1 2018-09-14 Bernd Edlinger msebor@redhat.com>
4 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
6 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
7 terminated string literal.
9 2018-09-14 Martin Sebor <msebor@redhat.com>
11 * builtins.c (unterminated_array): Handle ARRAY_REF.
12 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
13 * builtins.h (unterminated_array): Declare extern.
14 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
16 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
19 2018-09-14 Martin Sebor <msebor@redhat.com>
20 Jeff Law <law@redhat.com>
22 * builtins.c (unterminated_array): New.
23 (expand_builtin_strcpy): Adjust.
24 (expand_builtin_strcpy_args): Detect unterminated arrays.
25 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
27 * gimple-fold.h (get_maxval_strlen): Add argument.
28 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
30 * gimple-fold.c (get_range_strlen): Add argument.
31 (get_maxval_strlen): Adjust.
32 * gimple-fold.h (get_range_strlen): Add argument.
34 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
36 * config/i386/movdirintrin.h: Fix copyright year.
38 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
40 * reg-stack.c: Include regs.h.
41 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
42 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
43 FIRST_STACK_REG, not DFmode.
44 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
45 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
46 (change stack): Default register mode to the reg_raw_mode of
47 FIRST_STACK_REG, not DFmode.
48 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
49 (*swapxf): Rename from swapxf.
51 2018-09-14 Carl Love <cel@us.ibm.com>
53 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
54 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
56 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
59 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
62 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
65 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
67 2018-09-14 Jason Merrill <jason@redhat.com>
69 Fix --enable-gather-detailed-mem-stats.
70 * hash-table.c (hash_table_usage): Change from variable to function.
71 * hash-table.h: Adjust.
72 * Makefile.in: Add missing dependencies on hash-table.h.
74 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
76 PR tree-optimization/87259
78 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
79 execute_cse_reciprocals_1 has tried transforming.
81 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
83 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
84 VR_VARYING for PLUS/MINUS_EXPR.
86 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
88 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
91 2018-09-14 Richard Biener <rguenther@suse.de>
94 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
95 bits for the merged partition.
97 2018-09-13 Martin Sebor <msebor@redhat.com>
98 Bernd Edlinger <bernd.edlinger@hotmail.de>
100 * builtins.h (c_srlen): Add argument.
101 * builtins.c (warn_string_no_nul): New function.
102 (c_strlen): Add argument and use it. Update recursive calls.
103 Pass DECL argument to string_constant to get info on non
104 terminated strings. Update *NONSTR as needed.
105 (fold_builtin_strlen): Add argument to calls to c_strlen.
106 Warn for unterminated arrays.
107 (warn_string_no_null): Add prototype.
108 * expr.c (string_constant): Update arguments. Update recursive
109 calls appropriately. Detect missing NUL terminator and outermost
110 declaration its missing in.
111 Improve checks for arrays with nonzero lower bound or elements
112 that are not a single byte. Simplify offset computation.
113 Simplify checks for non-NUL terminated strings.
114 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
115 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
117 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
119 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
121 * fold-const.c (c_getstr): Fix function comment. Remove unused third
122 argument. Fix range checks.
123 * fold-const.h (c_getstr): Adjust protoype.
124 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
125 string is constant but contains no NUL byte.
127 * expr.c (string_constant): Adjust function comment.
128 Remove bogus check for zero termination.
130 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
132 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
133 (get_constant_size): Don't make STRING_CSTs larger than they are.
134 (check_string_literal): New check function for STRING_CSTs.
135 (output_constant): Use it.
137 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
140 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
142 2018-09-13 Richard Biener <rguenther@suse.de>
144 PR tree-optimization/87263
145 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
146 (struct unwind_state): Add max_rpo field.
147 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
148 Compute max_rpo, the max RPO number a block can be backwards reached
149 from. Re-write non-iterating mode to a RPO ordered worklist approach,
150 separating it from the iterating mode.
152 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
154 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
155 (rfs_decision): New scheduling decision.
157 2018-09-13 Richard Biener <rguenther@suse.de>
160 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
161 (vn_nary_op_insert_pieces_predicated): Do not write useless
162 valid_dominated_by_p entry outside of the allocated storage.
164 2018-09-13 Omar Sandoval <osandov@osandov.com>
165 Tom de Vries <tdevries@suse.de>
168 * dwarf2out.c (is_c): New function.
169 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
171 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
174 * config/aarch64/aarch64.md (*aarch64_bfxil):
176 * config/aarch64/constraints.md (Ulc): Define.
177 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
179 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
182 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
184 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
185 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
186 aarch64_layout_frame call.
187 (aarch64_expand_epilogue): Likewise.
188 (aarch64_initial_elimination_offset): Likewise.
189 (aarch64_get_separate_components): Likewise.
190 (aarch64_use_return_insn_p): Likewise.
191 (aarch64_layout_frame): Remove unneeded check.
193 2018-09-13 Jakub Jelinek <jakub@redhat.com>
195 * configure.ac: Only append
196 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
197 gcc_config_arguments if it was never reconfigured or last reconfigure
198 was with different arguments.
199 * configure: Regenerated.
201 2018-09-13 Jakub Jelinek <jakub@redhat.com>
202 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
205 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
206 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
208 2018-09-13 Jakub Jelinek <jakub@redhat.com>
210 PR tree-optimization/87287
211 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
212 X % C == 0 to X % (unsigned) C == 0 optimization to ...
213 * match.pd (X % C == 0): ... here. New optimization.
215 2018-09-12 Jakub Jelinek <jakub@redhat.com>
218 * expr.h (maybe_optimize_mod_cmp): Declare.
219 * expr.c (mod_inv): New function.
220 (maybe_optimize_mod_cmp): New function.
221 (do_store_flag): Use it.
222 * cfgexpand.c (expand_gimple_cond): Likewise.
224 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
225 Julian Brown <julian@codesourcery.com>
228 * gimplify.c (gimplify_scan_omp_clauses): Set
229 target_firstprivatize_array_bases in OpenACC parallel and kernels
230 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
231 OpenACC data regions.
233 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
235 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
236 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
238 2018-09-12 Richard Biener <rguenther@suse.de>
240 PR tree-optimization/87280
241 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
242 edge but unreachable target.
243 (do_rpo_vn): For conservatively handling a PHI only mark
244 the backedge executable but not the block reachable.
246 2018-09-12 Richard Biener <rguenther@suse.de>
248 PR tree-optimization/87266
249 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
252 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
254 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
256 ("trunc<BFP:mode><DFP_ALL:mode>2")
257 ("trunc<DFP_ALL:mode><BFP:mode>2")
258 ("extend<BFP:mode><DFP_ALL:mode>2")
259 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
260 according to the target operand type.
262 2018-09-12 Jakub Jelinek <jakub@redhat.com>
263 Andreas Krebbel <krebbel@linux.ibm.com>
265 PR tree-optimization/86844
266 * gimple-ssa-store-merging.c
267 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
268 there are any overlapping stores in between them, make sure they are
269 also coalesced or we give up completely.
271 2018-09-12 Jakub Jelinek <jakub@redhat.com>
274 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
275 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
277 2018-09-12 Tom de Vries <tdevries@suse.de>
279 * common.opt (gdescribe-dies): Add option.
280 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
281 attribute for artifical and nameless decls.
282 (dwarf2out_register_external_die): Add description attribute to
283 external reference die.
284 (add_desc_attribute): New functions.
285 (gen_subprogram_die): Add description attribute to
286 DW_TAG_call_site_parameter.
287 * tree-pretty-print.c (print_generic_expr_to_str): New function.
288 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
289 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
291 (@item -gdescribe-dies): Add.
293 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
295 * tree-vrp.c (vrp_shift_undefined_p): Remove.
296 (extract_range_from_binary_expr_1: Call
297 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
298 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
301 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
303 * gimple-ssa-warn-alloca.c
304 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
305 field for ALLOCA_BOUND_*_LARGE.
307 2018-09-11 Nathan Sidwell <nathan@acm.org>
309 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
311 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
313 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
314 for clobbers. Remove obsolete comment.
316 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
318 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
319 mpxchk, mpxld and mpxst types.
320 (define_attr length_immediate): Remove all processing of mpx types.
321 (define_attr prefix_0f): Ditto.
322 (define_attr memory): Ditto.
324 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
326 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
327 (log<mode>2): Change operand 1 predicate to general_operand.
328 Extend operand 1 to XFmode and generate logxf3 insn.
329 (log10<mode>2): Change operand 1 predicate to general_operand.
330 Extend operand 1 to XFmode and generate log10xf3 insn.
331 (log2<mode>2): Change operand 1 predicate to general_operand.
332 Extend operand 1 to XFmode and generate log2xf3 insn.
333 (fyl2xp1_extend<mode>xf3_i387): Remove.
334 (log1p<mode>2): Change operand 1 predicate to general_operand.
335 Extend operand 1 to XFmode and generate log1pxf3 insn.
336 (fxtract_extend<mode>xf3_i387): Remove.
337 (logb<mode>2): Change operand 1 predicate to general_operand.
338 Extend operand 1 to XFmode and generate logbxf3 insn.
339 (ilogb<mode>2): Change operand 1 predicate to general_operand.
340 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
341 (significand<mode>2): Change operand 1 predicate to general_operand.
342 Extend operand 1 to XFmode and generate significandxf3 insn.
344 2018-09-11 Nathan Sidwell <nathan@acm.org>
346 * gcc.c (perror_with_name, pfatal_with_name): Delete.
347 (load_specs): Use fatal_error.
348 (DELETE_IF_ORDINARY, process_command): Use error.
349 (execute, run_attempt): Use fatal_error.
351 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
353 * diagnostic-core.h (sorry_at): New prototype.
354 * diagnostic.c (sorry_at): New function.
356 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
358 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
359 by zero as VR_UNDEFINED.
361 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
363 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
364 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
365 (<sincos>mode2): New expander.
366 (sincos_extend<mode>xf3_i387): Remove insn pattern.
367 (sincos -> sin, cos splitters): Remove splitter patterns.
368 (sincos<mode>3): Change operand 2 predicate to general_operand.
369 Extend operand 2 to XFmode and generate sincosxf3 insn.
370 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
371 Change operand 3 predicate to const1_operand.
372 (fptan_extend<mode>xf4_i387): Remove insn pattern.
373 (tanxf2): Update operands in the call to fptanxf4_i387.
374 (tan<mode>2): Change operand 1 predicate to general_operand.
375 Extend operand 1 to XFmode and generate tanxf3 insn.
376 (atan2xf3): Rename from *fpatanxf3_i387.
377 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
378 (atan2xf3): Remove expander.
379 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
380 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
381 (atan<mode>2): Change operand 1 predicate to general_operand.
382 Extend operand 1 to XFmode and generate atanxf3 insn.
384 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
386 * config/i386/i386.md (x87/SSE constant load splitter): Use
387 memory_operand instead of nonimmediate_operand for input operand
390 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
392 * config/i386/i386.md (float partial SSE register stall splitter): Move
393 splitter near its instruction pattern.
394 (float_extend partial SSE register stall splitter): Ditto.
395 (float_truncate partial SSE register stall splitter): Ditto.
397 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
400 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
401 to speculation_safe_value_not_needed.
404 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
405 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
406 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
407 leaf_function_p, instead use has_hard_reg_initial_val.
409 2018-09-09 Nathan Sidwell <nathan@acm.org>
411 * gcc.h (pfatal_with_name): Don't declare here.
412 * gcc.c (pfatal_with_name): Make static.
414 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
416 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
419 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
421 PR rtl-optimization/85458
422 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
423 priority hook to reduce the priority of EXPR.
425 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
427 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
428 DImode for x87 on 32bit targets. Conditionally disable x87 modes
429 with X87_ENABLE_FLOAT. Remove preparation code.
430 (*float<SWI48:mode><MODEF:mode>2): Rename from
431 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
432 math using "enabled" attribute.
433 (*floatdi<MODEF:mode>2_i387): Rename from
434 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
435 enable for 32bit targets only.
436 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
438 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
439 as operand 1 predicate. Rewrite as define_insn_and_split.
440 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
442 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
444 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
445 to fallthru to FLOAT case.
447 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
450 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
451 around folding of vec_sl to handle out of range shift values.
453 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
455 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
456 Update callers to gen_fix_trunc<mode>_i387_fisttp
457 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
458 nonimmediate_operand.
459 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
460 and corresponding splitters.
461 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
462 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
463 (fix_truncdi_i387_with_temp): Remove insn pattern
464 and corresponding splitters.
465 (fix_trunc<mode>_i387): Change operand 0 predicate to
466 nonimmediate_operand.
467 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
468 and corresponding splitters.
469 (*fistdi2_1): Remove.
471 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
472 (lrintxfdi2): Remove expander. Reimplement as define_insn.
473 (*fist<mode>2_1): Remove.
474 (fist<mode>2): Ditto.
475 (fist<mode>2_with_temp): Remove insn pattern and corresponding
477 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
478 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
479 (fistdi2_<rounding>): Change operand 0 predicate to
480 nonimmediate_operand.
481 (fistdi2_<rounding>_with_temp): Remove insn pattern
482 and corresponding splitters.
483 (fist<mode>2_<rounding>): Change operand 0 predicate to
484 nonimmediate_operand.
485 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
486 and corresponding splitters.
488 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
490 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
492 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
495 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
497 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
498 early gimple folding of vec_splat().
499 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
500 * gimple-fold.h: Add an extern define for tree_vec_extract().
502 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
504 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
505 wrappers around TREE_TYPE comparisons.
507 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
510 * config/s390/predicates.md: Add nonsym_memory_operand.
511 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
512 contains a SYMBOL_REF, load it into an intermediate pseudo.
513 (s390_emit_compare_and_swap): Legitimize operand.
514 * config/s390/s390.md: Use the new nonsym_memory_operand
515 with UNSPECV_CAS patterns.
517 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
520 * config/s390/s390-passes.def: New file.
521 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
523 (make_pass_s390_early_mach): Add declaration.
524 * config/s390/s390.c (make_pass_s390_early_mach):
525 (s390_option_override): Remove dynamic registration.
526 * config/s390/t-s390: Add s390-passes.def.
528 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
530 * config/s390/s390.c (s390_decompose_constant_pool_ref):
531 Remove UNSPEC_LTREL_BASE check.
532 (annotate_constant_pool_refs): Likewise.
533 (find_constant_pool_ref): Likewise.
534 (find_ltrel_base): Removed.
535 (replace_ltrel_base): Removed.
536 (s390_mainpool_finish): Remove replace_ltrel_base call.
537 (s390_chunkify_start): Remove pending LTREL_BASE logic.
538 (s390_chunkify_finish): Remove replace_ltrel_base call.
539 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
541 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
544 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
545 to speculation_safe_value_not_needed.
547 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
548 Bernd Schmidt <bernds_cb1@t-online.de>
550 * config/nvptx/nvptx-opts.h: New file.
551 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
552 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
554 (TARGET_SM35): New macro.
555 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
557 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
560 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
562 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
564 * config/i386/i386.md (truncdfsf2): Remove expander.
565 (truncdfsf2_with_temp): Ditto.
566 (truncxf<mode>2): Ditto.
567 (*truncdfsf_fast_mixed): Remove insn pattern.
568 (*truncdfsf_fast_i387): Ditto.
569 (*truncdfsf_mixed): Ditto.
570 (*truncdfsf_i387): Ditto.
571 (*truncdfsf2_i387_1): Ditto.
572 (*truncxfsf2_mixed): Ditto.
573 (*truncxfdf2_mixed): Ditto.
574 (*truncxf<mode>2_i387_noop): Ditto. Update callers
575 to call gen_truncxf<mode>2 instead.
576 (*truncxf<mode>2_i387): Remove.
577 (reg->reg splitters): Remove splitter pattern.
578 (reg->mem splitters): Ditto.
580 (truncdfsf2): New insn pattern.
581 (truncxf<mode>2): Ditto.
583 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
585 * tree-ssa-math-opts.c (is_mult_by): New function.
586 (is_square_of): Use the above.
587 (optimize_recip_sqrt): New function.
588 (pass_cse_reciprocals::execute): Use the above.
590 2018-09-05 Richard Biener <rguenther@suse.de>
593 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
594 to zero-init the emplaced vec.
596 2018-09-05 Martin Liska <mliska@suse.cz>
598 PR tree-optimization/87205
599 * tree-switch-conversion.c (pass_lower_switch::execute):
600 Group cases for switch statements.
602 2018-09-05 Richard Biener <rguenther@suse.de>
604 PR tree-optimization/87217
605 * tree-ssa-sccvn.c (vuse_valueize): New.
606 (vn_reference_lookup_pieces): Use it.
607 (vn_reference_lookup): Likewise.
609 2018-09-05 Nathan Sidwell <nathan@acm.org>
612 * stor-layout.c (place_field): Scan forwards to check last
613 bitfield when ms_bitfield_placement is in effect.
615 2018-09-05 Richard Biener <rguenther@suse.de>
618 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
621 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
622 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
624 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
625 * config.gcc (extra_objs): Build it.
626 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
628 * config/aarch64/aarch64-passes.def
629 (pass_tag_collision_avoidance): New pass.
630 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
631 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
632 (aarch64_classify_address): Remove static qualifier.
633 (aarch64_address_info, aarch64_address_type): Move to...
634 * config/aarch64/aarch64-protos.h: ... here.
635 (make_pass_tag_collision_avoidance): New function.
636 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
639 2018-09-05 Martin Liska <mliska@suse.cz>
641 * doc/gcov.texi: Update documentation of humar
643 * gcov.c (format_count): Print one decimal place, it provides
644 more fine number of situations like '1G' vs. '1.4G'.
646 2018-09-05 Martin Liska <mliska@suse.cz>
649 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
650 * optc-gen.awk: Allow 'Var' for Deprecated options in order
651 to generate a MASK value.
653 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
656 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
657 if frame pointer isn't used.
658 (compute_frame_pointer_to_fb_displacement): Likewise.
660 2018-09-04 Jakub Jelinek <jakub@redhat.com>
663 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
664 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
665 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
666 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
667 and OPTION_MASK_ISA_XSAVEC_UNSET.
669 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
671 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
672 XOR operations in NAND case.
674 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
676 * wide-int-range.cc (wide_int_range_convert): New.
677 * wide-int-range.h (wide_int_range_convert): New.
678 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
679 code into wide_int_range_convert.
680 (extract_range_into_wide_ints): Do not munge anti range constants
681 into the entire domain. Just return the range back.
683 2018-09-04 Martin Liska <mliska@suse.cz>
685 * genmatch.c (output_line_directive): Add new argument
687 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
689 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
691 * doc/invoke.texi (Option Summary): Add whitespace.
693 * doc/invoke.texi (Option Summary): Add -Waligned-new.
695 2018-09-04 Richard Biener <rguenther@suse.de>
697 PR tree-optimization/87211
698 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
699 backedge value we're supposed to treat as VARYING also number
700 the PHI to VARYING in case it got a different value-number already.
702 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
704 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
705 (extract_range_from_binary_expr_1): Do not call
706 vrp_can_optimize_bit_op.
707 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
709 (wide_int_range_get_mask_and_bounds): New.
710 (wide_int_range_optimize_bit_op): New.
711 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
712 (wide_int_range_bit_and): Same.
713 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
714 (wide_int_range_optimize_bit_op): New.
715 (wide_int_range_get_mask_and_bounds): New.
717 2018-09-04 Richard Biener <rguenther@suse.de>
719 PR tree-optimization/87176
720 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
721 variable. When value-numbering a virtual PHI node make sure
722 to not value-number to the backedge value.
724 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
726 * doc/extend.texi (Long Long, Hex Floats): Document support for
727 long long and hex floats in more recent versions of ISO C++.
729 2018-09-03 Richard Biener <rguenther@suse.de>
731 PR tree-optimization/87177
732 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
735 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
737 * bb-reorder.c (edge_order): Convert to C-qsort-style
738 tri-state comparator.
739 (reorder_basic_blocks_simple): Change std::stable_sort to gcc_stablesort.
741 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
743 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
744 tri-state comparator.
745 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
747 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
749 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
750 (mergesort): ... here as maximum count for using netsort.
751 (gcc_qsort): Set nlim to 3 if stable sort is requested.
752 (gcc_stablesort): New.
753 * system.h (gcc_stablesort): Declare.
755 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
757 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
758 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
759 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
761 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
763 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
764 lxsdx and stxsdx alternatives.
765 (*mov<mode>_hardfloat64): Ditto.
766 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
768 2018-09-03 Richard Biener <rguenther@suse.de>
770 PR tree-optimization/87200
771 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
774 2018-09-03 Martin Liska <mliska@suse.cz>
776 PR tree-optimization/87201
777 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
778 Fix parenthesis in an expression.
780 2018-09-03 Richard Biener <rguenther@suse.de>
782 PR tree-optimization/87197
783 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
784 visited. CSE the VN_INFO hashtable lookup.
786 PR tree-optimization/87169
787 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
788 iterating make sure there's no extra backedges from irreducible
789 regions feeding the header. Mark the destination block
792 2018-09-03 Martin Liska <mliska@suse.cz>
795 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
796 * common/common-targhooks.c (default_get_valid_option_values):
798 * common/common-targhooks.h (default_get_valid_option_values):
800 * common/config/i386/i386-common.c: Move processor_target_table
802 (ix86_get_valid_option_values): New function.
803 (TARGET_GET_VALID_OPTION_VALUES): New macro.
804 * config/i386/i386.c (struct ptt): Move to i386-common.c.
805 (PTA_*): Move all defined masks into i386-common.c.
806 (ix86_function_specific_restore): Use new processor_cost_table.
807 * config/i386/i386.h (struct ptt): Moved from i386.c.
808 (struct pta): Likewise.
809 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
810 * doc/tm.texi.in: Likewise.
811 * opt-suggestions.c (option_proposer::suggest_option):
812 Pass prefix to build_option_suggestions.
813 (option_proposer::get_completions): Likewise.
814 (option_proposer::build_option_suggestions): Use the new target
816 * opts.c (struct option_help_tuple): New struct.
817 (print_filtered_help): Use the new target hook.
819 2018-09-03 Martin Liska <mliska@suse.cz>
822 * predict.c (set_even_probabilities): Add likely_edges
823 argument and handle cases where we have precisely one
825 (combine_predictions_for_bb): Catch also likely_edges.
826 (tree_predict_by_opcode): Handle gswitch statements.
827 * tree-cfg.h (find_case_label_for_value): New declaration.
828 (find_taken_edge_switch_expr): Likewise.
829 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
830 Find pivot in decision tree based on probabily, not by number of
833 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
835 * doc/standards.texi (Standards): Update Objective-C reference.
837 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
839 * doc/install.texi (Prerequisites): Update link for MPC.
841 2018-09-01 Michael Matz <matz@suse.de>
843 PR tree-optimization/87074
844 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
845 PHIs for outer-loop uses.
847 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
849 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
850 * doc/invoke.texi (C Dialect Options): Ditto.
852 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
854 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
856 2018-08-31 Richard Biener <rguenther@suse.de>
858 PR tree-optimization/87168
859 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
860 (rpo_elim::eliminate_avail): When OP was not visited it must
863 2018-08-31 David Malcolm <dmalcolm@redhat.com>
865 * tree-vrp.c (copy_value_range): Convert param "from" from
866 "value_range *" to "const value_range *".
867 (range_is_null): Likewise for param "vr".
868 (range_int_cst_p): Likewise.
869 (range_int_cst_singleton_p): Likewise.
870 (symbolic_range_p): Likewise.
871 (value_ranges_intersect_p): Likewise for both params.
872 (value_range_nonnegative_p): Likewise for param "vr".
873 (value_range_constant_singleton): Likewise.
874 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
875 (extract_range_into_wide_ints): Likewise for param "vr".
876 (extract_range_from_multiplicative_op): Likewise for params "vr0"
878 (vrp_can_optimize_bit_op): Likewise.
879 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
881 (extract_range_from_unary_expr): Likewise.
882 (debug_value_range): Likewise for param "vr".
883 (value_range::dump): Add "const" qualifier.
884 (vrp_prop::check_array_ref): Convert local "vr" from
885 "value_range *" to "const value_range *".
886 (vrp_prop::check_mem_ref): Likewise.
887 (vrp_prop::visit_stmt): Likewise for local "old_vr".
888 (vrp_intersect_ranges_1): Likewise for param "vr_1".
889 (vrp_intersect_ranges): Likewise.
890 (simplify_stmt_for_jump_threading): Likewise for local "vr".
891 (vrp_prop::vrp_finalize): Likewise.
892 * tree-vrp.h (value_range::dump): Add "const" qualifier.
893 (vrp_intersect_ranges): Add "const" qualifier to params as above.
894 (extract_range_from_unary_expr): Likewise.
895 (value_range_constant_singleton): Likewise.
896 (symbolic_range_p): Likewise.
897 (copy_value_range): Likewise.
898 (extract_range_from_binary_expr_1): Likewise.
899 (range_int_cst_p): Likewise.
900 (vrp_set_zero_nonzero_bits): Likewise.
901 (range_int_cst_singleton_p): Likewise.
903 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
905 * config/aarch64/arm_neon.h (vabsd_s64): New.
906 (vnegd_s64): Likewise.
908 2018-08-31 Martin Jambor <mjambor@suse.cz>
910 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
912 2018-08-31 Martin Liska <mliska@suse.cz>
914 * ipa-icf.c (sem_item::add_type): Use
915 sem_item::m_type_hash_cache.
916 * ipa-icf.h: Move the cache from sem_item_optimizer
919 2018-08-31 Nathan Sidwell <nathan@acm.org>
921 * doc/extend.texi (Backwards Compatibility): Remove implicit
922 extern C leeway of () being (...).
924 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
926 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
928 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
932 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
934 2018-08-31 Jakub Jelinek <jakub@redhat.com>
937 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
938 gen_int_mode. Formatting fixes.
940 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
942 * target.def (custom_function_descriptors): Improve documentation.
943 * doc/tm.texi.in (Trampolines): Expand discussion of function
944 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
945 beginning of the section.
946 * doc/tm.texi: Regenerated.
948 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
950 * cfg.h (class auto_edge_flag): Spell out the template-id of the
951 base class in the initializer list. This is a workaround for
952 building with older GCC.
953 (class auto_bb_flag): Likewise.
955 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
957 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
958 (altivec_vcmpequ<VI_char>_p): Remove star.
959 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
961 (expand_strncmp_vec_sequence): New function.
962 (emit_final_str_compare_vec): New function.
963 (expand_strn_compare): Add support for vector strncmp.
964 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
965 length specification to bytes.
966 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
967 (vcmpnezb_p): New pattern.
968 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
969 for option -mstring-compare-inline-limit.
971 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
973 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
974 (PTA_SKYLAKE): Add PTA_AES.
975 (PTA_GOLDMONT): Likewise.
977 2018-08-29 Jan Hubicka <jh@suse.cz>
980 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
981 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
983 2018-08-29 Jan Hubicka <jh@suse.cz>
985 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
987 (hash_tree): Do not visit TYPE_STUB_DECL.
988 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
989 stream TYPE_STUB_DECL.
990 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
991 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
992 after free_lang_data.
993 (type_in_anonymous_namespace_p): Likewise.
995 2018-08-29 Jan Hubicka <jh@suse.cz>
997 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
998 comment that it has to be even number.
999 (class sreal): Change m_sig type to int32_t.
1000 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1001 int64_t for temporary calculations.
1002 (sreal_verify_basics): Drop one bit from minimum and maximum.
1004 2018-08-30 Richard Biener <rguenther@suse.de>
1006 PR tree-optimization/87147
1007 * tree-ssa-sccvn.c (SSA_VISITED): New function.
1008 (visit_phi): When the degenerate result is from the backedge and
1009 we didn't visit its definition yet drop to VARYING.
1010 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
1012 2018-08-29 Jan Hubicka <jh@suse.cz>
1014 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
1016 (hash_tree): Likewise.
1018 2018-08-29 Jan Hubicka <jh@suse.cz>
1020 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
1021 and TYPE_NEXT_REF_TO.
1023 2018-08-29 Jan Hubicka <jh@suse.cz>
1025 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
1026 comment that it has to be even number.
1027 (class sreal): Change m_sig type to int32_t.
1028 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1029 int64_t for temporary calculations.
1030 (sreal_verify_basics): Drop one bit from minimum and maximum.
1032 2018-08-30 Tamar Christina <tamar.christina@arm.com>
1034 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
1036 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
1039 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
1040 if to_add is negative.
1042 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
1045 * builtins.c (c_strlen): Improve range checks.
1047 2018-08-29 Martin Sebor <msebor@redhat.com>
1048 Jeff Law <law@redhat.com>
1050 PR tree-optimization/86714
1051 PR tree-optimization/86711
1052 * builtins.c (c_strlen): Add arguments to call to string_constant.
1053 * expr.c (string_constant): Add argument. Detect missing nul
1054 terminator and outermost declaration it's missing in.
1055 * expr.h (string_constant): Add argument.
1056 * fold-const.c (read_from_constant_string): Add arguments to call to
1058 (c_getstr): Likewise.
1059 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1061 * tree-ssa-strlen.c (get_stridx): Likewise.
1063 2018-08-29 Jan Hubicka <jh@suse.cz>
1065 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
1066 Do not stream DECL_VINDEX.
1067 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
1068 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
1069 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
1071 2018-08-29 Richard Biener <rguenther@suse.de>
1073 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
1074 virtual operands that are not default defs to honor region
1076 (rpo_vn_valueize): Remove ineffective code here.
1078 2018-08-29 Richard Biener <rguenther@suse.de>
1080 PR tree-optimization/87132
1081 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
1082 when skipping defs reachable over backedges.
1084 2018-08-29 Richard Biener <rguenther@suse.de>
1086 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
1087 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
1088 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
1089 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
1090 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
1091 (vn_reference_lookup_3): Remove use of const_parms.
1092 (free_rpo_vn): Do not free const_parms.
1093 (do_rpo_vn): Do not call init_const_parms.
1094 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
1095 SSA_NAME_POINTS_TO_READONLY_MEMORY.
1096 (call_may_clobber_ref_p_1): Likewise.
1098 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
1101 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
1103 (-ftree-scev-cprop): Document.
1105 2018-08-29 Jan Hubicka <jh@suse.cz>
1107 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
1109 (sreal constructor): Update.
1110 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
1111 sreal:operator/): Update.
1113 2018-08-29 Martin Liska <mliska@suse.cz>
1115 * tree-switch-conversion.c (switch_conversion::expand):
1116 Strenghten assumption about gswitch statements.
1118 2018-08-29 Richard Biener <rguenther@suse.de>
1120 PR tree-optimization/87117
1121 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
1122 re-value-number released SSA VDEFs.
1124 2018-08-29 Richard Biener <rguenther@suse.de>
1126 PR tree-optimization/87126
1127 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
1129 2018-08-28 Jim Wilson <jimw@sifive.com>
1131 * config/riscv/pic.md: Rewrite.
1132 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
1134 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
1135 (SOFTF, default_load, softload, softstore): New.
1137 2018-08-28 Jeff Law <law@redhat.com>
1139 * fold-const.c (fold_binary_loc): Remove recently added assert.
1141 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
1143 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
1144 to OP parmeter of generated function.
1146 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
1148 PR tree-optimization/87009
1149 * match.pd: Add boolean optimizations.
1151 2018-08-28 Martin Sebor <msebor@redhat.com>
1154 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
1155 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
1156 (pass_walloca::gate): Use it.
1157 (alloca_call_type): Same.
1158 (pass_walloca::execute): Same.
1159 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
1161 2018-08-28 David Malcolm <dmalcolm@redhat.com>
1163 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
1164 GCC_VERSION for usage of "__gcc_dump_printf__" format from
1167 2018-08-28 Richard Biener <rguenther@suse.de>
1169 PR tree-optimization/87124
1170 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
1171 constants before looking up avail.
1173 2018-08-28 Jakub Jelinek <jakub@redhat.com>
1176 * calls.c (maybe_warn_nonstring_arg): Punt early if
1177 warn_stringop_overflow is zero. Don't call get_range_strlen
1178 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
1179 Swap comparison operands to have constants on rhs. Only use
1180 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
1181 increment lenrng[0].
1183 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
1185 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
1186 use of tree_to_shwi. Remove duplicated test for the size being
1187 a whole number of bytes.
1189 2018-08-28 Richard Biener <rguenther@suse.de>
1191 PR tree-optimization/87117
1192 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
1193 Handle removed stmt without LHS (GIMPLE_NOP).
1195 2018-08-28 Richard Biener <rguenther@suse.de>
1197 PR tree-optimization/87117
1198 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
1199 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
1201 2018-08-28 Richard Biener <rguenther@suse.de>
1203 PR tree-optimization/87117
1204 * tree-ssa-pre.c (compute_avail): Do not make expressions
1205 with predicated values available.
1206 (get_expr_value_id): Assert we do not run into predicated value
1209 2018-08-28 Richard Biener <rguenther@suse.de>
1211 PR tree-optimization/87117
1212 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
1213 get virtual operands.
1214 (get_expr_operands): Handle STRING_CST like other decls.
1216 2018-08-28 Martin Liska <mliska@suse.cz>
1218 * tree.h: Update documentation of fndecl_built_in_p
1222 2018-08-27 Jeff Law <law@redhat.com>
1223 PR tree-optimization/87110
1224 * tree-ssa-dse.c (compute_trims): Handle non-constant
1227 2018-08-27 Martin Sebor <msebor@redhat.com>
1229 PR tree-optimization/86914
1230 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
1232 2018-08-27 Martin Sebor <msebor@redhat.com>
1234 PR tree-optimization/87112
1235 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
1236 the type of the bound argument.
1238 2018-08-27 Jeff Law <law@redhat.com>
1240 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
1241 type does not have a TYPE_SIZE_UNIT.
1243 2018-08-27 Steve Ellcey <sellcey@cavium.com>
1245 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
1246 with include of backend.h.
1248 2018-08-27 Richard Biener <rguenther@suse.de>
1250 PR tree-optimization/86927
1251 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
1252 use const cond reduction code.
1254 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
1256 PR tree-optimization/85758
1257 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
1259 2018-08-27 David Malcolm <dmalcolm@redhat.com>
1262 * diagnostic-show-locus.c (class layout_range): Update for
1263 conversion of show_caret_p to a tri-state.
1264 (layout_range::layout_range): Likewise.
1265 (make_range): Likewise.
1266 (layout::maybe_add_location_range): Likewise.
1267 (layout::should_print_annotation_line_p): Don't show annotation
1268 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
1269 (layout::get_state_at_point): Update for conversion of
1270 show_caret_p to a tri-state. Bail out early for
1271 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
1272 underlining or source colorization.
1273 (gcc_rich_location::add_location_if_nearby): Update for conversion
1274 of show_caret_p to a tri-state.
1275 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
1276 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
1278 (selftest::test_one_liner_labels): Likewise.
1279 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
1280 conversion of show_caret_p to a tri-state.
1281 * pretty-print.c (text_info::set_location): Likewise.
1282 * pretty-print.h (text_info::set_location): Likewise.
1283 * substring-locations.c (format_warning_n_va): Likewise.
1284 * tree-diagnostic.c (default_tree_printer): Likewise.
1285 * tree-pretty-print.c (newline_and_indent): Likewise.
1287 2018-08-27 David Malcolm <dmalcolm@redhat.com>
1290 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
1291 line above for line-insertion fix-it hints.
1292 (selftest::test_fixit_insert_containing_newline): Update the
1293 expected results, and add a test with line-numbering enabled.
1295 2018-08-27 Martin Liska <mliska@suse.cz>
1298 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
1299 params with DECL_HAS_VALUE_EXPR_P.
1301 2018-08-27 Martin Liska <mliska@suse.cz>
1303 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
1304 selected expansion strategy.
1306 2018-08-27 Martin Liska <mliska@suse.cz>
1308 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
1309 * builtins.c (is_builtin_fn): Likewise.
1310 * attribs.c (diag_attr_exclusions): Use new function
1311 fndecl_built_in_p and remove check for FUNCTION_DECL if
1313 (builtin_mathfn_code): Likewise.
1314 (fold_builtin_expect): Likewise.
1315 (fold_call_expr): Likewise.
1316 (fold_builtin_call_array): Likewise.
1317 (fold_call_stmt): Likewise.
1318 (set_builtin_user_assembler_name): Likewise.
1319 (is_simple_builtin): Likewise.
1320 * calls.c (gimple_alloca_call_p): Likewise.
1321 (maybe_warn_nonstring_arg): Likewise.
1322 * cfgexpand.c (expand_call_stmt): Likewise.
1323 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
1324 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
1325 (cgraph_node::verify_node): Likewise.
1326 * cgraphclones.c (build_function_decl_skip_args): Likewise.
1327 (cgraph_node::create_clone): Likewise.
1328 * config/arm/arm.c (arm_insert_attributes): Likewise.
1329 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
1330 * dse.c (scan_insn): Likewise.
1331 * expr.c (expand_expr_real_1): Likewise.
1332 * fold-const.c (operand_equal_p): Likewise.
1333 (fold_binary_loc): Likewise.
1334 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
1335 * gimple-low.c (lower_stmt): Likewise.
1336 * gimple-pretty-print.c (dump_gimple_call): Likewise.
1337 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Likewise.
1338 * gimple.c (gimple_build_call_from_tree): Likewise.
1339 (gimple_call_builtin_p): Likewise.
1340 (gimple_call_combined_fn): Likewise.
1341 * gimplify.c (gimplify_call_expr): Likewise.
1342 (gimple_boolify): Likewise.
1343 (gimplify_modify_expr): Likewise.
1344 (gimplify_addr_expr): Likewise.
1345 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
1346 * ipa-cp.c (determine_versionability): Likewise.
1347 * ipa-fnsummary.c (compute_fn_summary): Likewise.
1348 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
1349 * ipa-split.c (visit_bb): Likewise.
1350 (split_function): Likewise.
1351 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1352 * lto-cgraph.c (input_node): Likewise.
1353 * lto-streamer-out.c (write_symbol): Likewise.
1354 * omp-low.c (setjmp_or_longjmp_p): Likewise.
1355 (lower_omp_1): Likewise.
1356 * predict.c (strip_predict_hints): Likewise.
1357 * print-tree.c (print_node): Likewise.
1358 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
1359 * trans-mem.c (is_tm_irrevocable): Likewise.
1360 (is_tm_load): Likewise.
1361 (is_tm_simple_load): Likewise.
1362 (is_tm_store): Likewise.
1363 (is_tm_simple_store): Likewise.
1364 (is_tm_abort): Likewise.
1365 (tm_region_init_1): Likewise.
1366 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
1367 * tree-cfg.c (verify_gimple_call): Likewise.
1368 (move_stmt_r): Likewise.
1369 (stmt_can_terminate_bb_p): Likewise.
1370 * tree-eh.c (lower_eh_constructs_2): Likewise.
1371 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
1372 * tree-inline.c (remap_gimple_stmt): Likewise.
1373 (copy_bb): Likewise.
1374 (estimate_num_insns): Likewise.
1375 (fold_marked_statements): Likewise.
1376 * tree-sra.c (scan_function): Likewise.
1377 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
1378 (optimize_stack_restore): Likewise.
1379 (pass_fold_builtins::execute): Likewise.
1380 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
1381 (mark_all_reaching_defs_necessary_1): Likewise.
1382 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
1383 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1384 (pass_forwprop::execute): Likewise.
1385 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1386 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
1387 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
1388 * tree-ssa-strlen.c (get_string_length): Likewise.
1389 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
1390 (find_func_aliases_for_call): Likewise.
1391 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
1392 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
1393 * tree-tailcall.c (find_tail_calls): Likewise.
1394 * tree.c (need_assembler_name_p): Likewise.
1395 (free_lang_data_in_decl): Likewise.
1396 (get_call_combined_fn): Likewise.
1397 * ubsan.c (is_ubsan_builtin_p): Likewise.
1398 * varasm.c (incorporeal_function_p): Likewise.
1399 * tree.h (DECL_BUILT_IN): Remove and replace with
1401 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
1402 (fndecl_built_in_p): New.
1404 2018-08-27 Martin Liska <mliska@suse.cz>
1406 PR tree-optimization/86847
1407 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
1408 Dump also subtree probability.
1409 (switch_decision_tree::do_jump_if_equal): New function.
1410 (switch_decision_tree::emit_case_nodes): Handle special
1411 situations in balanced tree that can be emitted much simpler.
1412 Fix calculation of probabilities that happen in tree expansion.
1413 * tree-switch-conversion.h (struct cluster): Add
1415 (struct simple_cluster): Likewise.
1416 (struct case_tree_node): Add new function has_child.
1417 (do_jump_if_equal): New.
1419 2018-08-27 Martin Liska <mliska@suse.cz>
1421 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
1422 Add new argument to bit_test_cluster constructor.
1423 (bit_test_cluster::emit): Set bits really number of values
1425 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
1426 probability argument.
1427 * tree-switch-conversion.h (struct bit_test_cluster):
1428 Add m_handles_entire_switch.
1430 2018-08-27 Martin Liska <mliska@suse.cz>
1432 PR tree-optimization/86702
1433 * tree-switch-conversion.c (jump_table_cluster::emit):
1434 Make probabilities even for values in jump table
1435 according to number of cases handled.
1436 (switch_decision_tree::compute_cases_per_edge): Pass
1437 argument to reset_out_edges_aux function.
1438 (switch_decision_tree::analyze_switch_statement): Likewise.
1439 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
1442 2018-08-27 Martin Liska <mliska@suse.cz>
1444 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
1445 cfun argument explicitly.
1446 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
1447 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
1448 function gimple_switch_default_bb.
1449 (convert_switch_statements):
1451 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
1452 * stmt.c (label_to_block_fn): Use label_to_block and pass
1453 cfun argument explicitly and use gimple_switch_label_bb.
1454 (expand_case): Likewise.
1455 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
1456 cfun argument explicitly. Likewise.
1457 (make_edges_bb): Likewise.
1458 (make_cond_expr_edges): Likewise.
1459 (get_cases_for_edge): Likewise.
1460 (make_gimple_switch_edges): Likewise.
1461 (label_to_block_fn): Likewise.
1462 (label_to_block): Likewise.
1463 (make_goto_expr_edges): Likewise.
1464 (make_gimple_asm_edges): Likewise.
1465 (main_block_label): Likewise.
1466 (group_case_labels_stmt): Likewise.
1467 (find_taken_edge_computed_goto): Likewise.
1468 (find_taken_edge_switch_expr): Likewise.
1469 (gimple_verify_flow_info): Likewise.
1470 (gimple_redirect_edge_and_branch): Likewise.
1471 (gimple_switch_label_bb): New function.
1472 (gimple_switch_default_bb): Likewise.
1473 (gimple_switch_edge): Likewise.
1474 (gimple_switch_default_edge): Likewise.
1475 * tree-cfg.h (label_to_block_fn): Remove and replace ...
1476 (label_to_block): ... with this.
1477 (gimple_switch_label_bb): New.
1478 (gimple_switch_default_bb): Likewise.
1479 (gimple_switch_edge): Likewise.
1480 (gimple_switch_default_edge): Likewise.
1481 * tree-cfgcleanup.c (convert_single_case_switch): Use
1482 new gimple functions and pass new argument to label_to_block.
1483 (cleanup_control_flow_bb):
1484 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
1485 cfun argument explicitly.
1486 (make_eh_edges): Likewise.
1487 (redirect_eh_dispatch_edge): Likewise.
1488 (lower_resx): Likewise.
1489 (lower_eh_dispatch): Likewise.
1490 (maybe_remove_unreachable_handlers): Likewise.
1491 (unsplit_eh): Likewise.
1492 (cleanup_empty_eh): Likewise.
1493 (verify_eh_edges): Likewise.
1494 (verify_eh_dispatch_edge): Likewise.
1495 * tree-ssa-dom.c (record_edge_info): Likewise.
1496 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
1497 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
1498 (thread_through_normal_block): Likewise.
1499 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
1500 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
1501 * tree-switch-conversion.c (switch_conversion::collect): Use new
1503 (switch_conversion::check_final_bb): Likewise.
1504 (switch_conversion::gather_default_values): Pass new argument
1506 (switch_conversion::build_constructors): Likewise.
1507 (switch_decision_tree::compute_cases_per_edge): Use new
1508 gimple_switch_edge function.
1509 (switch_decision_tree::analyze_switch_statement): Pass new argument
1511 (switch_decision_tree::try_switch_expansion): Use
1512 gimple_switch_default_edge.
1513 * tree-vrp.c (find_switch_asserts): Pass new argument
1515 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
1516 (vr_values::simplify_switch_using_ranges): Likewise.
1518 2018-08-27 Richard Biener <rguenther@suse.de>
1520 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
1521 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
1523 * tree-ssa-sccvn.h (struct vn_pval): New structure.
1524 (struct vn_nary_op_s): Add unwind_to member. Add
1525 predicated_values flag and put result into a union together
1526 with a linked list of vn_pval.
1527 (struct vn_ssa_aux): Add name member to make maintaining
1528 a map of SSA name to vn_ssa_aux possible. Remove no longer
1529 needed info, dfsnum, low, visited, on_sccstack, use_processed
1530 and range_info_anti_range_p members.
1531 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
1532 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
1534 (vn_valueize): New global.
1535 (vn_context_bb): Likewise.
1536 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
1537 VN_INFO_PTR_INFO): Remove.
1538 * tree-ssa-sccvn.c: ... (rewrite)
1539 (pass_fre::execute): For -O2+ initialize loops and run
1540 RPO VN in optimistic mode (iterating). For -O1 and -Og
1541 run RPO VN in non-optimistic mode.
1542 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
1543 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
1544 * doc/invoke.texi (sccvn-max-scc-size): Remove.
1545 (rpo-vn-max-loop-depth): Document.
1546 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
1547 when valuezing the VUSE signals we walked out of the region.
1548 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
1549 (phi_translate): Set VN context block to use for availability
1551 (compute_avail): Likewise.
1552 (pre_valueize): New function.
1553 (pass_pre::execute): Adjust to the RPO VN API.
1555 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
1556 (propagate_constants_for_unrolling): Remove.
1557 (tree_unroll_loops_completely): Perform value-numbering
1558 on the unrolled bodies loop parent.
1560 2018-08-27 Richard Biener <rguenther@suse.de>
1562 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
1563 for partial antic compute.
1565 2018-08-27 Jakub Jelinek <jakub@redhat.com>
1567 PR rtl-optimization/87065
1568 * combine.c (simplify_if_then_else): Formatting fix.
1569 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
1571 (known_cond): Don't return const_true_rtx for vector modes. Use
1572 CONST0_RTX instead of const0_rtx. Formatting fixes.
1574 2018-08-27 Martin Liska <mliska@suse.cz>
1576 PR gcov-profile/87069
1577 * gcov.c (process_file): Record files already processed
1578 and warn about a file being processed multiple times.
1580 2018-08-27 Martin Liska <mliska@suse.cz>
1583 * config/aarch64/aarch64.c (aarch64_override_options_internal):
1584 Set default values for x_aarch64_*_string strings.
1585 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
1586 prefix. For -mabi do not print '=ABI' in help and use
1587 <option_value> format for -msve-vector-bits and -moverride
1590 2018-08-26 Jeff Law <law@redhat.com>
1592 * config/mips/frame-header-opt.c: Include "backend.h" rather than
1595 2018-08-26 Marek Polacek <polacek@redhat.com>
1597 PR c++/87029, Implement -Wredundant-move.
1598 * doc/invoke.texi: Document -Wredundant-move.
1600 2018-08-25 Martin Sebor <msebor@redhat.com>
1602 PR tree-optimization/87059
1603 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
1604 to the same type as the other.
1605 * fold-const.c (fold_binary_loc): Assert expectation.
1607 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
1609 * config/darwin.c (machopic_legitimize_pic_address): Clean up
1610 extraneous parentheses, dead code section and formatting.
1612 2018-08-24 David Malcolm <dmalcolm@redhat.com>
1615 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
1616 wide enough for jumps in the line-numbering to be visible.
1617 (layout::print_gap_in_line_numbering): New member function.
1618 (layout::calculate_line_spans): When using line numbering, merge
1619 line spans that are only 1 line apart.
1620 (diagnostic_show_locus): When printing line numbers, show gaps in
1621 line numbering directly, rather than printing headers.
1622 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
1623 line-numbering with multiple line spans.
1624 (selftest::test_fixit_insert_containing_newline_2): Add test of
1625 line-numbering, in which the spans are close enough to be merged.
1627 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
1629 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
1630 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
1631 * tree-vrp.c (range_is_nonnull): Remove.
1632 (range_includes_zero_p): Accept value_range instead of min/max.
1633 (extract_range_from_binary_expr_1): Do not early bail on
1635 Use range_includes_zero_p instead of range_is_nonnull.
1636 (extract_range_from_unary_expr): Use range_includes_zero_p instead
1637 of range_is_nonnull.
1638 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
1639 special case VR_ANTI_RANGE.
1640 (vrp_finalize): Same.
1641 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
1643 (range_is_nonnull): Remove.
1644 * vr-values.c (vrp_stmt_computes_nonzero): Use
1645 range_includes_zero_p instead of range_is_nonnull.
1646 (extract_range_basic): Pass value_range to range_includes_zero_p
1647 instead of range_is_nonnull.
1649 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
1651 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
1652 * emit-rtl.h (rtl_data): Remove return_bnd.
1653 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
1654 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
1655 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
1656 (POINTER_BOUNDS_MODE): Remove definition.
1657 (make_pointer_bounds_mode): Remove.
1658 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
1659 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
1660 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
1661 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
1662 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
1663 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
1664 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
1666 * config/i386/i386-modes.def (BND32, BND64): Remove.
1667 * config/i386/i386.c (dbx_register_map): Remove bound registers.
1668 (dbx64_register_map): Ditto.
1669 (svr4_dbx_register_map): Ditto.
1670 (indirect_thunk_bnd_needed): Remove.
1671 (indirect_thunks_bnd_used): Ditto.
1672 (indirect_return_bnd_needed): Ditto.
1673 (indirect_return_via_cx_bnd): Ditto.
1674 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
1675 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
1676 (output_indirect_thunk): Ditto. Remove need_prefix argument.
1677 (output_indirect_thunk_function): Remove handling of
1678 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
1679 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
1680 (ix86_save_reg): Remove handling of crtl->return_bnd.
1681 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
1682 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
1683 and UNSPEC_BNDLX_ADDR.
1684 (ix86_output_indirect_branch_via_reg): Remove handling of
1685 indirect_thunk_prefix_bnd.
1686 (ix86_output_indirect_branch_via_push): Ditto.
1687 (ix86_output_function_return): Ditto.
1688 (ix86_output_indirect_function_return): Ditto.
1689 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
1690 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
1691 (CALL_USED_REGISTERS): Ditto.
1692 (REG_ALLOC_ORDER): Update for removal of bound registers.
1693 (HI_REGISTER_NAMES): Ditto.
1694 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
1695 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
1696 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
1697 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
1698 (FIRST_PSEUDO_REG): Update.
1699 (BND): Remove mode iterator.
1700 * config/i386/predicates.md (bnd_mem_operator): Remove.
1702 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
1704 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
1707 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
1709 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
1710 the case in which the permute needs only a single element and
1711 repeats for every vector of the result. Extend that case to
1712 handle variable-length vectors.
1713 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
1715 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
1718 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
1719 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
1721 2018-08-24 Richard Biener <rguenther@suse.de>
1723 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
1724 bb_flags_allocated members.
1725 (auto_flag): New RAII class for allocating flags.
1726 (auto_edge_flag): New RAII class for allocating edge flags.
1727 (auto_bb_flag): New RAII class for allocating bb flags.
1728 * cfgloop.c (verify_loop_structure): Allocate temporary edge
1730 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
1731 in favor of temporarily allocated BB flag.
1732 * hsa-brig.c: Re-order includes.
1733 * hsa-dump.c: Likewise.
1734 * hsa-regalloc.c: Likewise.
1735 * print-rtl.c: Likewise.
1736 * profile-count.c: Likewise.
1738 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
1741 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
1744 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
1747 * wide-int-range.cc (wide_int_range_div): Do not ignore result
1748 from wide_int_range_multiplicative_op.
1750 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1752 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
1755 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
1757 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
1760 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
1762 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
1763 full GENERIC expression used for replacement.
1765 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
1767 * tree-vrp.c (abs_extent_range): Remove.
1768 (extract_range_into_wide_ints): Pass wide ints by reference.
1769 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
1770 Pass wide ints by reference in all calls to
1771 extract_range_into_wide_ints.
1772 * wide-int-range.cc (wide_int_range_div): New.
1773 * wide-int-range.h (wide_int_range_div): New.
1774 (wide_int_range_includes_zero_p): New.
1775 (wide_int_range_zero_p): New.
1777 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
1779 * config/aarch64/aarch64.md (arches): New enum.
1780 (arch): New enum attr.
1781 (arch_enabled): New attr.
1782 (enabled): Now uses arch_enabled only.
1783 (simd, sve, fp16): Removed attribute.
1784 (fp): Attr now defined in terms of 'arch'.
1785 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
1786 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
1787 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
1788 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
1789 attributes into 'arch'.
1790 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
1791 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
1792 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
1793 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
1794 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
1795 'simd' attribute into 'arch'.
1796 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
1797 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
1798 Convert use of 'fp' attribute to 'arch'.
1799 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
1800 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
1802 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
1803 (different modes) Merge 'fp' and 'simd' into 'arch'.
1804 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
1807 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
1809 PR rtl-optimization/87026
1810 * expmed.c (canonicalize_comparison): If we can no longer create
1811 pseudoregisters, don't.
1813 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
1816 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
1818 * config/arm/arm.c (speculation_barrier_libfunc): New static
1820 (arm_init_libfuncs): Initialize it.
1821 (arm_emit_speculation_barrier): New function.
1822 * config/arm/arm.md (speculation_barrier): Call
1823 arm_emit_speculation_barrier for architectures that do not have
1825 (speculation_barrier_insn): Only match on Armv7 or later.
1827 2018-08-23 Richard Biener <rguenther@suse.de>
1830 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
1833 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
1835 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
1836 of single-vector TBLs.
1837 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
1840 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
1843 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
1844 aarch64_evpc_tbl guard.
1846 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
1848 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
1851 2018-08-22 Martin Sebor <msebor@redhat.com>
1854 * tree-pretty-print.c (pretty_print_string): Add argument.
1855 (dump_generic_node): Call to pretty_print_string with string size.
1857 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
1859 PR rtl-optimization/86771
1860 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
1861 of two SETs into those two SETs, one to be placed at i2, if that SETs
1862 destination is modified between i2 and i3.
1864 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1866 PR tree-optimization/86725
1867 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
1869 (vect_analyze_scalar_cycles_1): Check it.
1871 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1873 PR tree-optimization/86725
1874 * tree-vect-loop.c (vect_is_simple_reduction): When treating
1875 an outer loop phi as a double reduction, make sure that the
1876 single user of the phi result is an inner loop phi.
1878 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1880 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
1881 grouped stores with gaps to a strided group.
1883 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1885 * tree-vect-stmts.c (get_group_load_store_type)
1886 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
1887 first statement in a group.
1889 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1891 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
1892 the sequence used in gcc/gcc.c.
1894 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1897 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
1900 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1902 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
1903 Darwin10-specific unwinder-shim.
1904 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
1905 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
1906 New to cater for Darwin10 Rosetta.
1908 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1910 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
1911 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
1913 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1918 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
1919 (dwarf2out_switch_text_section): Generate a local label for the
1920 second function sub-section and apply it as the second FDE start
1922 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
1923 second sub-section start.
1925 2018-08-22 Richard Biener <rguenther@suse.de>
1927 PR tree-optimization/86988
1928 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
1930 2018-08-22 Richard Biener <rguenther@suse.de>
1932 PR tree-optimization/86945
1933 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
1935 2018-08-22 Alexandre Oliva <oliva@adacore.com>
1937 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
1938 a comment about how uses of r2 for .sdata2 come about.
1940 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
1942 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
1944 2018-08-21 Marek Polacek <polacek@redhat.com>
1946 PR c++/86981, Implement -Wpessimizing-move.
1947 * doc/invoke.texi: Document -Wpessimizing-move.
1949 2018-08-21 Jan Hubicka <jh@suse.cz>
1951 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
1952 * tree.h (is_redundant_typedef): Remove.
1953 * dwarf2out.c (is_redundant_typedef): Turn into static function.
1955 2018-08-21 Jan Hubicka <jh@suse.cz>
1957 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
1960 2018-08-21 Tamar Christina <tamar.christina@arm.com>
1962 * expmed.c (extract_low_bits): Reject invalid subregs early.
1964 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
1967 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
1970 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1972 * config/vxworks.h: Guard vxworks_asm_out_constructor and
1973 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
1974 * config/vxworks.c: Likewise.
1976 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1978 * config/vxworks.c: Set targetm.have_ctors_dtors
1979 if HAVE_INITFINI_ARRAY_SUPPORT.
1980 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
1981 if HAVE_INITFINI_ARRAY_SUPPORT.
1983 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1985 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
1986 default search path for VxWorks < 7.
1988 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
1990 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
1991 (get_format_string): Refer to c_getstr.
