-04-30 Nathan Sidwell <nathan@acm.org>
1484 Sandra Loosemore <sandra@codesourcery.com>
1486 * dumpfile.c (dump_open): Allow '-' for stdout.
1487 * doc/invoke.texi (Developer Options): Document dump filename
1488 determination early. Document stdin/stdout selection.
1490 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1492 Microblaze Target: PIC data text relative
1494 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1495 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1497 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1498 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1499 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1500 New addressing mode for data-text relative position indepenedent code.
1501 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1502 'ADDRESS_SYMBOLIC_TXT_REL'.
1503 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1504 (microblaze_legitimate_pic_operand): Exclude function calls from
1505 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1506 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1508 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1509 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1510 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1511 for 'address + offset'.
1512 (microblaze_expand_prologue): Add new function prologue call for
1514 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1515 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1516 table in case of TARGET_PIC_DATA_TEXT_REL.
1517 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1518 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1519 Add new macros 'UNSPEC_TEXT',
1520 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1521 + exclude function calls from 'UNSPEC_PLT' in case of data text
1523 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1524 new target hook for generating address diff vector tables in case of
1526 * doc/tm.texi : Regenerate.
1527 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1528 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1529 of addr diff vector generation.
1530 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1531 target hook definition.
1532 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1533 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1534 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1535 option for data text relative.
1537 2018-04-30 Richard Biener <rguenther@suse.de>
1539 * tree-chrec.h (evolution_function_is_constant_p): Remove
1541 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1543 2018-04-30 Richard Biener <rguenther@suse.de>
1546 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1548 2018-04-30 Richard Biener <rguenther@suse.de>
1550 PR tree-optimization/28364
1551 PR tree-optimization/85275
1552 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1553 copying first exit test.
1555 2018-04-28 Mark Wielaard <mark@klomp.org>
1557 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1559 (dwarf_AT): Handle DW_AT_addr_base.
1560 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1562 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1565 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1566 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1567 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1568 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1570 2018-04-28 Richard Biener <rguenther@suse.de>
1572 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1573 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1574 to reflect use. Only add interesting stmts.
1576 2018-04-27 Martin Jambor <mjambor@suse.cz>
1579 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1580 the jump function allows for passing through aggregate values.
1582 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1584 * input.h (in_system_header_at): Convert from macro to inline
1586 (from_macro_expansion_at): Likewise.
1587 (from_macro_definition_at): Likewise.
1589 2018-04-27 Jeff Law <law@redhat.com>
1591 * config.gcc: Mark tile* targets as deprecated/obsolete.
1593 2018-04-27 Richard Biener <rguenther@suse.de>
1595 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1598 2018-04-27 Richard Biener <rguenther@suse.de>
1600 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1602 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1604 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1605 with Yd constraint. Set "preferred_for_speed" attribute from
1606 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1608 (*movdi_internal): Ditto.
1609 (movti_interunit splitters): Remove
1610 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1611 (movdi_interunit splitters): Ditto.
1612 * config/i386/constraints.md (Ye): Remove.
1613 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1615 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1618 * config/aarch64/constraints.md (Usg): Limit to 31.
1621 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1623 PR tree-optimization/85529
1624 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1625 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1626 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1627 zero extension or masking of the MSB bit.
1628 (optimize_range_tests): Add FIRST_BB argument, pass it through
1629 to optimize_range_tests_var_bound.
1630 (maybe_optimize_range_tests, reassociate_bb): Adjust
1631 optimize_range_tests callers.
1633 2018-04-26 Richard Biener <rguenther@suse.de>
1634 Jakub Jelinek <jakub@redhat.com>
1636 * cgraph.h (symbol_table): Just declare debug method here.
1637 * symtab.c (symbol_table::debug): Define.
1639 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1641 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1643 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1645 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1646 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1647 (*movdi_internal): Substitute Yi and Yj constraint with x
1648 and Ym and Yn constraint with y constraint. Update "isa"
1649 attribute and set "preferred_for_speed" attribute from
1650 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1651 (*movsi_internal): Ditto.
1652 (*movdf_internal): Ditto.
1653 (*movsf_internal): Ditto.
1654 (*zero_extendsidi2): Ditto.
1655 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1656 (sse2_loadld): Ditto.
1657 (*vec_extract<ssevecmodelower>_0): Ditto.
1658 (*vec_extractv4si_0_zext_sse4): Ditto.
1659 (vec_concatv2di): Ditto.
1660 (*vec_dup<mode>): Ditto.
1661 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1662 * config/i386/constraints.md (Yi): Remove.
1667 2018-04-26 Nathan Sidwell <nathan@acm.org>
1669 * dumpfile.c (dump_open): New.
1670 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1671 (dump_finish): Detect stdio/stderr by value not name.
1673 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1675 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1677 2018-04-26 Tom de Vries <tom@codesourcery.com>
1680 * config/nvptx/nvptx.c (verify_neutering_jumps)
1681 (verify_neutering_labels): New function
1682 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1684 2018-04-26 Tom de Vries <tom@codesourcery.com>
1687 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1688 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1691 2018-04-26 Richard Biener <rguenther@suse.de>
1692 Tom de Vries <tom@codesourcery.com>
1695 * lto-streamer-out.c (output_function): Fixup loops if required to match
1696 discovery done in the reader.
1698 2018-04-26 Richard Biener <rguenther@suse.de>
1700 PR tree-optimization/85116
1701 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1702 have a loop exit from the single latch predecessor. Remove
1703 case of header with just condition.
1704 (ch_base::copy_headers): Exclude infinite loops from any
1706 (pass_ch::execute): Record exits.
1708 2018-04-26 Richard Biener <rguenther@suse.de>
1710 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1711 prologue cost vector and pass it to vect_get_load_cost.
1712 (vect_get_peeling_costs_all_drs): Likewise.
1713 (vect_peeling_hash_get_lowest_cost): Likewise.
1714 (vect_enhance_data_refs_alignment): Likewise.
1716 2018-04-26 Richard Biener <rguenther@suse.de>
1719 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1720 checking of integer<->pointer conversions.
1721 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1722 sign-/zero-extending pointer types.
1723 (expand_omp_for_static_chunk): Likewise.
1725 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1726 Jean Lee <xiaoyur347@gmail.com>
1728 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1729 (TARGET_ASAN_SHADOW_OFFSET): Define.
1730 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1731 true for -fsanitize=address.
1733 2018-04-25 Mark Wielaard <mark@klomp.org>
1735 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1738 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1740 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1741 than "alu", remove explicit "memory" and "imm_disp" attributes.
1742 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1745 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1746 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1747 gen_lowpart_no_emit.
1749 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1752 * config/i386/i386.c (ix86_expand_builtin): Change memory
1753 operand to XI, extend p0 to Pmode.
1754 * config/i386/i386.md: Change unspec volatile and operand
1755 1 mode to XI, change operand 0 mode to P.
1757 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1759 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1760 GET_MODE_MASK before any checking.
1761 (nds32_can_use_bset_p): Likewise.
1762 (nds32_can_use_btgl_p): Likewise.
1764 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1766 * config/nds32/nds32-doubleword.md: New define_split pattern for
1767 illegal register number.
1769 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1771 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1773 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1775 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1777 2018-04-25 Richard Biener <rguenther@suse.de>
1779 * lto-streamer.h (LTO_major_version): Bump to 8.
1781 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1783 * BASE-VER: Set to 9.0.0.
1785 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1787 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1788 in __abskf2 and __powikf2.
1790 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1793 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1794 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1795 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1796 Use the above on operand 2. Reindent.
1797 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1799 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1802 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1804 * config/i386/i386.opt (mcet): Removed.
1805 * doc/install.texi: Remove -mcet documentation.
1806 * doc/invoke.texi: Likewise.
1808 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1811 * doc/install.texi: Remove -mcet from bootstrap-cet.
1813 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1816 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1817 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1821 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1822 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1823 containing a CONST_VECTOR.
1825 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1827 * doc/install.texi: Update newlib dependency for nvptx.
1829 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1832 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1833 instead of INTVAL when shifting x left.
1835 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1837 PR tree-optimization/85478
1838 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1839 vect_grouped_store_supported for single element vectors.
1841 2018-04-24 Richard Biener <rguenther@suse.de>
1844 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1845 load cost increase to the case of non-constant step.
1847 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1850 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1851 destination if any_malformed_asm.
1853 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1856 * expr.c (store_field): In the bitfield case, if the value comes from
1857 a function call and is returned in registers by means of a PARALLEL,
1858 do not change the mode of the temporary unless BLKmode and VOIDmode.
1860 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1862 PR rtl-optimization/85423
1863 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1864 dependencies to debug insns when the previous insn is non-debug.
1866 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1868 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1869 enums into a single definition.
1870 (fls): Fix predicates and printing.
1873 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1875 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1876 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1877 and short u6 immediate.
1878 (check_if_valid_sleep_operand): Remove.
1879 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1881 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1883 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1884 flag_always_save_lp condition.
1885 * config/nds32/nds32.opt (malways-save-lp): New option.
1887 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1889 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1890 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1891 * config/nds32/nds32.h
1892 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1893 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1895 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1897 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1898 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1900 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1901 Chung-Ju Wu <jasonwucj@gmail.com>
1903 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1905 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1906 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1908 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1910 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1912 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1914 * config/nds32/nds32-protos.h (nds32_data_alignment,
1915 nds32_local_alignment): Declare.
1916 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1917 nds32_local_alignment): New functions.
1918 (TARGET_CONSTANT_ALIGNMENT): Define.
1919 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1921 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1923 * config/nds32/nds32.c
1924 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1925 (TARGET_MODES_TIEABLE_P): Likewise.
1927 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1929 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1932 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1933 Chung-Ju Wu <jasonwucj@gmail.com>
1935 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1936 for unaligned access.
1937 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1939 * config/nds32/nds32-intrinsic.md: Likewise.
1940 * config/nds32/nds32_intrinsic.h: Likewise.
1941 * config/nds32/nds32.h (nds32_builtins): Likewise.
1942 * config/nds32/nds32.opt (munaligned-access): New option.
1943 * config/nds32/nds32.c (nds32_asm_file_start): Display
1944 flag_unaligned_access status.
1946 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1948 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1949 -mno-relax is present.
1950 * config/riscv/linux.h (LINK_SPEC): Ditto.
1952 2018-04-20 Martin Sebor <msebor@redhat.com>
1955 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1957 (gimple_fold_builtin_stxcpy_chk): Same.
1958 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1960 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1963 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1964 __powikf2 when long double is IEEE 128-bit.
1966 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1968 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1969 step to make sure stack always aligned.
1971 2018-04-20 Carl Love <cel@us.ibm.com>
1974 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1975 size check for arg0.
1977 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1978 Tom de Vries <tom@codesourcery.com>
1981 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1982 Emit insns for calls too.
1983 (nvptx_find_par): Always look for worker-level predecessor insn.
1984 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1986 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1987 (nvptx_process_pars): Propagate frames for calls.
1989 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1992 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1994 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1995 (ix86_handle_option): Don't handle OPT_mibt.
1996 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1998 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2000 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2001 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2002 (ix86_target_macros): Define __CET__ with flag_cf_protection
2003 for -fcf-protection.
2004 * config/i386/i386.c (isa2_opts): Remove -mibt.
2005 * config/i386/i386.h (TARGET_IBT): Removed.
2006 (TARGET_IBT_P): Likewise.
2007 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2008 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2009 * config/i386/i386.opt (mcet): Update help message.
2012 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2013 -mcet as an alias for -mshstk.
2015 2018-04-20 Richard Biener <rguenther@suse.de>
2018 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2019 complexity by forcing a single use of the multiply operand.
2021 2018-04-20 Martin Jambor <mjambor@suse.cz>
2024 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2025 recursion dependency to only apply to non-clones.
2027 2018-04-20 Martin Jambor <mjambor@suse.cz>
2030 * ipa-cp.c (create_specialized_node): Check that clones of
2031 self-recursive edges exist during IPA-CP.
2033 2018-04-19 Toon Moene <toon@moene.org>
2035 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2038 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2040 PR tree-optimization/85467
2041 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2042 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2043 VECTOR_CST element to type.
2045 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2048 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2049 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2050 (builtin_longjmp): Likewise.
2051 (save_stack_nonlocal): New pattern.
2052 (restore_stack_nonlocal): Likewise.
2054 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2057 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2058 Replace ASM_OUTPUT_LABEL with fprintf.
2060 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2063 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2064 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2065 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2066 define __IBT__ and __SHSTK__ for -fcf-protection.
2067 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2069 (ix86_trampoline_init): Likewise.
2070 (x86_output_mi_thunk): Likewise.
2071 (ix86_notrack_prefixed_insn_p): Likewise.
2072 (ix86_option_override_internal): Don't disallow -fcf-protection.
2073 * config/i386/i386.md (rdssp<mode>): Also enable for
2075 (incssp<mode>): Likewise.
2076 (nop_endbr): Likewise.
2077 * config/i386/i386.opt (mcet): Change help message to built-in
2081 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2082 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2083 enable CET built-in functions.
2085 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2087 * common/config/i386/i386-common.c
2088 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2089 OPTION_MASK_ISA_MOVDIRI_UNSET,
2090 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2091 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2092 * config.gcc (movdirintrin.h): New header.
2093 * config/i386/cpuid.h (bit_MOVDIRI,
2094 bit_MOVDIR64B): New bits.
2095 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2097 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2098 (VOID, PVOID, PCVOID)): New function types.
2099 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2100 __builtin_ia32_directstoreu_u64,
2101 __builtin_ia32_movdir64b): New builtins.
2102 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2103 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2105 (ix86_valid_target_attribute_inner_p): Ditto.
2106 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2107 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2108 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2109 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2110 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2111 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2112 (movdiri<mode>, movdir64b_<mode>): New.
2113 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2114 * config/i386/immintrin.h: Include movdirintrin.h.
2115 * config/i386/movdirintrin.h: New file.
2116 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2118 2018-04-19 Richard Biener <rguenther@suse.de>
2121 * cfg.c (clear_bb_flags): When loop state says we have
2122 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2124 2018-04-19 Richard Biener <rguenther@suse.de>
2126 PR tree-optimization/84737
2127 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2128 copying restrict info.
2129 (vect_setup_realignment): Use it.
2130 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2131 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2132 the first DR to all generated stores.
2133 (vectorizable_load): Likewise for loads.
2135 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2137 PR tree-optimization/85446
2138 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2139 the integral and pointer types to have the same precision.
2141 * doc/install.texi: Document --disable-cet being the default and
2144 2018-04-18 Martin Liska <mliska@suse.cz>
2146 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2149 2018-04-18 Martin Liska <mliska@suse.cz>
2152 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2155 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2156 arguments if they are comparable.
2158 2018-04-18 Martin Liska <mliska@suse.cz>
2161 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2164 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2167 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2170 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2171 ENDBR after calling __morestack.
2173 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2176 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2177 by using gcc_base_ver to generate a gcc_driver_version, and use
2178 it when generating GCC_DRIVER_NAME.
2179 * configure: Regenerate.
2181 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2184 * config.gcc: Obsolete powerpc*-*-*spe*.
2186 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2189 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2190 (stabstr_D): Change type of unum from unsigned int to
2191 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2194 2018-04-17 Jim Wilson <jimw@sifive.com>
2197 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2198 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2199 Set arg_pointer_offset after using pretend_args_size.
2201 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2203 PR rtl-optimization/85431
2204 * dse.c (record_store): Ignore zero width stores.
2207 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2208 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2209 __builtin_stack_restore rather than after it.
2210 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2211 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2212 argument instead of virtual_dynamic_stack_rtx.
2214 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2216 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2218 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2219 Add note to error message to explain internal mapping of overloaded
2220 built-in function name to non-overloaded built-in function name.
2221 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2224 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2227 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2228 where the inputs overlap with the output.
2230 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2232 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2233 (=v, v) alternative and explicit "memory" attribute.
2234 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2235 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2237 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2238 "sselog1" type instead of "sselog".
2239 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2240 "sselog". Remove explicit "memory" attribute.
2241 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2242 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2244 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2245 "sselog1" type instead of "sselog". Remove explicit "memory"
2247 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2248 use "sselog1" type instead of "sselog". Remove explicit "memory"
2250 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2251 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2253 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2254 "sselog1" type instead of "sselog". Remove explicit "memory"
2256 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2257 use "sselog1" type instead of "sselog". Remove explicit "memory"
2261 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2264 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2267 2018-04-17 Martin Jambor <mjambor@suse.cz>
2270 * ipa-cp.c (create_specialized_node): Call
2271 expand_all_artificial_thunks if necessary.
2273 2018-04-17 Martin Liska <mliska@suse.cz>
2276 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2277 in message, remote space in between '_G' and '('.
2279 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2282 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2283 avx512f_vmcmp<mode>3<round_saeonly_name>,
2284 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2285 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2286 avx512f_rndscale<mode><round_saeonly_name>,
2287 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2288 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2289 Use %<iptr>2 instead of %2 for -masm=intel.
2290 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2291 avx512f_vcvttss2usi<round_saeonly_name>,
2292 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2294 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2295 avx512f_vcvttsd2usi<round_saeonly_name>,
2296 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2297 Use %q1 instead of %1 for -masm=intel.
2298 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2299 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2300 of %3 for -masm=intel.
2301 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2303 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2305 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2307 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2308 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2310 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2312 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2313 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2314 %g1 and one with %0 and %1.
2315 (avx512er_vmrcp28<mode><round_saeonly_name>,
2316 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2318 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2319 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2320 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2321 of %0 and %{%4%} for -masm=intel.
2322 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2323 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2324 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2325 order of %0 and %{%5%}%{z%} for -masm=intel.
2327 2018-04-17 Jan Hubicka <jh@suse.cz>
2330 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2332 2018-04-17 Martin Liska <mliska@suse.cz>
2335 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2336 for target_clone error message. Make default implementation
2337 clone to be a local declaration.
2338 (separate_attrs): Add new argument and check for an empty
2340 (expand_target_clones): Handle it.
2341 (ipa_target_clone): Make redirection just for target_clones
2344 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2345 Tom de Vries <tom@codesourcery.com>
2348 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2349 tiled basic blocks without omp continue statements.
2351 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2354 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2355 vec_extract expression as having side effects to make sure it gets
2358 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2361 * config/i386/i386.c (get_builtin_code_for_version): Check
2364 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2367 * gcc/config.gcc: Support "skylake".
2368 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2370 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2371 (processor_target_table): Add "skylake".
2372 (ix86_option_override_internal): Add "skylake".
2373 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2374 PROCESSOR_CANNONLAKE.
2375 (get_builtin_code_for_version): Fix priority for
2376 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2377 PROCESSOR_SKYLAKE-AVX512.
2378 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2379 (processor_type): Add PROCESSOR_SKYLAKE.
2381 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2382 Jason Merrill <jason@redhat.com>
2385 * convert.c (convert_to_integer_1): Use direct recursion for
2386 enumeral types and types with a precision less than the number
2387 of bits in their mode.
2389 2018-04-16 Julia Koval <julia.koval@intel.com>
2392 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2393 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2395 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2398 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2399 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2400 and -mno-direct-move.
2402 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2405 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2406 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2407 Ensure negative shifts result in {0}.
2409 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2411 PR rtl-optimization/79916
2412 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2413 regs (if any) to define how to gnerate SD moves when LRA is in
2416 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2418 PR rtl-optimization/85393
2419 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2420 * except.c (expand_dw2_landing_pad_for_region): Make static.
2421 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2422 a label and unconditional jump to old_bb, rather than
2423 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2426 PR rtl-optimization/85376
2427 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2428 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2429 instead of a specific value.
2431 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2432 Bin Cheng <bin.cheng@arm.com>
2434 PR tree-optimization/82965
2435 PR tree-optimization/83991
2436 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2437 by_profile_only parameter.
2438 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2439 information if the loop was predicted to iterate too many times.
2440 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2442 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2445 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2448 2018-04-13 Martin Liska <mliska@suse.cz>
2449 Jakub Jelinek <jakub@redhat.com>
2452 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2453 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2454 * builtins.c (expand_builtin_memory_copy_args): Use
2455 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2456 handle dest_addr == pc_rtx.
2458 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2461 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2462 asked to not generate direct moves.
2463 (fix_trunc<mode>si2_stfiwx): Similar.
2464 (fix_trunc<mode>si2_internal): Similar.
2466 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2469 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2470 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2471 lookup if dest in some wider mode is known to be const0_rtx and
2472 if so, record permanent equivalence for it to be ZERO_EXTEND of
2473 the narrower mode destination.
2475 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2477 * lto-streamer-out.c (output_function): Revert 259346.
2478 * omp-expand.c (expand_oacc_for): Likewise.
2480 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2482 PR rtl-optimization/85354
2483 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2484 * sel-sched.c (sel_global_init): ... here.
2486 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2489 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2490 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2491 mode for PE-COFF targets.
2492 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2493 (i386_pe_asm_lto_end): Likewise.
2494 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2495 (TARGET_ASM_LTO_END): Likewise.
2496 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2497 (i386_pe_asm_lto_start): New function.
2498 (i386_pe_asm_lto_end): Likewise.
2500 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2501 Richard Biener <rguenther@suse.de>
2504 * lto-streamer-out.c (output_function): Fix CFG loop state before
2506 * omp-expand.c (expand_oacc_for): Handle calls to internal
2507 functions like regular functions.
2509 2018-04-12 Richard Biener <rguenther@suse.de>
2512 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2513 for the early LTO debug to properly generate references to it
2514 during DIE emission. Do not re-use that for the skeleton for
2516 (dwarf2out_early_finish): Likewise.
2518 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2521 * config/i386/sse.md
2522 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2523 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2524 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2525 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2526 and output is a reg, avoid creating invalid lowpart subreg, but
2527 instead split into a 512-bit move. Don't split if not AVX512VL,
2528 input is xmm16+ reg and output is a mem.
2529 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2530 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2531 xmm16+ reg and output is a mem.
2533 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2535 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2536 also for flag_dwarf2_cfi_asm.
2538 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2540 PR rtl-optimization/85342
2541 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2542 a bool scalar var inside of the loop instead. Don't try to update
2543 recog_data.operand after failed apply_change_group.
2545 2018-04-12 Tom de Vries <tom@codesourcery.com>
2548 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2549 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2550 array with flexible array member as array without given dimension.
2551 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2552 argument for undefined param to true.
2554 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2557 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2558 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2559 from PowerPC section.
2560 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2561 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2562 help text that is too long.
2563 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2564 help text that is too long.
2565 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2566 help text that is too long.
2568 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2570 * config/alpha/alpha.md (stack_probe_internal): Rename
2571 from "probe_stack". Update all callers.
2573 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2575 PR rtl-optimization/84566
2576 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2577 sched_macro_fuse_insns.
2579 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2582 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2583 (compute_block_dependences): ... from here.
2585 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2587 PR tree-optimization/85331
2588 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2589 from int to HOST_WIDE_INT.
2591 2018-04-11 Martin Jambor <mjambor@suse.cz>
2594 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2595 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2596 not the same as the source val.
2597 (cgraph_edge_brings_value_p): New parameter.
2598 (gather_edges_for_value): Pass destination value to
2599 cgraph_edge_brings_value_p.
2600 (perhaps_add_new_callers): Likewise.
2601 (get_info_about_necessary_edges): Likewise and exclude values brought
2602 only by self-recursive edges.
2603 (create_specialized_node): Redirect only clones of self-calling edges.
2604 (+self_recursive_pass_through_p): New function.
2605 (find_more_scalar_values_for_callers_subset): Use it.
2606 (find_aggregate_values_for_callers_subset): Likewise.
2607 (known_aggs_to_agg_replacement_list): Removed.
2608 (decide_whether_version_node): Re-calculate known constants for all
2609 remaining context clones.
2611 2018-04-11 Richard Biener <rguenther@suse.de>
2614 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2615 from early DWARF output.
2616 (dwarf2out_early_finish): Output line info unconditionally into
2617 early DWARF and add reference to it.
2619 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2622 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2623 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2624 other than V2DFmode using iptr mode attribute.
2625 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2627 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2629 PR rtl-optimization/84659
2630 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2632 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2635 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2637 (output_loc_list): Pass address of a dummy size variable even in the
2638 locview handling loop.
2639 (index_location_lists): Add comment on why skip_loc_list_entry can't
2642 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2645 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2648 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2651 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2652 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2653 and -mstring-compare-inline-limit.
2655 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2658 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2659 for stack clash protection in a register whenever we need it to be in
2662 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2664 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2665 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2667 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2670 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2672 (mlong-double-): Ditto.
2673 * config/rs6000/sysv4.opt (msdata=): Ditto.
2674 (mtls-size=): Ditto.
2676 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2678 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2679 erroneous entries for
2680 "vector int vec_ldl (int, long int *)", and
2681 "vector unsigned int vec_ldl (int, unsigned long int *)".
2682 Add comments and entries for
2683 "vector bool char vec_ldl (int, bool char *)",
2684 "vector bool short vec_ldl (int, bool short *)",
2685 "vector bool int vec_ldl (int, bool int *)",
2686 "vector bool long long vec_ldl (int, bool long long *)",
2687 "vector pixel vec_ldl (int, pixel *)",
2688 "vector long long vec_ldl (int, long long *)",
2689 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2690 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2691 type tree bool_long_long_type_node and correct definition of
2692 bool_V2DI_type_node to make reference to this new type tree.
2693 (rs6000_mangle_type): Replace erroneous reference to
2694 bool_long_type_node with bool_long_long_type_node.
2695 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2696 comments to emphasize sign distinctions for char and int types and
2697 replace RS6000_BTI_bool_long constant with
2698 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2699 use of RS6000_BTI_pixel.
2700 (bool_long_type_node): Remove this macro definition.
2701 (bool_long_long_type_node): New macro definition
2703 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2705 PR rtl-optimization/85300
2706 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2707 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2708 simplify_unary_operation fails.
2710 2018-04-10 Martin Liska <mliska@suse.cz>
2712 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2713 cgraph_edge and ipa_ref.
2715 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2719 * config/i386/sse.md
2720 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2721 computation of the VEC_MERGE selector from mask.
2722 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2723 Fix decoding of the VEC_MERGE selector into mask.
2725 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2727 PR tree-optimization/85286
2728 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2730 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2732 * final.c (final_1): Set insn_last_address as well as
2733 insn_current_address.
2735 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2738 * explow.c (emit_stack_probe): Call validize_mem on memory location
2739 before passing it to gen_probe_stack. Create address operand and
2740 legitimize it for the probe_stack_address case.
2742 2018-04-09 Jan Hubicka <jh@suse.cz>
2745 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2746 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2747 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2749 (free_lang_data): Rebuild type inheritance graph.
2751 2018-04-09 Martin Sebor <msebor@redhat.com>
2753 * invoke.texi (-finline-small-functions): Mention other optimization
2755 (-findirect-inlining, -fpartial-inlining): Same.
2756 (-finline-functions-called-once): Same.
2757 (-freorder-blocks-and-partition): Same.
2759 2018-04-09 Jan Hubicka <jh@suse.cz>
2762 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2763 jumps; choose last target that matches the criteria (i.e.
2764 no partition changes for non-crossing jumps).
2765 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2766 support for redirecting crossing jumps to non-crossing.
2768 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2770 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2771 also for naked functions.
2773 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2775 * config/arc/arc.md (add_shift): New pattern.
2776 (add_shift2): Likewise.
2777 (sub_shift): Likewise.
2778 (sub_shift_cmp0_noout): Likewise.
2779 (compare_si_ashiftsi): Likewise.
2780 (xbfu_cmp0_noout): New combine pattern.
2781 (xbfu_cmp0"): Likewise.
2782 (movsi_set_cc_insn): Place the predicable variant first.
2783 (commutative_binary_cmp0_noout): Remove clobber.
2784 (commutative_binary_cmp0): New pattern.
2785 (noncommutative_binary_cmp0): Likewise.
2786 (noncommutative_binary_cmp0_noout): Likewise.
2787 (noncommutative_binary_comparison_result_used): Removed.
2788 (rsub_cmp0): New pattern.
2789 (rsub_cmp0_noout): Likewise.
2790 (extzvsi): Changed, keep only meaningful variants.
2791 (SQH, SEZ): New iterators.
2792 (SQH_postfix): New mode attribute.
2793 (SEZ_prefix): New code attribute.
2794 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2795 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2796 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2798 (noncommutative_operator): Check the availability of barrel
2801 2018-04-09 Richard Biener <rguenther@suse.de>
2803 PR tree-optimization/85284
2804 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2805 Only use the niter constraining form of simple_iv when the exit
2808 2018-04-09 Tom de Vries <tom@codesourcery.com>
2811 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2812 (define_expand "*memory_barrier"): New define_expand.
2813 (define_insn "memory_barrier"): New insn.
2815 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2817 PR rtl-optimization/80463
2818 PR rtl-optimization/83972
2819 PR rtl-optimization/83480
2821 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2822 correct producer for the insn.
2823 (tidy_control_flow): Fixup seqnos in case of debug insns.
2825 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2827 PR rtl-optimization/83913
2829 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2830 different sched-times when merging exprs.
2832 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2834 PR rtl-optimization/83962
2836 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2837 tidy_fallthru_edge and tidy_control_flow.
2839 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2841 PR rtl-optimization/83530
2843 * sel-sched.c (force_next_insn): New global variable.
2844 (remove_insn_for_debug): When force_next_insn is true, also leave only
2845 next insn in the ready list.
2846 (sel_sched_region): When the region wasn't scheduled, make another pass
2847 over it with force_next_insn set to 1.
2849 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2851 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2853 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2854 for interrupt control.
2855 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2856 functions for interrupt control.
2857 * config/nds32/nds32-intrinsic.md: Likewise.
2858 * config/nds32/nds32_intrinsic.h: Likewise.
2859 * config/nds32/nds32.h (nds32_builtins): Likewise.
2861 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2863 * config/nds32/nds32.c (nds32_init_machine_status,
2864 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2865 strict_aligned_p field.
2866 (nds32_expand_to_rtl_hook): New function.
2867 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2868 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2870 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2871 Chung-Ju Wu <jasonwucj@gmail.com>
2873 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2874 * config/nds32/nds32-n7.md: New file.
2875 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2876 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2878 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2879 * config/nds32/nds32.md (pipeline_model): Add n7.
2880 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2881 * config/nds32/pipelines.md: Include n7 settings.
2883 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2884 Chung-Ju Wu <jasonwucj@gmail.com>
2886 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2887 * config/nds32/nds32-e8.md: New file.
2888 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2889 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2891 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2892 * config/nds32/nds32.md (pipeline_model): Add e8.
2893 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2894 * config/nds32/pipelines.md: Include e8 settings.
2896 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2897 Chung-Ju Wu <jasonwucj@gmail.com>
2899 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2900 * config/nds32/nds32-n8.md: New file.
2901 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2902 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2904 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2905 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2906 * config/nds32/nds32.md (pipeline_model): Add n8.
2907 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2908 * config/nds32/pipelines.md: Include n8 settings.
2910 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2911 Chung-Ju Wu <jasonwucj@gmail.com>
2913 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2914 * config/nds32/nds32-n9-2r1w.md: New file.
2915 * config/nds32/nds32-n9-3r2w.md: New file.
2916 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2917 nds32_register_ports): New or modify for cpu n9.
2918 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2920 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2921 * config/nds32/nds32-utils.c: New file.
2922 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2923 TARGET_MUL_SLOW): Define.
2924 * config/nds32/nds32.md (pipeline_model): New attribute.
2925 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2926 New options that support cpu n9.
2927 * config/nds32/pipelines.md: Include n9 settings.
2928 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2930 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2932 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2933 information if necessary.
2934 (output_cond_branch_compare_zero): Likewise.
2935 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2936 (nds32_target_alignment): Refine for alignment.
2937 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2938 (FUNCTION_BOUNDARY): Modify.
2939 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2941 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2943 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2945 * config/nds32/constants.md (unspec_volatile_element): Add values for
2946 TLB operation and data prefetch.
2947 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2948 functions for TLB operation and data prefetch.
2949 * config/nds32/nds32-intrinsic.md: Likewise.
2950 * config/nds32/nds32_intrinsic.h: Likewise.
2951 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2952 (nds32_print_operand): Likewise.
2953 * config/nds32/nds32.h (nds32_builtins): Likewise.
2955 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2956 Andrew Pinski <pinsika@gcc.gnu.org>
2959 * match.pd: Use constant_boolean_node of correct type instead of
2960 boolean_true_node or boolean_false_node for simplifying
2961 pointer comparisons to zero.
2963 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2965 PR tree-optimization/80021
2966 * tree.c (verify_type_variant): Make error call in verify_variant_match
2967 translatable and remove final full stop.
2969 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2971 * config/nds32/constants.md (unspec_volatile_element): Add
2972 UNSPEC_VOLATILE_EH_RETURN.
2973 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2974 nds32_output_stack_pop): Support dwarf exception handling process.
2975 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2976 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2977 exception handling process.
2978 (nds32_compute_stack_frame): Likewise.
2979 (nds32_return_addr_rtx): Likewise.
2980 (nds32_initial_elimination_offset): Likewise.
2981 (nds32_expand_prologue): Likewise.
2982 (nds32_expand_epilogue): Likewise.
2983 (nds32_dynamic_chain_address): New function.
2984 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2986 (DYNAMIC_CHAIN_ADDRESS): Define.
2987 (EH_RETURN_DATA_REGNO): Define.
2988 (EH_RETURN_STACKADJ_RTX): Define.
2989 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2990 patterns for dwarf exception handling.
2992 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2994 * config/nds32/nds32.h: Clean up obsolete macros.
2996 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2998 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2999 Add enum values for particular instructions.
3000 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3001 particular intrinsic functions.
3002 * config/nds32/nds32-intrinsic.md: Likewise.
3003 * config/nds32/nds32_intrinsic.h: Likewise.
3004 * config/nds32/nds32.h (nds32_builtins): Likewise.
3005 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3006 (btst, ave): New patterns for particular instructions.
3008 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3010 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3011 Add enum values for atomic load/store and memory sync.
3012 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3014 * config/nds32/nds32-intrinsic.md: Likewise.
3015 * config/nds32/nds32_intrinsic.h: Likewise.
3016 * config/nds32/nds32.h (nds32_builtins): Likewise.
3018 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3020 PR tree-optimization/85257
3021 * fold-const.c (native_encode_vector): If not all elts could fit
3022 and off is -1, return 0 rather than offset.
3023 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3024 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3025 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3026 adjust buffer in native_interpret_expr call.
3028 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3030 * config/nds32/constants.md (unspec_volatile_element): Add cache
3031 control enum values.
3032 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3033 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3034 * config/nds32/nds32.c (nds32_cctl_names): New.
3035 (nds32_print_operand): Handle cache control register names.
3036 * config/nds32/nds32.h (nds32_builtins): New enum values.
3037 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3039 * config/nds32/nds32.md (type): Add mmu.
3040 * config/nds32/pipelines.md (simple_insn): Add mmu.
3042 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3044 * config/nds32/nds32.md (type): Remove call.
3045 * config/nds32/pipelines.md (simple_insn): Likewise.
3047 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3049 * config/nds32/constants.md (unspec_volatile_element): Add
3050 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3051 UNSPEC_VOLATILE_FMFCFG.
3052 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3053 description for fmfcfg and fmfcsr.
3054 (bdesc_1arg): Add fmtcsr.
3055 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3056 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3057 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3058 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3059 unspec_fmfcfg): New patterns.
3060 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3061 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3062 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3063 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3064 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3065 __nds32__fmfcfg): Define.
3067 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3069 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3070 intrinsic register names.
3071 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3072 intrinsic register enum values and macros.
3074 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3076 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3077 for load/store addressing form.
3078 (nds32_print_operand_address): Likewise.
3080 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3083 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3084 based on LABEL_REF. Remove useless assertion.
3085 (pic_address_needs_scratch): Fix formatting.
3086 (sparc_legitimize_pic_address): Minor tweaks.
3087 (sparc_delegitimize_address): Adjust assertion accordingly.
3088 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3089 into symbolic_operand.
3090 (movsi_high_pic_label_ref): Likewise.
3091 (movsi_lo_sum_pic_label_ref): Likewise.
3092 (movdi_pic_label_ref): Likewise.
3093 (movdi_high_pic_label_ref): Likewise.
3094 (movdi_lo_sum_pic_label_ref): Likewise.
3096 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3098 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3099 custom LIB_SPEC setup.
3101 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3102 Kito Cheng <kito.cheng@gmail.com>
3104 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3105 * config/riscv/freebsd.h: New.
3107 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3109 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3110 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3113 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3114 Kito Cheng <kito.cheng@gmail.com>
3116 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3117 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3118 * config/nds32/nds32-protos.h (nds32_output_call,
3119 nds32_output_return): Declare.
3120 * config/nds32/nds32.md: Refine all the call and return patterns.
3122 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3125 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3126 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3128 PR rtl-optimization/84872
3129 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3130 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3133 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3135 * expr.c (copy_blkmode_to_reg): Revert 254862.
3136 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3138 2018-04-06 Richard Biener <rguenther@suse.de>
3141 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3142 after seeing a component reference with an adjacent field. Treat
3143 refs to arrays at struct end of external decls similar to
3144 refs to unconstrained commons.
3146 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3149 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3150 look through SAVE_EXPRs with non-side-effects argument. Adjust
3152 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3155 2018-04-06 Richard Biener <rguenther@suse.de>
3158 * alias.c (find_base_term): New wrapper around find_base_term
3159 unwinding CSELIB_VAL_PTR changes.
3160 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3163 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3165 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3167 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3168 constant definitions.
3169 ("nop"): lr 0,0 -> nopr r0
3170 ("nop_lr0", "nop_lr1"): New insn definitions.
3172 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3174 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3175 NDS32_V3PUSH_AVAILABLE_P macro.
3177 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3178 Chung-Ju Wu <jasonwucj@gmail.com>
3180 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3181 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3182 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3183 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3184 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3185 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3186 * config/nds32/constraints.md: New constraints and checking for hard
3187 float configuration.
3188 * config/nds32/iterators.md: New mode iterator and attribute for hard
3189 float configuration.