1993 2018-08-21 Tom de Vries <tdevries@suse.de>
1995 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
1996 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
1997 (debuginfo_early_stop): Declare.
1998 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
1999 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2000 (debuginfo_early_stop): New function.
2001 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
2002 and debuginfo_early_stop.
2003 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
2004 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
2005 (general_init): Call debuginfo_early_init.
2006 (finalize): Call debuginfo_fini.
2007 (do_compile): Call debuginfo_init.
2008 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
2010 (@item -fdump-debug, @item -fdump-earlydebug): Add.
2012 2018-08-21 Tom de Vries <tdevries@suse.de>
2014 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
2015 flag_dump_noaddr and flag_dump_unnumbered.
2017 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
2019 * wide-int-range.cc (wide_int_range_abs): New.
2020 (wide_int_range_order_set): Rename from wide_int_range_min_max.
2021 * wide-int-range.h (wide_int_range_abs): New.
2022 (wide_int_range_min_max): New.
2023 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
2024 case to call wide_int_range_abs.
2025 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
2026 (extract_range_from_abs_expr): Delete.
2028 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
2031 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
2032 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
2035 2018-08-20 Nathan Sidwell <nathan@acm.org>
2036 Jeff Law <law@redhat.com>
2038 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
2039 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
2041 2018-08-20 David Malcolm <dmalcolm@redhat.com>
2044 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
2045 (decl_attributes): Likewise.
2046 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
2048 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2049 * diagnostic-core.h (class auto_diagnostic_group): New class.
2050 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
2051 (diagnostic_report_diagnostic): Handle the first diagnostics within
2053 (emit_diagnostic): Add auto_diagnostic_group instance.
2055 (inform_n): Likewise.
2056 (warning): Likewise.
2057 (warning_at): Likewise.
2058 (warning_n): Likewise.
2059 (pedwarn): Likewise.
2060 (permerror): Likewise.
2062 (error_n): Likewise.
2063 (error_at): Likewise.
2065 (fatal_error): Likewise.
2066 (internal_error): Likewise.
2067 (internal_error_no_backtrace): Likewise.
2068 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
2069 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
2070 * diagnostic.h (struct diagnostic_context): Add fields
2071 "diagnostic_group_nesting_depth",
2072 "diagnostic_group_emission_count", "begin_group_cb",
2074 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
2075 Add auto_diagnostic_group instance(s).
2076 (find_explicit_erroneous_behavior): Likewise.
2077 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
2078 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
2079 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
2080 (gimplify_va_arg_expr): Likewise.
2081 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
2082 (HSA_SORRY_AT): Likewise.
2083 * ipa-devirt.c (compare_virtual_tables): Likewise.
2084 (warn_odr): Likewise.
2085 * multiple_target.c (expand_target_clones): Likewise.
2086 * opts-common.c (cmdline_handle_error): Likewise.
2087 * reginfo.c (globalize_reg): Likewise.
2088 * substring-locations.c (format_warning_n_va): Likewise.
2089 * tree-inline.c (expand_call_inline): Likewise.
2090 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
2091 * tree-ssa-loop-niter.c
2092 (do_warn_aggressive_loop_optimizations): Likewise.
2093 * tree-ssa-uninit.c (warn_uninit): Likewise.
2094 * tree.c (warn_deprecated_use): Likewise.
2096 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
2099 * config/i386/i386.md (eh_return): Always update EH return
2100 address in word_mode.
2102 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
2104 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
2105 TARGET_SPLIT_COMPLEX_ARG is defined.
2107 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
2109 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
2111 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
2114 * expr.c (expand_assignment): Assert that bitpos is positive.
2115 (store_field): Likewise
2116 (expand_expr_real_1): Make sure that bitpos is positive.
2117 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
2120 2018-08-20 Nathan Sidwell <nathan@acm.org>
2122 * Makefile.in (CPP_ID_DATA_H): Delete.
2123 (CPP_INTERNAL_H): Don't add it.
2124 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
2125 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
2127 2018-08-20 Richard Biener <rguenther@suse.de>
2129 PR tree-optimization/78655
2130 * tree-vrp.c (extract_range_from_binary_expr_1): Make
2131 pointer + offset nonnull if either operand is nonnull work.
2133 2018-08-20 Tom de Vries <tdevries@suse.de>
2135 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
2136 unless the referenced die describes the added property using
2137 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
2138 Otherwise, add a DW_AT_location to the referenced die.
2140 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
2143 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
2144 register_operand when calling ix86_set_reg_reg_cost.
2145 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
2146 Set *total to 0 for operands that satisfy x86_64_immediate_operand
2147 predicate and to 1 otherwise.
2149 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
2151 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
2152 emit a diagnostic that it is not supported and reset the option.
2153 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
2154 supported and consume the option. (ASM_FINAL_SPEC): New.
2156 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
2158 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
2161 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
2163 C-SKY port: Documentation
2165 * doc/extend.texi (C-SKY Function Attributes): New section.
2166 * doc/invoke.texi (Option Summary): Add C-SKY options.
2167 (C-SKY Options): New section.
2168 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
2170 2018-08-17 Jojo <jijie_rong@c-sky.com>
2171 Huibin Wang <huibin_wang@c-sky.com>
2172 Sandra Loosemore <sandra@codesourcery.com>
2173 Chung-Lin Tang <cltang@codesourcery.com>
2175 C-SKY port: Backend implementation
2177 * config/csky/*: New.
2178 * common/config/csky/*: New.
2180 2018-08-17 Jojo <jijie_rong@c-sky.com>
2181 Huibin Wang <huibin_wang@c-sky.com>
2182 Sandra Loosemore <sandra@codesourcery.com>
2183 Chung-Lin Tang <cltang@codesourcery.com>
2184 Andrew Jenner <andrew@codesourcery.com>
2186 C-SKY port: Configury
2188 * config.gcc (csky-*-*): New.
2189 * configure.ac: Add csky to targets for dwarf2 debug_line support.
2190 * configure: Regenerated.
2192 2018-08-17 David Malcolm <dmalcolm@redhat.com>
2194 * dump-context.h: Include "dumpfile.h".
2195 (dump_context::dump_printf_va): Convert final param from va_list
2196 to va_list *. Convert from ATTRIBUTE_PRINTF to
2197 ATTRIBUTE_GCC_DUMP_PRINTF.
2198 (dump_context::dump_printf_loc_va): Likewise.
2199 * dumpfile.c: Include "stringpool.h".
2200 (make_item_for_dump_printf_va): Delete.
2201 (make_item_for_dump_printf): Delete.
2202 (class dump_pretty_printer): New class.
2203 (dump_pretty_printer::dump_pretty_printer): New ctor.
2204 (dump_pretty_printer::emit_items): New member function.
2205 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
2207 (dump_pretty_printer::emit_item): New member function.
2208 (dump_pretty_printer::stash_item): New member function.
2209 (dump_pretty_printer::format_decoder_cb): New member function.
2210 (dump_pretty_printer::decode_format): New member function.
2211 (dump_context::dump_printf_va): Reimplement in terms of
2212 dump_pretty_printer.
2213 (dump_context::dump_printf_loc_va): Convert final param from va_list
2215 (dump_context::begin_scope): Reimplement call to
2216 make_item_for_dump_printf.
2217 (dump_printf): Update for change to dump_printf_va.
2218 (dump_printf_loc): Likewise.
2219 (selftest::test_capture_of_dump_calls): Convert "stmt" from
2220 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
2221 with %T, %E, and %G.
2222 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
2223 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
2224 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
2225 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
2226 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
2227 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
2228 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
2229 within a dump_printf_loc call to "%wu".
2230 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
2231 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
2232 missing space after "=".
2233 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
2234 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
2235 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
2236 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
2237 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
2238 duplicate "vectorized" from message.
2240 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
2242 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
2243 polyNxK_t element's TYPE_STRING_FLAG.
2245 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
2247 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
2248 (they were unnamed before). Fix comments.
2250 2018-08-17 Nathan Sidwell <nathan@acm.org>
2252 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
2254 2018-08-17 Richard Biener <rguenther@suse.de>
2256 PR tree-optimization/86841
2257 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
2259 2018-08-17 Martin Liska <mliska@suse.cz>
2261 * common.opt: Remove Warn, Init and Report for options with
2262 Ignore/Deprecated flag. Warning is done automatically for
2264 * config/i386/i386.opt: Likewise.
2265 * config/ia64/ia64.opt: Likewise.
2266 * config/rs6000/rs6000.opt: Likewise.
2267 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
2268 Remove usage of flag_check_pointer_bounds.
2269 * lto-wrapper.c (merge_and_complain): Do not handle
2270 OPT_fcheck_pointer_bounds.
2271 (append_compiler_options): Likewise.
2272 * opt-functions.awk: Do not handle Deprecated.
2273 * optc-gen.awk: Check that Var, Report and Init are not
2274 used for an option with Ignore/Deprecated flag.
2275 * opts-common.c (decode_cmdline_option): Do not report
2277 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
2279 * opts.h (struct cl_option): Remove cl_deprecated flag.
2280 (CL_ERR_DEPRECATED): Remove error enum value.
2282 2018-08-17 Richard Biener <rguenther@suse.de>
2285 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
2286 across a va-arg-pack using call adjust its return value accordingly.
2288 2018-08-16 Martin Sebor <msebor@redhat.com>
2290 PR tree-optimization/86853
2291 * gimple-ssa-sprintf.c (struct format_result): Rename member.
2292 (struct fmtresult): Add member and initialize it in ctors.
2293 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
2294 (format_string): Handle %S the same as %ls. Set MAYFAIL.
2295 (format_directive): Set POSUNDER4K when MAYFAIL is set.
2296 (parse_directive): Handle %C same as %c.
2297 (sprintf_dom_walker::compute_format_length): Adjust.
2298 (is_call_safe): Adjust.
2300 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
2302 * builtins.c (c_strlen): Add new parameter eltsize. Use it
2303 for determining how to count the elements.
2304 * builtins.h (c_strlen): Adjust prototype.
2305 * expr.c (string_constant): Add new parameter mem_size.
2306 Set *mem_size appropriately.
2307 * expr.h (string_constant): Adjust protoype.
2308 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
2309 * gimple-fold.h (get_range_strlen): Adjust prototype.
2310 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
2311 (format_string): Call get_string_length with eltsize.
2313 2018-08-16 David Malcolm <dmalcolm@redhat.com>
2315 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
2316 to emit the span, rather than setting it as the prefix.
2318 2018-08-16 David Malcolm <dmalcolm@redhat.com>
2320 * diagnostic-show-locus.c (layout::start_annotation_line): Add
2321 "margin_char" parameter, defaulting to space. Use it in place
2322 of pp_space for the initial part of the margin.
2323 (layout::print_leading_fixits): Use '+' when filling the margin
2324 of line-insertion fix-it hints.
2326 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2328 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
2331 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2333 * config/rs6000/altivec.md: Don't set length attribute to the default
2335 * config/rs6000/darwin.md: Ditto.
2336 * config/rs6000/dfp.md: Ditto.
2337 * config/rs6000/htm.md: Ditto.
2338 * config/rs6000/rs6000.md: Ditto.
2339 * config/rs6000/sync.md: Ditto.
2340 * config/rs6000/vsx.md: Ditto.
2342 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2344 * config/rs6000/altivec.md: Don't set length attribute to the default
2345 value, for branch instructions.
2346 * config/rs6000/darwin.md: Ditto.
2347 * config/rs6000/rs6000.md: Ditto.
2349 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2351 * config/rs6000/rs6000.md (length): Always define as const_int 4.
2352 (unnamed conditional branch define_insn): Set length to 4 or 8
2353 depending on offset.
2354 (<bd>_<mode>): Similar, for alternative 0.
2355 (<bd>tf_<mode>): Ditto.
2357 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2359 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
2361 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
2363 * doc/rtl.texi: Replace old RTX class names with new names.
2366 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
2368 * expmed.h (canonicalize_comparison): New declaration.
2369 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
2370 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
2371 * optabs.c (prepare_cmp_insn): Likewise.
2372 * rtl.h (unsigned_condition_p): New function which checks if a
2373 comparison operator is unsigned.
2375 2018-08-16 Nathan Sidwell <nathan@acm.org>
2377 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
2378 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
2380 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2383 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
2384 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
2385 (mov<mov>): ..this and enable unconditionally.
2387 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2389 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
2391 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
2393 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
2394 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
2395 "Common" with "Target".
2397 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
2399 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
2400 * doc/invoke.texi (mmitigate-rop): Remove.
2401 * config/i386/i386.c: Do not include "regrename.h".
2402 (ix86_rop_should_change_byte_p, reg_encoded_number)
2403 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
2405 (ix86_reorg): Remove call to ix86_mitigate_rop.
2406 * config/i386/i386.md (attr "modrm_class"): Remove.
2407 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
2408 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
2409 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
2411 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
2413 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
2414 allow folding of mergeh() and mergel() for the float and double types.
2415 (fold_mergehl_helper): Rework to handle building a permute tree
2418 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
2420 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
2423 2018-08-15 David Malcolm <dmalcolm@redhat.com>
2425 * common.opt (fdiagnostics-show-labels): New option.
2426 * diagnostic-show-locus.c (class layout_range): Add field
2428 (class layout): Add field "m_show_labels_p".
2429 (layout_range::layout_range): Add param "label" and use it to
2431 (make_range): Pass in NULL for new "label" param of layout_range's
2433 (layout::layout): Initialize m_show_labels_p.
2434 (layout::maybe_add_location_range): Pass in loc_range->m_label
2435 when constructing layout_range instances.
2436 (struct line_label): New struct.
2437 (layout::print_any_labels): New member function.
2438 (layout::print_line): Call it if label-printing is enabled.
2439 (selftest::test_one_liner_labels): New test.
2440 (selftest::test_diagnostic_show_locus_one_liner): Call it.
2441 * diagnostic.c (diagnostic_initialize): Initialize
2442 context->show_labels_p.
2443 * diagnostic.h (struct diagnostic_context): Add field
2445 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
2446 -fno-diagnostics-show-labels.
2447 * dwarf2out.c (gen_producer_string): Add
2448 OPT_fdiagnostics_show_labels to the ignored options.
2449 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
2451 (gcc_rich_location::maybe_add_expr): Likewise.
2452 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
2453 label" param, defaulting to NULL.
2454 (gcc_rich_location::add_expr): Add "label" param.
2455 (gcc_rich_location::maybe_add_expr): Likewise.
2456 (class text_range_label): New class.
2457 (class range_label_for_type_mismatch): New class.
2458 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
2459 of format_warning_va.
2460 (fmtwarn_n): Likewise for new params of format_warning_n_va.
2461 * lto-wrapper.c (merge_and_complain): Add
2462 OPT_fdiagnostics_show_labels to the "pick one setting" options.
2463 (append_compiler_options): Likewise to the dropped options.
2464 (append_diag_options): Likewise to the passed-on options.
2465 * opts.c (common_handle_option): Handle the new option.
2466 * selftest-diagnostic.c
2467 (test_diagnostic_context::test_diagnostic_context): Enable
2469 * substring-locations.c: Include "gcc-rich-location.h".
2470 (format_warning_n_va): Add "fmt_label" and "param_label" params
2471 and use them as appropriate.
2472 (format_warning_va): Add "fmt_label" and "param_label" params,
2473 passing them on to format_warning_n_va.
2474 (format_warning_at_substring): Likewise.
2475 (format_warning_at_substring_n): Likewise.
2476 * substring-locations.h (format_warning_va): Add "fmt_label" and
2477 "param_label" params.
2478 (format_warning_n_va): Likewise.
2479 (format_warning_at_substring): Likewise.
2480 (format_warning_at_substring_n): Likewise.
2481 * toplev.c (general_init): Initialize global_dc->show_labels_p.
2483 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
2486 * builtins.c (expand_builtin_memcmp): Do not expand the call
2487 when overflow is detected.
2489 2018-08-15 Martin Sebor <msebor@redhat.com>
2491 PR tree-optimization/71625
2492 * config/aarch64/aarch64-builtins.c
2493 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
2495 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
2497 * config/s390/s390.c (s390_reorg): Remove loop.
2499 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
2502 (darwin_function_switched_text_sections): Delete.
2503 * gcc/config/darwin.h
2504 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
2506 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
2509 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
2510 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
2511 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
2513 2018-08-15 Martin Liska <mliska@suse.cz>
2515 PR tree-optimization/86925
2516 * predict.c (expr_expected_value_1): When taking
2517 later predictor, assign also probability.
2518 Use fold_build2_initializer_loc in order to fold
2519 the expression in -frounding-math.
2521 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
2523 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
2525 (expand_vec_perm_1): Try the new method.
2527 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
2530 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
2531 Check whether lra_live_max_point is 0 before dividing.
2533 2018-08-14 Martin Sebor <msebor@redhat.com>
2535 PR tree-optimization/86650
2536 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
2537 (vrp_prop::check_mem_ref): Same.
2539 2018-08-13 Liu Hao <lh_mouse@126.com>
2541 * pretty-print.c (eat_esc_sequence): Swap the foreground and
2542 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
2543 and clear it thereafter, as it only works for DBCS.
2545 2018-08-13 Liu Hao <lh_mouse@126.com>
2547 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
2548 handle returned by _get_osf_handle().
2550 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
2552 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
2553 for folding vec_perm.
2555 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
2557 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
2558 Add support for gimple-folding of vec_pack() and vec_unpack()
2561 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
2563 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
2564 vec_xst variants to the list.
2565 (rs6000_gimple_fold_builtin): Add support for folding unaligned
2566 vector loads and stores.
2568 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
2570 * config.gcc (rs6000-ibm-aix4.x): Delete.
2571 (rs6000-ibm-aix5.1): Delete.
2572 (rs6000-ibm-aix5.2): Delete.
2573 (rs6000-ibm-aix5.3): Delete.
2574 * config/rs6000/aix43.h: Delete.
2575 * config/rs6000/aix51.h: Delete.
2576 * config/rs6000/aix52.h: Delete.
2577 * config/rs6000/t-aix43: Delete.
2579 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
2581 * config/s390/s390.c (s390_decompose_constant_pool_ref):
2583 (s390_decompose_address): Factor out constant pool ref
2586 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
2588 * config/nds32/nds32-predicates.c
2589 (nds32_can_use_bclr_p): Change return type as bool.
2590 (nds32_can_use_bset_p): Ditto.
2591 (nds32_can_use_btgl_p): Ditto.
2592 (nds32_can_use_bitci_p): Ditto.
2593 * config/nds32/nds32-protos.h
2594 (nds32_can_use_bclr_p): Change declaration.
2595 (nds32_can_use_bset_p): Ditto.
2596 (nds32_can_use_btgl_p): Ditto.
2597 (nds32_can_use_bitci_p): Ditto.
2599 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
2601 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
2602 Support -msched-prolog-epilog option.
2603 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
2605 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
2607 * common/config/nds32/nds32-common.c
2608 (nds32_option_optimization_table): Enalbe -malways-align.
2610 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
2612 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
2614 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
2615 OPT_misr_secure_ case.
2616 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
2617 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
2618 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
2620 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
2621 (nds32_isr_info): New field security_level.
2622 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
2623 * config/nds32/nds32.md (return_internal): Consider critical attribute.
2624 * config/nds32/nds32.opt (misr-secure): New option.
2625 * config/nds32/nds32_init.inc: New file.
2626 * config/nds32/nds32_isr.h: New file.
2628 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
2630 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
2631 Update comment for atomic instructions.
2632 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
2633 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
2635 (atomic_loaddi): Revise fence expansion to only emit fence prior to
2636 load for __ATOMIC_SEQ_CST model.
2637 (atomic_loaddi_1): Remove float register target.
2638 (atomic_storedi): Handle CONST_INT values.
2639 (atomic_storedi_1): Remove float register source. Add special case
2641 (memory_barrier): New expander and insn.
2643 2018-08-11 Jakub Jelinek <jakub@redhat.com>
2645 PR tree-optimization/86835
2646 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
2647 new_stmt after def_gsi, make sure to insert new_square_stmt after
2648 that stmt, not 2 stmts before it.
2650 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
2653 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
2656 2018-08-10 Martin Liska <mliska@suse.cz>
2659 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
2661 * builtins.c (expand_builtin_expect_with_probability):
2663 (expand_builtin_expect_with_probability): New function.
2664 (build_builtin_expect_predicate): Add new argumnet probability
2665 for BUILT_IN_EXPECT_WITH_PROBABILITY.
2666 (fold_builtin_expect):
2669 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
2670 * builtins.h (fold_builtin_expect): Set new argument.
2671 * doc/extend.texi: Document __builtin_expect_with_probability.
2672 * doc/invoke.texi: Likewise.
2673 * gimple-fold.c (gimple_fold_call): Pass new argument.
2674 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
2675 also BUILT_IN_EXPECT_WITH_PROBABILITY.
2676 * predict.c (get_predictor_value): New function.
2677 (expr_expected_value): Add new argument probability. Assume
2678 that predictor and probability are always non-null.
2679 (expr_expected_value_1): Likewise. For __builtin_expect and
2680 __builtin_expect_with_probability set probability. Handle
2681 combination in binary expressions.
2682 (tree_predict_by_opcode): Simplify code by simply calling
2683 get_predictor_value.
2684 (pass_strip_predict_hints::execute): Add handling of
2685 BUILT_IN_EXPECT_WITH_PROBABILITY.
2686 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
2688 * tree.h (DECL_BUILT_IN_P): New function.
2690 2018-08-10 Martin Liska <mliska@suse.cz>
2692 PR tree-optimization/85799
2693 * passes.def: Add argument for pass_strip_predict_hints.
2694 * predict.c (class pass_strip_predict_hints): Add new argument
2696 (strip_predictor_early): New function.
2697 (pass_strip_predict_hints::execute): Call the function to
2699 (strip_predict_hints): New function.
2700 * predict.def: Fix comment.
2702 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
2704 * Makefile.in: Clarify which tm.texi to copy over to assert the
2705 right to grant a GFDL license for all.
2707 2018-08-09 Jeff Law <law@redhat.com>
2709 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
2712 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
2714 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
2717 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2719 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
2720 reductions for variable-length vectors.
2722 2018-08-09 David Malcolm <dmalcolm@redhat.com>
2725 * common.opt (fdiagnostics-show-line-numbers): New option.
2726 * diagnostic-show-locus.c (class layout): Add fields
2727 "m_show_line_numbers_p" and "m_linenum_width";
2728 (num_digits): New function.
2729 (test_num_digits): New function.
2730 (layout::layout): Initialize new fields. Update m_x_offset
2731 logic to handle any left margin.
2732 (layout::print_source_line): Print line number when requested.
2733 (layout::start_annotation_line): New member function.
2734 (layout::print_annotation_line): Call it.
2735 (layout::print_leading_fixits): Likewise.
2736 (layout::print_trailing_fixits): Likewise. Update calls to
2737 move_to_column for new parameter.
2738 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
2739 it to potentially call start_annotation_line.
2740 (layout::show_ruler): Call start_annotation_line.
2741 (selftest::test_line_numbers_multiline_range): New selftest.
2742 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
2743 and selftest::test_line_numbers_multiline_range.
2744 * diagnostic.c (diagnostic_initialize): Initialize
2745 show_line_numbers_p.
2746 * diagnostic.h (struct diagnostic_context): Add field
2747 "show_line_numbers_p".
2748 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
2749 -fno-diagnostics-show-line-numbers.
2750 * dwarf2out.c (gen_producer_string): Add
2751 OPT_fdiagnostics_show_line_numbers to the ignored options.
2752 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
2753 one setting" options.
2754 (append_compiler_options): Likewise to the dropped options.
2755 (append_diag_options): Likewise to the passed-on options.
2756 * opts.c (common_handle_option): Handle the new option.
2757 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
2759 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
2761 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
2762 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
2763 third argument of type "const signed char" to descriptions of
2764 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
2765 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
2766 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
2767 __builtin_bcdsub_ov functions.
2769 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2771 PR tree-optimization/86858
2772 * tree-vect-loop.c (vect_is_simple_reduction): Restore
2773 flow_bb_inside_loop_p calls.
2775 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2777 PR tree-optimization/86871
2778 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
2779 instead of gimple_assign_lhs.
2781 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
2784 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
2785 register constraint to operand 0.
2786 (add<mode>3_carryinC): Likewise.
2787 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
2789 2018-08-09 Martin Liska <mliska@suse.cz>
2792 * common.opt: Remove extra line.
2794 2018-08-09 Martin Liska <mliska@suse.cz>
2796 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
2797 at the end of a line, make first letter capital and end up
2798 a sentence with a dot.
2799 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
2800 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
2801 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
2802 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
2803 (PARAM_MAX_ISL_OPERATIONS): Likewise.
2804 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
2805 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
2806 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
2807 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
2808 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
2809 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
2810 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
2811 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
2812 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
2813 (PARAM_TREE_REASSOC_WIDTH): Likewise.
2814 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
2815 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
2816 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
2818 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
2821 * config/s390/s390.c (s390_option_override_internal): Reduce the
2822 stack-clash-protection-probe-interval param if it would be too big
2825 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
2828 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
2829 * config/m68k/m68k.c (handle_move_double): Don't call
2830 m68k_final_prescan_insn.
2831 (m68k_adjust_decorated_operand): Renamed from
2832 m68k_final_prescan_insn, remove first and third operand and
2834 (print_operand): Call it.
2835 (print_operand_address): Call it.
2837 2018-08-08 Nathan Sidwell <nathan@acm.org>
2839 * diagnostic.c (diagnostic_report_current_module): Use
2840 linemap_included_from & linemap_included_from_linemap.
2842 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
2844 * config/aarch64/aarch64-cores.def: Add phecda core.
2845 * config/aarch64/aarch64-tune.md: Regenerate.
2846 * doc/invoke.texi: Add phecda core.
2848 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
2851 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
2853 * config/s390/s390.md ("movti"): Add more alternatives for
2854 constant to GPR copies.
2856 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
2858 * config/s390/s390.c: Fix whitespace damage throughout the file.
2859 * config/s390/s390.h: Likewise.
2860 * config/s390/tpf.h: Likewise.
2862 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
2864 * config/s390/s390.c (s390_loadrelative_operand_p):
2865 Remove TARGET_CPU_ZARCH usages.
2866 (s390_rtx_costs): Likewise.
2867 (s390_legitimate_constant_p): Likewise.
2868 (s390_cannot_force_const_mem): Likewise.
2869 (legitimate_reload_constant_p): Likewise.
2870 (s390_preferred_reload_class): Likewise.
2871 (legitimize_pic_address): Likewise.
2872 (legitimize_tls_address): Likewise.
2873 (s390_split_branches): Removed.
2874 (s390_add_execute): Removed.
2875 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
2876 (s390_mainpool_start): Likewise.
2877 (s390_mainpool_finish): Likewise.
2878 (s390_mainpool_cancel): Removed.
2879 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
2880 (s390_chunkify_cancel): Likewise.
2881 (s390_return_addr_rtx): Likewise.
2882 (s390_register_info): Remove split_branches_pending_p uages.
2883 (s390_optimize_register_info): Likewise.
2884 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
2885 split_branches_pending_p usages.
2886 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
2887 (s390_load_got): Likewise.
2888 (s390_expand_split_stack_prologue): Likewise.
2889 (output_asm_nops): Likewise.
2890 (s390_function_profiler): Likewise.
2891 (s390_emit_call): Likewise.
2892 (s390_conditional_register_usage): Likewise.
2893 (s390_optimize_prologue): Likewise.
2894 (s390_reorg): Remove TARGET_CPU_ZARCH and
2895 split_branches_pending_p usages.
2896 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
2898 (s390_output_indirect_thunk_function): Likewise.
2899 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
2900 (TARGET_CPU_ZARCH_P): Removed.
2901 (struct machine_function): Remove split_branches_pending_p.
2902 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
2904 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
2906 * common/config/s390/s390-common.c (processor_flags_table):
2908 * config.gcc: Remove with_arch/with_tune support.
2909 * config/s390/2064.md: Remove cpu attribute comparisons.
2910 * config/s390/driver-native.c (s390_host_detect_local_cpu):
2912 * config/s390/linux.h (ASM_SPEC):
2913 Remove -march support.
2914 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
2915 Use a table to get an arch level.
2916 * config/s390/s390-opts.h (enum processor_type):
2918 * config/s390/s390.c
2919 (processor_table): Remove entries, add arch_level values.
2920 (s390_issue_rate): Remove cases.
2921 (s390_option_override): Adjust
2922 s390_option_override_internal() call.
2923 (s390_option_override_internal): Remove deprecation warning.
2924 (s390_valid_target_attribute_tree): Adjust
2925 s390_option_override_internal() call.
2926 * config/s390/s390.h (struct s390_processor):
2927 Share with s390-c.c, add arch_level field.
2928 * config/s390/s390.md:
2929 Remove occurrences in cpu attribute.
2930 * config/s390/s390.opt: Remove -march/-mtune support.
2931 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
2932 * doc/invoke.texi: Remove deprecation warning.
2934 2018-08-08 Luis Machado <luis.machado@linaro.org>
2936 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
2938 (qdf24xx_tunings): Set vector cost structure to
2939 qdf24xx_vector_cost.
2941 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
2942 <register_sextend>: Set to 3.
2944 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
2947 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
2948 * config/aarch64/aarch64-simd.md
2949 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
2950 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
2951 (aarch64_frecpx<mode>): ...this new pattern.
2952 * config/aarch64/aarch64-simd-builtins.def: Remove comment
2953 about aarch64_frecp<FRECP:frecp_suffix><mode>.
2955 2018-08-07 Martin Liska <mliska@suse.cz>
2958 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
2959 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
2960 * predict.def (PRED_MALLOC_NONNULL): New predictor.
2961 * doc/extend.texi: Document that malloc attribute adds
2964 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
2967 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2968 Define to speculation_safe_value_not_needed.
2970 2018-08-06 Jeff Law <law@redhat.com>
2972 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
2973 the vr_values instance to cprop_into_stmt.
2974 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
2975 (cprop_operand): Also query EVRP to determine if OP is a constant.
2977 2018-08-06 Nathan Sidwell <nathan@acm.org>
2979 * diagnostic.c (diagnostic_report_current_module): Reroll
2980 included-at loop. Translate text.
2982 2018-08-06 David Malcolm <dmalcolm@redhat.com>
2984 * function-tests.c (selftest::test_expansion_to_rtl): Call
2985 free_after_compilation.
2987 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2989 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
2991 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
2993 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
2994 loops with memory block operations from getting unrolled.
2996 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
2999 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3000 Define to speculation_safe_value_not_needed.
3002 2018-08-06 Jeff Law <law@redhat.com>
3004 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
3007 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3010 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
3011 with all enabled __intN types.
3013 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
3015 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3017 * alias.c (record_set): Check for clobber high.
3018 * cfgexpand.c (expand_gimple_stmt): Likewise.
3019 * combine-stack-adj.c (single_set_for_csa): Likewise.
3020 * combine.c (find_single_use_1): Likewise.
3021 (set_nonzero_bits_and_sign_copies): Likewise.
3022 (get_combine_src_dest): Likewise.
3023 (is_parallel_of_n_reg_sets): Likewise.
3024 (try_combine): Likewise.
3025 (record_dead_and_set_regs_1): Likewise.
3026 (reg_dead_at_p_1): Likewise.
3027 (reg_dead_at_p): Likewise.
3028 * dce.c (deletable_insn_p): Likewise.
3029 (mark_nonreg_stores_1): Likewise.
3030 (mark_nonreg_stores_2): Likewise.
3031 * df-scan.c (df_find_hard_reg_defs): Likewise.
3032 (df_uses_record): Likewise.
3033 (df_get_call_refs): Likewise.
3034 * dwarf2out.c (mem_loc_descriptor): Likewise.
3035 * haifa-sched.c (haifa_classify_rtx): Likewise.
3036 * ira-build.c (create_insn_allocnos): Likewise.
3037 * ira-costs.c (scan_one_insn): Likewise.
3038 * ira.c (equiv_init_movable_p): Likewise.
3039 (rtx_moveable_p): Likewise.
3040 (interesting_dest_for_shprep): Likewise.
3041 * jump.c (mark_jump_label_1): Likewise.
3042 * postreload-gcse.c (record_opr_changes): Likewise.
3043 * postreload.c (reload_cse_simplify): Likewise.
3044 (struct reg_use): Add source expr.
3045 (reload_combine): Check for clobber high.
3046 (reload_combine_note_use): Likewise.
3047 (reload_cse_move2add): Likewise.
3048 (move2add_note_store): Likewise.
3049 * print-rtl.c (print_pattern): Likewise.
3050 * recog.c (decode_asm_operands): Likewise.
3051 (store_data_bypass_p): Likewise.
3052 (if_test_bypass_p): Likewise.
3053 * regcprop.c (kill_clobbered_value): Likewise.
3054 (kill_set_value): Likewise.
3055 * reginfo.c (reg_scan_mark_refs): Likewise.
3056 * reload1.c (maybe_fix_stack_asms): Likewise.
3057 (eliminate_regs_1): Likewise.
3058 (elimination_effects): Likewise.
3059 (mark_not_eliminable): Likewise.
3060 (scan_paradoxical_subregs): Likewise.
3061 (forget_old_reloads_1): Likewise.
3062 * reorg.c (find_end_label): Likewise.
3063 (try_merge_delay_insns): Likewise.
3064 (redundant_insn): Likewise.
3065 (own_thread_p): Likewise.
3066 (fill_simple_delay_slots): Likewise.
3067 (fill_slots_from_thread): Likewise.
3068 (dbr_schedule): Likewise.
3069 * resource.c (update_live_status): Likewise.
3070 (mark_referenced_resources): Likewise.
3071 (mark_set_resources): Likewise.
3072 * rtl.c (copy_rtx): Likewise.
3073 * rtlanal.c (reg_referenced_p): Likewise.
3074 (single_set_2): Likewise.
3075 (noop_move_p): Likewise.
3076 (note_stores): Likewise.
3077 * sched-deps.c (sched_analyze_reg): Likewise.
3078 (sched_analyze_insn): Likewise.
3080 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3082 * cse.c (invalidate_reg): New function extracted from...
3083 (invalidate): ...here.
3084 (canonicalize_insn): Check for clobber high.
3085 (invalidate_from_clobbers): invalidate clobber highs.
3086 (invalidate_from_sets_and_clobbers): Likewise.
3087 (count_reg_usage): Check for clobber high.
3088 (insn_live_p): Likewise.
3089 * cselib.c (cselib_expand_value_rtx_1):Likewise.
3090 (cselib_invalidate_regno): Check for clobber in setter.
3091 (cselib_invalidate_rtx): Pass through setter.
3092 (cselib_invalidate_rtx_note_stores):
3093 (cselib_process_insn): Check for clobber high.
3094 * cselib.h (cselib_invalidate_rtx): Add operand.
3096 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3098 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
3099 (mark_not_eliminable): Likewise.
3100 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
3101 * lra-lives.c (process_bb_lives): Check for clobber high.
3102 * lra.c (new_insn_reg): Remember clobber highs.
3103 (collect_non_operand_hard_regs): Check for clobber high.
3104 (lra_set_insn_recog_data): Likewise.
3105 (add_regs_to_insn_regno_info): Likewise.
3106 (lra_update_insn_regno_info): Likewise.
3108 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3110 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
3111 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
3113 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3115 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
3116 (copy_insn_1): Likewise.
3117 (gen_hard_reg_clobber_high): New gen function.
3118 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
3119 * genemit.c (gen_exp): Likewise.
3120 (gen_emit_seq): Pass through info.
3121 (gen_insn): Check for CLOBBER_HIGH.
3122 (gen_expand): Pass through info.
3123 (gen_split): Likewise.
3124 (output_add_clobbers): Likewise.
3125 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
3126 (remove_clobbers): Likewise.
3127 * rtl.h (gen_hard_reg_clobber_high): New declaration.
3129 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3131 * doc/rtl.texi (clobber_high): Add.
3132 (parallel): Add in clobber high
3133 * rtl.c (rtl_check_failed_code3): Add function.
3134 * rtl.def (CLOBBER_HIGH): Add expression.
3135 * rtl.h (RTL_CHECKC3): Add macro.
3136 (rtl_check_failed_code3): Add declaration.
3137 (XC3EXP): Add macro.
3139 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
3142 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
3143 cfun->machine->max_used_stack_alignment if needed.
3145 2018-08-04 Martin Sebor <msebor@redhat.com>
3147 PR tree-optimization/86571
3148 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
3151 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
3154 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3157 2018-08-03 Jeff Law <law@redhat.com>
3160 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3161 Define to speculation_safe_value_not_needed.
3163 2018-08-03 David Malcolm <dmalcolm@redhat.com>
3165 * doc/gcov.texi (-x): Remove duplicate "to".
3166 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
3167 (-Wif-not-aligned): Remove duplicate "is".
3168 (-flto): Remove duplicate "the".
3169 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
3170 duplicate "v5.00.b".
3171 (MSP430 Options): Remove duplicate "and" from the description
3172 of "-mgprel-sec=regexp".
3173 (x86 Options): Remove duplicate copies of "vmldLog102" and
3174 vmlsLog104 from description of "-mveclibabi=type".
3176 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
3178 * internal-fn.h (first_commutative_argument): Declare.
3179 * internal-fn.c (first_commutative_argument): New function.
3180 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
3181 restrictions for pattern statements. Use first_commutative_argument
3182 to look for commutative operands in calls to internal functions.
3184 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
3186 * Makefile.in (wide-int-range.o): New.
3187 * tree-vrp.c: Move all the wide_int_* functions to...
3188 * wide-int-range.cc: ...here.
3189 * tree-vrp.h: Move all the wide_int_* prototypes to...
3190 * wide-int-range.h: ...here.
3192 2018-08-03 Tom de Vries <tdevries@suse.de>
3194 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
3196 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
3197 * except.c (output_function_exception_table): Do early exit if
3198 targetm_common.except_unwind_info (&global_options) == UI_NONE.
3200 2018-08-03 Martin Liska <mliska@suse.cz>
3202 * predict.c (dump_prediction): Change to 2 digits
3205 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
3207 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
3209 (falkor_am_1_gtov_gtov): ... a new insn reservation.
3211 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
3213 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
3214 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
3215 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
3216 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
3217 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
3219 2018-08-02 David Malcolm <dmalcolm@redhat.com>
3221 * diagnostic-show-locus.c (diagnostic_show_locus): Use
3222 pp_take_prefix when saving the existing prefix.
3223 * diagnostic.c (diagnostic_append_note): Likewise.
3224 * langhooks.c (lhd_print_error_function): Likewise.
3225 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
3226 param's type. Free the existing prefix.
3227 (pp_take_prefix): New function.
3228 (pretty_printer::pretty_printer): Drop the prefix parameter.
3229 Rename the length parameter to match the comment.
3230 (pretty_printer::~pretty_printer): Free the prefix.
3231 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
3233 (struct pretty_printer): Drop the "const" from "prefix" field's
3234 type and clarify memory management.
3235 (pp_set_prefix): Drop the "const" from the 2nd param.
3236 (pp_take_prefix): New decl.
3238 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
3240 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
3241 for word_mode_ok here instead of passing as argument.
3242 (expand_block_compare): Change select_block_compare_mode() call.
3243 (expand_strncmp_gpr_sequence): New function.
3244 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
3246 2018-08-02 Jeff Law <law@redhat.com>
3249 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3250 Define to speculation_safe_value_not_needed.
3253 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3254 Define to speculation_safe_value_not_needed.
3256 2018-08-02 Tom de Vries <tdevries@suse.de>
3259 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
3260 function. Return UI_TARGET unconditionally.
3261 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
3262 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
3264 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3266 * genemit.c (print_overload_test): Fix typo.
3268 2018-08-02 Richard Biener <rguenther@suse.de>
3270 PR tree-optimization/86816
3271 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
3272 which checks for value availability before querying it.
3273 (gvn_uses_equal): Use it.
3274 (same_succ_hash): Likewise.
3275 (gimple_equal_p): Likewise.
3277 2018-08-02 Nick Clifton <nickc@redhat.com>
3280 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3281 Define to speculation_safe_value_not_needed.
3284 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3285 Define to speculation_safe_value_not_needed.
3288 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3289 Define to speculation_safe_value_not_needed.
3292 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3293 Define to speculation_safe_value_not_needed.
3296 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3297 Define to speculation_safe_value_not_needed.
3300 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3301 Define to speculation_safe_value_not_needed.
3304 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3305 Define to speculation_safe_value_not_needed.
3308 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3309 Define to speculation_safe_value_not_needed.
3312 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
3313 speculation_safe_value_not_needed.
3316 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
3317 to speculation_safe_value_not_needed.
3319 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3321 * doc/md.texi: Expand the documentation of instruction names
3322 to mention port-local uses. Document '@' in pattern names.
3323 * read-md.h (overloaded_instance, overloaded_name): New structs.
3325 (md_reader::handle_overloaded_name): New member function.
3326 (md_reader::get_overloads): Likewise.
3327 (md_reader::m_first_overload): New member variable.
3328 (md_reader::m_next_overload_ptr): Likewise.
3329 (md_reader::m_overloads_htab): Likewise.
3330 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
3331 m_next_overload_ptr and m_overloads_htab.
3332 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
3333 (get_mode_token, get_code_token, get_int_token): New functions.
3334 (map_attr_string): Add an optional argument that passes back
3335 the associated iterator.
3336 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
3337 (md_reader::handle_overloaded_name, add_overload_instance): New
3339 (apply_iterators): Handle '@' names. Report an error if '@'
3340 is used without iterators.
3341 (initialize_iterators): Initialize the new iterator_group fields.
3342 * genopinit.c (handle_overloaded_code_for)
3343 (handle_overloaded_gen): New functions.
3344 (main): Use them to print declarations of maybe_code_for_* and
3345 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
3346 * genemit.c (print_overload_arguments, print_overload_test)
3347 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
3348 (main): Use it to print definitions of maybe_code_for_* and
3349 maybe_gen_* functions.
3350 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
3351 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
3352 instead of explicit mode checks.
3353 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
3354 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
3355 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
3356 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
3357 (aarch64_expand_compare_and_swap): Likewise
3358 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
3359 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
3360 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
3361 (aarch64_constant_pool_reload_icode): Delete.
3362 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
3363 instead of aarch64_constant_pool_reload_icode. Use
3364 code_for_aarch64_reload_mov instead of explicit mode checks.
3365 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
3366 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
3367 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
3368 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
3369 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
3370 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
3371 (aarch64_atomic_load_op_code): Delete.
3372 (aarch64_emit_atomic_load_op): Likewise.
3373 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
3374 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
3375 instead of aarch64_emit_atomic_load_op.
3376 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
3377 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
3378 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
3379 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
3380 character before the pattern name.
3381 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
3382 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
3383 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
3384 (aarch64_frecps<mode>): Likewise.
3385 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
3386 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
3387 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
3388 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
3389 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
3391 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3393 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
3394 Allow HFmode constants if TARGET_FP_F16INST.
3396 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
3399 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
3400 No longer check last store for clobber of address register.
3402 2018-08-02 Martin Liska <mliska@suse.cz>
3404 PR gcov-profile/86817
3405 * gcov.c (process_all_functions): New function.
3407 (process_file): Move functions processing to
3408 process_all_functions.
3410 2018-08-02 David Malcolm <dmalcolm@redhat.com>
3412 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
3413 "const" to the "gimple *" and "rtx_insn *" parameters.
3414 * dumpfile.h (dump_user_location_t::dump_user_location_t):
3416 (dump_location_t::dump_location_t): Likewise.
3418 2018-08-01 Martin Sebor <msebor@redhat.com>
3420 PR tree-optimization/86650
3421 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
3422 rather than a "gcall *". Directly pass the data of interest
3423 to percent_K_format, rather than building a temporary CALL_EXPR
3425 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
3426 (gimple_fold_builtin_strncat): Adjust.
3427 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
3428 gcall* argument with gimple*.
3429 * gimple-ssa-warn-restrict.c (check_call): Same.
3430 (wrestrict_dom_walker::before_dom_children): Same.
3431 (builtin_access::builtin_access): Same.
3432 (check_bounds_or_overlap): Same
3433 (maybe_diag_overlap): Same.
3434 (maybe_diag_offset_bounds): Same.
3435 * tree-diagnostic.c (default_tree_printer): Move usage of
3436 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
3438 * tree-pretty-print.c (percent_K_format): Add argument.
3439 * tree-pretty-print.h: Add argument.
3440 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
3441 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
3442 (maybe_diag_stxncpy_trunc): Same.
3443 (handle_builtin_stxncpy): Same.
3444 (handle_builtin_strcat): Same.
3446 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3448 * match.pd: Optimise pointer range checks.
3450 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3452 PR tree-optimization/86758
3453 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
3454 to remove pattern statements.
3456 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3458 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
3459 result of dfs_enumerate_from when constructing stmt_vec_infos,
3460 instead of additionally calling get_loop_body.
3462 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3464 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
3466 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
3467 When creating an iv, assert that the step is not known to be zero.
3468 (vect_setup_realignment): Update call accordingly.
3469 * tree-vect-stmts.c (vectorizable_store): Likewise.
3470 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
3472 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3474 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
3475 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
3476 (vectorizable_reduction): Likewise.
3477 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3478 (vect_detect_hybrid_slp_stmts): Likewise.
3479 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
3481 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
3483 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
3484 (wide_int_set_zero_nonzero_bits): ...this.
3485 (zero_nonzero_bits_from_vr): Rename to...
3486 (vrp_set_zero_nonzero_bits): ...this.
3487 (extract_range_from_multiplicative_op_1): Abstract wide int
3489 (wide_int_range_multiplicative_op): ...here.
3490 (extract_range_from_binary_expr_1): Extract wide int binary
3491 operations into their own functions.
3492 (wide_int_range_lshift): New.
3493 (wide_int_range_can_optimize_bit_op): New.
3494 (wide_int_range_shift_undefined_p): New.
3495 (wide_int_range_bit_xor): New.
3496 (wide_int_range_bit_ior): New.
3497 (wide_int_range_bit_and): New.
3498 (wide_int_range_trunc_mod): New.
3499 (extract_range_into_wide_ints): New.
3500 (vrp_shift_undefined_p): New.
3501 (extract_range_from_multiplicative_op): New.
3502 (vrp_can_optimize_bit_op): New.
3503 * tree-vrp.h (value_range::dump): New.
3504 (wide_int_range_multiplicative_op): New.
3505 (wide_int_range_lshift):New.
3506 (wide_int_range_shift_undefined_p): New.
3507 (wide_int_range_bit_xor): New.
3508 (wide_int_range_bit_ior): New.
3509 (wide_int_range_bit_and): New.
3510 (wide_int_range_trunc_mod): New.
3511 (zero_nonzero_bits_from_bounds): Rename to...
3512 (wide_int_set_zero_nonzero_bits): ...this.
3513 (zero_nonzero_bits_from_vr): Rename to...
3514 (vrp_set_zero_nonzero_bits): ...this.
3515 (range_easy_mask_min_max): Rename to...
3516 (wide_int_range_can_optimize_bit_op): this.
3517 * vr-values.c (simplify_bit_ops_using_ranges): Rename
3518 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
3520 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3522 * tree-vectorizer.h (vect_orig_stmt): New function.
3523 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
3524 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
3525 (vect_create_epilog_for_reduction): Likewise.
3526 (vectorizable_live_operation): Likewise.
3527 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
3528 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
3529 * tree-vect-stmts.c (vectorizable_call): Likewise.
3530 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
3532 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3534 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
3536 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
3537 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
3538 (vect_transform_loop): Likewise.
3539 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
3541 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3543 * tree-vectorizer.h (vect_schedule_slp): Return void.
3544 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
3545 (vect_schedule_slp): Likewise.
3547 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3549 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
3551 (vect_transform_loop): Update calls accordingly. Schedule SLP
3552 instances before the main loop, if any exist.
3554 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3556 PR tree-optimization/86749
3557 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
3558 If the lhs is used in a COND_EXPR, check that it is being used
3559 as the "then" or "else" value.
3561 2018-08-01 Tom de Vries <tdevries@suse.de>
3564 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
3565 speculation_safe_value_not_needed.
3567 2018-08-01 Richard Biener <rguenther@suse.de>
3569 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
3572 2018-08-01 Martin Liska <mliska@suse.cz>
3574 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
3575 format how successful transformation is dumped.
3576 (gimple_mod_pow2_value_transform): Likewise.
3577 (gimple_mod_subtract_transform): Likewise.
3578 (gimple_stringops_transform): Likewise.
3580 2018-08-01 Martin Liska <mliska@suse.cz>
3583 * value-prof.c (interesting_stringop_to_profile_p):
3584 Simplify the code and add BUILT_IN_MEMMOVE.
3585 (gimple_stringops_transform): Likewise.
3587 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
3589 * config/aarch64/aarch64-simd.md
3590 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
3591 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
3592 use GPI iterator instead of SI mode.
3594 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
3596 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
3597 rs6000_speculation_barrier.
3598 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
3599 new barrier pattern name.
3601 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
3603 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
3604 (speculation_barrier): New insn.
3606 2018-08-01 Richard Biener <rguenther@suse.de>
3609 * graphite.h: Include isl/id.h and isl/space.h to allow build
3612 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
3615 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3616 mode for COFF targets.
3617 * defaults.h (TARGET_COFF): Define.
3618 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
3619 TARGET_COFF): Define.
3620 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
3621 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
3622 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
3624 2018-07-31 Alexandre Oliva <oliva@adacore.com>
3626 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
3627 * gimple-streamer-out.c (output_bb): Save it.
3628 * lto-streamer-in.c (input_struct_function_base): Restore
3629 instance discriminator if available. Create map on demand.
3630 * lto-streamer-out.c (output_struct_function_base): Save it if
3632 * final.c (decl_to_instance_map): Document LTO strategy.
3634 2018-07-31 Alexandre Oliva <oliva@adacore.com>
3635 Olivier Hainque <hainque@adacore.com>
3637 * debug.h (decl_to_instance_map_t): New type.
3638 (decl_to_instance_map): Declare.
3639 (maybe_create_decl_to_instance_map): New inline function.
3640 * final.c (bb_discriminator, last_bb_discriminator): New statics,
3641 to track basic block discriminators.
3642 (final_start_function_1): Initialize them.
3643 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
3645 (decl_to_instance_map): New variable.
3646 (map_decl_to_instance, maybe_set_discriminator): New functions.
3647 (notice_source_line): Set discriminator.
3649 2018-07-31 Ian Lance Taylor <iant@golang.org>
3651 * targhooks.c (default_have_speculation_safe_value): Add
3654 2018-07-31 David Malcolm <dmalcolm@redhat.com>
3656 * dump-context.h: Include "pretty-print.h".
3657 (dump_context::refresh_dumps_are_enabled): New decl.
3658 (dump_context::emit_item): New decl.
3659 (class dump_context): Add fields "m_test_pp" and
3661 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
3662 (temp_dump_context::get_dumped_text): New decl.
3663 (class temp_dump_context): Add field "m_pp".
3664 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
3665 (dump_context::refresh_dumps_are_enabled): ...and add a test for
3667 (set_dump_file): Update for above change.
3668 (set_alt_dump_file): Likewise.
3669 (dump_loc): New overload, taking a pretty_printer *.
3670 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
3671 to any test pretty-printer.
3672 (make_item_for_dump_gimple_stmt): New function, adapted from
3673 optinfo::add_gimple_stmt.
3674 (dump_context::dump_gimple_stmt): Call it, and use the result,
3675 eliminating the direct usage of dump_file and alt_dump_file in
3676 favor of indirectly using them via emit_item.
3677 (make_item_for_dump_gimple_expr): New function, adapted from
3678 optinfo::add_gimple_expr.
3679 (dump_context::dump_gimple_expr): Call it, and use the result,
3680 eliminating the direct usage of dump_file and alt_dump_file in
3681 favor of indirectly using them via emit_item.
3682 (make_item_for_dump_generic_expr): New function, adapted from
3684 (dump_context::dump_generic_expr): Call it, and use the result,
3685 eliminating the direct usage of dump_file and alt_dump_file in
3686 favor of indirectly using them via emit_item.
3687 (make_item_for_dump_printf_va): New function, adapted from
3688 optinfo::add_printf_va.
3689 (make_item_for_dump_printf): New function.
3690 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
3691 and use the result, eliminating the direct usage of dump_file and
3692 alt_dump_file in favor of indirectly using them via emit_item.
3693 (make_item_for_dump_dec): New function.
3694 (dump_context::dump_dec): Call it, and use the result,
3695 eliminating the direct usage of dump_file and alt_dump_file in
3696 favor of indirectly using them via emit_item.
3697 (make_item_for_dump_symtab_node): New function, adapted from
3698 optinfo::add_symtab_node.
3699 (dump_context::dump_symtab_node): Call it, and use the result,
3700 eliminating the direct usage of dump_file and alt_dump_file in
3701 favor of indirectly using them via emit_item.
3702 (dump_context::begin_scope): Reimplement, avoiding direct usage
3703 of dump_file and alt_dump_file in favor of indirectly using them
3705 (dump_context::emit_item): New member function.
3706 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
3707 Set up test pretty-printer on the underlying context. Call
3708 refresh_dumps_are_enabled.
3709 (temp_dump_context::~temp_dump_context): Call
3710 refresh_dumps_are_enabled.
3711 (temp_dump_context::get_dumped_text): New member function.
3712 (selftest::verify_dumped_text): New function.
3713 (ASSERT_DUMPED_TEXT_EQ): New macro.
3714 (selftest::test_capture_of_dump_calls): Run all tests twice, with
3715 and then without optinfo enabled. Add uses of
3716 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
3717 * dumpfile.h: Update comment for the dump_* API.
3718 * optinfo-emit-json.cc
3719 (selftest::test_building_json_from_dump_calls): Update for new
3720 param for temp_dump_context ctor.
3721 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
3722 and "m_owned" field.
3723 (optinfo_item::~optinfo_item): Likewise.
3724 (optinfo::add_item): New member function.
3725 (optinfo::emit): Update comment.
3726 (optinfo::add_string): Delete.
3727 (optinfo::add_printf): Delete.
3728 (optinfo::add_printf_va): Delete.
3729 (optinfo::add_gimple_stmt): Delete.
3730 (optinfo::add_gimple_expr): Delete.
3731 (optinfo::add_tree): Delete.
3732 (optinfo::add_symtab_node): Delete.
3733 (optinfo::add_dec): Delete.
3734 * optinfo.h (class dump_context): New forward decl.
3735 (optinfo::add_item): New decl.
3736 (optinfo::add_string): Delete.
3737 (optinfo::add_printf): Delete.
3738 (optinfo::add_printf_va): Delete.
3739 (optinfo::add_gimple_stmt): Delete.
3740 (optinfo::add_gimple_expr): Delete.
3741 (optinfo::add_tree): Delete.
3742 (optinfo::add_symtab_node): Delete.
3743 (optinfo::add_dec): Delete.
3744 (optinfo::add_poly_int): Delete.
3745 (optinfo_item::optinfo_item): Remove "owned" param.
3746 (class optinfo_item): Remove field "m_owned".
3748 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3751 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
3752 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
3753 requested variable alignment.
3754 (expand_one_ssa_partition): Likewise.
3755 (expand_one_var): Likewise.
3757 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3759 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
3760 to speculation_safe_value_not_needed.
3762 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3764 * targhooks.h (speculation_safe_value_not_needed): New prototype.
3765 * targhooks.c (speculation_safe_value_not_needed): New function.
3766 * target.def (have_speculation_safe_value): Update documentation.
3767 * doc/tm.texi: Regenerated.
3769 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3771 * config/aarch64/iterators.md (ALLI_TI): New iterator.
3772 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
3774 (despeculate_copy<ALLI:mode>_insn): New insn.
3775 (despeculate_copyti_insn): New insn.
3776 (despeculate_simple<ALLI:mode>): New insn
3777 (despeculate_simpleti): New insn.
3778 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
3780 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
3781 aarch64_speculation_safe_value.
3782 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
3784 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3786 * config/aarch64/aarch64-speculation.cc: New file.
3787 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
3788 before pass_reorder_blocks.
3789 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
3791 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
3792 X14 and X15 when tracking speculation.
3793 * config/aarch64/aarch64.md (register name constants): Add
3794 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
3795 (unspec): Add UNSPEC_SPECULATION_TRACKER.
3796 (speculation_barrier): New insn attribute.
3797 (cmp<mode>): Allow SP in comparisons.
3798 (speculation_tracker): New insn.
3799 (speculation_barrier): Add speculation_barrier attribute.
3800 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
3801 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
3802 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
3804 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3806 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
3807 aarch64_track_speculation is true.
3808 (tb<optab><mode>1): Likewise.
3809 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
3810 generate CB[N]Z when tracking speculation.
3811 (aarch64_split_compare_and_swap): Likewise.
3812 (aarch64_split_atomic_op): Likewise.
3814 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3816 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
3818 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3820 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
3821 (speculation_barrier): New insn.
3823 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3825 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
3826 * config/arm/arm.md (speculation_barrier): New expand.
3827 (speculation_barrier_insn): New pattern.
3829 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3831 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
3832 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
3833 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
3834 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
3836 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
3837 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
3838 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
3839 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
3840 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
3841 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
3842 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
3843 * builtins.c (expand_speculation_safe_value): New function.
3844 (expand_builtin): Call it.
3845 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
3846 * doc/extend.texi: Document __builtin_speculation_safe_value.
3847 * doc/md.texi: Document "speculation_barrier" pattern.
3848 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
3849 TARGET_HAVE_SPECULATION_SAFE_VALUE.
3850 * doc/tm.texi: Regenerated.
3851 * target.def (have_speculation_safe_value, speculation_safe_value): New
3853 * targhooks.c (default_have_speculation_safe_value): New function.
3854 (default_speculation_safe_value): New function.
3855 * targhooks.h (default_have_speculation_safe_value): Add prototype.
3856 (default_speculation_safe_value): Add prototype.
3858 2018-07-31 David Malcolm <dmalcolm@redhat.com>
3860 * dump-context.h (dump_context::dump_loc): New decl.
3861 * dumpfile.c (dump_context::dump_loc): New member function.
3862 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
3863 and dump_gimple_stmt.
3864 (dump_context::dump_gimple_expr_loc): Likewise, using
3866 (dump_context::dump_generic_expr_loc): Likewise, using
3868 (dump_context::dump_printf_loc_va): Likewise, using
3870 (dump_context::begin_scope): Explicitly using the global function
3871 "dump_loc", rather than the member function.
3873 2018-07-31 Martin Sebor <msebor@redhat.com>
3875 PR tree-optimization/86741
3876 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
3878 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
3880 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
3881 depend on whether prefetch instructions will be emitted or not.
3882 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
3883 will be emitted or not.
3884 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
3885 (TARGET_SETMEM_PFD): New macros.
3887 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3889 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
3890 (NULL_STMT_VEC_INFO): Delete.
3891 (stmt_vec_info::operator*): Likewise.
3892 (stmt_vec_info::operator gimple *): Likewise.
3893 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
3894 of NULL_STMT_VEC_INFO.
3895 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
3896 (vect_reassociating_reduction_p): Likewise.
3897 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
3898 (vectorizable_store): Likewise.
3899 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
3900 (vec_info::free_stmt_vec_infos): Likewise.
3902 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3904 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
3905 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
3907 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3909 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
3910 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
3911 (vec_info::free_stmt_vec_info): New private member functions.
3912 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
3913 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
3914 * tree-parloops.c (gather_scalar_reductions): Remove calls to
3915 set_stmt_vec_info_vec and free_stmt_vec_infos.
3916 * tree-vect-loop.c (_loop_vec_info): Remove call to
3917 set_stmt_vec_info_vec.
3918 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
3919 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
3920 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
3921 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
3922 (vec_info::free_stmt_vec_info): ...these new functions. Remove
3923 assignments in {vec_info::,}new_stmt_vec_info that are redundant
3924 with the clearing in the xcalloc.
3925 (stmt_vec_info_vec): Delete.
3926 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
3927 (vectorize_loops): Likewise.
3928 (vec_info::~vec_info): Remove argument from call to
3929 free_stmt_vec_infos.
3930 (vec_info::add_stmt): Remove vinfo argument from call to
3933 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3935 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
3936 rather than a gimple stmt.
3937 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
3938 information for pattern statements when passed the original
3939 statement; instead wait to be passed the pattern statement itself.
3940 Don't call set_vinfo_for_stmt here.
3941 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
3942 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
3943 stmt_vec_infos here.
3944 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
3945 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
3946 stmt_vec_infos entry.
3948 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3950 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
3951 * tree-vectorizer.c (vec_info::replace_stmt): New function.
3952 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
3953 * tree-vect-stmts.c (vectorizable_call): Likewise.
3954 (vectorizable_simd_clone_call): Likewise.
3956 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3958 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
3959 * tree-vectorizer.c (vec_info::remove_stmt): New function.
3960 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
3961 * tree-vect-loop.c (vect_transform_loop): Likewise.
3962 * tree-vect-slp.c (vect_schedule_slp): Likewise.
3963 * tree-vect-stmts.c (vect_remove_stores): Likewise.
3965 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3967 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
3968 (vect_dr_stmt): Delete.
3969 * tree-vectorizer.c (vec_info::lookup_dr): New function.
3970 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
3972 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
3973 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
3974 (vect_verify_datarefs_alignment, vect_peeling_supportable)
3975 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
3976 (vect_analyze_data_refs): Likewise.
3977 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
3979 (vect_find_same_alignment_drs): Likewise.
3980 (vect_slp_analyze_node_dependences): Update calls accordingly.
3981 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
3982 instead of DR_VECT_AUX.
3983 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
3984 of a vector data references. Use vec_info::lookup_dr instead of
3986 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
3987 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
3988 instead of DR_VECT_AUX.
3990 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3992 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
3994 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
3996 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3997 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3998 (vect_gen_prolog_loop_niters): Likewise.
4000 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4002 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
4003 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
4004 (vect_known_alignment_in_bytes, vect_dr_behavior)
4005 (vect_get_scalar_dr_size): Take references as dr_vec_infos
4006 instead of data_references. Update calls to other routines for
4007 which the same change has been made.
4008 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
4009 dr_vec_infos instead of stmt_vec_infos.
4010 (vect_analyze_data_ref_dependence): Update call accordingly.
4011 (vect_slp_analyze_data_ref_dependence)
4012 (vect_record_base_alignments): Use DR_VECT_AUX.
4013 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
4014 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
4015 (vector_alignment_reachable_p, vect_get_data_access_cost)
4016 (vect_peeling_supportable, vect_analyze_group_access_1)
4017 (vect_analyze_group_access, vect_analyze_data_ref_access)
4018 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
4019 (vect_compile_time_alias, vect_small_gap_p)
4020 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
4021 (vect_supportable_dr_alignment): Take references as dr_vec_infos
4022 instead of data_references. Update calls to other routines for
4023 which the same change has been made.
4024 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
4025 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
4026 (vect_slp_analyze_and_verify_node_alignment)
4027 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
4028 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4029 (vect_setup_realignment): Use dr_vec_infos. Update calls after
4031 (_vect_peel_info::dr): Replace with...
4032 (_vect_peel_info::dr_info): ...this new field.
4033 (vect_peeling_hash_get_most_frequent)
4034 (vect_peeling_hash_choose_best_peeling): Update accordingly.
4035 (vect_peeling_hash_get_lowest_cost):
4036 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
4037 routines for which the same change has been made.
4038 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
4040 * tree-vect-loop-manip.c (get_misalign_in_elems)
4041 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
4043 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4044 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
4045 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
4046 (get_group_load_store_type, get_negative_load_store_type)
4047 (vect_get_data_ptr_increment, vectorizable_store)
4048 (vectorizable_load): Likewise.
4049 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
4050 Update calls to other routines for which the same change has been made.
4052 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4054 * tree-vectorizer.h (vec_info::move_dr): New member function.
4055 (dataref_aux): Rename to...
4056 (dr_vec_info): ...this and add "dr" and "stmt" fields.
4057 (_stmt_vec_info::dr_aux): Update accordingly.
4058 (_stmt_vec_info::data_ref_info): Delete.
4059 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
4060 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
4061 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
4062 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
4064 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
4065 (STMT_VINFO_DR_INFO): New macro.
4066 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
4067 (set_dr_misalignment): Update after rename of dataref_aux.
4068 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
4069 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
4070 initialization of STMT_VINFO_DATA_REF.
4071 * tree-vectorizer.c (vec_info::move_dr): New function.
4072 * tree-vect-patterns.c (vect_recog_bool_pattern)
4073 (vect_recog_mask_conversion_pattern)
4074 (vect_recog_gather_scatter_pattern): Use it.
4075 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
4076 the "dr" and "stmt" fields of dr_vec_info instead of
4077 STMT_VINFO_DATA_REF.
4079 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4081 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
4082 (is_pattern_stmt_p): Use it.
4083 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
4084 on pattern statements.
4086 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4088 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
4089 original stmt as a stmt_vec_info rather than a gimple stmt.
4090 (vect_pattern_recog_1): Take the statement directly as a
4091 stmt_vec_info, rather than via a gimple_stmt_iterator.
4092 Update call to vect_mark_pattern_stmts.
4093 (vect_pattern_recog): Update calls accordingly.
4095 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4097 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
4098 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
4099 a vect_def_type for the first argument.
4100 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
4101 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
4102 operand if it isn't defined by a vectorized statement.
4103 (vect_build_gather_load_calls): Remove the mask_dt argument and
4104 update calls to vect_get_vec_def_for_stmt_copy.
4105 (vectorizable_bswap): Likewise the dt argument.
4106 (vectorizable_call): Update calls to vectorizable_bswap and
4107 vect_get_vec_def_for_stmt_copy.
4108 (vectorizable_simd_clone_call, vectorizable_assignment)
4109 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
4110 (vectorizable_comparison): Update calls to
4111 vect_get_vec_def_for_stmt_copy.
4112 (vectorizable_store): Likewise. Remove now-unnecessary calls to
4114 (vect_get_loop_based_defs): Remove dt argument and update call
4115 to vect_get_vec_def_for_stmt_copy.
4116 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
4117 and vect_get_vec_def_for_stmt_copy.
4118 (vectorizable_load): Update calls to vect_build_gather_load_calls
4119 and vect_get_vec_def_for_stmt_copy.
4120 * tree-vect-loop.c (vect_create_epilog_for_reduction)
4121 (vectorizable_reduction, vectorizable_live_operation): Update calls
4122 to vect_get_vec_def_for_stmt_copy.
4124 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4126 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
4127 and gimple stmt arguments with a stmt_vec_info.
4128 (vect_record_base_alignments): Update calls accordingly.
4129 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
4130 and gimple stmt arguments with a stmt_vec_info.
4131 (vect_build_slp_tree_1): Remove vinfo argument and update call
4132 to vect_record_max_nunits.
4133 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
4134 and vect_record_max_nunits.
4136 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4138 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
4139 file and take a stmt_vec_info instead of a gimple stmt.
4140 (supportable_widening_operation, vect_finish_replace_stmt)
4141 (vect_finish_stmt_generation, vect_get_store_rhs)
4142 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
4143 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
4144 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
4145 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
4146 (vect_create_data_ref_ptr, bump_vector_ptr)
4147 (vect_permute_store_chain, vect_setup_realignment)
4148 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
4149 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
4150 (vectorizable_reduction, vectorizable_induction)
4151 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
4152 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
4153 than gimple stmts as arguments.
4154 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
4155 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
4156 (can_group_stmts_p, vect_check_gather_scatter)
4157 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4158 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
4159 (vect_permute_load_chain, vect_shift_permute_load_chain)
4160 (vect_transform_grouped_load)
4161 (vect_record_grouped_load_vectors): Likewise.
4162 * tree-vect-loop.c (vect_fixup_reduc_chain)
4163 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
4164 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
4165 (vectorizable_reduction, vectorizable_induction)
4166 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
4167 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
4168 (vect_get_load_store_mask): Likewise.
4169 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
4170 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
4171 * tree-vect-stmts.c (vect_mark_relevant)
4172 (is_simple_and_all_uses_invariant)
4173 (exist_non_indexing_operands_for_use_p, process_use)
4174 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
4175 (vect_get_vec_def_for_operand, vect_get_vec_defs)
4176 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
4177 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
4178 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
4179 (get_negative_load_store_type, get_load_store_type)
4180 (vect_check_load_store_mask, vect_check_store_rhs)
4181 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
4182 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4183 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4184 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4185 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
4186 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
4187 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
4188 (supportable_widening_operation): Likewise.
4190 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4192 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
4193 a stmt_vec_info instead of a gcall.
4194 (vect_check_gather_scatter): Update call accordingly.
4195 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
4197 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
4198 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
4199 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
4200 instead of a gimple stmt.
4201 (vect_transform_loop): Update calls accordingly.
4202 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
4203 stmt_vec_infos instead of gimple stmts.
4204 (vect_analyze_slp_instance): Update use accordingly.
4205 * tree-vect-stmts.c (read_vector_array, write_vector_array)
4206 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
4207 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
4208 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
4209 (vect_gen_widened_results_half, vect_get_loop_based_defs)
4210 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
4211 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
4212 down to subroutines.
4214 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4216 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
4217 of the worklist from a vector of gimple stmts to a vector of
4219 * tree-vect-stmts.c (vect_mark_relevant, process_use)
4220 (vect_mark_stmts_to_be_vectorized): Likewise
4222 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4224 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
4225 statement before passing it to vect_analyze_stmt.
4226 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
4227 the chain of phi vector definitions. Track the exit phi via its
4229 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
4230 STMT_VINFO_REDUC_DEF.
4231 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
4232 stmt_vec_infos to handle the statement chains.
4233 (vect_get_slp_defs): Record the first statement in the node
4234 using a stmt_vec_info.
4235 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
4236 statements here and pass their stmt_vec_info down to subroutines.
4237 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
4238 down to vect_finish_stmt_generation.
4239 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
4240 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
4241 stmt_vec_infos to subroutines.
4242 (vect_remove_stores): Use stmt_vec_infos to handle the statement
4245 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4247 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
4248 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
4249 (vect_permute_store_chain, vect_setup_realignment)
4250 (vect_permute_load_chain, vect_shift_permute_load_chain)
4251 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
4252 stmts internally, and when passing values to other vectorizer routines.
4253 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4254 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
4255 (vect_analyze_loop_operations, get_initial_def_for_reduction)
4256 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
4257 (vectorizable_reduction, vectorizable_induction)
4258 (vectorizable_live_operation, vect_transform_loop_stmt)
4259 (vect_transform_loop): Likewise.
4260 * tree-vect-patterns.c (vect_reassociating_reduction_p)
4261 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
4262 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
4263 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4264 (vect_slp_analyze_node_operations_1): Likewise.
4265 * tree-vect-stmts.c (vect_mark_relevant, process_use)
4266 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
4267 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
4268 (vect_finish_stmt_generation_1, get_group_load_store_type)
4269 (get_load_store_type, vect_build_gather_load_calls)
4270 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4271 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4272 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4273 (vectorizable_store, vectorizable_load, vectorizable_condition)
4274 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
4275 (supportable_widening_operation): Likewise.
4276 (vect_get_vector_types_for_stmt): Likewise.
4277 * tree-vectorizer.h (vect_dr_behavior): Likewise.
4279 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4281 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4282 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
4283 (vect_permute_store_chain, vect_permute_load_chain)
4284 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
4285 repeated stmt_vec_info lookups.
4286 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4287 (vect_update_ivs_after_vectorizer): Likewise.
4288 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
4289 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
4290 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
4291 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4292 (vect_bb_slp_scalar_cost): Likewise.
4293 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
4295 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4297 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
4298 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
4299 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
4300 to get gassigns and gcalls, rather than operating on generc gimple
4302 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
4303 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
4304 (vectorizable_load, vect_analyze_stmt): Likewise.
4305 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
4307 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4309 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
4310 return stmt_vec_infos rather than gimple stmts. Do not accept
4312 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
4314 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
4315 Update use of get_later_stmt.
4316 (vect_get_constant_vectors): Update call accordingly.
4317 (vect_schedule_slp_instance): Likewise
4318 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
4319 (vect_slp_analyze_instance_dependence): Likewise.
4320 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
4322 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4324 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
4325 (stmt_info_for_cost::stmt_info): ...this new field.
4326 (add_stmt_costs): Update accordingly.
4327 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
4328 (vect_get_known_peeling_cost): Likewise.
4329 (vect_estimate_min_profitable_iters): Likewise.
4330 * tree-vect-stmts.c (record_stmt_cost): Likewise.
4332 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4334 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
4335 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4336 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
4338 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
4340 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4342 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
4343 a gimple stmt to a stmt_vec_info.
4344 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
4346 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4348 * tree-vectorizer.h (vec_info::grouped_stores): Change from
4349 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4350 (_loop_vec_info::reduction_chains): Likewise.
4351 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
4353 * tree-vect-slp.c (vect_analyze_slp): Likewise.
4355 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4357 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
4358 a gimple stmt to a stmt_vec_info.
4359 (_stmt_vec_info::next_element): Likewise.
4360 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
4361 (vect_slp_analyze_and_verify_node_alignment)
4362 (vect_analyze_group_access_1, vect_analyze_group_access)
4363 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
4364 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
4365 (vect_supportable_dr_alignment): Update accordingly.
4366 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
4367 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
4368 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
4369 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4370 * tree-vect-slp.c (vect_build_slp_tree_1)
4371 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
4372 (vect_split_slp_store_group, vect_analyze_slp_instance)
4373 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
4374 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
4375 (get_group_load_store_type, get_load_store_type)
4376 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
4377 (vect_transform_stmt, vect_remove_stores): Likewise.
4379 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4381 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
4383 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4384 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
4385 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
4386 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
4387 (vector_alignment_reachable_p, vect_get_data_access_cost)
4388 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
4389 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
4390 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
4391 (vect_analyze_group_access_1, vect_analyze_group_access)
4392 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
4393 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
4394 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
4395 result of vect_dr_stmt and use the stmt_vec_info instead of
4396 the associated gimple stmt.
4397 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4398 (vect_gen_prolog_loop_niters): Likewise.
4399 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4401 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4403 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
4404 to a vec<stmt_vec_info>.
4405 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
4406 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
4408 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
4409 to a vec<stmt_vec_info>.
4410 (bst_traits::value_type, bst_traits::value_type): Likewise.
4411 (bst_traits::hash): Update accordingly.
4412 (vect_get_and_check_slp_defs): Change the stmts parameter from
4413 a vec<gimple *> to a vec<stmt_vec_info>.
4414 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
4415 (vect_build_slp_tree): Likewise.
4416 (vect_build_slp_tree_2): Likewise. Update uses of
4417 SLP_TREE_SCALAR_STMTS.
4418 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
4419 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
4420 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
4421 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
4422 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
4423 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
4424 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
4425 (vect_get_constant_vectors, vect_get_slp_defs)
4426 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
4427 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
4428 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
4429 instead of gimple stmts.
4430 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
4431 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
4432 (vect_slp_analyze_instance_dependence): Update uses of
4433 SLP_TREE_SCALAR_STMTS.
4434 (vect_slp_analyze_and_verify_node_alignment): Likewise.
4435 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
4436 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
4437 (get_initial_defs_for_reduction): Likewise.
4438 (vect_create_epilog_for_reduction): Likewise.
4439 (vectorize_fold_left_reduction): Likewise.
4440 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
4441 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
4442 (can_vectorize_live_stmts): Likewise.
4444 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4446 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
4447 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4448 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
4450 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
4452 (gather_scalar_reductions): Update after above interface changes.
4453 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
4454 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
4456 (vect_force_simple_reduction): Likewise.
4457 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
4458 LOOP_VINFO_REDUCTIONS.
4459 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4461 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4463 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
4464 a gimple stmt to a stmt_vec_info.
4465 * tree-vect-loop.c (vect_active_double_reduction_p)
4466 (vect_force_simple_reduction, vectorizable_reduction): Update
4469 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4471 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
4472 vec<gimple *> to a vec<stmt_vec_info>.
4473 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
4474 the reduction_phis argument from a vec<gimple *> to a
4476 (vectorizable_reduction): Likewise the phis local variable that
4477 is passed to vect_create_epilog_for_reduction. Update for new type
4478 of SLP_TREE_VEC_STMTS.
4479 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
4480 (vectorizable_live_operation): Likewise.
4481 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
4482 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
4484 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4486 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
4487 a gimple stmt to a stmt_vec_info.
4488 (vectorizable_condition, vectorizable_live_operation)
4489 (vectorizable_reduction, vectorizable_induction): Pass back the
4490 vectorized statement as a stmt_vec_info.
4491 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
4492 use of STMT_VINFO_VEC_STMT.
4493 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
4494 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
4495 as stmt_vec_infos rather than gimple stmts.
4496 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
4498 (vectorizable_live_operation): Likewise.
4499 (vectorizable_reduction, vectorizable_induction): Likewise,
4500 updating use of STMT_VINFO_VEC_STMT.
4501 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
4502 of STMT_VINFO_VEC_STMT.
4503 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
4504 (vectorizable_simd_clone_call, vectorizable_conversion)
4505 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4506 (vectorizable_store, vectorizable_load, vectorizable_condition)
4507 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
4508 from a gimple stmt to a stmt_vec_info.
4509 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
4510 pointer to a stmt_vec_info to the vectorizable_* routines.
4512 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4514 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
4515 a gimple stmt to a stmt_vec_info.
4516 (is_pattern_stmt_p): Update accordingly.
4517 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
4518 (vect_record_grouped_load_vectors): Likewise.
4519 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
4520 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
4521 (vect_model_reduction_cost): Likewise.
4522 (vect_create_epilog_for_reduction): Likewise.
4523 (vectorizable_reduction, vectorizable_induction): Likewise.
4524 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4525 Return the stmt_vec_info for the pattern statement.
4526 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
4527 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
4528 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
4529 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
4530 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
4531 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
4532 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
4533 (free_stmt_vec_info, vect_is_simple_use): Likewise.
4535 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4537 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
4538 (vect_finish_stmt_generation): Likewise.
4539 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
4540 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
4541 (vect_build_gather_load_calls): Use the return value of the above
4542 functions instead of a separate call to vinfo_for_stmt. Use narrow
4543 scopes for the input gimple stmt and wider scopes for the associated
4544 stmt_vec_info. Use vec_info::lookup_def when setting these
4545 stmt_vec_infos from an SSA_NAME definition.
4546 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4547 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4548 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4549 (vectorizable_store, vectorizable_load, vectorizable_condition)
4550 (vectorizable_comparison): Likewise.
4551 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
4552 (vectorizable_reduction): Likewise.
4554 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4556 * tree-vectorizer.h (vect_is_simple_use): Add an optional
4557 stmt_vec_info * parameter before the optional gimple **.
4558 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
4559 (process_use, vect_get_vec_def_for_operand_1): Update callers.
4560 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
4561 * tree-vect-loop.c (vectorizable_reduction): Likewise.
4562 (vectorizable_live_operation): Likewise.
4563 * tree-vect-patterns.c (type_conversion_p): Likewise.
4564 (vect_look_through_possible_promotion): Likewise.
4565 (vect_recog_rotate_pattern): Likewise.
4566 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
4568 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4570 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
4571 a typedef to a wrapper class.
4572 (NULL_STMT_VEC_INFO): New macro.
4573 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
4574 (stmt_vec_info::operator*): New function.
4575 (stmt_vec_info::operator gimple *): Likewise.
4576 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
4577 (add_stmt_costs): Likewise.
4578 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
4579 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
4580 (vect_get_known_peeling_cost): Likewise.
4581 (vect_estimate_min_profitable_iters): Likewise.
4582 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4583 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
4584 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
4585 (vectorizable_store, free_stmt_vec_infos): Likewise.
4586 (new_stmt_vec_info): Change return type of xcalloc to
4589 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4591 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
4592 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
4593 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
4594 a single_imm_use-based sequence.
4595 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
4597 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4599 * tree-vectorizer.h (vec_info::lookup_def): Declare.
4600 * tree-vectorizer.c (vec_info::lookup_def): New function.
4601 * tree-vect-patterns.c (vect_get_internal_def): Use it.
4602 (vect_widened_op_tree): Likewise.
4603 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
4604 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
4605 (vectorizable_reduction): Likewise.
4606 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
4608 (vect_is_slp_reduction): Update calls accordingly. Use
4609 vec_info::lookup_def.
4610 (vect_is_simple_reduction): Likewise
4611 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
4613 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4615 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
4616 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
4617 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
4619 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
4620 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
4621 (vect_update_vf_for_slp, vect_analyze_loop_operations)
4622 (vect_is_slp_reduction, vectorizable_induction)
4623 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
4624 * tree-vect-patterns.c (vect_init_pattern_stmt):
4625 (vect_determine_min_output_precision_1, vect_determine_precisions)
4626 (vect_pattern_recog): Likewise.
4627 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
4628 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
4629 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
4630 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
4631 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
4632 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
4633 info field from a loop to a loop_vec_info.
4635 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4637 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
4638 (vec_info::add_stmt): Declare.
4639 * tree-vectorizer.c (vec_info::add_stmt): New function.
4640 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
4641 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
4642 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
4643 (vectorizable_induction): Likewise.
4644 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
4645 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
4646 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
4647 (vectorizable_load): Likewise.
4648 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4649 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
4650 (vect_recog_gather_scatter_pattern): Likewise.
4651 (append_pattern_def_seq): Likewise. Remove a check that is
4652 performed by add_stmt itself.
4654 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4656 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
4657 which make_ssa_name was called with new_stmt before new_stmt
4660 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4662 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
4664 (vect_is_slp_reduction): ...here...
4665 (vect_is_simple_reduction): ...and here. Remove repetition of tests
4666 that are already known to be false.
4668 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4670 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
4671 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
4672 STMT_VINFO_NUM_SLP_USES when it's true.
4673 (vect_free_slp_instance): Add a final_p parameter and pass it to
4675 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
4676 (vect_analyze_slp_instance): Likewise.
4677 (vect_slp_analyze_operations): Likewise.
4678 (vect_slp_analyze_bb_1): Likewise.
4679 * tree-vectorizer.c (vec_info): Likewise.
4680 * tree-vect-loop.c (vect_transform_loop): Likewise.
4682 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4684 * tree-vect-loop.c (vectorizable_reduction): Assert that the
4685 function is not called for second and subsequent members of
4688 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4690 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
4691 cases for nested loops from here to ...
4692 (vect_create_epilog_for_reduction): ...here. Only call
4693 vect_is_simple_use for inner-loop reductions.
4695 2018-07-31 Martin Liska <mliska@suse.cz>
4697 PR gcov-profile/85338
4698 PR gcov-profile/85350
4699 PR gcov-profile/85372
4700 * profile.c (struct location_triplet): New.
4701 (struct location_triplet_hash): Likewise.
4702 (output_location): Do not output a BB that
4703 is already recorded for a line.
4704 (branch_prob): Use streamed_locations.
4706 2018-07-31 Martin Liska <mliska@suse.cz>
4708 PR gcov-profile/85370
4709 * coverage.c (coverage_begin_function): Do not mark target
4710 clones as artificial functions.
4712 2018-07-31 Martin Liska <mliska@suse.cz>
4714 PR gcov-profile/83813
4715 PR gcov-profile/84758
4716 PR gcov-profile/85217
4717 PR gcov-profile/85332
4718 * profile.c (branch_prob): Do not record GOTO expressions
4719 for GIMPLE statements which locations are already streamed.
4721 2018-07-31 Olivier Hainque <hainque@adacore.com>
4723 * gcc.c (handle_spec_function): Accept a soft_matched_part
4724 argument, as do_spec_1. Pass it down to ...
4725 (eval_spec_function): Accept a soft_matched_part argument,
4726 and pass it down to ...
4727 (do_spec_2): Accept a soft_matched_part argument, and pass
4728 it down to do_spec_1.
4729 (do_spec_1): Pass soft_matched_part to handle_spec_function.
4730 (handle_braces): Update call to handle_spec_function.
4731 (driver::set_up_specs): Update calls to do_spec_2.
4732 (compare_debug_dump_opt_spec_function): Likewise.
4733 (compare_debug_self_opt_spec_function): Likewise.
4735 2018-07-31 Olivier Hainque <hainque@adacore.com>
4737 * common.opt (nolibc): New option.
4738 * doc/invoke.texi (Link Options): Document it.
4739 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
4740 * config/alpha/linux.h: Likewise.
4741 * config/arc/elf.h: Likewise.
4742 * config/arm/uclinux-elf.h: Likewise.
4743 * config/arm/unknown-elf.h: Likewise.
4744 * config/avr/avrlibc.h: Likewise.
4745 * config/bfin/bfin.h: Likewise.
4746 * config/bfin/linux.h: Likewise.
4747 * config/bfin/uclinux.h: Likewise.
4748 * config/darwin.h: Likewise.
4749 * config/darwin10.h: Likewise.
4750 * config/darwin12.h: Likewise.
4751 * config/gnu-user.h: Likewise.
4752 * config/lm32/uclinux-elf.h: Likewise.
4753 * config/pa/pa-hpux11.h: Likewise.
4754 * config/pa/pa64-hpux.h: Likewise.
4755 * config/sparc/sparc.h: Likewise.
4757 2018-07-31 Olivier Hainque <hainque@adacore.com>
4759 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
4760 undefined variables.
4762 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
4765 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
4768 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
4770 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
4771 terminated string literal.
4773 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
4775 PR rtl-optimization/85160
4776 * combine.c (is_just_move): New function.
4777 (try_combine): Allow combining two instructions into two if neither of
4778 the original instructions was a move.
4780 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
4783 * doc/extend.texi (Global Register Variables): Discourage use of type
4785 (Local Register Variables): Likewise.
4787 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
4789 PR tree-optimization/86506
4790 * hwint.h (ceil_log2): Resync with hwint.c implementation.
4792 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
4795 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
4796 hard_regno, make sure no insn between `from` and `to` clobbers it.
4798 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
4799 Tom de Vries <tdevries@suse.de>
4801 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
4802 (PTX_DEFAULT_RUNTIME_DIM): ... this.
4803 (nvptx_goacc_validate_dims): Set default worker and gang dims to
4804 PTX_DEFAULT_RUNTIME_DIM.
4805 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
4807 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
4809 * config/pa/pa.c (pa_output_addr_vec): Align address table.
4810 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
4811 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
4813 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
4815 * config/rs6000/constraints.md (wG constraint): Delete, no longer
4817 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
4818 predicate to reflect toc fusion has been deleted.
4819 (toc_fusion_mem_raw): Delete, no longer used.
4820 (toc_fusion_mem_wrapped): Likewise.
4821 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
4823 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
4824 Delete, no longer used.
4825 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
4826 meant to be used for toc fusion.
4827 (rs6000_debug_print_mode): Delete toc fusion debugging.
4828 (rs6000_debug_reg_global): Likewise.
4829 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
4830 fusion and secondary reload support that were never used.
4831 (rs6000_option_override_internal): Delete TOC fusion, that was only
4832 partially defined, and it did not work unless you also used the
4834 (rs6000_legitimate_address_p): Delete TOC fusion support.
4835 (rs6000_opt_masks): Likewise.
4836 (fusion_wrap_memory_address): Delete function, no longer used.
4837 (fusion_split_address); Delete TOC fusion support.
4838 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
4839 longer used with toc fusion being deleted.
4840 (TARGET_TOC_FUSION_FP): Likewise.
4841 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
4843 (toc fusion spliter): Delete TOC fusion support.
4844 (toc_fusionload_<mode>): Likewise.
4845 (toc_fusionload_di): Likewise.
4846 (fusion_gpr_load_<mode>): Delete generator function, this insn no
4847 longer needs to be named. Rename predicate to delete TOC fusion.
4848 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
4849 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
4850 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
4851 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
4852 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
4854 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
4856 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
4857 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
4858 __int128 in built-in function prototypes.
4859 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
4860 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
4862 2018-07-27 Martin Sebor <msebor@redhat.com>
4864 PR tree-optimization/86696
4865 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
4866 types, including enums.
4867 (handle_char_store): Be prepared for the above function to fail.
4869 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
4871 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
4872 when optimization level is lower than 2 or optimize for size.
4874 2018-07-26 Martin Sebor <msebor@redhat.com>
4876 PR tree-optimization/86043
4877 PR tree-optimization/86042
4878 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
4879 (get_string_cst_length): Rename...
4880 (get_min_string_length): ...to this. Add argument.
4881 (handle_char_store): Extend to handle multi-character stores by
4883 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
4884 * tree.h (initializer_zerop): Add argument.
4886 2018-07-26 Jakub Jelinek <jakub@redhat.com>
4889 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
4890 declare target to variables if they have always,{to,from,tofrom} map
4893 2018-07-26 Martin Liska <mliska@suse.cz>
4896 * lto-wrapper.c: Add linker_output as prefix
4897 for ltrans_output_file.
4899 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
4901 PR rtl-optimization/85805
4902 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
4903 value for hard registers if that was written in the same mode.
4905 2018-07-26 Martin Liska <mliska@suse.cz>
4907 PR gcov-profile/86536
4908 * gcov.c (format_gcov): Use printf format %.*f directly
4909 and do not handle special values.
4911 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4913 * common/config/arc/arc-common.c (arc_option_optimization_table):
4914 Update default optimizations for size.
4916 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4918 * config/arc/arc.md (movsf_insn): Add short instruction selection.
4919 * config/arc/constraints.md (CfZ): New constraint.
4920 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
4921 (subsf3_fpu): Likewise.
4922 (cmpsf_fpu): Likewise.
4923 (cmpsf_fpu_uneq): Likewise.
4925 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4927 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
4929 (arc_is_uncached_mem_p): uncached applies to both the variable and
4932 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4934 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
4937 2018-07-25 David Malcolm <dmalcolm@redhat.com>
4939 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
4940 field "m_scopes" from vec to auto_vec.
4942 2018-07-25 Martin Liska <mliska@suse.cz>
4944 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
4947 2018-07-25 Richard Biener <rguenther@suse.de>
4950 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
4951 special wrt context_die late.
4952 (gen_subprogram_die): Re-use DIEs in local scope.
4954 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
4956 PR tree-optimization/86644
4957 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
4959 2018-07-25 Martin Liska <mliska@suse.cz>
4962 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
4963 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
4965 2018-07-25 Martin Liska <mliska@suse.cz>
4968 * params.def: Explain ASan abbreviation and provide
4969 a documentation link.
4971 2018-07-24 Martin Sebor <msebor@redhat.com>
4973 PR tree-optimization/86622
4974 PR tree-optimization/86532
4975 * builtins.h (string_length): Declare.
4976 * builtins.c (c_strlen): Correct handling of non-constant offsets.
4977 (check_access): Be prepared for non-constant length ranges.
4978 (string_length): Make extern.
4979 * expr.c (string_constant): Only handle the minor non-constant
4980 array index. Use string_constant to compute the length of
4981 a generic string constant.
4983 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
4985 PR tree-optimization/86618
4986 * tree-vect-stmts.c (vectorizable_call): Don't take the address
4987 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
4989 2018-07-24 David Malcolm <dmalcolm@redhat.com>
4991 PR tree-optimization/86636
4992 * json.cc (json::object::set): Fix comment. Add assertions.
4993 (json::array::append): Move here from json.h. Add comment and an
4995 (json::string::string): Likewise.
4996 * json.h (json::array::append): Move to json.cc.
4997 (json::string::string): Likewise.
4998 * optinfo-emit-json.cc
4999 (optrecord_json_writer::impl_location_to_json): Assert that we
5000 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
5001 wrapper around it. Expand the location once, rather than three
5003 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
5004 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
5006 (optrecord_json_writer::optinfo_to_json): Likewise, in four
5007 places. Fix some overlong lines.
5009 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
5011 * config/aarch64/aarch64-simd.md
5012 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
5013 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
5014 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
5015 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
5016 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
5017 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
5018 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
5019 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
5020 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
5022 2018-07-24 Jakub Jelinek <jakub@redhat.com>
5025 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
5026 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
5027 and abs_d == d, do the power of two handling if profitable.
5029 2018-07-24 Richard Biener <rguenther@suse.de>
5031 * match.pd: Add BIT_FIELD_REF canonicalizations.
5033 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5036 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
5038 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5040 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
5041 terminated STRING_CST object.
5043 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5045 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
5047 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
5049 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
5050 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
5051 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
5052 the elements into a register.
5053 (rs6000_split_v4si_init_di_reg): Delete.
5054 (rs6000_split_v4si_init): Delete.
5055 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
5056 (vsx_init_v4si): Rewrite as a define_expand.
5058 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
5060 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
5061 zero_extend argument from memory): New.
5063 2018-07-22 Martin Sebor <msebor@redhat.com>
5066 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
5067 diagnosing calls with unknown arguments unless -Walloca-larger-than
5068 is restricted to less than PTRDIFF_MAX bytes.
5070 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
5072 * doc/gcov.texi (Invoking Gcov): Editorial changes.
5074 2018-07-20 David Malcolm <dmalcolm@redhat.com>
5076 * pretty-print.c (text_info::set_location): Remove redundant
5077 "line_table" parameter from call to rich_location::set_range.
5079 2018-07-20 Martin Sebor <msebor@redhat.com>
5082 * builtins.c (expand_builtin_alloca): Adjust.
5083 * calls.c (alloc_max_size): Simplify.
5084 * cgraphunit.c (cgraph_node::expand): Adjust.
5085 * common.opt (larger_than_size, warn_frame_larger_than): Remove
5087 (frame_larger_than_size): Same.
5088 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
5089 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
5091 * doc/invoke.texi (GCC Command Options): Document option arguments.
5092 Explain byte-size arguments and suffixes.
5093 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
5094 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
5095 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
5096 * doc/options.texi (UInteger): Expand.
5097 (Host_Wide_Int, ByteSize): Document new properties.
5098 * final.c (final_start_function_1): Include sizes in an error message.
5099 * function.c (frame_offset_overflow): Same.
5100 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
5101 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
5102 Diagnose unbounded alloca calls only for limits of less than
5104 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
5105 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
5107 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
5108 only for limits of less than PTRDIFF_MAX.
5109 * langhooks-def.h (lhd_handle_option): Change function argument
5111 * langhooks.c (lhd_handle_option): Same.
5112 * langhooks.h (handle_option): Same.
5113 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
5115 (var_type, var_type_struct): Same.
5116 (var_set): Handle ByteSize flag.
5117 * optc-gen.awk: Add comments to output to ease debugging. Make
5118 use of HOST_WIDE_INT where appropriate.
5119 * opts-gen-save.awk: Use %lx to format unsigned long.
5120 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
5121 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
5122 arguments. Parse bytes-size suffixes.
5123 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
5124 (enum_value_to_arg): Same.
5125 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
5126 (handle_option): Adjust.
5127 (generate_option): Change function argument to HOST_WIDE_INT.
5128 (cmdline_handle_error): Adjust.
5129 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
5130 (set_option): Change function argument to HOST_WIDE_INT.
5131 (option_enabled): Handle cl_host_wide_int.
5132 (get_option_state): Handle CLVC_SIZE.
5133 (control_warning_option): Same.
5134 * opts.c (common_handle_option): Change function argument to
5135 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
5136 OPT_Wvla_larger_than_.
5137 * opts.h (enum cl_var_type): Add an enumerator.
5138 * stor-layout.c (layout_decl): Print a more meaningful warning.
5139 * toplev.c (output_stack_usage): Adjust.
5141 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
5143 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
5144 call to inline_expand_builtin_string_cmp.
5145 (expand_builtin_strcmp): Likewise.
5146 (expand_builtin_strncmp): Likewise.
5147 (inline_string_cmp): Delete the last parameter, change char_type_node
5148 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
5150 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
5151 the inlining expansion on target where the type of the call has same or
5152 narrower precision than unsigned char.
5154 2018-07-20 David Malcolm <dmalcolm@redhat.com>
5156 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
5157 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
5158 * common.opt (fsave-optimization-record): New option.
5159 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
5160 * doc/invoke.texi (-fsave-optimization-record): New option.
5161 * dumpfile.c: Include "optinfo-emit-json.h".
5162 (struct kv_pair): Move to coretypes.h.
5163 (optgroup_options): Make non-static.
5164 (dump_context::end_scope): Call
5165 optimization_records_maybe_pop_dump_scope.
5166 * dumpfile.h (optgroup_options): New decl.
5167 * json.cc: New file.
5169 * optinfo-emit-json.cc: New file.
5170 * optinfo-emit-json.h: New file.
5171 * optinfo.cc: Include "optinfo-emit-json.h".
5172 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
5173 (optinfo_enabled_p): Check optimization_records_enabled_p.
5174 (optinfo_wants_inlining_info_p): Likewise.
5175 * optinfo.h: Update comment.
5176 * profile-count.c (profile_quality_as_string): New function.
5177 * profile-count.h (profile_quality_as_string): New decl.
5178 (profile_count::quality): New accessor.
5179 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
5180 and optinfo_emit_json_cc_tests.
5181 * selftest.h (selftest::json_cc_tests): New decl.
5182 (selftest::optinfo_emit_json_cc_tests): New decl.
5183 * toplev.c: Include "optinfo-emit-json.h".
5184 (compile_file): Call optimization_records_finish.
5185 (do_compile): Call optimization_records_start.
5186 * tree-ssa-live.c: Include optinfo.h.
5187 (remove_unused_scope_block_p): Retain inlining information if
5188 optinfo_wants_inlining_info_p returns true.
5190 2018-07-20 Richard Biener <rguenther@suse.de>
5193 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
5194 to cover -flto-partition=none.
5196 2018-07-20 Martin Liska <mliska@suse.cz>
5198 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
5199 (get_decl_source_range): Remove unused function.
5201 2018-07-20 Richard Biener <rguenther@suse.de>
5203 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
5204 (struct vn_phi_s): Likewise.
5205 (struct vn_reference_s): Likewise.
5206 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
5207 for searching the slot of an entry known to be in the hash itself.
5208 (vn_phi_hasher::equal): Likewise.
5209 (vn_reference_hasher::equal): Likewise.
5210 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
5212 (optimistic_info, current_info): Remove, keeping only valid_info.
5213 (vn_reference_lookup_1): Remove fallback lookup.
5214 (vn_reference_lookup_2): Likewise.
5215 (vn_nary_op_lookup_1): Likewise.
5216 (vn_phi_lookup): Likewise.
5217 (vn_nary_build_or_lookup_1): Make sure to not chain the built
5219 (vn_reference_insert): Adjust, chain the inserted hash element
5220 at last_inserted_ref.
5221 (vn_reference_insert_pieces): Likewise.
5222 (visit_reference_op_call): Likewise.
5223 (vn_nary_op_insert_into): Chain the inserted hash element at
5225 (vn_nary_op_insert_pieces): Adjust.
5226 (vn_nary_op_insert): Likewise.
5227 (vn_nary_op_insert_stmt): Likewise.
5228 (vn_phi_insert): Adjust, chain the inserted hash element at
5230 (process_scc): Remove clearing and copying the optimistic
5231 table. Instead remove elements inserted during an optimistic
5232 iteration from the single table we maintain.
5233 (init_scc_vn): Adjust.
5234 (free_scc_vn): Likewise.
5235 (sccvn_dom_walker::record_cond): Likewise.
5236 (sccvn_dom_walker::after_dom_children): Likewise.
5238 2018-07-19 Martin Sebor <msebor@redhat.com>
5240 PR tree-optimization/84047
5241 PR tree-optimization/83776
5242 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
5243 (check_array_bounds): Call it.
5245 2018-07-19 Martin Sebor <msebor@redhat.com>
5247 * align.h (align_flags): Use member initialization.
5249 2018-07-19 David Malcolm <dmalcolm@redhat.com>
5251 * Makefile.in (OBJS): Add optinfo.o.
5252 * coretypes.h (class symtab_node): New forward decl.
5253 (struct cgraph_node): New forward decl.
5254 (class varpool_node): New forward decl.
5255 * dump-context.h: New file.
5256 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
5258 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
5259 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
5260 (set_alt_dump_file): Likewise.
5261 (dump_context::~dump_context): New dtor.
5262 (dump_gimple_stmt): Move implementation to...
5263 (dump_context::dump_gimple_stmt): ...this new member function.
5264 Add the stmt to any pending optinfo, creating one if need be.
5265 (dump_gimple_stmt_loc): Move implementation to...
5266 (dump_context::dump_gimple_stmt_loc): ...this new member function.
5267 Start a new optinfo and add the stmt to it.
5268 (dump_gimple_expr): Move implementation to...
5269 (dump_context::dump_gimple_expr): ...this new member function.
5270 Add the stmt to any pending optinfo, creating one if need be.
5271 (dump_gimple_expr_loc): Move implementation to...
5272 (dump_context::dump_gimple_expr_loc): ...this new member function.
5273 Start a new optinfo and add the stmt to it.
5274 (dump_generic_expr): Move implementation to...
5275 (dump_context::dump_generic_expr): ...this new member function.
5276 Add the tree to any pending optinfo, creating one if need be.
5277 (dump_generic_expr_loc): Move implementation to...
5278 (dump_context::dump_generic_expr_loc): ...this new member
5279 function. Add the tree to any pending optinfo, creating one if
5281 (dump_printf): Move implementation to...
5282 (dump_context::dump_printf_va): ...this new member function. Add
5283 the text to any pending optinfo, creating one if need be.
5284 (dump_printf_loc): Move implementation to...
5285 (dump_context::dump_printf_loc_va): ...this new member function.
5286 Start a new optinfo and add the stmt to it.
5287 (dump_dec): Move implementation to...
5288 (dump_context::dump_dec): ...this new member function. Add the
5289 value to any pending optinfo, creating one if need be.
5290 (dump_context::dump_symtab_node): New member function.
5291 (dump_context::get_scope_depth): New member function.
5292 (dump_context::begin_scope): New member function.
5293 (dump_context::end_scope): New member function.
5294 (dump_context::ensure_pending_optinfo): New member function.
5295 (dump_context::begin_next_optinfo): New member function.
5296 (dump_context::end_any_optinfo): New member function.
5297 (dump_context::s_current): New global.
5298 (dump_context::s_default): New global.
5299 (dump_scope_depth): Delete global.
5300 (dumpfile_ensure_any_optinfo_are_flushed): New function.
5301 (dump_symtab_node): New function.
5302 (get_dump_scope_depth): Reimplement in terms of dump_context.
5303 (dump_begin_scope): Likewise.
5304 (dump_end_scope): Likewise.
5305 (selftest::temp_dump_context::temp_dump_context): New ctor.
5306 (selftest::temp_dump_context::~temp_dump_context): New dtor.
5307 (selftest::verify_item): New function.
5308 (ASSERT_IS_TEXT): New macro.
5309 (ASSERT_IS_TREE): New macro.
5310 (ASSERT_IS_GIMPLE): New macro.
5311 (selftest::test_capture_of_dump_calls): New test.
5312 (selftest::dumpfile_c_tests): Call it.
5313 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
5314 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
5315 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
5316 descriptive comment.
5317 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
5318 (dump_node, dump_bb): Move these unrelated decls.
5319 (class dump_manager): Add leading comment.
5320 * optinfo.cc: New file.
5321 * optinfo.h: New file.
5323 2018-07-19 Michael Collison <michael.collison@arm.com>
5324 Richard Henderson <rth@redhat.com>
5326 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
5327 (subti): Handle op1 zero.
5328 (subvti4, usub4ti4): New.
5329 (*sub<GPI>3_compare1_imm): New.
5330 (sub<GPI>3_carryinCV): New.
5331 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
5332 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
5334 2018-07-19 Michael Collison <michael.collison@arm.com>
5335 Richard Henderson <rth@redhat.com>
5337 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
5338 (addti3): Create simpler code if low part is already known to be 0.
5339 (addvti4, uaddvti4): New.
5340 (*add<GPI>3_compareC_cconly_imm): New.
5341 (*add<GPI>3_compareC_cconly): New.
5342 (*add<GPI>3_compareC_imm): New.
5343 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
5344 handle constants within this pattern..
5345 (*add<GPI>3_compareV_cconly_imm): New.
5346 (*add<GPI>3_compareV_cconly): New.
5347 (*add<GPI>3_compareV_imm): New.
5348 (add<GPI>3_compareV): New.
5349 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
5350 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
5351 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
5352 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
5353 with 'comparison' operator.
5354 (*add<GPI>3_compareV_cconly_imm): Ditto.
5355 (*add<GPI>3_compareV_cconly): Ditto.
5356 (*add<GPI>3_compareV_imm): Ditto.
5357 (add<GPI>3_compareV): Ditto.
5358 (add<mode>3_carryinC): Ditto.
5359 (*add<mode>3_carryinC_zero): Ditto.
5360 (*add<mode>3_carryinC): Ditto.
5361 (add<mode>3_carryinV): Ditto.
5362 (*add<mode>3_carryinV_zero): Ditto.
5363 (*add<mode>3_carryinV): Ditto.
5365 2018-07-19 Michael Collison <michael.collison@arm.com>
5366 Richard Henderson <rth@redhat.com>
5368 * config/aarch64/aarch64-modes.def (CC_V): New.
5369 * config/aarch64/aarch64-protos.h
5370 (aarch64_addti_scratch_regs): Declare
5371 (aarch64_subvti_scratch_regs): Declare.
5372 (aarch64_expand_subvti): Declare.
5373 (aarch64_gen_unlikely_cbranch): Declare
5374 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
5375 for signed overflow using CC_Vmode.
5376 (aarch64_get_condition_code_1): Handle CC_Vmode.
5377 (aarch64_gen_unlikely_cbranch): New function.
5378 (aarch64_addti_scratch_regs): New function.
5379 (aarch64_subvti_scratch_regs): New function.
5380 (aarch64_expand_subvti): New function.
5382 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5384 * config/aarch64/aarch64-option-extensions.def: New entry for profile
5386 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
5387 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
5390 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5393 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5394 address check not strict.
5396 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5398 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
5400 (store_pair_lanes<mode>): Likewise.
5401 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
5402 enum value 'ADDR_QUERY_LDP_STP_N'.
5403 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
5404 (aarch64_print_address_internal): Add declaration.
5405 (aarch64_print_ldpstp_address): Remove.
5406 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
5407 (aarch64_print_operand): Change printing of 'y'.
5408 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
5409 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
5410 'true' rather than '1'.
5411 * config/aarch64/constraints.md (Uml): Likewise.
5412 (Uml): Rename to Umn.
5415 2018-07-19 Richard Biener <rguenther@suse.de>
5417 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
5419 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
5420 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
5421 (vn_reference_hasher): Likewise.
5422 (struct vn_tables_s): Remove obstack and alloc-pool members.
5423 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
5424 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
5425 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
5426 (vn_reference_insert_pieces): Likewise.
5427 (alloc_vn_nary_op_noinit): Adjust.
5428 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
5429 (vn_phi_eq): Adjust.
5430 (shared_lookup_phiargs): Remove.
5431 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
5432 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
5433 (visit_reference_op_call): Likewise.
5434 (copy_nary, copy_phi, copy_reference): Remove.
5435 (process_scc): Rewind the obstack when iterating. Do not
5436 copy the elements to valid_info but just move them from one
5437 hashtable to the other.
5438 (allocate_vn_table): Adjust.
5439 (free_vn_table): Likewise.
5440 (init_scc_vn): Likewise.
5441 (free_scc_vn): Likewise.
5443 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
5446 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
5447 indirect_return as function type attribute.
5448 (ix86_attribute_table): Change indirect_return to function
5450 * doc/extend.texi: Update indirect_return attribute.
5452 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
5454 * wide-int.h (widest2_int): New.
5455 * gimple-fold.c (arith_overflowed_p): Use it.
5456 * tree.h (widest2_int_cst): New.
5457 * tree-vrp.c (wide_int_binop_overflow): Rename from
5458 vrp_int_const_binop.
5459 Rewrite to work on trees.
5460 (extract_range_from_multiplicative_op_1): Abstract code to...
5461 (wide_int_range_min_max): ...here.
5462 (wide_int_range_cross_product): ...and here.
5463 (extract_range_from_binary_expr_1): Abstract overflow code to...
5464 (wide_int_range_mult_wrapping): ...here.
5465 * tree-vrp.h (wide_int_range_cross_product): New.
5466 (wide_int_range_mult_wrapping): New.
5468 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
5469 Julia Koval <julia.koval@intel.com>
5471 * config/i386/x86-tune-costs.h (skylake_memcpy,
5472 skylake_memset): Replace rep_prefix with unrolling for size 512.
5474 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
5477 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
5478 comparision with EQ_EXPR in last stmt.
5480 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
5482 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
5483 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
5484 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
5485 previously known as "PowerPC AltiVec Built-in Functions". Move
5486 some material to new subsubsections "PowerPC AltiVec Built-in
5487 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
5489 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
5490 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
5491 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
5492 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
5494 2018-07-18 Richard Biener <rguenther@suse.de>
5496 PR tree-optimization/86557
5497 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
5500 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
5502 * config/s390/s390.c (s390_function_profiler): Generate CFI.
5504 2018-07-17 Jeff Law <law@redhat.com>
5506 * config/arm/arm.c (get_label_padding): Update for recent
5507 changes to label_to_alignment.
5509 PR tree-optimization/86010
5510 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
5512 * config/mips/mips.c (vr4130_align_insns): Update for recent
5513 changes to label_to_alignment.
5515 * config/frv/frv.c (frv_label_align): Update for recent changes
5516 to label_to_alignment.
5518 * config/nios2/nios2.c (nios2_label_align): Update for recent
5519 changes which dropped ALIGN_LABELS_LOG.
5521 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
5523 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
5524 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
5526 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
5528 * config/arc/arc.c (arc_label_align): Use align_labels instead of
5529 deprecated align_labels_log.
5531 2018-07-17 Richard Biener <rguenther@suse.de>
5534 * dwarf2out.c (init_sections_and_labels): Always generate
5535 a debug_line_str_section for early LTO debug.
5536 (dwarf2out_finish): Reset debug_line_str_hash output early.
5537 Bump counter for extra dwarf5 .debug_loc labels to not conflict
5539 (dwarf2out_early_finish): Output debug_line_str.
5541 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
5543 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
5544 index register on z196 or later.
5546 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
5548 * config/s390/s390.c (s390_default_align): Set default function
5550 (s390_override_options_after_change): Call s390_default align.
5551 (s390_option_override_internal): Call s390_default_align.
5552 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
5554 2018-07-17 Jakub Jelinek <jakub@redhat.com>
5557 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
5558 to _looptemp_ clauses, other than the first two.
5560 2018-07-17 Martin Liska <mliska@suse.cz>
5562 * opts.c: Do not enable OPT_falign_* for -Os.
5564 2018-07-17 Martin Liska <mliska@suse.cz>
5566 * align.h (MAX_CODE_ALIGN): New.
5567 (MAX_CODE_ALIGN_VALUE): New.
5568 * common/config/i386/i386-common.c (ix86_handle_option):
5569 (MAX_CODE_ALIGN): Moved to align.h.
5570 * final.c (MAX_CODE_ALIGN): Likewise.
5571 * opts.c (parse_and_check_align_values):
5572 (MAX_CODE_ALIGN): Likewise.
5573 (MAX_CODE_ALIGN_VALUE): Likewise.
5575 2018-07-17 Martin Liska <mliska@suse.cz>
5577 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
5578 in order to fulfil coding style.
5579 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
5580 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5581 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
5582 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
5583 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
5584 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
5585 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
5586 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
5587 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5589 2018-07-17 Martin Liska <mliska@suse.cz>
5591 * align.h: New file.
5592 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
5594 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
5595 align_flags of label_to_alignment.
5596 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
5598 * config/m68k/m68k.c: Do not use removed align_labels_value and
5600 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
5601 (LOOP_ALIGN): Likewise.
5602 (LABEL_ALIGN): Likewise.
5603 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
5604 Remove not used macro.
5605 (rs6000_loop_align): Change return type to align_flags.
5606 (rs6000_loop_align_max_skip): Remove.
5607 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
5608 Change return type to align_flags.
5609 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
5610 Remove not used macro.
5611 (rs6000_loop_align): Change return type to align_flags.
5612 (rs6000_loop_align_max_skip): Remove.
5613 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
5614 * config/rx/rx-protos.h (rx_align_for_label): Make it
5616 * config/rx/rx.c (rx_align_for_label): Change return type
5618 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
5620 into align_flags class.
5621 (LABEL_ALIGN): Likewise.
5622 (LOOP_ALIGN): Likewise.
5623 * config/s390/s390.c (s390_label_align): Use align_flags
5625 (s390_asm_output_function_label): Likewise.
5626 * config/sh/sh.c (sh_override_options_after_change):
5627 Use align_flags class directly without macros.
5628 (find_barrier): Likewise.
5629 (barrier_align): Likewise.
5630 (sh_loop_align): Likewise.
5631 * config/spu/spu.c (spu_option_override):
5632 Use align_flags_tuple::get_value instead of removed macros.
5633 (spu_sched_init): Likewise.
5634 * config/spu/spu.h (GTY): Likewise.
5635 * config/visium/visium.c (visium_option_override):
5636 Set "8" as default secondary alignment.
5637 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
5638 in order to guarantee secondary alignment of 8.
5639 * coretypes.h: Include align.h header file.
5640 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
5641 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
5642 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
5643 * doc/tm.texi.in: Likewise.
5644 * final.c (struct label_alignment): Remove not used structure.
5645 (LABEL_ALIGN): Change type to align_flags.
5646 (LOOP_ALIGN): Likewise.
5647 (JUMP_ALIGN): Likewise.
5648 (default_loop_align_max_skip): Remove.
5649 (default_label_align_max_skip): Likewise.
5650 (default_jump_align_max_skip): Likewise.
5651 (default_label_align_after_barrier_max_skip):
5652 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
5653 (LABEL_TO_MAX_SKIP): Remove.
5654 (label_to_alignment): Return align_flags type instead of integer.
5655 (label_to_max_skip): Remove.
5656 (align_fuzz): Use align_flags type.
5657 (compute_alignments): Use align_flags type and use align_flags::max
5658 to combine multiple alignments.
5659 (grow_label_align): Grow vec instead of C array.
5660 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
5661 (shorten_branches): Use align_flags type and use align_flags::max
5662 to combine multiple alignments.