3190 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3192 * config/nds32/nds32-fpu.md: New file.
3193 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3194 deal with hard float code generation.
3195 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3197 (abi_type, float_reg_number): New enum type.
3198 * config/nds32/nds32-predicates.c: New predicates for hard float.
3199 * config/nds32/nds32-protos.h: Declare functions for hard float.
3200 * config/nds32/nds32.c: Implementation for hard float configuration.
3201 * config/nds32/nds32.h: Definitions for hard float configuration.
3202 * config/nds32/nds32.md: Include hard float machine description and
3203 modify patterns for hard float configuration.
3204 * config/nds32/nds32.opt: New options for hard float configuration.
3205 * config/nds32/predicates.md: New predicates for hard float
3208 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3210 * common/config/nds32/nds32-common.c
3211 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3213 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3216 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3217 CONSTRUCTOR_ELT (ctor, ...)->value.
3219 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3222 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3224 2018-04-05 Tom de Vries <tom@codesourcery.com>
3227 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3230 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3231 Kito Cheng <kito.cheng@gmail.com>
3233 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3234 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3235 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3236 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3238 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3239 Kito Cheng <kito.cheng@gmail.com>
3241 * config/nds32/constraints.md (Ufe): New memory constraint.
3242 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3243 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3244 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3246 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3247 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3249 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3251 * config/nds32/nds32.md: Use optimize_size in the condition for
3252 alu-shift instructions.
3254 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3256 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3258 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3260 * config/nds32/nds32.md (negsi2): Refine pattern.
3262 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3263 Chung-Ju Wu <jasonwucj@gmail.com>
3265 * config/nds32/iterators.md (shift_rotate): New code iterator.
3266 (shift): New code attribute.
3267 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3268 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3269 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3270 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3271 bit-wise operations.
3272 (andsi3, *andsi3): Ditto.
3273 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3274 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3275 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3276 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3277 nds32_ior_operand, nds32_xor_operand): New predicates.
3279 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3281 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3282 (addsi3, subsi3): ... this.
3284 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3286 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3288 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3290 * config/nds32/nds32.md: Adjust indention.
3292 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3294 * config/nds32/nds32.md (feature): New attribute.
3296 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3298 * config/nds32/nds32.md (subtype): New attribute.
3300 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3303 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3304 expansion to perform a bitwise AND of the argument followed by a
3305 boolean negation of the result.
3307 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3309 PR rtl-optimization/84878
3310 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3311 the basic block. Assert the use reference is not artificial and that
3312 it has an associated insn.
3314 2018-04-04 Michael Matz <matz@suse.de>
3316 * builtins.c (compute_objsize): Pass correct operand
3317 to array_at_struct_end_p.
3319 2018-04-04 Richard Biener <rguenther@suse.de>
3322 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3323 from contexts for DINFO_LEVEL_TERSE and below.
3325 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3327 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3328 resiter_operand condition.
3329 * config/nds32/nds32.md (*move<mode>): Ditto.
3331 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3332 Monk Chiang <sh.chiang04@gmail.com>
3334 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3336 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3338 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3340 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3341 Kito Cheng <kito.cheng@gmail.com>
3343 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3344 nds32_cond_code_str, output_cond_branch,
3345 output_cond_branch_compare_zero, nds32_expand_cbranch,
3346 nds32_expand_cstore, nds32_expand_movcc,
3347 nds32_output_cbranchsi4_equality_zero,
3348 nds32_output_cbranchsi4_equality_reg,
3349 nds32_output_cbranchsi4_equality_reg_or_const_int,
3350 nds32_output_cbranchsi4_greater_less_zero: New functions.
3351 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3352 nds32_expand_cstore, nds32_expand_movcc,
3353 nds32_output_cbranchsi4_equality_zero,
3354 nds32_output_cbranchsi4_equality_reg,
3355 nds32_output_cbranchsi4_equality_reg_or_const_int,
3356 nds32_output_cbranchsi4_greater_less_zero): Declare.
3357 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3358 nds32_rimm11s_operand): New predicates.
3359 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3360 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3363 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3365 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3366 * config/nds32/nds32.md: Ditto.
3367 * config/nds32/pipelines.md: Ditto.
3369 2018-04-04 Richard Biener <rguenther@suse.de>
3371 PR tree-optimization/85168
3372 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3373 propagating abnormals.
3375 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3377 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3379 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3380 Kito Cheng <kito.cheng@gmail.com>
3382 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3383 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3384 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3385 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3386 * config/nds32/nds32.md (sibcall_internal): New.
3387 (sibcall_register): Remove.
3388 (sibcall_immediate): Remove.
3389 (sibcall_value_internal): New.
3390 (sibcall_value_register): Remove.
3391 (sibcall_value_immediate): Remove.
3392 * config/nds32/predicates.md (nds32_general_register_operand): New.
3393 (nds32_call_address_operand): New.
3395 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3397 PR rtl-optimization/85167
3398 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3399 bb_defs if *split_p, instead preinitialize it to NULL.
3401 PR tree-optimization/85156
3402 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3403 evaluating the argument multiple times.
3405 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3407 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3409 (_mm_cvtpd_ps): Likewise.
3410 (_mm_cvttpd_epi32): Likewise.
3411 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3412 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3413 vector, pixel, and bool following altivec.h include.
3415 2018-04-03 Martin Sebor <msebor@redhat.com>
3417 * doc/extend.texi (Common Function Attributes): Clarify.
3418 (const attribute): Likewise.
3419 (pure attribute): Likewise.
3421 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3424 * config/i386/i386.c (ix86_expand_vector_set): Use
3425 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3427 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3429 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3430 instructions when changing rounding bits to preserve precision bits
3431 in the x87 control word.
3433 2018-04-03 Martin Liska <mliska@suse.cz>
3435 PR tree-optimization/82491
3436 * rtl.h (strip_offset_and_add): Replace += suboffset with
3437 poly_uint64 () + suboffset.
3439 2018-03-29 Martin Liska <mliska@suse.cz>
3440 Martin Jambor <mjambor@suse.cz>
3443 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3444 param_type is not an integral or pointer type.
3446 2018-04-03 Richard Biener <rguenther@suse.de>
3448 * sese.h (recompute_all_dominators): Remove.
3450 2018-04-02 Martin Sebor <msebor@redhat.com>
3452 * doc/invoke.texi (-Wrestrict): Fix typos.
3454 2018-04-02 Jim Wilson <jimw@sifive.com>
3456 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3457 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3458 (<optab>di3, <optab>si3_extend): Likewise.
3459 (<optab>si3_mask, <optab>si3_mask_1): New.
3460 (<optab>di3_mask, <optab>di3_mask_1): New.
3461 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3462 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3463 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3465 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3467 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3470 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3472 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3473 (nds32_canonicalize_comparison): New function.
3475 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3476 Kito Cheng <kito.cheng@gmail.com>
3477 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3479 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3480 * config/nds32/constants.md (unspec_volatile_element): Add
3481 UNSPEC_VOLATILE_RELAX_GROUP.
3482 * config/nds32/nds32-relax-opt.c: New file.
3483 * config/nds32/nds32-predicates.c
3484 (nds32_symbol_load_store_p): New function.
3485 * config/nds32/nds32-protos.h
3486 (nds32_symbol_load_store_p): Declare function.
3487 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3488 * config/nds32/nds32.c
3489 (nds32_register_pass): New function to register pass.
3490 (nds32_register_passes): New function to register passes.
3491 * config/nds32/nds32.md (relax_group): New pattern.
3492 * config/nds32/nds32.opt (mrelax-hint): New option.
3493 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3495 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3497 * config/nds32/t-nds32: Modify files dependency.
3499 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3501 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3502 (PROFILE_HOOK): Define its implementation.
3504 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3506 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3507 type and 32-bit size.
3509 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3512 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3513 (V_128_256): New mode iterator.
3514 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3515 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3516 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3518 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3519 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3521 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3524 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3525 NaN inputs correctly.
3527 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3530 * config/rs6000/vsx.md (??r): New mode attribute.
3531 (*vsx_mov<mode>_64bit): Use it.
3532 (*vsx_mov<mode>_32bit): Likewise.
3534 2018-03-30 Martin Sebor <msebor@redhat.com>
3536 PR tree-optimization/84818
3537 * builtins.c (check_access): Use warning_n.
3539 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3542 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3544 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3547 2018-03-30 Julia Koval <julia.koval@intel.com>
3550 * x86-tune.def (movx, partial_reg_dependency): Enable for
3553 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3556 * lra-constraints.c (process_alt_operands): Move setting
3557 this_alternative_matches below.
3559 2018-03-29 Martin Liska <mliska@suse.cz>
3562 * doc/invoke.texi: Document how LTO works with debug info.
3563 Describe auto-load support of binutils. Mention 'x86-64'
3564 as valid option value of -march option.
3566 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3568 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3571 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3572 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3573 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3576 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3579 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3580 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3581 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3582 for RS6000_BTM_POWERPC64.
3583 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3584 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3585 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3590 2018-03-28 Carl Love <cel@us.ibm.com>
3593 2017-09-27 Carl Love <cel@us.ibm.com>
3595 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3596 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3597 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3600 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3602 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3603 instead of __vector bool.
3604 (_mm_max_pu8): Likewise.
3605 (_mm_min_pi16): Likewise.
3607 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3610 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3611 (DIVWEUO): Likewise.
3613 (DIVDEUO): Likewise.
3614 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3615 DIVWEUO and DIVDEUO.
3616 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3617 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3618 (div_extend): Likewise.
3619 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3621 (__builtin_divweuo): Likewise.
3622 (__builtin_divdeo): Likewise.
3623 (__builtin_divdeuo): Likewise.
3625 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3628 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3629 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3631 PR tree-optimization/82004
3632 * gimple-match-head.c (optimize_pow_to_exp): New function.
3633 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3634 Don't fold to exp if optimize_pow_to_exp is false.
3636 2018-03-28 Martin Liska <mliska@suse.cz>
3639 * calls.c (initialize_argument_information): Fix trailing space.
3640 * common.opt: Fix typo and provide better explanation for
3641 -fsanitize-coverage option.
3642 * config/i386/i386.opt: Fix typo.
3644 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3645 Martin Liska <mliska@suse.cz>
3648 * gimplify.c (asan_poison_variable): Don't do the check for
3649 gimplify_omp_ctxp here.
3650 (gimplify_decl_expr): Do it here.
3651 (gimplify_target_expr): Likewise.
3653 2018-03-28 Martin Liska <mliska@suse.cz>
3656 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3657 chkp_type_bounds_count if MPX is not enabled.
3659 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3661 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3663 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3666 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3667 function to create the function decl for complex long double
3668 multiply and divide for -mabi=ieeelongdouble.
3669 (init_float128_ieee): Call it.
3671 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3674 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3675 -fcf-protection=branch -mibt.
3676 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3678 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3681 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3683 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3686 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3687 extern array declarations.
3689 2018-03-27 Richard Biener <rguenther@suse.de>
3692 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3693 explicit single_use checks.
3695 2018-03-27 Richard Biener <rguenther@suse.de>
3697 PR tree-optimization/85082
3698 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3701 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3703 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3704 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3705 Turn on fasynchronous-unwind-tables and funwind-tables.
3707 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3710 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3711 (*bmi_blsr_<mode>_ccz): Ditto.
3713 2018-03-26 Tom de Vries <tom@codesourcery.com>
3715 PR tree-optimization/85063
3716 * omp-general.c (offloading_function_p): New function. Factor out
3718 * omp-offload.c (pass_omp_target_link::gate): ... here.
3719 * omp-general.h (offloading_function_p): Declare.
3720 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3721 with attribute omp declare target for offloading functions.
3723 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3725 PR tree-optimization/84005
3726 * tree-data-ref.h (get_base_for_alignment): Declare.
3727 * tree-data-ref.c (get_base_for_alignment_1): New function.
3728 (get_base_for_alignment): Likewise.
3729 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3730 get_base_for_alignment to find a suitable base object, instead
3731 of always using drb->base_address.
3733 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3736 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3737 known size by default.
3739 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3742 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3743 and non BLKmode operands.
3745 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3748 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3749 Clean up attributes.
3751 2018-03-23 Richard Biener <rguenther@suse.de>
3754 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3755 we are going to emit early debug for LTO.
3757 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3760 * function.c (match_asm_constraints_1): Don't optimize if input
3761 doesn't satisfy general_operand predicate for output's mode.
3764 * alias.c (write_dependence_p): Don't require for x_canonicalized
3765 non-VOIDmode if x has VOIDmode.
3768 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3769 just don't try to optimize it rather than assert it never happens.
3771 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3773 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3774 macro expansions for definition of ST_INTERNAL_<mode> and
3775 LD_INTERNAL_<mode> builtins.
3776 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3778 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3780 (altivec_expand_st_builtin): Likewise.
3781 (altivec_expand_builtin): Remove calls to deleted functions.
3782 (rs6000_address_for_altivec): Delete this function.
3783 * config/rs6000/vector.md: Remove expands for
3784 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3786 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3789 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3790 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3791 re-computing once computed.
3792 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3793 (arm_init_machine_status): Initialize
3794 machine->static_chain_stack_bytes.
3796 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3799 * doc/extend.texi: Add four new prototypes for vec_ld.
3800 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3801 definitions for more logical presentation.
3802 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3803 entries for V1TI variants of __builtin_altivec_ld builtin.
3804 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3805 handling of V1TI variant of LVX icode pattern.
3806 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3807 (rs6000_gimple_fold_builtin): Likewise.
3808 (altivec_init_builtins): Add code to define
3809 __builtin_altivec_lvx_v1ti function.
3811 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3814 * function.c (match_asm_constraints_1): Don't do the optimization
3815 if input isn't a REG, SUBREG, MEM or constant.
3817 2018-03-22 Tom de Vries <tom@codesourcery.com>
3819 PR tree-optimization/84956
3820 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3821 bb_has_abnormal_pred.
3823 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3826 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3827 DECL_INITIAL (decl) to decl at the end.
3828 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3831 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3833 * doc/extend.texi (__builtin_tgmath): Document when complex
3834 integer types are treated as _Complex _Float64.
3836 2018-03-21 Tom de Vries <tom@codesourcery.com>
3838 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3840 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3842 PR tree-optimization/84960
3843 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3844 if it is ENTRY block, move them into single succ of ENTRY in that case.
3846 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3848 PR tree-optimization/84811
3849 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3850 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3851 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3852 (known_subrange_p): Update accordingly. Cast each value involved
3853 in the size comparison, rather than casting the result of the
3856 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3858 PR tree-optimization/84982
3859 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3860 by flipping the least significant bit rather than all bits from
3861 bitpos to bitpos + bitsize - 1.
3863 2018-03-21 Nathan Sidwell <nathan@acm.org>
3865 * doc/extend.texi (Deprecated Features): Remove mention of
3866 long-deleted deprecations.
3868 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3871 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3872 * configure: Regenerate.
3874 2018-03-21 Tom de Vries <tom@codesourcery.com>
3876 PR tree-optimization/83126
3877 * tree-parloops.c (num_phis): New function.
3878 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3880 2018-03-21 Nathan Sidwell <nathan@acm.org>
3882 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3883 mention anon-struct/union members and trailing attributes.
3885 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3887 PR tree-optimization/84969
3888 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3889 builtin memset partitions if they set different rhs values.
3891 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3893 PR rtl-optimization/84989
3894 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3895 VEC_DUPLICATE with scalar result mode.
3897 2018-03-21 Martin Liska <mliska@suse.cz>
3900 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3901 not intended return statement.
3903 2018-03-21 Martin Liska <mliska@suse.cz>
3906 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3907 (chkp_find_bound_slots_1): Limit number of iterations.
3909 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3912 * Minor grammar fixes for x86 options.
3914 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3917 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3918 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3920 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3923 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3924 (altivec_lvx_<mode>_1op): Likewise.
3925 (altivec_stvx_<mode>_2op): Likewise.
3926 (altivec_stvx_<mode>_1op): Likewise.
3927 (altivec_lvx_<VM2:mode>): New define_expand.
3928 (altivec_stvx_<VM2:mode>): Likewise.
3929 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3930 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3931 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3932 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3933 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3934 (rs6000_gen_lvx): Likewise.
3935 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3936 (altivec_expand_stv_builtin): Likewise.
3937 (altivec_expand_builtin): Likewise.
3938 * config/rs6000/vector.md: Likewise.
3940 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3943 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3946 2018-03-20 Richard Biener <rguenther@suse.de>
3949 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3950 sign-conversions as zero, fall back to standard scalar_stmt
3953 2018-03-20 Martin Liska <mliska@suse.cz>
3956 * predict.c (rebuild_frequencies): Handle case when we have
3957 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3959 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3962 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3963 flag_section_anchors.
3964 * varasm.c (use_blocks_for_decl_p): Remove hack for
3965 dw2_force_const_mem.
3968 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3970 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3971 be created, use lowpart_subreg of operands[0] rather than operands[0]
3973 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3974 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3975 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3976 and n constraint instead of aarch64_shift_imm_di and Usd.
3977 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3978 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3980 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3983 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3984 to favor GPR over NEON registers.
3985 (<shift>di3_neon): Likewise.
3987 2018-03-20 Tom de Vries <tom@codesourcery.com>
3990 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3991 (nvptx_process_pars): Emit bar.sync asap and alap.
3993 2018-03-20 Tom de Vries <tom@codesourcery.com>
3996 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3997 seen_label if seen_label is already set.
3999 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4002 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4003 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4004 Use 1U instead of 1. Formatting fixes.
4007 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4008 instead of TREE_TYPE (s1) for the return value.
4010 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4012 PR tree-optimization/84946
4013 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4014 bitsize + bitsize in poly_uint64 rather than poly_int64.
4017 * dwarf2asm.c: Include fold-const.c.
4018 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4019 of decl rather than decl itself.
4021 PR rtl-optimization/84643
4022 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4024 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4027 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4028 calling assemble_variable.
4030 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4033 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4034 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4036 2018-03-19 Jim Wilson <jimw@sifive.com>
4039 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4040 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4041 (riscv_first_stack_step): Likewise.
4042 (riscv_option_override): Use STACK_BOUNDARY instead of
4044 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4046 (BIGGEST_ALIGNMENT): Set to 128.
4047 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4048 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4051 2018-03-19 Richard Biener <rguenther@suse.de>
4053 PR tree-optimization/84933
4054 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4055 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4057 2018-03-19 Richard Biener <rguenther@suse.de>
4059 PR tree-optimization/84859
4060 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4061 (cond_if_else_store_replacement): Perform sinking operation on
4062 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4063 Generalize what a BB with a single eligible store is.
4065 2018-03-19 Richard Biener <rguenther@suse.de>
4067 PR tree-optimization/84929
4068 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4069 chrec_is_positive against non-chrec arg.
4071 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4074 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4076 2018-03-18 Martin Liska <mliska@suse.cz>
4078 PR rtl-optimization/84635
4079 * regrename.c (build_def_use): Use matches_mode only when
4082 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4084 PR tree-optimization/84913
4085 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4086 vectorize chains of COND_EXPRs.
4088 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4090 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4092 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4094 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4096 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4098 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4100 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4101 Kito Cheng <kito.cheng@gmail.com>
4103 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4104 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4105 (nds32_adjust_reg_alloc_order): New function.
4106 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4108 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4110 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4111 nds32_print_operand, nds32_print_operand_address): Use
4112 HOST_WIDE_INT_PRINT_DEC instead.
4114 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4116 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4118 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4121 * config/i386/i386.c (initial_ix86_tune_features,
4122 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4124 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4125 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4126 rather than 1u << ix86_tune. Formatting fix.
4127 (ix86_option_override_internal): Change ix86_arch_mask from
4128 unsigned int to unsigned HOST_WIDE_INT, initialize to
4129 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4130 (ix86_function_specific_restore): Likewise.
4132 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4135 * postreload.c (reload_combine_recognize_pattern): Perform
4136 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4137 truncate_int_for_mode the result for the destination's mode.
4140 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4141 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4143 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4146 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4149 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4150 regno_allocno_class_array and sorted_pseudos.
4151 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4152 insns where regno is used.
4154 2018-03-16 Martin Liska <mliska@suse.cz>
4157 * multiple_target.c (create_dispatcher_calls): Redirect
4158 reference in the symbol table.
4160 2018-03-16 Martin Liska <mliska@suse.cz>
4163 * multiple_target.c (create_dispatcher_calls): Redirect also
4166 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4170 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4171 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4172 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4174 2018-03-16 Julia Koval <julia.koval@intel.com>
4176 * doc/invoke.texi (Skylake Server): Add CLWB.
4177 Cannonlake): Remove CLWB.
4179 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4181 PR tree-optimization/84841
4182 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4184 (FLOAT_ONE_CONST_TYPE): Define.
4185 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4186 (sort_by_operand_rank): Put entries with higher constant_type last
4187 rather than first to match comments.
4189 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4191 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4194 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4197 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4198 check and use address of target always.
4200 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4203 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4204 (indirect_thunk_bnd_needed): Likewise.
4205 (indirect_thunks_used): Likewise.
4206 (indirect_thunks_bnd_used): Likewise.
4207 (indirect_return_needed): New.
4208 (indirect_return_bnd_needed): Likewise.
4209 (output_indirect_thunk_function): Add a bool argument for
4211 (output_indirect_thunk_function): Don't generate alias for
4212 function return thunk.
4213 (ix86_code_end): Call output_indirect_thunk_function to generate
4214 function return thunks.
4215 (ix86_output_function_return): Set indirect_return_bnd_needed
4216 and indirect_return_needed instead of indirect_thunk_bnd_needed
4217 and indirect_thunk_needed.
4219 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4221 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4222 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4223 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4225 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4226 Paul Hua <paul.hua.gm@gmail.com>
4229 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4231 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4233 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4234 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4237 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4240 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4241 instead of GET_MODE_SIZE when comparing Units.
4243 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4246 * varasm.c (hash_section): Return an unchangeble hash value
4247 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4248 Return !aarch64_can_use_per_function_literal_pools_p ().
4250 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4253 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4254 rather than address of cmode as last argument to prepare_cmp_insn.
4256 2018-03-15 Julia Koval <julia.koval@intel.com>
4258 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4259 F_AVX512VNNI, F_AVX512BITALG): New.
4261 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4264 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4265 insn for floating-point loads and stores.
4267 2018-03-14 Carl Love <cel@us.ibm.com>
4269 * config/rs6000/rs6000-c.c: Add macro definitions for
4270 ALTIVEC_BUILTIN_VEC_PERMXOR.
4271 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4272 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4273 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4274 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4276 * config/doc/extend.texi: Add prototypes for vec_permxor.
4278 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4281 * diagnostic-show-locus.c (class layout_point): Convert m_line
4282 from int to linenum_type.
4283 (line_span::comparator): Use linenum "compare" function when
4284 comparing line numbers.
4285 (test_line_span): New function.
4286 (layout_range::contains_point): Convert param "row" from int to
4288 (layout_range::intersects_line_p): Likewise.
4289 (layout::will_show_line_p): Likewise.
4290 (layout::print_source_line): Likewise.
4291 (layout::should_print_annotation_line_p): Likewise.
4292 (layout::print_annotation_line): Likewise.
4293 (layout::print_leading_fixits): Likewise.
4294 (layout::annotation_line_showed_range_p): Likewise.
4295 (struct line_corrections): Likewise for field m_row.
4296 (line_corrections::line_corrections): Likewise for param "row".
4297 (layout::print_trailing_fixits): Likewise.
4298 (layout::get_state_at_point): Likewise.
4299 (layout::get_x_bound_for_row): Likewise.
4300 (layout::print_line): Likewise.
4301 (diagnostic_show_locus): Likewise for locals "last_line" and
4303 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4304 * input.c (selftest::test_linenum_comparisons): New function.
4305 (selftest::input_c_tests): Call it.
4306 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4307 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4308 * selftest.h (ASSERT_GT): New macro.
4309 (ASSERT_GT_AT): New macro.
4310 (ASSERT_LT): New macro.
4311 (ASSERT_LT_AT): New macro.
4313 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4315 PR rtl-optimization/84780
4316 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4318 2018-03-14 Martin Liska <mliska@suse.cz>
4320 * tree.c (record_node_allocation_statistics): Use
4321 get_stats_node_kind.
4322 (get_stats_node_kind): New function extracted from
4323 record_node_allocation_statistics.
4324 (free_node): Use get_stats_node_kind.
4326 2018-03-14 Richard Biener <rguenther@suse.de>
4328 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4329 that the value-set of ANTIC_IN doesn't grow.
4332 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4334 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4335 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4337 2018-03-14 Julia Koval <julia.koval@intel.com>
4339 * config.gcc (icelake-client, icelake-server): New.
4341 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4342 (initial_ix86_arch_features): Ditto.
4343 (PTA_SKYLAKE): Add SGX.
4344 (PTA_ICELAKE): Remove.
4345 (PTA_ICELAKE_CLIENT): New.
4346 (PTA_ICELAKE_SERVER): New.
4347 (ix86_option_override_internal): Split up icelake on icelake client and
4349 (get_builtin_code_for_version): Ditto.
4350 (fold_builtin_cpu): Ditto.
4351 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4352 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4353 * config/i386/i386.h (processor_type): Ditto.
4354 * doc/invoke.texi: Ditto.
4356 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4359 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4360 INTEGER_CST offset, add it together with bitpos / 8 and
4361 sign extend based on POINTER_SIZE.
4365 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4368 * config/i386/constraints.md (Yc): New register constraint.
4369 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4370 Use Yc constraint for alternative 2 of operand 0. Remove
4371 preferred_for_speed attribute.
4373 2018-03-14 Richard Biener <rguenther@suse.de>
4375 PR tree-optimization/84830
4376 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4377 with the old one to avoid oscillations.
4379 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4382 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4384 (assign_by_spills): Return a flag of reload assignment failure.
4385 Do not process the reload assignment failures. Do not spill other
4386 reload pseudos if they has the same reg class. Update n if
4388 (lra_assign): Add a return arg. Set up from the result of
4389 assign_by_spills call.
4390 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4391 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4392 usage_insns if it is not NULL.
4393 (spill_hard_reg_in_range): New function.
4394 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4395 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4396 function prototypes.
4397 (lra_assign): Change prototype.
4398 * lra.c (lra): Add code to deal with fails by splitting hard reg
4401 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4403 * config/riscv/riscv.opt (mrelax): New option.
4404 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4405 "norelax" when riscv_mrelax is disabled.
4406 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4408 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4411 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4412 reassociation for int modes.
4414 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4416 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4417 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4419 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4420 * config/aarch64/aarch64-sve.md
4421 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4422 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4423 (*extend<mode><Vwide>2): Rename to...
4424 (aarch64_sve_extend<mode><Vwide>2): ...this.
4425 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4426 renaming the old pattern to...
4427 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4429 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4430 define_expand, renaming the old pattern to...
4431 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4432 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4433 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4434 account when deciding which SVE instruction the optab should use.
4435 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4437 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4439 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4440 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4441 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4442 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4443 (tlsdesc_small_<mode>): Turn a define_expand and use
4444 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4445 (tlsdesc_small_advsimd_<mode>): ...this.
4446 (tlsdesc_small_sve_<mode>): New pattern.
4448 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4450 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4451 (UNSPEC_UMUL_HIGHPART): New constants.
4452 (MUL_HIGHPART): New int iteraor.
4453 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4454 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4456 (*<su>mul<mode>3_highpart): New define_insn.
4458 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4461 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4464 2018-03-13 Martin Liska <mliska@suse.cz>
4467 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4469 (sem_item_optimizer::~sem_item_optimizer): Release it.
4470 (sem_item_optimizer::merge_classes): Register variable aliases.
4471 (sem_item_optimizer::fixup_pt_set): New function.
4472 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4473 * ipa-icf.h: Declare new variables and functions.
4475 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4478 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4479 integer_pow2p@2 and test integer_pow2p in condition.
4480 (A < 0 ? C : 0): Similarly for @1.
4483 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4484 characters starting at p contain '\0' character, don't look beyond
4488 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4489 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4492 * reg-stack.c (change_stack): Change update_end var from int to
4493 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4494 also call set_block_for_insn on the newly added insns and rescan.
4497 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4498 on the last operand.
4501 * tree.c (stabilize_reference_1): Return save_expr (e) for
4502 STATEMENT_LIST even if it doesn't have side-effects.
4504 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4506 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4508 2018-03-12 Renlin Li <renlin.li@arm.com>
4510 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4511 aarch64_output_scalar_simd_mov_immediate.
4513 2018-03-12 Martin Sebor <msebor@redhat.com>
4515 PR tree-optimization/83456
4516 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4517 for perfectly overlapping calls to memcpy.
4518 (gimple_fold_builtin_memory_chk): Same.
4519 (gimple_fold_builtin_strcpy): Handle no-warning.
4520 (gimple_fold_builtin_stxcpy_chk): Same.
4521 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4523 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4525 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4526 parameter. Use it for SFmode.
4527 (rs6000_function_arg_advance_1): Adjust.
4528 (rs6000_function_arg): Adjust.
4529 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4531 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4533 PR rtl-optimization/84169
4534 PR rtl-optimization/84780
4535 * combine.c (can_combine_p): Check for a 2-insn combination whether
4536 the destination register is used between the two insns, too.
4538 2018-03-12 Richard Biener <rguenther@suse.de>
4540 PR tree-optimization/84803
4541 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4542 for refs DR analysis didn't process.
4544 2018-03-12 Richard Biener <rguenther@suse.de>
4546 PR tree-optimization/84777
4547 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4548 force-vectorize loops ignore whether we are optimizing for size.
4550 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4552 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4553 (TARGET_MD_ASM_ADJUST): Define.
4555 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4556 Kito Cheng <kito.cheng@gmail.com>
4557 Chung-Ju Wu <jasonwucj@gmail.com>
4559 * config/nds32/nds32.c (nds32_compute_stack_frame,
4560 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4561 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4562 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4563 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4564 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4565 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4566 * config/nds32/nds32.md (prologue, epilogue): Use macro
4567 NDS32_V3PUSH_AVAILABLE_P to do checking.
4569 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4572 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4573 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4574 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4575 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4576 addition of most attributes on !orig_type_die or the attribute not
4577 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4579 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4580 Chung-Ju Wu <jasonwucj@gmail.com>
4582 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4584 * config/nds32/nds32.opt (mvh): New option.
4586 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4587 Chung-Ju Wu <jasonwucj@gmail.com>
4589 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4591 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4592 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4595 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4596 Chung-Ju Wu <jasonwucj@gmail.com>
4598 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4600 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4601 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4603 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4604 Kito Cheng <kito.cheng@gmail.com>
4605 Chung-Ju Wu <jasonwucj@gmail.com>
4607 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4608 UNSPEC_FFMISM and UNSPEC_FLMISM.
4609 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4610 for ffb, ffmism and flmism.
4611 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4612 (unspec_ffmism): Ditto.
4613 (unspec_flmism): Ditto.
4614 (nds32_expand_builtin_impl): Check if string extension is available.
4615 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4616 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4618 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4621 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4624 * lra-assigns.c (assign_by_spills): Return a flag of reload
4625 assignment failure. Do not process the reload assignment
4626 failures. Do not spill other reload pseudos if they has the same
4628 (lra_assign): Add a return arg. Set up from the result of
4629 assign_by_spills call.
4630 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4631 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4632 usage_insns if it is not NULL.
4633 (spill_hard_reg_in_range): New function.
4634 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4635 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4636 function prototypes.
4637 (lra_assign): Change prototype.
4638 * lra.c (lra): Add code to deal with fails by splitting hard reg
4641 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4644 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4646 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4649 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4650 (dw_val_node): Add val_symbolic_view.
4651 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4652 (symview_upper_bound): New.
4653 (new_line_info_table): Initialize symviews_since_reset.
4654 (dwarf2out_source_line): Count symviews_since_reset and set
4655 symview_upper_bound.
4656 (dw_val_equal_p): Handle symview.
4657 (add_AT_symview): New.
4658 (print_dw_val): Handle symview.
4659 (attr_checksum, attr_checksum_ordered): Likewise.
4660 (same_dw_val_p, size_of_die): Likewise.
4661 (value_format, output_die): Likewise.
4662 (add_high_low_attributes): Use add_AT_symview for entry_view.
4663 (dwarf2out_finish): Reset symview_upper_bound, clear
4666 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4669 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4670 Add strict argument and use it.
4671 (rs6000_split_multireg_move): Update for new strict argument.
4672 (mem_operand_gpr): Disallow all non-offsettable addresses.
4673 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4675 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4678 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4679 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4680 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4683 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4684 decl, use remap_type if we want to use the type.
4686 2018-03-09 Martin Sebor <msebor@redhat.com>
4688 PR tree-optimization/84526
4689 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4691 (builtin_access::generic_overlap): Be prepared to handle non-array
4694 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4696 PR rtl-optimization/84682
4697 * lra-constraints.c (process_address_1): Check is_address flag
4698 for address constraints.
4699 (process_alt_operands): Likewise.
4700 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4701 preprocess_constraints.
4702 * recog.h (preprocess_constraints): Add oploc parameter.
4704 * recog.c (preprocess_constraints): Test address_operand for
4705 CT_ADDRESS constraints.
4707 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4710 * lra-assigns.c (assign_by_spills): Return a flag of reload
4711 assignment failure. Do not process the reload assignment
4712 failures. Do not spill other reload pseudos if they has the same
4714 (lra_assign): Add a return arg. Set up from the result of
4715 assign_by_spills call.
4716 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4717 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4718 usage_insns if it is not NULL.
4719 (spill_hard_reg_in_range): New function.
4720 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4721 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4722 function prototypes.
4723 (lra_assign): Change prototype.
4724 * lra.c (lra): Add code to deal with fails by splitting hard reg
4727 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4730 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4731 Accept complain bool parameter. Only emit errors if it is true.
4732 (arm_parse_cpu_option_name): Likewise.
4733 (arm_target_thumb_only): Adjust callers of the above.
4734 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4735 prototype to take a default true bool parameter.
4736 (arm_parse_arch_option_name): Likewise.
4738 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4739 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4743 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4744 * configure: Regenerate.
4745 * configure.ac ("linker --version-script option"): New.
4746 ("linker soname option"): New.
4748 2018-03-09 Richard Biener <rguenther@suse.de>
4750 PR tree-optimization/84775
4751 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4752 immediate uses of predicate stmts and mark them modified.
4755 PR tree-optimization/84178
4756 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4758 (version_loop_for_if_conversion): Delay update_ssa call.
4759 (tree_if_conversion): Delay update_ssa until after predicate
4762 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4765 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4766 when the function accesses prior frames.
4768 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4771 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4772 gen_llsym, otherwise call maybe_gen_llsym.
4775 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4776 has ',' character inside of it.
4778 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4781 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4782 as clobbering CC_REGNUM.
4784 2018-03-08 Richard Biener <rguenther@suse.de>
4787 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4788 (follow_copies_to_constant): Do not follow SSA names registered
4791 2018-03-08 Richard Biener <rguenther@suse.de>
4793 PR tree-optimization/84178
4794 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4796 (version_loop_for_if_conversion): Delay update_ssa call.
4797 (tree_if_conversion): Delay update_ssa until after predicate
4800 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4802 PR tree-optimization/84178
4803 * tree-if-conv.c (release_bb_predicate): Remove the
4804 the assertion that the stmts have NULL use_ops.
4805 Discard the statements, asserting that they haven't
4806 yet been added to a BB.
4808 2018-03-08 Richard Biener <rguenther@suse.de>
4810 PR tree-optimization/84746
4811 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4812 (phi_translate): Pass in destination ANTIC_OUT set.
4813 (phi_translate_1): Likewise. For a simplified result lookup
4814 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4815 (phi_translate_set): Adjust.
4816 (do_pre_regular_insertion): Likewise.
4817 (do_pre_partial_partial_insertion): Likewise.
4819 2018-03-08 Martin Liska <mliska@suse.cz>
4821 PR gcov-profile/84735
4822 * doc/gcov.texi: Document usage of profile files.
4823 * gcov-io.h: Document changes in the format.
4825 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4829 * dwarf2out.c (struct dw_line_info_table): Update comments for
4831 (FORCE_RESET_NEXT_VIEW): New.
4832 (FORCE_RESETTING_VIEW_P): New.
4833 (RESETTING_VIEW_P): Check for -1 too.
4834 (ZERO_VIEW_P): Likewise.
4835 (new_line_info_table): Force-reset next view.
4836 (dwarf2out_begin_function): Likewise.
4837 (dwarf2out_source_line): Simplify zero_view_p initialization.
4838 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4839 view directly. Omit view when omitting .loc at line 0.
4841 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4843 PR tree-optimization/84740
4844 * tree-switch-conversion.c (process_switch): Call build_constructors
4845 only if info.phi_count is non-zero.
4847 PR tree-optimization/84739
4848 * tree-tailcall.c (find_tail_calls): Check call arguments against
4849 DECL_ARGUMENTS (current_function_decl) rather than
4850 DECL_ARGUMENTS (func) when checking for tail recursion.
4852 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4854 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4855 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4856 Volker Reichelt's entry and add entries for people that perform
4857 GCC fuzzy testing and report numerous bugs.
4859 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4862 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4863 readonly data in sdata, if that is disabled.
4864 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4865 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4866 -mreadonly-in-sdata option.
4868 2018-03-07 Martin Sebor <msebor@redhat.com>
4870 PR tree-optimization/84468
4871 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4872 basic block when looking for nul assignment.
4874 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4877 * except.h (output_function_exception_table): Adjust prototype.
4878 * except.c (output_function_exception_table): Remove FNNAME parameter
4879 and add SECTION parameter. Ouput one part of the table at a time.
4880 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4881 the first part of the exception table and emit unwind directives.
4882 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4883 (i386_pe_seh_cold_init): Likewise.
4884 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4885 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4886 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4887 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4888 * config/i386/winnt.c: Include except.h.
4889 (struct seh_frame_state): Add reg_offset, after_prologue and
4890 in_cold_section fields.
4891 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4892 (i386_pe_seh_cold_init): New function.
4893 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4894 to seh->in_cold_section.
4895 (seh_emit_push): Record the offset of the push.
4896 (seh_emit_save): Record the offet of the save.