5663 (final_scan_insn_1): Remove usage of secondary alignment that comes
5664 from label alignment, but instead use proper secondary alignment
5665 which is computed in grow_label_align.
5666 * flags.h (struct align_flags_tuple): Move to align.h.
5667 (struct align_flags): Likewise.
5668 (state_align_loops): Rename to align_loops.
5669 (state_align_jumps): Rename to align_jumps.
5670 (state_align_labels): Rename to align_labels.
5671 (state_align_functions): Rename to align_functions.
5672 (align_loops_log): Remove.
5673 (align_jumps_log): Remove.
5674 (align_labels_log): Remove.
5675 (align_functions_log): Remove.
5676 (align_loops_max_skip): Remove.
5677 (align_jumps_max_skip): Remove.
5678 (align_labels_max_skip): Remove.
5679 (align_functions_max_skip): Remove.
5680 (align_loops_value): Remove.
5681 (align_jumps_value): Remove.
5682 (align_labels_value): Remove.
5683 (align_functions_value): Remove.
5684 * output.h (label_to_alignment): Change return type to align_flags.
5685 (label_to_max_skip): Remove.
5686 * target.def: Remove loop_align_max_skip, label_align_max_skip,
5687 jump_align_max_skip macros.
5688 * targhooks.h (default_loop_align_max_skip): Remove.
5689 (default_label_align_max_skip): Likewise.
5690 (default_jump_align_max_skip): Likewise.
5691 (default_label_align_after_barrier_max_skip): Remove.
5692 * toplev.c (read_log_maxskip): Use ::normalize function.
5693 (parse_N_M): Remove not used argument and also call ::normalize.
5694 (parse_alignment_opts): Do not pass unused arguments.
5695 * varasm.c (assemble_start_function): Use directly align_functions
5696 instead of removed macros.
5697 * system.h: Do not poison removed macros.
5699 2018-07-17 Jakub Jelinek <jakub@redhat.com>
5702 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
5703 and cond temporaries don't have reference type if iterator has
5704 pointer type. For init use &for_pre_body instead of pre_p if
5705 for_pre_body is non-empty.
5707 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5709 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
5710 double-double modes to SFmode directly directly.
5711 (trunc<mode>sf2_fprs): Delete.
5713 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5715 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
5716 for conversions between IFmode and the decimal floating point modes.
5717 (init_float128_ieee): Use the correct names for conversions between
5718 KFmode and the decimal floating point modes.
5720 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5722 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
5723 for the conversions between TDmode and IFmode.
5724 (init_float128_ieee): Use more correct names for the conversions
5725 between TDmode and KFmode.
5727 2018-07-16 Jakub Jelinek <jakub@redhat.com>
5729 PR tree-optimization/86526
5730 * builtins.c (expand_builtin_memcmp): Formatting fixes.
5731 (inline_expand_builtin_string_cmp): Likewise.
5732 (inline_string_cmp): Likewise. Use c_readstr instead of
5733 builtin_memcpy_read_str. Add unit_mode temporary.
5735 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
5738 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
5739 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
5741 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5743 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
5744 Alphabetize prototypes of built-in functions, separating out
5745 built-in functions that are listed in this section but should be
5746 described elsewhere.
5748 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
5751 * expmed.c (emit_store_flag): Do not emit setcc followed by a
5752 conditional move when trapping comparison was split to a
5753 non-trapping one (and vice versa).
5755 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5757 * config/s390/s390.c (s390_function_profiler): Generate nops
5758 instead of profiler call sequences.
5759 * config/s390/s390.opt: Add the new option.
5761 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5763 * config/s390/s390.c (s390_function_profiler): Generate
5764 __mcount_loc section.
5765 * config/s390/s390.opt: Add the new option.
5767 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5769 * common.opt: Add the new warning.
5770 * config/s390/s390.c (s390_function_profiler): Emit "brasl
5771 %r0,__fentry__" when -mfentry is specified.
5772 (s390_option_override_internal): Disallow -mfentry for 31-bit
5774 * config/s390/s390.opt: Add the new option.
5776 2018-07-16 Richard Biener <rguenther@suse.de>
5779 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
5780 for function-local FUNCTION_DECL and RESULT_DECL immediately.
5782 2018-07-16 Martin Liska <mliska@suse.cz>
5785 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
5788 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
5790 * config/arc/arcHS.md: Update ARCHS scheduling rules.
5792 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
5794 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
5796 * config/arc/arc-cpus.def (hs4x): New cpu.
5798 * config/arc/arc-tables.opt: Regenerate.
5799 * config/arc/arc.c (arc_sched_issue_rate): New function.
5800 (TARGET_SCHED_ISSUE_RATE): Define.
5801 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
5802 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
5804 (attr tune): Add ARCHS4x tune values.
5805 (attr tune_dspmpy): Define.
5806 (*tst): Correct instruction type.
5807 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
5808 * config/arc/arcHS4x.md: New file.
5809 * config/arc/fpu.md: Update instruction type attributes.
5810 * config/arc/t-multilib: Regenerate.
5812 2018-07-16 Tom de Vries <tdevries@suse.de>
5815 * var-tracking.c (vt_initialize): Fix pre_dec handling.
5817 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5819 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
5822 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
5824 PR tree-optimization/86514
5825 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
5826 conversion to a boolean type from a type with greater precision.
5828 2018-07-16 Tom de Vries <tdevries@suse.de>
5830 * var-tracking.c (vt_initialize): Print adjusted insn slim if
5831 dump_flags request TDF_SLIM.
5833 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
5835 * fold-const.c (int_const_binop_1): Abstract...
5836 (wide_int_binop): ...wide int code here.
5837 (poly_int_binop): ...poly int code here.
5838 Abstract the rest of int_const_binop_1 into int_const_binop.
5839 * fold-const.h (wide_int_binop): New.
5840 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
5841 Remove useless PLUS/MINUS_EXPR case.
5842 (zero_nonzero_bits_from_vr): Move wide int code...
5843 (zero_nonzero_bits_from_bounds): ...here.
5844 (extract_range_from_binary_expr_1): Move mask optimization code...
5845 (range_easy_mask_min_max): ...here.
5846 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
5847 (range_easy_mask_min_max): New.
5849 2018-07-15 Jeff Law <law@redhat.com>
5852 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
5855 2018-07-14 Jim Wilson <jimw@sifive.com>
5857 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
5859 2018-07-14 Paul Koning <ni1d@arrl.net>
5861 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
5863 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
5865 * lto-streamer-out.c (copy_function_or_variable): Dump info about
5868 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
5869 Steve Munroe <munroesj52@gmail.com>
5871 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
5872 (_mm_andnot_si128): Likewise.
5873 (_mm_or_si128): Likewise.
5874 (_mm_xor_si128): Likewise.
5876 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
5879 * builtins.c (expand_builtin_memcmp): Inline the calls first
5880 when result_eq is false.
5881 (expand_builtin_strcmp): Inline the calls first.
5882 (expand_builtin_strncmp): Likewise.
5883 (inline_string_cmp): New routine. Expand a string compare
5884 call by using a sequence of char comparison.
5885 (inline_expand_builtin_string_cmp): New routine. Inline expansion
5886 a call to str(n)cmp/memcmp.
5887 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
5889 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
5891 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
5893 * config/arm/driver-arm.c: Include arm-native.h.
5894 (host_detect_local_cpu): Use auto-generated data tables.
5895 (vendors, arm_cpu_table): Delete. Move part information to ...
5896 * config/arm/arm-cpus.in: ... here.
5897 * config/arm/parsecpu.awk (gen_native): New function.
5898 (vendor, part): New CPU fields.
5899 (END): Add support for building the native CPU detection tables.
5900 * config/arm/t-arm (arm-native.h): Add build rule.
5901 (driver-arm.o): Add dependency on arm-native.h.
5903 2018-07-13 Richard Biener <rguenther@suse.de>
5906 * match.pd (addr1 - addr2): Allow either of the operand to
5909 2018-07-13 Tom de Vries <tdevries@suse.de>
5911 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
5914 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
5916 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
5917 arrays instead of numbered variables.
5919 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
5921 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
5922 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
5924 2018-07-13 Richard Biener <rguenther@suse.de>
5927 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
5928 instead of get_context_die.
5930 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
5931 Richard Biener <rguenther@suse.de>
5934 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
5935 that the loop latch destination where phi is defined.
5937 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
5939 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
5940 (riscv_expand_epilogue): Add assertion to check interrupt mode.
5941 (riscv_set_current_function): Extract getting interrupt type to new
5943 (riscv_get_interrupt_type): New function.
5944 (riscv_merge_decl_attributes): New function, checking interrupt type is
5946 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
5948 2018-07-12 Paul Koning <ni1d@arrl.net>
5950 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
5953 2018-07-12 Paul Koning <ni1d@arrl.net>
5955 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
5956 zero reference, add doloop_end instead.
5957 * doc/md.texi (decrement_and_branch_until_zero): Remove.
5958 (Looping patterns): Remove decrement_and_branch_until_zero. Add
5959 detail for doloop_end.
5961 2018-07-12 Martin Sebor <msebor@redhat.com>
5964 * attribs.c (decl_attributes): Reject conflicting attributes before
5965 calling attribute handlers.
5967 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
5969 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
5971 (gcc::dump_manager::get_dump_file_name): likewise.
5972 (dump_begin): Likewise.
5973 * dumpfile.h (dump_begin): Update prototype.
5974 (gcc::dump_manager::get_dump_file_name,
5975 gcc::dump_manager::get_dump_file_name): Update prototype.
5977 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5979 * internal-fn.h (vectorizable_internal_fn_p): New function.
5980 * tree-vect-slp.c (compatible_calls_p): Likewise.
5981 (vect_build_slp_tree_1): Remove nops argument. Handle calls
5982 to internal functions.
5983 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
5985 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5987 * fold-const.h (inverse_conditions_p): Declare.
5988 * fold-const.c (inverse_conditions_p): New function.
5989 * match.pd: Use inverse_conditions_p. Add folds of view_converts
5990 that test the inverse condition of a conditional internal function.
5991 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
5992 * internal-fn.c (internal_fn_mask_index): Handle conditional
5994 (vectorized_internal_fn_supported_p): New function.
5995 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
5996 (any_pred_load_store): Replace with...
5997 (need_to_predicate): ...this new variable.
5998 (redundant_ssa_names): New variable.
5999 (ifcvt_can_use_mask_load_store): Move initial checks to...
6000 (ifcvt_can_predicate): ...this new function. Handle tree codes
6001 for which a conditional internal function exists.
6002 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
6003 instead of ifcvt_can_use_mask_load_store. Update after variable
6005 (predicate_load_or_store): New function, split out from
6006 predicate_mem_writes.
6007 (check_redundant_cond_expr): New function.
6008 (value_available_p): Likewise.
6009 (predicate_rhs_code): Likewise.
6010 (predicate_mem_writes): Rename to...
6011 (predicate_statements): ...this. Use predicate_load_or_store
6012 and predicate_rhs_code.
6013 (combine_blocks, tree_if_conversion): Update after above name changes.
6014 (ifcvt_local_dce): Handle redundant_ssa_names.
6015 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
6016 general conditional functions.
6017 * tree-vect-stmts.c (vectorizable_call): Likewise.
6019 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6020 Alan Hayward <alan.hayward@arm.com>
6021 David Sherwood <david.sherwood@arm.com>
6023 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
6024 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
6025 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
6026 plus and minus and convert them into IFN_COND_FMA-based sequences.
6027 (convert_mult_to_fma): Handle conditional plus and minus.
6029 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6031 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
6032 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
6033 (cond_fnms_optab): New optabs.
6034 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
6036 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
6037 * internal-fn.h (get_conditional_internal_fn): Declare.
6038 (get_unconditional_internal_fn): Likewise.
6039 * internal-fn.c (cond_ternary_direct): New macro.
6040 (expand_cond_ternary_optab_fn): Likewise.
6041 (direct_cond_ternary_optab_supported_p): Likewise.
6042 (FOR_EACH_COND_FN_PAIR): Likewise.
6043 (get_conditional_internal_fn): New function.
6044 (get_unconditional_internal_fn): Likewise.
6045 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
6046 (gimple_match_op::gimple_match_op): Add a new overload for 5
6048 (gimple_match_op::set_op): Likewise.
6049 (gimple_resimplify5): Declare.
6050 * genmatch.c (decision_tree::gen): Generate simplifications for
6052 * gimple-match-head.c (gimple_simplify): Define an overload for
6053 5 operands. Handle calls with 5 arguments in the top-level overload.
6054 (convert_conditional_op): Handle conversions from unconditional
6055 internal functions to conditional ones.
6056 (gimple_resimplify5): New function.
6057 (build_call_internal): Pass a fifth operand.
6058 (maybe_push_res_to_seq): Likewise.
6059 (try_conditional_simplification): Try converting conditional
6060 internal functions to unconditional internal functions.
6061 Handle 3-operand unconditional forms.
6062 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
6063 Define ternary equivalents of the current rules for binary conditional
6065 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
6067 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
6068 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
6069 (optab): Handle them.
6070 (SVE_COND_FP_TERNARY): New int iterator.
6071 (sve_fmla_op, sve_fmad_op): New int attributes.
6072 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
6073 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
6074 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
6076 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6078 * target.def (preferred_else_value): New target hook.
6079 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
6080 * doc/tm.texi: Regenerate.
6081 * targhooks.h (default_preferred_else_value): Declare.
6082 * targhooks.c (default_preferred_else_value): New function.
6083 * internal-fn.h (conditional_internal_fn_code): Declare.
6084 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
6085 (get_conditional_internal_fn): Use it.
6086 (conditional_internal_fn_code): New function.
6087 * gimple-match.h (gimple_match_cond): New struct.
6088 (gimple_match_op): Add a cond member function.
6089 (gimple_match_op::gimple_match_op): Update all forms to take a
6091 * genmatch.c (expr::gen_transform): Use the same condition as res_op
6092 for the suboperation, but don't specify a particular else_value.
6093 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
6094 (visit_nary_op, visit_reference_op_load): Pass
6095 gimple_match_cond::UNCOND to the gimple_match_op constructor.
6096 * gimple-match-head.c: Include tree-eh.h
6097 (convert_conditional_op): New function.
6098 (maybe_resimplify_conditional_op): Likewise.
6099 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
6100 (gimple_resimplify2): Likewise.
6101 (gimple_resimplify3): Likewise.
6102 (gimple_resimplify4): Likewise.
6103 (maybe_push_res_to_seq): Return null for conditional operations.
6104 (try_conditional_simplification): New function.
6105 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
6107 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
6109 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
6111 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
6113 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
6115 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
6117 (hash_tree): Do not hash DECL_FCONTEXT
6118 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
6119 Do not stream DECL_FCONTEXT.
6120 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
6121 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
6123 2018-07-12 Richard Biener <rguenther@suse.de>
6126 * dwarf2out.c (gen_block_die): Only output blocks when they have
6127 at least one !DECL_IGNORED_P variable.
6129 2018-07-12 Richard Biener <rguenther@suse.de>
6132 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
6133 Remove -mieee-fp handling.
6135 2018-07-12 Richard Biener <rguenther@suse.de>
6137 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
6138 left-over from last patch.
6140 2018-07-12 Jakub Jelinek <jakub@redhat.com>
6142 PR tree-optimization/86492
6143 * gimple-ssa-store-merging.c
6144 (imm_store_chain_info::coalesce_immediate_stores): Call
6145 check_no_overlap even for the merge_overlapping case. Formatting fix.
6147 2018-07-12 Richard Biener <rguenther@suse.de>
6150 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
6151 move possibly trapping operations into the conditional.
6153 2018-07-12 Richard Biener <rguenther@suse.de>
6155 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
6156 (vn_lookup_simplify_result): Remove recursion limit applied
6158 (vn_nary_build_or_lookup_1): Adjust.
6159 (try_to_simplify): Likewise.
6160 * gimple-match-head.c (gimple_resimplify1): Instead apply one
6162 (gimple_resimplify2): Likewise.
6163 (gimple_resimplify3): Likewise.
6164 (gimple_resimplify4): Likewise.
6166 2018-07-11 Jakub Jelinek <jakub@redhat.com>
6168 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
6169 Use __mmask64 type instead of __mmask8 for __M argument.
6170 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
6171 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
6173 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
6174 __mmask16 for __M argument.
6175 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
6176 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
6177 to __mmask16 instead of __mmask8.
6178 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
6179 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
6180 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
6181 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
6182 instead of __mmask16 for __U argument.
6183 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
6184 __mmask16 instead of __mmask8 for __U argument.
6185 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
6187 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
6189 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
6191 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
6193 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
6195 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
6197 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
6199 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
6201 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
6203 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
6205 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
6206 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
6207 return type as well as __M argument type and all casts from __mmask8
6209 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
6210 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
6211 return type as well as __M argument type and all casts from __mmask8
6213 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
6214 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
6215 return type as well as __M argument type and all casts from __mmask8
6217 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
6218 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
6219 return type as well as __M argument type and all casts from __mmask8
6221 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
6222 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
6225 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
6227 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
6228 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
6231 2018-07-11 Paul Koning <ni1d@arrl.net>
6233 * doc/md.texi (define_subst): Document how multiple occurrences of
6234 the same argument in the replacement pattern are handled.
6236 2018-07-11 Paul Koning <ni1d@arrl.net>
6238 * doc/extend.texi (Common Variable Attributes): Move "mode" into
6240 (Common Type Attributes): Add "mode" attribute.
6242 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
6244 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
6245 stream DECL_ORIGINAL_TYPE.
6246 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
6247 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
6248 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
6249 Do not walk original type.
6250 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
6252 (write_ts_decl_non_common_tree_pointers): Do not stream
6254 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
6255 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
6257 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
6259 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
6260 thread twice from the same starting edge.
6262 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
6264 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
6265 * gimple.c (gimple_call_nonnull_result_p): ...here...
6266 (gimple_call_nonnull_arg): ...and here.
6267 * gimple.h (gimple_call_nonnull_result_p): New.
6268 (gimple_call_nonnull_arg): New.
6270 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
6272 * config/arm/arm-cpus.in: Move information from fpu field of each
6273 cpu definition to the isa field.
6274 * config/arm/parsecpu.awk (fpu): Delete match rule.
6275 (gen_comm_data): Don't add bits from the CPU's FPU entry.
6277 2018-07-11 Richard Biener <rguenther@suse.de>
6280 * dwarf2out.c (init_sections_and_labels): Use
6281 output_asm_line_debug_info consistently.
6282 (dwarf2out_early_finish): Likewise.
6283 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
6286 2018-07-11 Richard Biener <rguenther@suse.de>
6288 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
6289 Rework father_bb setting in a way to avoid propagating constants
6290 multiple times on a loop body.
6292 2018-07-10 Mark Wielaard <mark@klomp.org>
6295 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
6297 2018-07-10 Richard Biener <rguenther@suse.de>
6299 * hash-map.h (hash_map::iterator::operator*): Return
6300 references to key and value.
6302 2018-07-10 Jakub Jelinek <jakub@redhat.com>
6305 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
6306 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
6307 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
6308 (gimplify_omp_for): For composite loops, move outer
6309 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
6310 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
6311 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
6312 TREE_LIST for both the original class iterator and the "last" helper
6313 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
6314 loop, remember has_decl_expr from outer composite loops for the
6315 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
6317 2018-07-09 Martin Sebor <msebor@redhat.com>
6321 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
6322 accessing implicitly initialized array elements.
6323 * expr.c (string_constant): Handle string initializers of
6324 character arrays within aggregates.
6325 * gimple-fold.c (fold_array_ctor_reference): Add argument.
6326 Store element offset. As a special case, handle zero size.
6327 (fold_nonarray_ctor_reference): Same.
6328 (fold_ctor_reference): Add argument. Store subobject offset.
6329 * gimple-fold.h (fold_ctor_reference): Add argument.
6331 2018-07-09 Paul Koning <ni1d@arrl.net>
6333 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
6334 (pdp11_insn_cost): New function.
6335 (pdp11_md_asm_adjust): New function.
6336 (TARGET_INVALID_WITHIN_DOLOOP): Define.
6337 (pdp11_rtx_costs): Update to match machine better.
6338 (output_addr_const_pdp11): Correct format mismatch warnings.
6339 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
6340 * config/pdp11/pdp11.md: General change to add base_cost and/or
6341 length attributes for use by new pdp11_insn_cost function.
6342 (MIN_BRANCH): Correct definition.
6344 (doloop_end): Use standard pattern name for looping pattern.
6345 (doloop_end_nocc): New.
6346 (movsf): Add another constraint alternative.
6347 (zero_extendqihi2): Add constraint alternatives for not in place
6349 (zero_extendhisi2): Remove.
6350 (shift patterns): Add CC handling variants.
6354 (define_peephole2): New peephole to recognize mov that sets CC for
6357 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6359 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
6360 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
6363 2018-07-09 Jakub Jelinek <jakub@redhat.com>
6366 * real.c (real_nextafter): Return true if result is denormal.
6368 2018-07-09 Martin Liska <mliska@suse.cz>
6370 * common.opt: Add back wrongly removed attribute.
6372 2018-07-09 Richard Biener <rguenther@suse.de>
6375 * dwarf2out.c (gen_block_die): For an early generated DIE
6376 always output high/low PC attributes.
6378 2018-07-09 Tom de Vries <tdevries@suse.de>
6380 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
6381 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
6382 onto VAR_DECL with abstract origin.
6384 2018-07-07 Jim Wilson <jimw@sifive.com>
6386 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
6388 2018-07-07 Tom de Vries <tdevries@suse.de>
6390 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
6392 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
6394 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
6396 (combine_bound): Use wide-int overflow calculation instead of
6398 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
6400 * fold-const.c (int_const_binop_2): Same.
6401 (extract_muldiv_1): Same.
6402 (fold_div_compare): Same.
6403 (fold_abs_const): Same.
6405 * poly-int.h (add): Same.
6409 * predict.c (predict_iv_comparison): Same.
6410 * profile-count.c (slow_safe_scale_64bit): Same.
6411 * simplify-rtx.c (simplify_const_binary_operation): Same.
6412 * tree-chrec.c (tree_fold_binomial): Same.
6413 * tree-data-ref.c (split_constant_offset_1): Same.
6414 * tree-if-conv.c (idx_within_array_bound): Same.
6415 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
6416 * tree-ssa-phiopt.c (minmax_replacement): Same.
6417 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
6418 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
6419 * vr-values.c (vr_values::adjust_range_with_scev): Same.
6420 * wide-int.cc (wi::add_large): Same.
6421 (wi::mul_internal): Same.
6422 (wi::sub_large): Same.
6423 (wi::divmod_internal): Same.
6424 * wide-int.h: Change overflow type to overflow_type for neg, add,
6425 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
6426 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
6427 mul_internal, divmod_internal.
6428 (overflow_type): New enum.
6429 (accumulate_overflow): New.
6431 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
6433 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
6434 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
6436 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
6438 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
6439 argument is checked for zero before entering loop, avoid checking again.
6441 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
6443 * gimplify.h (generic_expr_could_trap_p): Set as global function.
6444 * gimplify.h (generic_expr_could_trap_p): Likwise.
6445 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
6447 2018-07-06 Jakub Jelinek <jakub@redhat.com>
6449 PR tree-optimization/86401
6450 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
6451 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
6452 (fold_bit_and_mask): ... here. New helper function for match.pd.
6453 * fold-const.h (fold_bit_and_mask): Declare.
6454 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
6456 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
6459 * target.def (translate_mode_attribute): New hook.
6460 * targhooks.h (default_translate_mode_attribute): Declare.
6461 * targhooks.c (default_translate_mode_attribute): New function.
6462 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
6463 * doc/tm.texi: Regenerate.
6464 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
6465 (rs6000_translate_mode_attribute): New function.
6467 2018-07-06 Paul Koning <ni1d@arrl.net>
6469 * doc/md.texi (define_split): Document DONE and FAIL.
6470 (define_peephole2): Ditto.
6472 2018-07-05 Jeff Law <law@redhat.com>
6474 PR tree-optimization/86010
6475 * tree-ssa-dse.c (compute_trims): More aggressively trim at
6476 both the head and tail of mem* and str* calls.
6478 2018-07-05 Jim Wilson <jimw@sifive.com>
6480 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
6482 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
6484 * config/aarch64/aarch64-simd.md: correct flags text for
6485 MIN_EXPR replacement.
6487 2018-07-05 James Clarke <jrtc27@jrtc27.com>
6489 * configure: Regenerated.
6491 2018-07-05 Carl Love <cel@us.ibm.com>
6493 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
6494 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
6495 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
6496 VSX_BUILTIN_DOUBLEL_V4SF.
6498 2018-07-05 Martin Sebor <msebor@redhat.com>
6501 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
6502 than its domain to compute its the upper bound of a char array.
6504 2018-07-05 Nathan Sidwell <nathan@acm.org>
6506 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
6507 * doc/cpp.texi: Update comment.
6508 * doc/tm.texi: Rebuilt.
6509 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
6510 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
6511 * doc/extend.texi (Backwards Compatibility): Clarify it is system
6512 headers affected by extern "C".
6513 * system.h: Poison NO_IMPLICIT_EXTERN_C.
6514 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
6515 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
6516 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
6517 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
6518 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
6519 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
6520 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
6521 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
6522 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
6523 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
6524 config/sparc/sp64-elf.h, config/spu/spu.h,
6525 config/stormy16/stormy16.h, config/v850/v850.h,
6526 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
6527 define NO_IMPLICIT_EXTERN_C.
6528 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
6530 2018-07-05 Tamar Christina <tamar.christina@arm.com>
6533 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6534 instead of GET_MODE_SIZE when comparing Units.
6536 2018-07-05 Tamar Christina <tamar.christina@arm.com>
6539 * rtlanal.c (set_noop_p): Constrain on mode change,
6540 include hard-reg-set.h
6542 2018-07-05 Tamar Christina <tamar.christina@arm.com>
6544 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
6546 2018-07-05 Jakub Jelinek <jakub@redhat.com>
6549 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
6552 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
6554 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
6556 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
6559 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
6561 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
6563 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
6566 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
6568 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
6570 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
6571 neg_*_op* variables.
6573 2018-07-04 Martin Liska <mliska@suse.cz>
6575 * tree-switch-conversion.c: Define
6576 max_ratio_for_speed and max_ratio_for_size constants.
6578 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
6579 Martin Liska <mliska@suse.cz>
6584 * common.opt: Rename align options with 'str_' prefix.
6585 * common/config/i386/i386-common.c (set_malign_value): New
6587 (ix86_handle_option): Use it to set -falign-* options/
6588 * config/aarch64/aarch64-protos.h (struct tune_params): Change
6589 type from int to string.
6590 * config/aarch64/aarch64.c: Update default values from int
6592 * config/alpha/alpha.c (alpha_override_options_after_change):
6594 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
6595 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6596 max skip conditionally.
6597 * config/i386/freebsd.h (SUBALIGN_LOG): New.
6598 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6599 max skip conditionally.
6600 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6601 max skip conditionally.
6602 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
6603 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6604 max skip conditionally.
6605 * config/i386/i386.c (struct ptt): Change type from int to
6607 (ix86_default_align): Set default values.
6608 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
6609 max skip conditionally.
6610 * config/i386/iamcu.h (SUBALIGN_LOG): New.
6611 (ASM_OUTPUT_MAX_SKIP_ALIGN):
6612 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
6613 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6614 max skip conditionally.
6615 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
6616 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
6617 * config/i386/x86-64.h (SUBALIGN_LOG): New.
6618 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6619 max skip conditionally.
6620 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6621 * config/ia64/ia64.c (ia64_option_override): Set default values
6622 for alignment options.
6623 * config/m68k/m68k.c: Handle new str_align_* options.
6624 * config/mips/mips.c (mips_set_compression_mode): Change
6626 (mips_option_override): Set default values for options.
6627 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
6629 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6631 * config/rx/rx.c (rx_option_override): Likewise.
6632 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
6633 (LABEL_ALIGN): Use align_labels_log.
6634 (LOOP_ALIGN): Use align_loops_align.
6635 * config/s390/s390.c (s390_asm_output_function_label): Use new
6637 * config/sh/sh.c (sh_override_options_after_change):
6638 Change type of constants.
6639 * config/spu/spu.c (spu_sched_init): Likewise.
6640 * config/sparc/sparc.c (sparc_option_override): Set default
6642 * config/visium/visium.c (visium_option_override): Likewise.
6643 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
6644 emit p2align format with last argument if it's not needed.
6645 * doc/invoke.texi: Document extended format of -falign-*.
6646 * final.c: Use align_labels alignment.
6647 * flags.h (struct target_flag_state): Change type to use
6649 (struct align_flags_tuple): New.
6650 (struct align_flags): Likewise.
6651 (align_loops_log): Redefine macro to use new types.
6652 (align_loops_max_skip): Redefine macro to use new types.
6653 (align_jumps_log): Redefine macro to use new types.
6654 (align_jumps_max_skip): Redefine macro to use new types.
6655 (align_labels_log): Redefine macro to use new types.
6656 (align_labels_max_skip): Redefine macro to use new types.
6657 (align_functions_log): Redefine macro to use new types.
6658 (align_loops): Redefine macro to use new types.
6659 (align_jumps): Redefine macro to use new types.
6660 (align_labels): Redefine macro to use new types.
6661 (align_functions): Redefine macro to use new types.
6662 (align_functions_max_skip): Redefine macro to use new types.
6663 (align_loops_value): New macro.
6664 (align_jumps_value): New macro.
6665 (align_labels_value): New macro.
6666 (align_functions_value): New macro.
6667 * function.c (invoke_set_current_function_hook): Propagate
6668 alignment values from flags to global variables default in
6670 * ipa-icf.c (sem_function::equals_wpa): Use
6671 cl_optimization_option_eq instead of memcmp.
6672 * lto-streamer.h (cl_optimization_stream_out): Support streaming
6674 (cl_optimization_stream_in): Likewise.
6675 * optc-save-gen.awk: Support strings in cl_optimization.
6676 * opth-gen.awk: Likewise.
6677 * opts.c (finish_options): Remove error checking of invalid
6679 (MAX_CODE_ALIGN): Remove.
6680 (MAX_CODE_ALIGN_VALUE): Likewise.
6681 (parse_and_check_align_values): New function.
6682 (check_alignment_argument): Likewise.
6683 (common_handle_option): Use check_alignment_argument.
6684 * opts.h (parse_and_check_align_values): Declare.
6685 * toplev.c (init_alignments): Remove.
6686 (read_log_maxskip): New.
6687 (parse_N_M): Likewise.
6688 (parse_alignment_opts): Likewise.
6689 (backend_init_target): Remove usage of init_alignments.
6690 * toplev.h (parse_alignment_opts): Declare.
6691 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
6693 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
6694 * tree.c (cl_option_hasher::equal): New.
6695 * varasm.c: Use new global macros.
6697 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
6699 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
6700 Use a simpler align directive also if MAXSKIP = ALIGN-1.
6701 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6702 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6703 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6704 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6705 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
6706 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
6707 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
6708 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6709 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6710 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6711 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6713 2018-07-04 Martin Liska <mliska@suse.cz>
6714 Jonathan Wakely <jwakely@redhat.com>
6716 * coverage.c: Use correct type.
6717 * doc/invoke.texi: Language correction.
6719 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
6722 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
6723 ENDBRANCH for non-tail call which may return via indirect branch.
6724 * doc/extend.texi: Document indirect_return attribute.
6726 2018-07-03 Martin Sebor <msebor@redhat.com>
6728 PR tree-optimization/86274
6729 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
6731 (format_floating): Correct handling of infinities and NaNs.
6733 2018-07-03 Martin Sebor <msebor@redhat.com>
6735 * print-tree.c (print_real_cst): New function.
6736 (print_node_brief): Call it.
6737 (print_node): Ditto.
6739 2018-07-03 Jeff Law <law@redhat.com>
6741 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
6742 into a single pattern.
6744 * config/h8300/h8300.md (ors code_iterator): New.
6745 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
6746 a single pattern and single splitter.
6747 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
6748 (iorqi3_1, xorqi3_1): Likewise.
6749 (iorqi3, xorqi3 expanders): Similarly.
6751 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
6752 (movmd_internal) into a single pattern using the P mode iterator.
6753 (movmd splitters): Similarly.
6754 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
6755 (movsd splitters): Similarly.
6757 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
6758 ADDB, ADDW and ADDL into a single ADD attribute which selects the
6759 right table based on the size of the operand.
6760 * config/h8300/h8300.md (length_table): Corresponding changes. All
6761 references to "addb", "addw" and "addl" changed to "add".
6762 (btst patterns): Merge two variants into a single pattern.
6763 (tstqi, tsthi): Likewise.
6764 (addhi3_incdec, addsi3_incdec): Likewise.
6765 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
6766 (mulhi3, mulsi3): Likewise.
6767 (udivhi3, udivsi3): Likewise.
6768 (divhi3, divsi3): Likewise.
6769 (andorqi3, andorhi3, andorsi3): Likewise.
6771 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
6774 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
6775 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
6777 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6779 PR tree-optimization/85694
6780 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
6781 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
6783 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
6784 (<u>avg<mode>3_ceil): New patterns.
6786 2018-07-03 David Malcolm <dmalcolm@redhat.com>
6788 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
6789 scan-tree-dump directive.
6790 * gcc.dg/vect/slp-perm-2.c: Likewise.
6791 * gcc.dg/vect/slp-perm-3.c: Likewise.
6792 * gcc.dg/vect/slp-perm-5.c: Likewise.
6793 * gcc.dg/vect/slp-perm-6.c: Likewise.
6794 * gcc.dg/vect/slp-perm-7.c: Likewise.
6795 * gcc.dg/vect/slp-perm-8.c: Likewise.
6797 2018-07-03 Marek Polacek <polacek@redhat.com>
6800 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
6802 2018-07-03 Richard Biener <rguenther@suse.de>
6805 * tree-ssa-structalias.c (find_func_clobbers): Properly
6806 handle indirect calls.
6808 2018-07-03 Jeff Law <law@redhat.com>
6810 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
6811 (shifts): New code iterator.
6812 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
6813 expander. Fix HImode handling on H8/SX.
6814 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
6815 (subqi3, subhi3, subsi3 expanders): Likewise.
6816 (andqi3, andhi3, andsi3 expanders): Likewise.
6817 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
6818 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
6819 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
6820 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
6821 (zero_extendqihi2, zero_extendqisi2): Likewise.
6822 (extendqihi2, extendqisi2): Likewise.
6823 (rotlqi3, rotlhi3, rotlsi3): Likewise.
6824 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
6825 (rotlqi3_1, rotlhi3_1): Likewise.
6826 (logicalhi3_sn, logicalsi3_sn): Likewise.
6827 (logicalhi3, logicalsi3): Likewise.
6829 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6831 * tree-vect-patterns.c (vect_recog_rotate_pattern)
6832 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
6833 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
6834 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
6835 type to append_pattern_def_seq instead of creating a stmt_vec_info
6837 (build_mask_conversion): Likewise. Remove vinfo argument.
6838 (vect_add_conversion_to_patterm): Likewise, renaming to...
6839 (vect_add_conversion_to_pattern): ...this.
6840 (vect_recog_mask_conversion_pattern): Update call to
6841 build_mask_conversion. Pass the vector type to
6842 append_pattern_def_seq here too.
6843 (vect_recog_gather_scatter_pattern): Update call to
6844 vect_add_conversion_to_pattern.
6846 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6848 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
6849 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
6850 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
6851 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
6852 STMT_VINFO_PATTERN_DEF_SEQ to null here.
6853 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
6854 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
6855 append_pattern_def_seq instead of new_pattern_def_seq.
6856 (vect_recog_divmod_pattern): Do both of the above.
6857 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
6860 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6862 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
6863 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
6864 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
6865 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
6866 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
6867 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
6868 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
6869 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
6870 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
6871 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
6872 parameter with a single stmt_vec_info.
6873 (vect_recog_func_ptr): Likewise.
6874 (vect_recog_gather_scatter_pattern): Likewise, folding in...
6875 (vect_try_gather_scatter_pattern): ...this.
6876 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
6877 the stmt_vec_info of the statement to be matched. Don't clear
6878 STMT_VINFO_RELATED_STMT.
6879 (vect_pattern_recog): Update call accordingly.
6881 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6883 PR tree-optimization/85694
6884 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
6885 (uavgM3_ceil): Document new optabs.
6886 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
6887 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
6889 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
6890 (savg_ceil_optab): New optabs.
6891 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
6892 (vect_vect_recog_func_ptrs): Add it.
6893 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
6894 constant directly from the associated lhs.
6896 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6898 * tree-vect-patterns.c (vect_split_statement): New function.
6899 (vect_convert_input): Use it to try to split an existing cast.
6901 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6903 * poly-int.h (print_hex): New function.
6904 * dumpfile.h (dump_dec, dump_hex): Declare.
6905 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
6906 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
6907 min_input_precision, operation_precision and operation_sign.
6908 * tree-vect-patterns.c (vect_get_range_info): New function.
6909 (vect_same_loop_or_bb_p, vect_single_imm_use)
6910 (vect_operation_fits_smaller_type): Delete.
6911 (vect_look_through_possible_promotion): Add an optional
6912 single_use_p parameter.
6913 (vect_recog_over_widening_pattern): Rewrite to use new
6914 stmt_vec_info infomration. Handle one operation at a time.
6915 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
6916 (vect_truncatable_operation_p, vect_set_operation_type)
6917 (vect_set_min_input_precision): New functions.
6918 (vect_determine_min_output_precision_1): Likewise.
6919 (vect_determine_min_output_precision): Likewise.
6920 (vect_determine_precisions_from_range): Likewise.
6921 (vect_determine_precisions_from_users): Likewise.
6922 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
6923 (vect_vect_recog_func_ptrs): Put over_widening first.
6925 (vect_pattern_recog): Call vect_determine_precisions.
6927 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6929 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
6930 statements that have been replaced by further pattern statements.
6931 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
6933 2018-07-03 Richard Biener <rguenther@suse.de>
6935 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
6936 always set *dt. Dump vectype in vectype overload.
6937 * dumpfile.h (dump_gimple_expr): New function.
6938 (dump_gimple_expr_loc): Likewise.
6939 * dumpfile.c (dump_gimple_expr): New function.
6940 (dump_gimple_expr_loc): Likewise.
6942 2018-07-02 Jeff Law <law@redhat.com>
6944 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
6945 the H8/300, H8/300H and H8/S variants into a single pattern.
6946 (movhi_h8300, movqi_h8300hs): Similarly.
6947 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
6948 (QHI mode iterator): New.
6950 * config/h8300/h8300.md: Remove trailing whitespace.
6952 2018-07-02 Jim Wilson <jimw@sifive.com>
6954 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
6955 instead of emit_insn for interrupt returns.
6956 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
6957 (riscv_sret, riscv_uret): Likewise.
6959 2018-07-02 David Malcolm <dmalcolm@redhat.com>
6961 * pretty-print.c (selftest::test_pp_format): Move save and restore
6962 of quotes to class auto_fix_quotes, and add an instance.
6963 * selftest.c: Include "intl.h".
6964 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
6965 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
6966 * selftest.h (selftest::auto_fix_quotes): New class.
6968 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
6970 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
6971 (aarch64_sve_prepare_conditional_op): Remove.
6972 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
6973 Allow aarch64_simd_reg_or_zero as select operand; remove
6974 the aarch64_sve_prepare_conditional_op call.
6975 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
6976 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
6977 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
6978 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
6979 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
6980 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
6981 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
6982 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
6983 and a splitters to match all of the *_any patterns.
6984 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
6986 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
6987 (SVE_COND_FP_BINARY_REV): Remove.
6988 (sve_int_op_rev, sve_fp_op_rev): New.
6989 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
6990 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
6991 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
6992 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
6993 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
6994 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
6995 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
6996 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
6997 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
6999 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
7000 Remove match_dup 1 from the inner unspec.
7001 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7003 * config/aarch64/aarch64.md (movprfx): New attr.
7004 (length): Default movprfx to 8.
7005 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
7006 (*madd<SVE_I>, *msub<SVE_I): Likewise.
7007 (*<su>mul<SVE_I>3_highpart): Likewise.
7008 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
7009 (*v<ASHIFT><SVE_I>3): Likewise.
7010 (*<su><MAXMIN><SVE_I>3): Likewise.
7011 (*<su><MAXMIN><SVE_F>3): Likewise.
7012 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
7013 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
7014 (*div<SVE_F>4): Likewise.
7016 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
7018 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
7021 2018-07-02 Richard Biener <rguenther@suse.de>
7023 PR tree-optimization/86363
7024 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
7025 memset argument refers to a non-variable address.
7027 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
7029 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
7030 {PLUS,MINUS}_EXPR code to...
7031 (adjust_symbolic_bound): ...here,
7032 (combine_bound): ...here,
7033 (set_value_range_with_overflow): ...and here.
7035 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
7037 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
7039 (extract_range_from_abs_expr): ...here.
7041 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
7043 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
7044 -fno-omit-frame-pointer when not optimizing.
7046 2018-07-02 Martin Liska <mliska@suse.cz>
7049 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
7050 (propagate_nothrow): Likewise.
7052 2018-07-02 Martin Liska <mliska@suse.cz>
7055 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
7057 2018-07-02 David Malcolm <dmalcolm@redhat.com>
7059 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
7060 function in r262149, changing "loc" param from source_location to
7061 const dump_location_t &.
7062 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
7063 declaration, as above.
7065 2018-07-01 Paul Koning <ni1d@arrl.net>
7067 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
7068 -munit-asm, -mgnu-asm, -mdec-asm.
7069 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
7070 (pdp11_output_labelref): New.
7071 (pdp11_output_def): New.
7072 (pdp11_output_addr_vec_elt): New.
7073 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
7074 %# and %@ format codes.
7075 (pdp11_option_override): New.
7076 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
7077 (pdp11_output_ident): New.
7078 (pdp11_asm_named_section): New.
7079 (pdp11_asm_init_sections): New.
7080 (pdp11_file_start): New.
7081 (pdp11_file_end): New.
7082 (output_ascii): Use .ascii/.asciz for -mdec-asm.
7083 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
7084 %o, like %c but octal.
7085 (pdp11_option_override): New.
7086 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
7088 (DATA_SECTION_ASM_OP): Ditto.
7089 (READONLY_DATA_SECTION_ASM_OP): New.
7090 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
7091 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
7092 (ASM_OUTPUT_LABELREF): Ditto.
7093 (ASM_OUTPUT_DEF): Ditto.
7094 (ASM_OUTPUT_EXTERNAL): New.
7095 (ASM_OUTPUT_SOURCE_FILENAME): New.
7096 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
7097 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
7098 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
7099 %# and %@ format codes.
7100 * config/pdp11/pdp11.opt (mgnu-asm): New.
7101 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
7102 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
7103 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
7105 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
7107 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
7108 dereferencing path[] beyond its length.
7110 (debug_all_paths): New.
7111 (rewire_first_differing_edge): New.
7112 (adjust_paths_after_duplication): New.
7113 (duplicate_thread_path): Call adjust_paths_after_duplication.
7115 (thread_through_all_blocks): Add new argument to
7116 duplicate_thread_path.
7118 2018-06-30 Jim Wilson <jimw@sifive.com>
7120 * config/riscv/predicates.md (p2m1_shift_operand): New.
7121 (high_mask_shift_operand): New.
7122 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
7123 pattern using p2m1_shift_operand.
7124 (lshsi3_zero_extend_3+2): New combiner pattern using
7125 high_mask_shift_operand.
7127 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7129 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
7131 (vect_recog_rotate_pattern): ...here.
7132 (vect_convert_input): Try to insert casts of invariants in the
7134 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
7135 preheader to be empty.
7137 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7139 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
7140 vector type. If given, install it in the new statement's
7142 (vect_element_precision): New function.
7143 (vect_unpromoted_value): New struct.
7144 (vect_unpromoted_value::vect_unpromoted_value): New function.
7145 (vect_unpromoted_value::set_op): Likewise.
7146 (vect_look_through_possible_promotion): Likewise.
7147 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
7148 (vect_widened_op_tree, vect_convert_input): Likewise.
7149 (vect_convert_inputs, vect_convert_output): Likewise.
7150 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
7151 to handle the optional cast of the multiplication result and
7152 vect_widened_op_tree to detect the widened multiplication itself.
7153 Do not require the input and output of promotion casts to have
7154 the same sign, but base the signedness of the operation on the
7155 input rather than the result. If the pattern includes two
7156 promotions, check that those promotions have the same sign.
7157 Do not restrict the MULT_EXPR handling to a double-width result;
7158 handle quadruple-width results and wider. Use vect_convert_inputs
7159 to convert the inputs to the common type.
7160 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
7161 to handle the optional cast of the ABS result. Also allow a sign
7162 change or a sign extension between the ABS and MINUS.
7163 Use vect_widened_op_tree to detect the widened subtraction and use
7164 vect_convert_inputs to convert the inputs to the common type.
7165 (vect_handle_widen_op_by_const): Delete.
7166 (vect_recog_widen_op_pattern): New function.
7167 (vect_recog_widen_mult_pattern): Use it.
7168 (vect_recog_widen_shift_pattern): Likewise.
7169 (vect_recog_widen_sum_pattern): Use
7170 vect_look_through_possible_promotion to handle the promoted
7173 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7175 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
7176 the containing gimple_seq *.
7177 * gimple-iterator.h (gsi_for_stmt): Declare it.
7178 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
7179 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
7180 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7181 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
7182 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
7184 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
7186 (vect_mark_pattern_stmts): ...here. Handle cases in which the
7187 statement being replaced is part of an existing pattern
7188 definition sequence, inserting the new pattern statements before
7190 (vect_pattern_recog_1): Don't return a bool. If the statement
7191 is already part of a pattern, instead apply pattern matching
7192 to the pattern definition statements. Don't clear the
7193 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
7194 (vect_pattern_recog): Don't break after the first match;
7195 continue processing the pattern definition statements instead.
7196 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
7198 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7200 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7201 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7202 (vect_recog_widen_sum_pattern): Use it.
7204 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7206 * tree-vect-loop.c (vectorizable_reduction): Assert that the
7207 phi is not a pattern statement and has not been replaced by
7208 a pattern statement.
7209 * tree-vect-patterns.c (type_conversion_p): Don't check
7210 STMT_VINFO_IN_PATTERN_P.
7211 (vect_recog_vector_vector_shift_pattern): Likewise.
7212 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
7213 the pattern statement rather than the original statement; check
7214 directly for a WIDEN_MULT_EXPR here.
7215 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
7216 vect_is_simple_use to return the pattern statement rather
7217 than the original statement; use is_pattern_stmt_p to check
7218 for such a pattern statement.
7219 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
7220 to return the pattern statement rather than the original statement;
7221 don't do the same transformation here.
7222 (vect_is_simple_use): If the defining statement has been replaced
7223 by a pattern statement, return the pattern statement instead.
7224 Remove the corresponding (local) transformation from the vectype
7227 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7229 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
7230 end and default to null.
7231 * tree-vect-loop.c (vect_create_epilog_for_reduction)
7232 (vectorizable_reduction): Update calls accordingly, dropping the
7233 gimple ** argument if the passed-back statement isn't needed.
7234 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
7235 (vect_recog_rotate_pattern): Likewise.
7236 (vect_recog_mask_conversion_pattern): Likewise.
7237 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7238 (vect_mask_constant_operand_p): Likewise.
7239 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
7240 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
7241 (get_group_load_store_type, get_load_store_type): Likewise.
7242 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
7243 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
7244 (vectorizable_conversion, vectorizable_assignment): Likewise.
7245 (vectorizable_shift, vectorizable_operation): Likewise.
7246 (vectorizable_store, vect_is_simple_cond): Likewise.
7247 (vectorizable_condition, vectorizable_comparison): Likewise.
7248 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
7249 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
7250 and move it to the end. Cope with null def_stmt_outs.
7252 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
7254 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
7256 2018-06-29 Jeff Law <law@redhat.com>
7258 * config/v850/v850.c (v850_legitimate_address_p): Handle large
7259 displacements for TARGET_V850E2V3 and newer.
7260 (TARGET_LRA_P): Remove. Defaults to LRA now.
7261 * config/v850/v850.md (sign23byte_load): Remove.
7262 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
7263 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
7265 2018-06-29 Martin Liska <mliska@suse.cz>
7268 * coverage.c (coverage_init): Mangle full path name.
7269 * doc/invoke.texi: Document the change.
7270 * gcov-io.c (mangle_path): New.
7271 * gcov-io.h (mangle_path): Likewise.
7272 * gcov.c (mangle_name): Use mangle_path for path mangling.
7274 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7276 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
7277 if starting source register is not even.
7279 2018-06-29 Martin Liska <mliska@suse.cz>
7281 PR tree-optimization/86263
7282 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
7283 Make edge redirection.
7285 2018-06-29 David Malcolm <dmalcolm@redhat.com>
7287 * dumpfile.c (dump_loc): Add indentation based on scope depth.
7288 (dump_scope_depth): New variable.
7289 (get_dump_scope_depth): New function.
7290 (dump_begin_scope): New function.
7291 (dump_end_scope): New function.
7292 * dumpfile.h (get_dump_scope_depth): New declaration.
7293 (dump_begin_scope): New declaration.
7294 (dump_end_scope): New declaration.
7295 (class auto_dump_scope): New class.
7296 (AUTO_DUMP_SCOPE): New macro.
7297 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
7300 2018-06-29 Richard Biener <rguenther@suse.de>
7302 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
7303 compute_all_dependences succeeds.
7304 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
7305 exceed --param loop-max-datarefs-for-datadeps.
7307 2018-06-29 Jakub Jelinek <jakub@redhat.com>
7309 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
7311 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
7314 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
7315 alternative 0 in preferred_for_speed attribute.
7317 2018-06-28 Paul Koning <ni1d@arrl.net>
7319 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
7320 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
7321 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
7322 * config/pdp11/pdp11.md: Correct "length" attribute calculation
7323 for shift insn patterns.
7325 2018-06-28 David Malcolm <dmalcolm@redhat.com>
7327 * cgraph.c (cgraph_node::get_body): Replace assignments to
7328 "dump_file" with calls to set_dump_file.
7329 * dumpfile.c (alt_dump_file): Make static, and group with...
7330 (alt_flags): ...this definition.
7331 (dumps_are_enabled): New variable.
7332 (refresh_dumps_are_enabled): New function.
7333 (set_dump_file): New function.
7334 (set_alt_dump_file): New function.
7335 (gcc::dump_manager::dump_start): Replace assignments to
7336 "dump_file" and "alt_dump_file" with calls to set_dump_file and
7338 (gcc::dump_manager::dump_finish): Likewise.
7339 * dumpfile.h (alt_dump_file): Delete decl.
7340 (dumps_are_enabled): New variable decl.
7341 (set_dump_file): New function decl.
7342 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
7344 * tree-nested.c (lower_nested_functions): Replace assignments to
7345 "dump_file" with calls to set_dump_file.
7347 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
7349 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
7350 goto_locus of each outgoing edge of each basic block.
7352 2018-06-28 Richard Biener <rguenther@suse.de>
7354 * dwarf2out.c (decl_scope_table): Remove.
7355 (push_decl_scope): Likewise.
7356 (pop_decl_scope): Likewise.
7357 (gen_type_die_for_member): Do not call push/pop_decl_scope.
7358 (gen_struct_or_union_type_die): Likewise.
7359 (gen_tagged_type_die): Likewise.
7360 (dwarf2out_init): Do not initialize decl_scope_table.
7361 (dwarf2out_c_finalize): Do not free it.
7363 2018-06-28 Richard Biener <rguenther@suse.de>
7365 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
7366 deciding whether to not re-use a DIE.
7368 2018-06-28 Richard Biener <rguenther@suse.de>
7370 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
7371 DW_AT_abstract_origin attribute.
7373 2018-06-28 Martin Liska <mliska@suse.cz>
7375 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
7376 Use newly introduced constants.
7377 * tree-switch-conversion.h (struct jump_table_cluster):
7378 Define max_ratio_for_size and max_ratio_for_speed.
7380 2018-06-28 Martin Liska <mliska@suse.cz>
7382 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
7383 Add new checking assert to catch invalid state.
7384 (jump_table_cluster::can_be_handled): Handle single case
7386 (jump_table_cluster::is_beneficial): Bail out for such case.
7387 (bit_test_cluster::find_bit_tests):
7388 Add new checking assert to catch invalid state.
7389 (bit_test_cluster::can_be_handled): Handle single case
7391 (bit_test_cluster::is_beneficial): Bail out for such case.
7392 (switch_decision_tree::analyze_switch_statement):
7395 2018-06-28 Martin Liska <mliska@suse.cz>
7397 * common.opt: Introduce -completion option.
7398 * gcc.c (driver_handle_option): Handle it.
7399 (driver::main): Print completions if completion
7401 * opt-suggestions.c (option_proposer::get_completions):
7403 (option_proposer::suggest_completion): Likewise.
7404 (option_proposer::find_param_completions): Likewise.
7405 (verify_autocompletions): Likewise.
7406 (test_completion_valid_options): Likewise.
7407 (test_completion_valid_params): Likewise.
7408 (in_completion_p): Likewise.
7409 (empty_completion_p): Likewise.
7410 (test_completion_partial_match): Likewise.
7411 (test_completion_garbage): Likewise.
7412 (opt_proposer_c_tests): Likewise.
7413 * opt-suggestions.h: Declare new functions.
7414 * opts.c (common_handle_option): Handle OPT__completion_.
7415 * selftest-run-tests.c (selftest::run_tests): Add
7416 opt_proposer_c_tests.
7417 * selftest.c (assert_str_startswith): New.
7418 * selftest.h (assert_str_startswith): Likewise.
7419 (opt_proposer_c_tests): New.
7420 (ASSERT_STR_STARTSWITH): Likewise.
7422 2018-06-28 Martin Liska <mliska@suse.cz>
7424 * Makefile.in: Add opt-suggestions.o.
7425 * gcc-main.c: Include opt-suggestions.h.
7426 * gcc.c (driver::driver): Likewise.
7427 (driver::~driver): Remove m_option_suggestions.
7428 (driver::build_option_suggestions): Moved to option_proposer.
7429 (driver::suggest_option): Likewise.
7430 (driver::handle_unrecognized_options): Use option_proposer.
7431 * gcc.h (class driver): Add new memver m_option_proposer.
7432 * opt-suggestions.c: New file.
7433 * opt-suggestions.h: New file.
7435 2018-06-28 Martin Liska <mliska@suse.cz>
7437 * vec.h (class auto_string_vec): New (moved from auto_argvec).
7438 (auto_string_vec::~auto_string_vec): Likewise.
7440 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
7442 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
7443 prevent_decl_creation_for_types fields up and add reset_location field.
7444 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
7445 statement if id->reset_location is true.
7446 (copy_edges_for_bb): Do not set goto_locus on the new edges if
7447 id->reset_location is true.
7448 (copy_phis_for_bb): Force input_location on the arguments if
7449 id->reset_location is true.
7450 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
7451 is set on the function to be inlined.
7453 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
7455 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
7457 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
7459 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
7460 registers for Pmode.
7461 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
7462 hard registers for the clobbered pseudo.
7464 2018-06-27 Paul Koning <ni1d@arrl.net>
7466 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
7467 mutually exclusive options.
7468 * config/pdp11/constraints.md (h): New constraint.
7469 (O): Update definition to match shift code generation.
7470 (D): New constraint.
7471 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
7473 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
7475 (output_jump): Change arguments.
7476 (pdp11_fixed_cc_regs): New function.
7477 (pdp11_cc_mode): Ditto.
7478 (pdp11_expand_shift): Ditto.
7479 (pdp11_assemble_shift): Ditto.
7480 (pdp11_small_shift): Ditto.
7481 (pdp11_branch_cost): Remove.
7482 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
7484 (pdp11_register_move_cost): Update for CC registers.
7485 (pdp11_rtx_costs): Add case for LSHIFTRT.
7486 (pdp11_output_jump): Add CCNZ mode conditional branches.
7487 (notice_update_cc_on_set): Remove.
7488 (pdp11_cc_mode): New function.
7489 (simple_memory_operand): Correct pre/post decrement case.
7490 (no_side_effect_operand): New function.
7491 (pdp11_regno_reg_class): Add CC_REGS class.
7492 (pdp11_fixed_cc_regs): New function.
7493 (pdp11_small_shift): New function.
7494 (pdp11_expand_shift): New function to expand shift insns.
7495 (pdp11_assemble_shift): New function to output shifts.
7496 (pdp11_branch_cost): Remove.
7497 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
7498 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
7499 (WCHAR_TYPE): Ditto.
7500 (PTRDIFF_TYPE): Ditto.
7501 (ADJUST_INSN_LENGTH): New macro.
7502 (FIXED_REGISTERS): Add CC registers.
7503 (CALL_USED_REGISTERS): Ditto.
7505 (REG_CLASS_NAMES): Ditto.
7506 (REG_CLASS_CONTENTS): Ditto.
7507 (SELECT_CC_MODE): Use new function.
7508 (TARGET_FLAGS_REGNUM): New macro.