4897 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4898 Test seh->after_prologue to disregard the epilogue.
4899 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4900 (i386_pe_end_cold_function): New function.
4902 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4905 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4906 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4909 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4911 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4912 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4915 * multiple_target.c: Include tree-inline.h and intl.h.
4916 (expand_target_clones): Diagnose and fail if node->definition and
4917 !tree_versionable_function_p (node->decl).
4919 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4921 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4923 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4924 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4925 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4927 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4930 * combine.c (try_combine): Use reg_or_subregno instead of handling
4931 just paradoxical SUBREGs and REGs.
4933 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4935 * config/arc/arc.c (arc_finalize_pic): Remove function.
4936 (arc_must_save_register): We use single base PIC register, remove
4937 checks to save/restore the PIC register.
4938 (arc_expand_prologue): Likewise.
4939 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4941 (arc_verify_short): Likewise.
4942 (arc_attr_type): Likewise.
4943 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4944 (walk_stores): Likewise.
4945 (arc_address_cost): Make it static.
4946 (arc_verify_short): Likewise.
4947 (branch_dest): Likewise.
4948 (arc_attr_type): Likewise.
4949 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4950 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4951 (arc_final_prescan_insn): Remove inserting the nops due to
4952 hardware hazards. It is done in reorg step.
4953 (insn_length_variant_t): Remove.
4954 (insn_length_parameters_t): Likewise.
4955 (arc_insn_length_parameters): Likewise.
4956 (arc_get_insn_variants): Likewise.
4957 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4959 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4962 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4965 PR tree-optimization/84687
4966 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4968 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4970 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4972 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4973 Rename to ppc_speculation_barrier.
4974 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4975 __builtin_ppc_speculation_barrier.
4977 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4980 * combine.c (combine_simplify_rtx): Don't try to simplify if
4981 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4984 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4986 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4987 to 32 bytes when compiling for POWER9.
4989 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4992 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4993 regparm >= 3 with no arg reg available also for calls with
4994 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4997 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4999 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5001 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5004 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5006 2018-03-05 Richard Biener <rguenther@suse.de>
5008 PR tree-optimization/84486
5009 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5010 When inserting a __builtin_assume_aligned call set the LHS
5011 SSA name alignment info accordingly.
5013 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5015 PR tree-optimization/84114
5016 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5017 Avoid reassociation of FLOAT_MODE addition.
5019 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5021 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5022 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5023 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5024 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5025 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5026 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5027 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5029 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5030 __builtin_ia32_wbinvd): New builtins.
5031 (SPECIAL_ARGS2): New.
5032 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5033 (SPECIAL_ARGS2): New.
5034 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5035 (ix86_valid_target_attribute_inner_p): Ditto.
5036 (ix86_init_mmx_sse_builtins): Add special_args2.
5037 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5038 TARGET_WBNOINVD_P): New.
5039 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5040 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5041 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5042 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5043 * config/i386/pconfigintrin.h: New file.
5044 * config/i386/wbnoinvdintrin.h: Ditto.
5045 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5046 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5048 2018-03-05 Richard Biener <rguenther@suse.de>
5050 PR tree-optimization/84670
5051 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5053 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5054 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5055 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5056 doesn't grow if all successors (recursively) were visited at least
5059 2018-03-05 Richard Biener <rguenther@suse.de>
5061 PR tree-optimization/84650
5062 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5063 if executed in the loop pipeline.
5065 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5067 * doc/configfiles.texi (Configuration Files): Move info about
5068 conditionalizing $target-protos.h to...
5069 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5070 differs from $target-protos.h.
5072 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5073 Chung-Ju Wu <jasonwucj@gmail.com>
5075 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5076 * config/nds32/nds32-multiple.md (setmemsi): Define.
5077 * config/nds32/nds32-memory-manipulation.c
5078 (nds32_gen_dup_4_byte_to_word_value): New.
5079 (emit_setmem_word_loop): New.
5080 (emit_setmem_byte_loop): New.
5081 (nds32_expand_setmem_loop): New.
5082 (nds32_expand_setmem_loop_v3m): New.
5083 (nds32_expand_setmem_unroll): New.
5084 (nds32_expand_setmem): New.
5086 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5087 Chung-Ju Wu <jasonwucj@gmail.com>
5089 * config/nds32/nds32-memory-manipulation.c
5090 (nds32_emit_load_store): New.
5091 (nds32_emit_post_inc_load_store): New.
5092 (nds32_emit_mem_move): New.
5093 (nds32_emit_mem_move_block): New.
5094 (nds32_expand_movmemsi_loop_unknown_size): New.
5095 (nds32_expand_movmemsi_loop_known_size): New.
5096 (nds32_expand_movmemsi_loop): New.
5097 (nds32_expand_movmemsi_unroll): New.
5098 (nds32_expand_movmemqi): Rename ...
5099 (nds32_expand_movmemsi): ... to this.
5100 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5101 (movmemsi): ... to this.
5102 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5103 (nds32_expand_movmemsi): ... to this.
5105 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5106 Monk Chiang <sh.chiang04@gmail.com>
5107 Chung-Ju Wu <jasonwucj@gmail.com>
5109 * config/nds32/nds32-protos.h
5110 (nds32_expand_load_multiple): New arguments.
5111 (nds32_expand_store_multiple): Ditto.
5112 (nds32_valid_multiple_load_store): Rename ...
5113 (nds32_valid_multiple_load_store_p): ... to this.
5114 * config/nds32/nds32-memory-manipulation.c
5115 (nds32_expand_load_multiple): Refine implementation.
5116 (nds32_expand_store_multiple): Ditto.
5117 * config/nds32/nds32-multiple.md
5118 (load_multiple): Update nds32_expand_load_multiple interface.
5119 (store_multiple): Update nds32_expand_store_multiple interface.
5120 * config/nds32/nds32-predicates.c
5121 (nds32_valid_multiple_load_store): Rename ...
5122 (nds32_valid_multiple_load_store_p): ... to this and refine
5124 * config/nds32/predicates.md
5125 (nds32_load_multiple_and_update_address_operation): New predicate.
5126 (nds32_store_multiple_and_update_address_operation): New predicate.
5128 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5129 Chung-Ju Wu <jasonwucj@gmail.com>
5131 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5132 (combo): New attribute.
5133 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5135 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5137 * config/nds32/nds32.opt: Change -mcmodel= default value.
5139 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5140 Monk Chiang <sh.chiang04@gmail.com>
5141 Chung-Ju Wu <jasonwucj@gmail.com>
5143 * config/nds32/constants.md (unspec_element): New enum.
5144 * config/nds32/constraints.md (Umw): New constraint.
5145 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5146 * config/nds32/nds32-intrinsic.md: Likewise.
5147 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5148 (nds32_valid_smw_lwm_base_p): New.
5149 (nds32_output_smw_single_word): New.
5150 (nds32_output_lmw_single_word): New.
5151 (nds32_expand_unaligned_load): New.
5152 (nds32_expand_unaligned_store): New.
5153 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5154 (nds32_output_smw_single_word): Declare.
5155 (nds32_output_lmw_single_word): Declare.
5156 (nds32_expand_unaligned_load): Declare.
5157 (nds32_expand_unaligned_store): Declare.
5158 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5159 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5160 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5161 NDS32_BUILTIN_UASTORE_DW.
5162 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5165 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5166 Kito Cheng <kito.cheng@gmail.com>
5167 Chung-Ju Wu <jasonwucj@gmail.com>
5169 * config/nds32/nds32-intrinsic.c
5170 (nds32_expand_builtin_null_ftype_reg): Delete.
5171 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5172 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5173 (nds32_read_argument): New.
5174 (nds32_legitimize_target): Ditto.
5175 (nds32_legitimize_argument): Ditto.
5176 (nds32_check_constant_argument): Ditto.
5177 (nds32_expand_unop_builtin): Ditto.
5178 (nds32_expand_unopimm_builtin): Ditto.
5179 (nds32_expand_binop_builtin): Ditto.
5180 (nds32_builtin_decl_impl): Ditto.
5181 (builtin_description): Ditto.
5182 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5183 (nds32_init_builtins_impl): Ditto.
5184 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5185 (nds32_builtin_decl): New.
5186 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5187 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5189 2018-03-02 Jeff Law <law@redhat.com>
5191 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5192 (redundant_insn, fill_simple_delay_slots): Likewise.
5193 (fill_slots_from_thread): Likewise.
5194 * resource.c (mark_referenced_resources): Likewise.
5195 (mark_set_resources, find_dead_or_set_registers): Likewise.
5197 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5199 * substring-locations.h (format_warning_va): Formatting fix for
5201 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5203 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5204 * substring-locations.c: Include intl.h.
5205 (format_warning_va): Turned into small wrapper around
5206 format_warning_n_va, renamed to ...
5207 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5208 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5210 (format_warning_at_substring_n): New function.
5211 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5212 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5213 format_warning_at_substring with just a shorter name instead of
5214 const function pointer.
5215 (fmtwarn_n): New function.
5216 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5217 appropriate, get rid of all the fmtstr temporaries, move conditionals
5218 with G_() wrapped string literals directly into fmtwarn arguments,
5219 cast dir.len to (int), formatting fixes.
5221 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5223 * doc/invoke.texi: Remove "Cilk Plus" references.
5225 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5226 Richard Biener <rguenther@suse.de>
5229 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5230 for error or warning attributes if CALL_FROM_THUNK_P is set.
5233 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5236 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5237 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5240 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5241 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5243 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5244 instead of -1U in last predictors element's probability member.
5246 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5249 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5250 arguments if they are comparable.
5252 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5254 PR tree-optimization/84634
5255 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5256 masks and masked_loop_p with a single loop_masks, making sure it's
5257 null for bb vectorization.
5259 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5261 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5262 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5263 loop->force_vectorize to check whether there is no alias.
5265 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5268 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5270 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5272 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5273 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5274 instead of a loop around prev_real_insn.
5275 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5279 * config/i386/i386.c (ix86_print_operand): Use conditional
5280 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5283 2018-03-02 Richard Biener <rguenther@suse.de>
5285 PR tree-optimization/84427
5286 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5287 (bitmap_set_subtract_values): Rewrite to handle multiple
5290 (prune_clobbered_mems): Likewise.
5291 (phi_translate): Take edge instead of pred/phiblock.
5292 (phi_translate_1): Likewise.
5293 (phi_translate_set): Likewise. Insert all translated
5294 exprs for a value into the set, keeping possibly multiple
5295 expressions per value.
5296 (compute_antic_aux): Adjust for phi_translate changes.
5297 When intersecting union the expressions and prune those
5298 not in the final value set, keeping possibly multiple
5299 expressions per value. Do not use value-insertion
5300 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5301 all expressions. Add verification that the value-sets
5302 only shrink during iteration.
5303 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5304 (do_pre_regular_insertion): Likewise.
5305 (do_pre_partial_partial_insertion): Likewise.
5307 2018-03-02 Richard Biener <rguenther@suse.de>
5310 * config/darwin.c (saved_debug_info_level): New static global.
5311 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5312 (darwin_asm_lto_end): Restore debug info generation settings.
5314 2018-03-01 Martin Liska <mliska@suse.cz>
5317 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5320 2018-03-01 Richard Biener <rguenther@suse.de>
5323 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5324 type annotation with LTO when debug was disabled at compile-time.
5326 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5328 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5330 (mips_final_postscan_insn): Likewise.
5332 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5334 PR rtl-optimization/84528
5335 * alias.c (init_alias_target): Add commentary.
5336 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5337 a unique base value if the frame pointer is not eliminated
5338 to the stack pointer.
5340 2018-03-01 Tom de Vries <tom@codesourcery.com>
5342 PR rtl-optimization/83327
5343 * lra-int.h (hard_regs_spilled_into): Declare.
5344 * lra.c (hard_regs_spilled_into): Define.
5345 (init_reg_info): Init hard_regs_spilled_into.
5346 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5347 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5348 (process_bb_lives): Handle hard_regs_spilled_into.
5349 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5350 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5352 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5354 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5355 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5356 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5357 * config/rs6000/aix72.h: New file.
5359 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5361 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5362 instead of warning_at with conditional singular and plural messages
5366 * stor-layout.c (update_alignment_for_field): For
5367 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5368 && !DECL_PACKED (field), do the alignment update, just use
5369 only desired_align instead of MAX (type_align, desired_align)
5371 (place_field): Don't do known_align < desired_align handling
5372 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5373 is non-NULL, instead do it after rli->prev_field handling and
5374 only if not within a bitfield word. For DECL_PACKED (field)
5375 use type_align of BITS_PER_UNIT.
5377 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5379 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5380 superfluous parentheses and trailing spaces.
5382 2018-02-28 Richard Biener <rguenther@suse.de>
5384 PR tree-optimization/84584
5385 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5386 SCoPs with fake exit edge.
5388 2018-02-28 Martin Liska <mliska@suse.cz>
5391 * timevar.c (timer::print): Fix format to properly print 100%
5394 2018-02-28 Richard Biener <rguenther@suse.de>
5397 * genmatch.c (capture_info::walk_match): Do not mark
5398 captured expressions without operands as expr_p given
5399 they act more like predicates and should be subject to
5400 "lost tail" side-effect preserving.
5402 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5404 PR rtl-optimization/81611
5405 * auto-inc-dec.c (attempt_change): Move dead note from
5406 mem_insn if it's the next use of regno
5407 (find_address): Take address use of reg holding
5408 non-incremented value. Add parm to limit search to the named
5410 (merge_in_block): Attempt to use a mem insn that is the next
5411 use of the original regno.
5413 2018-02-27 Martin Sebor <msebor@redhat.com>
5416 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5417 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5419 2018-02-27 Martin Sebor <msebor@redhat.com>
5421 PR translation/84207
5422 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5423 n argument to unsigned HOST_WIDE_INT.
5424 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5425 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5426 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5427 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5429 2018-02-27 Richard Biener <rguenther@suse.de>
5431 PR tree-optimization/84512
5432 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5433 Do not use the estimate returned from record_stmt_cost for
5434 the scalar iteration cost but sum properly using add_stmt_cost.
5436 2018-02-27 Richard Biener <rguenther@suse.de>
5438 PR tree-optimization/84466
5439 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5440 Adjust last change to less strictly validate use operands.
5442 2018-02-27 Martin Liska <mliska@suse.cz>
5444 PR gcov-profile/84548
5445 * gcov.c (process_file): Allow partial overlap and consider it
5446 also as group functions.
5447 (output_lines): Properly calculate range of lines for a group.
5449 2018-02-27 Martin Liska <mliska@suse.cz>
5451 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5452 'ggc' suffixes. Change first column width.
5453 (timer::print): Fix formatting of the column.
5455 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5457 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5458 preserve inline entry blocks for the sake of debug inline
5459 entry point markers alone.
5460 (remove_unused_locals): Suggest in comments a better place to
5461 force the preservation of inline entry blocks that are
5462 otherwise unused, but do not preserve them.
5464 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5466 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5468 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5471 * config/i386/constraints.md (Bs): Replace
5472 ix86_indirect_branch_register with
5473 TARGET_INDIRECT_BRANCH_REGISTER.
5475 * config/i386/i386.md (indirect_jump): Likewise.
5476 (tablejump): Likewise.
5477 (*sibcall_memory): Likewise.
5478 (*sibcall_value_memory): Likewise.
5479 Peepholes of indirect call and jump via memory: Likewise.
5480 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5481 (*sibcall_value_GOT_32): Likewise.
5482 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5483 (GOT_memory_operand): Likewise.
5484 (call_insn_operand): Likewise.
5485 (sibcall_insn_operand): Likewise.
5486 (GOT32_symbol_operand): Likewise.
5487 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5489 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5491 PR rtl-optimization/83496
5492 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5493 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5494 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5495 redundant insn, if any.
5496 (relax_delay_slots): Likewise.
5497 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5499 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5501 PR tree-optimization/83965
5502 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5503 that grouped statements are part of a reduction chain. Return
5504 true if the statement is not marked as a reduction itself but
5506 (vect_recog_dot_prod_pattern): Don't check whether the statement
5507 is part of a group here.
5508 (vect_recog_sad_pattern): Likewise.
5509 (vect_recog_widen_sum_pattern): Likewise.
5511 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5514 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5516 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5519 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5521 (ix86_output_indirect_function_return): New prototype.
5522 (ix86_split_simple_return_pop_internal): Likewise.
5523 * config/i386/i386.c (indirect_return_via_cx): New.
5524 (indirect_return_via_cx_bnd): Likewise.
5525 (indirect_thunk_name): Handle return va CX_REG.
5526 (output_indirect_thunk_function): Create alias for
5527 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5528 (ix86_output_indirect_jmp): Remove the bool argument.
5529 (ix86_output_indirect_function_return): New function.
5530 (ix86_split_simple_return_pop_internal): Likewise.
5531 * config/i386/i386.md (*indirect_jump): Don't pass false
5532 to ix86_output_indirect_jmp.
5533 (*tablejump_1): Likewise.
5534 (simple_return_pop_internal): Change it to define_insn_and_split.
5535 Call ix86_split_simple_return_pop_internal to split it for
5537 (simple_return_indirect_internal): Call
5538 ix86_output_indirect_function_return instead of
5539 ix86_output_indirect_jmp.
5541 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5544 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5545 memset and value initialization afterwards.
5547 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5549 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5551 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5554 * common/config/aarch64/aarch64-common.c
5555 (aarch_option_optimization_table[]): Switch
5556 off fomit-frame-pointer
5558 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5559 Chung-Ju Wu <jasonwucj@gmail.com>
5561 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5563 (store_multiple): Ditto.
5565 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5567 * config.gcc: Add --with-cpu support for nds32 target.
5568 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5569 * config/nds32/nds32.opt: Add -mcpu= option.
5571 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5573 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5574 isel=yes): Warn for these deprecated options.
5576 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5578 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5579 ISA_2_5_MASKS_EMBEDDED.
5581 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5583 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5584 p->max as pointers rather than using iterative_hash_expr.
5586 2018-02-23 Carl Love <cel@us.ibm.com>
5588 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5589 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5591 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5592 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5593 P8V_BUILTIN_VEC_VUNSIGNED2.
5595 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5598 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5599 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5601 * lra-constraints.c (curr_insn_transform): Set up
5602 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5604 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5605 * lra-eliminations.c (spill_pseudos): Ditto.
5606 (process_insn_for_elimination): Ditto.
5607 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5608 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5609 LRA_NON_CLOBBERED_ALT.
5611 2018-02-22 Martin Sebor <msebor@redhat.com>
5613 PR tree-optimization/84480
5614 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5615 to maybe_diag_stxncpy_trunc. Call it.
5616 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5617 from gimple_fold_builtin_strcpy. Print inlining stack.
5618 (handle_builtin_stxncpy): Print inlining stack.
5619 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5621 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5624 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5625 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5626 and -fcheck-pointer-bounds are used together.
5627 (indirect_thunk_prefix): New enum.
5628 (indirect_thunk_need_prefix): New function.
5629 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5630 "_nt" instead of "_bnd" for NOTRACK prefix.
5631 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5632 (output_indirect_thunk_function): Likewise.
5634 (ix86_code_end): Update output_indirect_thunk_function calls.
5635 (ix86_output_indirect_branch_via_reg): Replace
5636 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5637 (ix86_output_indirect_branch_via_push): Likewise.
5638 (ix86_output_function_return): Likewise.
5639 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5640 incompatible with -fcf-protection=branch and
5641 -fcheck-pointer-bounds.
5643 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5646 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5647 Change gcc_assert call to output_operand_lossage.
5649 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5651 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5653 2018-02-22 DJ Delorie <dj@redhat.com>
5654 Sebastian Perta <sebastian.perta@renesas.com>
5655 Oleg Endo <olegendo@gcc.gnu.org>
5657 * config/rx/rx.c (rx_rtx_costs): New function.
5658 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5660 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5662 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5664 2018-02-22 Martin Liska <mliska@suse.cz>
5667 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5668 Add "native" as a possible value.
5670 2018-02-22 Martin Liska <mliska@suse.cz>
5673 * config/i386/i386.c (ix86_option_override_internal):
5674 Add "native" as a possible value for -march and -mtune.
5676 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5679 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5680 to all type variants.
5682 PR tree-optimization/84503
5683 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5684 width as info->bitpos + info->bitsize - start.
5685 (merged_store_group::merge_overlapping): Simplify width computation.
5686 (check_no_overlap): New function.
5687 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5688 start + width and last_order of the group, fail if check_no_overlap
5690 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5691 to group if check_no_overlap fails.
5693 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5695 * config/rs6000/altivec.md: Delete contraint arguments to
5696 define_expand, define_split, and define_peephole2, and in
5697 define_insn_and_split if always unused.
5698 * config/rs6000/darwin.md: Ditto.
5699 * config/rs6000/dfp.md: Ditto.
5700 * config/rs6000/rs6000.md: Ditto.
5701 * config/rs6000/sync.md: Ditto.
5702 * config/rs6000/vector.md: Ditto.
5703 * config/rs6000/vsx.md: Ditto.
5705 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5707 * config/rs6000/altivec.md: Write output control strings as braced
5708 blocks instead of double-quoted strings.
5709 * config/rs6000/darwin.md: Ditto.
5710 * config/rs6000/rs6000.md: Ditto.
5711 * config/rs6000/vector.md: Ditto.
5712 * config/rs6000/vsx.md: Ditto.
5714 2018-02-21 Jason Merrill <jason@redhat.com>
5716 PR c++/84314 - ICE with templates and fastcall attribute.
5717 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5719 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5721 * ipa-cp.c (determine_versionability): Fix comment typos.
5723 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5726 * ipa-cp.c (determine_versionability): Do not version functions caling
5729 2018-02-21 Martin Liska <mliska@suse.cz>
5732 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5733 Add "native" as a possible value.
5734 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5735 the macro when native cpu detection is available.
5737 2018-02-21 Martin Liska <mliska@suse.cz>
5740 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5741 Add "native" as a possible value.
5742 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5743 when native cpu detection is available.
5745 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5746 Martin Sebor <msebor@redhat.com>
5748 PR tree-optimization/84478
5749 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5751 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5752 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5753 support which is conservatively correct, for 2 only stay conservative
5754 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5755 argument to the 2 argument get_range_strlen, adjust 6 arg
5756 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5758 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5759 (gimple_fold_builtin_strlen): Pass true as last argument to
5762 2018-02-20 Martin Sebor <msebor@redhat.com>
5765 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5766 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5767 (builtin_memref::builtin_memref): Factor out parts into
5768 set_base_and_offset and call it.
5770 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5773 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5774 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5775 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5776 search at the associated MODE_INT.
5778 2018-02-20 Jeff Law <law@redhat.com>
5781 PR tree-optimization/81592
5783 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5784 for range data rather than using global data.
5785 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5786 range data rather than using global data.
5787 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5788 pass it to children as needed.
5789 (struct directive::fmtresult): Similarly.
5790 (struct directive::set_width): Similarly.
5791 (struct directive::set_precision): Similarly.
5792 (format_integer, format_directive, parse_directive): Similarly.
5793 (format_none): Accept unnamed vr_values parameter.
5794 (format_percent, format_floating, format_character): Similarly.
5795 (format_string, format_plain): Similarly.
5796 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5797 the EVRP range analyzer for range data rather than using global data.
5798 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5799 gimple-ssa-evrp-analyze.h
5800 (class sprintf_dom_walker): Add after_dom_children member function.
5801 Add evrp_range_analyzer member.
5802 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5803 range analyzer as needed.
5804 (sprintf_dom_walker::after_dom_children): New member function.
5805 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5807 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5808 (evrp_range_analyzer::pop_to_marker): Likewise.
5810 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5812 PR tree-optimization/84419
5813 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5814 with the required type if its current type is compatible but
5817 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5820 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5821 after vectorization.
5823 2018-02-20 Martin Liska <mliska@suse.cz>
5826 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5827 possible values if we don't have a hint.
5829 2018-02-20 Martin Liska <mliska@suse.cz>
5833 * final.c (shorten_branches): Build align_tab array with one
5835 * opts.c (finish_options): Add alignment option limit check.
5836 (MAX_CODE_ALIGN): Likewise.
5837 (MAX_CODE_ALIGN_VALUE): Likewise.
5838 * doc/invoke.texi: Document maximum allowed option value for
5839 all -falign-* options.
5841 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5844 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5845 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5846 * var-tracking.c (emit_note_insn_var_location): Remove all references
5847 to NOTE_INSN_CALL_ARG_LOCATION.
5848 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5849 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5850 Use copy_rtx_if_shared.
5851 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5852 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5853 (dwarf2out_var_location): Remove handling of
5854 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5856 * final.c (final_scan_insn): Remove all references to
5857 NOTE_INSN_CALL_ARG_LOCATION.
5858 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5859 before dumping final insns.
5860 * except.c (emit_note_eh_region_end): Remove all references to
5861 NOTE_INSN_CALL_ARG_LOCATION.
5862 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5863 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5864 * config/arc/arc.c (hwloop_optimize): Likewise.
5865 * config/arm/arm.c (create_fix_barrier): Likewise.
5866 * config/s390/s390.c (s390_chunkify_start): Likewise.
5867 * config/sh/sh.c (find_barrier): Likewise.
5868 * config/i386/i386.c (rest_of_insert_endbranch,
5869 ix86_seh_fixup_eh_fallthru): Likewise.
5870 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5871 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5872 * config/frv/frv.c (frv_function_prologue): Likewise.
5873 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5875 (note_outside_basic_block_p): Remove all references to
5876 NOTE_INSN_CALL_ARG_LOCATION.
5877 * gengtype.c (adjust_field_rtx_def): Likewise.
5878 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5880 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5881 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5884 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5887 PR tree-optimization/84452
5888 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5889 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5892 2018-02-19 Martin Liska <mliska@suse.cz>
5895 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5897 2018-02-19 Martin Liska <mliska@suse.cz>
5898 Richard Sandiford <richard.sandiford@linaro.org>
5900 PR tree-optimization/82491
5901 * gimple-fold.c (get_base_constructor): Make earlier bail out
5904 2018-02-19 Carl Love <cel@us.ibm.com>
5906 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5907 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5909 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5910 P8V_BUILTIN_VEC_NEG.
5912 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5914 * config/rl78/rl78.md (movdf): New define expand.
5916 2018-02-19 Martin Liska <mliska@suse.cz>
5919 * doc/invoke.texi: Fix typo.
5920 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5922 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5924 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5925 handle rs6000_single_float and rs6000_double_float specially for
5928 2018-02-16 Jeff Law <law@redhat.com>
5930 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5933 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5936 * ipa-inline.c (inline_small_functions): Fix a typo.
5938 2018-02-16 Nathan Sidwell <nathan@acm.org>
5940 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5942 2018-02-16 Carl Love <cel@us.ibm.com>
5944 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5945 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5946 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5947 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5948 expansion to P8V_BUILTIN_VEC_FLOAT2.
5950 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5952 PR rtl-optimization/70023
5953 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5954 src_regno into account.
5956 2018-02-16 Carl Love <cel@us.ibm.com>
5958 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5959 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5960 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5961 * config/rs6000/rs6000.c: Remove case statements for
5962 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5963 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5964 and P9V_BUILTIN_VEC_VINSERT4B.
5965 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5966 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5967 * config/rs6000/vsx.md:
5968 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5971 2018-02-16 Carl Love <cel@us.ibm.com>
5973 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5975 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5977 * config/rs6000/rs6000-c.c: Add the definitions for
5978 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5979 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5980 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5981 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5982 definition for insert4b and define insn *insert3b_internal.
5983 * doc/extend.texi: Add documentation for vec_extract4b.
5985 2018-02-16 Nathan Sidwell <nathan@acm.org>
5987 * doc/extend.texi (Backwards Compatibility): Mention friend
5988 injection. Note for-scope is deprecated.
5989 * doc/invoke.texi (-ffriend-injection): Deprecate.
5991 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5993 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5994 that moved to I2, also allow destinations that are a paradoxical
5995 subreg (instead of a normal reg).
5997 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6000 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6003 2018-02-16 Richard Biener <rguenther@suse.de>
6005 PR tree-optimization/84037
6006 PR tree-optimization/84016
6008 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6009 Adjust vec_construct for the fact we need additional higher latency
6010 128bit inserts for AVX256 and AVX512 vector builds.
6011 (ix86_add_stmt_cost): Scale vector construction cost for
6014 2018-02-16 Richard Biener <rguenther@suse.de>
6016 PR tree-optimization/84417
6017 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6018 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6019 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6021 2018-02-16 Martin Liska <mliska@suse.cz>
6024 * internal-fn.def (ASAN_CHECK): Set proper flags.
6025 (ASAN_MARK): Likewise.
6027 2018-02-16 Julia Koval <julia.koval@intel.com>
6029 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6030 from PTA_CANNONLAKE.
6032 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6035 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6036 Use ++iter rather than iter++ for std::list iterators.
6037 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6038 defer deleting them until all nodes in the forest are processed. Do
6039 free even leaf nodes. Change to_process into auto_vec.
6042 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6043 * vec.h (vec_default_construct): Use memset instead of placement new
6044 if BROKEN_VALUE_INITIALIZATION is defined.
6045 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6046 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6049 PR rtl-optimization/83723
6050 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6051 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6052 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6054 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6056 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6058 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6060 PR rtl-optimization/81443
6061 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6062 from inner REGs to paradoxical SUBREGs.
6064 2018-02-16 Richard Biener <rguenther@suse.de>
6066 PR tree-optimization/84399
6067 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6068 For operands we can analyze at their definition make sure we can
6069 analyze them at each use as well.
6071 2018-02-16 Richard Biener <rguenther@suse.de>
6073 PR tree-optimization/84190
6074 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6075 volatile accesses if the decl isn't volatile.
6077 2018-02-15 Jason Merrill <jason@redhat.com>
6079 PR c++/84314 - ICE with templates and fastcall attribute.
6080 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6081 TYPE_CANONICAL on an existing type.
6083 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6085 PR tree-optimization/84383
6086 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6087 dstoff nor call operand_equal_p if dstbase is NULL.
6089 PR tree-optimization/84334
6090 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6091 also a CONSTANT_CLASS_P, punt.
6093 2018-02-14 Jim Wilson <jimw@sifive.com>
6095 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6096 first SMALL_OPERAND check. New local min_second_step. Move assert
6097 to where locals are set. Add TARGET_RVC support.
6098 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6100 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6102 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6104 2018-02-14 Martin Sebor <msebor@redhat.com>
6106 PR tree-optimization/83698
6107 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6108 arrays constrain the offset range to their bounds.
6109 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6110 (builtin_access::overlap): Avoid setting the size of overlap if it's
6112 (maybe_diag_overlap): Also consider arrays when deciding what values
6113 of offsets to include in diagnostics.
6115 2018-02-14 Martin Sebor <msebor@redhat.com>
6118 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6119 that correspond to the kind of a declaration.
6121 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6124 * config/pa/pa.md: Load address of PIC label using the linkage table
6125 if the label is nonlocal.
6127 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6129 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6130 warning message if user requests -maltivec=be.
6131 * doc/invoke.texi: Document deprecation of -maltivec=be.
6133 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6136 * config/rs6000/rs6000-c.c: Update definitions for
6137 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6138 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6140 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6143 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6144 add _get_ssp intrinsics. Remove argument from
6145 __builtin_ia32_rdssp[d|q].
6146 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6147 * config/i386/i386-builtin.def: Remove argument from
6148 __builtin_ia32_rdssp[d|q].
6149 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6150 ix86_expand_special_args_builtin for _rdssp[d|q].
6151 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6152 Clear register before usage.
6153 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6154 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6156 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6158 PR tree-optimization/84357
6159 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6160 operand 1 of an ARRAY_REF too.
6162 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6165 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6166 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6168 (set_of_reg): New struct.
6169 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6170 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6171 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6173 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6174 Split into bitclr, bitset, bitinvert patterns if appropriate.
6175 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6176 use rx_fuse_in_memory_bitop.
6177 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6178 to named insn, correct maximum insn length.
6180 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6183 * machmode.def: Define a complex mode for PARTIAL_INT.
6184 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6186 * doc/rtl.texi: Document CSPImode.
6187 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6189 (msp430_hard_regno_nregs_with_padding): Likewise.
6191 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6194 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6196 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6198 PR rtl-optimization/84169
6199 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6200 we generated a parallel as new i3 and we split that to new i2 and i3
6201 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6202 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6203 those to i2, not i1. Partially rewrite this scan code.
6205 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6208 * stor-layout.c (place_field): For variable length fields, adjust
6209 offset_align afterwards not just based on the field's alignment,
6210 but also on the size.
6213 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6214 of exps and logs in the use_exp2 case.
6216 2018-02-13 Jeff Law <law@redhat.com>
6218 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6221 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6224 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6228 * common.opt (gas-loc-support, gas-locview-support): New.
6229 (ginline-points, ginternal-reset-location-views): New.
6230 * doc/invoke.texi: Document them. Use @itemx where intended.
6231 (gvariable-location-views): Adjust.
6232 * target.def (reset_location_view): New.
6233 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6234 (TARGET_RESET_LOCATION_VIEW): New.
6235 * doc/tm.texi: Rebuilt.
6236 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6237 (dwarf2out_default_as_locview_support): New.
6238 (output_asm_line_debug_info): Use option variables.
6239 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6240 (output_loc_list): Likewise.
6241 (add_high_low_attributes): Check option variables.
6242 Don't output entry view attribute in strict mode.
6243 (gen_inlined_subroutine_die): Check option variables.
6244 (dwarf2out_inline_entry): Likewise.
6245 (init_sections_and_labels): Likewise.
6246 (dwarf2out_early_finish): Likewise.
6247 (maybe_reset_location_view): New, from...
6248 (dwarf2out_var_location): ... here. Call it.
6249 * debug.h (dwarf2out_default_as_loc_support): Declare.
6250 (dwarf2out_default_as_locview_support): Declare.
6251 * hooks.c (hook_int_rtx_insn_0): New.
6252 * hooks.h (hook_int_rtx_insn_0): Declare.
6253 * toplev.c (process_options): Take -gas-loc-support and
6254 -gas-locview-support from dwarf2out. Enable
6255 -gvariable-location-views by default only with locview
6256 assembler support. Enable -ginternal-reset-location-views by
6257 default only if the target defines the corresponding hook.
6258 Enable -ginline-points by default if location views are
6259 enabled; force it disabled if statement frontiers are
6261 * tree-inline.c (expand_call_inline): Check option variables.
6262 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6264 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6266 PR tree-optimization/84321
6267 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6268 handling. Also check whether the anti-range contains any values
6269 that satisfy the mask; switch to a VR_RANGE if not.
6271 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6274 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6276 2018-02-13 Martin Jambor <mjambor@suse.cz>
6279 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6280 of call statements, also set location of a load to a temporary.
6282 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6284 * config/rl78/rl78.c (add_vector_labels): New function.
6285 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6286 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6287 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6288 which checks that no arguments are passed.
6289 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6290 * doc/extend.texi: Documentation for the new attribute.
6292 2018-02-13 Andreas Schwab <schwab@suse.de>
6294 * config/riscv/linux.h (CPP_SPEC): Define.
6296 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6299 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6300 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6301 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6302 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6303 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6304 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6305 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6306 temporarily for AES and PCLMUL builtins.
6308 PR tree-optimization/84339
6309 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6310 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6314 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6315 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6316 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6318 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6320 * omp-simd-clone.h: New file.
6321 * omp-simd-clone.c: Include omp-simd-clone.h.
6322 (expand_simd_clones): No longer static.
6323 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6324 cgraph.h and omp-simd-clone.h.
6325 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6326 (vect_recog_widen_shift_pattern): Formatting fix.
6327 (vect_pattern_recog_1): Don't check optab for calls.
6330 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6331 operands[2] into a REG before using gen_lowpart on it.
6333 2018-02-12 Jeff Law <law@redhat.com>
6336 * config/sh/sh.c (find_barrier): Consider a sibling call
6339 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6340 successfully back substituting a reg.
6342 2018-02-12 Richard Biener <rguenther@suse.de>
6344 PR tree-optimization/84037
6345 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6346 parameter, move visited init to caller.
6347 (vect_slp_analyze_operations): Separate cost from validity
6348 check, initialize visited once for all instances.
6349 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6351 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6353 (vect_model_promotion_demotion_cost): Likewise.
6354 (vectorizable_bswap): Guard cost modeling with !slp_node
6355 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6357 (vectorizable_call): Likewise.
6358 (vectorizable_conversion): Likewise.
6359 (vectorizable_assignment): Likewise.
6360 (vectorizable_shift): Likewise.
6361 (vectorizable_operation): Likewise.
6362 (vectorizable_store): Likewise.
6363 (vectorizable_load): Likewise.
6364 (vectorizable_condition): Likewise.
6365 (vectorizable_comparison): Likewise.
6367 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6370 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6371 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6372 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6374 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6377 * params.def (inline-min-speedup): Increase from 8 to 15.
6378 (max-inline-insns-auto): Decrease from 40 to 30.
6379 * ipa-split.c (consider_split): Add some buffer for function to
6380 be considered inlining candidate.
6381 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6384 2018-02-12 Richard Biener <rguenther@suse.de>
6386 PR tree-optimization/84037
6387 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6388 matched stmts if we cannot swap the non-matched ones.
6390 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6392 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6393 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6394 _mm_maskz_scalef_round_ss): New intrinsics.
6395 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6396 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6397 __builtin_ia32_scalefss_round): Remove.
6398 (__builtin_ia32_scalefsd_mask_round,
6399 __builtin_ia32_scalefss_mask_round): New intrinsics.
6400 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6401 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6402 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6403 "<round_constraint>")): Changed to ...
6404 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6405 "<round_scalar_constraint>")): ... this.
6406 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6407 %0, %1, %2<round_op3>}"): Changed to ...
6408 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6409 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6410 %2<round_scalar_mask_op3>}"): ... this.
6411 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6413 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6415 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6416 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6417 (_mm_maskz_sqrt_round_ss): New intrinsics.
6418 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6419 (__builtin_ia32_sqrtsd_mask_round)
6420 (__builtin_ia32_sqrtss_mask_round): New builtins.
6421 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6422 (__builtin_ia32_sqrtss_round): Remove.
6423 (__builtin_ia32_sqrtsd_mask_round)
6424 (__builtin_ia32_sqrtss_mask_round): New builtins.
6425 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6426 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6427 ((match_operand:VF_128 1 "vector_operand"
6428 "xBm,<round_constraint>")): Changed to ...
6429 ((match_operand:VF_128 1 "vector_operand"
6430 "xBm,<round_scalar_constraint>")): ... this.
6431 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6432 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6433 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6434 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6435 %<iptr>1<round_scalar_mask_op3>}): ... this.
6436 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6437 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6439 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6442 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6443 Cast vec_cmpeq result to correct type.
6444 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6445 Cast vec_cmpgt result to correct type.
6447 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6449 * final.c (final_scan_insn_1): Renamed from...
6450 (final_scan_insn): ... this. New wrapper, to recover
6451 seen from the outermost call in recursive ones.
6452 * config/sparc/sparc.c (output_return): Drop seen from call.
6453 (output_sibcall): Likewise.
6454 * config/visium/visium.c (output_branch): Likewise.
6456 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6458 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6461 2018-02-10 Alan Modra <amodra@gmail.com>
6464 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6465 Specify LR as an input.
6467 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6470 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6471 remove_member_access_dummy_vars): New functions.
6472 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6473 lower_omp_1, execute_lower_omp): Use them.
6475 PR rtl-optimization/84308
6476 * shrink-wrap.c (spread_components): Release todo vector.
6478 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6480 PR rtl-optimization/57193
6481 * ira-color.c (struct allocno_color_data): Add member
6482 conflict_allocno_hard_prefs.
6483 (update_conflict_allocno_hard_prefs): New.
6484 (bucket_allocno_compare_func): Add a preference based on
6485 conflict_allocno_hard_prefs.
6486 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6487 (color_allocnos): Remove a dead code. Initiate
6488 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6490 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6493 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6494 constraint from =wa to wa. Avoid a subreg on the output operand,
6495 instead use a pseudo and subreg it in a move.
6496 (p9_xxbrd_<mode>): Changed to ...
6497 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6498 (p9_xxbrd_v2df): New expander.
6499 (p9_xxbrw_<mode>): Changed to ...
6500 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6501 (p9_xxbrw_v4sf): New expander.
6503 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6505 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6507 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6510 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6511 multiply in 32-bit mode.
6512 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6513 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6516 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6518 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6519 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6520 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6521 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6523 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6526 * dwarf2out.c (is_trivial_indirect_ref): New function.
6527 (dwarf2out_late_global_decl): Do not generate a location
6528 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6529 and that are not defined in the current unit.
6531 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6533 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6534 instead of a libcall for UNORDERED.
6536 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6539 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6540 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6542 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6545 * config/s390/s390.c (s390_set_current_function): Invoke
6546 s390_indirect_branch_settings also if fndecl didn't change.
6548 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6550 * config/rs6000/rs6000.md (blockage): Set length to zero.
6552 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6554 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6556 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6559 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6560 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6564 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6565 PARALLEL incoming that failed vt_get_decl_and_offset check.
6568 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6569 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6571 (get_variable_section): For decls in named .bss* sections pass true as
6572 second argument to bss_initializer_p.
6574 2018-02-09 Marek Polacek <polacek@redhat.com>
6575 Jakub Jelinek <jakub@redhat.com>
6578 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6579 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6580 Sync some changes from cxx_fold_indirect_ref.
6582 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6584 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6586 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6587 (BLOCK_INLINE_ENTRY_LABEL): New.
6588 (dwarf2out_var_location): Disregard inline entry markers.
6589 (inline_entry_data): New struct.
6590 (inline_entry_data_hasher): New hashtable type.
6591 (inline_entry_data_hasher::hash): New.
6592 (inline_entry_data_hasher::equal): New.
6593 (inline_entry_data_table): New variable.
6594 (add_high_low_attributes): Add DW_AT_entry_pc and
6595 DW_AT_GNU_entry_view attributes if a pending entry is found
6596 in inline_entry_data_table. Add old entry_pc attribute only
6597 if debug nonbinding markers are disabled.
6598 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6599 markers are enabled.
6600 (block_within_block_p, dwarf2out_inline_entry): New.
6601 (dwarf2out_finish): Check that no entries remained in
6602 inline_entry_data_table.
6603 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6604 (final_scan_insn, notice_source_line): Likewise.
6605 (rest_of_clean_state): Skip inline entry markers.
6606 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6608 * gimple.c (gimple_build_debug_inline_entry): New.
6609 * gimple.h (enum gimple_debug_subcode): Add
6610 GIMPLE_DEBUG_INLINE_ENTRY.
6611 (gimple_build_debug_inline_entry): Declare.
6612 (gimple_debug_inline_entry_p): New.
6613 (gimple_debug_nonbind_marker_p): Adjust.
6614 * insn-notes.def (INLINE_ENTRY): New.
6615 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6616 inline entry marker notes.
6617 (print_insn): Likewise.
6618 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6619 (INSN_DEBUG_MARKER_KIND): Likewise.
6620 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6621 * tree-inline.c (expand_call_inline): Build and insert
6622 debug_inline_entry stmt.
6623 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6624 inline entry blocks early, if nonbind markers are enabled.
6625 (dump_scope_block): Dump fragment info.
6626 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6627 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6628 (gimple_build_debug_inline_entry): New.
6629 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6630 Enable/disable inline entry points too.
6631 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6632 (DEBUG_INSN): Describe inline entry markers.
6634 * common.opt (gvariable-location-views): New.
6635 (gvariable-location-views=incompat5): New.
6636 * config.in: Rebuilt.
6637 * configure: Rebuilt.
6638 * configure.ac: Test assembler for view support.
6639 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6640 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6641 * dwarf2out.c (var_loc_view): New typedef.
6642 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6643 (dwarf2out_locviews_in_attribute): New.
6644 (dwarf2out_locviews_in_loclist): New.
6645 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6646 (enum dw_line_info_opcode): Add LI_adv_address.
6647 (struct dw_line_info_table): Add view.
6648 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6649 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6650 (zero_view_p): New variable.
6651 (ZERO_VIEW_P): New macro.
6652 (output_asm_line_debug_info): New.
6653 (struct var_loc_node): Add view.
6654 (add_AT_view_list, AT_loc_list): New.
6655 (add_var_loc_to_decl): Add view param. Test it against last.
6656 (new_loc_list): Add view params. Record them.
6657 (AT_loc_list_ptr): Handle loc and view lists.
6658 (view_list_to_loc_list_val_node): New.
6659 (print_dw_val): Handle dw_val_class_view_list.
6660 (size_of_die): Likewise.
6661 (value_format): Likewise.
6662 (loc_list_has_views): New.
6663 (gen_llsym): Set vl_symbol too.
6664 (maybe_gen_llsym, skip_loc_list_entry): New.
6665 (dwarf2out_maybe_output_loclist_view_pair): New.
6666 (output_loc_list): Output view list or entries too.
6667 (output_view_list_offset): New.
6668 (output_die): Handle dw_val_class_view_list.
6669 (output_dwarf_version): New.
6670 (output_compilation_unit_header): Use it.
6671 (output_skeleton_debug_sections): Likewise.
6672 (output_rnglists, output_line_info): Likewise.
6673 (output_pubnames, output_aranges): Update version comments.
6674 (output_one_line_info_table): Output view numbers in asm comments.
6675 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6676 Call maybe_gen_llsym.
6677 (loc_list_from_tree_1): Adjust.
6678 (add_AT_location_description): Create view list attribute if
6679 needed, check it's absent otherwise.
6680 (convert_cfa_to_fb_loc_list): Adjust.
6681 (maybe_emit_file): Call output_asm_line_debug_info for test.
6682 (dwarf2out_var_location): Reset views as needed. Precompute
6683 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6684 attribute. Set view.
6685 (new_line_info_table): Reset next view.
6686 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6687 (dwarf2out_source_line): Likewise. Output view resets and labels to
6688 the assembler, or select appropriate line info opcodes.
6689 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6690 (optimize_string_length): Catch it. Adjust.
6691 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6692 dw_val_class_view_list, and remove it if no longer needed.
6693 (hash_loc_list): Hash view numbers.
6694 (loc_list_hasher::equal): Compare them.
6695 (optimize_location_lists): Check whether a view list symbol is
6696 needed, and whether the locview attribute is present, and
6697 whether they match. Remove the locview attribute if no longer
6699 (index_location_lists): Call skip_loc_list_entry for test.
6700 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6701 Use output_dwarf_version.
6702 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6703 (struct dw_val_node): Add val_view_list.
6704 * final.c (SEEN_NEXT_VIEW): New.
6705 (set_next_view_needed): New.
6706 (clear_next_view_needed): New.
6707 (maybe_output_next_view): New.
6708 (final_start_function): Rename to...
6709 (final_start_function_1): ... this. Take pointer to FIRST,
6710 add SEEN parameter. Emit param bindings in the initial view.
6711 (final_start_function): Reintroduce SEEN-less interface.
6712 (final): Rename to...
6713 (final_1): ... this. Take SEEN parameter. Output final pending
6714 next view at the end.
6715 (final): Reintroduce seen-less interface.
6716 (final_scan_insn): Output pending next view before switching
6717 sections or ending a block. Mark the next view as needed when
6718 outputting variable locations. Notify debug backend of section
6719 changes, and of location view changes.
6720 (rest_of_handle_final): Adjust.
6721 * toplev.c (process_options): Autodetect value for debug variable
6722 location views option. Warn on incompat5 without -gdwarf-5.
6723 * doc/invoke.texi (gvariable-location-views): New.
6724 (gvariable-location-views=incompat5): New.
6725 (gno-variable-location-views): New.
6727 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6729 PR tree-optimization/84136
6730 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6731 that the result of find_edge is non-NULL.
6733 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6736 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6737 storing integer register in SImode. Fix cost of 256 and 512
6738 byte aligned SSE register store.
6740 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6742 * config/i386/i386.c (ix86_multiplication_cost): Fix
6743 multiplication cost for TARGET_AVX512DQ.
6745 2018-02-08 Marek Polacek <polacek@redhat.com>
6747 PR tree-optimization/84238
6748 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6751 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6753 PR tree-optimization/84265
6754 * tree-vect-stmts.c (vectorizable_store): Don't treat
6755 VMAT_CONTIGUOUS accesses as grouped.
6756 (vectorizable_load): Likewise.
6758 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6760 PR tree-optimization/81635
6761 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6762 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6763 (test_round_for_mask): New functions.
6764 (wide_int_cc_tests): Call test_round_for_mask.
6765 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6766 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6767 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6768 range returned by get_range_info.
6770 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6773 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6774 * symtab.c: Include builtins.h
6775 (symtab_node::output_to_lto_symbol_table_p): Move here
6776 from lto-streamer-out.c:output_symbol_p.
6777 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6778 (output_symbol_p): Move all logic to symtab.c
6779 (produce_symtab): Update.
6781 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6783 * config/s390/s390-opts.h (enum indirect_branch): Define.
6784 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6785 (s390_indirect_branch_via_thunk)
6786 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6787 (enum s390_indirect_branch_type): Define.
6788 * config/s390/s390.c (struct s390_frame_layout, struct
6789 machine_function): Remove.
6790 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6791 (indirect_branch_table_label_no, indirect_branch_table_name):
6793 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6794 (enum s390_indirect_branch_option): Define.
6795 (s390_return_addr_from_memory): New function.
6796 (s390_handle_string_attribute): New function.
6797 (s390_attribute_table): Add new attribute handler.
6798 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6799 (s390_indirect_branch_via_thunk): New function.
6800 (s390_indirect_branch_via_inline_thunk): New function.
6801 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6802 sibling call optimization for non z10 compiles.
6803 (s390_emit_call): Force indirect branch target to be a single
6804 register. Add r1 clobber for non-z10 compiles.
6805 (s390_emit_epilogue): Emit return jump via return_use expander.
6806 (s390_reorg): Handle JUMP_INSNs as execute targets.
6807 (s390_option_override_internal): Perform validity checks for the
6808 new command line options.
6809 (s390_indirect_branch_attrvalue): New function.
6810 (s390_indirect_branch_settings): New function.
6811 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6812 (s390_output_indirect_thunk_function): New function.
6813 (s390_code_end): Implement target hook.
6814 (s390_case_values_threshold): Implement target hook.
6815 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6817 * config/s390/s390.h (struct s390_frame_layout)
6818 (struct machine_function): Move here from s390.c.
6819 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6820 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6821 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6822 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6823 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6824 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6825 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6826 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6827 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6828 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6829 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6830 (mnemonic attribute): Add values which aren't recognized
6832 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6833 pattern for branch conversion. Fix mnemonic attribute.
6834 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6835 indirect branch via thunk if requested.
6836 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6837 ("*indirect_jump"): Disable for branch conversion using out of
6839 ("indirect_jump_via_thunk<mode>_z10")
6840 ("indirect_jump_via_thunk<mode>")
6841 ("indirect_jump_via_inlinethunk<mode>_z10")
6842 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6843 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6844 ("casesi_jump_via_inlinethunk<mode>_z10")
6845 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6846 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6847 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6848 ("*indirect2_jump"): Disable for branch conversion.
6849 ("casesi_jump"): Turn into expander and expand patterns for branch
6851 ("return_use"): New expander.
6852 ("*return"): Emit return via thunk and rename it to ...
6853 ("*return<mode>"): ... this one.
6854 * config/s390/s390.opt: Add new options and and enum for the
6857 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6859 * lra-constraints.c (match_reload): Unconditionally use
6860 gen_lowpart_SUBREG, rather than selecting between that
6861 and equivalent gen_rtx_SUBREG code.
6863 2018-02-08 Richard Biener <rguenther@suse.de>
6865 PR tree-optimization/84233
6866 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6867 changed flag instead of boguously re-using phi_inserted.
6869 2018-02-08 Martin Jambor <mjambor@suse.cz>
6871 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6872 static local variables.
6874 2018-02-08 Richard Biener <rguenther@suse.de>
6876 PR tree-optimization/84278
6877 * tree-vect-stmts.c (vectorizable_store): When looking for
6878 smaller vector types to perform grouped strided loads/stores
6879 make sure the mode is supported by the target.
6880 (vectorizable_load): Likewise.
6882 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6884 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6885 Increase LDP/STP opportunities by adding adjacent callee-saves.
6887 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6889 PR rtl-optimization/84068
6890 PR rtl-optimization/83459
6891 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6893 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6895 PR tree-optimization/84224
6896 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6897 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6900 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6903 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6904 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6905 count, remove one USE.
6907 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6909 * doc/install.texi (Configuration): Document the
6910 --with-long-double-format={ibm,ieee} PowerPC configuration
6914 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6915 Convert from define_expand to be define_insn_and_split. Rework
6916 float/double/_Float128 conversions to QI/HI/SImode to work with
6917 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6918 conversions to QI/HImode types did a store and then a load to
6919 truncate the value. For conversions to VSX registers, don't split
6920 the insn, instead emit the code directly. Use the code iterator
6921 any_fix to combine signed and unsigned conversions.
6922 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6923 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6924 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6925 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6926 (fix_<mode>di2_hw): Likewise.
6927 (fixuns_<mode>di2_hw): Likewise.
6928 (fix_<mode>si2_hw): Likewise.
6929 (fixuns_<mode>si2_hw): Likewise.
6930 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6931 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6932 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6933 fix<uns>_trunc<SFDF:mode>si2_p8.
6934 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6936 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6937 (fix<uns>_<mode>_mem): Likewise.
6938 (fctiw<u>z_<mode>_mem): Likewise.
6939 (fix<uns>_<mode>_mem): Likewise.
6940 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6941 the register allocator from doing a direct move to the GPRs to do
6942 a store, and instead use the ISA 3.0 store byte/half-word from
6943 vector register instruction. For IEEE 128-bit floating point,
6944 also optimize stores of 32-bit ints.
6945 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6947 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6949 * genextract.c (push_pathstr_operand): New function to support
6951 (walk_rtx): Call push_pathstr_operand.
6952 (print_path): Support [a-zA-Z].
6954 2018-02-07 Richard Biener <rguenther@suse.de>
6956 PR tree-optimization/84037
6957 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6958 (cse_and_gimplify_to_preheader): Declare.
6959 (vect_get_place_in_interleaving_chain): Likewise.
6960 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6962 (_loop_vec_info::~_loop_vec_info): Delete it.
6963 (cse_and_gimplify_to_preheader): New function.
6964 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6965 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6966 (vectorizable_load): Likewise. For grouped stores always base
6967 the IV on the first element.
6968 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6969 condition before gimplifying.
6971 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6973 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6974 *DIV_EXPR and *MOD_EXPR.
6976 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6979 * config/i386/i386.c (ix86_option_override_internal): Mask out
6980 the CF_SET bit when checking -fcf-protection.
6982 2018-02-07 Tom de Vries <tom@codesourcery.com>
6985 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6988 2018-02-07 Richard Biener <rguenther@suse.de>
6990 PR tree-optimization/84204
6991 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6994 PR tree-optimization/84205
6995 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6996 special-case isl_ast_op_zdiv_r.
6998 PR tree-optimization/84223
6999 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7000 Only add conditions from within the region.
7001 (gather_bbs::after_dom_children): Adjust.
7003 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7006 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7007 * config/avr/avr.md: Only post-reload split REG-REG moves if
7008 either register is GENERAL_REG_P.
7010 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7012 PR tree-optimization/84235
7013 * tree-ssa-scopedtables.c
7014 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7015 if the subtraction is performed in floating point type where NaNs are
7016 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7017 build 1. Formatting fix.
7019 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7022 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7023 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7024 and skip it regardless of bb boundaries. Use CALL_P macro,
7025 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7026 unnecessarily, formatting fix.
7028 2018-02-06 Michael Collison <michael.collison@arm.com>
7030 * config/arm/thumb2.md:
7031 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7032 (*thumb_mov_notscc): Ditto.
7034 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7037 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7038 unsigned_fix, not "s".
7040 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7042 * configure.ac (gcc_fn_eh_frame_ro): New function.
7043 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7044 correct .eh_frame permissions.
7045 * configure: Regenerate.
7047 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7049 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7052 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7054 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7055 Display warning message for -mno-speculate-indirect-jumps.
7057 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7059 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7061 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7063 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7065 PR tree-optimization/84225
7066 * tree-eh.c (find_trapping_overflow): Only call
7067 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7069 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7072 * config/i386/i386.c: Reimplement the check of possible options
7073 -mibt/-mshstk conbination. Change error messages.
7074 * doc/invoke.texi: Fix a typo: remove extra '='.
7076 2018-02-06 Marek Polacek <polacek@redhat.com>
7078 PR tree-optimization/84228
7079 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7081 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7084 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7085 emitted arch directives.
7086 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7087 __ARM_FEATURE_COPROC before changing architectures.
7089 2018-02-06 Richard Biener <rguenther@suse.de>
7091 * config/i386/i386.c (print_reg): Fix typo.
7092 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7094 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7096 * configure: Regenerate.
7098 2018-02-05 Martin Sebor <msebor@redhat.com>
7100 PR tree-optimization/83369
7101 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7104 2018-02-05 Martin Liska <mliska@suse.cz>
7106 * doc/invoke.texi: Cherry-pick upstream r323995.
7108 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7110 * ira.c (ira_init_register_move_cost): Adjust comment.
7112 2018-02-05 Martin Liska <mliska@suse.cz>
7114 PR gcov-profile/84137
7115 * doc/gcov.texi: Fix typo in documentation.
7117 2018-02-05 Martin Liska <mliska@suse.cz>
7119 PR gcov-profile/83879
7120 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7121 using dlopen functionality.
7123 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7125 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7126 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7127 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7128 _mm_maskz_range_round_ss): New intrinsics.
7129 (__builtin_ia32_rangesd128_round)
7130 (__builtin_ia32_rangess128_round): Remove.
7131 (__builtin_ia32_rangesd128_mask_round,
7132 __builtin_ia32_rangess128_mask_round): New builtins.
7133 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7134 __builtin_ia32_rangess128_round): Remove.
7135 (__builtin_ia32_rangesd128_mask_round,
7136 __builtin_ia32_rangess128_mask_round): New builtins.
7137 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7138 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7139 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7140 "<round_saeonly_constraint>")): Changed to ...
7141 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7142 "<round_saeonly_scalar_constraint>")): ... this.
7143 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7144 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7145 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7146 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7147 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7149 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7151 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7153 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7154 Remove all values except native, 8540 and 8548.
7156 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7158 * config/i386/i386.c (ix86_output_function_return): Pass
7159 INVALID_REGNUM, instead of -1, as invalid register number to
7160 indirect_thunk_name and output_indirect_thunk.
7162 2018-02-02 Julia Koval <julia.koval@intel.com>
7164 * config.gcc: Add -march=icelake.
7165 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7166 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7167 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7168 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7169 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7170 (processor_target_table): Add icelake.
7171 (ix86_option_override_internal): Handle new PTAs.
7172 (get_builtin_code_for_version): Handle icelake.
7173 (M_INTEL_COREI7_ICELAKE): New.
7174 (fold_builtin_cpu): Handle icelake.
7175 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7176 * doc/invoke.texi: Add -march=icelake.
7178 2018-02-02 Julia Koval <julia.koval@intel.com>
7180 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7181 to wide_int_bitmask.
7182 * wide-int-bitmask.h: New.
7184 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7187 * config/i386/i386.md: Replace Pmode with word_mode in
7188 builtin_setjmp_setup and builtin_longjmp to support x32.
7190 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7194 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7196 (rs6000_supported_cpu_names): New static variable.
7197 (linux_cpu_translation_table): Likewise.
7198 (elf_platform) <cpu>: Define new static variable and use it.
7199 Translate kernel AT_PLATFORM name to canonical name if needed.
7200 Error if platform name is unknown.
7202 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7205 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7207 2018-02-01 Jeff Law <law@redhat.com>
7210 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7211 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7212 the scratch if RELEASE_VIA_POP is false.
7213 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7214 If we have to save a temporary register, decrement SIZE appropriately.
7215 Pass new arguments to release_scratch_register_on_entry.
7216 (ix86_adjust_stack_and_probe): Likewise.
7217 (ix86_emit_probe_stack_range): Pass new arguments to
7218 release_scratch_register_on_entry.
7220 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7222 PR rtl-optimization/84157
7223 * combine.c (change_zero_ext): Use REG_P predicate in
7224 front of HARD_REGISTER_P predicate.
7226 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7228 * config/avr/avr.c (avr_option_override): Move disabling of
7229 -fdelete-null-pointer-checks to...
7230 * common/config/avr/avr-common.c (avr_option_optimization_table):
7233 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7235 PR tree-optimization/81635
7236 * tree-data-ref.c (split_constant_offset_1): For types that
7237 wrap on overflow, try to use range info to prove that wrapping
7240 2018-02-01 Renlin Li <renlin.li@arm.com>
7243 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7245 (aarch64_register_move_cost): Likewise.
7246 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7248 (REG_CLASS_NAMES): Likewise.
7249 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7250 TAILCALL_ADDR_REGS. Remove IP registers.
7251 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7253 2018-02-01 Richard Biener <rguenther@suse.de>
7255 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7256 for specifying RPO order and allow NULL for that.
7257 * domwalk.c (dom_walker::dom_walker): Likewise.
7258 (dom_walker::walk): Handle NULL RPO order.
7259 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7261 (rewrite_update_dom_walker): Likewise.
7262 (mark_def_dom_walker): Likewise.
7264 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7266 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7267 (aarch64_maybe_expand_sve_subreg_move): Declare.
7268 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7269 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7271 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7272 that are semantically a reverse operation.
7273 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7274 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7275 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7277 (aarch64_can_change_mode_class): For big-endian, forbid changes
7278 between two SVE modes if they have different element sizes.
7280 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7282 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7283 the TImode handling for big-endian targets.
7285 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7287 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7288 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7290 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7291 Remove BSWAP handing for big-endian targets and use the form of
7292 LD1RQ appropariate for the mode.
7294 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7296 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7297 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7300 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7303 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7304 check for operands that need to go through aarch64_sve_reload_be.
7306 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7308 PR tree-optimization/81661
7309 PR tree-optimization/84117
7310 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7311 * tree-eh.c: Include gimplify.h.
7312 (find_trapping_overflow, replace_trapping_overflow,
7313 rewrite_to_non_trapping_overflow): New functions.
7314 * tree-vect-loop.c: Include tree-eh.h.
7315 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7316 * tree-data-ref.c: Include tree-eh.h.
7317 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7319 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7321 PR rtl-optimization/84123
7322 * combine.c (change_zero_ext): Check if hard register satisfies
7323 can_change_dest_mode before calling gen_lowpart_SUBREG.
7325 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7328 * ira.c (ira_init_register_move_cost): Remove assert.
7330 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7332 PR rtl-optimization/84071
7333 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7334 * doc/tm.texi: Regenerate.
7336 2018-01-31 Richard Biener <rguenther@suse.de>
7338 PR tree-optimization/84132
7339 * tree-data-ref.c (analyze_miv_subscript): Properly
7340 check whether evolution_function_is_affine_multivariate_p
7341 before calling gcd_of_steps_may_divide_p.
7343 2018-01-31 Julia Koval <julia.koval@intel.com>
7346 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7347 * config/i386/i386.md (rdpid_rex64) New.
7348 (rdpid): Make 32bit only.
7350 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7353 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7354 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7356 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7359 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7361 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7363 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7365 PR rtl-optimization/84071
7366 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7367 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7369 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7371 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7372 (arc_attribute_table): Add 'aux' attribute.
7373 (arc_in_small_data_p): Consider aux like variables.
7374 (arc_is_aux_reg_p): New function.
7375 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7376 (arc_get_aux_arg): New function.
7377 (prepare_move_operands): Handle aux-register access.
7378 (arc_handle_aux_attribute): New function.
7379 * doc/extend.texi (ARC Variable attributes): Add subsection.
7381 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7383 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7384 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7385 (arc_attribute_table): Add 'uncached' attribute.
7386 (arc_print_operand): Print '.di' flag for uncached memory
7388 (arc_in_small_data_p): Do not consider for small data the uncached
7390 (arc_is_uncached_mem_p): New function.
7391 * config/arc/predicates.md (compact_store_memory_operand): Check
7392 for uncached memory accesses.
7393 (nonvol_nonimm_operand): Likewise.
7394 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7396 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7399 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7400 falign-loops=): Add Optimization flag.
7402 2018-01-30 Jeff Law <law@redhat.com>
7405 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7406 INT_REGISTERS_SAVED. Check it prior to calling
7407 get_scratch_register_on_entry.
7408 (ix86_adjust_stack_and_probe): Similarly.
7409 (ix86_emit_probe_stack_range): Similarly.
7410 (ix86_expand_prologue): Corresponding changes.
7412 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7415 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7416 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7418 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7421 * lra-constraints.c (curr_insn_transform): Process AND in the
7424 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7426 PR rtl-optimization/83986
7427 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7428 dependence against last_pending_memory_flush in addition to
7431 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7433 PR tree-optimization/81611
7434 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7437 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7440 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7443 2018-01-30 Richard Biener <rguenther@suse.de>
7444 Jakub Jelinek <jakub@redhat.com>
7446 PR tree-optimization/84111
7447 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7448 inner loops added during recursion, as they don't have up-to-date
7451 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7454 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7455 (can_inline_edge_by_limits_p): ... here.
7456 (can_early_inline_edge_p, check_callers,
7457 update_caller_keys, update_callee_keys, recursive_inlining,
7458 add_new_edges_to_heap, speculation_useful_p,
7459 inline_small_functions,
7460 inline_small_functions, flatten_function,
7461 inline_to_all_callers_1): Update.
7463 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7465 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7468 2018-01-30 Richard Biener <rguenther@suse.de>
7470 PR tree-optimization/83008
7471 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7472 invariant and constant vector uses in stmts when they need
7475 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7478 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7479 * configure: Regenerate.
7481 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7483 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7485 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7486 Use gen_rtx_REG rather than gen_lowpart.
7488 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7490 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7491 rather than 0 when creating partial subregs.
7493 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7495 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7498 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7501 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7502 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7503 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7504 flags. This restores the settings used before the 2017-07-24.
7505 Turning off pre increment/decrement/modify allows IVOPTS to
7506 optimize DF/SF loops where the index is an int.
7508 2018-01-29 Richard Biener <rguenther@suse.de>
7509 Kelvin Nilsen <kelvin@gcc.gnu.org>
7512 * tree-vect-stmts.c (vectorizable_call): Don't call
7513 targetm.vectorize_builtin_md_vectorized_function if callee is
7516 2018-01-22 Carl Love <cel@us.ibm.com>
7518 * doc/extend.tex: Fix typo in second arg in
7519 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7521 2018-01-29 Richard Biener <rguenther@suse.de>
7523 PR tree-optimization/84086
7524 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7525 (flush_ssaname_freelist): When SSA names were released reset
7526 the SCEV hash table.
7528 2018-01-29 Richard Biener <rguenther@suse.de>
7530 PR tree-optimization/84057
7531 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7532 removed paths when removing edges.
7534 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7536 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7537 -mfunction-return=@var{choice}.
7539 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7542 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7544 (layout::print_source_line): Likewise.
7545 (test_get_line_width_without_trailing_whitespace): Add test cases.
7547 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7550 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7553 2018-01-26 Jim Wilson <jimw@sifive.com>
7555 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7557 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7560 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7562 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7563 and CMP + SUB-immediate -> SUBS.
7565 2018-01-26 Martin Sebor <msebor@redhat.com>
7567 PR tree-optimization/83896
7568 * tree-ssa-strlen.c (get_string_len): Rename...
7569 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7570 Avoid assuming length is constant.
7571 (handle_char_store): Use HOST_WIDE_INT for string length.
7573 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7576 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7577 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7579 2018-01-26 Richard Biener <rguenther@suse.de>
7581 PR rtl-optimization/84003
7582 * dse.c (record_store): Only record redundant stores when
7583 the earlier store aliases at least all accesses the later one does.
7585 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7587 PR rtl-optimization/83985
7588 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7589 REG_CFA_RESTORE insns.
7590 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7591 REG_CFA_RESTORE insns here.
7594 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7595 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7597 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7599 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7600 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7601 (arc_init): Likewise.
7602 (arc_override_options): Likewise.
7603 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7605 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7607 * config/arc/arc.h (TARGET_DBNZ): Define.
7608 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7609 properly set the tune attribute.
7610 (dbnz): Use TARGET_DBNZ guard.
7611 * config/arc/arc.opt (mtune): Add core3 option.
7613 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7615 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7616 recognize new pic like addresses.
7617 (arc_delegitimize_address): Clean up.
7619 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7621 * config/arc/arc-arches.def: Option mrf16 valid for all
7623 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7624 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7625 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7626 * config/arc/arc-tables.opt: Regenerate.
7627 * config/arc/arc.c (arc_conditional_register_usage): Handle
7628 reduced register file case.
7629 (arc_file_start): Set must have build attributes.
7630 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7632 * config/arc/arc.opt (mrf16): Add new option.
7633 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7634 * config/arc/genmultilib.awk: Handle new mrf16 option.
7635 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7636 * config/arc/t-multilib: Regenerate.
7637 * doc/invoke.texi (ARC Options): Document mrf16 option.
7639 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7641 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7642 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7643 (arc_attribute_table): Add 'secure_call' attribute.
7644 (arc_print_operand): Print secure call operand.
7645 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7647 (arc_is_secure_call_p): New function. * config/arc/arc.md
7648 (call_i): Add support for sjli instruction.
7649 (call_value_i): Likewise.
7650 * config/arc/constraints.md (Csc): New constraint.
7652 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7653 John Eric Martin <John.Martin@emmicro-us.com>
7655 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7656 * config/arc/arc.c (_arc_jli_section): New struct.
7657 (arc_jli_section): New type.
7658 (rc_jli_sections): New static variable.
7659 (arc_handle_jli_attribute): New function.
7660 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7661 (arc_file_end): New function.
7662 (TARGET_ASM_FILE_END): Define.
7663 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7664 (arc_add_jli_section): New function.
7665 (jli_call_scan): Likewise.
7666 (arc_reorg): Call jli_call_scan.
7667 (arc_output_addsi): Remove 'S' from printing asm operand.
7668 (arc_is_jli_call_p): New function.
7669 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7671 (movhi_insn): Likewise.
7672 (movsi_insn): Likewise.
7673 (movsi_set_cc_insn): Likewise.
7674 (loadqi_update): Likewise.
7675 (load_zeroextendqisi_update): Likewise.
7676 (load_signextendqisi_update): Likewise.
7677 (loadhi_update): Likewise.
7678 (load_zeroextendhisi_update): Likewise.
7679 (load_signextendhisi_update): Likewise.
7680 (loadsi_update): Likewise.
7681 (loadsf_update): Likewise.
7682 (movsicc_insn): Likewise.
7683 (bset_insn): Likewise.
7684 (bxor_insn): Likewise.
7685 (bclr_insn): Likewise.
7686 (bmsk_insn): Likewise.
7687 (bicsi3_insn): Likewise.
7688 (cmpsi_cc_c_insn): Likewise.
7689 (movsi_ne): Likewise.
7690 (movsi_cond_exec): Likewise.
7691 (clrsbsi2): Likewise.
7702 (call_i): Remove 'S' asm letter, add jli instruction.
7703 (call_value_i): Likewise.
7704 * config/arc/arc.op (mjli-always): New option.
7705 * config/arc/constraints.md (Cji): New constraint.
7706 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7708 (subsf3_fpx): Likewise.
7709 (mulsf3_fpx): Likewise.
7710 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7712 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7714 * doc/invoke.texi (ARC): Document mjli-always option.
7716 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7718 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7719 avoid addition with 0 and use incw and decw where possible.
7721 2018-01-26 Richard Biener <rguenther@suse.de>
7723 PR tree-optimization/81082
7724 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7725 association if it requires casting to unsigned.
7726 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7727 from fold_plusminus_mult_expr to catch important cases late when
7728 range info is available.
7730 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7732 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7733 * configure.ac (hidden_linkonce): New test.
7734 * configure: Regenerate.
7735 * config.in: Regenerate.
7737 2018-01-26 Julia Koval <julia.koval@intel.com>
7739 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7740 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7741 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7742 _mm_mask_bitshuffle_epi64_mask): Fix type.
7743 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7744 USI_FTYPE_V4DI_V4DI_USI): Remove.
7745 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7746 __builtin_ia32_vpshufbitqmb256_mask,
7747 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7748 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7749 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7751 2018-01-26 Alan Modra <amodra@gmail.com>
7754 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7755 UNSPEC_VBPERMQ. Sort other unspecs.
7757 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7759 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7761 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7764 * predict.c (drop_profile): Do not push/pop cfun; update also
7766 (handle_missing_profiles): Fix logic looking for zero profiles.
7768 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7771 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7772 on functions with #pragma omp declare simd or functions with simd
7774 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7775 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7776 Remove trailing \n from warning_at calls.
7778 2018-01-25 Tom de Vries <tom@codesourcery.com>
7781 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7782 for neutered workers.
7784 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7787 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7788 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7790 2018-01-24 Jeff Law <law@redhat.com>
7793 * i386.c (get_probe_interval): Move to earlier point.
7794 (ix86_compute_frame_layout): If -fstack-clash-protection and
7795 the frame is larger than the probe interval, then use pushes
7796 to save registers rather than reg->mem moves.
7797 (ix86_expand_prologue): Remove conditional for int_registers_saved
7800 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7803 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7804 min/max for never referenced object.
7806 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7809 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7811 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7812 attributes from DECL_ATTRIBUTES (decl) without affecting
7813 DECL_ATTRIBUTES (current_function_decl).
7814 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7815 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7817 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7819 PR tree-optimization/83979
7820 * fold-const.c (fold_comparison): Use constant_boolean_node
7821 instead of boolean_{true,false}_node.
7823 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7825 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7828 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7830 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7831 Simplify the clause that sets the length attribute.
7832 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7833 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7834 clause that sets the length attribute.
7835 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7837 2018-01-24 Tom de Vries <tom@codesourcery.com>
7840 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7841 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7842 Add strict parameter.
7843 (prevent_branch_around_nothing): Insert dummy insn between branch to
7844 label and label with no ptx insn inbetween.
7845 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7847 2018-01-24 Tom de Vries <tom@codesourcery.com>
7850 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7851 for neutered threads in warp.
7852 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7854 2018-01-24 Richard Biener <rguenther@suse.de>
7856 PR tree-optimization/83176
7857 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7860 2018-01-24 Richard Biener <rguenther@suse.de>
7862 PR tree-optimization/82819
7863 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7864 code generating pluses that are no-ops in the target precision.
7866 2018-01-24 Richard Biener <rguenther@suse.de>
7869 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7871 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7873 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7874 to merge probabilities.
7875 * predict.c (probably_never_executed): Also mark as cold functions
7876 with global 0 profile and guessed local profile.
7877 * profile-count.c (profile_probability::combine_with_count): New
7879 * profile-count.h (profile_probability::operator*,
7880 profile_probability::operator*=, profile_probability::operator/,
7881 profile_probability::operator/=): Reduce precision to adjusted
7882 and set value to guessed on contradictory divisions.
7883 (profile_probability::combine_with_freq): Remove.
7884 (profile_probability::combine_wiht_count): Declare.
7885 (profile_count::force_nonzero):: Set to adjusted.
7886 (profile_count::probability_in):: Set quality to adjusted.
7887 * tree-ssa-tail-merge.c (replace_block_by): Use
7890 2018-01-23 Andrew Waterman <andrew@sifive.com>
7891 Jim Wilson <jimw@sifive.com>
7893 * config/riscv/riscv.c (riscv_stack_boundary): New.
7894 (riscv_option_override): Set riscv_stack_boundary. Handle
7895 riscv_preferred_stack_boundary_arg.
7896 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7897 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7898 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7899 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7900 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7901 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7903 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7906 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7907 of struct ix86_frame.
7908 (ix86_expand_epilogue): Likewise. Add a local variable for
7909 the reg_save_offset field in struct ix86_frame.