7509 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
7510 (cc0_reg_rtx): Remove.
7511 (CC_STATUS_MDEP): Remove.
7512 (CC_STATUS_MDEFP_INIT): Remove.
7513 (CC_IN_FPU): Remove.
7514 (NOTICE_UPDATE_CC): Remove.
7515 (REGISTER_NAMES): Add CC registers.
7516 (BRANCH_COST): Change to constant 1.
7517 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
7519 * config/pdp11/pdp11.opt (mbcopy): Remove.
7520 (mbcopy-builtin): Remove.
7521 (mbranch-cheap): Remove.
7522 (mbranch-expensive): Remove.
7523 * config/pdp11/predicates.md (expand_shift_operand): Update to
7524 match shift code generation.
7525 (ccnz_operator): New predicate.
7526 * doc/invoke.texi (PDP-11 Options): Remove deleted options
7527 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
7528 Remove non-existent option -mabshi, -mno-abshi. Document mutually
7530 * doc/md.texi (PDP-11): Document new D and h constraints. Update
7531 description of O constraint.
7533 2018-06-27 Jeff Law <law@redhat.com>
7534 Austin Law <austinklaw@gmail.com>
7536 * config/v850/v850.md (addsi3_set_flags): New pattern.
7537 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
7538 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
7539 (zero_extendhisi2_v850_set_flags): Likewise.
7540 (zero_extendqisi2_v850_set_flags): Likewise.
7541 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
7542 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
7543 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
7545 * config/v850/v850-protos.h (notice_update_cc): Remove.
7546 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
7547 (v850_print_operand): Handle 'D' and "d".
7548 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
7549 Add handling of arithmetic/logical operations compared against zero.
7550 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
7551 Do not look at v850_compare_op, instead get mode from last argument.
7552 (v850_gen_compare): Remove
7553 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
7554 after reload for prologue insns.
7555 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
7557 (construct_save_jarl): Likewise.
7558 (TARGET_FLAGS_REGNUM): Define.
7559 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
7560 (NOTICE_UPDATE_CC): Remove.
7561 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
7562 than cc0. Conditionalize on reload_completed.
7563 (cmpsi_insn, setfcc_insn): Likewise.
7564 (tst1 splitter): Turn into define_and_split which sets the flags
7566 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
7567 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
7568 (cstoresf4, cstoredf4): Clobber the flags.
7569 (cmpsi, cmpsf, cmpdf): Remove expanders.
7570 (setf_insn): Remove pattern.
7571 (addsi3): Turn into define_and_split which clobbers the flags after
7572 reload and a suitable pattern (addsi3_clobber_flags) for use after
7574 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
7575 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
7576 (ashrsi3, ashrsi3_v850e2): Likewise.
7577 (bins): Clobber the flags.
7578 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
7579 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
7580 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
7581 (call_value_internal_short, call_value_internal_long): Likewise.
7582 (callt_save_interrupt, callt_return_interrupt): Likewise.
7583 (save_interrupt, return_interrupt): Likewise.
7584 (callt_save_all_interrupt, save_all_interrupt): Likewise.
7585 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
7586 (restore_all_interrupt, _restore_all_interrupt): Likewise.
7587 (All FP comparisons): Only allow after reload has completed.
7589 (divh, divhu): Tweak output template.
7590 (branch_z_normal, branch_z_invert): Remove
7591 (branch_nz_normal, branch_nz_invert): Likewise.
7592 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
7594 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
7595 * config/v850/v850.c (notice_update_cc): Remove.
7596 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
7597 (CC_NO_CARRY): Likewise.
7598 (NOTICE_UPDATE_CC): Define to nothing.
7599 * config/v850/v850.md: Remove block comment on cc0 handling
7600 Remove "cc" attribute from all patterns. Remove cc_status handling
7601 from all patterns. Minor formatting fixes.
7603 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7605 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
7606 (cortex-a76.cortex-a55): Likewise.
7607 * config/aarch64/aarch64-tune.md: Regenerate.
7608 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
7609 cortex-a76.cortex-a55.
7611 2018-06-27 Jeff Law <law@redhat.com>
7613 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
7614 (MULTILIB_DIRNAMES): Similarly.
7616 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
7618 * gimple.h (gimple_return_retbnd): Delete.
7619 (gimple_return_set_retbnd): Likewise.
7620 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
7621 gimple_return_set_retbnd.
7622 * gimple-pretty-print.c (dump_gimple_return): Remove call to
7623 gimple_return_retbnd and adjust.
7624 * tree-inline.h (struct copy_body_data): Remove retbnd field.
7625 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
7626 Explicitly return NULL in a couple more cases. Move assertion
7627 on debug statements and remove unreachable code.
7628 (reset_debug_binding): Do not test id->retbnd.
7629 (expand_call_inline): Do not set it.
7631 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
7633 * configure.ac: Add --disable-gcov option.
7634 * configure: Regenerate.
7635 * Makefile.in: Honour @enable_gcov@.
7636 * doc/install.texi: Document --disable-gcov.
7638 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7640 * config/arm/arm-cpus.in (cortex-a76): New entry.
7641 (cortex-a76.cortex-a55): Likewise.
7642 * config/arm/arm-tables.opt: Regenerate.
7643 * config/arm/arm-tune.md: Likewise.
7644 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
7645 * doc/invoke.texi (ARM Options): Document cortex-a76 and
7646 cortex-a76.cortex-a55.
7648 2018-06-27 Tamar Christina <tamar.christina@arm.com>
7651 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
7653 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
7655 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
7657 (EPILOGUE_USES): Likewise.
7659 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
7661 * tree-inline.c (remap_location): New function extracted from...
7662 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
7663 (copy_phis_for_bb): ...here. Call remap_location.
7664 (copy_cfg_body): Adjust call to copy_edges_for_bb.
7666 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
7668 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
7669 unaligned vsx for 16B memset.
7671 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
7674 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
7675 ieee128_float_type_node to long_double_type_node unless
7676 TARGET_LONG_DOUBLE_128 is set.
7678 2018-06-26 David Malcolm <dmalcolm@redhat.com>
7680 * cfgloop.c (get_loop_location): Convert return type from
7681 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
7682 by implicit construction from rtx_insn *, and using
7683 dump_user_location_t::from_function_decl for the fallback case.
7684 * cfgloop.h (get_loop_location): Convert return type from
7685 location_t to dump_user_location_t.
7686 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
7687 dump_printf_loc to pass in a dump_location_t rather than a
7688 location_t, via the gimple stmt.
7689 * coverage.c (get_coverage_counts): Update calls to
7690 dump_printf_loc to pass in dump_location_t rather than a
7692 * doc/optinfo.texi (Dump types): Convert example of
7693 dump_printf_loc from taking "locus" to taking "insn". Update
7694 description of the "_loc" calls to cover dump_location_t.
7695 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
7697 (dump_user_location_t::dump_user_location_t): New constructors,
7698 from gimple *stmt and rtx_insn *.
7699 (dump_user_location_t::from_function_decl): New function.
7700 (dump_loc): Make static.
7701 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
7702 const dump_location_t &.
7703 (dump_generic_expr_loc): Delete.
7704 (dump_printf_loc): Convert param "loc" from location_t to
7705 const dump_location_t &.
7706 (selftest::test_impl_location): New function.
7707 (selftest::dumpfile_c_tests): New function.
7708 * dumpfile.h: Include "profile-count.h".
7709 (class dump_user_location_t): New class.
7710 (struct dump_impl_location_t): New struct.
7711 (class dump_location_t): New class.
7712 (dump_printf_loc): Convert 2nd param from source_location to
7713 const dump_location_t &.
7714 (dump_generic_expr_loc): Delete.
7715 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
7716 const dump_location_t &.
7717 * gimple-fold.c (fold_gimple_assign): Update call to
7718 dump_printf_loc to pass in a dump_location_t rather than a
7719 location_t, via the gimple stmt.
7720 (gimple_fold_call): Likewise.
7721 * gimple-loop-interchange.cc
7722 (loop_cand::analyze_iloop_reduction_var): Update for change to
7723 check_reduction_path.
7724 (tree_loop_interchange::interchange): Update for change to
7726 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
7727 change in return-type of find_loop_location.
7728 (graphite_regenerate_ast_isl): Likewise.
7729 * graphite-optimize-isl.c (optimize_isl): Likewise.
7730 * graphite.c (graphite_transform_loops): Likewise.
7731 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
7732 pass in a dump_location_t rather than a location_t, via the
7734 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7735 * ipa.c (walk_polymorphic_call_targets): Likewise.
7736 * loop-unroll.c (report_unroll): Convert "locus" param from
7737 location_t to dump_location_t.
7738 (decide_unrolling): Update for change to get_loop_location's
7740 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
7741 location_t to dump_user_location_t.
7742 (grid_find_single_omp_among_assignments_1): Updates calls to
7743 dump_printf_loc to pass in a dump_location_t rather than a
7744 location_t, via the gimple stmt.
7745 (grid_parallel_clauses_gridifiable): Convert "tloc" from
7746 location_t to dump_location_t. Updates calls to dump_printf_loc
7747 to pass in a dump_location_t rather than a location_t, via the
7749 (grid_inner_loop_gridifiable_p): Likewise.
7750 (grid_dist_follows_simple_pattern): Likewise.
7751 (grid_gfor_follows_tiling_pattern): Likewise.
7752 (grid_target_follows_gridifiable_pattern): Likewise.
7753 (grid_attempt_target_gridification): Convert initialization
7754 of local "grid" from memset to zero-initialization; FIXME: does
7755 this require C++11? Update call to dump_printf_loc to pass in a
7756 optinfo_location rather than a location_t, via the gimple stmt.
7757 * profile.c (read_profile_edge_counts): Updates call to
7758 dump_printf_loc to pass in a dump_location_t rather than a
7760 (compute_branch_probabilities): Likewise.
7761 * selftest-run-tests.c (selftest::run_tests): Call
7763 * selftest.h (dumpfile_c_tests): New decl.
7764 * tree-loop-distribution.c (pass_loop_distribution::execute):
7765 Update for change in return type of find_loop_location.
7766 * tree-parloops.c (parallelize_loops): Likewise.
7767 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
7768 "locus" from location_t to dump_user_location_t.
7769 (canonicalize_loop_induction_variables): Likewise.
7770 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
7771 for change in return type of find_loop_location.
7772 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
7773 to dump_printf_loc to pass in a dump_location_t rather than a
7774 location_t, via the stmt.
7775 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
7777 * tree-vect-loop-manip.c (find_loop_location): Convert return
7778 type from source_location to dump_user_location_t.
7779 (vect_do_peeling): Update for above change.
7780 (vect_loop_versioning): Update for change in type of
7782 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
7783 from location_t to dump_user_location_t.
7784 (vect_estimate_min_profitable_iters): Update for change in type
7786 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
7787 location_t to dump_location_t.
7788 (vect_slp_bb): Update for change in type of vect_location.
7789 * tree-vectorizer.c (vect_location): Convert from source_location
7790 to dump_user_location_t.
7791 (try_vectorize_loop_1): Update for change in vect_location's type.
7792 (vectorize_loops): Likewise.
7793 (increase_alignment): Likewise.
7794 * tree-vectorizer.h (vect_location): Convert from source_location
7795 to dump_user_location_t.
7796 (find_loop_location): Convert return type from source_location to
7797 dump_user_location_t.
7798 (check_reduction_path): Convert 1st param from location_t to
7799 dump_user_location_t.
7800 * value-prof.c (check_counter): Update call to dump_printf_loc to
7801 pass in a dump_user_location_t rather than a location_t; update
7802 call to error_at for change in type of "locus".
7803 (check_ic_target): Update call to dump_printf_loc to
7804 pass in a dump_user_location_t rather than a location_t, via the
7807 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
7809 * config/s390/s390.h (enum processor_flags): Do not use
7810 default tune parameter when -march was specified.
7812 2018-06-26 Jakub Jelinek <jakub@redhat.com>
7815 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
7816 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
7819 2018-06-26 Richard Biener <rguenther@suse.de>
7821 PR tree-optimization/86287
7823 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
7824 (vect_analyze_loop): Initialize n_stmts.
7826 2018-06-26 Richard Biener <rguenther@suse.de>
7829 * fold-const.c (fold_convertible_p): Pointer extension
7832 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
7835 * dwarf2out.c (loc_list_has_views): Adjust comments.
7836 (dw_loc_list): Split single cross-partition range with
7839 2018-06-25 Jeff Law <law@redhat.com>
7841 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
7842 on -mbig-switch by default.
7844 * config/v850/predicates.md (const_float_1_operand): Fix match_code
7846 (const_float_0_operand): Remove unused predicate.
7847 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
7848 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
7849 (recipsf2): New expander. Original pattern now called
7851 (recipdf2, recipdf2_insn): Similarly.
7852 (rsqrtsf2, rsqrtsf2_insn): Similarly
7853 (rsqrtdf2, rsqrtdf2_insn): Similarly
7855 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
7857 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
7858 Simplify logic for FreeBSD (twice).
7860 2018-06-25 Martin Sebor <msebor@redhat.com>
7862 PR tree-optimization/86204
7863 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
7864 a strnlen result if it's less than the length of the string.
7866 2018-06-25 Martin Sebor <msebor@redhat.com>
7868 PR tree-optimization/85700
7869 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
7870 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
7871 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
7873 2018-06-25 Martin Sebor <msebor@redhat.com>
7875 * doc/extend.texi (Zero-length arrays): Update and clarify.
7877 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
7879 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
7880 added IEEE/IBM long double multilib support on PowerPC little
7881 endian Linux systems.
7882 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
7883 (MULTILIB_DEFAULTS): Likewise.
7884 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7886 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
7887 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
7888 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
7890 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
7893 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
7894 (REORDER_45): Likewise.
7896 2018-06-25 Jeff Law <law@redhat.com>
7898 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
7899 dividend to 32 bits. Adjust length.
7900 (udivmodhi4): Cleanup output template. Fix length.
7902 2018-06-25 Carl Love <cel@us.ibm.com>
7904 * config/rs6000/vsx.md: Change word selector to prefered location.
7906 2018-06-25 Richard Biener <rguenther@suse.de>
7908 PR tree-optimization/86304
7909 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
7910 epilogue-if-converted loops as well.
7912 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
7914 * lto-section-out.c (lto_begin_section): Do not print section
7915 name for noaddr and unnumbered dumps.
7917 2018-06-25 Richard Biener <rguenther@suse.de>
7919 * tree-vectorizer.h (struct vec_info_shared): New structure
7920 with parts split out from struct vec_info and loop_nest from
7921 struct _loop_vec_info.
7922 (struct vec_info): Adjust accordingly.
7923 (struct _loop_vec_info): Likewise.
7924 (LOOP_VINFO_LOOP_NEST): Adjust.
7925 (LOOP_VINFO_DATAREFS): Likewise.
7926 (LOOP_VINFO_DDRS): Likewise.
7927 (struct _bb_vec_info): Likewise.
7928 (BB_VINFO_DATAREFS): Likewise.
7929 (BB_VINFO_DDRS): Likewise.
7930 (struct _stmt_vec_info): Add dr_aux member.
7931 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
7932 (DR_MISALIGNMENT_UNINITIALIZED): New.
7933 (set_dr_misalignment): Adjust.
7934 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
7935 (vect_analyze_loop): Adjust prototype.
7936 (vect_analyze_loop_form): Likewise.
7937 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
7938 Compute dependences lazily.
7939 (vect_record_base_alignments): Use shared datarefs/ddrs.
7940 (vect_verify_datarefs_alignment): Likewise.
7941 (vect_analyze_data_refs_alignment): Likewise.
7942 (vect_analyze_data_ref_accesses): Likewise.
7943 (vect_analyze_data_refs): Likewise.
7944 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
7945 constructor parameter for shared part.
7946 (vect_analyze_loop_form): Pass in shared part and adjust.
7947 (vect_analyze_loop_2): Pass in storage for the number of
7948 stmts. Move loop nest finding to the caller. Compute
7950 (vect_analyze_loop): Pass in shared part.
7951 (vect_transform_loop): Verify shared datarefs are unchanged.
7952 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
7953 constructor parameter for shared part.
7954 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
7955 (vect_slp_bb): Verify shared datarefs are unchanged before
7957 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
7959 (new_stmt_vec_info): Initialize DR_AUX misalignment to
7960 DR_MISALIGNMENT_UNINITIALIZED.
7961 * tree-vectorizer.c (vec_info::vec_info): Add constructor
7962 parameter for shared part.
7963 (vec_info::~vec_info): Adjust.
7964 (vec_info_shared::vec_info_shared): New.
7965 (vec_info_shared::~vec_info_shared): Likewise.
7966 (vec_info_shared::save_datarefs): Likewise.
7967 (vec_info_shared::check_datarefs): Likewise.
7968 (try_vectorize_loop_1): Construct shared part live for analyses
7969 of a single loop for multiple vector sizes.
7970 * tree-parloops.c (gather_scalar_reductions): Adjust.
7972 2018-06-25 Richard Biener <rguenther@suse.de>
7974 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
7975 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
7976 (vect_analyze_data_refs): Remove similar code from here and
7977 simplify accordingly.
7979 2018-06-25 Richard Biener <rguenther@suse.de>
7981 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
7982 for reverse storage order accesses rather than asserting
7983 they cannot happen here.
7985 2018-06-25 Tom de Vries <tdevries@suse.de>
7988 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
7989 Use data16 instead of .byte for insn prefix.
7991 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
7994 * parser.c (make_char_string_pack): Pass this literal chars
7995 through cpp_interpret_string.
7996 (cp_parser_userdef_numeric_literal): Check the result of
7997 make_char_string_pack.
7999 2018-06-24 Maya Rashish <coypu@sdf.org>
8001 * ginclude/stddef.h: Simplify conditions around avoiding
8002 re-definition of __size_t.
8004 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8006 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
8007 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
8009 2018-06-22 Maya Rashish <coypu@sdf.org>
8011 * doc/invoke.texi (mno-fancy-math-387): Update for changes
8012 made to OpenBSD and NetBSD through the years.
8014 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8016 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8017 behavior of vec_pack (vector double, vector double) to match
8018 behavior of vec_float2 (vector double, vector double).
8020 2018-06-22 Olivier Hainque <hainque@adacore.com>
8022 * gimplify.c (gimplify_function_tree): Prevent creation
8023 of a trampoline for the address of the current function
8024 passed to entry/exit instrumentation hooks.
8026 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
8029 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
8032 2018-06-22 Martin Liska <mliska@suse.cz>
8034 PR tree-optimization/86263
8035 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8036 Bail out if is_enabled is false.
8037 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
8039 (jump_table_cluster::is_enabled): New function.
8041 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8043 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
8044 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
8045 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
8046 (lto_input_ts_binfo_tree_pointers): Likewise.
8047 * tree-streamer-out.c (streamer_write_tree_bitfields,
8048 write_ts_binfo_tree_pointers): Likewise.
8049 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
8051 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8053 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
8055 2018-06-22 Martin Liska <mliska@suse.cz>
8057 * symbol-summary.h (get): Make it pure and inline move
8058 functionality from ::get function.
8059 (get): Remove and inline into ::get and ::get_create.
8060 (get_create): Move code from ::get function.
8062 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8065 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
8066 -x assembler-with-cpp.
8068 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8070 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
8071 _FILE_OFFSET_BITS=64 for C++.
8073 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
8075 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
8076 conversion insn that shows up when pr85657-3.c is compiled using
8077 IEEE 128-bit long double.
8078 (neg<mode>2_internal): Use the correct mode to check whether the
8079 mode is IBM extended.
8080 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
8081 multiply and divide external functions from being created more
8084 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8086 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
8088 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
8089 the edge can be forwarded.
8090 (cfg_layout_merge_blocks): Likewise.
8092 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8094 * except.c (finish_eh_generation): Commit edge insertions only after
8095 the EH edges have been redirected from post-landing to landing pads.
8097 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8099 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
8100 create_tmp_var_for to create the FRAME decl.
8101 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
8103 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8105 * tree-inline.c (copy_edges_for_bb): Minor tweak.
8106 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
8107 debug statement when resetting its value.
8108 (expand_call_inline): Copy the locus of the call onto the assignment
8109 of the return value, if any. Use local variable in more cases.
8111 2018-06-21 Martin Liska <mliska@suse.cz>
8113 * ipa-pure-const.c (propagate_nothrow): Use
8114 funct_state_summaries->get.
8115 (dump_malloc_lattice): Likewise.
8116 (propagate_malloc): Likewise.
8118 2018-06-21 Richard Biener <rguenther@suse.de>
8120 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
8121 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
8122 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
8124 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
8125 BLOCK_ABSTRACT_ORIGIN unconditionally.
8127 2018-06-21 David Malcolm <dmalcolm@redhat.com>
8129 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
8131 * ipa-reference.c (ipa_reference_c_finalize): Delete
8132 ipa_ref_opt_sum_summaries and set it to NULL.
8134 2018-06-21 Tom de Vries <tdevries@suse.de>
8136 PR tree-optimization/85859
8137 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
8138 test with comment from bb_no_side_effects_p.
8140 2018-06-21 Richard Biener <rguenther@suse.de>
8142 PR tree-optimization/86232
8143 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
8144 max for constant niter.
8146 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
8148 * config/aarch64/aarch64-simd.md
8149 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
8151 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
8153 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
8154 Make opernads of the unspec commutative.
8156 2018-06-21 Richard Biener <rguenther@suse.de>
8158 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
8159 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
8160 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
8161 (vect_analyze_data_ref_dependence): Re-order checks to deal with
8163 (vect_record_base_alignments): Do not record base alignment
8164 for gathers or scatters.
8165 (vect_compute_data_ref_alignment): Drop return value that is always
8166 true. Bail out early for gathers or scatters.
8167 (vect_enhance_data_refs_alignment): Bail out early for gathers
8169 (vect_find_same_alignment_drs): Likewise.
8170 (vect_analyze_data_refs_alignment): Remove dead code.
8171 (vect_slp_analyze_and_verify_node_alignment): Likewise.
8172 (vect_analyze_data_refs): For possible gathers or scatters do
8173 not create an alternate DR, just check their possible validity
8174 and mark them. Adjust DECL_NONALIASED handling to not rely
8176 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
8177 update inits of gathers or scatters.
8178 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
8179 Also copy gather/scatter flag to pattern vinfo.
8181 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
8183 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8184 behavior of vec_packsu (vector unsigned long long, vector unsigned
8185 long long) to match behavior of vec_packs with same signature.
8187 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
8188 Thomas Schwinge <thomas@codesourcery.com>
8189 Cesar Philippidis <cesar@codesourcery.com>
8191 * gimplify.c (gimplify_scan_omp_clauses): Add support for
8192 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8193 (gimplify_adjust_omp_clauses): Likewise.
8194 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
8195 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
8196 (gimplify_omp_target_update): Update handling of acc update and
8198 * omp-low.c (install_var_field): Remove unused parameter
8199 base_pointers_restrict.
8200 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
8201 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
8203 (omp_target_base_pointers_restrict_p): Delete.
8204 (scan_omp_target): Update call to scan_sharing_clauses.
8205 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
8207 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
8208 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8209 (convert_local_omp_clauses): Likewise.
8210 * tree-pretty-print.c (dump_omp_clause): Likewise.
8211 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
8213 (omp_clause_code_name): Likewise.
8215 2018-06-20 Jakub Jelinek <jakub@redhat.com>
8218 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
8221 PR tree-optimization/86231
8222 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
8223 anti-range don't overwrite *vr0min before using it to compute *vr0max.
8225 2018-06-20 Tom de Vries <tdevries@suse.de>
8227 PR tree-optimization/86097
8228 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
8229 iv type if signedness of iv type is not the same as that of *nit.
8231 2018-06-20 Jakub Jelinek <jakub@redhat.com>
8233 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
8234 EDGE_EH edges, verify they are all EDGE_EH.
8236 2018-06-20 Maya Rashish <coypu@sdf.org>
8238 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
8240 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8242 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
8243 * config/aarch64/aarch64.c (xgene1_tunings): Add
8244 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
8245 (aarch64_mode_valid_for_sched_fusion_p):
8246 Allow 16-byte modes.
8247 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
8248 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
8250 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
8252 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
8253 * config/aarch64/iterators.md (VQ2): New mode iterator.
8255 2018-06-20 Martin Liska <mliska@suse.cz>
8257 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
8258 Change default ratio from 10 to 8.
8260 2018-06-20 Martin Liska <mliska@suse.cz>
8262 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8264 (bit_test_cluster::find_bit_tests): Likewise.
8265 (switch_decision_tree::analyze_switch_statement): Find clusters.
8266 * tree-switch-conversion.h (struct jump_table_cluster): Document
8269 2018-06-20 Martin Liska <mliska@suse.cz>
8271 * tree-switch-conversion.c (switch_conversion::collect):
8272 Record m_uniq property.
8273 (switch_conversion::expand): Bail out for special conditions.
8274 (group_cluster::~group_cluster): New.
8275 (group_cluster::group_cluster): Likewise.
8276 (group_cluster::dump): Likewise.
8277 (jump_table_cluster::emit): New.
8278 (switch_decision_tree::fix_phi_operands_for_edges): New.
8279 (struct case_node): Remove struct.
8280 (jump_table_cluster::can_be_handled): New.
8281 (case_values_threshold): Moved to header.
8282 (reset_out_edges_aux): Likewise.
8283 (jump_table_cluster::is_beneficial): New.
8284 (bit_test_cluster::can_be_handled): Likewise.
8285 (add_case_node): Remove.
8286 (bit_test_cluster::is_beneficial): New.
8287 (case_bit_test::cmp): New.
8288 (bit_test_cluster::emit): New.
8289 (expand_switch_as_decision_tree_p): Remove.
8290 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
8291 (fix_phi_operands_for_edge): Likewise.
8292 (switch_decision_tree::analyze_switch_statement): New.
8293 (compute_cases_per_edge): Move ...
8294 (switch_decision_tree::compute_cases_per_edge): ... here.
8295 (try_switch_expansion): Likewise.
8296 (switch_decision_tree::try_switch_expansion): Likewise.
8297 (record_phi_operand_mapping): Likewise.
8298 (switch_decision_tree::record_phi_operand_mapping): Likewise.
8299 (emit_case_decision_tree): Likewise.
8300 (switch_decision_tree::emit): Likewise.
8301 (balance_case_nodes): Likewise.
8302 (switch_decision_tree::balance_case_nodes): Likewise.
8303 (dump_case_nodes): Likewise.
8304 (switch_decision_tree::dump_case_nodes): Likewise.
8305 (emit_jump): Likewise.
8306 (switch_decision_tree::emit_jump): Likewise.
8307 (emit_cmp_and_jump_insns): Likewise.
8308 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
8309 (emit_case_nodes): Likewise.
8310 (switch_decision_tree::emit_case_nodes): Likewise.
8311 (conditional_probability): Remove.
8312 * tree-switch-conversion.h (enum cluster_type): New.
8314 (struct cluster): Likewise.
8315 (cluster::cluster): Likewise.
8316 (struct simple_cluster): Likewise.
8317 (simple_cluster::simple_cluster): Likewise.
8318 (struct group_cluster): Likewise.
8319 (struct jump_table_cluster): Likewise.
8320 (struct bit_test_cluster): Likewise.
8321 (struct min_cluster_item): Likewise.
8322 (struct case_tree_node): Likewise.
8323 (case_tree_node::case_tree_node): Likewise.
8324 (jump_table_cluster::case_values_threshold): Likewise.
8325 (struct case_bit_test): Likewise.
8326 (struct switch_decision_tree): Likewise.
8327 (struct switch_conversion): Likewise.
8328 (switch_decision_tree::reset_out_edges_aux): Likewise.
8330 2018-06-20 Martin Liska <mliska@suse.cz>
8332 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
8333 (hoist_edge_and_branch_if_true): Likewise.
8334 (expand_switch_using_bit_tests_p): Likewise.
8335 (struct case_bit_test): Likewise.
8336 (case_bit_test_cmp): Likewise.
8337 (emit_case_bit_tests): Likewise.
8338 (switch_conversion::switch_conversion): New class.
8339 (struct switch_conv_info): Remove old struct.
8340 (collect_switch_conv_info): More to ...
8341 (switch_conversion::collect): ... this.
8342 (check_range): Likewise.
8343 (switch_conversion::check_range): Likewise.
8344 (check_all_empty_except_final): Likewise.
8345 (switch_conversion::check_all_empty_except_final): Likewise.
8346 (check_final_bb): Likewise.
8347 (switch_conversion::check_final_bb): Likewise.
8348 (create_temp_arrays): Likewise.
8349 (switch_conversion::create_temp_arrays): Likewise.
8350 (free_temp_arrays): Likewise.
8351 (gather_default_values): Likewise.
8352 (switch_conversion::gather_default_values): Likewise.
8353 (build_constructors): Likewise.
8354 (switch_conversion::build_constructors): Likewise.
8355 (constructor_contains_same_values_p): Likewise.
8356 (switch_conversion::contains_same_values_p): Likewise.
8357 (array_value_type): Likewise.
8358 (switch_conversion::array_value_type): Likewise.
8359 (build_one_array): Likewise.
8360 (switch_conversion::build_one_array): Likewise.
8361 (build_arrays): Likewise.
8362 (switch_conversion::build_arrays): Likewise.
8363 (gen_def_assigns): Likewise.
8364 (switch_conversion::gen_def_assigns): Likewise.
8365 (prune_bbs): Likewise.
8366 (switch_conversion::prune_bbs): Likewise.
8367 (fix_phi_nodes): Likewise.
8368 (switch_conversion::fix_phi_nodes): Likewise.
8369 (gen_inbound_check): Likewise.
8370 (switch_conversion::gen_inbound_check): Likewise.
8371 (process_switch): Use the newly created class.
8372 (switch_conversion::expand): New.
8373 (switch_conversion::~switch_conversion): New.
8374 * tree-switch-conversion.h: New file.
8376 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8378 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
8379 tree-vect-patterns.c.
8380 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
8381 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
8382 (vect_recog_sad_pattern): Likewise.
8383 (vect_recog_widen_sum_pattern): Likewise.
8384 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
8385 (vect_recog_widen_shift_pattern): Remove the type_in argument.
8386 (vect_recog_rotate_pattern): Likewise.
8387 (vect_recog_mult_pattern): Likewise.
8388 (vect_recog_vector_vector_shift_pattern): Likewise.
8389 (vect_recog_divmod_pattern): Likewise.
8390 (vect_recog_mixed_size_cond_pattern): Likewise.
8391 (vect_recog_bool_pattern): Likewise.
8392 (vect_recog_mask_conversion_pattern): Likewise.
8393 (vect_try_gather_scatter_pattern): Likewise.
8394 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
8395 (vect_recog_over_widening_pattern): Likewise.
8396 (vect_recog_gather_scatter_pattern): Likewise.
8397 (vect_recog_func_ptr): Move from tree-vectorizer.h
8398 (vect_vect_recog_func_ptrs): Move further down the file.
8399 (vect_recog_func): Likewise. Remove the third argument.
8400 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
8401 (vect_pattern_recog_1): Expect the pattern function to do any
8402 necessary target tests. Also expect it to provide a vector type.
8403 Remove the type_in handling.
8405 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8407 * tree-vect-patterns.c (vect_pattern_detected): New function.
8408 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
8409 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
8410 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
8411 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
8412 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
8413 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
8414 (vect_recog_mask_conversion_pattern)
8415 (vect_try_gather_scatter_pattern): Likewise.
8417 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8419 * tree-vect-patterns.c (vect_get_internal_def): New function.
8420 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
8421 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
8422 (search_type_for_mask_1): Use it.
8424 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8426 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
8427 redundant WIDEN_SUM_EXPR handling.
8428 (vect_recog_sad_pattern): Likewise.
8430 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8432 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
8433 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
8434 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
8435 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
8436 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
8438 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8440 * tree-vect-stmts.c (vectorizable_call): Make sure that we
8441 use the stmt_vec_info of the original bb statement for the
8442 new zero assignment, even if the call is part of a pattern.
8444 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8446 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
8447 that the sequence is attached to the original statement rather
8448 than the pattern statement.
8449 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
8450 PATTERN_DEF_SEQ from the original statement rather than
8451 the main pattern statement.
8452 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
8453 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
8454 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
8456 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8458 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
8459 definition statements before the early exit for statements that aren't
8461 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
8463 (vect_transform_loop): ...here. Process pattern definition
8464 statements without first checking whether the main pattern
8465 statement is live or relevant.
8467 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
8469 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
8470 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
8472 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
8474 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
8475 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
8476 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
8477 (expand_block_compare): Change select_block_compare_mode call.
8478 (expand_strncmp_align_check): Use new functions, fix comment.
8479 (emit_final_str_compare_gpr): New function.
8480 (expand_strn_compare): Refactor and clean up code.
8481 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
8483 2018-06-19 Tony Reix <tony.reix@atos.com>
8484 Damien Bergamini <damien.bergamini@atos.com>
8485 David Edelsohn <dje.gcc@gmail.com>
8487 * collect2.c (static_obj): New variable.
8488 (static_libs): New variable.
8489 (is_in_list): Uncomment declaration.
8490 (main): Track AIX libraries linked statically.
8491 (is_in_list): Uncomment definition.
8492 (scan_prog_file): Don't add AIX shared libraries initializer
8493 to constructor list if linking statically.
8495 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
8497 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
8499 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
8501 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
8503 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
8506 2018-06-19 Martin Liska <mliska@suse.cz>
8508 * config/i386/i386.c (ix86_can_inline_p): Do not use
8509 ipa_fn_summaries::get_create.
8510 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
8512 (devirtualization_time_bonus): Likewise.
8513 (ipcp_propagate_stage): Likewise.
8514 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
8515 (edge_set_predicate): Likewise.
8516 (evaluate_conditions_for_known_args): Likewise.
8517 (evaluate_properties_for_edge): Likewise.
8518 (ipa_call_summary::reset): Tranform to ...
8519 (ipa_call_summary::~ipa_call_summary): ... this.
8520 (ipa_fn_summary::reset): Transform to ...
8521 (ipa_fn_summary::~ipa_fn_summary): ... this.
8522 (ipa_fn_summary_t::remove): Rename to ...
8523 (ipa_fn_summary_t::remove_callees): ... this.
8524 (ipa_fn_summary_t::duplicate): Use placement new
8525 instead of memory copy.
8526 (ipa_call_summary_t::duplicate): Likewise.
8527 (ipa_call_summary_t::remove): Remove.
8528 (dump_ipa_call_summary): Change get_create to get.
8529 (ipa_dump_fn_summary): Dump only when summary exists.
8530 (analyze_function_body): Use symbol_summary::get instead
8532 (compute_fn_summary): Likewise.
8533 (estimate_edge_devirt_benefit): Likewise.
8534 (estimate_edge_size_and_time): Likewise.
8535 (inline_update_callee_summaries): Likewise.
8536 (remap_edge_change_prob): Likewise.
8537 (remap_edge_summaries): Likewise.
8538 (ipa_merge_fn_summary_after_inlining): Likewise.
8539 (write_ipa_call_summary): Likewise.
8540 (ipa_fn_summary_write): Likewise.
8541 (ipa_free_fn_summary): Likewise.
8542 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
8543 (struct ipa_call_summary): Likewise.
8544 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
8546 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
8547 (estimate_size_after_inlining): Likewise.
8548 (estimate_growth): Likewise.
8549 (growth_likely_positive): Likewise.
8550 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
8551 (inline_call): Likewise.
8552 * ipa-inline.c (caller_growth_limits): Likewise.
8553 (can_inline_edge_p): Likewise.
8554 (can_inline_edge_by_limits_p): Likewise.
8555 (compute_uninlined_call_time): Likewise.
8556 (compute_inlined_call_time): Likewise.
8557 (want_inline_small_function_p): Likewise.
8558 (edge_badness): Likewise.
8559 (update_caller_keys): Likewise.
8560 (update_callee_keys): Likewise.
8561 (inline_small_functions): Likewise.
8562 (inline_to_all_callers_1): Likewise.
8563 (dump_overall_stats): Likewise.
8564 (early_inline_small_functions): Likewise.
8565 (early_inliner): Likewise.
8566 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
8567 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
8568 * ipa-pure-const.c (malloc_candidate_p): Likewise.
8569 * ipa-split.c (execute_split_functions): Likewise.
8570 * symbol-summary.h: Likewise.
8571 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
8573 2018-06-19 Richard Biener <rguenther@suse.de>
8575 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
8576 (vectorize_loops): ... here. Fix dbgcnt handling.
8577 (try_vectorize_loop): Wrap try_vectorize_loop_1.
8579 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
8582 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
8583 ieee128 argument takes up only one (vector) register, not two (floating
8586 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
8588 * gimplify.c (gimplify_init_constructor): Really never clear for an
8589 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
8591 2018-06-19 Richard Biener <rguenther@suse.de>
8593 PR tree-optimization/86179
8594 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
8595 after failed recognition.
8597 2018-06-18 Martin Sebor <msebor@redhat.com>
8600 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
8601 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
8602 Handle integer subtraction.
8603 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
8604 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
8606 2018-06-18 David Malcolm <dmalcolm@redhat.com>
8608 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
8609 param from rtx to rtx_insn *.
8610 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
8612 (frv_ifcvt_modify_insn): Likwise.
8613 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
8614 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
8615 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
8616 as_a <rtx_insn *> cast to local "unprotected_region" once
8617 it's been established that it's not NULL or pc_rtx.
8618 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
8619 param "sethi" from rtx to rtx_insn *.
8620 (nds32_group_float_insns): Likewise for param "insn".
8621 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
8623 (vax_output_int_subtract): Likewise.
8624 * config/vax/vax.c (vax_output_int_add): Likewise for param
8626 (vax_output_int_subtract): Likewise.
8627 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
8628 (emit_pattern_after): Likewise for param "after".
8629 (emit_insn_after): Likewise.
8630 (emit_jump_insn_after): Likewise.
8631 (emit_call_insn_after): Likewise.
8632 (emit_debug_insn_after): Likewise.
8633 (emit_pattern_before): Likewise for param "before".
8634 (emit_insn_before): Likewise.
8635 (emit_jump_insn_before): Likewise.
8636 * final.c (get_insn_template): Likewise for param "insn", removing
8638 * output.h (get_insn_template): Likewise for 2nd param.
8639 * rtl.h (emit_insn_before): Likewise.
8640 (emit_jump_insn_before): Likewise.
8641 (emit_debug_insn_before_noloc): Likewise.
8642 (emit_insn_after): Likewise.
8643 (emit_jump_insn_after): Likewise.
8644 (emit_call_insn_after): Likewise.
8645 (emit_debug_insn_after): Likewise.
8646 (set_insn_deleted): Likewise for param.
8648 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
8651 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
8652 floating point modes, so that IFmode is numerically greater than
8653 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
8654 to declare the ordering. This prevents IFmode from being
8655 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
8656 machine. Include rs6000-modes.h to share the fractional values
8657 between genmodes* and the rest of the compiler.
8661 * config/rs6000/rs6000-modes.h: New file.
8662 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
8663 meaning of rs6000_long_double_size so that 126..128 selects an
8664 appropriate 128-bit floating point type.
8665 (rs6000_option_override_internal): Likewise.
8666 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
8667 (TARGET_LONG_DOUBLE_128): Change the meaning of
8668 rs6000_long_double_size so that 126..128 selects an appropriate
8669 128-bit floating point type.
8670 (LONG_DOUBLE_TYPE_SIZE): Update comment.
8671 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
8672 source and destination to match the standard usage.
8673 (truncifkf2): Likewise.
8674 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
8675 ISA 2.07 to use an explicit clobber, instead of passing in a
8677 (copysign<mode>3_soft): Likewise.
8679 2018-06-18 David Malcolm <dmalcolm@redhat.com>
8681 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
8682 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
8683 (vect_slp_analyze_instance_dependence): Likewise.
8684 (vect_enhance_data_refs_alignment): Likewise.
8685 (vect_analyze_data_refs_alignment): Likewise.
8686 (vect_slp_analyze_and_verify_instance_alignment
8687 (vect_analyze_data_ref_accesses): Likewise.
8688 (vect_prune_runtime_alias_test_list): Likewise.
8689 (vect_analyze_data_refs): Likewise.
8690 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
8691 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8692 (vect_analyze_scalar_cycles_1): Likewise.
8693 (vect_get_loop_niters): Likewise.
8694 (vect_analyze_loop_form_1): Likewise.
8695 (vect_update_vf_for_slp): Likewise.
8696 (vect_analyze_loop_operations): Likewise.
8697 (vect_analyze_loop): Likewise.
8698 (vectorizable_induction): Likewise.
8699 (vect_transform_loop): Likewise.
8700 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
8701 * tree-vect-slp.c (vect_analyze_slp): Likewise.
8702 (vect_make_slp_decision): Likewise.
8703 (vect_detect_hybrid_slp): Likewise.
8704 (vect_slp_analyze_operations): Likewise.
8705 (vect_slp_bb): Likewise.
8706 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
8707 (vectorizable_bswap): Likewise.
8708 (vectorizable_call): Likewise.
8709 (vectorizable_simd_clone_call): Likewise.
8710 (vectorizable_conversion): Likewise.
8711 (vectorizable_assignment): Likewise.
8712 (vectorizable_shift): Likewise.
8713 (vectorizable_operation): Likewise.
8714 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
8716 2018-06-18 Martin Sebor <msebor@redhat.com>
8718 PR tree-optimization/81384
8719 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
8720 * builtins.c (expand_builtin_strnlen): New function.
8721 (expand_builtin): Call it.
8722 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
8723 * builtins.def (BUILT_IN_STRNLEN): New.
8724 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
8725 Warn for bounds in excess of maximum object size.
8726 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
8727 single-value ranges. Handle strnlen.
8728 (handle_builtin_strlen): Handle strnlen.
8729 (strlen_check_and_optimize_stmt): Same.
8730 * doc/extend.texi (Other Builtins): Document strnlen.
8732 2018-06-18 Maya Rashish <coypu@sdf.org>
8734 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
8735 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
8736 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
8738 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
8740 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
8742 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8744 * tree.c (escaped_string::escape): Replace cast to char * by
8745 const_cast<char *> (unescaped).
8747 2018-06-18 Nick Clifton <nickc@redhat.com>
8750 * tree.c (escaped_string): New class. Converts an unescaped
8751 string into its escaped equivalent.
8752 (warn_deprecated_use): Use the new class to convert the
8753 deprecation message, if present.
8754 (test_escaped_strings): New self test.
8755 (test_c_tests): Add test_escaped_strings.
8756 * doc/extend.texi (deprecated): Add a note that the
8757 deprecation message is affected by the -fmessage-length
8758 option, and that control characters will be escaped.
8759 (#pragma GCC error): Document this pragma.
8760 (#pragma GCC warning): Likewise.
8761 * doc/invoke.texi (-fmessage-length): Document this option's
8762 effect on the #warning and #error preprocessor directives and
8763 the deprecated attribute.
8765 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
8767 * tree.c (decl_value_expr_lookup): Revert latest change.
8768 (decl_value_expr_insert): Likewise.
8770 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
8772 * gimplify.c (nonlocal_vlas): Delete.
8773 (nonlocal_vla_vars): Likewise.
8774 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
8776 (gimplify_body): Do not create and destroy nonlocal_vlas.
8777 * tree-nested.c: Include diagnostic.h.
8778 (use_pointer_in_frame): Tweak.
8779 (lookup_field_for_decl): Add assertion and declare the transformation.
8780 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
8781 internal error when the reference is in a wrong context. Do not
8782 create a debug decl by default.
8783 (note_nonlocal_block_vlas): Delete.
8784 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
8785 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
8786 create a debug decl by default.
8787 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
8788 call is in a wrong context.
8789 (fixup_vla_decls): New function.
8790 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
8791 debug variables were created.
8792 * tree.c (decl_value_expr_lookup): Add checking assertion.
8793 (decl_value_expr_insert): Likewise.
8795 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
8798 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
8799 * tree-scalar-evolution.c (interpret_expr): Likewise.
8800 (expression_expensive_p): Likewise.
8801 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
8802 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
8803 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
8804 (ssa_defined_by_minus_one_stmt_p): New.
8806 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
8809 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
8810 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
8811 * dojump.c (do_jump): Likewise.
8812 * expr.c (expand_expr_real_2): Check operand type's sign.
8813 * fold-const.c (const_unop): Handle ABSU_EXPR.
8814 (fold_abs_const): Likewise.
8815 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
8816 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
8817 (strip_sign_op_1): Likesise.
8818 * match.pd: Add new pattern to generate ABSU_EXPR.
8819 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
8820 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
8821 * tree-eh.c (operation_could_trap_helper_p): Likewise.
8822 * tree-inline.c (estimate_operator_cost): Likewise.
8823 * tree-pretty-print.c (dump_generic_node): Likewise.
8824 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
8825 * tree.def (ABSU_EXPR): New.
8827 2018-06-16 Jakub Jelinek <jakub@redhat.com>
8830 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
8831 documented as preserved for backward compatibility only.
8832 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
8834 PR rtl-optimization/86108
8835 * bb-reorder.c (create_forwarder_block): Renamed to ...
8836 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
8837 jump from new landing pad to the second part.
8838 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
8841 2018-06-15 Jakub Jelinek <jakub@redhat.com>
8844 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
8845 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
8846 Only call store_expr for halves if the mode is the same.
8849 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
8850 Fix up comment formatting.
8852 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
8854 * typed-splay-tree.h (typed_splay_tree::remove): New function.
8855 (typed_splay_tree::closure,
8856 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
8857 (typed_splay_tree::typed_splay_tree,
8858 typed_splay_tree::operator =): Declared private.
8859 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
8860 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
8861 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
8862 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
8863 typed_splay_tree::splay_tree_splay,
8864 typed_splay_tree::splay_tree_foreach_helper,
8865 typed_splay_tree::splay_tree_insert,
8866 typed_splay_tree::splay_tree_remove,
8867 typed_splay_tree::splay_tree_lookup,
8868 typed_splay_tree::splay_tree_predecessor,
8869 typed_splay_tree::splay_tree_successor,
8870 typed_splay_tree::splay_tree_min,
8871 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
8872 (typed_splay_tree::root, typed_splay_tree::comp,
8873 typed_splay_tree::delete_key,
8874 typed_splay_tree::delete_value): New data members.
8875 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
8876 typed_splay_tree::remove.
8878 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
8880 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
8881 -mginv and -mno-ginv to the assembler.
8882 * config/mips/mips.opt (-mcrc): New option.
8884 * doc/invoke.text (-mcrc): Document.
8887 2018-06-15 Nick Clifton <nickc@redhat.com>
8890 * tree.c (escaped_string): New class. Converts an unescaped
8891 string into its escaped equivalent.
8892 (warn_deprecated_use): Use the new class to convert the
8893 deprecation message, if present.
8894 (test_escaped_strings): New self test.
8895 (test_c_tests): Add test_escaped_strings.
8896 * doc/extend.texi (deprecated): Add a note that the
8897 deprecation message is affected by the -fmessage-length
8898 option, and that control characters will be escaped.
8899 (#pragma GCC error): Document this pragma.
8900 (#pragma GCC warning): Likewise.
8901 * doc/invoke.texi (-fmessage-length): Document this option's
8902 effect on the #warning and #error preprocessor directives and
8903 the deprecated attribute.
8905 2018-06-15 Richard Biener <rguenther@suse.de>
8907 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
8908 here, also noting vector size used.
8909 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
8910 size used in MSG_OPTIMIZED_LOCATIONS dump.
8911 (pass_slp_vectorize::execute): Adjust.
8913 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
8916 * config/arc/arc.c (arc_return_address_register): Fix
8919 2018-06-15 Richard Biener <rguenther@suse.de>
8922 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
8923 leave useless conversion stripping to force_gimple_operand_gsi.
8924 (gimplify_build2): Likewise.
8925 (gimplify_build1): Likewise.
8927 2018-06-15 Richard Biener <rguenther@suse.de>
8930 * tree-cfg.c (move_stmt_op): unshare invariant addresses
8931 before adjusting their block.
8933 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
8935 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
8936 multilibs for *-*-rtems*.
8937 * config/riscv/t-rtems: New file.
8939 2018-06-14 Jakub Jelinek <jakub@redhat.com>
8942 * match.pd ((A +- CST1) +- CST2): Punt if last resort
8943 unsigned_type_for returns NULL.
8946 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
8947 subregs of multi-word pseudos unless the float mode has word size.
8949 2018-06-14 Richard Biener <rguenther@suse.de>
8952 * tree-vect-generic.c (build_word_mode_vector_type): Remove
8953 duplicate and harmful type_hash_canon.
8954 * tree.c (type_hash_canon): Assert we didn't find ourselves.
8956 2018-06-14 Richard Biener <rguenther@suse.de>
8959 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
8962 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
8964 * config/rtems.h (STDINT_LONG32): Define.
8966 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
8967 Prachi Godbole <prachi.godbole@imgtec.com>
8969 * config/mips/mips-cpus.def: Define P6600.
8970 * config/mips/mips-tables.opt: Regenerate.
8971 * config/mips/mips.c (mips_ucbranch_type): New enum.
8972 (mips_rtx_cost_data): Add support for P6600.
8973 (mips_issue_rate): Likewise.
8974 (mips_multipass_dfa_lookahead): Likewise.
8975 (mips_avoid_hazard): Likewise.
8976 (mips_reorg_process_insns): Likewise.
8977 (mips_classify_branch_p6600): New function.
8978 * config/mips/mips.h (TUNE_P6600): New define.
8979 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
8980 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
8981 * config/mips/mips.md: Include p6600.md.
8982 (processor): Add p6600.
8983 * config/mips/p6600.md: New file.
8984 * doc/invoke.texi: Add p6600 to supported architectures.
8986 2018-06-13 Martin Sebor <msebor@redhat.com>
8988 PR tree-optimization/86114
8989 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
8991 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
8993 2018-06-13 Richard Biener <rguenther@suse.de>
8995 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
8996 Properly set vector type of the intermediate stmt.
8997 * tree-vect-stmts.c (vectorizable_operation): The destination
8998 var always has vectype_out type.
9000 2018-06-13 Jeff Law <law@redhat.com>
9002 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
9003 integer 0 for argument to print_rtl_with_bb.
9004 (rl78_reorg): Likewise.
9006 2018-06-13 David Malcolm <dmalcolm@redhat.com>
9008 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
9009 from rtx to rtx_insn *.
9010 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
9012 (add_sched_insns_for_speculation): Likewise for local "target",
9013 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9014 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
9015 from rtx_insn ** to rtx_code_label **.
9016 (reorg_emit_nops): Likewise.
9017 (c6x_reorg): Likewise for local "call_labels".
9018 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
9020 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
9021 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
9022 the loops over LABEL_REFS.
9023 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
9025 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9026 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
9027 (split_branches): Strengthen local "olabel" from rtx to
9028 rtx_insn *, adding a safe_as_a cast.
9029 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
9031 (add_insn_after): Likewise for first two params.
9032 (add_insn_before): Likewise.
9033 (remove_insn): Likewise for param.
9034 (emit_pattern_before_noloc): Likewise for second and third params.
9035 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
9036 (emit_call_insn_before_noloc): Likewise.
9037 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
9039 (emit_barrier_before): Likewise.
9040 (emit_label_before): Strengthen "label" param from "rtx" to
9041 "rtx_code_label *". Strengthen "before" param from "rtx" to
9043 (emit_insn_after_1): Strengthen "after" param from "rtx" to
9045 (emit_pattern_after_noloc): Likewise.
9046 (emit_insn_after_noloc): Likewise.
9047 (emit_jump_insn_after_noloc): Likewise.
9048 (emit_call_insn_after_noloc): Likewise.
9049 (emit_debug_insn_after_noloc): Likewise.
9050 (emit_barrier_after): Likewise.
9051 (emit_label_after): Likewise for both params.
9052 (emit_pattern_after_setloc): Likewise for "after" param. Convert
9053 "loc" param from "int" to "location_t".
9054 (emit_insn_after_setloc): Likewise.
9055 (emit_jump_insn_after_setloc): Likewise.
9056 (emit_call_insn_after_setloc): Likewise.
9057 (emit_debug_insn_after_setloc): Likewise.
9058 (emit_pattern_before_setloc): Likewise for "before" param. Convert
9059 "loc" param from "int" to "location_t".
9060 (emit_pattern_before): Convert NULL_RTX to NULL.
9061 (emit_insn_before_setloc): Convert "loc" param from "int" to
9063 (emit_jump_insn_before_setloc): Likewise.
9064 (emit_call_insn_before_setloc): Likewise.
9065 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
9066 rtx_insn *. Convert "loc" param from "int" to "location_t".
9067 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
9068 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
9069 Convert 3rd param from "int" to "location_t".
9070 (emit_barrier_before, emit_barrier_after, next_real_insn):
9071 Strengthen param from rtx to rtx_insn *.
9072 (emit_label_before): Strengthen 1st param from "rtx" to
9073 "rtx_code_label *". Strengthen 2nd param from "rtx" to
9075 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
9076 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
9077 Strengthen 2nd param from "rtx" to "rtx_insn *".
9078 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
9079 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
9080 Likewise. Convert 3rd param from "int" to "location_t".
9081 (emit_label_after): Strengthen 1st param from "rtx" to
9083 (next_real_insn, remove_insn): Strengthen param from "rtx" to
9085 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
9086 from "rtx" to "rtx_insn *".
9088 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
9090 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
9091 bodies streamed in with -Q.
9092 * dumpfile.c (dump_files): Add lto-stream-out dump file.
9093 * dumpfile.h (tree_dump_index): Add lto_stream_out.
9094 * gimple-streamer-out.c: Include gimple-pretty-print.h
9095 (output_bb): Dump stmts streamed.
9096 * lto-section-out.c: Include print-tree.h
9097 (lto_begin_section): Dump sections created.
9098 (lto_output_decl_index): Dump decl encoded.
9099 * lto-streamer-out.c: Include print-tree.h
9100 (create_output_block): Dump output block created.
9101 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
9102 (output_function): Dump function output.
9103 (output_constructor): Dump constructor streamed.
9104 (write_global_stream): Output indexes encoded.
9105 (produce_asm_for_decls): Dump streams encoded.
9106 * lto-streamer.c (streamer_dump_file): New global var.
9107 * lto-streamer.h (streamer_dump_file): Declare.
9108 * passes.c (ipa_write_summaries): Initialize streamer dump.
9109 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
9112 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
9115 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
9116 offsets for register save directives. Emit a second batch of save
9117 directives, if need be, when the function accesses prior frames.
9119 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9121 * config/arc/fpu.md (fmasf4): Force operand to register.
9122 (fnmasf4): Likewise.
9124 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9126 * config/arc/arc-protos.h (arc_pad_return): Remove.
9127 * config/arc/arc.c (machine_function): Remove force_short_suffix
9129 (arc_print_operand): Adjust printing of '&'.
9130 (arc_verify_short): Remove conditional printing of short suffix.
9131 (arc_final_prescan_insn): Remove reference to size_reason.
9132 (pad_return): New function.
9133 (arc_reorg): Call pad_return.
9134 (arc_pad_return): Remove.
9135 (arc_init_machine_status): Remove reference to force_short_suffix.
9136 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
9137 (attr length): When attribute iscompact is true force to 2
9138 regardless; in the case of maybe check if we want to force the
9139 instruction to have 4 bytes length.
9140 (nopv): Change it to generate 4 byte long nop as well.
9141 (blockage): New pattern.
9142 (simple_return): Remove call to arc_pad_return.
9143 (p_return_i): Likewise.
9145 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9147 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
9149 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9151 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
9154 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9156 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
9157 for ARC700 and ARCv2.
9159 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
9162 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
9163 operands[2] instead of operands[1].
9166 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
9168 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
9169 case, check whether the outer register overlaps an unallocatable
9170 register, not just whether it fits the required class.
9172 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
9174 * poly-int.h (can_div_trunc_p): Add new overload in which all values
9176 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
9177 (memrefs_conflict_p): Likewise.
9178 (init_alias_analysis): Likewise.
9179 * cfgexpand.c (expand_debug_expr): Likewise.
9180 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
9181 * cse.c (fold_rtx): Likewise.
9182 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
9183 * expr.c (emit_block_move_hints): Likewise.
9184 (clear_storage_hints, push_block, emit_push_insn): Likewise.
9185 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
9186 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
9187 (emit_group_store): Likewise.
9188 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
9189 to read the PRE/POST_MODIFY increment.
9190 * calls.c (store_one_arg): Use strip_offset.
9191 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
9193 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
9195 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
9196 (simplify_binary_operation_1): Extend CONST_INT handling to
9198 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
9199 than a HOST_WIDE_INT.
9200 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
9202 (adjust_mems, add_stores): Update accodingly.
9203 (vt_canonicalize_addr): Track polynomial offsets.
9204 (emit_note_insn_var_location): Likewise.
9205 (vt_add_function_parameter): Likewise.
9206 (vt_initialize): Likewise.
9208 2018-06-12 Jeff Law <law@redhat.com>
9210 * config.gcc (alpha*-*-freebsd*): Remove.
9211 * config/alpha/freebsd.h: Remove.