7911 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7913 PR tree-optimization/82604
7914 * tree-loop-distribution.c (enum partition_kind): New enum item
7915 PKIND_PARTIAL_MEMSET.
7916 (partition_builtin_p): Support above new enum item.
7917 (generate_code_for_partition): Ditto.
7918 (compute_access_range): Differentiate cases that equality can be
7919 proven at all loops, the innermost loops or no loops.
7920 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7921 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7922 (finalize_partitions, distribute_loop): Don't fuse partition of
7923 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7924 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7927 2018-01-23 Martin Liska <mliska@suse.cz>
7929 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7930 order to ignore the predictor.
7931 (PRED_POLYMORPHIC_CALL): Likewise.
7932 (PRED_RECURSIVE_CALL): Likewise.
7934 2018-01-23 Martin Liska <mliska@suse.cz>
7936 * tree-profile.c (tree_profiling): Print function header to
7937 aware reader which function we are working on.
7938 * value-prof.c (gimple_find_values_to_profile): Do not print
7939 not interesting value histograms.
7941 2018-01-23 Martin Liska <mliska@suse.cz>
7943 * profile-count.h (enum profile_quality): Add
7944 profile_uninitialized as the first value. Do not number values
7945 as they are zero based.
7946 (profile_count::verify): Update sanity check.
7947 (profile_probability::verify): Likewise.
7949 2018-01-23 Nathan Sidwell <nathan@acm.org>
7951 * doc/invoke.texi (ffor-scope): Deprecate.
7953 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7955 PR tree-optimization/83510
7956 * domwalk.c (set_all_edges_as_executable): New function.
7957 (dom_walker::dom_walker): Convert bool param
7958 "skip_unreachable_blocks" to enum reachability. Move setup of
7959 edge flags to set_all_edges_as_executable and only do it when
7960 reachability is REACHABLE_BLOCKS.
7961 * domwalk.h (enum dom_walker::reachability): New enum.
7962 (dom_walker::dom_walker): Convert bool param
7963 "skip_unreachable_blocks" to enum reachability.
7964 (set_all_edges_as_executable): New decl.
7965 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7966 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7968 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7969 but converting true to REACHABLE_BLOCKS.
7970 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7972 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7973 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7974 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7976 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7977 if check_all_array_refs will be called.
7979 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7981 * tree.c (selftest::test_location_wrappers): Add more test
7984 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7986 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7987 (selftest::test_bit_in_range): Likewise.
7989 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7992 * doc/sourcebuild.texi (vect_float): Say that the selector
7993 only describes the situation when -funsafe-math-optimizations is on.
7994 (vect_float_strict): Document.
7996 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7998 PR tree-optimization/83965
7999 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8000 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8001 instead of checking only for a reduction.
8002 (vect_recog_widen_sum_pattern): Likewise.
8004 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8006 * predict.c (probably_never_executed): Only use precise profile info.
8007 (compute_function_frequency): Skip after inlining hack since we now
8008 have quality checking.
8010 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8012 * profile-count.h (profile_probability::very_unlikely,
8013 profile_probability::unlikely, profile_probability::even): Set
8014 precision to guessed.
8016 2018-01-23 Richard Biener <rguenther@suse.de>
8018 PR tree-optimization/83963
8019 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8020 Properly terminate dominator walk when crossing the exit edge not
8021 when visiting its source block.
8023 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8026 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8027 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8029 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8031 PR tree-optimization/83957
8032 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8033 semicolon after for body surrounded by braces.
8035 PR tree-optimization/83081
8036 * profile-count.h (profile_probability::split): New method.
8037 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8038 Use profile_probability::split.
8039 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8040 when splitting a single conditional jump into 2.
8042 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8044 PR tree-optimization/69452
8045 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8048 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8050 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8051 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8052 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8054 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8056 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8057 * config/rl78/rl78.md (movdi): New define_expand.
8058 * config/rl78/rl78.c (rl78_split_movdi): New function.
8060 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8063 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8065 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8066 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8067 128-bit to produce an UNSPEC move to get the double word with the
8068 signbit and then a shift directly to do signbit.
8069 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8070 implementation with a new version that just does either a direct
8071 move or a regular move. Move memory interface to separate insns.
8072 Move insns so they are next to the expander.
8073 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8074 with signbit move. Split big and little endian case.
8075 (signbit<mode>2_dm_mem_le): Likewise.
8076 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8077 (signbit<mode>2_dm2): Likewise.
8079 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8081 * config/rl78/rl78.md (anddi3): New define_expand.
8083 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8085 * config/rl78/rl78.md (umindi3): New define_expand.
8087 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8089 * config/rl78/rl78.md (smindi3): New define_expand.
8091 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8093 * config/rl78/rl78.md (smaxdi3): New define_expand.
8095 2018-01-22 Carl Love <cel@us.ibm.com>
8097 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8098 LVX_V1TI): Add macro expansion.
8099 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8100 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8101 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8102 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8103 Change check to determine if the instruction is a byte reversing
8104 entry. Fix typo in comment.
8105 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8106 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8107 Add def_builtin calls for new builtins.
8108 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8109 Add define_insn expansion.
8111 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8113 * config/rl78/rl78.md (umaxdi3): New define_expand.
8115 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8117 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8118 for non-QImode registers.
8120 2018-01-22 Richard Biener <rguenther@suse.de>
8122 PR tree-optimization/83963
8123 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8124 including the loop exit block.
8125 (scop_detection::merge_sese): Likewise.
8126 (scop_detection::add_scop): Do it here instead.
8128 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8130 * doc/sourcebuild.texi (arm_softfloat): Document.
8132 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8135 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8136 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8137 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8139 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8140 David Edelsohn <dje.gcc@gmail.com>
8143 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8144 Change "crset eq" to "crset 2".
8145 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8146 (*call_indirect_aix<mode>_nospec): Likewise.
8147 (*call_value_indirect_aix<mode>_nospec): Likewise.
8148 (*call_indirect_elfv2<mode>_nospec): Likewise.
8149 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8150 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8151 change assembly output from . to $.
8152 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8153 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8154 (*tablejump<mode>_internal1_nospec): Likewise.
8156 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8159 * config/sh/sh_optimize_sett_clrt.cc:
8160 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8162 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8164 PR tree-optimization/83940
8165 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8166 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8167 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8168 use it to pass back the definition type.
8169 (vect_check_store_rhs): Likewise rhs_dt_out.
8170 (vect_build_gather_load_calls): Add a mask_dt argument and use
8171 it instead of a call to vect_is_simple_use.
8172 (vectorizable_store): Update calls to vect_check_load_store_mask
8173 and vect_check_store_rhs. Use the dt returned by the latter instead
8174 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8175 instead of calls to vect_is_simple_use. Pass the scalar rather
8176 than the vector operand to vect_is_simple_use when handling
8177 second and subsequent copies of an rhs value.
8178 (vectorizable_load): Update calls to vect_check_load_store_mask
8179 and vect_build_gather_load_calls. Use the cached mask_dt and
8180 gs_info.offset_dt instead of calls to vect_is_simple_use.
8182 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8185 * tree-emutls.c: Include gimplify.h.
8186 (lower_emutls_2): New function.
8187 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8188 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8189 it before further processing.
8192 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8193 UINTVAL (trueop1) instead of INTVAL (op1).
8195 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8199 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8200 INCOMING_FRAME_SP_OFFSET if not defined.
8201 (scan_trace): Add ENTRY argument. If true and
8202 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8203 emit a note to adjust the CFA offset.
8204 (create_cfi_notes): Adjust scan_trace callers.
8205 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8206 INCOMING_FRAME_SP_OFFSET in the CIE.
8207 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8208 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8210 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8211 * doc/tm.texi: Regenerated.
8213 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8215 PR rtl-optimization/83147
8216 * lra-constraints.c (remove_inheritance_pseudos): Use
8217 lra_substitute_pseudo_within_insn.
8219 2018-01-19 Tom de Vries <tom@codesourcery.com>
8220 Cesar Philippidis <cesar@codesourcery.com>
8223 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8225 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8228 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8229 spaces for function labels.
8231 2018-01-19 Martin Liska <mliska@suse.cz>
8233 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8234 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8235 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8236 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8237 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8238 (PRED_CONST_RETURN): Change from 69 to 65.
8239 (PRED_NULL_RETURN): Change from 91 to 71.
8240 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8241 (PRED_LOOP_GUARD): Change from 66 to 73.
8243 2018-01-19 Martin Liska <mliska@suse.cz>
8245 * predict.c (predict_insn_def): Add new assert.
8246 (struct branch_predictor): Change type to signed integer.
8247 (test_prediction_value_range): Amend test to cover
8249 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8250 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8251 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8252 (PRED_LOOP_IV_COMPARE): Likewise.
8253 * predict.h (PROB_UNINITIALIZED): Define new constant.
8255 2018-01-19 Martin Liska <mliska@suse.cz>
8257 * predict.c (dump_prediction): Add new format for
8258 analyze_brprob.py script which is enabled with -details
8260 * profile-count.h (precise_p): New function.
8262 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8264 PR tree-optimization/83922
8265 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8266 there are no statements that need masking.
8267 (vect_active_double_reduction_p): New function.
8268 (vect_analyze_loop_operations): Use it when handling phis that
8269 are not in the loop header.
8271 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8273 PR tree-optimization/83914
8274 * tree-vect-loop.c (vectorizable_induction): Don't convert
8275 init_expr or apply the peeling adjustment for inductions
8276 that are nested within the vectorized loop.
8278 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8280 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8283 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8287 * function.h (gimplify_parameters): Add gimple_seq * argument.
8288 * function.c: Include gimple.h and options.h.
8289 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8290 for the added local temporaries if needed.
8291 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8292 if there are any parameter cleanups, wrap whole body into a
8293 try/finally with the cleanups.
8295 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8298 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8299 Use GET_MODE_CLASS for scalar floating point.
8301 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8305 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8306 Fix call of call_cgraph_insertion_hooks.
8308 2018-01-18 Martin Sebor <msebor@redhat.com>
8310 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8312 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8315 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8318 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8321 * common.opt: (-ffile-prefix-map): New option.
8322 * opts.c (common_handle_option): Defer it.
8323 * opts-global.c (handle_common_deferred_options): Handle it.
8324 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8325 * file-prefix-map.h: New file.
8326 (remap_debug_filename, add_debug_prefix_map): ...here.
8327 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8328 * final.c (debug_prefix_map, add_debug_prefix_map
8329 remap_debug_filename): Move to...
8330 * file-prefix-map.c: New file.
8331 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8332 generalize, get rid of alloca(), use strrchr() instead of strchr().
8333 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8334 Implement in terms of add_prefix_map().
8335 (remap_macro_filename, remap_debug_filename): Implement in term of
8337 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8338 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8339 * dbxout.c: Include file-prefix-map.h.
8340 * varasm.c: Likewise.
8341 * vmsdbgout.c: Likewise.
8342 * xcoffout.c: Likewise.
8343 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8344 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8345 * doc/invoke.texi (-ffile-prefix-map): Document.
8346 (-fdebug-prefix-map): Update description.
8348 2018-01-18 Martin Liska <mliska@suse.cz>
8350 * config/i386/i386.c (indirect_thunk_name): Document that also
8352 (output_indirect_thunk): Document why both instructions
8353 (pause and lfence) are generated.
8355 2018-01-18 Richard Biener <rguenther@suse.de>
8357 PR tree-optimization/83887
8358 * graphite-scop-detection.c
8359 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8360 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8361 (scop_detection::merge_sese): Re-implement with a flood-fill
8362 algorithm that properly finds a SESE region if it exists.
8364 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8367 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8368 pointer_diff optimizations use view_convert instead of convert.
8370 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8372 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8373 Generate different code for -mno-speculate-indirect-jumps.
8374 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8375 (*call_indirect_aix<mode>): Disable for
8376 -mno-speculate-indirect-jumps.
8377 (*call_indirect_aix<mode>_nospec): New define_insn.
8378 (*call_value_indirect_aix<mode>): Disable for
8379 -mno-speculate-indirect-jumps.
8380 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8381 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8382 -mno-speculate-indirect-jumps.
8383 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8385 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8387 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8388 long double type, set the flags for noting the default long double
8389 type, even if we don't pass or return a long double type.
8391 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8394 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8397 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8399 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8400 support for merge[hl].
8401 (fold_mergehl_helper): New helper function.
8402 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8403 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8404 (altivec_vmrglw_direct): Add xxmrglw insn.
8406 2018-01-17 Andrew Waterman <andrew@sifive.com>
8408 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8409 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8411 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8414 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8415 call the lto_location_cache before reading the
8416 DECL_SOURCE_LOCATION of the types.
8418 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8419 Richard Sandiford <richard.sandiford@linaro.org>
8421 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8422 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8423 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8424 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8425 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8427 * config/aarch64/constraints.md (aarch64_movti_operand):
8429 * config/aarch64/predicates.md (Uti): Add new constraint.
8431 2018-01-17 Carl Love <cel@us.ibm.com>
8433 * config/rs6000/vsx.md (define_expand xl_len_r,
8434 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8435 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8437 (define_expand, define_insn): Move the shift left from the
8438 define_insn to the define_expand for lxvl and stxvl instructions.
8439 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8440 and XL_LEN_R definitions to PURE.
8442 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8444 * config/i386/i386.c (indirect_thunk_name): Declare regno
8445 as unsigned int. Compare regno with INVALID_REGNUM.
8446 (output_indirect_thunk): Ditto.
8447 (output_indirect_thunk_function): Ditto.
8448 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8449 in the call to output_indirect_thunk_function.
8451 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8454 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8455 rather than the size of inner_type to determine the stack slot size
8456 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8458 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8461 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8464 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8466 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8467 endian Linux systems to optionally enable multilibs for selecting
8468 the long double type if the user configured an explicit type.
8469 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8470 have no long double multilibs if not defined.
8471 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8472 warn if the user used -mabi={ieee,ibm}longdouble and we built
8473 multilibs for long double.
8474 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8475 appropriate multilib option.
8476 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8478 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8479 for building long double multilibs.
8480 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8482 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8484 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8487 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8489 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8492 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8495 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8498 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8500 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8501 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8503 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8505 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8506 different rtl trees depending on TARGET_64BIT.
8507 (rs6000_gen_lvx): Likewise.
8509 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8511 * config/visium/visium.md (nop): Tweak comment.
8512 (hazard_nop): Likewise.
8514 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8516 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8517 -mspeculate-indirect-jumps.
8518 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8519 for -mno-speculate-indirect-jumps.
8520 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8521 (*call_value_indirect_elfv2<mode>): Disable for
8522 -mno-speculate-indirect-jumps.
8523 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8524 (indirect_jump): Emit different RTL for
8525 -mno-speculate-indirect-jumps.
8526 (*indirect_jump<mode>): Disable for
8527 -mno-speculate-indirect-jumps.
8528 (*indirect_jump<mode>_nospec): New define_insn.
8529 (tablejump): Emit different RTL for
8530 -mno-speculate-indirect-jumps.
8531 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8532 (tablejumpsi_nospec): New define_expand.
8533 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8534 (tablejumpdi_nospec): New define_expand.
8535 (*tablejump<mode>_internal1): Disable for
8536 -mno-speculate-indirect-jumps.
8537 (*tablejump<mode>_internal1_nospec): New define_insn.
8538 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8541 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8543 * caller-save.c (insert_save): Drop unnecessary parameter. All
8546 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8547 Richard Biener <rguenth@suse.de>
8550 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8551 return early, inline manually is_gimple_sizepos. Make sure if we
8552 call gimplify_expr we don't end up with a gimple constant.
8553 * tree.c (variably_modified_type_p): Don't return true for
8554 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8555 * gimplify.h (is_gimple_sizepos): Remove.
8557 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8559 PR tree-optimization/83857
8560 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8561 vectorizable_live_operation for pure SLP statements.
8562 (vectorizable_live_operation): Handle PHIs.
8564 2018-01-16 Richard Biener <rguenther@suse.de>
8566 PR tree-optimization/83867
8567 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8568 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8570 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8573 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8574 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8575 If off is not INTEGER_CST, issue a may not be aligned warning
8576 rather than isn't aligned. Use isn%'t rather than isn't.
8577 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8579 <case MULT_EXPR>: Improve the case when bottom and one of the
8580 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8581 operand, in that case check if the other operand is multiple of
8582 bottom divided by the INTEGER_CST operand.
8584 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8587 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8588 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8589 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8590 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8591 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8592 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8593 (pa_function_arg_size): New function.
8595 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8597 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8598 in a separate statement.
8600 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8602 PR tree-optimization/83847
8603 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8604 group gathers and scatters.
8606 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8608 PR rtl-optimization/86620
8609 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8611 PR rtl-optimization/83213
8612 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8613 to last if both are JUMP_INSNs.
8615 PR tree-optimization/83843
8616 * gimple-ssa-store-merging.c
8617 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8618 store_immediate_info for bswap/nop orig_stores.
8620 2018-01-15 Andrew Waterman <andrew@sifive.com>
8622 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8624 <UDIV>: Increase cost if !TARGET_DIV.
8626 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8628 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8629 (define_attr "cr_logical_3op"): New.
8630 (cceq_ior_compare): Adjust.
8631 (cceq_ior_compare_complement): Adjust.
8632 (*cceq_rev_compare): Adjust.
8633 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8634 (is_cracked_insn): Adjust.
8635 (insn_must_be_first_in_group): Adjust.
8636 * config/rs6000/40x.md: Adjust.
8637 * config/rs6000/440.md: Adjust.
8638 * config/rs6000/476.md: Adjust.
8639 * config/rs6000/601.md: Adjust.
8640 * config/rs6000/603.md: Adjust.
8641 * config/rs6000/6xx.md: Adjust.
8642 * config/rs6000/7450.md: Adjust.
8643 * config/rs6000/7xx.md: Adjust.
8644 * config/rs6000/8540.md: Adjust.
8645 * config/rs6000/cell.md: Adjust.
8646 * config/rs6000/e300c2c3.md: Adjust.
8647 * config/rs6000/e500mc.md: Adjust.
8648 * config/rs6000/e500mc64.md: Adjust.
8649 * config/rs6000/e5500.md: Adjust.
8650 * config/rs6000/e6500.md: Adjust.
8651 * config/rs6000/mpc.md: Adjust.
8652 * config/rs6000/power4.md: Adjust.
8653 * config/rs6000/power5.md: Adjust.
8654 * config/rs6000/power6.md: Adjust.
8655 * config/rs6000/power7.md: Adjust.
8656 * config/rs6000/power8.md: Adjust.
8657 * config/rs6000/power9.md: Adjust.
8658 * config/rs6000/rs64.md: Adjust.
8659 * config/rs6000/titan.md: Adjust.
8661 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8663 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8664 ix86_indirect_branch_register logic.
8666 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8668 * config/i386/constraints.md (Bs): Update
8669 ix86_indirect_branch_register check. Don't check
8670 ix86_indirect_branch_register with GOT_memory_operand.
8672 * config/i386/predicates.md (GOT_memory_operand): Don't check
8673 ix86_indirect_branch_register here.
8674 (GOT32_symbol_operand): Likewise.
8676 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8678 * config/i386/predicates.md (constant_call_address_operand):
8679 Rewrite ix86_indirect_branch_register logic.
8680 (sibcall_insn_operand): Likewise.
8682 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8684 * config/i386/constraints.md (Bs): Replace
8685 ix86_indirect_branch_thunk_register with
8686 ix86_indirect_branch_register.
8688 * config/i386/i386.md (indirect_jump): Likewise.
8689 (tablejump): Likewise.
8690 (*sibcall_memory): Likewise.
8691 (*sibcall_value_memory): Likewise.
8692 Peepholes of indirect call and jump via memory: Likewise.
8693 * config/i386/i386.opt: Likewise.
8694 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8695 (GOT_memory_operand): Likewise.
8696 (call_insn_operand): Likewise.
8697 (sibcall_insn_operand): Likewise.
8698 (GOT32_symbol_operand): Likewise.
8700 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8703 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8704 type rather than type addr's type points to.
8705 (expand_omp_atomic_mutex): Likewise.
8706 (expand_omp_atomic): Likewise.
8708 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8711 * config/i386/i386.c (output_indirect_thunk_function): Use
8712 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8713 for __x86_return_thunk.
8715 2018-01-15 Richard Biener <rguenther@suse.de>
8718 * expmed.c (extract_bit_field_1): Fix typo.
8720 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8723 * config/arm/iterators.md (VF): New mode iterator.
8724 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8725 Remove integer-related logic from pattern.
8726 (neon_vabd<mode>_3): Likewise.
8728 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8731 * common.opt (fstrict-overflow): No longer an alias.
8732 (fwrapv-pointer): New option.
8733 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8734 also for pointer types based on flag_wrapv_pointer.
8735 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8736 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8737 opts->x_flag_wrapv got set.
8738 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8739 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8740 POINTER_TYPE_OVERFLOW_UNDEFINED.
8741 * match.pd: Likewise in address comparison pattern.
8742 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8744 2018-01-15 Richard Biener <rguenther@suse.de>
8747 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8748 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8749 Reset type names to their identifier if their TYPE_DECL doesn't
8750 have linkage (and thus is used for ODR and devirt).
8751 (save_debug_info_for_decl): Remove.
8752 (save_debug_info_for_type): Likewise.
8753 (add_tree_to_fld_list): Adjust.
8754 * tree-pretty-print.c (dump_generic_node): Make dumping of
8755 type names more robust.
8757 2018-01-15 Richard Biener <rguenther@suse.de>
8759 * BASE-VER: Bump to 8.0.1.
8761 2018-01-14 Martin Sebor <msebor@redhat.com>
8764 * builtins.c (check_access): Avoid warning when the no-warning bit
8767 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8769 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8770 * ira-color (allocno_hard_regs_compare): Likewise.
8772 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8775 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8776 Use .pushsection/.popsection.
8778 2018-01-14 Martin Sebor <msebor@redhat.com>
8781 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8783 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8785 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8786 entry from extra_headers.
8787 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8788 extra_headers, make the list bitwise identical to the i?86-*-* one.
8790 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8792 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8793 -mcmodel=large with -mindirect-branch=thunk,
8794 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8795 -mfunction-return=thunk-extern.
8796 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8797 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8798 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8800 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8802 * config/i386/i386.c (print_reg): Print the name of the full
8803 integer register without '%'.
8804 (ix86_print_operand): Handle 'V'.
8805 * doc/extend.texi: Document 'V' modifier.
8807 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8809 * config/i386/constraints.md (Bs): Disallow memory operand for
8810 -mindirect-branch-register.
8812 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8813 (GOT_memory_operand): Likewise.
8814 (call_insn_operand): Likewise.
8815 (sibcall_insn_operand): Likewise.
8816 (GOT32_symbol_operand): Likewise.
8817 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8818 for -mindirect-branch-register.
8819 (tablejump): Likewise.
8820 (*sibcall_memory): Likewise.
8821 (*sibcall_value_memory): Likewise.
8822 Disallow peepholes of indirect call and jump via memory for
8823 -mindirect-branch-register.
8824 (*call_pop): Replace m with Bw.
8825 (*call_value_pop): Likewise.
8826 (*sibcall_pop_memory): Replace m with Bs.
8827 * config/i386/i386.opt (mindirect-branch-register): New option.
8828 * doc/invoke.texi: Document -mindirect-branch-register option.
8830 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8832 * config/i386/i386-protos.h (ix86_output_function_return): New.
8833 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8834 set function_return_type.
8835 (indirect_thunk_name): Add ret_p to indicate thunk for function
8837 (output_indirect_thunk_function): Pass false to
8838 indirect_thunk_name.
8839 (ix86_output_indirect_branch_via_reg): Likewise.
8840 (ix86_output_indirect_branch_via_push): Likewise.
8841 (output_indirect_thunk_function): Create alias for function
8842 return thunk if regno < 0.
8843 (ix86_output_function_return): New function.
8844 (ix86_handle_fndecl_attribute): Handle function_return.
8845 (ix86_attribute_table): Add function_return.
8846 * config/i386/i386.h (machine_function): Add
8847 function_return_type.
8848 * config/i386/i386.md (simple_return_internal): Use
8849 ix86_output_function_return.
8850 (simple_return_internal_long): Likewise.
8851 * config/i386/i386.opt (mfunction-return=): New option.
8852 (indirect_branch): Mention -mfunction-return=.
8853 * doc/extend.texi: Document function_return function attribute.
8854 * doc/invoke.texi: Document -mfunction-return= option.
8856 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8858 * config/i386/i386-opts.h (indirect_branch): New.
8859 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8860 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8861 with local indirect jump when converting indirect call and jump.
8862 (ix86_set_indirect_branch_type): New.
8863 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8864 (indirectlabelno): New.
8865 (indirect_thunk_needed): Likewise.
8866 (indirect_thunk_bnd_needed): Likewise.
8867 (indirect_thunks_used): Likewise.
8868 (indirect_thunks_bnd_used): Likewise.
8869 (INDIRECT_LABEL): Likewise.
8870 (indirect_thunk_name): Likewise.
8871 (output_indirect_thunk): Likewise.
8872 (output_indirect_thunk_function): Likewise.
8873 (ix86_output_indirect_branch_via_reg): Likewise.
8874 (ix86_output_indirect_branch_via_push): Likewise.
8875 (ix86_output_indirect_branch): Likewise.
8876 (ix86_output_indirect_jmp): Likewise.
8877 (ix86_code_end): Call output_indirect_thunk_function if needed.
8878 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8880 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8881 (ix86_attribute_table): Add indirect_branch.
8882 * config/i386/i386.h (machine_function): Add indirect_branch_type
8883 and has_local_indirect_jump.
8884 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8886 (tablejump): Likewise.
8887 (*indirect_jump): Use ix86_output_indirect_jmp.
8888 (*tablejump_1): Likewise.
8889 (simple_return_indirect_internal): Likewise.
8890 * config/i386/i386.opt (mindirect-branch=): New option.
8891 (indirect_branch): New.
8894 (thunk-inline): Likewise.
8895 (thunk-extern): Likewise.
8896 * doc/extend.texi: Document indirect_branch function attribute.
8897 * doc/invoke.texi: Document -mindirect-branch= option.
8899 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8902 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8904 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8906 * ipa-inline.c (want_inline_small_function_p): Return false if
8907 inlining has already failed with CIF_FINAL_ERROR.
8908 (update_caller_keys): Call want_inline_small_function_p before
8910 (update_callee_keys): Likewise.
8912 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8914 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8916 (rs6000_quadword_masked_address_p): Likewise.
8917 (quad_aligned_load_p): Likewise.
8918 (quad_aligned_store_p): Likewise.
8919 (const_load_sequence_p): Add comment to describe the outer-most loop.
8920 (mimic_memory_attributes_and_flags): New function.
8921 (rs6000_gen_stvx): Likewise.
8922 (replace_swapped_aligned_store): Likewise.
8923 (rs6000_gen_lvx): Likewise.
8924 (replace_swapped_aligned_load): Likewise.
8925 (replace_swapped_load_constant): Capitalize argument name in
8926 comment describing this function.
8927 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8928 and stores that access quad-word aligned addresses and replace
8929 with stvx or lvx instructions when appropriate.
8930 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8931 New function prototype.
8932 (rs6000_quadword_masked_address_p): Likewise.
8933 (rs6000_gen_lvx): Likewise.
8934 (rs6000_gen_stvx): Likewise.
8935 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8936 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8937 when memory address is aligned.
8938 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8939 this split to select lvx instruction when memory address is aligned.
8940 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8941 instruction when memory address is aligned.
8942 (*vsx_le_perm_load_v16qi): Likewise.
8943 (four unnamed splitters): Modify to select the stvx instruction
8944 when memory is aligned.
8946 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8948 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8949 which appears in the queue multiple times.
8951 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8952 Alan Hayward <alan.hayward@arm.com>
8953 David Sherwood <david.sherwood@arm.com>
8955 * tree-vectorizer.h (vec_lower_bound): New structure.
8956 (_loop_vec_info): Add check_nonzero and lower_bounds.
8957 (LOOP_VINFO_CHECK_NONZERO): New macro.
8958 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8959 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8960 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8961 fields. Make seg_len the distance travelled, not including the
8963 (dr_direction_indicator): Declare.
8964 (dr_zero_step_indicator): Likewise.
8965 (dr_known_forward_stride_p): Likewise.
8966 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8968 (runtime_alias_check_p): Allow runtime alias checks with
8970 (operator ==): Compare access_size and align.
8971 (prune_runtime_alias_test_list): Rework for new distinction between
8972 the access_size and seg_len.
8973 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8975 (get_segment_min_max): New function.
8976 (create_intersect_range_checks): Use it.
8977 (dr_step_indicator): New function.
8978 (dr_direction_indicator): Likewise.
8979 (dr_zero_step_indicator): Likewise.
8980 (dr_known_forward_stride_p): Likewise.
8981 * tree-loop-distribution.c (data_ref_segment_size): Return
8982 DR_STEP * (niters - 1).
8983 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8985 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8986 (vect_preserves_scalar_order_p): New function, split out from...
8987 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8988 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8989 (vect_vfa_access_size): New function.
8990 (vect_vfa_align): Likewise.
8991 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8992 (dump_lower_bound): New function.
8993 (vect_check_lower_bound): Likewise.
8994 (vect_small_gap_p): Likewise.
8995 (vectorizable_with_step_bound_p): Likewise.
8996 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8997 depencies if the vectorization factor is 1. Convert the checks
8998 for nonzero steps into checks on the bounds of DR_STEP. Try using
8999 a bunds check for variable steps if the minimum required step is
9000 relatively small. Update calls to the dr_with_seg_len
9001 constructor and to vect_compile_time_alias.
9002 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9004 (vect_loop_versioning): Call it.
9005 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9007 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9009 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9010 Alan Hayward <alan.hayward@arm.com>
9011 David Sherwood <david.sherwood@arm.com>
9013 * doc/sourcebuild.texi (vect_scatter_store): Document.
9014 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9016 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9018 * genopinit.c (main): Add supports_vec_scatter_store and
9019 supports_vec_scatter_store_cached to target_optabs.
9020 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9021 IFN_MASK_SCATTER_STORE.
9022 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9024 * internal-fn.h (internal_store_fn_p): Declare.
9025 (internal_fn_stored_value_index): Likewise.
9026 * internal-fn.c (scatter_store_direct): New macro.
9027 (expand_scatter_store_optab_fn): New function.
9028 (direct_scatter_store_optab_supported_p): New macro.
9029 (internal_store_fn_p): New function.
9030 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9031 IFN_MASK_SCATTER_STORE.
9032 (internal_fn_mask_index): Likewise.
9033 (internal_fn_stored_value_index): New function.
9034 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9036 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9037 * optabs-query.c (supports_vec_scatter_store_p): New function.
9038 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9039 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9040 true for scatter stores.
9041 (vect_gather_scatter_fn_p): Handle scatter stores too.
9042 (vect_check_gather_scatter): Consider using scatter stores if
9043 supports_vec_scatter_store_p.
9044 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9046 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9047 internal_fn_stored_value_index.
9048 (check_load_store_masking): Handle scatter stores too.
9049 (vect_get_store_rhs): Make public.
9050 (vectorizable_call): Use internal_store_fn_p.
9051 (vectorizable_store): Handle scatter store internal functions.
9052 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9053 when deciding whether the end of the group has been reached.
9054 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9055 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9056 (mask_scatter_store<mode>): New insns.
9058 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9059 Alan Hayward <alan.hayward@arm.com>
9060 David Sherwood <david.sherwood@arm.com>
9062 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9063 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9064 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9066 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9067 Use vect_truncate_gather_scatter_offset if we can't treat the
9068 operation as a normal gather load or scatter store.
9069 (get_group_load_store_type): Take the gather_scatter_info
9070 as argument. Try using a gather load or scatter store for
9071 single-element groups.
9072 (get_load_store_type): Update calls to get_group_load_store_type
9073 and vect_use_strided_gather_scatters_p.
9075 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9076 Alan Hayward <alan.hayward@arm.com>
9077 David Sherwood <david.sherwood@arm.com>
9079 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9080 optional tree argument.
9081 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9083 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9084 but continue to use the current value as a fallback.
9085 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9086 to compare the updates.
9087 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9088 (get_load_store_type): Use it when handling a strided access.
9089 (vect_get_strided_load_store_ops): New function.
9090 (vect_get_data_ptr_increment): Likewise.
9091 (vectorizable_load): Handle strided gather loads. Always pass
9092 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9094 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9095 Alan Hayward <alan.hayward@arm.com>
9096 David Sherwood <david.sherwood@arm.com>
9098 * doc/md.texi (gather_load@var{m}): Document.
9099 (mask_gather_load@var{m}): Likewise.
9100 * genopinit.c (main): Add supports_vec_gather_load and
9101 supports_vec_gather_load_cached to target_optabs.
9102 * optabs-tree.c (init_tree_optimization_optabs): Use
9103 ggc_cleared_alloc to allocate target_optabs.
9104 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9105 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9107 * internal-fn.h (internal_load_fn_p): Declare.
9108 (internal_gather_scatter_fn_p): Likewise.
9109 (internal_fn_mask_index): Likewise.
9110 (internal_gather_scatter_fn_supported_p): Likewise.
9111 * internal-fn.c (gather_load_direct): New macro.
9112 (expand_gather_load_optab_fn): New function.
9113 (direct_gather_load_optab_supported_p): New macro.
9114 (direct_internal_fn_optab): New function.
9115 (internal_load_fn_p): Likewise.
9116 (internal_gather_scatter_fn_p): Likewise.
9117 (internal_fn_mask_index): Likewise.
9118 (internal_gather_scatter_fn_supported_p): Likewise.
9119 * optabs-query.c (supports_at_least_one_mode_p): New function.
9120 (supports_vec_gather_load_p): Likewise.
9121 * optabs-query.h (supports_vec_gather_load_p): Declare.
9122 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9123 and memory_type field.
9124 (NUM_PATTERNS): Bump to 15.
9125 * tree-vect-data-refs.c: Include internal-fn.h.
9126 (vect_gather_scatter_fn_p): New function.
9127 (vect_describe_gather_scatter_call): Likewise.
9128 (vect_check_gather_scatter): Try using internal functions for
9129 gather loads. Recognize existing calls to a gather load function.
9130 (vect_analyze_data_refs): Consider using gather loads if
9131 supports_vec_gather_load_p.
9132 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9133 (vect_get_gather_scatter_offset_type): Likewise.
9134 (vect_convert_mask_for_vectype): Likewise.
9135 (vect_add_conversion_to_patterm): Likewise.
9136 (vect_try_gather_scatter_pattern): Likewise.
9137 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9138 (vect_vect_recog_func_ptrs): Add it.
9139 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9140 internal_fn_mask_index and internal_gather_scatter_fn_p.
9141 (check_load_store_masking): Take the gather_scatter_info as an
9142 argument and handle gather loads.
9143 (vect_get_gather_scatter_ops): New function.
9144 (vectorizable_call): Check internal_load_fn_p.
9145 (vectorizable_load): Likewise. Handle gather load internal
9147 (vectorizable_store): Update call to check_load_store_masking.
9148 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9149 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9150 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9151 (aarch64_gather_scale_operand_d): New predicates.
9152 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9153 (mask_gather_load<mode>): New insns.
9155 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9156 Alan Hayward <alan.hayward@arm.com>
9157 David Sherwood <david.sherwood@arm.com>
9159 * optabs.def (fold_left_plus_optab): New optab.
9160 * doc/md.texi (fold_left_plus_@var{m}): Document.
9161 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9162 * internal-fn.c (fold_left_direct): Define.
9163 (expand_fold_left_optab_fn): Likewise.
9164 (direct_fold_left_optab_supported_p): Likewise.
9165 * fold-const-call.c (fold_const_fold_left): New function.
9166 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9167 * tree-parloops.c (valid_reduction_p): New function.
9168 (gather_scalar_reductions): Use it.
9169 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9170 (vect_finish_replace_stmt): Declare.
9171 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9172 (needs_fold_left_reduction_p): New function, split out from...
9173 (vect_is_simple_reduction): ...here. Accept reductions that
9174 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9175 (vect_force_simple_reduction): Also store the reduction type in
9176 the assignment's STMT_VINFO_REDUC_TYPE.
9177 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9178 (merge_with_identity): New function.
9179 (vect_expand_fold_left): Likewise.
9180 (vectorize_fold_left_reduction): Likewise.
9181 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9182 scalar phi in place for it. Check for target support and reject
9183 cases that would reassociate the operation. Defer the transform
9184 phase to vectorize_fold_left_reduction.
9185 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9186 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9187 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9189 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9191 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9192 call to ifc_temp_var.
9194 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9195 Alan Hayward <alan.hayward@arm.com>
9196 David Sherwood <david.sherwood@arm.com>
9198 * target.def (legitimize_address_displacement): Take the original
9199 offset as a poly_int.
9200 * targhooks.h (default_legitimize_address_displacement): Update
9202 * targhooks.c (default_legitimize_address_displacement): Likewise.
9203 * doc/tm.texi: Regenerate.
9204 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9205 as an argument, moving assert of ad->disp == ad->disp_term to...
9206 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9207 Try calling targetm.legitimize_address_displacement before expanding
9208 the address rather than afterwards, and adjust for the new interface.
9209 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9210 Match the new hook interface. Handle SVE addresses.
9211 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9214 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9216 * Makefile.in (OBJS): Add early-remat.o.
9217 * target.def (select_early_remat_modes): New hook.
9218 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9219 * doc/tm.texi: Regenerate.
9220 * targhooks.h (default_select_early_remat_modes): Declare.
9221 * targhooks.c (default_select_early_remat_modes): New function.
9222 * timevar.def (TV_EARLY_REMAT): New timevar.
9223 * passes.def (pass_early_remat): New pass.
9224 * tree-pass.h (make_pass_early_remat): Declare.
9225 * early-remat.c: New file.
9226 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9228 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9230 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9231 Alan Hayward <alan.hayward@arm.com>
9232 David Sherwood <david.sherwood@arm.com>
9234 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9235 vfm1 with a bound_epilog parameter.
9236 (vect_do_peeling): Update calls accordingly, and move the prologue
9237 call earlier in the function. Treat the base bound_epilog as 0 for
9238 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9239 this base when peeling for gaps.
9240 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9241 with fully-masked loops.
9242 (vect_estimate_min_profitable_iters): Handle the single peeled
9243 iteration in that case.
9245 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9246 Alan Hayward <alan.hayward@arm.com>
9247 David Sherwood <david.sherwood@arm.com>
9249 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9250 single-element interleaving even if the size is not a power of 2.
9251 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9252 accesses for single-element interleaving if the group size is
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 * doc/md.texi (fold_extract_last_@var{m}): Document.
9260 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9261 * optabs.def (fold_extract_last_optab): New optab.
9262 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9263 * internal-fn.c (fold_extract_direct): New macro.
9264 (expand_fold_extract_optab_fn): Likewise.
9265 (direct_fold_extract_optab_supported_p): Likewise.
9266 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9267 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9268 EXTRACT_LAST_REDUCTION.
9269 (get_initial_def_for_reduction): Do not create an initial vector
9270 for EXTRACT_LAST_REDUCTION reductions.
9271 (vectorizable_reduction): Leave the scalar phi in place for
9272 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9273 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9274 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9275 transform phase to vectorizable_condition.
9276 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9278 (vect_finish_stmt_generation): ...here.
9279 (vect_finish_replace_stmt): New function.
9280 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9281 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9283 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9285 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9286 Alan Hayward <alan.hayward@arm.com>
9287 David Sherwood <david.sherwood@arm.com>
9289 * doc/md.texi (extract_last_@var{m}): Document.
9290 * optabs.def (extract_last_optab): New optab.
9291 * internal-fn.def (EXTRACT_LAST): New internal function.
9292 * internal-fn.c (cond_unary_direct): New macro.
9293 (expand_cond_unary_optab_fn): Likewise.
9294 (direct_cond_unary_optab_supported_p): Likewise.
9295 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9296 loops using EXTRACT_LAST.
9297 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9298 (extract_last_<mode>): ...this optab.
9299 (vec_extract<mode><Vel>): Update accordingly.
9301 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9302 Alan Hayward <alan.hayward@arm.com>
9303 David Sherwood <david.sherwood@arm.com>
9305 * target.def (empty_mask_is_expensive): New hook.
9306 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9307 * doc/tm.texi: Regenerate.
9308 * targhooks.h (default_empty_mask_is_expensive): Declare.
9309 * targhooks.c (default_empty_mask_is_expensive): New function.
9310 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9311 if the target says that empty masks are expensive.
9312 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9314 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9316 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9317 Alan Hayward <alan.hayward@arm.com>
9318 David Sherwood <david.sherwood@arm.com>
9320 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9321 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9322 (vect_use_loop_mask_for_alignment_p): New function.
9323 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9324 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9325 niters_skip argument. Make sure that the first niters_skip elements
9326 of the first iteration are inactive.
9327 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9328 Update call to vect_set_loop_masks_directly.
9329 (get_misalign_in_elems): New function, split out from...
9330 (vect_gen_prolog_loop_niters): ...here.
9331 (vect_update_init_of_dr): Take a code argument that specifies whether
9332 the adjustment should be added or subtracted.
9333 (vect_update_init_of_drs): Likewise.
9334 (vect_prepare_for_masked_peels): New function.
9335 (vect_do_peeling): Skip prologue peeling if we're using a mask
9336 instead. Update call to vect_update_inits_of_drs.
9337 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9339 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9340 alignment. Do not include the number of peeled iterations in
9341 the minimum threshold in that case.
9342 (vectorizable_induction): Adjust the start value down by
9343 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9344 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9345 Take the number of skipped iterations into account when calculating
9347 * tree-vect-stmts.c (vect_gen_while_not): New function.
9349 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9350 Alan Hayward <alan.hayward@arm.com>
9351 David Sherwood <david.sherwood@arm.com>
9353 * doc/sourcebuild.texi (vect_fully_masked): Document.
9354 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9356 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9358 (vect_analyze_loop_2): ...here. Don't check the vectorization
9359 factor against the number of loop iterations if the loop is
9362 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9363 Alan Hayward <alan.hayward@arm.com>
9364 David Sherwood <david.sherwood@arm.com>
9366 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9367 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9368 (dump_groups): Update accordingly.
9369 (iv_use::mem_type): New member variable.
9370 (address_p): New function.
9371 (record_use): Add a mem_type argument and initialize the new
9373 (record_group_use): Add a mem_type argument. Use address_p.
9374 Remove obsolete null checks of base_object. Update call to record_use.
9375 (find_interesting_uses_op): Update call to record_group_use.
9376 (find_interesting_uses_cond): Likewise.
9377 (find_interesting_uses_address): Likewise.
9378 (get_mem_type_for_internal_fn): New function.
9379 (find_address_like_use): Likewise.
9380 (find_interesting_uses_stmt): Try find_address_like_use before
9381 calling find_interesting_uses_op.
9382 (addr_offset_valid_p): Use the iv mem_type field as the type
9383 of the addressed memory.
9384 (add_autoinc_candidates): Likewise.
9385 (get_address_cost): Likewise.
9386 (split_small_address_groups_p): Use address_p.
9387 (split_address_groups): Likewise.
9388 (add_iv_candidate_for_use): Likewise.
9389 (autoinc_possible_for_pair): Likewise.
9390 (rewrite_groups): Likewise.
9391 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9392 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9393 (get_alias_ptr_type_for_ptr_address): New function.
9394 (rewrite_use_address): Rewrite address uses in calls that were
9395 identified by find_address_like_use.
9397 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9398 Alan Hayward <alan.hayward@arm.com>
9399 David Sherwood <david.sherwood@arm.com>
9401 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9403 * gimple-expr.h (is_gimple_addressable: Likewise.
9404 * gimple-expr.c (is_gimple_address): Likewise.
9405 * internal-fn.c (expand_call_mem_ref): New function.
9406 (expand_mask_load_optab_fn): Use it.
9407 (expand_mask_store_optab_fn): Likewise.
9409 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9410 Alan Hayward <alan.hayward@arm.com>
9411 David Sherwood <david.sherwood@arm.com>
9413 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9414 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9415 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9416 (cond_umax@var{mode}): Document.
9417 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9418 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9419 (cond_umin_optab, cond_umax_optab): New optabs.
9420 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9421 (COND_IOR, COND_XOR): New internal functions.
9422 * internal-fn.h (get_conditional_internal_fn): Declare.
9423 * internal-fn.c (cond_binary_direct): New macro.
9424 (expand_cond_binary_optab_fn): Likewise.
9425 (direct_cond_binary_optab_supported_p): Likewise.
9426 (get_conditional_internal_fn): New function.
9427 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9428 Cope with reduction statements that are vectorized as calls rather
9430 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9431 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9432 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9433 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9434 (UNSPEC_COND_EOR): New unspecs.
9435 (optab): Add mappings for them.
9436 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9437 (sve_int_op, sve_fp_op): New int attributes.
9439 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9440 Alan Hayward <alan.hayward@arm.com>
9441 David Sherwood <david.sherwood@arm.com>
9443 * optabs.def (while_ult_optab): New optab.
9444 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9445 * internal-fn.def (WHILE_ULT): New internal function.
9446 * internal-fn.h (direct_internal_fn_supported_p): New override
9447 that takes two types as argument.
9448 * internal-fn.c (while_direct): New macro.
9449 (expand_while_optab_fn): New function.
9450 (convert_optab_supported_p): Likewise.
9451 (direct_while_optab_supported_p): New macro.
9452 * wide-int.h (wi::udiv_ceil): New function.
9453 * tree-vectorizer.h (rgroup_masks): New structure.
9454 (vec_loop_masks): New typedef.
9455 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9457 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9458 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9459 (vect_max_vf): New function.
9460 (slpeel_make_loop_iterate_ntimes): Delete.
9461 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9462 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9463 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9464 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9465 internal-fn.h, stor-layout.h and optabs-query.h.
9466 (vect_set_loop_mask): New function.
9467 (add_preheader_seq): Likewise.
9468 (add_header_seq): Likewise.
9469 (interleave_supported_p): Likewise.
9470 (vect_maybe_permute_loop_masks): Likewise.
9471 (vect_set_loop_masks_directly): Likewise.
9472 (vect_set_loop_condition_masked): Likewise.
9473 (vect_set_loop_condition_unmasked): New function, split out from
9474 slpeel_make_loop_iterate_ntimes.
9475 (slpeel_make_loop_iterate_ntimes): Rename to..
9476 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9477 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9478 (vect_do_peeling): Update call accordingly.
9479 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9481 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9482 mask_compare_type, can_fully_mask_p and fully_masked_p.
9483 (release_vec_loop_masks): New function.
9484 (_loop_vec_info): Use it to free the loop masks.
9485 (can_produce_all_loop_masks_p): New function.
9486 (vect_get_max_nscalars_per_iter): Likewise.
9487 (vect_verify_full_masking): Likewise.
9488 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9489 retries, and free the mask rgroups before retrying. Check loop-wide
9490 reasons for disallowing fully-masked loops. Make the final decision
9491 about whether use a fully-masked loop or not.
9492 (vect_estimate_min_profitable_iters): Do not assume that peeling
9493 for the number of iterations will be needed for fully-masked loops.
9494 (vectorizable_reduction): Disable fully-masked loops.
9495 (vectorizable_live_operation): Likewise.
9496 (vect_halve_mask_nunits): New function.
9497 (vect_double_mask_nunits): Likewise.
9498 (vect_record_loop_mask): Likewise.
9499 (vect_get_loop_mask): Likewise.
9500 (vect_transform_loop): Handle the case in which the final loop
9501 iteration might handle a partial vector. Call vect_set_loop_condition
9502 instead of slpeel_make_loop_iterate_ntimes.
9503 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9504 (check_load_store_masking): New function.
9505 (prepare_load_store_mask): Likewise.
9506 (vectorizable_store): Handle fully-masked loops.
9507 (vectorizable_load): Likewise.
9508 (supportable_widening_operation): Use vect_halve_mask_nunits for
9510 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9511 (vect_gen_while): New function.
9512 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9513 (aarch64_uqdec<mode>): New insn.
9515 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9516 Alan Hayward <alan.hayward@arm.com>
9517 David Sherwood <david.sherwood@arm.com>
9519 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9520 (reduc_xor_scal_optab): New optabs.
9521 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9522 (reduc_xor_scal_@var{m}): Document.
9523 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9524 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9526 * fold-const-call.c (fold_const_call): Handle them.
9527 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9528 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9529 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9530 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9531 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9532 (UNSPEC_XORV): New unspecs.
9533 (optab): Add entries for them.
9534 (BITWISEV): New int iterator.
9535 (bit_reduc_op): New int attributes.
9537 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9538 Alan Hayward <alan.hayward@arm.com>
9539 David Sherwood <david.sherwood@arm.com>
9541 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9542 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9543 * optabs.def (vec_shl_insert_optab): New optab.
9544 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9545 (duplicate_and_interleave): Likewise.
9546 * tree-vect-loop.c: Include internal-fn.h.
9547 (neutral_op_for_slp_reduction): New function, split out from
9548 get_initial_defs_for_reduction.
9549 (get_initial_def_for_reduction): Handle option 2 for variable-length
9550 vectors by loading the neutral value into a vector and then shifting
9551 the initial value into element 0.
9552 (get_initial_defs_for_reduction): Replace the code argument with
9553 the neutral value calculated by neutral_op_for_slp_reduction.
9554 Use gimple_build_vector for constant-length vectors.
9555 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9556 but the first group_size elements have a neutral value.
9557 Use duplicate_and_interleave otherwise.
9558 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9559 Update call to get_initial_defs_for_reduction. Handle SLP
9560 reductions for variable-length vectors by creating one vector
9561 result for each scalar result, with the elements associated
9562 with other scalar results stubbed out with the neutral value.
9563 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9564 Require IFN_VEC_SHL_INSERT for double reductions on
9565 variable-length vectors, or SLP reductions that have
9566 a neutral value. Require can_duplicate_and_interleave_p
9567 support for variable-length unchained SLP reductions if there
9568 is no neutral value, such as for MIN/MAX reductions. Also require
9569 the number of vector elements to be a multiple of the number of
9570 SLP statements when doing variable-length unchained SLP reductions.
9571 Update call to vect_create_epilog_for_reduction.
9572 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9573 and remove initial values.
9574 (duplicate_and_interleave): Make public.
9575 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9576 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9578 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9579 Alan Hayward <alan.hayward@arm.com>
9580 David Sherwood <david.sherwood@arm.com>
9582 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9583 (can_duplicate_and_interleave_p): New function.
9584 (vect_get_and_check_slp_defs): Take the vector of statements
9585 rather than just the current one. Remove excess parentheses.
9586 Restriction rejectinon of vect_constant_def and vect_external_def
9587 for variable-length vectors to boolean types, or types for which
9588 can_duplicate_and_interleave_p is false.
9589 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9590 (duplicate_and_interleave): New function.
9591 (vect_get_constant_vectors): Use gimple_build_vector for
9592 constant-length vectors and suitable variable-length constant
9593 vectors. Use duplicate_and_interleave for other variable-length
9594 vectors. Don't defer the update when inserting new statements.
9596 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9597 Alan Hayward <alan.hayward@arm.com>
9598 David Sherwood <david.sherwood@arm.com>
9600 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9601 min_profitable_iters doesn't go negative.
9603 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9604 Alan Hayward <alan.hayward@arm.com>
9605 David Sherwood <david.sherwood@arm.com>
9607 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9608 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9609 * optabs.def (vec_mask_load_lanes_optab): New optab.
9610 (vec_mask_store_lanes_optab): Likewise.
9611 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9612 (MASK_STORE_LANES): Likewise.
9613 * internal-fn.c (mask_load_lanes_direct): New macro.
9614 (mask_store_lanes_direct): Likewise.
9615 (expand_mask_load_optab_fn): Handle masked operations.
9616 (expand_mask_load_lanes_optab_fn): New macro.
9617 (expand_mask_store_optab_fn): Handle masked operations.
9618 (expand_mask_store_lanes_optab_fn): New macro.
9619 (direct_mask_load_lanes_optab_supported_p): Likewise.
9620 (direct_mask_store_lanes_optab_supported_p): Likewise.
9621 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9623 (vect_load_lanes_supported): Likewise.
9624 * tree-vect-data-refs.c (strip_conversion): New function.
9625 (can_group_stmts_p): Likewise.
9626 (vect_analyze_data_ref_accesses): Use it instead of checking
9627 for a pair of assignments.
9628 (vect_store_lanes_supported): Take a masked_p parameter.
9629 (vect_load_lanes_supported): Likewise.
9630 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9631 vect_store_lanes_supported and vect_load_lanes_supported.
9632 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9633 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9634 parameter. Don't allow gaps for masked accesses.
9635 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9636 and vect_load_lanes_supported.
9637 (get_load_store_type): Take a masked_p parameter and update
9638 call to get_group_load_store_type.
9639 (vectorizable_store): Update call to get_load_store_type.
9640 Handle IFN_MASK_STORE_LANES.
9641 (vectorizable_load): Update call to get_load_store_type.
9642 Handle IFN_MASK_LOAD_LANES.
9644 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9645 Alan Hayward <alan.hayward@arm.com>
9646 David Sherwood <david.sherwood@arm.com>
9648 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9650 * config/aarch64/aarch64-protos.h
9651 (aarch64_sve_struct_memory_operand_p): Declare.
9652 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9653 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9654 (VPRED, vpred): Handle SVE structure modes.
9655 * config/aarch64/constraints.md (Utx): New constraint.
9656 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9657 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9658 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9659 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9660 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9661 structure modes. Split into pieces after RA.
9662 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9663 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9665 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9666 SVE structure modes.
9667 (aarch64_classify_address): Likewise.
9668 (sizetochar): Move earlier in file.
9669 (aarch64_print_operand): Handle SVE register lists.
9670 (aarch64_array_mode): New function.
9671 (aarch64_sve_struct_memory_operand_p): Likewise.
9672 (TARGET_ARRAY_MODE): Redefine.
9674 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9675 Alan Hayward <alan.hayward@arm.com>
9676 David Sherwood <david.sherwood@arm.com>
9678 * target.def (array_mode): New target hook.
9679 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9680 * doc/tm.texi: Regenerate.
9681 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9682 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9683 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9685 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9688 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9689 Alan Hayward <alan.hayward@arm.com>
9690 David Sherwood <david.sherwood@arm.com>
9692 * fold-const.c (fold_binary_loc): Check the argument types
9693 rather than the result type when testing for a vector operation.
9695 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9697 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9698 * doc/tm.texi: Regenerate.
9700 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9701 Alan Hayward <alan.hayward@arm.com>
9702 David Sherwood <david.sherwood@arm.com>
9704 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9705 (sve): Document new AArch64 extension.
9706 * doc/md.texi (w): Extend the description of the AArch64
9707 constraint to include SVE vectors.
9708 (Upl, Upa): Document new AArch64 predicate constraints.
9709 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9711 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9712 (msve-vector-bits=): New option.
9713 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9714 SVE when these are disabled.
9715 (sve): New extension.
9716 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9717 modes. Adjust their number of units based on aarch64_sve_vg.
9718 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9719 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9720 aarch64_addr_query_type.
9721 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9722 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9723 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9724 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9725 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9726 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9727 (aarch64_simd_imm_zero_p): Delete.
9728 (aarch64_check_zero_based_sve_index_immediate): Declare.
9729 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9730 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9731 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9732 (aarch64_sve_float_mul_immediate_p): Likewise.
9733 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9735 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9736 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9737 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9738 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9739 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9740 (aarch64_regmode_natural_size): Likewise.
9741 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9742 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9744 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9745 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9746 for VG and the SVE predicate registers.
9747 (V_ALIASES): Add a "z"-prefixed alias.
9748 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9749 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9750 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9751 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9752 (REG_CLASS_NAMES): Add entries for them.
9753 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9754 and the predicate registers.
9755 (aarch64_sve_vg): Declare.
9756 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9757 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9758 (REGMODE_NATURAL_SIZE): Define.
9759 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9761 * config/aarch64/aarch64.c: Include cfgrtl.h.
9762 (simd_immediate_info): Add a constructor for series vectors,
9763 and an associated step field.
9764 (aarch64_sve_vg): New variable.
9765 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9766 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9767 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9768 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9769 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9770 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9771 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9772 (aarch64_get_mask_mode): New functions.
9773 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9774 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9775 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9776 predicate modes and predicate registers. Explicitly restrict
9777 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9778 to store a vector mode if it is recognized by
9779 aarch64_classify_vector_mode.
9780 (aarch64_regmode_natural_size): New function.
9781 (aarch64_hard_regno_caller_save_mode): Return the original mode
9783 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9784 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9785 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9786 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9788 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9789 does not overlap dest if the function is frame-related. Handle
9791 (aarch64_split_add_offset): New function.
9792 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9793 them aarch64_add_offset.
9794 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9795 and update call to aarch64_sub_sp.
9796 (aarch64_add_cfa_expression): New function.
9797 (aarch64_expand_prologue): Pass extra temporary registers to the
9798 functions above. Handle the case in which we need to emit new
9799 DW_CFA_expressions for registers that were originally saved
9800 relative to the stack pointer, but now have to be expressed
9801 relative to the frame pointer.
9802 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9804 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9805 IP0 and IP1 values for SVE frames.
9806 (aarch64_expand_vec_series): New function.
9807 (aarch64_expand_sve_widened_duplicate): Likewise.
9808 (aarch64_expand_sve_const_vector): Likewise.
9809 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9810 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9811 into the register, rather than emitting a SET directly.
9812 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9813 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9814 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9815 (offset_9bit_signed_scaled_p): New functions.
9816 (aarch64_replicate_bitmask_imm): New function.
9817 (aarch64_bitmask_imm): Use it.
9818 (aarch64_cannot_force_const_mem): Reject expressions involving
9819 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9820 (aarch64_classify_index): Handle SVE indices, by requiring
9821 a plain register index with a scale that matches the element size.
9822 (aarch64_classify_address): Handle SVE addresses. Assert that
9823 the mode of the address is VOIDmode or an integer mode.
9824 Update call to aarch64_classify_symbol.
9825 (aarch64_classify_symbolic_expression): Update call to
9826 aarch64_classify_symbol.
9827 (aarch64_const_vec_all_in_range_p): New function.
9828 (aarch64_print_vector_float_operand): Likewise.
9829 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9830 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9831 and the FP immediates 1.0 and 0.5.
9832 (aarch64_print_address_internal): Handle SVE addresses.
9833 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9834 (aarch64_regno_regclass): Handle predicate registers.
9835 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9837 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9838 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9839 (aarch64_convert_sve_vector_bits): New function.
9840 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9841 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9843 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9844 Handle SVE vector and predicate modes. Accept VL-based constants
9845 that need only one temporary register, and VL offsets that require
9846 no temporary registers.
9847 (aarch64_conditional_register_usage): Mark the predicate registers
9848 as fixed if SVE isn't available.
9849 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9850 Return true for SVE vector and predicate modes.
9851 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9852 rather than an unsigned int. Handle SVE modes.
9853 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9855 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9857 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9858 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9859 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9860 (aarch64_sve_float_mul_immediate_p): New functions.
9861 (aarch64_sve_valid_immediate): New function.
9862 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9863 Explicitly reject structure modes. Check for INDEX constants.
9864 Handle PTRUE and PFALSE constants.
9865 (aarch64_check_zero_based_sve_index_immediate): New function.
9866 (aarch64_simd_imm_zero_p): Delete.
9867 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9868 vector modes. Accept constants in the range of CNT[BHWD].
9869 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9870 ask for an Advanced SIMD mode.
9871 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9872 (aarch64_simd_vector_alignment): Handle SVE predicates.
9873 (aarch64_vectorize_preferred_vector_alignment): New function.
9874 (aarch64_simd_vector_alignment_reachable): Use it instead of
9876 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9877 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9879 (MAX_VECT_LEN): Delete.
9880 (expand_vec_perm_d): Add a vec_flags field.
9881 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9882 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9883 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9885 (aarch64_evpc_rev): Rename to...
9886 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9887 (aarch64_evpc_rev_global): New function.
9888 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9889 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9891 (aarch64_evpc_sve_tbl): New function.
9892 (aarch64_expand_vec_perm_const_1): Update after rename of
9893 aarch64_evpc_rev. Handle SVE permutes too, trying
9894 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9895 than aarch64_evpc_tbl.
9896 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9897 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9898 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9899 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9900 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9901 (aarch64_expand_sve_vcond): New functions.
9902 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9903 of aarch64_vector_mode_p.
9904 (aarch64_dwarf_poly_indeterminate_value): New function.
9905 (aarch64_compute_pressure_classes): Likewise.
9906 (aarch64_can_change_mode_class): Likewise.
9907 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9908 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9909 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9910 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9911 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9912 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9913 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9914 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9916 (Dn, Dl, Dr): Accept const as well as const_vector.
9917 (Dz): Likewise. Compare against CONST0_RTX.
9918 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9919 of "vector" where appropriate.
9920 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9921 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9922 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9923 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9924 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9925 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9926 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9927 (v_int_equiv): Extend to SVE modes.
9928 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9930 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9931 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9932 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9933 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9934 (SVE_COND_FP_CMP): New int iterators.
9935 (perm_hilo): Handle the new unpack unspecs.
9936 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9938 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9939 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9940 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9941 (aarch64_equality_operator, aarch64_constant_vector_operand)
9942 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9943 (aarch64_sve_nonimmediate_operand): Likewise.
9944 (aarch64_sve_general_operand): Likewise.
9945 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9946 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9947 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9948 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9949 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9950 (aarch64_sve_float_arith_immediate): Likewise.
9951 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9952 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9953 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9954 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9955 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9956 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9957 (aarch64_sve_float_arith_operand): Likewise.
9958 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9959 (aarch64_sve_float_mul_operand): Likewise.
9960 (aarch64_sve_vec_perm_operand): Likewise.
9961 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9962 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9963 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9964 as well as const_vector.
9965 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9966 in file. Use CONST0_RTX and CONSTM1_RTX.
9967 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9968 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9969 Use aarch64_simd_imm_zero.
9970 * config/aarch64/aarch64-sve.md: New file.
9971 * config/aarch64/aarch64.md: Include it.
9972 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9973 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9974 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9975 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9976 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9977 (sve): New attribute.
9978 (enabled): Disable instructions with the sve attribute unless
9980 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9981 aarch64_expand_mov_immediate.
9982 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9983 CNT[BHSD] immediates.
9984 (movti): Split CONST_POLY_INT moves into two halves.
9985 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9986 Split additions that need a temporary here if the destination
9987 is the stack pointer.
9988 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9989 (*add<mode>3_poly_1): New instruction.
9990 (set_clobber_cc): New expander.
9992 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9994 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9995 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9996 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9997 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9998 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9999 Change innermode from fixed_mode_size to machine_mode.
10000 (simplify_subreg): Update call accordingly. Handle a constant-sized
10001 subreg of a variable-length CONST_VECTOR.
10003 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10004 Alan Hayward <alan.hayward@arm.com>
10005 David Sherwood <david.sherwood@arm.com>
10007 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10008 (add_offset_to_base): New function, split out from...
10009 (create_mem_ref): ...here. When handling a scale other than 1,
10010 check first whether the address is valid without the offset.
10011 Add it into the base if so, leaving the index and scale as-is.
10013 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10016 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10017 fold_for_warn before checking if arg2 is INTEGER_CST.
10019 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10021 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10022 (store_multiple_operation): Delete.
10023 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10024 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10025 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10026 guarded by TARGET_STRING.
10027 (rs6000_output_load_multiple): Delete.
10028 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10029 OPTION_MASK_STRING / TARGET_STRING handling.
10030 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10031 (const rs6000_opt_masks) <"string">: Change mask to 0.
10032 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10033 (MASK_STRING): Delete.
10034 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10036 (load_multiple): Delete.
10043 (store_multiple): Delete.
10050 (movmemsi_8reg): Delete.
10051 (corresponding unnamed define_insn): Delete.
10052 (movmemsi_6reg): Delete.
10053 (corresponding unnamed define_insn): Delete.
10054 (movmemsi_4reg): Delete.
10055 (corresponding unnamed define_insn): Delete.
10056 (movmemsi_2reg): Delete.
10057 (corresponding unnamed define_insn): Delete.
10058 (movmemsi_1reg): Delete.
10059 (corresponding unnamed define_insn): Delete.
10060 * config/rs6000/rs6000.opt (mno-string): New.
10061 (mstring): Replace by deprecation warning stub.
10062 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10064 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10066 * regrename.c (regrename_do_replace): If replacing the same
10067 reg multiple times, try to reuse last created gen_raw_REG.
10070 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10071 main to workaround a bug in GDB.
10073 2018-01-12 Tom de Vries <tom@codesourcery.com>
10076 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10078 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10080 PR rtl-optimization/80481
10081 * ira-color.c (get_cap_member): New function.
10082 (allocnos_conflict_by_live_ranges_p): Use it.
10083 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10084 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10086 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10089 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10090 (*saddl_se_1): Ditto.
10091 (*ssubsi_1): Ditto.
10092 (*ssubl_se_1): Ditto.
10094 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10096 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10097 rather than wi::to_widest for DR_INITs.
10098 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10099 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10100 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10102 (vect_analyze_group_access_1): Note that here.
10104 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10106 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10107 polynomial type sizes.
10109 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10111 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10112 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10113 (gimple_add_tmp_var): Likewise.
10115 2018-01-12 Martin Liska <mliska@suse.cz>
10117 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10118 (gimple_alloc_sizes): Likewise.
10119 (dump_gimple_statistics): Use PRIu64 in printf format.
10120 * gimple.h: Change uint64_t to int.
10122 2018-01-12 Martin Liska <mliska@suse.cz>
10124 * tree-core.h: Use uint64_t instead of int.
10125 * tree.c (tree_node_counts): Likewise.
10126 (tree_node_sizes): Likewise.
10127 (dump_tree_statistics): Use PRIu64 in printf format.
10129 2018-01-12 Martin Liska <mliska@suse.cz>
10131 * Makefile.in: As qsort_chk is implemented in vec.c, add
10132 vec.o to linkage of gencfn-macros.
10133 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10134 passing the info to record_node_allocation_statistics.
10135 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10137 * ggc-common.c (struct ggc_usage): Add operator== and use
10138 it in operator< and compare function.
10139 * mem-stats.h (struct mem_usage): Likewise.
10140 * vec.c (struct vec_usage): Remove operator< and compare
10141 function. Can be simply inherited.
10143 2018-01-12 Martin Jambor <mjambor@suse.cz>
10146 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10147 * tree-ssa-math-opts.c: Include domwalk.h.
10148 (convert_mult_to_fma_1): New function.
10149 (fma_transformation_info): New type.
10150 (fma_deferring_state): Likewise.
10151 (cancel_fma_deferring): New function.
10152 (result_of_phi): Likewise.
10153 (last_fma_candidate_feeds_initial_phi): Likewise.
10154 (convert_mult_to_fma): Added deferring logic, split actual
10155 transformation to convert_mult_to_fma_1.
10156 (math_opts_dom_walker): New type.
10157 (math_opts_dom_walker::after_dom_children): New method, body moved
10158 here from pass_optimize_widening_mul::execute, added deferring logic
10160 (pass_optimize_widening_mul::execute): Moved most of code to
10161 math_opts_dom_walker::after_dom_children.
10162 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10163 * config/i386/i386.c (ix86_option_override_internal): Added
10164 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10166 2018-01-12 Richard Biener <rguenther@suse.de>
10169 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10170 inline instance vars.
10172 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10175 * config/rx/rx.c (rx_is_restricted_memory_address):
10176 Handle SUBREG case.
10178 2018-01-12 Richard Biener <rguenther@suse.de>
10180 PR tree-optimization/80846
10181 * target.def (split_reduction): New target hook.
10182 * targhooks.c (default_split_reduction): New function.
10183 * targhooks.h (default_split_reduction): Declare.
10184 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10185 target requests first reduce vectors by combining low and high
10187 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10188 (get_vectype_for_scalar_type_and_size): Export.
10189 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10190 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10191 * doc/tm.texi: Regenerate.
10192 * config/i386/i386.c (ix86_split_reduction): Implement
10193 TARGET_VECTORIZE_SPLIT_REDUCTION.
10195 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10198 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10199 in PIC mode except for TARGET_VXWORKS_RTP.
10200 * config/sparc/sparc.c: Include cfgrtl.h.
10201 (TARGET_INIT_PIC_REG): Define.
10202 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10203 (sparc_pic_register_p): New predicate.
10204 (sparc_legitimate_address_p): Use it.
10205 (sparc_legitimize_pic_address): Likewise.
10206 (sparc_delegitimize_address): Likewise.
10207 (sparc_mode_dependent_address_p): Likewise.
10208 (gen_load_pcrel_sym): Remove 4th parameter.
10209 (load_got_register): Adjust call to above. Remove obsolete stuff.
10210 (sparc_expand_prologue): Do not call load_got_register here.
10211 (sparc_flat_expand_prologue): Likewise.
10212 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10213 (sparc_use_pseudo_pic_reg): New function.
10214 (sparc_init_pic_reg): Likewise.
10215 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10216 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10218 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10220 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10221 Add item for branch_cost.
10223 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10225 PR rtl-optimization/83565
10226 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10227 not extend the result to a larger mode for rotate operations.
10228 (num_sign_bit_copies1): Likewise.
10230 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10233 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10235 Use values-Xc.o for -pedantic.
10236 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10238 2018-01-12 Martin Liska <mliska@suse.cz>
10241 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10243 (possible_polymorphic_call_targets): Use it.
10244 (ipa_devirt): Likewise.
10246 2018-01-12 Martin Liska <mliska@suse.cz>
10248 * profile-count.h (enum profile_quality): Use 0 as invalid
10249 enum value of profile_quality.
10251 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10253 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10254 -mext-string options.
10256 2018-01-12 Richard Biener <rguenther@suse.de>
10258 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10259 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10260 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10262 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10264 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10266 * configure.ac (--with-long-double-format): Add support for the
10267 configuration option to change the default long double format on
10269 * config.gcc (powerpc*-linux*-*): Likewise.
10270 * configure: Regenerate.
10271 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10272 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10273 used without modification.
10275 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10277 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10278 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10279 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10280 MISC_BUILTIN_SPEC_BARRIER.
10281 (rs6000_init_builtins): Likewise.
10282 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10284 (speculation_barrier): New define_insn.
10285 * doc/extend.texi: Document __builtin_speculation_barrier.
10287 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10290 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10291 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10292 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10294 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10295 integral modes instead of "ss" and "sd".
10296 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10297 vectors with 32-bit and 64-bit elements.
10298 (vecdupssescalarmodesuffix): New mode attribute.
10299 (vec_dup<mode>): Use it.
10301 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10304 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10305 frame if argument is passed on stack.
10307 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10310 * ree.c (combine_reaching_defs): Optimize also
10311 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10312 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10314 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10316 PR middle-end/83189
10317 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10319 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10321 PR middle-end/83718
10322 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10323 after they are computed.
10325 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10327 PR tree-optimization/83695
10328 * gimple-loop-linterchange.cc
10329 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10330 reset cached scev information after interchange.
10331 (pass_linterchange::execute): Remove call to scev_reset_htab.
10333 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10335 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10336 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10337 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10338 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10339 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10340 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10341 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10342 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10343 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10344 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10345 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10346 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10347 (V_lane_reg): Likewise.
10348 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10350 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10351 (vfmal_lane_low<mode>_intrinsic,
10352 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10353 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10354 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10355 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10356 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10357 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10359 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10361 * config/arm/arm-cpus.in (fp16fml): New feature.
10362 (ALL_SIMD): Add fp16fml.
10363 (armv8.2-a): Add fp16fml as an option.
10364 (armv8.3-a): Likewise.
10365 (armv8.4-a): Add fp16fml as part of fp16.
10366 * config/arm/arm.h (TARGET_FP16FML): Define.
10367 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10369 * config/arm/arm-modes.def (V2HF): Define.
10370 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10371 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10372 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10373 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10374 vfmsl_low, vfmsl_high): New set of builtins.
10375 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10376 (vfml_op): New code attribute.
10377 (VFMLHALVES): New int iterator.
10378 (VFML, VFMLSEL): New mode attributes.
10379 (V_reg): Define mapping for V2HF.
10380 (V_hi, V_lo): New mode attributes.
10381 (VF_constraint): Likewise.
10382 (vfml_half, vfml_half_selector): New int attributes.
10383 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10385 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10386 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10388 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10389 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10390 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10391 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10393 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10394 Document new effective target and option set.
10396 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10398 * config/arm/arm-cpus.in (armv8_4): New feature.
10399 (ARMv8_4a): New fgroup.
10400 (armv8.4-a): New arch.
10401 * config/arm/arm-tables.opt: Regenerate.
10402 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10403 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10404 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10405 Add matching rules for -march=armv8.4-a and extensions.
10406 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10408 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10411 * config/rx/rx.md (BW): New mode attribute.
10412 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10414 2018-01-11 Richard Biener <rguenther@suse.de>
10416 PR tree-optimization/83435
10417 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10418 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10419 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10421 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10422 Alan Hayward <alan.hayward@arm.com>
10423 David Sherwood <david.sherwood@arm.com>
10425 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10427 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10428 (aarch64_print_address_internal): Use it to check for a zero offset.
10430 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10431 Alan Hayward <alan.hayward@arm.com>
10432 David Sherwood <david.sherwood@arm.com>
10434 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10435 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10436 Return a poly_int64 rather than a HOST_WIDE_INT.
10437 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10438 rather than a HOST_WIDE_INT.
10439 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10440 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10441 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10442 final_offset from HOST_WIDE_INT to poly_int64.
10443 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10444 to_constant when getting the number of units in an Advanced SIMD
10446 (aarch64_builtin_vectorized_function): Check for a constant number
10448 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10450 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10451 attribute instead of GET_MODE_NUNITS.
10452 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10453 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10454 GET_MODE_SIZE for fixed-size registers.
10455 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10456 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10457 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10458 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10459 (aarch64_print_operand, aarch64_print_address_internal)
10460 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10461 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10462 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10463 Handle polynomial GET_MODE_SIZE.
10464 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10465 wider than SImode without modification.
10466 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10467 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10468 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10469 passing and returning SVE modes.
10470 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10471 rather than GEN_INT.
10472 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10473 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10474 (aarch64_allocate_and_probe_stack_space): Likewise.
10475 (aarch64_layout_frame): Cope with polynomial offsets.
10476 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10477 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10478 polynomial offsets.
10479 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10480 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10481 poly_int64 rather than a HOST_WIDE_INT.
10482 (aarch64_get_separate_components, aarch64_process_components)
10483 (aarch64_expand_prologue, aarch64_expand_epilogue)
10484 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10485 (aarch64_anchor_offset): New function, split out from...
10486 (aarch64_legitimize_address): ...here.
10487 (aarch64_builtin_vectorization_cost): Handle polynomial
10488 TYPE_VECTOR_SUBPARTS.
10489 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10491 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10492 number of elements from the PARALLEL rather than the mode.
10493 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10494 rather than GET_MODE_BITSIZE.
10495 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10496 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10497 (aarch64_expand_vec_perm_const_1): Handle polynomial
10498 d->perm.length () and d->perm elements.
10499 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10500 Apply to_constant to d->perm elements.
10501 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10502 polynomial CONST_VECTOR_NUNITS.
10503 (aarch64_move_pointer): Take amount as a poly_int64 rather
10505 (aarch64_progress_pointer): Avoid temporary variable.
10506 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10507 the mode attribute instead of GET_MODE.
10509 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10510 Alan Hayward <alan.hayward@arm.com>
10511 David Sherwood <david.sherwood@arm.com>
10513 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10514 x exists before using it.
10515 (aarch64_add_constant_internal): Rename to...
10516 (aarch64_add_offset_1): ...this. Replace regnum with separate
10517 src and dest rtxes. Handle the case in which they're different,
10518 including when the offset is zero. Replace scratchreg with an rtx.
10519 Use 2 additions if there is no spare register into which we can
10520 move a 16-bit constant.
10521 (aarch64_add_constant): Delete.
10522 (aarch64_add_offset): Replace reg with separate src and dest
10523 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10524 Use aarch64_add_offset_1.