9213 2018-06-12 David Malcolm <dmalcolm@redhat.com>
9216 * spellcheck-tree.c (levenshtein_distance): Rename to...
9217 (get_edit_distance): ...this, and update for underlying renaming.
9218 * spellcheck-tree.h (levenshtein_distance): Rename to...
9219 (get_edit_distance): ...this.
9220 * spellcheck.c (levenshtein_distance): Rename to...
9221 (get_edit_distance): ...this. Convert from Levenshtein distance
9222 to Damerau-Levenshtein distance by supporting transpositions of
9223 adjacent characters. Rename "v1" to "v_next" and "v0" to
9225 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
9226 (selftest::test_edit_distance_unit_test_oneway): ...this, and
9227 update for underlying renaming.
9228 (selftest::levenshtein_distance_unit_test): Rename to...
9229 (selftest::test_get_edit_distance_unit): ...this, and update for
9230 underlying renaming.
9231 (selftest::test_find_closest_string): Add example from PR 69968
9232 where transposition helps
9233 (selftest::test_metric_conditions): Update for renaming.
9234 (selftest::test_metric_conditions): Likewise.
9235 (selftest::spellcheck_c_tests): Likewise.
9236 * spellcheck.h (levenshtein_distance): Rename both overloads to...
9237 (get_edit_distance): ...this.
9238 (best_match::consider): Update for renaming.
9240 2018-06-12 Martin Sebor <msebor@redhat.com>
9242 PR tree-optimization/85259
9243 * builtins.c (compute_objsize): Handle constant offsets.
9244 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
9245 true iff a warning has been issued.
9246 * gimple.h (gimple_nonartificial_location): New function.
9247 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
9248 gimple_nonartificial_location and handle -Wno-system-headers.
9249 (handle_builtin_stxncpy): Same.
9251 2018-06-12 Martin Sebor <msebor@redhat.com>
9254 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
9256 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
9258 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
9259 BUILTIN_VEC_XST entries for pointer to double and long long.
9261 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
9264 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9266 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9269 2018-06-12 Martin Liska <mliska@suse.cz>
9271 * doc/options.texi: Document IntegerRange.
9273 2018-06-12 Martin Liska <mliska@suse.cz>
9275 * config/i386/i386.opt: Make MPX-related options as Deprecated.
9276 * opt-functions.awk: Handle Deprecated flag.
9277 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
9279 (read_cmdline_option): Report warning for a deprecated option.
9280 * opts.h (struct cl_option): Add new field cl_deprecated.
9281 (CL_ERR_DEPRECATED): New.
9283 2018-06-12 Martin Liska <mliska@suse.cz>
9285 * doc/options.texi: Document Deprecated option flag.
9287 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9289 * config/arc/arc-arch.h (arc_extras): New enum.
9290 (arc_cpu_t):Add field extra.
9291 (arc_cpu_types): Consider the extras.
9292 * config/arc/arc-cpus.def: Add extras info.
9293 * config/arc/arc-opts.h (processor_type): Consider extra field.
9294 * config/arc/arc.c (arc_override_options): Handle extra field.
9296 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9298 * config/arc/arc-arch.h: Update ARC_OPTX macro.
9299 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
9301 * config/arc/arc.c (arc_init): Update pic warning.
9302 (irq_range): Update irq range parsing warnings.
9303 (arc_override_options): Update various warning messages.
9304 (arc_handle_aux_attribute): Likewise.
9306 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
9308 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
9310 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9312 * doc/sourcebuild.texi: Document usage of line number 0 in verify
9313 compiler messages directives.
9315 2018-06-12 Matthew Fortune <mfortune@gmail.com>
9317 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
9318 * config/mips/mips-tables.opt: Regenerate.
9319 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
9321 * doc/invoke.texi: Document -march=i6500.
9323 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
9325 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
9326 (i6400_gpmul): Add cpu_unit.
9327 (i6400_gpdiv): Likewise.
9328 (i6400_msa_add_d): Update reservations.
9329 (i6400_msa_int_add) Likewise.
9330 (i6400_msa_short_logic3) Likewise.
9331 (i6400_msa_short_logic2) Likewise.
9332 (i6400_msa_short_logic) Likewise.
9333 (i6400_msa_move) Likewise.
9334 (i6400_msa_cmp) Likewise.
9335 (i6400_msa_short_float2) Likewise.
9336 (i6400_msa_div_d) Likewise.
9337 (i6400_msa_long_logic1) Likewise.
9338 (i6400_msa_long_logic2) Likewise.
9339 (i6400_msa_mult) Likewise.
9340 (i6400_msa_long_float2) Likewise.
9341 (i6400_msa_long_float4) Likewise.
9342 (i6400_msa_long_float5) Likewise.
9343 (i6400_msa_long_float8) Likewise.
9344 (i6400_fpu_fadd): Include frint type.
9345 (i6400_fpu_store): New define_insn_reservation.
9346 (i6400_fpu_load): Likewise.
9347 (i6400_fpu_move): Likewise.
9348 (i6400_fpu_fcmp): Likewise.
9349 (i6400_fpu_fmadd): Likewise.
9350 (i6400_int_mult): Include imul3nc type and update reservation.
9351 (i6400_int_div): Include idiv3 type and update reservation.
9352 (i6400_int_load): Update to check type not move_type.
9353 (i6400_int_store): Likewise.
9354 (i6400_int_prefetch): Set zero latency.
9356 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
9358 * gcc.c: Document new %@{...} sequence.
9359 (LINK_COMMAND_SPEC): Use it for the -L switches.
9360 (cpp_unique_options): Use it for the -I switches.
9361 (at_file_argbuf): New global variable.
9362 (in_at_file): Likewise.
9363 (alloc_args): Create at_file_argbuf.
9364 (clear_args): Truncate at_file_argbuf.
9365 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
9366 (open_at_file): New function.
9367 (close_at_file): Likewise.
9368 (create_at_file): Delete.
9369 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
9372 (validate_switches_from_spec): Deal with %@{...} sequence.
9373 (validate_switches): Likewise.
9374 (driver::finalize): Call clear_args.
9376 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
9378 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
9380 2018-06-11 Martin Sebor <msebor@redhat.com>
9382 * doc/invoke.texi (-Wall): List -Wc++17-compat.
9383 (Wno-class-memaccess): Add @opindex.
9384 (Wno-templates, Wno-multiple-inheritance): Same.
9385 (Wno-virtual-inheritance, Wno-namespaces): Same.
9386 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
9387 (Wno-format-overflow, Wno-format-truncation): Same.
9388 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
9389 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
9390 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
9391 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
9392 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
9393 (Wno-misspelled-isr): Same.
9395 2018-06-11 Martin Sebor <msebor@redhat.com>
9397 * PR tree-optimization/86083
9398 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
9400 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
9402 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
9404 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
9407 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
9408 on the correct operand.
9409 (*movdi_internal64): Ditto.
9411 2018-06-11 Martin Liska <mliska@suse.cz>
9413 PR tree-optimization/86089
9414 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
9416 2018-06-11 Julia Koval <julia.koval@intel.com>
9418 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
9419 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
9420 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
9422 2018-06-11 Olivier Hainque <hainque@adacore.com>
9424 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
9425 for Ada with strict dwarf2.
9427 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
9430 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
9433 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
9435 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
9437 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
9439 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
9442 2018-06-08 Martin Liska <mliska@suse.cz>
9444 * tree-cfg.h (debug_function): Fix argument type to match
9447 2018-06-08 Martin Liska <mliska@suse.cz>
9449 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
9450 Remove usage of MPX-related (and removed) fields.
9451 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
9453 2018-06-08 David Malcolm <dmalcolm@redhat.com>
9455 * cfg.c (debug): Use TDF_NONE rather than 0.
9456 * cfghooks.c (debug): Likewise.
9457 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
9458 (struct dump_option_value_info): Convert to...
9459 (struct kv_pair): ...this template type.
9460 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
9462 (optinfo_verbosity_options): Likewise.
9463 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
9465 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
9466 than int for "optgroup_flags" param.
9467 (dump_generic_expr_loc): Use dump_flags_t rather than int for
9469 (dump_dec): Likewise.
9470 (dump_finish): Use TDF_NONE rather than 0.
9471 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
9472 rather than int for "optgroup_flags" param. Use TDF_NONE rather
9473 than 0. Update for change to option_ptr.
9474 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
9475 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
9476 0. Update for changes to optinfo_verbosity_options and
9478 (opt_info_switch_p): Convert optgroup_flags from int to
9480 (dump_basic_block): Use dump_flags_t rather than int
9481 for "dump_kind" param.
9482 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
9483 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
9484 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
9485 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
9486 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
9487 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
9488 TDF_NONE): Convert from macros to...
9489 (enum dump_flag): ...this new enum.
9490 (dump_flags_t): Update to use enum.
9491 (operator|, operator&, operator~, operator|=, operator&=):
9492 Implement for dump_flags_t.
9493 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
9494 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
9495 Convert from macros to...
9496 (enum optgroup_flag): ...this new enum.
9497 (optgroup_flags_t): New typedef.
9498 (operator|, operator|=): Implement for optgroup_flags_t.
9499 (struct dump_file_info): Convert field "alt_flags" to
9500 dump_flags_t. Convert field "optgroup_flags" to
9502 (dump_basic_block): Use dump_flags_t rather than int for param.
9503 (dump_generic_expr_loc): Likewise.
9504 (dump_dec): Likewise.
9505 (dump_register): Convert param "optgroup_flags" to
9507 (opt_info_enable_passes): Likewise.
9508 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
9510 * gimple-pretty-print.c (debug): Likewise.
9511 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
9512 (merged_store_group::apply_stores): Likewise.
9513 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
9514 * gimple.c (verify_gimple_pp): Likewise.
9515 * graphite-poly.c (print_pbb_body): Likewise.
9516 * passes.c (pass_manager::register_one_dump_file): Convert
9517 local "optgroup_flags" to optgroup_flags_t.
9518 * print-tree.c (print_node): Use TDF_NONE rather than 0.
9520 (debug_body): Likewise.
9521 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
9522 to optgroup_flags_t.
9523 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
9525 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
9526 (convert_mult_to_fma): Likewise.
9527 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
9528 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
9529 * tree-vect-data-refs.c (dump_lower_bound): Convert param
9530 "dump_kind" to dump_flags_t.
9532 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
9534 * config/rs6000/rs6000.c (min, max): Delete.
9536 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
9538 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
9539 -mabi=spe and -mabi=no-spe.
9541 2018-06-08 Martin Liska <mliska@suse.cz>
9543 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
9544 where we expect an existing summary.
9546 2018-06-08 Martin Liska <mliska@suse.cz>
9548 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
9549 * ipa-inline.h (estimate_edge_growth): Likewise.
9551 2018-06-08 Martin Liska <mliska@suse.cz>
9553 * cgraph.c (function_version_hasher::hash): Use
9554 cgraph_node::get_uid ().
9555 (function_version_hasher::equal):
9556 * cgraph.h (cgraph_node::get_uid): New method.
9557 * ipa-inline.c (update_caller_keys): Use
9558 cgraph_node::get_uid ().
9559 (update_callee_keys): Likewise.
9560 * ipa-utils.c (searchc): Likewise.
9561 (ipa_reduced_postorder): Likewise.
9562 * lto-cgraph.c (input_node): Likewise.
9563 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
9564 * symbol-summary.h (symtab_insertion): Likewise.
9565 (symtab_removal): Likewise.
9566 (symtab_duplication): Likewise.
9567 * tree-pretty-print.c (dump_function_header): Likewise.
9568 * tree-sra.c (convert_callers_for_node): Likewise.
9570 2018-06-08 Martin Liska <mliska@suse.cz>
9572 * cgraph.c (symbol_table::create_edge): Always assign a new
9574 (symbol_table::free_edge): Do not recycle numbers.
9575 * cgraph.h (cgraph_edge::get): New method.
9576 * symbol-summary.h (symtab_removal): Use it.
9577 (symtab_duplication): Likewise.
9578 (call_summary::hashable_uid): Remove.
9580 2018-06-08 Martin Liska <mliska@suse.cz>
9582 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
9583 (initialize_growth_caches): Remove.
9584 (free_growth_caches): Likewise.
9585 (do_estimate_edge_time): Use edge_growth_cache.
9586 (do_estimate_edge_size): Likewise.
9587 (do_estimate_edge_hints): Likewise.
9588 * ipa-inline.c (reset_edge_caches): Likewise.
9589 (recursive_inlining): Likewise.
9590 (inline_small_functions): Likewise.
9591 * ipa-inline.h (initialize_growth_caches): Remove.
9592 (estimate_edge_size): Likewise.
9593 (estimate_edge_time): Likewise.
9594 (estimate_edge_hints): Likewise.
9595 (reset_edge_growth_cache): Likewise.
9596 * symbol-summary.h (call_summary::remove): New method.
9598 2018-06-08 Martin Liska <mliska@suse.cz>
9600 * ipa-cp.c (class edge_clone_summary): New summary.
9601 (grow_edge_clone_vectors): Remove.
9602 (ipcp_edge_duplication_hook): Remove.
9603 (class edge_clone_summary_t): New call_summary class.
9604 (ipcp_edge_removal_hook): Remove.
9605 (edge_clone_summary_t::duplicate): New function.
9606 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
9607 (create_specialized_node): Likewise.
9608 (ipcp_driver): Initialize edge_clone_summaries and do not
9611 2018-06-08 Martin Liska <mliska@suse.cz>
9613 * symbol-summary.h (get): New function.
9614 (call_summary::m_initialize_when_cloning): New class member.
9616 2018-06-08 Martin Liska <mliska@suse.cz>
9618 * cgraph.c (cgraph_node::remove): Do not recycle uid.
9619 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
9620 (symbol_table::allocate_cgraph_symbol): Do not set uid.
9621 * passes.c (uid_hash_t): Record removed_nodes by their uids.
9622 (remove_cgraph_node_from_order): Use the removed_nodes set.
9623 (do_per_function_toporder): Likwise.
9624 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
9625 instead of summary_uid.
9626 (symtab_removal): Likewise.
9627 (symtab_duplication): Likewise.
9629 2018-06-08 Martin Liska <mliska@suse.cz>
9631 * ipa-cp.c (ipcp_store_bits_results): Use
9632 ipcp_transformation_sum.
9633 (ipcp_store_vr_results): Likewise.
9634 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
9636 (ipcp_transformation_initialize): ... this.
9637 (ipa_set_node_agg_value_chain):
9638 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
9639 (write_ipcp_transformation_info): Likewise.
9640 (read_ipcp_transformation_info): Likewise.
9641 (ipcp_update_bits): Likewise.
9642 (ipcp_update_vr): Likewise.
9643 (ipcp_transform_function): Likewise.
9644 * ipa-prop.h: Rename ipcp_transformation_summary to
9645 ipcp_transformation.
9646 (class ipcp_transformation_t): New function summary.
9647 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
9648 (ipa_get_agg_replacements_for_node): Likewise.
9650 2018-06-08 Martin Liska <mliska@suse.cz>
9652 * ipa-pure-const.c (struct funct_state_d): Do it class instead
9654 (class funct_state_summary_t): New function_summary class.
9655 (has_function_state): Remove.
9656 (get_function_state): Likewise.
9657 (set_function_state): Likewise.
9658 (add_new_function): Likewise.
9659 (funct_state_summary_t::insert): New function.
9660 (duplicate_node_data): Remove.
9661 (remove_node_data): Remove.
9662 (funct_state_summary_t::duplicate): New function.
9663 (register_hooks): Create new funct_state_summaries.
9664 (pure_const_generate_summary): Use it.
9665 (pure_const_write_summary): Likewise.
9666 (pure_const_read_summary): Likewise.
9667 (propagate_pure_const): Likewise.
9668 (propagate_nothrow): Likewise.
9669 (dump_malloc_lattice): Likewise.
9670 (propagate_malloc): Likewise.
9671 (execute): Do not register hooks, just remove summary
9673 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
9676 2018-06-08 Martin Liska <mliska@suse.cz>
9678 * ipa-reference.c (remove_node_data): Remove.
9679 (duplicate_node_data): Likewise.
9680 (class ipa_ref_var_info_summary_t): New class.
9681 (class ipa_ref_opt_summary_t): Likewise.
9682 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
9683 (get_reference_optimization_summary): Use
9684 ipa_ref_opt_sum_summaries.
9685 (set_reference_vars_info): Remove.
9686 (set_reference_optimization_summary): Likewise.
9687 (ipa_init): Create summaries.
9688 (init_function_info): Use function summary.
9689 (ipa_ref_opt_summary_t::duplicate): New function.
9690 (ipa_ref_opt_summary_t::remove): New function.
9691 (get_read_write_all_from_node): Fix GNU coding style.
9692 (propagate): Use function summary.
9693 (write_node_summary_p): Fix GNU coding style.
9694 (stream_out_bitmap): Likewise.
9695 (ipa_reference_read_optimization_summary): Use function summary.
9696 (ipa_reference_c_finalize): Do not release hooks.
9698 2018-06-08 Martin Liska <mliska@suse.cz>
9700 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
9701 (analyze_function_body): Extract multiple calls of get_create.
9702 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
9703 * ipa-inline.c (recursive_inlining): Use ::get method.
9704 * ipa-inline.h (estimate_edge_growth): Likewise.
9706 2018-06-08 Martin Liska <mliska@suse.cz>
9708 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
9710 (hsa_function_summary::hsa_function_summary): Use the new enum
9712 (hsa_gpu_implementation_p): Use hsa_summaries::get.
9713 * hsa-gen.c (hsa_get_host_function): Likewise.
9714 (get_brig_function_name): Likewise.
9715 * ipa-hsa.c (process_hsa_functions): Likewise.
9716 (ipa_hsa_write_summary): Likewise.
9717 * symbol-summary.h (symtab_duplication): Use ::get function/
9718 (get): New function.
9720 2018-06-08 Martin Liska <mliska@suse.cz>
9722 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
9724 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
9725 (hsa_register_kernel): Likewise.
9726 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
9727 * hsa-gen.c (hsa_get_host_function): Likewise.
9728 (get_brig_function_name): Likewise.
9729 (generate_hsa): Likewise.
9730 (pass_gen_hsail::execute): Likewise.
9731 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
9732 (devirtualization_time_bonus): Likewise.
9733 (ipcp_propagate_stage): Likewise.
9734 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
9735 (edge_set_predicate): Likewise.
9736 (evaluate_conditions_for_known_args): Likewise.
9737 (evaluate_properties_for_edge): Likewise.
9738 (ipa_fn_summary::reset): Likewise.
9739 (ipa_fn_summary_t::duplicate): Likewise.
9740 (dump_ipa_call_summary): Likewise.
9741 (ipa_dump_fn_summary): Likewise.
9742 (analyze_function_body): Likewise.
9743 (compute_fn_summary): Likewise.
9744 (estimate_edge_devirt_benefit): Likewise.
9745 (estimate_edge_size_and_time): Likewise.
9746 (estimate_calls_size_and_time): Likewise.
9747 (estimate_node_size_and_time): Likewise.
9748 (inline_update_callee_summaries): Likewise.
9749 (remap_edge_change_prob): Likewise.
9750 (remap_edge_summaries): Likewise.
9751 (ipa_merge_fn_summary_after_inlining): Likewise.
9752 (ipa_update_overall_fn_summary): Likewise.
9753 (read_ipa_call_summary): Likewise.
9754 (inline_read_section): Likewise.
9755 (write_ipa_call_summary): Likewise.
9756 (ipa_fn_summary_write): Likewise.
9757 (ipa_free_fn_summary): Likewise.
9758 * ipa-hsa.c (process_hsa_functions): Likewise.
9759 (ipa_hsa_write_summary): Likewise.
9760 (ipa_hsa_read_section): Likewise.
9761 * ipa-icf.c (sem_function::merge): Likewise.
9762 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
9763 (do_estimate_edge_time): Likewise.
9764 (estimate_size_after_inlining): Likewise.
9765 (estimate_growth): Likewise.
9766 (growth_likely_positive): Likewise.
9767 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
9768 (inline_call): Likewise.
9769 * ipa-inline.c (caller_growth_limits): Likewise.
9770 (can_inline_edge_p): Likewise.
9771 (can_inline_edge_by_limits_p): Likewise.
9772 (compute_uninlined_call_time): Likewise.
9773 (compute_inlined_call_time): Likewise.
9774 (want_inline_small_function_p): Likewise.
9775 (edge_badness): Likewise.
9776 (update_caller_keys): Likewise.
9777 (update_callee_keys): Likewise.
9778 (recursive_inlining): Likewise.
9779 (inline_small_functions): Likewise.
9780 (inline_to_all_callers_1): Likewise.
9781 (dump_overall_stats): Likewise.
9782 (early_inline_small_functions): Likewise.
9783 (early_inliner): Likewise.
9784 * ipa-inline.h (estimate_edge_growth): Likewise.
9785 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
9786 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
9787 * ipa-prop.h (IPA_NODE_REF): Likewise.
9788 (IPA_EDGE_REF): Likewise.
9789 * ipa-pure-const.c (malloc_candidate_p): Likewise.
9790 (propagate_malloc): Likewise.
9791 * ipa-split.c (execute_split_functions): Likewise.
9792 * symbol-summary.h: Rename get to get_create.
9794 (get_create): Likewise.
9795 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
9797 2018-06-08 Martin Liska <mliska@suse.cz>
9799 * symbol-summary.h (release): Move definition out of class
9801 (symtab_removal): Likewise.
9802 (symtab_duplication): Likewise.
9804 2018-06-08 Martin Liska <mliska@suse.cz>
9806 * symbol-summary.h (function_summary): Move constructor
9807 implementation out of class declaration.
9808 (release): Likewise.
9809 (symtab_insertion): Likewise.
9810 (symtab_removal): Likewise.
9811 (symtab_duplication): Likewise.
9814 2018-06-08 Martin Liska <mliska@suse.cz>
9816 * Makefile.in: Remove support for MPX (macros, related functions,
9817 fields in cgraph_node, ...).
9818 * builtin-types.def (BT_BND): Likewise.
9819 (BT_FN_BND_CONST_PTR): Likewise.
9820 (BT_FN_CONST_PTR_BND): Likewise.
9821 (BT_FN_VOID_PTR_BND): Likewise.
9822 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
9823 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
9824 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
9825 (expand_builtin_mempcpy_with_bounds): Likewise.
9826 (expand_builtin_memset_with_bounds): Likewise.
9827 (expand_builtin_memset_args): Likewise.
9828 (std_expand_builtin_va_start): Likewise.
9829 (expand_builtin): Likewise.
9830 (expand_builtin_with_bounds): Likewise.
9831 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
9832 (DEF_LIB_BUILTIN_CHKP): Likewise.
9833 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
9834 (DEF_CHKP_BUILTIN): Likewise.
9835 (BUILT_IN_MEMCPY): Likewise.
9836 (BUILT_IN_MEMMOVE): Likewise.
9837 (BUILT_IN_MEMPCPY): Likewise.
9838 (BUILT_IN_MEMSET): Likewise.
9839 (BUILT_IN_STPCPY): Likewise.
9840 (BUILT_IN_STRCAT): Likewise.
9841 (BUILT_IN_STRCHR): Likewise.
9842 (BUILT_IN_STRCPY): Likewise.
9843 (BUILT_IN_STRLEN): Likewise.
9844 (BUILT_IN_MEMCPY_CHK): Likewise.
9845 (BUILT_IN_MEMMOVE_CHK): Likewise.
9846 (BUILT_IN_MEMPCPY_CHK): Likewise.
9847 (BUILT_IN_MEMSET_CHK): Likewise.
9848 (BUILT_IN_STPCPY_CHK): Likewise.
9849 (BUILT_IN_STRCAT_CHK): Likewise.
9850 (BUILT_IN_STRCPY_CHK): Likewise.
9851 * calls.c (store_bounds): Likewise.
9852 (emit_call_1): Likewise.
9853 (special_function_p): Likewise.
9854 (maybe_warn_nonstring_arg): Likewise.
9855 (initialize_argument_information): Likewise.
9856 (finalize_must_preallocate): Likewise.
9857 (compute_argument_addresses): Likewise.
9858 (expand_call): Likewise.
9859 * cfgexpand.c (expand_call_stmt): Likewise.
9860 (expand_return): Likewise.
9861 (expand_gimple_stmt_1): Likewise.
9862 (pass_expand::execute): Likewise.
9863 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
9864 (cgraph_node::remove): Likewise.
9865 (cgraph_node::dump): Likewise.
9866 (cgraph_node::verify_node): Likewise.
9867 * cgraph.h (chkp_function_instrumented_p): Likewise.
9868 (symtab_node::get_alias_target): Likewise.
9869 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
9870 (cgraph_local_p): Likewise.
9871 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
9872 (cgraph_edge::rebuild_references): Likewise.
9873 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
9874 (walk_polymorphic_call_targets): Likewise.
9875 (cgraph_node::expand_thunk): Likewise.
9876 (symbol_table::output_weakrefs): Likewise.
9877 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
9878 (ix86_handle_option): Likewise.
9879 * config/i386/constraints.md: Likewise.
9880 * config/i386/i386-builtin-types.def (BND): Likewise.
9884 * config/i386/i386-builtin.def (BDESC_END): Likewise.
9885 (BDESC_FIRST): Likewise.
9887 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
9888 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
9889 * config/i386/i386.c (enum reg_class): Likewise.
9890 (ix86_target_string): Likewise.
9891 (ix86_option_override_internal): Likewise.
9892 (ix86_conditional_register_usage): Likewise.
9893 (ix86_valid_target_attribute_inner_p): Likewise.
9894 (ix86_set_indirect_branch_type): Likewise.
9895 (ix86_set_current_function): Likewise.
9896 (ix86_function_arg_regno_p): Likewise.
9897 (init_cumulative_args): Likewise.
9898 (ix86_function_arg_advance): Likewise.
9899 (ix86_function_arg): Likewise.
9900 (ix86_pass_by_reference): Likewise.
9901 (ix86_function_value_regno_p): Likewise.
9902 (ix86_function_value_1): Likewise.
9903 (ix86_function_value_bounds): Likewise.
9904 (ix86_return_in_memory): Likewise.
9905 (ix86_setup_incoming_vararg_bounds): Likewise.
9906 (ix86_va_start): Likewise.
9907 (indirect_thunk_need_prefix): Likewise.
9908 (print_reg): Likewise.
9909 (ix86_print_operand): Likewise.
9910 (ix86_expand_call): Likewise.
9911 (ix86_output_function_return): Likewise.
9912 (reg_encoded_number): Likewise.
9913 (BDESC_VERIFYS): Likewise.
9914 (ix86_init_mpx_builtins): Likewise.
9915 (ix86_init_builtins): Likewise.
9916 (ix86_emit_cmove): Likewise.
9917 (ix86_emit_move_max): Likewise.
9918 (ix86_expand_builtin): Likewise.
9919 (ix86_builtin_mpx_function): Likewise.
9920 (ix86_get_arg_address_for_bt): Likewise.
9921 (ix86_load_bounds): Likewise.
9922 (ix86_store_bounds): Likewise.
9923 (ix86_load_returned_bounds): Likewise.
9924 (ix86_store_returned_bounds): Likewise.
9925 (ix86_class_likely_spilled_p): Likewise.
9926 (ix86_hard_regno_mode_ok): Likewise.
9927 (x86_order_regs_for_local_alloc): Likewise.
9928 (ix86_mitigate_rop): Likewise.
9929 (ix86_bnd_prefixed_insn_p): Likewise.
9930 (ix86_mpx_bound_mode): Likewise.
9931 (ix86_make_bounds_constant): Likewise.
9932 (ix86_initialize_bounds): Likewise.
9933 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
9934 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
9935 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
9936 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
9937 (TARGET_CHKP_BOUND_MODE): Likewise.
9938 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
9939 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
9940 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
9941 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
9942 * config/i386/i386.h (TARGET_MPX): Likewise.
9943 (TARGET_MPX_P): Likewise.
9944 (VALID_BND_REG_MODE): Likewise.
9945 (FIRST_BND_REG): Likewise.
9946 (LAST_BND_REG): Likewise.
9947 (enum reg_class): Likewise.
9948 (BND_REG_P): Likewise.
9949 (BND_REGNO_P): Likewise.
9950 (BNDmode): Likewise.
9951 (ADJUST_INSN_LENGTH): Likewise.
9952 * config/i386/i386.md: Likewise.
9953 * config/i386/i386.opt: Likewise.
9954 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
9955 (defined): Likewise.
9956 (LINK_MPX): Likewise.
9957 (MPX_SPEC): Likewise.
9958 (LIBMPX_SPEC): Likewise.
9959 (LIBMPXWRAPPERS_SPEC): Likewise.
9960 (CHKP_SPEC): Likewise.
9961 * config/i386/predicates.md: Likewise.
9962 * dbxout.c (dbxout_type): Likewise.
9963 * doc/extend.texi: Likewise.
9964 * doc/invoke.texi: Likewise.
9965 * doc/md.texi: Likewise.
9966 * doc/tm.texi: Likewise.
9967 * doc/tm.texi.in: Likewise.
9968 * dwarf2out.c (is_base_type): Likewise.
9969 (gen_formal_types_die): Likewise.
9970 (gen_subprogram_die): Likewise.
9971 (gen_type_die_with_usage): Likewise.
9972 (gen_decl_die): Likewise.
9973 (dwarf2out_late_global_decl): Likewise.
9974 * expr.c (expand_assignment): Likewise.
9975 (emit_storent_insn): Likewise.
9976 (store_expr_with_bounds): Likewise.
9977 (store_expr): Likewise.
9978 (expand_expr_real_1): Likewise.
9979 * expr.h (store_expr_with_bounds): Likewise.
9980 * function.c (use_register_for_decl): Likewise.
9981 (struct bounds_parm_data): Likewise.
9982 (assign_parms_augmented_arg_list): Likewise.
9983 (assign_parm_find_entry_rtl): Likewise.
9984 (assign_parm_is_stack_parm): Likewise.
9985 (assign_parm_load_bounds): Likewise.
9986 (assign_bounds): Likewise.
9987 (assign_parms): Likewise.
9988 (expand_function_start): Likewise.
9989 * gcc.c (CHKP_SPEC): Likewise.
9990 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
9991 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
9992 (wrestrict_dom_walker::check_call): Likewise.
9993 * gimple.c (gimple_build_call_from_tree): Likewise.
9994 * gimple.h (enum gf_mask): Likewise.
9995 (gimple_call_with_bounds_p): Likewise.
9996 (gimple_call_set_with_bounds): Likewise.
9997 * gimplify.c (gimplify_init_constructor): Likewise.
9998 * ipa-cp.c (initialize_node_lattices): Likewise.
9999 (propagate_constants_across_call): Likewise.
10000 (find_more_scalar_values_for_callers_subset): Likewise.
10001 * ipa-hsa.c (process_hsa_functions): Likewise.
10002 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
10003 * ipa-icf.c (sem_function::merge): Likewise.
10004 * ipa-inline.c (early_inliner): Likewise.
10005 * ipa-pure-const.c (warn_function_noreturn): Likewise.
10006 (warn_function_cold): Likewise.
10007 (propagate_pure_const): Likewise.
10008 * ipa-ref.h (enum GTY): Likewise.
10009 * ipa-split.c (find_retbnd): Likewise.
10010 (consider_split): Likewise.
10011 (split_function): Likewise.
10012 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
10013 * ipa.c (walk_polymorphic_call_targets): Likewise.
10014 (symbol_table::remove_unreachable_nodes): Likewise.
10015 (process_references): Likewise.
10016 (cgraph_build_static_cdtor_1): Likewise.
10017 * lto-cgraph.c (lto_output_node): Likewise.
10018 (output_refs): Likewise.
10019 (compute_ltrans_boundary): Likewise.
10020 (input_overwrite_node): Likewise.
10021 (input_node): Likewise.
10022 (input_cgraph_1): Likewise.
10023 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
10024 * passes.c (pass_manager::execute_early_local_passes): Likewise.
10025 (class pass_chkp_instrumentation_passes): Likewise.
10026 (make_pass_chkp_instrumentation_passes): Likewise.
10027 * passes.def: Likewise.
10028 * rtl.h (struct GTY): Likewise.
10029 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
10030 * stor-layout.c (layout_type): Likewise.
10031 * symtab.c: Likewise.
10032 * target.def: Likewise.
10033 * targhooks.c (default_chkp_bound_type): Likewise.
10034 (default_chkp_bound_mode): Likewise.
10035 (default_builtin_chkp_function): Likewise.
10036 (default_chkp_function_value_bounds): Likewise.
10037 (default_chkp_make_bounds_constant): Likewise.
10038 (default_chkp_initialize_bounds): Likewise.
10039 * targhooks.h (default_chkp_bound_type): Likewise.
10040 (default_chkp_bound_mode): Likewise.
10041 (default_builtin_chkp_function): Likewise.
10042 (default_chkp_function_value_bounds): Likewise.
10043 (default_chkp_make_bounds_constant): Likewise.
10044 (default_chkp_initialize_bounds): Likewise.
10045 * toplev.c (compile_file): Likewise.
10046 (process_options): Likewise.
10047 * tree-core.h (DEF_BUILTIN): Likewise.
10048 (DEF_BUILTIN_CHKP): Likewise.
10049 * tree-inline.c (declare_return_variable): Likewise.
10050 (remap_gimple_stmt): Likewise.
10051 (copy_bb): Likewise.
10052 (initialize_inlined_parameters): Likewise.
10053 (expand_call_inline): Likewise.
10054 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
10055 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
10056 (make_pass_ipa_chkp_produce_thunks): Likewise.
10057 (make_pass_chkp): Likewise.
10058 (make_pass_chkp_opt): Likewise.
10059 (make_pass_chkp_instrumentation_passes): Likewise.
10060 * tree-pretty-print.c (dump_generic_node): Likewise.
10061 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
10062 * tree-ssa-dce.c (propagate_necessity): Likewise.
10063 (eliminate_unnecessary_stmts): Likewise.
10064 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
10065 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
10066 * tree-ssa-sccvn.h: Likewise.
10067 * tree-ssa-strlen.c (get_string_length): Likewise.
10068 (valid_builtin_call): Likewise.
10069 (adjust_last_stmt): Likewise.
10070 (handle_builtin_strchr): Likewise.
10071 (handle_builtin_strcpy): Likewise.
10072 (handle_builtin_stxncpy): Likewise.
10073 (handle_builtin_memcpy): Likewise.
10074 (handle_builtin_strcat): Likewise.
10075 (strlen_check_and_optimize_stmt): Likewise.
10076 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
10077 * tree-streamer-in.c: Likewise.
10078 * tree-streamer.c (record_common_node): Likewise.
10079 * tree.c (tree_code_size): Likewise.
10080 (wide_int_to_tree_1): Likewise.
10081 (type_contains_placeholder_1): Likewise.
10082 (build_common_tree_nodes): Likewise.
10083 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
10084 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
10085 (POINTER_BOUNDS_P): Likewise.
10086 (BOUNDED_TYPE_P): Likewise.
10087 (BOUNDED_P): Likewise.
10088 (CALL_WITH_BOUNDS_P): Likewise.
10089 (pointer_bounds_type_node): Likewise.
10090 * value-prof.c (gimple_ic): Likewise.
10091 * var-tracking.c (vt_add_function_parameters): Likewise.
10092 * varasm.c (make_decl_rtl): Likewise.
10093 (assemble_start_function): Likewise.
10094 (output_constant): Likewise.
10095 (maybe_assemble_visibility): Likewise.
10096 * varpool.c (ctor_for_folding): Likewise.
10097 * chkp-builtins.def: Remove.
10098 * ipa-chkp.c: Remove.
10099 * ipa-chkp.h: Remove.
10100 * rtl-chkp.c: Remove.
10101 * rtl-chkp.h: Remove.
10102 * tree-chkp-opt.c: Remove.
10103 * tree-chkp.c: Remove.
10104 * tree-chkp.h: Remove.
10106 2018-06-07 Carl Love <cel@us.ibm.com>
10108 * config/rs6000/vsx.md (vextract_fp_from_shorth,
10109 vextract_fp_from_shortl): Add BE support.
10111 2018-06-07 Paul Koning <ni1d@arrl.net>
10113 * compare-elim.c (try_merge_compare): Don't merge compare if
10114 address contains a side effect.
10115 (try_eliminate_compare): Likewise.
10117 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
10119 * config.gcc: Support "tremont".
10120 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
10121 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10123 * config/i386/i386.c (m_TREMONT): Define.
10124 (processor_target_table): Add "tremont".
10125 (PTA_TREMONT): Define.
10126 (ix86_lea_outperforms): Add TARGET_TREMONT.
10127 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
10128 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
10129 and M_INTEL_GOLDMONT_PLUS.
10130 (fold_builtin_cpu): Add "tremont".
10131 (ix86_add_stmt_cost): Add TARGET_TREMONT.
10132 (ix86_option_override_internal): Add "tremont".
10133 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
10134 (processor_type): Add PROCESSOR_TREMONT.
10135 * config/i386/x86-tune.def: Add m_TREMONT.
10136 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
10138 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10140 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
10141 symbol defined for msp430i* devices to be lower case.
10143 2018-06-07 Richard Biener <rguenther@suse.de>
10145 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
10146 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
10147 Properly wrap signed arithmetic if overflow wraps.
10149 2018-06-07 Jakub Jelinek <jakub@redhat.com>
10151 PR tree-optimization/69615
10152 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
10153 of a cast from a same precision integral SSA_NAME in a bb dominated
10154 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
10155 cast to utype if rhs2 has already a compatible type.
10157 2018-06-07 Richard Biener <rguenther@suse.de>
10159 PR tree-optimization/85935
10160 * graphite-scop-detection.c (find_params_in_bb): Analyze
10161 condition operands with respect to the correct loop. Assert
10162 the analysis doesn't fail.
10164 2018-06-04 Carl Love <cel@us.ibm.com>
10166 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
10167 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
10168 as it is slightly cheaper.
10169 (first_match_or_eos_index_<mode>):
10170 Calculate index using natural element order.
10171 (first_match_index_<mode>):
10172 Calculate index using natural element order.
10173 (first_match_or_eos_index_<mode>):
10174 Calculate index using natural order.
10175 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
10176 for BE and LE modes.
10177 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
10178 P9V_BUILTIN_VCLZLSBB_V16QI.
10179 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
10182 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
10184 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
10185 indentation and line wrap for many prototypes. Add missing
10186 @smallexample directives around block of prototypes for vec_xl and
10189 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
10191 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
10192 track if we pass or return IEEE 128-bit floating point.
10193 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
10194 C++ mangling that is compatible with GCC 8.1.
10195 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
10196 (init_cumulative_args): Note if we pass or return IEEE 128-bit
10197 floating point types.
10198 (rs6000_function_arg_advance_1): Likewise.
10199 (rs6000_mangle_type): Optionally generate mangled names that match
10200 what GCC 8.1 generated for IEEE 128-bit floating point types.
10201 (rs6000_globalize_decl_name): If we have an external function that
10202 passes or returns IEEE 128-bit types, generate a weak reference
10203 from the mangled name used in GCC 8.1 to the current mangled
10205 (rs6000_init_builtins): Make __ibm128 use the long double type if
10206 long double is IBM extended double. Make __float128 use the long
10207 double type if long double is IEEE 128-bit.
10210 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
10211 macro for __ibm128 built-in functions.
10212 (PACK_IF): Add __ibm128 pack/unpack functions.
10213 (UNPACK_IF): Likewise.
10214 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
10215 enable long double built-in functions if long double is IEEE
10216 128-bit floating point.
10217 (rs6000_invalid_builtin): Update long double built-in function
10219 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
10220 functions, adjust the built-in function to use the long double
10221 built-in function if __ibm128 and long double are the same type.
10222 * doc/extend.texi (PowerPC builtins): Update documention for
10223 __builtin_{,un}pack_longdouble. Add documentation for
10224 __builtin_{,un}pack_ibm128.
10226 2018-06-06 Jim Wilson <jimw@sifive.com>
10228 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
10229 (struct machine_function): New field interrupt_mode.
10230 (riscv_handle_type_attribute): New function. Add forward declaration.
10231 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
10232 (riscv_expand_epilogue): Check interrupt_mode field.
10233 (riscv_set_current_function): Check interrupt attribute args and
10234 set interrupt_mode field.
10235 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
10236 (riscv_sret, riscv_uret): New.
10237 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
10238 new arguments to interrupt attribute.
10240 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
10243 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
10244 Don't handle -mcpu=power8 if -mpower9-vector is also used.
10246 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
10248 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
10249 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
10250 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
10251 several redundant entries.
10253 2018-06-06 David Malcolm <dmalcolm@redhat.com>
10255 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
10256 type from "rtx" to "rtx_insn *".
10257 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
10258 for local "call_insn", removing cast.
10259 (ix86_expand_call): Likewise, introducing a "call_insn" local.
10261 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
10263 PR tree-optimization/86066
10264 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
10265 for BIT_INSERT_EXPR stores.
10267 2018-06-06 Richard Biener <rguenther@suse.de>
10269 PR tree-optimization/86062
10270 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
10271 component refs ontop
10272 of to be offsetted base.
10274 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10276 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
10277 to be static and remove check on interrupt attribute name.
10279 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
10281 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
10282 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
10284 2018-06-05 Steve Ellcey <sellcey@cavium.com>
10287 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
10289 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
10290 Remove second argument, change how error is called.
10291 (aarch64_layout_arg): Remove second argument from
10292 aarch64_err_no_fpadvsimd call.
10293 (aarch64_init_cumulative_args): Ditto.
10294 (aarch64_gimplify_va_arg_expr): Ditto.
10295 * config/aarch64/aarch64.md (mov<mode>): Ditto.
10297 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
10299 * config/i386/i386.md (simple_return_indirect_internal): New expander.
10300 (*simple_return_indirect_internal<mode>): Rename from
10301 simple_return_indirect_internal. Use W mode iterator.
10302 (rstorssp): New expander.
10303 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
10304 (clrssbsy): New expander.
10305 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
10307 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
10309 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
10311 (cmse_check_pointed_object): Likewise.
10313 2018-06-05 Martin Liska <mliska@suse.cz>
10315 PR gcov-profile/47618
10316 * doc/invoke.texi: Document how -fprofile-dir format
10319 2018-06-05 Richard Biener <rguenther@suse.de>
10321 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
10322 removal pretend DOM info isn't available so we do not update
10323 it and only remove edges, not dominated blocks. Actually free
10324 DOM info in case we removed something. Remove unreachable blocks.
10325 (mfb_keep_latches): Work with either DOM info or marked backedges.
10326 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
10327 first. Mark backedges if DOM info isn't available.
10328 (Re-)compute DOM info after cleanup_control_flow_pre.
10330 2018-06-05 Richard Biener <rguenther@suse.de>
10332 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
10333 (locus_discrim_hasher::hash): Adjust.
10334 (locus_discrim_hasher::equal): Likewise.
10335 (next_discriminator_for_locus): Work on line directly.
10336 (same_line_p): Pass in expanded locus1 as well.
10337 (assign_discriminators): Avoid redundant location expansions.
10339 2018-06-05 Richard Biener <rguenther@suse.de>
10341 PR tree-optimization/86046
10342 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
10343 if required after clearing TREE_ADDRESSABLE.
10345 2018-06-05 Richard Biener <rguenther@suse.de>
10347 PR tree-optimization/86047
10348 * tree-ssa-loop.c (for_each_index): Glob handling of all
10349 decls and constants and really handle all of them.
10351 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10354 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
10355 qualifier_void_pointer and qualifier_const_void_pointer.
10356 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
10357 (arm_init_builtins): Handle the above.
10358 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
10359 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
10360 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
10363 2018-06-05 Martin Liska <mliska@suse.cz>
10365 * auto-profile.c (read_autofdo_file): Do not use
10366 gcov_ctr_summary struct.
10367 (afdo_callsite_hot_enough_for_early_inline): Likewise.
10368 * coverage.c (struct counts_entry): Likewise.
10369 (read_counts_file): Read just single summary entry.
10370 (get_coverage_counts): Use gcov_summary struct.
10371 * coverage.h (get_coverage_counts): Likewise.
10372 * gcov-dump.c (dump_working_sets): Likewise.
10373 (tag_summary): Dump just single summary.
10374 * gcov-io.c (gcov_write_summary): Write just histogram
10376 (gcov_read_summary): Read just single summary.
10377 (compute_working_sets): Use gcov_summary struct.
10378 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
10379 of GCOV_COUNTERS_SUMMABLE.
10380 (GCOV_COUNTERS_SUMMABLE): Remove.
10381 (GCOV_FIRST_VALUE_COUNTER): Replace with
10382 GCOV_COUNTER_V_INTERVAL.
10383 (struct gcov_ctr_summary): Remove.
10384 (struct gcov_summary): Directly use fields of former
10386 (compute_working_sets): Use gcov_summary struct.
10387 * gcov.c (read_count_file): Do not use ctrs fields.
10388 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
10390 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
10392 * profile.c: Likewise.
10393 * profile.h: Likewise.
10395 2018-06-05 Martin Liska <mliska@suse.cz>
10397 PR gcov-profile/84846
10398 * gcov.c (output_lines): Print working directory only
10399 in intermediate format.
10401 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
10403 * config/s390/s390-builtin-types.def: Add void function type.
10404 * config/s390/s390-builtins.def: Use the function type for the
10407 2018-06-04 Jim Wilson <jimw@sifive.com>
10409 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
10411 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
10412 and maybe_eh_return. Change regno to unsigned int. Use new args to
10413 handle EH_RETURN_DATA_REGNO registers properly.
10414 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
10415 (riscv_expand_epilogue): Update comment. Change argument name and
10416 type. Update code to use new name and type. Pass new args to
10417 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
10419 * config/riscv/riscv.md (NORMAL_RETURN): New.
10420 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
10421 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
10422 (eh_return): Call gen_eh_return_internal and emit barrier.
10423 (eh_return_internal): Call riscv_expand_epilogue.
10425 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
10427 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
10428 bit_insertion field and declare can_be_merged_into method.
10429 (merged_store_group::can_be_merged_into): New method.
10430 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
10431 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
10432 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
10434 2018-06-04 Richard Biener <rguenther@suse.de>
10436 PR tree-optimization/85955
10437 * builtins.c (fold_builtin_sincos): Convert pointers to
10438 destination to appropriate type before dereferencing.
10440 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
10442 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
10444 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
10446 * expr.c (expand_expr_real_1): Force the operand into memory if
10447 its TYPE_MODE is BLKmode and if there is no integer mode for
10448 the number of bits being extracted.
10450 2018-06-04 Jakub Jelinek <jakub@redhat.com>
10454 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
10455 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
10457 2018-06-04 Richard Biener <rguenther@suse.de>
10459 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
10460 (cleanup_tree_cfg_noloop): ... single caller. Do
10461 start_recording_case_labels later.
10463 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
10465 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
10466 to _IMMINTRIN_H_INCLUDED.
10467 * config/i386/pconfigintrin.h: Ditto.
10468 * config/i386/waitpkgintrin.h: Ditto.
10469 * config/i386/immintrin.h: Add includes for sgxintrin.h,
10470 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
10471 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
10472 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
10473 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
10474 waitpkgintrin.h and cldemoteintrin.h.
10476 2018-06-04 Richard Biener <rguenther@suse.de>
10478 PR tree-optimization/86038
10479 * tracer.c (find_best_successor): Check probability for
10480 being initialized, bail out if not.
10482 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
10485 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
10486 of bits to ignore when comparing architectures.
10488 2018-06-04 Jakub Jelinek <jakub@redhat.com>
10490 PR tree-optimization/69615
10491 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
10492 maximum or minimum of the type, try to merge it also as if
10493 range1 is + [-, x - 1] or + [x + 1, -].
10496 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
10498 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
10500 PR tree-optimization/86034
10501 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
10502 the unsigned bitfield type in a bit insertion sequence if it does not
10503 have a larger precision than the bitfield size.
10504 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
10506 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
10508 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
10510 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
10512 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
10513 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
10514 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
10515 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
10517 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
10519 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
10520 Disable -fdelete-null-pointer-checks for ELF toolchain.
10522 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
10523 Kito Cheng <kito.cheng@gmail.com>
10525 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
10526 (nds32le-*-*, nds32be-*-*): Integrate checking process.
10527 (nds32*-*-*): Add glibc and uclibc conditions.
10528 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
10529 (TARGET_EXCEPT_UNWIND_INFO): Define.
10530 * config/nds32/elf.h: New file.
10531 * config/nds32/linux.h: New file.
10532 * config/nds32/nds32-elf.opt: New file.
10533 * config/nds32/nds32-linux.opt: New file.
10534 * config/nds32/nds32-fp-as-gp.c
10535 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
10536 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
10538 (nds32_asm_file_end): Ditto.
10539 (nds32_print_operand): Ditto.
10540 (nds32_insert_attributes): Ditto.
10541 (nds32_init_libfuncs): New function.
10542 (TARGET_HAVE_TLS): Define.
10543 (TARGET_INIT_LIBFUNCS): Define.
10544 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
10546 (TARGET_ELF): Apply different mcmodel setting.
10547 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
10548 been migrated into elf.h and linux.h files.
10549 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
10550 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
10551 (mcmodel): The content has been migrated into nds32-elf.opt and
10552 nds32-linux.opt files.
10553 * config/nds32/t-elf: New file.
10554 * config/nds32/t-linux: New file.
10556 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
10557 Shiva Chen <shiva0217@gmail.com>
10559 * config/nds32/constants.md (unspec_volatile_element): Add
10560 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
10561 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
10563 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
10564 (make_pass_nds32_fp_as_gp): Declare.
10565 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
10567 (nds32_asm_function_end_prologue): Remove unused asm output.
10568 (nds32_asm_function_begin_epilogue): Remove unused asm output.
10569 (nds32_asm_file_start): Output necessary fp_as_gp information.
10570 (nds32_option_override): Adjust register usage.
10571 (nds32_expand_prologue): Consider fp_as_gp situation.
10572 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
10573 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
10576 (simple_return): Ditto.
10577 (omit_fp_begin): Output special directive for fp_as_gp.
10578 (omit_fp_end): Output special directive for fp_as_gp.
10579 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
10580 mforbid-fp-as-gp): New options.
10582 2018-06-01 Mark Wielaard <mark@klomp.org>
10584 * dwarf2out.c (dwarf2out_finish): Remove generation of
10585 DW_AT_loclists_base.
10587 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
10589 * gimple-ssa-store-merging.c: Include gimple-fold.h.
10590 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
10591 (struct merged_store_group): Add bit_insertion field.
10592 (dump_char_array): Use standard hexadecimal format.
10593 (merged_store_group::merged_store_group): Set bit_insertion to false.
10594 (merged_store_group::apply_stores): Use optimal buffer size. Deal
10595 with BIT_INSERT_EXPR stores. Move up code updating the mask and
10596 also print the mask in the dump file.
10597 (pass_store_merging::gate): Minor tweak.
10598 (imm_store_chain_info::coalesce_immediate): Fix wrong association
10599 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
10600 stores with INTEGER_CST stores.
10601 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
10602 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
10603 and use it throughout. Generate bit insertion sequences if need be.
10604 (pass_store_merging::process_store): Remove redundant condition.
10605 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
10607 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
10609 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
10610 the 128-bit floating point types. Fix function comment.
10612 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10614 * config/aarch64/aarch64-simd.md
10615 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
10617 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
10620 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
10622 PR tree-optimization/85989
10623 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
10625 (backprop::intersect_uses): Check it when deciding whether this
10626 is a backedge reference.
10627 (backprop::process_block): Add each phi to m_visited_phis
10628 after visiting it, then clear it at the end.
10630 2018-06-01 Richard Biener <rguenther@suse.de>
10632 * tree-vectorizer.h (vect_dr_stmt): New function.
10633 (vect_get_load_cost): Adjust.
10634 (vect_get_store_cost): Likewise.
10635 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
10636 Use vect_dr_stmt instead of DR_SMTT.
10637 (vect_record_base_alignments): Likewise.
10638 (vect_calculate_target_alignment): Likewise.
10639 (vect_compute_data_ref_alignment): Likewise and make static.
10640 (vect_update_misalignment_for_peel): Likewise.
10641 (vect_verify_datarefs_alignment): Likewise.
10642 (vector_alignment_reachable_p): Likewise.
10643 (vect_get_data_access_cost): Likewise. Pass down
10644 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
10645 (vect_get_peeling_costs_all_drs): Likewise.
10646 (vect_peeling_hash_get_lowest_cost): Likewise.
10647 (vect_enhance_data_refs_alignment): Likewise.
10648 (vect_find_same_alignment_drs): Likewise.
10649 (vect_analyze_data_refs_alignment): Likewise.
10650 (vect_analyze_group_access_1): Likewise.
10651 (vect_analyze_group_access): Likewise.
10652 (vect_analyze_data_ref_access): Likewise.
10653 (vect_analyze_data_ref_accesses): Likewise.
10654 (vect_vfa_segment_size): Likewise.
10655 (vect_small_gap_p): Likewise.
10656 (vectorizable_with_step_bound_p): Likewise.
10657 (vect_prune_runtime_alias_test_list): Likewise.
10658 (vect_analyze_data_refs): Likewise.
10659 (vect_supportable_dr_alignment): Likewise.
10660 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10661 (vect_gen_prolog_loop_niters): Likewise.
10662 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10663 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
10665 (vect_recog_mask_conversion_pattern): Likewise.
10666 (vect_try_gather_scatter_pattern): Likewise.
10667 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
10668 to vect_get_store_cost.
10669 (vect_get_store_cost): Get stmt_info instead of DR.
10670 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
10671 (vect_get_load_cost): Get stmt_info instead of DR.
10673 2018-06-01 Richard Biener <rguenther@suse.de>
10675 PR middle-end/86017
10676 * gimple-fold.c (var_decl_component_p): Also allow offsetted
10677 vars wrapped in MEM_REFs.
10679 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
10681 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
10682 Fix subreg tests so that we only return a choice between
10683 GENERAL_REGS and FP_REGS if the original classes included both.
10685 2018-06-01 Richard Biener <rguenther@suse.de>
10688 * tree-ssa-structalias.c (get_function_part_constraint):
10689 Handle NULL fi->decl.
10690 (find_func_aliases_for_call): Properly handle indirect
10691 fi from direct call.
10692 (find_func_clobbers): Likewise.
10693 (ipa_pta_execute): Likewise.
10694 (create_variable_info_for): For functions that are ifunc_resolver
10695 resolve to a varinfo that contains the result of the resolver call.
10696 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
10699 2018-05-31 Michael Collison <michael.collison@arm.com>
10701 * config/aarch64/aarch64.md:
10702 (*fix_to_zero_extenddfdi2): New pattern.
10703 * gcc.target/aarch64/fix_extend1.c: New testcase.
10705 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
10707 PR middle-end/78809
10708 PR middle-end/83026
10709 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
10710 and BUILT_IN_STRNCMP_EQ.
10711 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
10712 BUILT_IN_STRNCMP_EQ.
10713 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
10714 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
10715 (gimple_fold_builtin): Likewise.
10716 * tree-ssa-strlen.c (compute_string_length): New function.
10717 (determine_min_obsize): New function.
10718 (handle_builtin_string_cmp): New function to handle calls to
10719 string compare functions.
10720 (strlen_optimize_stmt): Add handling to builtin string compare
10722 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
10723 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
10724 * tree.c (build_common_builtin_nodes): Add new defines of
10725 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
10727 2018-05-31 Jakub Jelinek <jakub@redhat.com>
10730 * bb-reorder.c (pass_partition_blocks::gate): Return false for
10731 functions with naked attribute.
10733 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
10735 * config/i386/sse.md (avx_vec_concat<mode>):
10736 Substitute concat_tg_mode mode attribute with xtg_mode.
10737 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
10738 (concat_tg_mode): Remove mode attribute.
10740 2018-05-31 Martin Sebor <msebor@redhat.com>
10743 * calls.c (alloc_max_size): Correct a logic error/typo.
10744 Treat excessive arguments as infinite. Warn for invalid arguments.
10745 * doc/invoke.texi (-Walloc-size-larger-than): Update.
10747 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
10750 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
10751 and movx for Haswell.
10753 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
10754 Cesar Philippidis <cesar@codesourcery.com>
10756 PR middle-end/85879
10757 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
10758 when emitting error on private/firstprivate reductions.
10759 * omp-low.c (lower_omp_target): Avoid reference-type processing
10760 on pointers for firstprivate clause.
10762 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
10764 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
10767 * config/aarch64/aarch64-simd.md
10768 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
10769 (aarch64_ld1_x3_<mode>): Likewise
10770 (aarch64_st1x2<VALLDIF:mode>): Likewise
10771 (aarch64_st1_x2_<mode>): Likewise
10772 (aarch64_st1x3<VALLDIF:mode>): Likewise
10773 (aarch64_st1_x3_<mode>): Likewise
10774 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
10775 (vld1_s8_x3): Likewise.
10776 (vld1_u16_x3): Likewise.
10777 (vld1_s16_x3): Likewise.
10778 (vld1_u32_x3): Likewise.
10779 (vld1_s32_x3): Likewise.
10780 (vld1_u64_x3): Likewise.
10781 (vld1_s64_x3): Likewise.