10525 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10526 an rtx rather than an int. Take the delta as a poly_int64
10527 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10528 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10529 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10530 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10531 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10532 and aarch64_add_sp.
10533 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10534 aarch64_add_constant.
10536 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10538 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10539 Use scalar_float_mode.
10541 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10543 * config/aarch64/aarch64-simd.md
10544 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10545 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10546 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10547 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10548 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10549 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10550 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10551 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10552 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10553 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10555 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10558 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10559 targ_options->x_arm_arch_string is non NULL.
10561 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10563 * config/aarch64/aarch64.h
10564 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10566 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10569 * expmed.c (emit_store_flag_force): Swap if const op0
10570 and change VOIDmode to mode of op0.
10572 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10574 PR rtl-optimization/83761
10575 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10576 than bytes to mode_for_size.
10578 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10580 PR middle-end/83189
10581 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10582 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10585 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10587 PR middle-end/83575
10588 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10589 when in layout mode.
10590 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10591 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10594 2018-01-10 Michael Collison <michael.collison@arm.com>
10596 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10597 * config/aarch64/aarch64-option-extension.def: Add
10598 AARCH64_OPT_EXTENSION of 'fp16fml'.
10599 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10600 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10601 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10602 * config/aarch64/constraints.md (Ui7): New constraint.
10603 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10604 (VFMLA_SEL_W): Ditto.
10607 (VFMLA16_LOW): New int iterator.
10608 (VFMLA16_HIGH): Ditto.
10609 (UNSPEC_FMLAL): New unspec.
10610 (UNSPEC_FMLSL): Ditto.
10611 (UNSPEC_FMLAL2): Ditto.
10612 (UNSPEC_FMLSL2): Ditto.
10613 (f16mac): New code attribute.
10614 * config/aarch64/aarch64-simd-builtins.def
10615 (aarch64_fmlal_lowv2sf): Ditto.
10616 (aarch64_fmlsl_lowv2sf): Ditto.
10617 (aarch64_fmlalq_lowv4sf): Ditto.
10618 (aarch64_fmlslq_lowv4sf): Ditto.
10619 (aarch64_fmlal_highv2sf): Ditto.
10620 (aarch64_fmlsl_highv2sf): Ditto.
10621 (aarch64_fmlalq_highv4sf): Ditto.
10622 (aarch64_fmlslq_highv4sf): Ditto.
10623 (aarch64_fmlal_lane_lowv2sf): Ditto.
10624 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10625 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10626 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10627 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10628 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10629 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10630 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10631 (aarch64_fmlal_lane_highv2sf): Ditto.
10632 (aarch64_fmlsl_lane_highv2sf): Ditto.
10633 (aarch64_fmlal_laneq_highv2sf): Ditto.
10634 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10635 (aarch64_fmlalq_lane_highv4sf): Ditto.
10636 (aarch64_fmlsl_lane_highv4sf): Ditto.
10637 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10638 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10639 * config/aarch64/aarch64-simd.md:
10640 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10641 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10642 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10643 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10644 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10645 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10646 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10647 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10648 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10649 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10650 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10651 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10652 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10653 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10654 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10655 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10656 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10657 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10658 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10659 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10660 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10661 (vfmlsl_low_u32): Ditto.
10662 (vfmlalq_low_u32): Ditto.
10663 (vfmlslq_low_u32): Ditto.
10664 (vfmlal_high_u32): Ditto.
10665 (vfmlsl_high_u32): Ditto.
10666 (vfmlalq_high_u32): Ditto.
10667 (vfmlslq_high_u32): Ditto.
10668 (vfmlal_lane_low_u32): Ditto.
10669 (vfmlsl_lane_low_u32): Ditto.
10670 (vfmlal_laneq_low_u32): Ditto.
10671 (vfmlsl_laneq_low_u32): Ditto.
10672 (vfmlalq_lane_low_u32): Ditto.
10673 (vfmlslq_lane_low_u32): Ditto.
10674 (vfmlalq_laneq_low_u32): Ditto.
10675 (vfmlslq_laneq_low_u32): Ditto.
10676 (vfmlal_lane_high_u32): Ditto.
10677 (vfmlsl_lane_high_u32): Ditto.
10678 (vfmlal_laneq_high_u32): Ditto.
10679 (vfmlsl_laneq_high_u32): Ditto.
10680 (vfmlalq_lane_high_u32): Ditto.
10681 (vfmlslq_lane_high_u32): Ditto.
10682 (vfmlalq_laneq_high_u32): Ditto.
10683 (vfmlslq_laneq_high_u32): Ditto.
10684 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10685 (AARCH64_FL_FOR_ARCH8_4): New.
10686 (AARCH64_ISA_F16FML): New ISA flag.
10687 (TARGET_F16FML): New feature flag for fp16fml.
10688 (doc/invoke.texi): Document new fp16fml option.
10690 2018-01-10 Michael Collison <michael.collison@arm.com>
10692 * config/aarch64/aarch64-builtins.c:
10693 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10694 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10695 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10696 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10697 (AARCH64_ISA_SHA3): New ISA flag.
10698 (TARGET_SHA3): New feature flag for sha3.
10699 * config/aarch64/iterators.md (sha512_op): New int attribute.
10700 (CRYPTO_SHA512): New int iterator.
10701 (UNSPEC_SHA512H): New unspec.
10702 (UNSPEC_SHA512H2): Ditto.
10703 (UNSPEC_SHA512SU0): Ditto.
10704 (UNSPEC_SHA512SU1): Ditto.
10705 * config/aarch64/aarch64-simd-builtins.def
10706 (aarch64_crypto_sha512hqv2di): New builtin.
10707 (aarch64_crypto_sha512h2qv2di): Ditto.
10708 (aarch64_crypto_sha512su0qv2di): Ditto.
10709 (aarch64_crypto_sha512su1qv2di): Ditto.
10710 (aarch64_eor3qv8hi): Ditto.
10711 (aarch64_rax1qv2di): Ditto.
10712 (aarch64_xarqv2di): Ditto.
10713 (aarch64_bcaxqv8hi): Ditto.
10714 * config/aarch64/aarch64-simd.md:
10715 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10716 (aarch64_crypto_sha512su0qv2di): Ditto.
10717 (aarch64_crypto_sha512su1qv2di): Ditto.
10718 (aarch64_eor3qv8hi): Ditto.
10719 (aarch64_rax1qv2di): Ditto.
10720 (aarch64_xarqv2di): Ditto.
10721 (aarch64_bcaxqv8hi): Ditto.
10722 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10723 (vsha512h2q_u64): Ditto.
10724 (vsha512su0q_u64): Ditto.
10725 (vsha512su1q_u64): Ditto.
10726 (veor3q_u16): Ditto.
10727 (vrax1q_u64): Ditto.
10728 (vxarq_u64): Ditto.
10729 (vbcaxq_u16): Ditto.
10730 * config/arm/types.md (crypto_sha512): New type attribute.
10731 (crypto_sha3): Ditto.
10732 (doc/invoke.texi): Document new sha3 option.
10734 2018-01-10 Michael Collison <michael.collison@arm.com>
10736 * config/aarch64/aarch64-builtins.c:
10737 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10738 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10739 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10740 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10741 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10742 (AARCH64_ISA_SM4): New ISA flag.
10743 (TARGET_SM4): New feature flag for sm4.
10744 * config/aarch64/aarch64-simd-builtins.def
10745 (aarch64_sm3ss1qv4si): Ditto.
10746 (aarch64_sm3tt1aq4si): Ditto.
10747 (aarch64_sm3tt1bq4si): Ditto.
10748 (aarch64_sm3tt2aq4si): Ditto.
10749 (aarch64_sm3tt2bq4si): Ditto.
10750 (aarch64_sm3partw1qv4si): Ditto.
10751 (aarch64_sm3partw2qv4si): Ditto.
10752 (aarch64_sm4eqv4si): Ditto.
10753 (aarch64_sm4ekeyqv4si): Ditto.
10754 * config/aarch64/aarch64-simd.md:
10755 (aarch64_sm3ss1qv4si): Ditto.
10756 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10757 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10758 (aarch64_sm4eqv4si): Ditto.
10759 (aarch64_sm4ekeyqv4si): Ditto.
10760 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10761 (sm3part_op): Ditto.
10762 (CRYPTO_SM3TT): Ditto.
10763 (CRYPTO_SM3PART): Ditto.
10764 (UNSPEC_SM3SS1): New unspec.
10765 (UNSPEC_SM3TT1A): Ditto.
10766 (UNSPEC_SM3TT1B): Ditto.
10767 (UNSPEC_SM3TT2A): Ditto.
10768 (UNSPEC_SM3TT2B): Ditto.
10769 (UNSPEC_SM3PARTW1): Ditto.
10770 (UNSPEC_SM3PARTW2): Ditto.
10771 (UNSPEC_SM4E): Ditto.
10772 (UNSPEC_SM4EKEY): Ditto.
10773 * config/aarch64/constraints.md (Ui2): New constraint.
10774 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10775 * config/arm/types.md (crypto_sm3): New type attribute.
10776 (crypto_sm4): Ditto.
10777 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10778 (vsm3tt1aq_u32): Ditto.
10779 (vsm3tt1bq_u32): Ditto.
10780 (vsm3tt2aq_u32): Ditto.
10781 (vsm3tt2bq_u32): Ditto.
10782 (vsm3partw1q_u32): Ditto.
10783 (vsm3partw2q_u32): Ditto.
10784 (vsm4eq_u32): Ditto.
10785 (vsm4ekeyq_u32): Ditto.
10786 (doc/invoke.texi): Document new sm4 option.
10788 2018-01-10 Michael Collison <michael.collison@arm.com>
10790 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10791 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10792 (AARCH64_FL_FOR_ARCH8_4): New.
10793 (AARCH64_FL_V8_4): New flag.
10794 (doc/invoke.texi): Document new armv8.4-a option.
10796 2018-01-10 Michael Collison <michael.collison@arm.com>
10798 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10799 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10800 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10801 * config/aarch64/aarch64-option-extension.def: Add
10802 AARCH64_OPT_EXTENSION of 'sha2'.
10803 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10804 (crypto): Disable sha2 and aes if crypto disabled.
10805 (crypto): Enable aes and sha2 if enabled.
10806 (simd): Disable sha2 and aes if simd disabled.
10807 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10809 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10810 (TARGET_SHA2): New feature flag for sha2.
10811 (TARGET_AES): New feature flag for aes.
10812 * config/aarch64/aarch64-simd.md:
10813 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10814 conditional on TARGET_AES.
10815 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10816 (aarch64_crypto_sha1hsi): Make pattern conditional
10818 (aarch64_crypto_sha1hv4si): Ditto.
10819 (aarch64_be_crypto_sha1hv4si): Ditto.
10820 (aarch64_crypto_sha1su1v4si): Ditto.
10821 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10822 (aarch64_crypto_sha1su0v4si): Ditto.
10823 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10824 (aarch64_crypto_sha256su0v4si): Ditto.
10825 (aarch64_crypto_sha256su1v4si): Ditto.
10826 (doc/invoke.texi): Document new aes and sha2 options.
10828 2018-01-10 Martin Sebor <msebor@redhat.com>
10830 PR tree-optimization/83781
10831 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10834 2018-01-11 Martin Sebor <msebor@gmail.com>
10835 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10837 PR tree-optimization/83501
10838 PR tree-optimization/81703
10840 * tree-ssa-strlen.c (get_string_cst): Rename...
10841 (get_string_len): ...to this. Handle global constants.
10842 (handle_char_store): Adjust.
10844 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10845 Jim Wilson <jimw@sifive.com>
10847 * config/riscv/riscv-protos.h (riscv_output_return): New.
10848 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10849 (riscv_attribute_table, riscv_output_return),
10850 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10851 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10852 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10854 (riscv_expand_prologue): Add early return for naked function.
10855 (riscv_expand_epilogue): Likewise.
10856 (riscv_function_ok_for_sibcall): Return false for naked function.
10857 (riscv_set_current_function): New.
10858 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10859 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10860 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10861 * doc/extend.texi (RISC-V Function Attributes): New.
10863 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10865 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10866 check for 128-bit long double before checking TCmode.
10867 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10868 128-bit long doubles before checking TFmode or TCmode.
10869 (FLOAT128_IBM_P): Likewise.
10871 2018-01-10 Martin Sebor <msebor@redhat.com>
10873 PR tree-optimization/83671
10874 * builtins.c (c_strlen): Unconditionally return zero for the empty
10876 Use -Warray-bounds for warnings.
10877 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10878 for non-constant array indices with COMPONENT_REF, arrays of
10879 arrays, and pointers to arrays.
10880 (gimple_fold_builtin_strlen): Determine and set length range for
10881 non-constant character arrays.
10883 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10885 PR middle-end/81897
10886 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10889 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10891 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10893 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10896 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10897 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10898 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10899 indexed_or_indirect_operand predicate.
10900 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10901 (*vsx_le_perm_load_v8hi): Likewise.
10902 (*vsx_le_perm_load_v16qi): Likewise.
10903 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10904 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10905 (*vsx_le_perm_store_v8hi): Likewise.
10906 (*vsx_le_perm_store_v16qi): Likewise.
10907 (eight unnamed splitters): Likewise.
10909 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10911 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10912 * config/rs6000/emmintrin.h: Likewise.
10913 * config/rs6000/mmintrin.h: Likewise.
10914 * config/rs6000/xmmintrin.h: Likewise.
10916 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10919 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10921 * tree.c (tree_nop_conversion): Return true for location wrapper
10923 (maybe_wrap_with_location): New function.
10924 (selftest::check_strip_nops): New function.
10925 (selftest::test_location_wrappers): New function.
10926 (selftest::tree_c_tests): Call it.
10927 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10928 (maybe_wrap_with_location): New decl.
10929 (EXPR_LOCATION_WRAPPER_P): New macro.
10930 (location_wrapper_p): New inline function.
10931 (tree_strip_any_location_wrapper): New inline function.
10933 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10936 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10937 stack_realign_offset for the largest alignment of stack slot
10939 (ix86_find_max_used_stack_alignment): New function.
10940 (ix86_finalize_stack_frame_flags): Use it. Set
10941 max_used_stack_alignment if we don't realign stack.
10942 * config/i386/i386.h (machine_function): Add
10943 max_used_stack_alignment.
10945 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10947 * config/arm/arm.opt (-mbranch-cost): New option.
10948 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10951 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10954 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10955 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10957 2018-01-10 Richard Biener <rguenther@suse.de>
10960 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10961 early out so it also covers the case where we have a non-NULL
10964 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10966 PR tree-optimization/83753
10967 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10968 for non-strided grouped accesses if the number of elements is 1.
10970 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10973 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10974 * i386.h (TARGET_USE_GATHER): Define.
10975 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10977 2018-01-10 Martin Liska <mliska@suse.cz>
10980 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10981 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10983 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10984 CLEANUP_NO_PARTITIONING is not set.
10986 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10988 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10989 for vectors, as a partial revert of r254296.
10990 * rtl.h (const_vec_p): Delete.
10991 (const_vec_duplicate_p): Don't test for vector CONSTs.
10992 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10993 * expmed.c (make_tree): Likewise.
10996 * common.md (E, F): Use CONSTANT_P instead of checking for
10998 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10999 checking for CONST_VECTOR.
11001 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11003 PR middle-end/83575
11004 * predict.c (force_edge_cold): Handle in more sane way edges
11005 with no prediction.
11007 2018-01-09 Carl Love <cel@us.ibm.com>
11009 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11011 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11012 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11013 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11014 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11015 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11016 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11017 * config/rs6000/rs6000-protos.h: Add extern defition for
11018 rs6000_generate_float2_double_code.
11019 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11021 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11022 (float2_v2df): Add define_expand.
11024 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11027 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11028 op_mode in the force_to_mode call.
11030 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11032 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11033 instead of checking each element individually.
11034 (aarch64_evpc_uzp): Likewise.
11035 (aarch64_evpc_zip): Likewise.
11036 (aarch64_evpc_ext): Likewise.
11037 (aarch64_evpc_rev): Likewise.
11038 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11039 instead of checking each element individually. Return true without
11041 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11042 whether all selected elements come from the same input, instead of
11043 checking each element individually. Remove calls to gen_rtx_REG,
11044 start_sequence and end_sequence and instead assert that no rtl is
11047 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11049 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11050 order of HIGH and CONST checks.
11052 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11054 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11055 if the destination isn't an SSA_NAME.
11057 2018-01-09 Richard Biener <rguenther@suse.de>
11059 PR tree-optimization/83668
11060 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11062 (canonicalize_loop_form): ... here, renamed from ...
11063 (canonicalize_loop_closed_ssa_form): ... this and amended to
11064 swap successor edges for loop exit blocks to make us use
11065 the RPO order we need for initial schedule generation.
11067 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11069 PR tree-optimization/64811
11070 * match.pd: When optimizing comparisons with Inf, avoid
11071 introducing or losing exceptions from comparisons with NaN.
11073 2018-01-09 Martin Liska <mliska@suse.cz>
11076 * asan.c (shadow_mem_size): Add gcc_assert.
11078 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11080 Don't save registers in main().
11083 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11084 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11085 * config/avr/avr.c (avr_set_current_function): Don't error if
11086 naked, OS_task or OS_main are specified at the same time.
11087 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11089 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11091 * common/config/avr/avr-common.c (avr_option_optimization_table):
11092 Switch on -mmain-is-OS_task for optimizing compilations.
11094 2018-01-09 Richard Biener <rguenther@suse.de>
11096 PR tree-optimization/83572
11097 * graphite.c: Include cfganal.h.
11098 (graphite_transform_loops): Connect infinite loops to exit
11099 and remove fake edges at the end.
11101 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11103 * ipa-inline.c (edge_badness): Revert accidental checkin.
11105 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11108 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11109 symbols; not inline clones.
11111 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11114 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11115 hard registers. Formatting fixes.
11117 PR preprocessor/83722
11118 * gcc.c (try_generate_repro): Pass
11119 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11120 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11123 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11124 Kito Cheng <kito.cheng@gmail.com>
11126 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11127 (riscv_leaf_function_p): Delete.
11128 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11130 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11132 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11134 (do_ifelse): New function.
11135 (do_isel): New function.
11136 (do_sub3): New function.
11137 (do_add3): New function.
11138 (do_load_mask_compare): New function.
11139 (do_overlap_load_compare): New function.
11140 (expand_compare_loop): New function.
11141 (expand_block_compare): Call expand_compare_loop() when appropriate.
11142 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11143 option description.
11144 (-mblock-compare-inline-loop-limit): New option.
11146 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11149 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11150 Reverse order of second and third operands in first alternative.
11151 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11152 of first and second elements in UNSPEC_VPERMR vector.
11153 (altivec_expand_vec_perm_le): Likewise.
11155 2018-01-08 Jeff Law <law@redhat.com>
11157 PR rtl-optimizatin/81308
11158 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11159 (process_switch): If group_case_labels makes a change, then set
11161 (pass_convert_switch::execute): If a switch is converted, then
11162 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11164 PR rtl-optimization/81308
11165 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11168 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11170 PR target/83663 - Revert r255946
11171 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11172 generation for cases where splatting a value is not useful.
11173 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11174 across a vec_duplicate and a paradoxical subreg forming a vector
11175 mode to a vec_concat.
11177 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11179 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11180 -march=armv8.3-a variants.
11181 * config/arm/t-multilib: Likewise.
11182 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11184 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11186 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11188 (cceq_ior_compare_complement): Give it a name so I can use it, and
11189 change boolean_or_operator predicate to boolean_operator so it can
11190 be used to generate a crand.
11191 (eqne): New code iterator.
11192 (bd/bd_neg): New code_attrs.
11193 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11194 a single define_insn.
11195 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11196 decrement (bdnzt/bdnzf/bdzt/bdzf).
11197 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11198 with the new names of the branch decrement patterns, and added the
11199 names of the branch decrement conditional patterns.
11201 2018-01-08 Richard Biener <rguenther@suse.de>
11203 PR tree-optimization/83563
11204 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11207 2018-01-08 Richard Biener <rguenther@suse.de>
11209 PR middle-end/83713
11210 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11212 2018-01-08 Richard Biener <rguenther@suse.de>
11214 PR tree-optimization/83685
11215 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11216 references to abnormals.
11218 2018-01-08 Richard Biener <rguenther@suse.de>
11221 * dwarf2out.c (output_indirect_strings): Handle empty
11222 skeleton_debug_str_hash.
11223 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11225 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11227 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11228 (emit_store_direct): Likewise.
11229 (arc_trampoline_adjust_address): Likewise.
11230 (arc_asm_trampoline_template): New function.
11231 (arc_initialize_trampoline): Use asm_trampoline_template.
11232 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11233 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11234 * config/arc/arc.md (flush_icache): Delete pattern.
11236 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11238 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11239 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11242 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11245 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11246 by not USED_FOR_TARGET.
11247 (make_pass_resolve_sw_modes): Likewise.
11249 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11251 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11254 2018-01-08 Richard Biener <rguenther@suse.de>
11256 PR middle-end/83580
11257 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11259 2018-01-08 Richard Biener <rguenther@suse.de>
11261 PR middle-end/83517
11262 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11264 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11266 PR middle-end/81897
11267 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11268 basic blocks with a small number of successors.
11269 (convert_control_dep_chain_into_preds): Improve handling of
11271 (dump_predicates): Split apart into...
11272 (dump_pred_chain): ...here...
11273 (dump_pred_info): ...and here.
11274 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11275 (can_chain_union_be_invalidated_p): Improve check for invalidation
11277 (uninit_uses_cannot_happen): Avoid unnecessary if
11278 convert_control_dep_chain_into_preds yielded nothing.
11280 2018-01-06 Martin Sebor <msebor@redhat.com>
11282 PR tree-optimization/83640
11283 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11284 subtracting negative offset from size.
11285 (builtin_access::overlap): Adjust offset bounds of the access to fall
11286 within the size of the object if possible.
11288 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11290 PR rtl-optimization/83699
11291 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11292 extract_bit_field_as_subreg to cases in which the extracted
11293 value is also a vector.
11295 * lra-constraints.c (process_alt_operands): Test for the equivalence
11296 substitutions when detecting a possible reload cycle.
11298 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11301 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11302 by default if flag_selective_schedling{,2}. Formatting fixes.
11304 PR rtl-optimization/83682
11305 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11306 if it has non-VECTOR_MODE element mode.
11307 (vec_duplicate_p): Likewise.
11309 PR middle-end/83694
11310 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11311 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11313 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11316 * config/i386/i386-builtin.def
11317 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11318 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11319 Require also OPTION_MASK_ISA_AVX512F in addition to
11320 OPTION_MASK_ISA_GFNI.
11321 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11322 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11323 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11324 to OPTION_MASK_ISA_GFNI.
11325 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11326 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11327 OPTION_MASK_ISA_AVX512BW.
11328 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11329 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11330 addition to OPTION_MASK_ISA_GFNI.
11331 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11332 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11333 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11334 to OPTION_MASK_ISA_GFNI.
11335 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11336 a requirement for all ISAs rather than any of them with a few
11338 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11340 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11341 bitmasks to be enabled with 3 exceptions, instead of requiring any
11342 enabled ISA with lots of exceptions.
11343 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11344 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11345 Change avx512bw in isa attribute to avx512f.
11346 * config/i386/sgxintrin.h: Add license boilerplate.
11347 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11348 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11349 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11350 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11352 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11353 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11354 temporarily sse2 rather than sse if not enabled already.
11357 * config/i386/sse.md (VI248_VLBW): Rename to ...
11358 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11359 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11360 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11361 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11362 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11363 mode iterator instead of VI248_VLBW.
11365 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11367 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11368 (record_modified): Skip clobbers; add debug output.
11369 (param_change_prob): Use sreal frequencies.
11371 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11373 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11374 punt for user-aligned variables.
11376 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11378 * tree-chrec.c (chrec_contains_symbols): Return true for
11381 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11384 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11385 of (x|y) == x for BICS pattern.
11387 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11389 PR tree-optimization/83605
11390 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11391 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11394 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11396 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11397 * config/epiphany/rtems.h: New file.
11399 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11400 Uros Bizjak <ubizjak@gmail.com>
11403 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11404 QIreg_operand instead of register_operand predicate.
11405 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11406 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11407 comments instead of -fmitigate[-_]rop.
11409 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11412 * cgraphunit.c (symbol_table::compile): Switch to text_section
11413 before calling assembly_start debug hook.
11414 * run-rtl-passes.c (run_rtl_passes): Likewise.
11417 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11419 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11420 range_int_cst_p rather than !symbolic_range_p before calling
11421 extract_range_from_multiplicative_op_1.
11423 2018-01-04 Jeff Law <law@redhat.com>
11425 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11426 redundant test in assertion.
11428 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11430 * doc/rtl.texi: Document machine_mode wrapper classes.
11432 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11434 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11435 using tree_to_uhwi.
11437 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11439 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11440 the VEC_PERM_EXPR fold to fail.
11442 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11445 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11446 to switched_sections.
11448 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11451 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11452 test for d.testing.
11454 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11457 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11458 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11460 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11463 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11464 is BLKmode and bitpos not zero or mode change is needed.
11466 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11469 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11472 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11475 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11476 instead of MULT rtx. Update all corresponding splitters.
11477 (*saddl_se): Ditto.
11478 (*ssub<modesuffix>): Ditto.
11479 (*ssubl_se): Ditto.
11480 (*cmp_sadd_di): Update split patterns.
11481 (*cmp_sadd_si): Ditto.
11482 (*cmp_sadd_sidi): Ditto.
11483 (*cmp_ssub_di): Ditto.
11484 (*cmp_ssub_si): Ditto.
11485 (*cmp_ssub_sidi): Ditto.
11486 * config/alpha/predicates.md (const23_operand): New predicate.
11487 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11488 Look for ASHIFT, not MULT inner operand.
11489 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11491 2018-01-04 Martin Liska <mliska@suse.cz>
11493 PR gcov-profile/83669
11494 * gcov.c (output_intermediate_file): Add version to intermediate
11496 * doc/gcov.texi: Document new field 'version' in intermediate
11497 file format. Fix location of '-k' option of gcov command.
11499 2018-01-04 Martin Liska <mliska@suse.cz>
11502 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11504 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11506 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11508 2018-01-03 Martin Sebor <msebor@redhat.com>
11510 PR tree-optimization/83655
11511 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11512 checking calls with invalid arguments.
11514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11516 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11517 (vectorizable_mask_load_store): Delete.
11518 (vectorizable_call): Return false for masked loads and stores.
11519 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11520 instead of gimple_assign_rhs1.
11521 (vectorizable_load): Handle IFN_MASK_LOAD.
11522 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11524 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11526 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11528 (vectorizable_mask_load_store): ...here.
11529 (vectorizable_load): ...and here.
11531 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11533 * tree-vect-stmts.c (vect_build_all_ones_mask)
11534 (vect_build_zero_merge_argument): New functions, split out from...
11535 (vectorizable_load): ...here.
11537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11539 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11541 (vectorizable_mask_load_store): ...here.
11542 (vectorizable_store): ...and here.
11544 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11546 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11548 (vectorizable_mask_load_store): ...here.
11550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11552 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11553 (vect_model_store_cost): Take a vec_load_store_type instead of a
11555 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11556 (vect_model_store_cost): Take a vec_load_store_type instead of a
11558 (vectorizable_mask_load_store): Update accordingly.
11559 (vectorizable_store): Likewise.
11560 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11564 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11565 IFN_MASK_LOAD calls here rather than...
11566 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11569 Alan Hayward <alan.hayward@arm.com>
11570 David Sherwood <david.sherwood@arm.com>
11572 * expmed.c (extract_bit_field_1): For vector extracts,
11573 fall back to extract_bit_field_as_subreg if vec_extract
11576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11577 Alan Hayward <alan.hayward@arm.com>
11578 David Sherwood <david.sherwood@arm.com>
11580 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11581 they are variable or constant sized.
11582 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11583 slots for constant-sized data.
11585 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11586 Alan Hayward <alan.hayward@arm.com>
11587 David Sherwood <david.sherwood@arm.com>
11589 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11590 handling COND_EXPRs with boolean comparisons, try to find a better
11591 basis for the mask type than the boolean itself.
11593 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11595 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11596 is calculated and how it can be overridden.
11597 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11598 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11600 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11604 Alan Hayward <alan.hayward@arm.com>
11605 David Sherwood <david.sherwood@arm.com>
11607 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11608 Remove the mode argument.
11609 (aarch64_simd_valid_immediate): Remove the mode and inverse
11611 * config/aarch64/iterators.md (bitsize): New iterator.
11612 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11613 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11614 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11615 aarch64_simd_valid_immediate.
11616 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11617 (aarch64_reg_or_bic_imm): Likewise.
11618 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11619 with an insn_type enum and msl with a modifier_type enum.
11620 Replace element_width with a scalar_mode. Change the shift
11621 to unsigned int. Add constructors for scalar_float_mode and
11622 scalar_int_mode elements.
11623 (aarch64_vect_float_const_representable_p): Delete.
11624 (aarch64_can_const_movi_rtx_p)
11625 (aarch64_simd_scalar_immediate_valid_for_move)
11626 (aarch64_simd_make_constant): Update call to
11627 aarch64_simd_valid_immediate.
11628 (aarch64_advsimd_valid_immediate_hs): New function.
11629 (aarch64_advsimd_valid_immediate): Likewise.
11630 (aarch64_simd_valid_immediate): Remove mode and inverse
11631 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11632 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11633 and aarch64_float_const_representable_p on the result.
11634 (aarch64_output_simd_mov_immediate): Remove mode argument.
11635 Update call to aarch64_simd_valid_immediate and use of
11636 simd_immediate_info.
11637 (aarch64_output_scalar_simd_mov_immediate): Update call
11640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11641 Alan Hayward <alan.hayward@arm.com>
11642 David Sherwood <david.sherwood@arm.com>
11644 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11645 (mode_nunits): Likewise CONST_MODE_NUNITS.
11646 * machmode.def (ADJUST_NUNITS): Document.
11647 * genmodes.c (mode_data::need_nunits_adj): New field.
11648 (blank_mode): Update accordingly.
11649 (adj_nunits): New variable.
11650 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11652 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11653 listed in adj_nunits.
11654 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11655 listed in adj_nunits. Don't emit case statements for such modes.
11656 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11657 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11658 nothing if adj_nunits is nonnull.
11659 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11660 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11661 (emit_mode_fbit): Update use of print_maybe_const_decl.
11662 (emit_move_size): Likewise. Treat the array as non-const
11664 (emit_mode_adjustments): Handle adj_nunits.
11666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11668 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11669 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11670 (VECTOR_MODES): Use it.
11671 (make_vector_modes): Take the prefix as an argument.
11673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11674 Alan Hayward <alan.hayward@arm.com>
11675 David Sherwood <david.sherwood@arm.com>
11677 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11678 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11679 for MODE_VECTOR_BOOL.
11680 * machmode.def (VECTOR_BOOL_MODE): Document.
11681 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11682 (make_vector_bool_mode): New function.
11683 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11685 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11686 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11688 * stor-layout.c (int_mode_for_mode): Likewise.
11689 * tree.c (build_vector_type_for_mode): Likewise.
11690 * varasm.c (output_constant_pool_2): Likewise.
11691 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11692 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11693 for MODE_VECTOR_BOOL.
11694 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11695 of mode class checks.
11696 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11697 instead of a list of mode class checks.
11698 (expand_vector_scalar_condition): Likewise.
11699 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11702 Alan Hayward <alan.hayward@arm.com>
11703 David Sherwood <david.sherwood@arm.com>
11705 * machmode.h (mode_size): Change from unsigned short to
11707 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11708 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11709 or if measurement_type is not polynomial.
11710 (fixed_size_mode::includes_p): Check for constant-sized modes.
11711 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11712 return a poly_uint16 rather than an unsigned short.
11713 (emit_mode_size): Change the type of mode_size from unsigned short
11714 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11715 (emit_mode_adjustments): Cope with polynomial vector sizes.
11716 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11718 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11720 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11721 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11722 * caller-save.c (setup_save_areas): Likewise.
11723 (replace_reg_with_saved_mem): Likewise.
11724 * calls.c (emit_library_call_value_1): Likewise.
11725 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11726 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11727 (gen_lowpart_for_combine): Likewise.
11728 * convert.c (convert_to_integer_1): Likewise.
11729 * cse.c (equiv_constant, cse_insn): Likewise.
11730 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11731 (cselib_subst_to_values): Likewise.
11732 * dce.c (word_dce_process_block): Likewise.
11733 * df-problems.c (df_word_lr_mark_ref): Likewise.
11734 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11735 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11736 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11737 (rtl_for_decl_location): Likewise.
11738 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11739 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11740 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11741 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11742 (expand_expr_real_1): Likewise.
11743 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11744 (pad_below): Likewise.
11745 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11746 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11747 * ira.c (get_subreg_tracking_sizes): Likewise.
11748 * ira-build.c (ira_create_allocno_objects): Likewise.
11749 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11750 (ira_sort_regnos_for_alter_reg): Likewise.
11751 * ira-costs.c (record_operand_costs): Likewise.
11752 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11753 (resolve_simple_move): Likewise.
11754 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11755 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11756 (lra_constraints): Likewise.
11757 (CONST_POOL_OK_P): Reject variable-sized modes.
11758 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11759 (add_pseudo_to_slot, lra_spill): Likewise.
11760 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11761 * optabs-query.c (get_best_extraction_insn): Likewise.
11762 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11763 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11764 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11765 * recog.c (offsettable_address_addr_space_p): Likewise.
11766 * regcprop.c (maybe_mode_change): Likewise.
11767 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11768 * regrename.c (build_def_use): Likewise.
11769 * regstat.c (dump_reg_info): Likewise.
11770 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11771 (find_reloads, find_reloads_subreg_address): Likewise.
11772 * reload1.c (eliminate_regs_1): Likewise.
11773 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11774 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11775 (simplify_binary_operation_1, simplify_subreg): Likewise.
11776 * targhooks.c (default_function_arg_padding): Likewise.
11777 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11778 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11779 (verify_gimple_assign_ternary): Likewise.
11780 * tree-inline.c (estimate_move_cost): Likewise.
11781 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11782 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11783 (get_address_cost_ainc): Likewise.
11784 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11785 (vect_supportable_dr_alignment): Likewise.
11786 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11787 (vectorizable_reduction): Likewise.
11788 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11789 (vectorizable_operation, vectorizable_load): Likewise.
11790 * tree.c (build_same_sized_truth_vector_type): Likewise.
11791 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11792 * var-tracking.c (emit_note_insn_var_location): Likewise.
11793 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11794 (ADDR_VEC_ALIGN): Likewise.
11796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11797 Alan Hayward <alan.hayward@arm.com>
11798 David Sherwood <david.sherwood@arm.com>
11800 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11802 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11803 or if measurement_type is polynomial.
11804 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11805 * combine.c (make_extraction): Likewise.
11806 * dse.c (find_shift_sequence): Likewise.
11807 * dwarf2out.c (mem_loc_descriptor): Likewise.
11808 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11809 (extract_bit_field, extract_low_bits): Likewise.
11810 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11811 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11812 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11813 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11814 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11815 * reload.c (find_reloads): Likewise.
11816 * reload1.c (alter_reg): Likewise.
11817 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11818 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11819 * tree-if-conv.c (predicate_mem_writes): Likewise.
11820 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11821 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11822 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11823 * valtrack.c (dead_debug_insert_temp): Likewise.
11824 * varasm.c (mergeable_constant_section): Likewise.
11825 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11828 Alan Hayward <alan.hayward@arm.com>
11829 David Sherwood <david.sherwood@arm.com>
11831 * expr.c (expand_assignment): Cope with polynomial mode sizes
11832 when assigning to a CONCAT.
11834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11835 Alan Hayward <alan.hayward@arm.com>
11836 David Sherwood <david.sherwood@arm.com>
11838 * machmode.h (mode_precision): Change from unsigned short to
11840 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11842 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11843 or if measurement_type is not polynomial.
11844 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11845 in which the mode is already known to be a scalar_int_mode.
11846 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11847 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11849 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11850 for GET_MODE_PRECISION.
11851 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11852 for GET_MODE_PRECISION.
11853 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11855 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11856 (expand_field_assignment, make_extraction): Likewise.
11857 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11858 (get_last_value): Likewise.
11859 * convert.c (convert_to_integer_1): Likewise.
11860 * cse.c (cse_insn): Likewise.
11861 * expr.c (expand_expr_real_1): Likewise.
11862 * lra-constraints.c (simplify_operand_subreg): Likewise.
11863 * optabs-query.c (can_atomic_load_p): Likewise.
11864 * optabs.c (expand_atomic_load): Likewise.
11865 (expand_atomic_store): Likewise.
11866 * ree.c (combine_reaching_defs): Likewise.
11867 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11868 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11869 * tree.h (type_has_mode_precision_p): Likewise.
11870 * ubsan.c (instrument_si_overflow): Likewise.
11872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11873 Alan Hayward <alan.hayward@arm.com>
11874 David Sherwood <david.sherwood@arm.com>
11876 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11877 polynomial numbers of units.
11878 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11879 (valid_vector_subparts_p): New function.
11880 (build_vector_type): Remove temporary shim and take the number
11881 of units as a poly_uint64 rather than an int.
11882 (build_opaque_vector_type): Take the number of units as a
11883 poly_uint64 rather than an int.
11884 * tree.c (build_vector_from_ctor): Handle polynomial
11885 TYPE_VECTOR_SUBPARTS.
11886 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11887 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11888 (build_vector_from_val): If the number of units is variable,
11889 use build_vec_duplicate_cst for constant operands and
11890 VEC_DUPLICATE_EXPR otherwise.
11891 (make_vector_type): Remove temporary is_constant ().
11892 (build_vector_type, build_opaque_vector_type): Take the number of
11893 units as a poly_uint64 rather than an int.
11894 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11896 * cfgexpand.c (expand_debug_expr): Likewise.
11897 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11898 (store_constructor, expand_expr_real_1): Likewise.
11899 (const_scalar_mask_from_tree): Likewise.
11900 * fold-const-call.c (fold_const_reduction): Likewise.