10782 (vld1_f16_x3): Likewise.
10783 (vld1_f32_x3): Likewise.
10784 (vld1_f64_x3): Likewise.
10785 (vld1_p8_x3): Likewise.
10786 (vld1_p16_x3): Likewise.
10787 (vld1_p64_x3): Likewise.
10788 (vld1q_u8_x3): Likewise.
10789 (vld1q_s8_x3): Likewise.
10790 (vld1q_u16_x3): Likewise.
10791 (vld1q_s16_x3): Likewise.
10792 (vld1q_u32_x3): Likewise.
10793 (vld1q_s32_x3): Likewise.
10794 (vld1q_u64_x3): Likewise.
10795 (vld1q_s64_x3): Likewise.
10796 (vld1q_f16_x3): Likewise.
10797 (vld1q_f32_x3): Likewise.
10798 (vld1q_f64_x3): Likewise.
10799 (vld1q_p8_x3): Likewise.
10800 (vld1q_p16_x3): Likewise.
10801 (vld1q_p64_x3): Likewise.
10802 (vst1_s64_x2): Likewise.
10803 (vst1_u64_x2): Likewise.
10804 (vst1_f64_x2): Likewise.
10805 (vst1_s8_x2): Likewise.
10806 (vst1_p8_x2): Likewise.
10807 (vst1_s16_x2): Likewise.
10808 (vst1_p16_x2): Likewise.
10809 (vst1_s32_x2): Likewise.
10810 (vst1_u8_x2): Likewise.
10811 (vst1_u16_x2): Likewise.
10812 (vst1_u32_x2): Likewise.
10813 (vst1_f16_x2): Likewise.
10814 (vst1_f32_x2): Likewise.
10815 (vst1_p64_x2): Likewise.
10816 (vst1q_s8_x2): Likewise.
10817 (vst1q_p8_x2): Likewise.
10818 (vst1q_s16_x2): Likewise.
10819 (vst1q_p16_x2): Likewise.
10820 (vst1q_s32_x2): Likewise.
10821 (vst1q_s64_x2): Likewise.
10822 (vst1q_u8_x2): Likewise.
10823 (vst1q_u16_x2): Likewise.
10824 (vst1q_u32_x2): Likewise.
10825 (vst1q_u64_x2): Likewise.
10826 (vst1q_f16_x2): Likewise.
10827 (vst1q_f32_x2): Likewise.
10828 (vst1q_f64_x2): Likewise.
10829 (vst1q_p64_x2): Likewise.
10830 (vst1_s64_x3): Likewise.
10831 (vst1_u64_x3): Likewise.
10832 (vst1_f64_x3): Likewise.
10833 (vst1_s8_x3): Likewise.
10834 (vst1_p8_x3): Likewise.
10835 (vst1_s16_x3): Likewise.
10836 (vst1_p16_x3): Likewise.
10837 (vst1_s32_x3): Likewise.
10838 (vst1_u8_x3): Likewise.
10839 (vst1_u16_x3): Likewise.
10840 (vst1_u32_x3): Likewise.
10841 (vst1_f16_x3): Likewise.
10842 (vst1_f32_x3): Likewise.
10843 (vst1_p64_x3): Likewise.
10844 (vst1q_s8_x3): Likewise.
10845 (vst1q_p8_x3): Likewise.
10846 (vst1q_s16_x3): Likewise.
10847 (vst1q_p16_x3): Likewise.
10848 (vst1q_s32_x3): Likewise.
10849 (vst1q_s64_x3): Likewise.
10850 (vst1q_u8_x3): Likewise.
10851 (vst1q_u16_x3): Likewise.
10852 (vst1q_u32_x3): Likewise.
10853 (vst1q_u64_x3): Likewise.
10854 (vst1q_f16_x3): Likewise.
10855 (vst1q_f32_x3): Likewise.
10856 (vst1q_f64_x3): Likewise.
10857 (vst1q_p64_x3): Likewise.
10859 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10861 * config/msp430/msp430.c (msp430_output_labelref): Prepend
10862 user_label_prefix to name.
10864 * tree-core.h: Update comment about the format of NAME string
10865 passed to handler in attribute_spec.
10867 * config/msp430/msp430.md: Remove erroneous subreg expression from
10868 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
10869 zero_extend{q,h}isi2.
10871 2018-05-30 Borislav Petkov <bp@suse.de>
10873 * doc/extend.texi: Document some architecture specific
10874 constraints and sort entries.
10876 2018-05-30 Martin Sebor <msebor@redhat.com>
10878 PR middle-end/85369
10879 * builtins.c (expand_builtin_stpcpy_1): New function.
10880 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
10881 only if the former succeeds.
10883 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
10885 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
10888 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10890 * doc/invoke.texi (-flinker-output): Document
10892 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10894 * passes.c (ipa_write_summaries): Only modify statements if body
10896 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
10897 incrementally linking.
10898 (ipa_passes): Likewise.
10899 * lto-cgraph.c (lto_output_node): When incrementally linking do not
10900 pass down resolution info.
10901 * common.opt (flag_incremental_link): Update info.
10902 * gcc.c (plugin specs): Turn flinker-output=* to
10903 -plugin-opt=-linker-output-known
10904 * toplev.c (compile_file): Also cut compilation when doing incremental
10906 * flag-types. (enum lto_partition_model): Add
10907 LTO_LINKER_OUTPUT_NOLTOREL.
10908 (invoke.texi): Add -flinker-output docs.
10909 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
10910 link same way as WPA; do not stream in dead initializers.
10912 * dwarf2out.c (dwarf2out_die_ref_for_decl,
10913 darf2out_register_external_decl): Support incremental link.
10915 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10917 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
10919 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10921 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
10922 it down to simple_object_copy_lto_debug_sections.
10923 (run_gcc): Determine incremental LTO link time and configure
10924 lto1 into non-wpa mode, disable renaming of debug sections.
10926 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
10928 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
10929 descriptions of various incorrectly documented functions.
10931 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
10934 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
10935 address check not strict.
10937 2018-05-30 Richard Biener <rguenther@suse.de>
10939 PR tree-optimization/85964
10940 * tracer.c (better_p): Drop initialized count check, we only
10941 call the function with initialized counts now.
10942 (find_best_successor): Do find a best edge if one
10943 has uninitialized count.
10944 (find_best_predecessor): Likewise. Do BB frequency check only
10945 if count is initialized.
10947 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
10949 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
10950 (aarch64_ldrstr_offset_compare): New.
10951 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
10952 load/store orderings.
10953 (aarch64_gen_adjusted_ldpstp): Likewise.
10955 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
10957 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
10958 Check for subset of GENERAL_REGS and FP_REGS.
10959 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
10962 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
10964 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
10965 and wi::to_poly_offset. Add the current offset and then check
10966 whether the sum fits, rather than using an unchecked addition of
10967 a checked term. Check for a shwi rather than a uhwi.
10968 * expr.c (get_bit_range): Use tree_to_poly_uint64.
10969 (store_constructor): Use poly_int_tree_p.
10970 (expand_expr_real_1): Likewise.
10971 * function.c (assign_temp): Likewise.
10972 * fold-const.c (const_binop): Use poly_int_tree_p and
10973 wi::to_poly_offset.
10974 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
10976 * ipa-icf-gimple.c (func_checker::compare_operand): Use
10977 to_poly_offset for MEM offsets.
10978 * ipa-icf.c (sem_variable::equals): Likewise.
10979 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
10980 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
10981 wi::to_poly_offset for BIT_FIELD_REF offsets.
10982 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
10983 wi::to_poly_offset.
10984 * var-tracking.c (emit_note_insn_var_location): Use
10985 tree_to_poly_uint64.
10987 2018-05-29 Jim Wilson <jimw@sifive.com>
10989 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
10991 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
10994 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
10995 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
10997 (sse4_1_round<mode>2): Use nonimmediate_operand
10998 for operand 1 predicate.
11000 2018-05-29 Martin Sebor <msebor@redhat.com>
11001 Richard Biener <rguenther@suse.de>
11004 * calls.c (get_size_range): Call determine_value_range instead
11005 of get_value_range..
11006 * tree-vrp.h (determine_value_range): Declared new function.
11007 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
11009 2018-05-29 Richard Biener <rguenther@suse.de>
11011 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
11012 sure to use non-pattern stmts for get_earlier_stmt arguments.
11013 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
11014 called on pattern stmts.
11015 (get_later_stmt): Likewise.
11017 2018-05-29 Martin Liska <mliska@suse.cz>
11019 PR gcov-profile/85759
11020 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
11023 2018-05-29 Jakub Jelinek <jakub@redhat.com>
11025 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
11027 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
11031 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11032 VEC_PACK_FLOAT_EXPR): New tree codes.
11033 * tree-pretty-print.c (op_code_prio): Handle
11034 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11035 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11036 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11037 * tree-inline.c (estimate_operator_cost): Likewise.
11038 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
11039 * fold-const.c (const_binop): Likewise.
11040 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
11041 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11042 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
11043 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
11044 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11045 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11046 * expr.c (expand_expr_real_2): Likewise.
11047 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
11048 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
11049 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
11051 * optabs.c (expand_widen_pattern_expr): For
11052 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
11053 sign from result type rather than operand's type.
11054 (expand_binop_directly): For vec_packu_float_optab and
11055 vec_packs_float_optab allow result type to be different from operand's
11057 * optabs-tree.c (optab_for_tree_code): Handle
11058 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11059 VEC_PACK_FLOAT_EXPR. Formatting fixes.
11060 * tree-vect-generic.c (expand_vector_operations_1): Handle
11061 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11062 VEC_PACK_FLOAT_EXPR.
11063 * tree-vect-stmts.c (supportable_widening_operation): Handle
11065 (supportable_narrowing_operation): Handle FLOAT_EXPR.
11066 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
11067 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
11068 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
11069 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
11071 (vec_pack<floatprefix>_float_<mode>): New expander.
11072 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
11074 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
11075 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
11076 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
11077 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
11078 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
11080 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
11081 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
11082 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11083 VEC_PACK_FLOAT_EXPR): Document.
11085 2018-05-29 Richard Biener <rguenther@suse.de>
11087 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
11089 (stmt_vec_info_vec): Make pointer.
11090 (init_stmt_vec_info_vec): Remove.
11091 (free_stmt_vec_info_vec): Likewise.
11092 (set_stmt_vec_info_vec): New function.
11093 (free_stmt_vec_infos): Likewise.
11094 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
11095 (set_vinfo_for_stmt): Likewise.
11096 (get_earlier_stmt): Likewise.
11097 (get_later_stmt): Likewise.
11098 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
11099 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
11100 (vec_info::~vec_info): Free stmt_vec_infos.
11101 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
11102 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
11103 (pass_slp_vectorize::execute): Likewise.
11104 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
11105 (free_stmt_vec_info_vec): Likewise.
11106 (set_stmt_vec_info_vec): New function.
11107 (free_stmt_vec_infos): Likewise.
11108 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
11109 the global stmt_vec_info_vec.
11110 * tree-parloops.c (gather_scalar_reductions): Use
11111 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
11114 2018-05-29 Richard Biener <rguenther@suse.de>
11116 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
11118 2018-05-29 Martin Liska <mliska@suse.cz>
11119 David Malcolm <dmalcolm@redhat.com>
11121 * vec.c (test_reverse): New.
11122 (vec_c_tests): Add new test.
11123 * vec.h (vl_ptr>::reverse): New function.
11125 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
11127 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
11129 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
11132 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
11134 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
11136 2018-05-28 Richard Biener <rguenther@suse.de>
11138 PR tree-optimization/85933
11139 * tree-vect-data-refs.c (vect_record_base_alignments): Only
11140 look at stmts marked as vectorizable.
11142 2018-05-28 Richard Biener <rguenther@suse.de>
11144 PR tree-optimization/85934
11145 * tree-vect-generic.c (expand_vector_operations_1): Hoist
11146 vector boolean check before scalar optimization.
11148 2018-05-28 Jakub Jelinek <jakub@redhat.com>
11150 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
11153 2018-05-28 Mark Wielaard <mark@klomp.org>
11155 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
11156 if it is an expression containing a minus sign.
11158 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
11160 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
11162 2018-05-27 Paul Koning <ni1d@arrl.net>
11164 * config/pdp11/pdp11.md (truncsihi2): Remove.
11166 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
11167 Chung-Ju Wu <jasonwucj@gmail.com>
11169 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
11171 (unaligned_store_dw): Ditto.
11172 * config/nds32/nds32-memory-manipulation.c
11173 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
11174 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
11175 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
11176 (emit_setmem_word_loop): Rename to ...
11177 (emit_setmem_doubleword_loop): ... this.
11178 (nds32_gen_dup_4_byte_to_word_value): New function.
11179 (nds32_gen_dup_8_byte_to_double_word_value): New function.
11180 (nds32_expand_setmem_loop): Refine implementation.
11181 (nds32_expand_setmem_loop_v3m): Ditto.
11182 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
11185 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
11187 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
11189 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
11191 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
11192 (nds32_init_machine_status): Initialize machine->attr_naked_p and
11193 machine->attr_no_prologue_p.
11194 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
11195 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
11196 (nds32_expand_epilogue): Consider attr_naked_p.
11197 (nds32_expand_epilogue_v3pop): Likewise.
11198 (nds32_can_use_return_insn): Likewise.
11199 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
11200 attr_no_prologue_p fields.
11201 * config/nds32/nds32.opt (mret-in-naked-func): New option.
11203 2018-05-27 Jakub Jelinek <jakub@redhat.com>
11206 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
11208 * config/i386/sse.md
11209 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
11211 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
11213 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
11215 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
11217 (*<floatsuffix>floatv2div2sf2): Rename to ...
11218 (*float<floatunssuffix>v2div2sf2): ... this.
11219 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
11220 (float<floatunssuffix>v2div2sf2_mask): ... this.
11221 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
11222 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
11223 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
11225 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
11227 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11229 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11231 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11233 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11235 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
11236 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
11237 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
11238 gen_ufix_truncv8dfv8si2.
11239 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
11240 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
11241 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
11242 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
11243 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
11244 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
11245 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
11246 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
11248 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
11252 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
11254 2018-05-25 Jim Wilson <jimw@sifive.com>
11256 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
11257 * config/riscv/riscv.c (struct machine_function): Add
11258 interrupt_handler_p and attribute_checked_p fields.
11259 (riscv_attribute_table): Add interrupt.
11260 (riscv_interrupt_type_p): New.
11261 (riscv_save_reg_p): Save extra regs for interrupt handler.
11262 (riscv_use_save_libcall): Return false for interrupt handler.
11263 (riscv_first_stack_step): Add forward declaration.
11264 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
11265 for interrupt handler with large frame. Use it for saved reg list.
11266 (riscv_expand_prologue): Move flag_stack_usage_info support to
11267 eliminate duplication.
11268 (riscv_expand_epilogue): Generate mret for interrupt handler.
11269 (riscv_epilogue_uses): New.
11270 (riscv_can_use_return_insn): Return false for interrupt handler.
11271 (riscv_function_ok_for_sibcall): Likewise.
11272 (riscv_set_current_function): Add interrupt handler support.
11273 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
11274 * config/riscv/riscv.md (UNSPECV_MRET): New.
11276 (riscv_frflags, riscv_fsflags): Use tab after opcode.
11278 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
11280 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
11282 PR tree-optimization/85712
11283 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
11284 this candidate has already been replaced in-situ by a copy.
11286 2018-05-25 Jason Merrill <jason@redhat.com>
11288 PR c++/80485 - inline function non-zero address.
11289 * symtab.c (nonzero_address): Check DECL_COMDAT.
11291 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
11294 * config/alpha/alpha.md (ashlsi3): New insn pattern.
11295 (*ashlsi_se): Rename from *ashldi_se. Define as sign
11296 extension of SImode operation. Use const123_operand predicate.
11297 (*saddsi_1): Remove.
11298 (*saddl_se_1): Ditto.
11299 (*ssubsi_1): Ditto.
11300 (*ssubl_se_1): Ditto.
11301 * config/alpha/predicates.md (const123_operand): New predicate.
11302 * config/alpha/constraints.md (P): Use IN_RANGE.
11304 2018-05-25 Richard Biener <rguenther@suse.de>
11306 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
11308 (ref_maybe_used_by_stmt_p): Likewise.
11309 (stmt_may_clobber_ref_p): Likewise.
11310 (stmt_may_clobber_ref_p_1): Likewise.
11311 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
11313 (ref_maybe_used_by_stmt_p): Likewise.
11314 (stmt_may_clobber_ref_p): Likewise.
11315 (stmt_may_clobber_ref_p_1): Likewise.
11316 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
11317 the alias oracle to disambiguate DRs with stmts DR analysis
11319 (vect_analyze_data_refs): Do not give up on not analyzable
11320 DRs for BB vectorization. Remove code truncating the dataref
11323 2018-05-25 Jakub Jelinek <jakub@redhat.com>
11326 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
11327 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
11328 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
11330 2018-05-25 Richard Biener <rguenther@suse.de>
11332 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
11333 function, combining stmt data ref gathering and fatal analysis
11335 (vect_analyze_data_refs): Remove now redudnant code and simplify.
11336 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
11337 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
11338 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
11339 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
11341 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11343 PR tree-optimization/85720
11344 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
11345 SCC if all partitions are builtins.
11346 (version_loop_by_alias_check): New parameter. Generate cancelable
11347 runtime alias check if all partitions are builtins.
11348 (distribute_loop): Update call to above function.
11350 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11352 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
11353 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
11354 (parm_default_def_partition_arg): Ditto.
11355 (set_parm_default_def_partition): Ditto.
11356 (get_parm_default_def_partitions): Ditto and make it static.
11357 (get_undefined_value_partitions): Ditto and make it static.
11358 (remove_ssa_form): Refactor call to init_var_map here.
11359 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
11360 computation for loop region.
11361 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
11362 (register_default_def): Delete.
11363 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
11364 (parm_default_def_partition_arg): Ditto.
11365 (set_parm_default_def_partition): Ditto.
11366 (get_parm_default_def_partitions): Ditto and make it static.
11367 (get_undefined_value_partitions): Ditto and make it static.
11368 (coalesce_with_default, coalesce_with_default): Update comment.
11369 (create_coalesce_list_for_region): New func factored out from
11370 create_outofssa_var_map.
11371 (populate_coalesce_list_for_outofssa): New func factored out from
11372 create_outofssa_var_map and coalesce_ssa_name.
11373 (create_outofssa_var_map): Delete.
11374 (coalesce_ssa_name): Refactor to support live range computation.
11375 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
11376 (get_parm_default_def_partitions): Delete.
11377 (get_undefined_value_partitions): Ditto.
11378 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
11379 computation for loop region.
11380 (new_tree_live_info, loe_visit_block): Ditto.
11381 (live_worklist, set_var_live_on_entry): Ditto.
11382 (calculate_live_on_exit, verify_live_on_entry): Ditto.
11383 * tree-ssa-live.h (struct _var_map): New fields.
11384 (init_var_map): Change decl.
11385 (region_contains_p): New.
11387 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11389 * tree-ssa-live.h (live_merge_and_clear): Delete.
11391 2018-05-25 Richard Biener <rguenther@suse.de>
11394 * tree-dump.c (dequeue_and_dump): Remove access to removed
11395 operand 2 of a SWITCH_EXPR.
11397 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11399 * doc/sourcebuild.texi (vect_double_cond_arith): Include
11400 multiplication and division.
11401 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
11402 (cond_udiv@var{m}, cond_umod@var{m}): Document.
11403 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
11404 (cond_udiv_optab, cond_umod_optab): New optabs.
11405 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
11406 (IFN_COND_RDIV): New internal functions.
11407 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
11408 TRUNC_MOD_EXPR and RDIV_EXPR.
11409 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
11410 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
11412 (SVE_INT_BINARY): Include mult.
11413 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
11414 (optab, sve_int_op): Handle mult.
11415 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
11417 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
11418 for SVE_INT_BINARY_SD.
11420 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11422 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
11423 (optab, sve_int_op): Handle div and udiv.
11424 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
11425 for SVE_INT_BINARY_SD.
11426 (*<optab><mode>3): New insn for the same.
11428 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11430 * tree-vect-patterns.c: Include predict.h.
11431 (vect_recog_divmod_pattern): Restrict check for division support
11432 to when optimizing for size.
11434 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11436 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
11437 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
11438 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
11439 (gimple_match_op::set_op): Likewise.
11440 (gimple_resimplify4): Declare.
11441 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
11442 (expr::gen_transform): Likewise.
11443 (decision_tree::gen): Generate a simplification routine for 4 operands.
11444 * gimple-match-head.c (gimple_simplify): Add an overload for
11445 4 operands. In the top-level function, handle up to 4 call
11446 arguments and call gimple_resimplify4.
11447 (gimple_resimplify4): New function.
11448 (build_call_internal): Pass a fourth operand.
11449 (maybe_push_to_seq): Likewise.
11450 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
11451 Fold VEC_COND_EXPRs of an operation and a default value into
11452 an IFN_COND_* function if possible.
11453 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
11455 (SVE_COND_FP_BINARY): Include them.
11456 (optab, sve_fp_op): Handle them.
11457 (SVE_INT_BINARY_REV): New code iterator.
11458 (SVE_COND_FP_BINARY_REV): New int iterator.
11459 (commutative): New int attribute.
11460 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
11462 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
11464 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
11465 (*cond_<optab><mode>): New patterns for reversed operands.
11467 2018-05-25 Richard Biener <rguenther@suse.de>
11469 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
11470 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
11471 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
11472 (STMT_VINFO_GROUPED_ACCESS): Adjust.
11473 * tree-vect-data-refs.c (everywhere): Adjust users.
11474 * tree-vect-loop.c (everywhere): Likewise.
11475 * tree-vect-slp.c (everywhere): Likewise.
11476 * tree-vect-stmts.c (everywhere): Likewise.
11477 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
11479 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11481 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
11483 (gcc_cv_as_section_exclude): ... this.
11484 Try Solaris as #exclude syntax.
11485 * configure: Regenerate.
11486 * config.in: Regenerate.
11487 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
11489 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
11490 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
11492 * varasm.c (default_elf_asm_named_section): Don't check if
11493 HAVE_GAS_SECTION_EXCLUDE is defined.
11495 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11497 * doc/md.texi: Update the documentation of the cond_* optabs
11498 to mention the new final operand. Fix GET_MODE_NUNITS call.
11499 Describe the scalar case too.
11500 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
11501 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
11503 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
11504 (get_conditional_internal_fn): Update comment.
11505 * tree-vect-loop.c (vectorizable_reduction): Pass the original
11506 accumulator value as a final argument to conditional functions.
11507 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
11508 a define_expand and add an "else" operand. Assert for now that
11509 the else operand is equal to operand 2. Use SVE_INT_BINARY and
11510 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
11511 (*cond_<optab><mode>): New patterns.
11512 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
11513 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11514 (UNSPEC_COND_EOR): Delete.
11515 (optab): Remove associated mappings.
11516 (SVE_INT_BINARY): New code iterator.
11517 (sve_int_op): Remove int attribute and add "minus" to the code
11519 (SVE_COND_INT_OP): Delete.
11520 (SVE_COND_FP_OP): Rename to...
11521 (SVE_COND_FP_BINARY): ...this.
11523 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11525 * optabs.c (can_reuse_operands_p): New function.
11526 (maybe_legitimize_operands): Try to reuse the results for
11529 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
11531 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
11532 Add {q} suffix to insn mnemonic.
11534 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11536 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
11537 (msp430_warn_func_return): New.
11539 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
11541 * fold-const.c (tree_nonzero_bits): New function.
11542 * fold-const.h (tree_nonzero_bits): Likewise.
11543 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
11544 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
11546 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
11550 * varasm.c (assemble_alias): Check ifunc_resolver only on
11553 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
11556 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
11557 when memory input operand is handled.
11559 2018-05-24 Luis Machado <luis.machado@linaro.org>
11561 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
11563 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
11565 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
11567 * match.pd: Delay FMA folds until after vectorization.
11569 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
11572 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11573 address check not strict.
11575 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
11577 * gimple-match.h (gimple_match_op): New class.
11578 (mprts_hook): Replace parameters with a gimple_match_op *.
11579 (maybe_build_generic_op): Likewise.
11580 (gimple_simplified_result_is_gimple_val): Replace parameters with
11581 a const gimple_match_op *.
11582 (gimple_simplify): Replace code_helper * and tree * parameters with
11583 a gimple_match_op * parameter.
11584 (gimple_resimplify1): Replace code_helper *, tree and tree *
11585 parameters with a gimple_match_op * parameter.
11586 (gimple_resimplify2): Likewise.
11587 (gimple_resimplify3): Likewise.
11588 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
11589 parameters with a gimple_match_op * parameter.
11590 * gimple-match-head.c (gimple_simplify): Change prototypes of
11591 auto-generated functions to take a gimple_match_op * instead of
11592 separate code_helper * and tree * parameters. Make the same
11593 change in the top-level overload and update calls to the
11594 gimple_resimplify routines. Update calls to the auto-generated
11595 functions and to maybe_push_res_to_seq in the publicly-facing
11596 operation-specific gimple_simplify overloads.
11597 (gimple_match_op::MAX_NUM_OPS): Define.
11598 (gimple_resimplify1): Replace rcode and ops with a single res_op
11599 parameter. Update call to gimple_simplify.
11600 (gimple_resimplify2): Likewise.
11601 (gimple_resimplify3): Likewise.
11602 (mprts_hook): Replace parameters with a gimple_match_op *.
11603 (maybe_build_generic_op): Likewise.
11604 (build_call_internal): Replace type, nargs and ops with
11605 a gimple_match_op *.
11606 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
11607 with a single gimple_match_op *. Update calls to mprts_hook,
11608 build_call_internal and gimple_simplified_result_is_gimple_val.
11609 Factor out code that is common to the tree_code and combined_fn cases.
11610 * genmatch.c (expr::gen_transform): Replace tem_code and
11611 tem_ops with a gimple_match_op called tem_op. Update calls
11612 to the gimple_resimplify functions and maybe_push_res_to_seq.
11613 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
11614 res_ops. Update call to the gimple_resimplify functions.
11615 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
11616 (decision_tree::gen): Make the functions take a gimple_match_op *
11617 called res_op instead of separate res_code and res_ops parameters.
11618 Update call accordingly.
11619 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
11620 and ops with a single res_op parameter. Update calls to
11621 maybe_build_generic_op and maybe_push_res_to_seq.
11622 (fold_stmt_1): Update calls to gimple_simplify and
11623 replace_stmt_with_simplification.
11624 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
11625 and gimple_simplified_result_is_gimple_val.
11626 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
11628 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
11629 with a gimple_match_op *.
11630 (vn_nary_build_or_lookup): Likewise. Update call to
11631 vn_nary_build_or_lookup_1.
11632 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
11633 gimple_match_op *. Update calls to the gimple_resimplify routines
11634 and to gimple_simplified_result_is_gimple_val.
11635 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
11636 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
11637 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
11638 (visit_nary_op): Likewise.
11639 (visit_reference_op_load): Likewise.
11641 2018-05-23 Luis Machado <luis.machado@linaro.org>
11643 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
11644 modifier for printing the step amount.
11646 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
11649 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
11652 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
11654 * doc/sourcebuild.texi (Endianness): New subsubsection.
11656 2018-05-23 Luis Machado <luis.machado@linaro.org>
11658 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11659 <prefetch_dynamic_strides>: New const bool field.
11660 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11661 prefetch_dynamic_strides.
11662 (exynosm1_prefetch_tune): Likewise.
11663 (thunderxt88_prefetch_tune): Likewise.
11664 (thunderx_prefetch_tune): Likewise.
11665 (thunderx2t99_prefetch_tune): Likewise.
11666 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
11668 (aarch64_override_options_internal): Update to set
11669 PARAM_PREFETCH_DYNAMIC_STRIDES.
11670 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11671 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11672 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11673 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11674 prefetch-dynamic-strides setting.
11676 2018-05-23 Luis Machado <luis.machado@linaro.org>
11678 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11679 <minimum_stride>: New const int field.
11680 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11681 minimum_stride field defaulting to -1.
11682 (exynosm1_prefetch_tune): Likewise.
11683 (thunderxt88_prefetch_tune): Likewise.
11684 (thunderx_prefetch_tune): Likewise.
11685 (thunderx2t99_prefetch_tune): Likewise.
11686 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
11687 <default_opt_level>: Set to 3.
11688 (aarch64_override_options_internal): Update to set
11689 PARAM_PREFETCH_MINIMUM_STRIDE.
11690 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11691 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11692 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11693 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11694 stride is constant and is below the minimum stride threshold.
11696 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11698 * config/arm/arm-cpus.in (mode26): Delete.
11699 (armv4): Delete mode26 reference.
11700 * config/arm/arm.c (arm_configure_build_target): Delete use of
11703 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
11705 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
11707 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
11708 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
11710 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
11711 (floatunsdidf2): Ditto.
11713 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
11715 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
11716 (fixuns_trunc<mode>si2_avx512f): Ditto.
11717 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
11718 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
11719 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
11721 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
11723 PR rtl-optimization/79985
11724 * df-scan.c (df_insn_refs_collect): Remove special case for
11725 global registers and asm statements.
11727 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
11729 * extend.texi (Global Register Variables): Rewrite the bullet list.
11730 Note that the register is available for allocation. Note that access
11731 via inline asm must use constraints. Add note about async-signal
11732 handlers. Remove paragraph about automagic register selection.
11734 2018-05-23 Richard Biener <rguenther@suse.de>
11736 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
11737 of fixed offset from memset VN.
11739 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
11741 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
11742 first_interp field.
11743 (alloc_cand_and_find_basis): Initialize first_interp field.
11744 (slsr_process_mul): Modify first_interp field.
11745 (slsr_process_add): Likewise.
11746 (slsr_process_cast): Modify first_interp field for each new
11748 (slsr_process_copy): Likewise.
11749 (dump_candidate): Dump first_interp field.
11750 (replace_mult_candidate): Process all interpretations, not just
11752 (replace_rhs_if_not_dup): Likewise.
11753 (replace_one_candidate): Likewise.
11755 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
11757 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
11759 (aarch64_needs_frame_chain): New function.
11760 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
11762 2018-05-23 Sudakshina Das <sudi.das@arm.com>
11765 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
11766 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
11767 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
11768 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
11769 as true for strict-align.
11770 (aarch64_can_inline_p): Perform checks even when callee has no
11771 attributes to check for strict alignment.
11772 * doc/extend.texi (AArch64 Function Attributes): Document
11774 * doc/invoke.texi: (AArch64 Options): Likewise.
11776 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
11778 PR tree-optimization/85853
11779 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
11780 the handling of the root of the node to...
11781 (vect_slp_analyze_node_operations_1): ...this new function,
11782 and run the whole thing with the child nodes' def types
11783 set according to their SLP node's def type.
11785 2018-05-23 Richard Biener <rguenther@suse.de>
11787 PR middle-end/85874
11788 * tree-data-ref.c (create_runtime_alias_checks): Defer
11789 and ignore overflow warnings.
11791 2018-05-23 Yury Gribov <tetra2005@gmail.com>
11793 PR tree-optimization/85822
11794 * tree-vrp.c (is_masked_range_test): Fix handling of negative
11797 2018-05-23 Richard Biener <rguenther@suse.de>
11799 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
11800 memset constants via native_interpret_expr.
11802 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
11805 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
11807 (cgraph_node::create_alias): Likewise.
11808 (cgraph_node::get_availability): Check ifunc_resolver instead
11809 of looking up ifunc attribute.
11810 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
11811 * varasm.c (do_assemble_alias): Likewise.
11812 (assemble_alias): Likewise.
11813 (default_binds_local_p_3): Likewise.
11814 * cgraph.h (cgraph_node): Add ifunc_resolver.
11815 (cgraph_node::only_called_directly_or_aliased_p): Return false
11816 for IFUNC resolver.
11817 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
11819 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
11820 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
11821 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
11822 instead of looking up ifunc attribute.
11824 2018-05-22 Luis Machado <luis.machado@linaro.org>
11826 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
11828 2018-05-22 Martin Sebor <msebor@redhat.com>
11830 PR middle-end/85359
11831 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
11832 only when expasion succeeds.
11833 (expand_builtin_strcmp): Same.
11834 (expand_builtin_strncmp): Same.
11836 2018-05-22 Martin Sebor <msebor@redhat.com>
11838 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
11840 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
11841 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11843 * config/aarch64/aarch64-ldpstp.md: Replace uses of
11844 aarch64_mem_pair_operand with memory_operand and delete operand swapping
11846 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
11847 Add check for legitimate_address.
11848 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
11849 (aarch64_swap_ldrstr_operands): New.
11850 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
11853 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
11854 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11856 * config/aarch64/aarch64.md: New patterns to generate stp
11858 (store_pair_sw, store_pair_dw): New patterns to generate stp for
11859 single words and double words.
11860 (load_pair_sw, load_pair_dw): Likewise.
11861 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
11863 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
11865 * config/aarch64/aarch64-ldpstp.md: Modify peephole
11866 for different mode ldpstp and add peephole for merged zero stores.
11867 Likewise for loads.
11868 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
11870 (aarch64_gen_store_pair): Rename calls to match new patterns.
11871 (aarch64_gen_load_pair): Rename calls to match new patterns.
11872 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
11873 (load_pair<DREG:mode><DREG2:mode>): ... This.
11874 (store_pair<mode>): Rename to...
11875 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
11876 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
11877 New mode iterators.
11878 (V_INT_EQUIV): Handle SImode.
11879 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
11882 2018-05-22 Martin Sebor <msebor@redhat.com>
11885 * calls.c (maybe_warn_nonstring_arg): Use string length to set
11886 or ajust the presumed bound on an operation to avoid unnecessary
11889 2018-05-22 Martin Sebor <msebor@redhat.com>
11891 PR tree-optimization/85826
11892 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
11893 assuming that a DECL necesarily has a constant size.
11895 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
11897 PR middle-end/85862
11898 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
11900 2018-05-22 Richard Biener <rguenther@suse.de>
11902 PR tree-optimization/85834
11903 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
11904 non-constant and non-zero memset arguments.
11906 2018-05-22 Martin Liska <mliska@suse.cz>
11909 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
11911 2018-05-22 Richard Biener <rguenther@suse.de>
11913 PR tree-optimization/85863
11914 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
11915 comparisons when vectype is specified.
11916 (vectorizable_condition): Do not specify vectype for
11917 vect_is_simple_cond when SLP vectorizing.
11919 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
11922 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
11923 define __ibm128 as long double.
11924 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
11925 as a distinct type when IEEE 128-bit support is enabled.
11926 (init_float128_ieee): Fix up conversions between IFmode and IEEE
11927 128-bit types to use the correct functions.
11928 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
11929 convert between 128-bit floating point types that have different
11930 modes but the same representation, instead of using gen_lowpart to
11932 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
11934 (IFKF_reg): New attributes to give the register constraints for
11936 (extend<mode>tf2_internal): New insns to mark an explicit
11937 conversion between 128-bit floating point types that have a
11938 different mode but share the same representation.
11940 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
11942 PR tree-optimization/85814
11943 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
11944 a null return from get_strinfo when unsharing the next
11945 strinfo in the chain.
11947 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
11950 * varasm.c (weak_finish): Clean up weak_decls.
11952 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11954 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
11955 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
11956 UNSPEC_UADALP values.
11957 * config/aarch64/iterators.md (ABAL): New int iterator.
11960 (sur): Add mappings for the above.
11961 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
11963 (aarch64_<sur>abal<mode>_4): Likewise.
11964 (aarch64_<sur>adalp<mode>_3): Likewise.
11965 (<sur>sadv16qi): New define_expand.
11967 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
11969 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
11970 (*movdf_internal): Ditto.
11971 (*rcpsf2_sse): Ditto.
11972 (*rsqrtsf2_sse): Ditto.
11973 (*sqrt<mode>2_sse): Ditto.
11975 2018-05-21 Tamar Christina <tamar.christina@arm.com>
11977 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
11979 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
11980 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
11981 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
11982 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
11984 * config/aarch64/arm_neon.h: Likewise.
11985 * config/aarch64/iterators.md (VQ_I): New.
11987 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
11989 * config.gcc: Add arc/t-multilib-linux to tmake_file for
11991 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
11994 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
11996 * config/nds32/constraints.md (S): New constraint.
11997 * config/nds32/nds32.md (call_internal): Use constraint S.
11998 (call_value_internal): Likewise.
11999 (sibcall_internal): Likewise.
12000 (sibcall_value_internal): Likewise.
12002 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
12003 Chung-Ju Wu <jasonwucj@gmail.com>
12005 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
12006 into consideration.
12008 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
12009 Chung-Ju Wu <jasonwucj@gmail.com>
12011 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
12012 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
12013 (nds32_rtx_costs_impl): Simplify.
12014 (nds32_address_cost_impl): Simplify.
12015 (nds32_init_rtx_costs): New function.
12016 (nds32_rtx_costs_speed_prefer): Likewise.
12017 (nds32_rtx_costs_size_prefer): Likewise.
12018 (nds32_address_cost_speed_prefer): Likewise.
12019 (nds32_address_cost_speed_fwprop): Likewise.
12020 (nds32_address_cost_size_prefer): Likewise.
12021 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
12022 * config/nds32/nds32.c (nds32_option_override): Use
12023 nds32_init_rtx_costs function.
12025 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
12027 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
12028 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
12029 (TARGET_PIPELINE_N8): Likewise.
12030 (TARGET_PIPELINE_N10): Likewise.
12031 (TARGET_PIPELINE_N13): Likewise.
12032 (TARGET_PIPELINE_GRAYWOLF): Likewise.
12034 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
12036 * config/nds32/nds32-fpu.md: Update copyright year.
12038 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12040 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
12042 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12044 * config/nds32/nds32.c
12045 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
12046 * config/nds32/nds32.opt (minline-asm-r15): New option.
12048 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12050 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
12052 * config/nds32/nds32.md (abssi2): New pattern.
12054 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
12056 * config/i386/i386.md (rex64namesuffix): New mode attribute.
12057 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
12058 Merge insn pattern from sse_cvtsi2ss<round_name> and
12059 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
12060 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
12061 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
12062 using SWI48 mode iterator.
12063 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
12064 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
12065 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
12066 pattern from sse_cvttss2si<round_saeonly_name>
12067 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
12068 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
12069 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
12070 using SWI48 mode iterator.
12071 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
12072 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
12073 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
12074 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
12075 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
12076 using SWI48 mode iterator.
12077 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
12078 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
12079 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
12080 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
12081 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
12082 SWI48 mode iterator.
12083 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
12084 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
12085 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
12086 pattern from sse_cvttsd2si<round_saeonly_name>
12087 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
12089 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12091 * config/nds32/nds32-md-auxiliary.c
12092 (nds32_valid_smw_lwm_base_p): Refine.
12093 (nds32_output_smw_single_word): Refine.
12094 (nds32_output_smw_double_word): New.
12095 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
12097 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12099 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
12100 (nds32_output_stack_pop): Refine.
12101 (nds32_expand_unaligned_load): Refine.
12102 (nds32_expand_unaligned_store): Refine.
12104 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12105 Chung-Ju Wu <jasonwucj@gmail.com>
12107 * config/nds32/constants.md: Add TP_REGNUM constant.
12108 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
12109 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
12111 * config/nds32/nds32-doubleword.md: Consider flag_pic.
12112 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
12113 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
12114 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
12115 and PIC code generation.
12116 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
12118 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
12120 * config/nds32/nds32.md: Support TLS and PIC.
12121 * config/nds32/nds32.c: Support TLS and PIC.
12122 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
12123 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
12126 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12128 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
12129 mode with E_ prefix.
12131 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12132 Chung-Ju Wu <jasonwucj@gmail.com>
12134 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
12135 * config/nds32/nds32-md-auxiliary.c
12136 (symbolic_reference_mentioned_p): New.
12137 (nds32_legitimize_ict_address): New.
12138 (nds32_expand_ict_move): New.
12139 (nds32_indirect_call_referenced_p): New.
12140 (nds32_symbol_binds_local_p): Delete.
12141 (nds32_long_call_p): Modify.
12142 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
12143 * config/nds32/nds32-protos.h
12144 (symbolic_reference_mentioned_p): Declare.
12145 (nds32_legitimize_ict_address): Declare.
12146 (nds32_expand_ict_move): Declare.
12147 (nds32_indirect_call_referenced_p): Declare.
12148 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
12149 (nds32_relax_group): Use nds32_ict_const_p as condition.
12150 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
12151 (nds32_asm_file_start): Output ict_model directive in asm code.
12152 (nds32_legitimate_address_p): Consider indirect call.
12153 (nds32_print_operand): Consider indirect call.
12154 (nds32_print_operand_address): Consider indirect call.
12155 (nds32_insert_attributes): Handle "indirect_call" attribute.
12156 (TARGET_LEGITIMATE_ADDRESS_P): Define.
12157 (TARGET_LEGITIMATE_CONSTANT_P): Define.
12158 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
12159 (TARGET_DELEGITIMIZE_ADDRESS): Define.
12160 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
12161 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
12162 (TARGET_ICT_MODEL_SMALL): Define.
12163 (TARGET_ICT_MODEL_LARGE): Define.
12164 * config/nds32/nds32.md (movsi): Consider ict model.
12165 (call, call_value): Consider ict model.
12166 (sibcall, sibcall_value): Consider ict model.
12167 * config/nds32/nds32.opt (mict-model): New option.
12168 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
12171 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
12172 Monk Chiang <sh.chiang04@gmail.com>
12173 Jim Wilson <jimw@sifive.com>
12175 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
12176 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
12177 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
12178 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
12179 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
12180 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
12181 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
12182 compute save_libcall_adjustment properly.
12183 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
12184 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
12185 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
12186 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
12187 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
12188 (ABI_SPEC): Handle mabi=ilp32e.
12189 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
12190 (RVE): Add RVE mask.
12191 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
12192 <-march>: Add rv32e as an example.
12194 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
12197 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
12198 (intra_create_variable_infos): Handle C++ constructors.
12200 2018-05-18 Martin Liska <mliska@suse.cz>
12202 * passes.def: Remove a redundant pass.
12204 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
12207 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
12209 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12211 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
12213 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
12215 (armv2, armv2a, armv3, armv3m): Delete architectures.
12216 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
12217 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
12218 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
12220 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
12221 (*mulsidi3adddi): Likewise.
12222 (mulsidi3): Likewise.
12223 (*mulsidi3_nov6): Likewise.
12224 (umulsidi3): Likewise.
12225 (umulsidi3_nov6): Likewise.
12226 (umaddsidi4): Likewise.
12227 (*umulsidi3adddi): Likewise.
12228 (smulsi3_highpart): Likewise.
12229 (*smulsi3_highpart_nov6): Likewise.
12230 (umulsi3_highpart): Likewise.
12231 (*umulsi3_highpart_nov6): Likewise.
12232 * config/arm/arm.h (arm_arch3m): Delete.
12233 * config/arm/arm.c (arm_arch3m): Delete.
12234 (arm_option_override_internal): Update armv3-related comment.
12235 (arm_configure_build_target): Delete use of isa_bit_mode32.
12236 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
12237 (arm_rtx_costs_internal): Delete check of arm_arch3m.
12238 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
12239 (mulsa3): Likewise.
12240 (mulusa3): Likewise.
12241 * config/arm/arm-protos.h (arm_arch3m): Delete.
12242 * config/arm/arm-tables.opt: Regenerate.
12243 * config/arm/arm-tune.md: Likewise.
12244 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
12245 deleted architectures.
12247 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12249 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
12250 (armv5t, armv5te): New features.
12251 (ARMv5, ARMv5e): Delete fgroups.
12252 (ARMv5t, ARMv5te): Adjust for above changes.
12253 (ARMv6m): Likewise.
12254 (armv5, armv5e): Delete arches.
12255 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
12257 (*call_reg_arm): Likewise.
12258 (*call_value_reg_armv5): Likewise.
12259 (*call_value_reg_arm): Likewise.
12260 (*call_symbol): Likewise.
12261 (*call_value_symbol): Likewise.
12262 (*sibcall_insn): Likewise.
12263 (*sibcall_value_insn): Likewise.
12264 (clzsi2): Likewise.
12265 (prefetch): Likewise.
12266 (define_split and define_peephole2 dependent on arm_arch5):
12268 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
12270 (TARGET_ARM_QBIT): Likewise.
12271 (TARGET_DSP_MULTIPLY): Likewise.
12272 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
12273 (arm_arch5, arm_arch5e): Delete.
12274 (arm_arch5t, arm_arch5te): Declare.
12275 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
12276 (arm_arch5t): Declare.
12277 (arm_option_reconfigure_globals): Update for the above.
12278 (arm_options_perform_arch_sanity_checks): Update comment, replace
12279 use of arm_arch5 with arm_arch5t.
12280 (use_return_insn): Likewise.
12281 (arm_emit_call_insn): Likewise.
12282 (output_return_instruction): Likewise.
12283 (arm_final_prescan_insn): Likewise.
12284 (arm_coproc_builtin_available): Likewise.
12285 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
12286 arm_arch5e with arm_arch5t and arm_arch5te.
12287 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
12288 (arm_arch5t, arm_arch5te): Declare.
12289 * config/arm/arm-tables.opt: Regenerate.
12290 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
12291 * config/arm/t-multilib: Likewise.
12292 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
12293 instead of arm_arch5.
12294 (*call_reg_thumb1): Likewise.
12295 (*call_value_reg_thumb1_v5): Likewise.
12296 (*call_value_reg_thumb1): Likewise.
12297 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
12299 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
12301 2018-05-18 Martin Liska <mliska@suse.cz>
12303 PR gcov-profile/84846
12304 * doc/gcov.texi: Document -t option of gcov tool.
12306 2018-05-18 Martin Liska <mliska@suse.cz>
12308 PR gcov-profile/84846
12309 * gcov.c (print_usage): Add new -t option.
12310 (process_args): Handle the option.
12311 (generate_results): Use stdout as output when requested by
12314 2018-05-18 Martin Liska <mliska@suse.cz>
12316 PR gcov-profile/84846
12317 * coverage.c (coverage_init): Write PWD to .gcno file.
12318 * doc/gcov.texi: Document how working directory is printed.
12319 * gcov-dump.c (dump_gcov_file): Print PWD.
12320 * gcov.c (output_intermediate_file): Likewise.
12321 (read_graph_file): Read PWD string.
12322 (output_lines): Print PWD.
12324 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12326 PR middle-end/85817
12327 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
12328 for retval and return false if all args to phi are zero.
12330 2018-05-18 Richard Biener <rguenther@suse.de>
12332 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
12334 (evrp_dom_walker::before_dom_children): Call it.
12336 2018-05-18 Richard Biener <rguenther@suse.de>
12338 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
12339 results when processing array refs with variable index.
12341 2018-05-18 Toon Moene <toon@moene.org>
12343 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
12344 directly after that of -floop-interchange. Indicate that both
12345 options are enabled by default when specifying -O3.
12347 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12349 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
12350 iterator. Delete separate integer-mode vec_set<mode> expander.
12351 (aarch64_simd_vec_setv2di): Delete.
12352 (vec_setv2di): Delete.
12353 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
12354 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
12355 the "w, r" alternative.
12357 2018-05-18 Martin Liska <mliska@suse.cz>
12359 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
12360 * tree-pass.h (make_pass_lower_switch_O0): New function.
12361 * tree-switch-conversion.c (node_has_low_bound): Remove.
12362 (node_has_high_bound): Likewise.
12363 (node_is_bounded): Likewise.
12364 (class pass_lower_switch): Make it a template type and create
12366 (pass_lower_switch::execute): Add template argument.
12367 (make_pass_lower_switch): New function.
12368 (make_pass_lower_switch_O0): New function.
12369 (do_jump_if_equal): Remove.
12370 (emit_case_nodes): Simplify to just handle all 3 cases and leave
12371 all the hard work to tree optimization passes.
12373 2018-05-18 Martin Liska <mliska@suse.cz>
12375 * dbgcnt.c (limit_low): Renamed from limit.
12376 (limit_high): New variable.
12377 (dbg_cnt_is_enabled): Check for upper limit.
12378 (dbg_cnt): Adjust dumping.
12379 (dbg_cnt_set_limit_by_index): Add new argument for high
12381 (dbg_cnt_set_limit_by_name): Likewise.
12382 (dbg_cnt_process_single_pair): Parse new format.
12383 (dbg_cnt_process_opt): Use strtok.
12384 (dbg_cnt_list_all_counters): Remove 'value' and add
12386 * doc/invoke.texi: Document changes.
12388 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
12390 * doc/sourcebuild.texi (scalar_all_fma): Document.
12391 * tree.def (FMA_EXPR): Delete.
12392 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
12393 * internal-fn.c (ternary_direct): New macro.
12394 (expand_ternary_optab_fn): Likewise.
12395 (direct_ternary_optab_supported_p): Likewise.
12396 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
12397 * builtins.c (fold_builtin_fma): Delete.
12398 (fold_builtin_3): Don't call it.
12399 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
12400 * expr.c (expand_expr_real_2): Likewise.
12401 * fold-const.c (operand_equal_p): Likewise.
12402 (fold_ternary_loc): Likewise.
12403 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
12404 * gimple.c (DEFTREECODE): Likewise.
12405 * gimplify.c (gimplify_expr): Likewise.
12406 * optabs-tree.c (optab_for_tree_code): Likewise.
12407 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
12408 * tree-eh.c (operation_could_trap_p): Likewise.
12409 (stmt_could_throw_1_p): Likewise.
12410 * tree-inline.c (estimate_operator_cost): Likewise.
12411 * tree-pretty-print.c (dump_generic_node): Likewise.
12412 (op_code_prio): Likewise.
12413 * tree-ssa-loop-im.c (stmt_cost): Likewise.
12414 * tree-ssa-operands.c (get_expr_operands): Likewise.
12415 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
12416 * fold-const-call.h (fold_fma): Delete.
12417 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
12418 CFN_FNMA and CFN_FNMS.
12419 (fold_fma): Delete.
12420 * genmatch.c (combined_fn): New enum.
12421 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
12422 (commutative_op): New function.
12423 (commutate): Use it. Handle more than 2 operands.
12424 (dt_operand::gen_gimple_expr): Use commutative_op.
12425 (parser::parse_expr): Allow :c to be used with non-binary
12426 operators if the commutative operand is known.
12427 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
12428 CFN_FMS, CFN_FNMA and CFN_FNMS.
12429 (backprop::process_assign_use): Remove FMA_EXPR handling.
12430 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
12431 (gen_hsa_fma): New function.
12432 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
12433 IFN_FNMA and IFN_FNMS.
12434 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
12435 * gimple-fold.h (follow_all_ssa_edges): Declare.
12436 * gimple-fold.c (follow_all_ssa_edges): New function.
12437 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
12438 gimple_build interface and use follow_all_ssa_edges to fold the result.
12439 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
12440 instead of checking for optabs directly.
12441 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
12442 rather than FMA_EXPRs.
12443 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
12444 call to IFN_FMA instead of an FMA_EXPR.
12446 2018-05-17 Jim Wilson <jimw@sifive.com>
12448 * expr.c (do_tablejump): When converting index to Pmode, if we have a
12449 sign extended promoted subreg, and the range does not have the sign bit
12450 set, then do a sign extend.
12452 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
12453 test, check for sign extended subreg and/or constant operands, and
12454 do a sign extend in that case.
12456 2018-05-17 Steve Ellcey <sellcey@cavium.com>
12458 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
12459 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
12461 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
12462 Change logics_shift_reg to logics_shift_imm.
12463 (thunderx2t99_fp_loadpair_basic): Delete.
12464 (thunderx2t99_fp_storepair_basic): Delete.
12465 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
12466 (thunderx2t99_asimd_polynomial): Delete.
12467 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
12468 and neon_fp_mul_d_scalar_q.
12469 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
12470 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
12471 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
12472 (thunderx2t99_asimd_lut): Add missing tbl types.
12473 (thunderx2t99_asimd_ext): Delete.
12474 (thunderx2t99_asimd_load1_1_mult): Delete.
12475 (thunderx2t99_asimd_load1_2_mult): Delete.
12476 (thunderx2t99_asimd_load1_ldp): New.
12477 (thunderx2t99_asimd_load1): New.
12478 (thunderx2t99_asimd_load2): Add missing *load2* types.
12479 (thunderx2t99_asimd_load3): New.
12480 (thunderx2t99_asimd_load4): New.
12481 (thunderx2t99_asimd_store1_1_mult): Delete.
12482 (thunderx2t99_asimd_store1_2_mult): Delete.
12483 (thunderx2t99_asimd_store2_mult): Delete.
12484 (thunderx2t99_asimd_store2_onelane): Delete.
12485 (thunderx2t99_asimd_store_stp): New.
12486 (thunderx2t99_asimd_store1): New.
12487 (thunderx2t99_asimd_store2): New.
12488 (thunderx2t99_asimd_store3): New.
12489 (thunderx2t99_asimd_store4): New.
12491 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
12493 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
12494 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
12496 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
12497 Segher Boessenkool <segher@kernel.crashing.org>
12500 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
12503 2018-05-17 Richard Biener <rguenther@suse.de>
12505 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
12506 for pruning loop and prune defs feeding only already visited PHIs.
12508 2018-05-17 Richard Biener <rguenther@suse.de>
12510 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
12512 2018-05-17 Bin Cheng <bin.cheng@arm.com>
12513 Richard Biener <rguenther@suse.de>
12515 PR tree-optimization/85793
12516 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
12517 for VMAT_ELEMENTWISE.
12519 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
12521 * internal-fn.h (lookup_internal_fn): Declare
12522 * internal-fn.c (lookup_internal_fn): New function.
12523 * gimple.c (gimple_build_call_from_tree): Handle calls to
12524 internal functions.
12525 * gimple-pretty-print.c (dump_gimple_call): Print "." before
12526 internal function names.
12527 * tree-pretty-print.c (dump_generic_node): Likewise.
12528 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
12530 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
12532 * gimple-fold.h (gimple_build): Make the function forms take
12533 combined_fn rather than built_in_function.
12534 (gimple_simplify): Likewise.
12535 * gimple-match-head.c (gimple_simplify): Likewise.
12536 * gimple-fold.c (gimple_build): Likewise.
12537 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
12538 rather than gimple_build_call_internal.
12539 (get_initial_defs_for_reduction): Likewise.
12540 (vect_create_epilog_for_reduction): Likewise.
12541 (vectorizable_live_operation): Likewise.
12543 2018-05-17 Martin Liska <mliska@suse.cz>
12545 * gimple-ssa-sprintf.c (format_directive): Do not use
12546 space in between 'G_' and '('.
12548 2018-05-17 Jakub Jelinek <jakub@redhat.com>
12551 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
12552 even if the mask is not all ones.
12555 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
12557 (ix86_gimple_fold_builtin): Likewise.
12560 * config/i386/i386.c: Include tree-vector-builder.h.
12561 (ix86_vector_shift_count): New function.
12562 (ix86_fold_builtin): Fold shift builtins by scalar count.
12563 (ix86_gimple_fold_builtin): Likewise.
12565 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
12566 _mm512_setzero): New intrinsics.
12568 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
12569 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12571 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
12572 code generation for cases where splatting a value is not useful.
12573 * simplify-rtx.c (simplify_ternary_operation): Simplify
12574 vec_merge across a vec_duplicate and a paradoxical subreg forming
12575 a vector mode to a vec_concat.
12577 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
12579 * config.gcc: Support "goldmont-plus".
12580 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
12582 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12583 PROCESSOR_GOLDMONT_PLUS.
12584 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
12585 (processor_target_table): Add "goldmont-plus".
12586 (PTA_GOLDMONT_PLUS): Define.
12587 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
12588 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
12589 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
12590 (fold_builtin_cpu): Add "goldmont-plus".
12591 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
12592 (ix86_option_override_internal): Add "goldmont-plus".
12593 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
12594 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
12595 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
12596 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
12598 2018-05-17 Richard Biener <rguenther@suse.de>
12600 PR tree-optimization/85757
12601 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
12602 remove defs that only feed that PHI from further processing.
12604 2018-05-16 Jim Wilson <jimw@sifive.com>
12606 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
12608 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
12610 2018-05-16 Mark Wielaard <mark@klomp.org>
12612 * dwarf2out.c (count_index_strings): New function.
12613 (output_indirect_strings): Call count_index_strings and generate
12614 header for dwarf_version >= 5.
12616 2018-05-16 Mark Wielaard <mark@klomp.org>
12618 * dwarf2out.c (dwarf_FORM): New function.
12619 (set_indirect_string): Use dwarf_FORM.
12620 (reset_indirect_string): Likewise.
12621 (size_of_die): Likewise.
12622 (value_format): Likewise.
12623 (output_die): Likewise.
12624 (add_skeleton_AT_string): Likewise.
12625 (output_macinfo_op): Likewise.
12626 (index_string): Likewise.
12627 (output_index_string_offset): Likewise.
12628 (output_index_string): Likewise.
12629 (count_index_strings): Likewise.
12631 2018-05-16 Carl Love <cel@us.ibm.com>
12633 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
12634 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
12636 2018-05-16 Martin Jambor <mjambor@suse.cz>
12638 * ipa-prop.c (ipa_free_all_edge_args): Remove.