11901 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11902 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11903 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11904 (fold_relational_const): Likewise.
11905 (native_interpret_vector): Likewise. Change the size from an
11906 int to an unsigned int.
11907 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11908 TYPE_VECTOR_SUBPARTS.
11909 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11910 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11911 duplicating a non-constant operand into a variable-length vector.
11912 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11913 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11914 * ipa-icf.c (sem_variable::equals): Likewise.
11915 * match.pd: Likewise.
11916 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11917 * print-tree.c (print_node): Likewise.
11918 * stor-layout.c (layout_type): Likewise.
11919 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11920 * tree-cfg.c (verify_gimple_comparison): Likewise.
11921 (verify_gimple_assign_binary): Likewise.
11922 (verify_gimple_assign_ternary): Likewise.
11923 (verify_gimple_assign_single): Likewise.
11924 * tree-pretty-print.c (dump_generic_node): Likewise.
11925 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11926 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11927 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11928 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11929 (vect_shift_permute_load_chain): Likewise.
11930 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11931 (expand_vector_condition, optimize_vector_constructor): Likewise.
11932 (lower_vec_perm, get_compute_type): Likewise.
11933 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11934 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11935 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11936 (vect_recog_mask_conversion_pattern): Likewise.
11937 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11938 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11939 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11940 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11941 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11942 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11943 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11944 (supportable_widening_operation): Likewise.
11945 (supportable_narrowing_operation): Likewise.
11946 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11948 * varasm.c (output_constant): Likewise.
11950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11951 Alan Hayward <alan.hayward@arm.com>
11952 David Sherwood <david.sherwood@arm.com>
11954 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11955 so that both the length == 3 and length != 3 cases set up their
11956 own permute vectors. Add comments explaining why we know the
11957 number of elements is constant.
11958 (vect_permute_load_chain): 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 * machmode.h (mode_nunits): Change from unsigned char to
11966 (ONLY_FIXED_SIZE_MODES): New macro.
11967 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11968 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11969 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11971 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11972 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11973 or if measurement_type is not polynomial.
11974 * genmodes.c (ZERO_COEFFS): New macro.
11975 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11977 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11978 Use ZERO_COEFFS when emitting initializers.
11979 * data-streamer.h (bp_pack_poly_value): New function.
11980 (bp_unpack_poly_value): Likewise.
11981 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11982 for GET_MODE_NUNITS.
11983 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11984 for GET_MODE_NUNITS.
11985 * tree.c (make_vector_type): Remove temporary shim and make
11986 the real function take the number of units as a poly_uint64
11987 rather than an int.
11988 (build_vector_type_for_mode): Handle polynomial nunits.
11989 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11990 * emit-rtl.c (const_vec_series_p_1): Likewise.
11991 (gen_rtx_CONST_VECTOR): Likewise.
11992 * fold-const.c (test_vec_duplicate_folding): Likewise.
11993 * genrecog.c (validate_pattern): Likewise.
11994 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11995 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11996 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11997 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11998 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11999 * rtlanal.c (subreg_get_info): Likewise.
12000 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12001 (vect_grouped_load_supported): Likewise.
12002 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12003 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12004 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12005 (simplify_const_unary_operation, simplify_binary_operation_1)
12006 (simplify_const_binary_operation, simplify_ternary_operation)
12007 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12008 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12009 instead of CONST_VECTOR_NUNITS.
12010 * varasm.c (output_constant_pool_2): Likewise.
12011 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12012 explicit-encoded elements in the XVEC for variable-length vectors.
12014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12016 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12019 Alan Hayward <alan.hayward@arm.com>
12020 David Sherwood <david.sherwood@arm.com>
12022 * coretypes.h (fixed_size_mode): Declare.
12023 (fixed_size_mode_pod): New typedef.
12024 * builtins.h (target_builtins::x_apply_args_mode)
12025 (target_builtins::x_apply_result_mode): Change type to
12026 fixed_size_mode_pod.
12027 * builtins.c (apply_args_size, apply_result_size, result_vector)
12028 (expand_builtin_apply_args_1, expand_builtin_apply)
12029 (expand_builtin_return): Update accordingly.
12031 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12033 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12034 * cselib.c (cselib_hash_rtx): Likewise.
12035 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12036 CONST_VECTOR encoding.
12038 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12039 Jeff Law <law@redhat.com>
12042 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12043 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12044 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12045 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12048 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12049 explicitly probe *sp in a noreturn function if there were any callee
12050 register saves or frame pointer is needed.
12052 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12055 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12056 BLKmode for ternary, binary or unary expressions.
12059 * var-tracking.c (delete_vta_debug_insn): New inline function.
12060 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12061 insns from get_insns () to NULL instead of each bb separately.
12062 Use delete_vta_debug_insn. No longer static.
12063 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12064 delete_vta_debug_insns callers.
12065 * rtl.h (delete_vta_debug_insns): Declare.
12066 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12067 instead of variable_tracking_main.
12069 2018-01-03 Martin Sebor <msebor@redhat.com>
12071 PR tree-optimization/83603
12072 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12073 arguments past the endof the argument list in functions declared
12074 without a prototype.
12075 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12076 Avoid checking when arguments are null.
12078 2018-01-03 Martin Sebor <msebor@redhat.com>
12081 * doc/extend.texi (attribute const): Fix a typo.
12082 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12083 issuing -Wsuggest-attribute for void functions.
12085 2018-01-03 Martin Sebor <msebor@redhat.com>
12087 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12088 offset_int::from instead of wide_int::to_shwi.
12089 (maybe_diag_overlap): Remove assertion.
12090 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12091 * gimple-ssa-sprintf.c (format_directive): Same.
12092 (parse_directive): Same.
12093 (sprintf_dom_walker::compute_format_length): Same.
12094 (try_substitute_return_value): Same.
12096 2018-01-03 Jeff Law <law@redhat.com>
12098 PR middle-end/83654
12099 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12100 non-constant residual for zero at runtime and avoid probing in
12101 that case. Reorganize code for trailing problem to mirror handling
12104 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12106 PR tree-optimization/83501
12107 * tree-ssa-strlen.c (get_string_cst): New.
12108 (handle_char_store): Call get_string_cst.
12110 2018-01-03 Martin Liska <mliska@suse.cz>
12112 PR tree-optimization/83593
12113 * tree-ssa-strlen.c: Include tree-cfg.h.
12114 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12115 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12116 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12118 (strlen_dom_walker::before_dom_children): Call
12119 gimple_purge_dead_eh_edges. Dump tranformation with details
12121 (strlen_dom_walker::before_dom_children): Update call by adding
12122 new argument cleanup_eh.
12123 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12125 2018-01-03 Martin Liska <mliska@suse.cz>
12128 * cif-code.def (VARIADIC_THUNK): New enum value.
12129 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12132 2018-01-03 Jan Beulich <jbeulich@suse.com>
12134 * sse.md (mov<mode>_internal): Tighten condition for when to use
12135 vmovdqu<ssescalarsize> for TI and OI modes.
12137 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12139 Update copyright years.
12141 2018-01-03 Martin Liska <mliska@suse.cz>
12144 * ipa-visibility.c (function_and_variable_visibility): Skip
12145 functions with noipa attribure.
12147 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12149 * gcc.c (process_command): Update copyright notice dates.
12150 * gcov-dump.c (print_version): Ditto.
12151 * gcov.c (print_version): Ditto.
12152 * gcov-tool.c (print_version): Ditto.
12153 * gengtype.c (create_file): Ditto.
12154 * doc/cpp.texi: Bump @copying's copyright year.
12155 * doc/cppinternals.texi: Ditto.
12156 * doc/gcc.texi: Ditto.
12157 * doc/gccint.texi: Ditto.
12158 * doc/gcov.texi: Ditto.
12159 * doc/install.texi: Ditto.
12160 * doc/invoke.texi: Ditto.
12162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12164 * vector-builder.h (vector_builder::m_full_nelts): Change from
12165 unsigned int to poly_uint64.
12166 (vector_builder::full_nelts): Update prototype accordingly.
12167 (vector_builder::new_vector): Likewise.
12168 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12169 (vector_builder::operator ==): Likewise.
12170 (vector_builder::finalize): Likewise.
12171 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12172 Take the number of elements as a poly_uint64 rather than an
12174 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12175 from unsigned int to poly_uint64.
12176 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12177 (vec_perm_indices::new_vector): Likewise.
12178 (vec_perm_indices::length): Likewise.
12179 (vec_perm_indices::nelts_per_input): Likewise.
12180 (vec_perm_indices::input_nelts): Likewise.
12181 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12182 number of elements per input as a poly_uint64 rather than an
12183 unsigned int. Use the original encoding for variable-length
12184 vectors, rather than clamping each individual element.
12185 For the second and subsequent elements in each pattern,
12186 clamp the step and base before clamping their sum.
12187 (vec_perm_indices::series_p): Handle polynomial element counts.
12188 (vec_perm_indices::all_in_range_p): Likewise.
12189 (vec_perm_indices_to_tree): Likewise.
12190 (vec_perm_indices_to_rtx): Likewise.
12191 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12192 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12193 (tree_vector_builder::new_binary_operation): Handle polynomial
12194 element counts. Return false if we need to know the number
12195 of elements at compile time.
12196 * fold-const.c (fold_vec_perm): Punt if the number of elements
12197 isn't known at compile time.
12199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12201 * vec-perm-indices.h (vec_perm_builder): Change element type
12202 from HOST_WIDE_INT to poly_int64.
12203 (vec_perm_indices::element_type): Update accordingly.
12204 (vec_perm_indices::clamp): Handle polynomial element_types.
12205 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12206 (vec_perm_indices::all_in_range_p): Likewise.
12207 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12209 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12210 polynomial vec_perm_indices element types.
12211 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12212 * fold-const.c (fold_vec_perm): Likewise.
12213 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12214 * tree-vect-generic.c (lower_vec_perm): Likewise.
12215 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12216 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12217 element type to HOST_WIDE_INT.
12219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12220 Alan Hayward <alan.hayward@arm.com>
12221 David Sherwood <david.sherwood@arm.com>
12223 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12224 rather than an int. Use plus_constant.
12225 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12226 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12228 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12229 Alan Hayward <alan.hayward@arm.com>
12230 David Sherwood <david.sherwood@arm.com>
12232 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12233 a HOST_WIDE_INT to a poly_int64.
12235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12236 Alan Hayward <alan.hayward@arm.com>
12237 David Sherwood <david.sherwood@arm.com>
12239 * calls.c (load_register_parameters): Cope with polynomial
12240 mode sizes. Require a constant size for BLKmode parameters
12241 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12242 forces a parameter to be padded at the lsb end in order to
12243 fill a complete number of words, require the parameter size
12244 to be ordered wrt UNITS_PER_WORD.
12246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12247 Alan Hayward <alan.hayward@arm.com>
12248 David Sherwood <david.sherwood@arm.com>
12250 * reload1.c (spill_stack_slot_width): Change element type
12251 from unsigned int to poly_uint64_pod.
12252 (alter_reg): Treat mode sizes as polynomial.
12254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12255 Alan Hayward <alan.hayward@arm.com>
12256 David Sherwood <david.sherwood@arm.com>
12258 * reload.c (complex_word_subreg_p): New function.
12259 (reload_inner_reg_of_subreg, push_reload): Use it.
12261 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12262 Alan Hayward <alan.hayward@arm.com>
12263 David Sherwood <david.sherwood@arm.com>
12265 * lra-constraints.c (process_alt_operands): Reject matched
12266 operands whose sizes aren't ordered.
12267 (match_reload): Refer to this check here.
12269 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12270 Alan Hayward <alan.hayward@arm.com>
12271 David Sherwood <david.sherwood@arm.com>
12273 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12274 that the mode size is in the set {1, 2, 4, 8, 16}.
12276 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12277 Alan Hayward <alan.hayward@arm.com>
12278 David Sherwood <david.sherwood@arm.com>
12280 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12281 Use plus_constant instead of gen_rtx_PLUS.
12283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12284 Alan Hayward <alan.hayward@arm.com>
12285 David Sherwood <david.sherwood@arm.com>
12287 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12288 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12289 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12290 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12291 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12292 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12293 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12294 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12295 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12296 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12298 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12299 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12300 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12301 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12302 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12303 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12304 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12305 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12306 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12307 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12309 * expr.c (emit_move_resolve_push): Treat the input and result
12310 of PUSH_ROUNDING as a poly_int64.
12311 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12312 (emit_push_insn): Likewise.
12313 * lra-eliminations.c (mark_not_eliminable): Likewise.
12314 * recog.c (push_operand): Likewise.
12315 * reload1.c (elimination_effects): Likewise.
12316 * rtlanal.c (nonzero_bits1): Likewise.
12317 * calls.c (store_one_arg): Likewise. Require the padding to be
12318 known at compile time.
12320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12321 Alan Hayward <alan.hayward@arm.com>
12322 David Sherwood <david.sherwood@arm.com>
12324 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12325 Use plus_constant instead of gen_rtx_PLUS.
12327 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12328 Alan Hayward <alan.hayward@arm.com>
12329 David Sherwood <david.sherwood@arm.com>
12331 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12332 rather than an int.
12334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12335 Alan Hayward <alan.hayward@arm.com>
12336 David Sherwood <david.sherwood@arm.com>
12338 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12339 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12340 via stack temporaries. Treat the mode size as polynomial too.
12342 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12343 Alan Hayward <alan.hayward@arm.com>
12344 David Sherwood <david.sherwood@arm.com>
12346 * expr.c (expand_expr_real_2): When handling conversions involving
12347 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12348 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12349 as a poly_uint64 too.
12351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12352 Alan Hayward <alan.hayward@arm.com>
12353 David Sherwood <david.sherwood@arm.com>
12355 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12358 Alan Hayward <alan.hayward@arm.com>
12359 David Sherwood <david.sherwood@arm.com>
12361 * combine.c (can_change_dest_mode): Handle polynomial
12362 REGMODE_NATURAL_SIZE.
12363 * expmed.c (store_bit_field_1): Likewise.
12364 * expr.c (store_constructor): Likewise.
12365 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12366 and polynomial REGMODE_NATURAL_SIZE.
12367 (gen_lowpart_common): Likewise.
12368 * reginfo.c (record_subregs_of_mode): Likewise.
12369 * rtlanal.c (read_modify_subreg_p): Likewise.
12371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12372 Alan Hayward <alan.hayward@arm.com>
12373 David Sherwood <david.sherwood@arm.com>
12375 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12376 numbers of elements.
12378 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12379 Alan Hayward <alan.hayward@arm.com>
12380 David Sherwood <david.sherwood@arm.com>
12382 * match.pd: Cope with polynomial numbers of vector elements.
12384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12385 Alan Hayward <alan.hayward@arm.com>
12386 David Sherwood <david.sherwood@arm.com>
12388 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12389 in a POINTER_PLUS_EXPR.
12391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12392 Alan Hayward <alan.hayward@arm.com>
12393 David Sherwood <david.sherwood@arm.com>
12395 * omp-simd-clone.c (simd_clone_subparts): New function.
12396 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12397 (ipa_simd_modify_function_body): Likewise.
12399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12400 Alan Hayward <alan.hayward@arm.com>
12401 David Sherwood <david.sherwood@arm.com>
12403 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12404 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12405 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12406 (expand_vector_condition, vector_element): Likewise.
12407 (subparts_gt): New function.
12408 (get_compute_type): Use subparts_gt.
12409 (count_type_subparts): Delete.
12410 (expand_vector_operations_1): Use subparts_gt instead of
12411 count_type_subparts.
12413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12414 Alan Hayward <alan.hayward@arm.com>
12415 David Sherwood <david.sherwood@arm.com>
12417 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12418 (vect_compile_time_alias): ...this new function. Do the calculation
12419 on poly_ints rather than trees.
12420 (vect_prune_runtime_alias_test_list): Update call accordingly.
12422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12423 Alan Hayward <alan.hayward@arm.com>
12424 David Sherwood <david.sherwood@arm.com>
12426 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12428 (vect_schedule_slp_instance): Likewise.
12430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12431 Alan Hayward <alan.hayward@arm.com>
12432 David Sherwood <david.sherwood@arm.com>
12434 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12435 constant and extern definitions for variable-length vectors.
12436 (vect_get_constant_vectors): Note that the number of units
12437 is known to be constant.
12439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12440 Alan Hayward <alan.hayward@arm.com>
12441 David Sherwood <david.sherwood@arm.com>
12443 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12444 of units as polynomial. Choose between WIDE and NARROW based
12447 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12448 Alan Hayward <alan.hayward@arm.com>
12449 David Sherwood <david.sherwood@arm.com>
12451 * tree-vect-stmts.c (simd_clone_subparts): New function.
12452 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12455 Alan Hayward <alan.hayward@arm.com>
12456 David Sherwood <david.sherwood@arm.com>
12458 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12459 vectors as polynomial. Use build_index_vector for
12460 IFN_GOMP_SIMD_LANE.
12462 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12463 Alan Hayward <alan.hayward@arm.com>
12464 David Sherwood <david.sherwood@arm.com>
12466 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12467 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12468 for variable-length vectors.
12469 (vectorizable_mask_load_store): Treat the number of units as
12470 polynomial, asserting that it is constant if the condition has
12471 already been enforced.
12472 (vectorizable_store, vectorizable_load): Likewise.
12474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12475 Alan Hayward <alan.hayward@arm.com>
12476 David Sherwood <david.sherwood@arm.com>
12478 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12479 of units as polynomial. Punt if we can't tell at compile time
12480 which vector contains the final result.
12482 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12483 Alan Hayward <alan.hayward@arm.com>
12484 David Sherwood <david.sherwood@arm.com>
12486 * tree-vect-loop.c (vectorizable_induction): Treat the number
12487 of units as polynomial. Punt on SLP inductions. Use an integer
12488 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12489 cast of such a series for variable-length floating-point
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.h (build_index_vector): Declare.
12497 * tree.c (build_index_vector): New function.
12498 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12499 of units as polynomial, forcibly converting it to a constant if
12500 vectorizable_reduction has already enforced the condition.
12501 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12502 to create a {1,2,3,...} vector.
12503 (vectorizable_reduction): Treat the number of units as polynomial.
12504 Choose vectype_in based on the largest scalar element size rather
12505 than the smallest number of units. Enforce the restrictions
12508 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12509 Alan Hayward <alan.hayward@arm.com>
12510 David Sherwood <david.sherwood@arm.com>
12512 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12513 number of units as polynomial.
12515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12516 Alan Hayward <alan.hayward@arm.com>
12517 David Sherwood <david.sherwood@arm.com>
12519 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12520 * target.def (autovectorize_vector_sizes): Return the vector sizes
12521 by pointer, using vector_sizes rather than a bitmask.
12522 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12523 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12524 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12526 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12527 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12528 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12529 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12530 * omp-general.c (omp_max_vf): Likewise.
12531 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12532 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12533 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12534 * tree-vect-slp.c (vect_slp_bb): Likewise.
12535 * doc/tm.texi: Regenerate.
12536 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12538 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12539 the vector size as a poly_uint64 rather than an unsigned int.
12540 (current_vector_size): Change from an unsigned int to a poly_uint64.
12541 (get_vectype_for_scalar_type): Update accordingly.
12542 * tree.h (build_truth_vector_type): Take the size and number of
12543 units as a poly_uint64 rather than an unsigned int.
12544 (build_vector_type): Add a temporary overload that takes
12545 the number of units as a poly_uint64 rather than an unsigned int.
12546 * tree.c (make_vector_type): Likewise.
12547 (build_truth_vector_type): Take the number of units as a poly_uint64
12548 rather than an unsigned int.
12550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12551 Alan Hayward <alan.hayward@arm.com>
12552 David Sherwood <david.sherwood@arm.com>
12554 * target.def (get_mask_mode): Take the number of units and length
12555 as poly_uint64s rather than unsigned ints.
12556 * targhooks.h (default_get_mask_mode): Update accordingly.
12557 * targhooks.c (default_get_mask_mode): Likewise.
12558 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12559 * doc/tm.texi: Regenerate.
12561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12562 Alan Hayward <alan.hayward@arm.com>
12563 David Sherwood <david.sherwood@arm.com>
12565 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12566 * omp-general.c (omp_max_vf): Likewise.
12567 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12568 (expand_omp_simd): Handle polynomial safelen.
12569 * omp-low.c (omplow_simd_context): Add a default constructor.
12570 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12571 (lower_rec_simd_input_clauses): Update accordingly.
12572 (lower_rec_input_clauses): Likewise.
12574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12575 Alan Hayward <alan.hayward@arm.com>
12576 David Sherwood <david.sherwood@arm.com>
12578 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12579 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12580 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12581 (vect_analyze_slp_cost): Likewise.
12582 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12583 (vect_model_load_cost): Likewise.
12585 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12586 Alan Hayward <alan.hayward@arm.com>
12587 David Sherwood <david.sherwood@arm.com>
12589 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12590 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12591 from an unsigned int * to a poly_uint64_pod *.
12592 (calculate_unrolling_factor): New function.
12593 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
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-vectorizer.h (_slp_instance::unrolling_factor): Change
12600 from an unsigned int to a poly_uint64.
12601 (_loop_vec_info::slp_unrolling_factor): Likewise.
12602 (_loop_vec_info::vectorization_factor): Change from an int
12604 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12605 (vect_get_num_vectors): New function.
12606 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12607 (vect_get_num_copies): Use vect_get_num_vectors.
12608 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12609 to an unsigned int *.
12610 (vect_analyze_data_refs): Change min_vf from an int * to a
12612 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12613 than an unsigned HOST_WIDE_INT.
12614 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12615 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12616 to an unsigned int *.
12617 (vect_analyze_data_ref_dependences): Likewise.
12618 (vect_compute_data_ref_alignment): Handle polynomial vf.
12619 (vect_enhance_data_refs_alignment): Likewise.
12620 (vect_prune_runtime_alias_test_list): Likewise.
12621 (vect_shift_permute_load_chain): Likewise.
12622 (vect_supportable_dr_alignment): Likewise.
12623 (dependence_distance_ge_vf): Take the vectorization factor as a
12624 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12625 (vect_analyze_data_refs): Change min_vf from an int * to a
12627 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12628 vfm1 as a poly_uint64 rather than an int. Make the same change
12629 for the returned bound_scalar.
12630 (vect_gen_vector_loop_niters): Handle polynomial vf.
12631 (vect_do_peeling): Likewise. Update call to
12632 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12633 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12635 * tree-vect-loop.c (vect_determine_vectorization_factor)
12636 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12637 (vect_get_known_peeling_cost): Likewise.
12638 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12639 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12640 (vect_transform_loop): Likewise. Use the lowest possible VF when
12641 updating the upper bounds of the loop.
12642 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12643 rather than an int.
12644 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12645 polynomial unroll factors.
12646 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12647 (vect_make_slp_decision): Likewise.
12648 (vect_supported_load_permutation_p): Likewise, and polynomial
12650 (vect_analyze_slp_cost): Handle polynomial vf.
12651 (vect_slp_analyze_node_operations): Likewise.
12652 (vect_slp_analyze_bb_1): Likewise.
12653 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12654 than an unsigned HOST_WIDE_INT.
12655 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12656 (vectorizable_load): Handle polynomial vf.
12657 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12659 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12661 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12662 Alan Hayward <alan.hayward@arm.com>
12663 David Sherwood <david.sherwood@arm.com>
12665 * match.pd: Handle bit operations involving three constants
12666 and try to fold one pair.
12668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12670 * tree-vect-loop-manip.c: Include gimple-fold.h.
12671 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12672 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12673 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12674 Add a path that uses a step of VF instead of 1, but disable it
12676 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12677 and niters_no_overflow parameters. Update calls to
12678 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12679 Create a new SSA name if the latter choses to use a ste other
12680 than zero, and return it via niters_vector_mult_vf_var.
12681 * tree-vect-loop.c (vect_transform_loop): Update calls to
12682 vect_do_peeling, vect_gen_vector_loop_niters and
12683 slpeel_make_loop_iterate_ntimes.
12684 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12685 (vect_gen_vector_loop_niters): Update declarations after above changes.
12687 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12689 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12690 128-bit round to integer instructions.
12691 (ceil<mode>2): Likewise.
12692 (btrunc<mode>2): Likewise.
12693 (round<mode>2): Likewise.
12695 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12697 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12698 unaligned VSX load/store on P8/P9.
12699 (expand_block_clear): Allow the use of unaligned VSX
12700 load/store on P8/P9.
12702 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12704 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12706 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12707 swap associated with both a load and a store.
12709 2018-01-02 Andrew Waterman <andrew@sifive.com>
12711 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12712 * config/riscv/riscv.md (clear_cache): Use it.
12714 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12716 * web.c: Remove out-of-date comment.
12718 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12720 * expr.c (fixup_args_size_notes): Check that any existing
12721 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12722 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12723 (emit_single_push_insn): ...here.
12725 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12727 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12728 (const_vector_encoded_nelts): New function.
12729 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12730 (const_vector_int_elt, const_vector_elt): Declare.
12731 * emit-rtl.c (const_vector_int_elt_1): New function.
12732 (const_vector_elt): Likewise.
12733 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12734 of CONST_VECTOR_ELT.
12736 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12738 * expr.c: Include rtx-vector-builder.h.
12739 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12740 directly on the tree encoding.
12741 (const_vector_from_tree): Likewise.
12742 * optabs.c: Include rtx-vector-builder.h.
12743 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12744 sequence of "u" values.
12745 * vec-perm-indices.c: Include rtx-vector-builder.h.
12746 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12747 directly on the vec_perm_indices encoding.
12749 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12751 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12752 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12753 * rtx-vector-builder.h: New file.
12754 * rtx-vector-builder.c: Likewise.
12755 * rtl.h (rtx_def::u2): Add a const_vector field.
12756 (CONST_VECTOR_NPATTERNS): New macro.
12757 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12758 (CONST_VECTOR_DUPLICATE_P): Likewise.
12759 (CONST_VECTOR_STEPPED_P): Likewise.
12760 (CONST_VECTOR_ENCODED_ELT): Likewise.
12761 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12762 (unwrap_const_vec_duplicate): Likewise.
12763 (const_vec_series_p): Check for a non-duplicated vector encoding.
12764 Say that the function only returns true for integer vectors.
12765 * emit-rtl.c: Include rtx-vector-builder.h.
12766 (gen_const_vec_duplicate_1): Delete.
12767 (gen_const_vector): Call gen_const_vec_duplicate instead of
12768 gen_const_vec_duplicate_1.
12769 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12770 (gen_const_vec_duplicate): Use rtx_vector_builder.
12771 (gen_const_vec_series): Likewise.
12772 (gen_rtx_CONST_VECTOR): Likewise.
12773 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12774 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12775 Build a new vector rather than modifying a CONST_VECTOR in-place.
12776 (handle_special_swappables): Update call accordingly.
12777 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12778 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12779 Build a new vector rather than modifying a CONST_VECTOR in-place.
12780 (handle_special_swappables): Update call accordingly.
12782 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12784 * simplify-rtx.c (simplify_const_binary_operation): Use
12785 CONST_VECTOR_ELT instead of XVECEXP.
12787 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12789 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12790 the selector elements to be different from the data elements
12791 if the selector is a VECTOR_CST.
12792 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12793 ssizetype for the selector.
12795 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12797 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12798 before testing each element individually.
12799 * tree-vect-generic.c (lower_vec_perm): Likewise.
12801 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12803 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12804 * selftest-run-tests.c (selftest::run_tests): Call it.
12805 * vector-builder.h (vector_builder::operator ==): New function.
12806 (vector_builder::operator !=): Likewise.
12807 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12808 (vec_perm_indices::all_from_input_p): New function.
12809 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12810 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12811 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12812 instead of reading the VECTOR_CST directly. Detect whether both
12813 vector inputs are the same before constructing the vec_perm_indices,
12814 and update the number of inputs argument accordingly. Use the
12815 utility functions added above. Only construct sel2 if we need to.
12817 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12819 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12820 the broadcast of the low byte.
12821 (expand_mult_highpart): Use an explicit encoding for the permutes.
12822 * optabs-query.c (can_mult_highpart_p): Likewise.
12823 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12824 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12825 (vectorizable_bswap): Likewise.
12826 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12827 explicit encoding for the power-of-2 permutes.
12828 (vect_permute_store_chain): Likewise.
12829 (vect_grouped_load_supported): Likewise.
12830 (vect_permute_load_chain): Likewise.
12832 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12834 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12835 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12836 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12837 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12838 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12839 (vect_gen_perm_mask_any): Likewise.
12841 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12843 * int-vector-builder.h: New file.
12844 * vec-perm-indices.h: Include int-vector-builder.h.
12845 (vec_perm_indices): Redefine as an int_vector_builder.
12846 (auto_vec_perm_indices): Delete.
12847 (vec_perm_builder): Redefine as a stand-alone class.
12848 (vec_perm_indices::vec_perm_indices): New function.
12849 (vec_perm_indices::clamp): Likewise.
12850 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12851 (vec_perm_indices::new_vector): New function.
12852 (vec_perm_indices::new_expanded_vector): Update for new
12853 vec_perm_indices class.
12854 (vec_perm_indices::rotate_inputs): New function.
12855 (vec_perm_indices::all_in_range_p): Operate directly on the
12856 encoded form, without computing elided elements.
12857 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12858 encoding. Update for new vec_perm_indices class.
12859 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12860 the given vec_perm_builder.
12861 (expand_vec_perm_var): Update vec_perm_builder constructor.
12862 (expand_mult_highpart): Use vec_perm_builder instead of
12863 auto_vec_perm_indices.
12864 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12865 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12866 or double series encoding as appropriate.
12867 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12868 vec_perm_indices instead of auto_vec_perm_indices.
12869 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12870 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12871 (vect_permute_store_chain): Likewise.
12872 (vect_grouped_load_supported): Likewise.
12873 (vect_permute_load_chain): Likewise.
12874 (vect_shift_permute_load_chain): Likewise.
12875 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12876 (vect_transform_slp_perm_load): Likewise.
12877 (vect_schedule_slp_instance): Likewise.
12878 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12879 (vectorizable_mask_load_store): Likewise.
12880 (vectorizable_bswap): Likewise.
12881 (vectorizable_store): Likewise.
12882 (vectorizable_load): Likewise.
12883 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12884 vec_perm_indices instead of auto_vec_perm_indices. Use
12885 tree_to_vec_perm_builder to read the vector from a tree.
12886 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12887 vec_perm_builder instead of a vec_perm_indices.
12888 (have_whole_vector_shift): Use vec_perm_builder and
12889 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12890 truncation to calc_vec_perm_mask_for_shift.
12891 (vect_create_epilog_for_reduction): Likewise.
12892 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12893 from auto_vec_perm_indices to vec_perm_indices.
12894 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12895 instead of changing individual elements.
12896 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12897 the vector in d.perm.
12898 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12899 from auto_vec_perm_indices to vec_perm_indices.
12900 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12901 instead of changing individual elements.
12902 (arm_vectorize_vec_perm_const): Use new_vector to install
12903 the vector in d.perm.
12904 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12905 Update vec_perm_builder constructor.
12906 (rs6000_expand_interleave): Likewise.
12907 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12908 (rs6000_expand_interleave): Likewise.
12910 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12912 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12913 to qimode could truncate the indices.
12914 * optabs.c (expand_vec_perm_var): Likewise.
12916 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12918 * Makefile.in (OBJS): Add vec-perm-indices.o.
12919 * vec-perm-indices.h: New file.
12920 * vec-perm-indices.c: Likewise.
12921 * target.h (vec_perm_indices): Replace with a forward class
12923 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12924 * optabs.h: Include vec-perm-indices.h.
12925 (expand_vec_perm): Delete.
12926 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12927 (expand_vec_perm_const): Declare.
12928 * target.def (vec_perm_const_ok): Replace with...
12929 (vec_perm_const): ...this new hook.
12930 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12931 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12932 * doc/tm.texi: Regenerate.
12933 * optabs.def (vec_perm_const): Delete.
12934 * doc/md.texi (vec_perm_const): Likewise.
12935 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12936 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12937 expand_vec_perm for constant permutation vectors. Assert that
12938 the mode of variable permutation vectors is the integer equivalent
12939 of the mode that is being permuted.
12940 * optabs-query.h (selector_fits_mode_p): Declare.
12941 * optabs-query.c: Include vec-perm-indices.h.
12942 (selector_fits_mode_p): New function.
12943 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12944 is defined, instead of checking whether the vec_perm_const_optab
12945 exists. Use targetm.vectorize.vec_perm_const instead of
12946 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12947 fit in the vector mode before using a variable permute.
12948 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12949 vec_perm_indices instead of an rtx.
12950 (expand_vec_perm): Replace with...
12951 (expand_vec_perm_const): ...this new function. Take the selector
12952 as a vec_perm_indices rather than an rtx. Also take the mode of
12953 the selector. Update call to shift_amt_for_vec_perm_mask.
12954 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12955 Use vec_perm_indices::new_expanded_vector to expand the original
12956 selector into bytes. Check whether the indices fit in the vector
12957 mode before using a variable permute.
12958 (expand_vec_perm_var): Make global.
12959 (expand_mult_highpart): Use expand_vec_perm_const.
12960 * fold-const.c: Includes vec-perm-indices.h.
12961 * tree-ssa-forwprop.c: Likewise.
12962 * tree-vect-data-refs.c: Likewise.
12963 * tree-vect-generic.c: Likewise.
12964 * tree-vect-loop.c: Likewise.
12965 * tree-vect-slp.c: Likewise.
12966 * tree-vect-stmts.c: Likewise.
12967 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12969 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12970 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12971 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12972 (aarch64_vectorize_vec_perm_const): ...this new function.
12973 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12974 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12975 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12976 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12977 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12978 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12979 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12981 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12982 check for NEON modes.
12983 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12984 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12985 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12986 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12988 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12989 the old VEC_PERM_CONST conditions.
12990 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12991 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12992 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12993 (ia64_vectorize_vec_perm_const_ok): Merge into...
12994 (ia64_vectorize_vec_perm_const): ...this new function.
12995 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12996 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12997 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12998 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12999 * config/mips/mips.c (mips_expand_vec_perm_const)
13000 (mips_vectorize_vec_perm_const_ok): Merge into...
13001 (mips_vectorize_vec_perm_const): ...this new function.
13002 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13003 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13004 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13005 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13006 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13007 (rs6000_expand_vec_perm_const): Delete.
13008 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13010 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13011 (altivec_expand_vec_perm_const_le): Take each operand individually.
13012 Operate on constant selectors rather than rtxes.
13013 (altivec_expand_vec_perm_const): Likewise. Update call to
13014 altivec_expand_vec_perm_const_le.
13015 (rs6000_expand_vec_perm_const): Delete.
13016 (rs6000_vectorize_vec_perm_const_ok): Delete.
13017 (rs6000_vectorize_vec_perm_const): New function.
13018 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13019 an element count and rtx array.
13020 (rs6000_expand_extract_even): Update call accordingly.
13021 (rs6000_expand_interleave): Likewise.
13022 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13023 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13024 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13025 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13026 (rs6000_expand_vec_perm_const): Delete.
13027 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13028 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13029 (altivec_expand_vec_perm_const_le): Take each operand individually.
13030 Operate on constant selectors rather than rtxes.
13031 (altivec_expand_vec_perm_const): Likewise. Update call to
13032 altivec_expand_vec_perm_const_le.
13033 (rs6000_expand_vec_perm_const): Delete.
13034 (rs6000_vectorize_vec_perm_const_ok): Delete.
13035 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13036 reference to the SPE evmerge intructions.
13037 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13038 an element count and rtx array.
13039 (rs6000_expand_extract_even): Update call accordingly.
13040 (rs6000_expand_interleave): Likewise.
13041 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13042 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13044 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13046 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13048 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13049 vector mode and that that mode matches the mode of the data
13051 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13052 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13053 directly using expand_vec_perm_1 when forcing selectors into
13055 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13057 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13059 * optabs-query.h (can_vec_perm_p): Delete.
13060 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13061 * optabs-query.c (can_vec_perm_p): Split into...
13062 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13063 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13064 particular selector is valid.
13065 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13066 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13067 (vect_grouped_load_supported): Likewise.
13068 (vect_shift_permute_load_chain): Likewise.
13069 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13070 (vect_transform_slp_perm_load): Likewise.
13071 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13072 (vectorizable_bswap): Likewise.
13073 (vect_gen_perm_mask_checked): Likewise.
13074 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13075 implementations of variable permutation vectors into account
13076 when deciding which selector to use.
13077 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13078 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13079 with a false third argument.
13080 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13081 to test whether the constant selector is valid and can_vec_perm_var_p
13082 to test whether a variable selector is valid.
13084 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13086 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13087 * optabs-query.c (can_vec_perm_p): Likewise.
13088 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13089 instead of vec_perm_indices.
13090 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13091 (vect_gen_perm_mask_checked): Likewise,
13092 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13093 (vect_gen_perm_mask_checked): Likewise,
13095 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13097 * optabs-query.h (qimode_for_vec_perm): Declare.
13098 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13099 (qimode_for_vec_perm): ...this new function.
13100 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13102 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13104 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13105 does not have a conditional at the top.
13107 2018-01-02 Richard Biener <rguenther@suse.de>
13109 * ipa-inline.c (big_speedup_p): Fix expression.
13111 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13114 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13117 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13121 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13122 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13123 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13124 cond_taken_branch_cost 3->4.
13126 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13128 PR tree-optimization/83581
13129 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13130 TODO_cleanup_cfg if any changes have been made.
13132 PR middle-end/83608
13133 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13134 convert_modes if target mode has the right side, but different mode
13137 PR middle-end/83609
13138 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13139 last argument when extracting from CONCAT. If either from_real or
13140 from_imag is NULL, use expansion through memory. If result is not
13141 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13142 the parts directly to inner mode, if even that fails, use expansion
13145 PR middle-end/83623
13146 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13147 check for bswap in mode rather than HImode and use that in expand_unop
13150 Copyright (C) 2018 Free Software Foundation, Inc.
13152 Copying and distribution of this file, with or without modification,
13153 are permitted in any medium without royalty provided the copyright
13154 notice and this notice are preserved.