12639 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
12641 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
12643 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
12644 (fnma<mode>4): Likewise.
12645 (fms<mode>4): Likewise.
12646 (fnms<mode>4): Likewise.
12647 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
12648 (aarch64_fnma<mode>4): Likewise.
12649 (aarch64_fms<mode>4): Likewise.
12650 (aarch64_fnms<mode>4): Likewise.
12651 (aarch64_fnmadd<mode>4): Likewise.
12653 2018-05-16 Jason Merrill <jason@redhat.com>
12655 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
12657 2018-05-16 Richard Biener <rguenther@suse.de>
12659 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
12660 (dump_stmt_cost): Declare.
12661 (add_stmt_cost): Dump cost we add.
12662 (add_stmt_costs): New function.
12663 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
12664 No longer exported.
12665 (vect_analyze_stmt): Adjust prototype.
12666 (vectorizable_condition): Likewise.
12667 (vectorizable_live_operation): Likewise.
12668 (vectorizable_reduction): Likewise.
12669 (vectorizable_induction): Likewise.
12670 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
12671 cost vector to pass to vectorizable_ and record afterwards.
12672 (vect_model_reduction_cost): Take cost vector argument and adjust.
12673 (vect_model_induction_cost): Likewise.
12674 (vectorizable_reduction): Likewise.
12675 (vectorizable_induction): Likewise.
12676 (vectorizable_live_operation): Likewise.
12677 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
12678 SLP_TREE_NUMBER_OF_VEC_STMTS.
12679 (vect_analyze_slp_cost_1): Remove.
12680 (vect_analyze_slp_cost): Likewise.
12681 (vect_slp_analyze_node_operations): Take visited args and
12682 a target cost vector. Avoid processing already visited stmt sets.
12683 (vect_slp_analyze_operations): Use a local cost vector to gather
12684 costs and register those of non-discarded instances.
12685 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
12686 (vect_schedule_slp_instance): Remove copying of
12687 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
12689 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
12690 adding cost. Record cost entry location.
12691 (vect_prologue_cost_for_slp_op): Function to compute cost of
12692 a constant or invariant generated for SLP vect in the prologue,
12693 split out from vect_analyze_slp_cost_1.
12694 (vect_model_simple_cost): Make static. Adjust for SLP costing.
12695 (vect_model_promotion_demotion_cost): Likewise.
12696 (vect_model_store_cost): Likewise, make static.
12697 (vect_model_load_cost): Likewise.
12698 (vectorizable_bswap): Add cost vector arg and adjust.
12699 (vectorizable_call): Likewise.
12700 (vectorizable_simd_clone_call): Likewise.
12701 (vectorizable_conversion): Likewise.
12702 (vectorizable_assignment): Likewise.
12703 (vectorizable_shift): Likewise.
12704 (vectorizable_operation): Likewise.
12705 (vectorizable_store): Likewise.
12706 (vectorizable_load): Likewise.
12707 (vectorizable_condition): Likewise.
12708 (vectorizable_comparison): Likewise.
12709 (can_vectorize_live_stmts): Likewise.
12710 (vect_analyze_stmt): Likewise.
12711 (vect_transform_stmt): Adjust calls to vectorizable_*.
12712 * tree-vectorizer.c: Include gimple-pretty-print.h.
12713 (dump_stmt_cost): New function.
12715 2018-05-16 Richard Biener <rguenther@suse.de>
12717 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
12718 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
12719 * tree-ssa-dse.c: Include tree-ssa-loop.h.
12720 (check_name): New callback.
12721 (dse_classify_store): Track cycles via a visited bitmap of PHI
12722 defs and simplify handling of in-loop and across loop dead stores
12723 and properly fail for loop-variant refs. Handle byte-tracking with
12724 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
12727 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
12729 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
12730 (vect_get_mask_type_for_stmt): Likewise.
12731 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
12733 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
12734 to determine the statement's vector type and the vector type that
12735 should be used for calculating nunits. Deal with cases in which
12736 the type has to be deferred.
12737 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
12738 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
12739 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
12740 (vect_determine_vf_for_stmt): New functions, split out from...
12741 (vect_determine_vectorization_factor): ...here.
12742 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
12743 (vect_get_mask_type_for_stmt): New functions, split out from
12744 vect_determine_vectorization_factor.
12746 2018-05-16 Richard Biener <rguenther@suse.de>
12748 * tree-cfg.c (verify_gimple_assign_ternary): Properly
12749 verify the [VEC_]COND_EXPR embedded comparison.
12751 2018-05-15 Martin Sebor <msebor@redhat.com>
12753 PR tree-optimization/85753
12754 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
12755 RECORD_TYPE in addition to ARRAY_TYPE.
12757 2018-05-15 Martin Sebor <msebor@redhat.com>
12759 PR middle-end/85643
12760 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
12762 2018-05-15 Richard Biener <rguenther@suse.de>
12764 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
12765 add by_clobber_p one. Change algorithm to collect all defs
12766 representing uses we need to walk and try reducing them to
12767 a single one before failing.
12768 (dse_dom_walker::dse_optimize_stmt): Adjust.
12770 2018-05-13 Mark Wielaard <mark@klomp.org>
12772 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
12773 (size_of_loc_descr): Likewise.
12774 (output_loc_operands): Likewise.
12775 (output_loc_operands_raw): Likewise.
12776 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
12777 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
12778 (hash_loc_operands): Likewise.
12779 (compare_loc_operands): Likewise.
12781 2018-05-14 Mark Wielaard <mark@klomp.org>
12783 * dwarf2out.c (count_index_addrs): New function.
12784 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
12786 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12788 PR tree-optimization/83648
12789 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
12790 return value as malloc candidate.
12792 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12795 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
12796 param as true in call to suggest_attribute.
12798 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
12800 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
12801 -mreadonly-in-sdata.
12803 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12805 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
12807 (aarch64_crypto_aesd_fused): Likewise.
12809 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
12811 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
12812 (movsi_aarch64): Likewise.
12813 (load_pairsi): Likewise.
12814 (load_pairdi): Likewise.
12815 (store_pairsi): Likewise.
12816 (store_pairdi): Likewise.
12817 (load_pairsf): Likewise.
12818 (load_pairdf): Likewise.
12819 (store_pairsf): Likewise.
12820 (store_pairdf): Likewise.
12821 (zero_extend): Likewise.
12822 (trunc): Swap alternatives.
12823 (fcvt_target): Add '?' to prefer w over r.
12825 2018-05-14 Jakub Jelinek <jakub@redhat.com>
12828 * config/i386/i386.md: Disallow non-commutative arithmetics in
12829 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
12830 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
12831 in the peephole2 before it.
12833 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
12835 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
12836 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
12837 (ix86_handle_option): Handle -mcldemote.
12838 * config.gcc: New header.
12839 * config/i386/cldemoteintrin.h: New file.
12840 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
12841 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
12843 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12844 OPTION_MASK_ISA_CLDEMOTE.
12845 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
12846 (ix86_valid_target_attribute_inner_p): Ditto.
12847 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
12848 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
12849 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
12850 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
12851 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
12853 * config/i386/i386.opt: Add -mcldemote.
12854 * config/i386/x86intrin.h: New header.
12855 * doc/invoke.texi: Add -mcldemote.
12857 2018-05-14 Richard Biener <rguenther@suse.de>
12859 * doc/match-and-simplify.texi: Adjust :s documentation.
12861 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
12863 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
12864 intended memcpy size.
12865 (REORDER_45): Likewise.
12867 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
12869 * sort.cc: New file.
12870 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
12871 * vec.c (qsort_chk): Use gcc_qsort.
12872 * Makefile.in (OBJS-libcommon): Add sort.o.
12873 (build/sort.o): New target. Use it...
12874 (BUILD_RTL): ... here, and...
12875 (build/gencfn-macros): ... here, and...
12876 (build/genmatch): ... here.
12878 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
12879 Chung-Ju Wu <jasonwucj@gmail.com>
12881 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
12882 * config/nds32/nds32-graywolf.md: New file.
12883 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
12884 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
12886 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
12887 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
12888 * config/nds32/nds32.md (pipeline_model): Add graywolf.
12889 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
12890 * config/nds32/pipelines.md: Include n15 settings.
12892 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
12893 Chung-Ju Wu <jasonwucj@gmail.com>
12895 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
12896 * config/nds32/nds32-n13.md: New file.
12897 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
12898 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
12900 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
12901 * config/nds32/nds32.md (pipeline_model): Add n13.
12902 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
12903 * config/nds32/pipelines.md: Include n13 settings.
12905 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
12906 Chung-Ju Wu <jasonwucj@gmail.com>
12908 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
12909 * config/nds32/nds32-n10.md: New file.
12910 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
12911 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
12913 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
12914 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
12915 * config/nds32/nds32.md (pipeline_model): Add n10.
12916 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
12917 * config/nds32/pipelines.md: Include n10 settings.
12919 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
12920 Kito Cheng <kito.cheng@gmail.com>
12921 Chung-Ju Wu <jasonwucj@gmail.com>
12923 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
12924 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
12925 Add enum values for DSP extension instructions.
12926 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
12928 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
12929 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
12930 New code iterators.
12931 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
12932 * config/nds32/nds32-dspext.md: New file for DSP implementation.
12933 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
12934 * config/nds32/nds32-intrinsic.md: Likewise.
12935 * config/nds32/nds32_intrinsic.h: Likewise.
12936 * config/nds32/nds32-md-auxiliary.c: Likewise.
12937 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
12938 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
12939 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
12940 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
12941 * config/nds32/nds32-protos.h: New declarations for DSP extension.
12942 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
12943 TYPE_DMAC in switch statement.
12944 * config/nds32/nds32.c: New checking and implementation for DSP
12945 extension instructions.
12946 * config/nds32/nds32.h: Likewise.
12947 * config/nds32/nds32.md: Likewise.
12948 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
12949 * config/nds32/predicates.md: Implement new predicates for DSP
12952 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12954 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
12955 Reformat alternatives and attributes so it is easier to identify
12956 which constraints/attributes go with which instruction.
12957 (mov<mode>_hardfloat32, FMOVE64): Likewise.
12958 (mov<mode>_softfloat32, FMOVE64): Likewise.
12959 (mov<mode>_hardfloat64, FMOVE64): Likewise.
12960 (mov<mode>_softfloat64, FMOVE64): Likewise.
12962 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
12964 * doc/extend.texi (PowerPC Built-in Functions): Rename this
12966 (Basic PowerPC Built-in Functions): The new name of the
12967 subsection previously known as "PowerPC Built-in Functions".
12968 (Basic PowerPC Built-in Functions Available on all Configurations):
12970 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
12971 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
12972 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
12973 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
12975 2018-05-11 Martin Jambor <mjambor@suse.cz>
12978 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
12981 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
12984 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
12986 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
12988 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
12989 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
12990 (ix86_handle_option): Handle -mwaitpkg.
12991 * config.gcc: New header.
12992 * config/i386/cpuid.h (bit_WAITPKG): New bit.
12993 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
12994 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
12996 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12997 OPTION_MASK_ISA_WAITPKG.
12998 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
12999 (ix86_option_override_internal): Add PTA_WAITPKG.
13000 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
13001 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
13002 IX86_BUILTIN_TPAUSE.
13003 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
13004 __builtin_ia32_umwait and __builtin_ia32_tpause.
13005 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
13006 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
13007 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
13008 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
13009 UNSPECV_TPAUSE): New.
13010 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
13011 * config/i386/i386.opt: Add -mwaitpkg.
13012 * config/i386/waitpkgintrin.h: New file.
13013 * config/i386/x86intrin.h: New header.
13014 * doc/invoke.texi: Add -mwaitpkg.
13016 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
13019 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
13021 (cortex-m0): Use armv6s-m isa.
13022 (cortex-m0plus): Likewise.
13023 (cortex-m1): Likewise.
13024 (cortex-m0.small-multiply): Likewise.
13025 (cortex-m0plus.small-multiply): Likewise.
13026 (cortex-m1.small-multiply): Likewise.
13028 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
13029 Jakub Jelinek <jakub@redhat.com>
13031 PR tree-optimization/85692
13032 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
13033 source permute as well.
13035 2018-05-11 Martin Liska <mliska@suse.cz>
13038 * doc/extend.texi: Document LLVM style format for no_sanitize
13041 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13043 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
13044 mode_supports_vsx_dform_quad to mode_supports_dq_form.
13045 (mode_supports_vsx_dform_quad): Likewise.
13046 (mode_supports_vmx_dform): Move these functions to be next to the
13047 other mode_supports functions.
13048 (mode_supports_dq_form): Likewise.
13049 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
13050 mode_supports_dq_form.
13051 (reg_offset_addressing_ok_p): Likewise.
13052 (offsettable_ok_by_alignment): Likewise.
13053 (rs6000_legitimate_offset_address_p): Likewise.
13054 (legitimate_lo_sum_address_p): Likewise.
13055 (rs6000_legitimize_address): Likewise.
13056 (rs6000_legitimize_reload_address): Likewise.
13057 (rs6000_secondary_reload_inner): Likewise.
13058 (rs6000_preferred_reload_class): Likewise.
13059 (rs6000_output_move_128bit): Likewise.
13061 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
13063 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
13064 Generate SImode target register for null target.
13065 <case IX86_BUILTIN_XGETBV>: Ditto.
13066 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
13067 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
13069 2018-05-10 Carl Love <cel@us.ibm.com>
13071 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
13072 dcbtt and dcbtstt if operands[2] is 0.
13074 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
13077 * config/i386/sse.md (usadv64qi): New expander.
13079 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
13081 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
13082 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
13083 -maltivec=be support.
13084 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
13085 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
13086 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
13087 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
13088 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
13089 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
13090 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
13091 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
13092 altivec_vsumsws): Adjust.
13093 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
13094 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
13095 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
13097 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
13098 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
13099 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
13100 (altivec_lve<VI_char>x): Delete expand.
13101 (*altivec_lve<VI_char>x_internal): Rename to...
13102 (altivec_lve<VI_char>x): ... this.
13103 (altivec_lvxl_<mode>): Delete expand.
13104 (*altivec_lvxl_<mode>_internal): Rename to ...
13105 (altivec_lvxl_<mode>): ... this.
13106 (altivec_stvxl_<mode>): Delete expand.
13107 (*altivec_stvxl_<mode>_internal): Rename to ...
13108 (altivec_stvxl_<mode>): ... this.
13109 (altivec_stve<VI_char>x): Delete expand.
13110 (*altivec_stve<VI_char>x_internal): Rename to ...
13111 (altivec_stve<VI_char>x): ... this.
13112 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
13113 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
13114 reduc_plus_scal_<mode>): Adjust.
13115 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
13117 (rs6000_cpu_cpp_builtins): Adjust.
13118 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
13119 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
13120 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
13121 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
13122 -maltivec=be support.
13123 (rs6000_split_vec_extract_var): Adjust.
13124 (rs6000_split_v4si_init): Adjust.
13125 (swap_selector_for_mode): Delete.
13126 (altivec_expand_lvx_be, altivec_expand_stvx_be,
13127 altivec_expand_stvex_be): Delete.
13128 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
13129 -maltivec=be support.
13130 (rs6000_gimple_fold_builtin): Ditto.
13131 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
13133 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
13134 (TARGET_DIRECT_MOVE_64BIT): Adjust.
13135 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
13136 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
13137 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
13138 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
13139 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
13140 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
13141 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
13142 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
13143 anonymous split): Adjust.
13144 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
13145 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
13147 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
13149 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
13150 when --with-gxx-include-dir is also specified.
13151 * configure: Regenerate.
13153 2018-05-09 Jim Wilson <jimw@sifive.com>
13156 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
13157 * config/riscv/t-withmultilib: New.
13158 * config/riscv/withmultilib.h: New.
13159 * doc/install.texi: Document RISC-V --with-multilib-list support.
13161 2018-05-09 Richard Biener <rguenther@suse.de>
13163 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
13165 (vect_bb_vectorization_profitable_p): Adjust. Compute
13166 actual scalar cost using the cost vector and the add_stmt_cost
13169 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13171 PR rtl-optimization/85645
13172 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
13173 in the REG_CFA_REGISTER note for LR, don't leave it empty.
13175 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13177 PR rtl-optimization/85645
13178 * shrink-wrap.c (spread_components): Return a boolean saying if
13179 anything was changed.
13180 (try_shrink_wrapping_separate): Iterate spread_components until
13181 nothing changes anymore.
13183 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13185 PR rtl-optimization/85645
13186 * regrename.c (build_def_use): Also kill the chains that include the
13187 destination of a REG_CFA_REGISTER note.
13189 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13191 PR rtl-optimization/85645
13192 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
13193 insn that has a REG_CFA_REGISTER note.
13195 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
13197 * cfgexpand.c (expand_clobber): New function.
13198 (expand_gimple_stmt_1): Use it.
13199 * tree-vect-stmts.c (vect_clobber_variable): New function,
13201 (vectorizable_simd_clone_call): ...here.
13202 (vectorizable_store): Emit a clobber either side of an
13203 IFN_STORE_LANES sequence.
13204 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
13206 2018-05-09 Tom de Vries <tom@codesourcery.com>
13209 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
13210 (define_insn "trap_if_false"): Add exit after trap.
13212 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
13214 PR rtl-optimization/85638
13215 * bb-reorder.c: Include common/common-target.h.
13216 (create_forwarder_block): New function extracted from...
13217 (fix_up_crossing_landing_pad): ...here. Rename into...
13218 (dw2_fix_up_crossing_landing_pad): ...this.
13219 (sjlj_fix_up_crossing_landing_pad): New function.
13220 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
13221 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
13222 from both partitions and exit the loop after one iteration.
13224 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
13227 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13229 (Basic PowerPC Built-in Functions): The new name of the
13230 subsection previously known as "PowerPC Built-in Functions".
13231 (Basic PowerPC Built-in Functions Available on all Configurations):
13233 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13235 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13237 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13239 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13242 2018-05-08 Jim Wilson <jimw@sifive.com>
13244 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
13245 (LD_EMUL_SUFFIX): New.
13246 (LINK_SPEC): Use it.
13248 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
13250 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13252 (Basic PowerPC Built-in Functions): The new name of the
13253 subsection previously known as "PowerPC Built-in Functions".
13254 (Basic PowerPC Built-in Functions Available on all Configurations):
13256 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13258 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13260 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13262 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13265 2018-05-08 Jakub Jelinek <jakub@redhat.com>
13268 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
13269 after cmpelim optimization.
13271 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
13273 * config.gcc: Support "goldmont".
13274 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
13275 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13276 PROCESSOR_GOLDMONT.
13277 * config/i386/i386.c (m_GOLDMONT): Define.
13278 (processor_target_table): Add "goldmont".
13279 (PTA_GOLDMONT): Define.
13280 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
13281 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
13282 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
13283 (fold_builtin_cpu): Add "goldmont".
13284 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
13285 (ix86_option_override_internal): Add "goldmont".
13286 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
13287 (processor_type): Add PROCESSOR_GOLDMONT.
13288 * config/i386/i386.md: Add CPU "glm".
13289 * config/i386/glm.md: New file.
13290 * config/i386/x86-tune.def: Add m_GOLDMONT.
13291 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
13293 2018-05-08 Jakub Jelinek <jakub@redhat.com>
13296 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
13298 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
13299 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
13300 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
13303 * config/i386/i386.c (ix86_fold_builtin): Handle
13304 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
13307 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
13308 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
13310 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
13313 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
13314 (check_arch): Likewise.
13315 (check_fpu): Return the result rather than printing it.
13316 (end arch): Fix operator precedence.
13317 (end cpu): Likewise.
13318 (END): Print the result from check_fpu.
13320 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
13321 Alan Hayward <alan.hayward@arm.com>
13322 David Sherwood <david.sherwood@arm.com>
13324 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
13325 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
13326 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
13327 (*fcmuo<mode>_and): New patterns.
13329 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
13331 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
13332 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
13333 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
13334 (cmp_op, sve_imm_con): New code attributes.
13335 (SVE_COND_INT_CMP, imm_con): Delete.
13336 (cmp_op): Remove above unspecs from int attribute.
13337 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
13339 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
13340 comparison-specific unspecs.
13341 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
13342 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
13343 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
13344 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
13345 (*vec_fcm<cmp_op><mode>): Rename to...
13346 (*fcm<cmp_op><mode>): ...this and adjust likewise.
13347 (*vec_fcmuo<mode>): Rename to...
13348 (*fcmuo<mode>): ...this and adjust likewise.
13349 (*pred_fcm<cmp_op><mode>): New pattern.
13350 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
13351 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
13353 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
13355 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
13356 (aarch64_emit_sve_predicated_cond): New function.
13357 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
13358 (aarch64_emit_unspec_cond_or): Replace with...
13359 (aarch64_emit_sve_or_conds): ...this new function. Use
13360 aarch64_emit_sve_ptrue_op for the individual comparisons and
13361 aarch64_emit_binop to OR them together.
13362 (aarch64_emit_inverted_unspec_cond): Replace with...
13363 (aarch64_emit_sve_inverted_cond): ...this new function. Use
13364 aarch64_emit_sve_ptrue_op for the comparison and
13365 aarch64_emit_unop to invert the result.
13366 (aarch64_expand_sve_vec_cmp_float): Update after the above
13367 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
13369 2018-05-07 Nathan Sidwell <nathan@acm.org>
13371 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
13372 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
13373 (Backwards Compatibility): Likewise.
13375 2018-05-07 Luis Machado <luis.machado@linaro.org>
13379 2018-05-07 Luis Machado <luis.machado@linaro.org>
13381 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13382 <prefetch_dynamic_strides>: New const bool field.
13383 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13384 prefetch_dynamic_strides.
13385 (exynosm1_prefetch_tune): Likewise.
13386 (thunderxt88_prefetch_tune): Likewise.
13387 (thunderx_prefetch_tune): Likewise.
13388 (thunderx2t99_prefetch_tune): Likewise.
13389 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
13391 (aarch64_override_options_internal): Update to set
13392 PARAM_PREFETCH_DYNAMIC_STRIDES.
13393 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13394 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13395 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13396 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13397 prefetch-dynamic-strides setting.
13399 2018-05-07 Luis Machado <luis.machado@linaro.org>
13401 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13402 <minimum_stride>: New const int field.
13403 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13404 minimum_stride field.
13405 (exynosm1_prefetch_tune): Likewise.
13406 (thunderxt88_prefetch_tune): Likewise.
13407 (thunderx_prefetch_tune): Likewise.
13408 (thunderx2t99_prefetch_tune): Likewise.
13409 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
13410 (aarch64_override_options_internal): Update to set
13411 PARAM_PREFETCH_MINIMUM_STRIDE.
13412 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13413 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13414 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13415 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13416 stride is constant and is below the minimum stride threshold.
13418 2018-05-07 Luis Machado <luis.machado@linaro.org>
13420 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
13423 2018-05-07 Luis Machado <luis.machado@linaro.org>
13425 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13426 <prefetch_dynamic_strides>: New const bool field.
13427 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13428 prefetch_dynamic_strides.
13429 (exynosm1_prefetch_tune): Likewise.
13430 (thunderxt88_prefetch_tune): Likewise.
13431 (thunderx_prefetch_tune): Likewise.
13432 (thunderx2t99_prefetch_tune): Likewise.
13433 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
13435 (aarch64_override_options_internal): Update to set
13436 PARAM_PREFETCH_DYNAMIC_STRIDES.
13437 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13438 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13439 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13440 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13441 prefetch-dynamic-strides setting.
13443 2018-05-07 Luis Machado <luis.machado@linaro.org>
13445 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13446 <minimum_stride>: New const int field.
13447 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13448 minimum_stride field.
13449 (exynosm1_prefetch_tune): Likewise.
13450 (thunderxt88_prefetch_tune): Likewise.
13451 (thunderx_prefetch_tune): Likewise.
13452 (thunderx2t99_prefetch_tune): Likewise.
13453 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
13454 (aarch64_override_options_internal): Update to set
13455 PARAM_PREFETCH_MINIMUM_STRIDE.
13456 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13457 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13458 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13459 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13460 stride is constant and is below the minimum stride threshold.
13462 2018-05-06 Jakub Jelinek <jakub@redhat.com>
13465 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
13466 the type is addressable. Don't force op into register if it has
13469 2018-05-05 Roland McGrath <mcgrathr@google.com>
13472 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
13473 any section for which we don't know a specific type it should have,
13474 regardless of name. Previously this was done only for the exact
13475 names ".init_array", ".fini_array", and ".preinit_array".
13476 (default_elf_asm_named_section): Add comment about
13477 relationship with default_section_type_flags and SECTION_NOTYPE.
13478 (get_section): Don't consider it a type conflict if one side has
13479 SECTION_NOTYPE and the other doesn't, as long as neither has the
13480 SECTION_BSS et al used in the default_section_type_flags logic.
13482 2018-05-05 Tom de Vries <tom@codesourcery.com>
13485 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
13486 (workaround_barsyncs): New function.
13487 (nvptx_reorg): Use workaround_barsyncs.
13488 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
13489 (define_expand "nvptx_membar_cta"): New define_expand.
13490 (define_insn "*nvptx_membar_cta"): New insn.
13492 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
13494 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
13495 To improve optimization opportunities.
13496 * builtin-types.def: The new needed builtin types for the above.
13498 2018-05-04 Richard Biener <rguenther@suse.de>
13500 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
13501 * gimple-ssa-store-merging.c
13502 (imm_store_chain_info::output_merged_store): Remove redundant create,
13503 release split_store vector contents on failure.
13504 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
13505 scalar stmt vector on cache hit.
13507 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
13509 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
13511 * config.gcc (powerpc-xilinx-eabi*): Remove.
13512 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
13514 (fusion_addis_mem_combo_load): Ditto.
13515 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
13517 (rs6000_cpu_cpp_builtins): Ditto.
13518 * config/rs6000/rs6000-linux.c
13519 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
13520 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
13521 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
13523 (rs6000_setup_reg_addr_masks): Ditto.
13524 (rs6000_init_hard_regno_mode_ok): Ditto.
13525 (rs6000_option_override_internal): Ditto.
13526 (legitimate_lo_sum_address_p): Ditto.
13527 (rs6000_legitimize_address): Ditto.
13528 (rs6000_legitimize_reload_address): Ditto.
13529 (rs6000_legitimate_address_p): Ditto.
13530 (abi_v4_pass_in_fpr): Ditto.
13531 (setup_incoming_varargs): Ditto.
13532 (rs6000_gimplify_va_arg): Ditto.
13533 (rs6000_split_multireg_move): Ditto.
13534 (rs6000_savres_strategy): Ditto.
13535 (rs6000_emit_prologue_components): Ditto.
13536 (rs6000_emit_epilogue_components): Ditto.
13537 (rs6000_emit_prologue): Ditto.
13538 (rs6000_emit_epilogue): Ditto.
13539 (rs6000_elf_file_end): Ditto.
13540 (rs6000_function_value): Ditto.
13541 (rs6000_libcall_value): Ditto.
13542 * config/rs6000/rs6000.h: Ditto.
13543 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
13544 (TARGET_MINMAX): ... this. New.
13545 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
13546 * config/rs6000/rs6000.md: Remove Xilinx FP support.
13547 (*movsi_internal1_single): Delete.
13548 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
13549 mfpu=, mxilinx-fpu): Delete.
13550 * config/rs6000/singlefp.h: Delete.
13551 * config/rs6000/sysv4.h: Remove Xilinx FP support.
13552 * config/rs6000/t-rs6000: Ditto.
13553 * config/rs6000/t-xilinx: Delete.
13554 * config/rs6000/titan.md: Adjust for fp_type removal.
13555 * config/rs6000/vsx.md: Remove Xilinx FP support.
13556 (VStype_simple): Delete.
13557 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
13558 * config/rs6000/xfpu.h: Delete.
13559 * config/rs6000/xfpu.md: Delete.
13560 * config/rs6000/xilinx.h: Delete.
13561 * config/rs6000/xilinx.opt: Delete.
13562 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
13563 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
13565 2018-05-04 Tom de Vries <tom@codesourcery.com>
13568 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
13571 2018-05-04 Richard Biener <rguenther@suse.de>
13573 PR middle-end/85627
13574 * tree-complex.c (update_complex_assignment): We are always in SSA form.
13575 (expand_complex_div_wide): Likewise.
13576 (expand_complex_operations_1): Likewise.
13577 (expand_complex_libcall): Preserve EH info of the original stmt.
13578 (tree_lower_complex): Handle removed blocks.
13579 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
13580 on complex multiplication and division libcall builtins.
13582 2018-05-04 Richard Biener <rguenther@suse.de>
13584 PR middle-end/85574
13585 * fold-const.c (negate_expr_p): Restrict negation of operand
13586 zero of a division to when we know that can happen without
13588 (fold_negate_expr_1): Likewise.
13590 2018-05-04 Jakub Jelinek <jakub@redhat.com>
13593 * real.h (real_nextafter): Declare.
13594 * real.c (real_nextafter): New function.
13595 * fold-const-call.c (fold_const_nextafter): New function.
13596 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
13597 CASE_CFN_NEXTTOWARD.
13598 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
13599 even when arg1_mode is different from arg0_mode.
13601 2018-05-03 Nathan Sidwell <nathan@acm.org>
13603 * doc/extend.texi (Deprecated Features): Remove
13604 -ffriend-injection.
13605 (Backwards Compatibility): Likewise.
13606 * doc/invoke.texi (C++ Language Options): Likewise.
13607 (C++ Dialect Options): Likewise.
13609 2018-05-03 Jakub Jelinek <jakub@redhat.com>
13612 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
13613 _mm512_mask_mullox_epi64): New intrinsics.
13615 2018-05-03 Tom de Vries <tom@codesourcery.com>
13618 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13619 dump files): Add offload-tree.
13621 2018-05-03 Richard Biener <rguenther@suse.de>
13623 PR tree-optimization/85615
13624 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
13625 to loops not nested in BBs loop father to avoid creating multi-entry
13628 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13630 PR tree-optimization/70291
13631 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
13632 arguments. Change return type to tree. Emit libcall as a new
13633 statement rather than replacing existing one when inplace_p is true.
13634 (expand_complex_multiplication_components): New function.
13635 (expand_complex_multiplication): Expand floating-point complex
13636 multiplication using the above.
13637 (expand_complex_division): Rename inner_type parameter to type.
13638 Update expand_complex_libcall call-site.
13639 (expand_complex_operations_1): Update expand_complex_multiplication
13640 and expand_complex_division call-sites.
13642 2018-05-02 Jakub Jelinek <jakub@redhat.com>
13645 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
13646 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
13647 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
13648 the highest significant bit of the shift count mask is clear. In
13649 check whether and[sq]i3 is needed verify that all significant bits
13650 of the shift count other than the highest are set.
13652 2018-05-02 Tom de Vries <tom@codesourcery.com>
13655 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
13656 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
13657 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
13658 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
13659 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
13660 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
13661 __builtin_goacc_parlevel_size.
13663 2018-05-02 Richard Biener <rguenther@suse.de>
13665 PR tree-optimization/85597
13666 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
13667 do not use split vect_get_vec_defs call but call vect_get_slp_defs
13670 2018-05-02 Tom de Vries <tom@codesourcery.com>
13673 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13674 dump files): Add ltrans-tree.
13676 2018-05-02 Tom de Vries <tom@codesourcery.com>
13679 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13680 dump files): Add wpa-ipa.
13682 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
13684 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
13685 powerpc*-*-linux*paired* target.
13686 * config/rs6000/750cl.h: Delete.
13687 * config/rs6000/paired.h: Delete.
13688 * config/rs6000/paired.md: Delete.
13689 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
13691 * config/rs6000/rs6000-builtin.def: Remove paired float support.
13692 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
13693 comment. Remove paired float support.
13694 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
13695 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
13697 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
13698 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
13700 * config/rs6000/rs6000.c: Remove paired float support.
13701 (paired_expand_vector_init, paired_expand_vector_move,
13702 paired_emit_vector_compare, paired_emit_vector_cond_expr,
13703 (paired_expand_lv_builtin, paired_expand_stv_builtin,
13704 paired_expand_builtin, paired_expand_predicate_builtin,
13705 paired_init_builtins): Delete.
13706 * config/rs6000/rs6000.h: Remove paired float support.
13707 * config/rs6000/rs6000.md: Remove paired float support.
13708 (move_from_CR_ov_bit): Delete.
13709 * config/rs6000/rs6000.opt (mpaired): Delete.
13710 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
13711 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
13713 2018-05-02 Richard Biener <rguenther@suse.de>
13715 PR middle-end/85567
13716 * gimplify.c (gimplify_save_expr): When in SSA form allow
13717 SAVE_EXPRs to compute to SSA vars.
13719 2018-05-02 Jakub Jelinek <jakub@redhat.com>
13722 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
13723 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
13724 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
13725 clobber operands[2], instead use a new pseudo. Formatting fixes.
13727 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
13729 PR tree-optimization/85586
13730 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
13731 exit early for statements in the same group if the accesses are
13734 2018-05-02 Tom de Vries <tom@codesourcery.com>
13737 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
13740 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
13742 PR tree-optimization/85143
13743 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
13745 2018-05-01 Tom de Vries <tom@codesourcery.com>
13748 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
13749 not found" error message.
13751 2018-05-01 Tom de Vries <tom@codesourcery.com>
13754 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
13755 * vec.c (test_ordered_remove_if): New function.
13756 (vec_c_tests): Call test_ordered_remove_if.
13757 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
13758 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
13759 * tree-vect-patterns.c (vect_pattern_recog_1): Use
13760 VEC_ORDERED_REMOVE_IF.
13762 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13764 PR tree-optimization/82665
13765 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
13766 pointer subtraction where arguments come from a memchr call.
13768 2018-05-01 Jakub Jelinek <jakub@redhat.com>
13770 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
13771 --push-state --as-needed and --pop-state instead of --as-needed and
13772 --no-as-needed if ld supports it.
13773 * configure: Regenerated.
13776 * doc/install.texi2html: Replace _002d with - and _002a with * in
13777 generated html files using sed.
13779 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13782 * gcc-rich-location.c (blank_line_before_p): New function.
13783 (use_new_line): New function.
13784 (gcc_rich_location::add_fixit_insert_formatted): New function.
13785 * gcc-rich-location.h
13786 (gcc_rich_location::add_fixit_insert_formatted): New function.
13788 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13790 * selftest.c (assert_streq): Rename "expected" and "actual" to
13791 "val1" and "val2". Extend NULL-handling to cover both inputs
13792 symmetrically, while still requiring both to be non-NULL for a pass.
13793 * selftest.h (assert_streq): Rename "expected" and "actual" to
13795 (ASSERT_EQ): Likewise.
13796 (ASSERT_EQ_AT): Likewise.
13797 (ASSERT_KNOWN_EQ): Likewise.
13798 (ASSERT_KNOWN_EQ_AT): Likewise.
13799 (ASSERT_NE): Likewise.
13800 (ASSERT_MAYBE_NE): Likewise.
13801 (ASSERT_MAYBE_NE_AT): Likewise.
13802 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
13803 the assertion to pass.
13804 (ASSERT_STREQ_AT): Likewise.
13806 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
13808 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
13809 interaction with -pie.
13811 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13813 * selftest.h: Fix alphabetization of per-source-file selftest
13816 2018-04-30 Jason Merrill <jason@redhat.com>
13818 PR c++/61982 - dead stores to destroyed objects.
13819 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
13822 2018-04-30 Jason Merrill <jason@redhat.com>
13824 * tree.c (build_clobber): New.
13825 * tree.h: Declare it.
13826 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
13828 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13830 * diagnostic-show-locus.c (layout::layout): Update for
13831 location_get_source_line returning a char_span.
13832 (struct char_span): Move to input.h.
13833 (struct correction): Update for fields in char_span becoming
13835 (struct source_line): Update for location_get_source_line
13836 returning a char_span.
13837 (layout::print_line): Likewise.
13838 * edit-context.c (edited_file::print_content): Likewise.
13839 (edited_file::print_diff_hunk): Likewise.
13840 (edited_file::print_run_of_changed_lines): Likewise.
13841 (edited_file::get_num_lines): Likewise.
13842 (edited_line::edited_line): Likewise.
13843 * final.c (asm_show_source): Likewise.
13844 * input.c (location_get_source_line): Convert return type
13845 from const char * to char_span, losing the final "line_len"
13847 (dump_location_info): Update for the above.
13848 (get_substring_ranges_for_loc): Likewise. Use a char_span
13849 when handling the literal within the line.
13850 (test_reading_source_line): Update for location_get_source_line
13851 returning a char_span.
13852 * input.h (class char_span): Move here from
13853 diagnostic-show-locus.c, converting from a struct to a class.
13854 Make data members private.
13855 (char_span::operator bool): New.
13856 (char_span::length): New.
13857 (char_span::get_buffer): New.
13858 (char_span::operator[]): New.
13859 (char_span::subspan): Make const.
13860 (char_span::xstrdup): New.
13861 (location_get_source_line): Convert return type from const char *
13862 to char_span, losing the final "line_size" param.
13864 2018-04-30 Jan Hubicka <jh@suse.cz>
13866 * lto-wrapper.c (ltrans_priorities): New static var.
13867 (cmp_priority): New.
13868 (run_gcc): Read priorities and if doing parallel build order
13869 the Makefile by them.
13871 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13873 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
13875 2018-04-30 Richard Biener <rguenther@suse.de>
13877 * tree-cfg.c (verify_address): Remove base argument, add
13878 flag whether to check TREE_ADDRESSABLE and do that.
13879 (verify_expr): Remove.
13880 (verify_types_in_gimple_reference): Add pieces from verify_expr.
13881 (verify_gimple_assign_single): Likewise.
13882 (verify_gimple_switch): Likewise.
13883 (verify_expr_location_1): Dereference tp once. Add (disabled)
13884 piece from verify_expr.
13885 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
13887 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
13889 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
13891 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
13893 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
13894 (small_data_pattern): Likewise.
13895 (arc_rewrite_small_data): Likewise.
13896 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
13897 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
13898 (get_symbol_alignment): New function.
13899 (legitimate_small_data_address_p): Likewise.
13900 (legitimate_scaled_address): Update, call
13901 legitimate_small_data_address_p.
13902 (output_sdata): New static variable.
13903 (arc_print_operand): Update how we handle small data operands.
13904 (arc_print_operand_address): Likewise.
13905 (arc_legitimate_address_p): Update, use
13906 legitimate_small_data_address_p.
13907 (arc_rewrite_small_data_p): Remove.
13908 (arc_rewrite_small_data_1): Likewise.
13909 (arc_rewrite_small_data): Likewise.
13910 (small_data_pattern): Likewise.
13911 (compact_sda_memory_operand): Update to use
13912 legitimate_small_data_address_p and get_symbol_alignment.
13913 (prepare_move_operands): Don't rewite sdata pattern.
13914 (prepare_extend_operands): Remove.
13915 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
13917 (zero_extendqisi2): Likewise.
13918 (zero_extendhisi2): Likewise.
13919 (extendqihi2): Likewise.
13920 (extendqisi2): Likewise.
13921 (extendhisi2): Likewise.
13922 (addsi3): Likewise.
13923 (subsi3): Likewise.
13924 (andsi3): Likewise.
13925 * config/arc/constraints.md (Usd): Change it to memory constraint.
13927 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
13929 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
13930 as source of std instructions.
13931 * config/arc/arc.md (movsi_insn): Update pattern predicate to
13932 allow 6-bit constants as source for store instructions.
13933 (movdi_insn): Update instruction pattern to allow 6-bit constants
13934 as source for store instructions.
13936 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
13938 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
13940 2018-04-30 Nathan Sidwell <nathan@acm.org>
13941 Sandra Loosemore <sandra@codesourcery.com>
13943 * dumpfile.c (dump_open): Allow '-' for stdout.
13944 * doc/invoke.texi (Developer Options): Document dump filename
13945 determination early. Document stdin/stdout selection.
13947 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
13949 Microblaze Target: PIC data text relative
13951 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
13952 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
13954 * config/microblaze/microblaze.h (microblaze_constant_address_p):
13955 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
13956 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
13957 New addressing mode for data-text relative position indepenedent code.
13958 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
13959 'ADDRESS_SYMBOLIC_TXT_REL'.
13960 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
13961 (microblaze_legitimate_pic_operand): Exclude function calls from
13962 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
13963 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
13965 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
13966 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
13967 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
13968 for 'address + offset'.
13969 (microblaze_expand_prologue): Add new function prologue call for
13971 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
13972 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
13973 table in case of TARGET_PIC_DATA_TEXT_REL.
13974 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
13975 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
13976 Add new macros 'UNSPEC_TEXT',
13977 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
13978 + exclude function calls from 'UNSPEC_PLT' in case of data text
13980 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
13981 new target hook for generating address diff vector tables in case of
13983 * doc/tm.texi : Regenerate.
13984 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
13985 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
13986 of addr diff vector generation.
13987 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
13988 target hook definition.
13989 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
13990 Add default function for generate_pic_addr_diff_vec -> flag_pic.
13991 * doc/invoke.texi (Add new pic option): Add new microblaze pic
13992 option for data text relative.
13994 2018-04-30 Richard Biener <rguenther@suse.de>
13996 * tree-chrec.h (evolution_function_is_constant_p): Remove
13998 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
14000 2018-04-30 Richard Biener <rguenther@suse.de>
14003 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
14005 2018-04-30 Richard Biener <rguenther@suse.de>
14007 PR tree-optimization/28364
14008 PR tree-optimization/85275
14009 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
14010 copying first exit test.
14012 2018-04-28 Mark Wielaard <mark@klomp.org>
14014 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
14015 dwarf_version >= 5.
14016 (dwarf_AT): Handle DW_AT_addr_base.
14017 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
14019 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
14022 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
14023 (*ashl<dwi>3_doubleword_mask_1): Ditto.
14024 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
14025 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
14027 2018-04-28 Richard Biener <rguenther@suse.de>
14029 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
14030 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
14031 to reflect use. Only add interesting stmts.
14033 2018-04-27 Martin Jambor <mjambor@suse.cz>
14036 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
14037 the jump function allows for passing through aggregate values.
14039 2018-04-27 David Malcolm <dmalcolm@redhat.com>
14041 * input.h (in_system_header_at): Convert from macro to inline
14043 (from_macro_expansion_at): Likewise.
14044 (from_macro_definition_at): Likewise.
14046 2018-04-27 Jeff Law <law@redhat.com>
14048 * config.gcc: Mark tile* targets as deprecated/obsolete.
14050 2018-04-27 Richard Biener <rguenther@suse.de>
14052 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
14055 2018-04-27 Richard Biener <rguenther@suse.de>
14057 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
14059 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
14061 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
14062 with Yd constraint. Set "preferred_for_speed" attribute from
14063 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
14064 with Yd constraint.
14065 (*movdi_internal): Ditto.
14066 (movti_interunit splitters): Remove
14067 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
14068 (movdi_interunit splitters): Ditto.
14069 * config/i386/constraints.md (Ye): Remove.
14070 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
14072 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14075 * config/aarch64/constraints.md (Usg): Limit to 31.
14076 (Usj): Limit to 63.
14078 2018-04-27 Jakub Jelinek <jakub@redhat.com>
14080 PR tree-optimization/85529
14081 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
14082 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
14083 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
14084 zero extension or masking of the MSB bit.
14085 (optimize_range_tests): Add FIRST_BB argument, pass it through
14086 to optimize_range_tests_var_bound.
14087 (maybe_optimize_range_tests, reassociate_bb): Adjust
14088 optimize_range_tests callers.
14090 2018-04-26 Richard Biener <rguenther@suse.de>
14091 Jakub Jelinek <jakub@redhat.com>
14093 * cgraph.h (symbol_table): Just declare debug method here.
14094 * symtab.c (symbol_table::debug): Define.
14096 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
14098 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
14100 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
14102 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
14103 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
14104 (*movdi_internal): Substitute Yi and Yj constraint with x
14105 and Ym and Yn constraint with y constraint. Update "isa"
14106 attribute and set "preferred_for_speed" attribute from
14107 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
14108 (*movsi_internal): Ditto.
14109 (*movdf_internal): Ditto.
14110 (*movsf_internal): Ditto.
14111 (*zero_extendsidi2): Ditto.
14112 * config/i386/sse.md (vec_set<mode>_0): Ditto.
14113 (sse2_loadld): Ditto.
14114 (*vec_extract<ssevecmodelower>_0): Ditto.
14115 (*vec_extractv4si_0_zext_sse4): Ditto.
14116 (vec_concatv2di): Ditto.
14117 (*vec_dup<mode>): Ditto.
14118 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
14119 * config/i386/constraints.md (Yi): Remove.
14124 2018-04-26 Nathan Sidwell <nathan@acm.org>
14126 * dumpfile.c (dump_open): New.
14127 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
14128 (dump_finish): Detect stdio/stderr by value not name.
14130 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
14132 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
14134 2018-04-26 Tom de Vries <tom@codesourcery.com>
14137 * config/nvptx/nvptx.c (verify_neutering_jumps)
14138 (verify_neutering_labels): New function
14139 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
14141 2018-04-26 Tom de Vries <tom@codesourcery.com>
14144 * config/nvptx/nvptx.c (needs_neutering_p): New function.
14145 (nvptx_single): Use needs_neutering_p to skip over insns that do not
14148 2018-04-26 Richard Biener <rguenther@suse.de>
14149 Tom de Vries <tom@codesourcery.com>
14152 * lto-streamer-out.c (output_function): Fixup loops if required to match
14153 discovery done in the reader.
14155 2018-04-26 Richard Biener <rguenther@suse.de>
14157 PR tree-optimization/85116
14158 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
14159 have a loop exit from the single latch predecessor. Remove
14160 case of header with just condition.
14161 (ch_base::copy_headers): Exclude infinite loops from any
14163 (pass_ch::execute): Record exits.
14165 2018-04-26 Richard Biener <rguenther@suse.de>
14167 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
14168 prologue cost vector and pass it to vect_get_load_cost.
14169 (vect_get_peeling_costs_all_drs): Likewise.
14170 (vect_peeling_hash_get_lowest_cost): Likewise.
14171 (vect_enhance_data_refs_alignment): Likewise.
14173 2018-04-26 Richard Biener <rguenther@suse.de>
14175 PR middle-end/85450
14176 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
14177 checking of integer<->pointer conversions.
14178 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
14179 sign-/zero-extending pointer types.
14180 (expand_omp_for_static_chunk): Likewise.
14182 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
14183 Jean Lee <xiaoyur347@gmail.com>
14185 * config/mips/mips.c (mips_asan_shadow_offset): New function.
14186 (TARGET_ASAN_SHADOW_OFFSET): Define.
14187 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
14188 true for -fsanitize=address.
14190 2018-04-25 Mark Wielaard <mark@klomp.org>
14192 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
14195 2018-04-25 Jakub Jelinek <jakub@redhat.com>
14197 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
14198 than "alu", remove explicit "memory" and "imm_disp" attributes.
14199 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
14201 PR middle-end/85414
14202 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
14203 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
14204 gen_lowpart_no_emit.
14206 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
14209 * config/i386/i386.c (ix86_expand_builtin): Change memory
14210 operand to XI, extend p0 to Pmode.
14211 * config/i386/i386.md: Change unspec volatile and operand
14212 1 mode to XI, change operand 0 mode to P.
14214 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14216 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
14217 GET_MODE_MASK before any checking.
14218 (nds32_can_use_bset_p): Likewise.
14219 (nds32_can_use_btgl_p): Likewise.
14221 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14223 * config/nds32/nds32-doubleword.md: New define_split pattern for
14224 illegal register number.
14226 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14228 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
14230 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14232 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
14234 2018-04-25 Richard Biener <rguenther@suse.de>
14236 * lto-streamer.h (LTO_major_version): Bump to 8.
14238 2018-04-25 Jakub Jelinek <jakub@redhat.com>
14240 * BASE-VER: Set to 9.0.0.
14242 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
14244 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
14245 in __abskf2 and __powikf2.
14247 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14250 * config/aarch64/constraints.md (Usg, Usj): New constraints.
14251 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
14252 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
14253 Use the above on operand 2. Reindent.
14254 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
14256 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
14259 * common/config/i386/i386-common.c (ix86_handle_option): Don't
14261 * config/i386/i386.opt (mcet): Removed.
14262 * doc/install.texi: Remove -mcet documentation.
14263 * doc/invoke.texi: Likewise.
14265 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
14268 * doc/install.texi: Remove -mcet from bootstrap-cet.
14270 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14273 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
14274 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
14278 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
14279 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
14280 containing a CONST_VECTOR.
14282 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
14284 * doc/install.texi: Update newlib dependency for nvptx.
14286 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14289 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
14290 instead of INTVAL when shifting x left.
14292 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
14294 PR tree-optimization/85478
14295 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
14296 vect_grouped_store_supported for single element vectors.
14298 2018-04-24 Richard Biener <rguenther@suse.de>
14301 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
14302 load cost increase to the case of non-constant step.
14304 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14307 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
14308 destination if any_malformed_asm.
14310 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
14312 PR middle-end/85496
14313 * expr.c (store_field): In the bitfield case, if the value comes from
14314 a function call and is returned in registers by means of a PARALLEL,
14315 do not change the mode of the temporary unless BLKmode and VOIDmode.
14317 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
14319 PR rtl-optimization/85423
14320 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
14321 dependencies to debug insns when the previous insn is non-debug.
14323 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
14325 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
14326 enums into a single definition.
14327 (fls): Fix predicates and printing.
14330 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
14332 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
14333 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
14334 and short u6 immediate.
14335 (check_if_valid_sleep_operand): Remove.
14336 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
14338 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14340 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
14341 flag_always_save_lp condition.
14342 * config/nds32/nds32.opt (malways-save-lp): New option.
14344 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14346 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
14347 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
14348 * config/nds32/nds32.h
14349 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
14350 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
14352 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14354 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
14355 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
14357 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14358 Chung-Ju Wu <jasonwucj@gmail.com>
14360 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
14362 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
14363 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
14365 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14367 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
14369 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14371 * config/nds32/nds32-protos.h (nds32_data_alignment,
14372 nds32_local_alignment): Declare.
14373 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
14374 nds32_local_alignment): New functions.
14375 (TARGET_CONSTANT_ALIGNMENT): Define.
14376 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
14378 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14380 * config/nds32/nds32.c
14381 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
14382 (TARGET_MODES_TIEABLE_P): Likewise.
14384 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14386 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
14387 level Ofast and Og.
14389 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
14390 Chung-Ju Wu <jasonwucj@gmail.com>
14392 * config/nds32/constants.md (unspec_volatile_element): Add enum values
14393 for unaligned access.
14394 * config/nds32/nds32-intrinsic.c: Implementation of expanding
14396 * config/nds32/nds32-intrinsic.md: Likewise.
14397 * config/nds32/nds32_intrinsic.h: Likewise.
14398 * config/nds32/nds32.h (nds32_builtins): Likewise.
14399 * config/nds32/nds32.opt (munaligned-access): New option.
14400 * config/nds32/nds32.c (nds32_asm_file_start): Display
14401 flag_unaligned_access status.
14403 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
14405 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
14406 -mno-relax is present.
14407 * config/riscv/linux.h (LINK_SPEC): Ditto.
14409 2018-04-20 Martin Sebor <msebor@redhat.com>
14412 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
14414 (gimple_fold_builtin_stxcpy_chk): Same.
14415 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
14417 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
14420 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
14421 __powikf2 when long double is IEEE 128-bit.
14423 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
14425 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
14426 step to make sure stack always aligned.
14428 2018-04-20 Carl Love <cel@us.ibm.com>
14431 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
14432 size check for arg0.
14434 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
14435 Tom de Vries <tom@codesourcery.com>
14438 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
14439 Emit insns for calls too.
14440 (nvptx_find_par): Always look for worker-level predecessor insn.
14441 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
14443 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
14444 (nvptx_process_pars): Propagate frames for calls.
14446 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
14449 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
14451 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
14452 (ix86_handle_option): Don't handle OPT_mibt.
14453 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
14455 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
14457 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
14458 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
14459 (ix86_target_macros): Define __CET__ with flag_cf_protection
14460 for -fcf-protection.
14461 * config/i386/i386.c (isa2_opts): Remove -mibt.
14462 * config/i386/i386.h (TARGET_IBT): Removed.
14463 (TARGET_IBT_P): Likewise.
14464 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
14465 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
14466 * config/i386/i386.opt (mcet): Update help message.
14467 (mshstk): Likewise.
14469 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
14470 -mcet as an alias for -mshstk.
14472 2018-04-20 Richard Biener <rguenther@suse.de>
14474 PR middle-end/85475
14475 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
14476 complexity by forcing a single use of the multiply operand.
14478 2018-04-20 Martin Jambor <mjambor@suse.cz>
14481 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
14482 recursion dependency to only apply to non-clones.
14484 2018-04-20 Martin Jambor <mjambor@suse.cz>
14487 * ipa-cp.c (create_specialized_node): Check that clones of
14488 self-recursive edges exist during IPA-CP.
14490 2018-04-19 Toon Moene <toon@moene.org>
14492 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
14495 2018-04-19 Jakub Jelinek <jakub@redhat.com>
14497 PR tree-optimization/85467
14498 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
14499 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
14500 VECTOR_CST element to type.
14502 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
14505 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
14506 * config/i386/i386.md (builtin_setjmp_setup): Removed.
14507 (builtin_longjmp): Likewise.
14508 (save_stack_nonlocal): New pattern.
14509 (restore_stack_nonlocal): Likewise.
14511 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
14514 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
14515 Replace ASM_OUTPUT_LABEL with fprintf.
14517 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
14520 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
14521 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
14522 * config/i386/i386-c.c (ix86_target_macros_internal): Also
14523 define __IBT__ and __SHSTK__ for -fcf-protection.
14524 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
14526 (ix86_trampoline_init): Likewise.
14527 (x86_output_mi_thunk): Likewise.
14528 (ix86_notrack_prefixed_insn_p): Likewise.
14529 (ix86_option_override_internal): Don't disallow -fcf-protection.
14530 * config/i386/i386.md (rdssp<mode>): Also enable for
14532 (incssp<mode>): Likewise.
14533 (nop_endbr): Likewise.
14534 * config/i386/i386.opt (mcet): Change help message to built-in
14537 (mshstk): Likewise.
14538 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
14539 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
14540 enable CET built-in functions.
14542 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
14544 * common/config/i386/i386-common.c
14545 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
14546 OPTION_MASK_ISA_MOVDIRI_UNSET,
14547 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
14548 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
14549 * config.gcc (movdirintrin.h): New header.
14550 * config/i386/cpuid.h (bit_MOVDIRI,
14551 bit_MOVDIR64B): New bits.
14552 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
14554 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
14555 (VOID, PVOID, PCVOID)): New function types.
14556 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
14557 __builtin_ia32_directstoreu_u64,
14558 __builtin_ia32_movdir64b): New builtins.
14559 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
14560 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
14562 (ix86_valid_target_attribute_inner_p): Ditto.
14563 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
14564 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
14565 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
14566 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
14567 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
14568 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
14569 (movdiri<mode>, movdir64b_<mode>): New.
14570 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
14571 * config/i386/immintrin.h: Include movdirintrin.h.
14572 * config/i386/movdirintrin.h: New file.
14573 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
14575 2018-04-19 Richard Biener <rguenther@suse.de>
14577 PR middle-end/85455
14578 * cfg.c (clear_bb_flags): When loop state says we have
14579 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
14581 2018-04-19 Richard Biener <rguenther@suse.de>
14583 PR tree-optimization/84737
14584 * tree-vect-data-refs.c (vect_copy_ref_info): New function
14585 copying restrict info.
14586 (vect_setup_realignment): Use it.
14587 * tree-vectorizer.h (vect_copy_ref_info): Declare.
14588 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
14589 the first DR to all generated stores.
14590 (vectorizable_load): Likewise for loads.
14592 2018-04-19 Jakub Jelinek <jakub@redhat.com>
14594 PR tree-optimization/85446
14595 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
14596 the integral and pointer types to have the same precision.
14598 * doc/install.texi: Document --disable-cet being the default and
14601 2018-04-18 Martin Liska <mliska@suse.cz>
14603 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
14606 2018-04-18 Martin Liska <mliska@suse.cz>
14609 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
14612 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
14613 arguments if they are comparable.
14615 2018-04-18 Martin Liska <mliska@suse.cz>
14618 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
14621 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
14624 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
14627 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
14628 ENDBR after calling __morestack.
14630 2018-04-18 David Malcolm <dmalcolm@redhat.com>
14633 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
14634 by using gcc_base_ver to generate a gcc_driver_version, and use
14635 it when generating GCC_DRIVER_NAME.
14636 * configure: Regenerate.
14638 2018-04-18 Jakub Jelinek <jakub@redhat.com>
14641 * config.gcc: Obsolete powerpc*-*-*spe*.
14643 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14646 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
14647 (stabstr_D): Change type of unum from unsigned int to
14648 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
14651 2018-04-17 Jim Wilson <jimw@sifive.com>
14654 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
14655 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
14656 Set arg_pointer_offset after using pretend_args_size.
14658 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14660 PR rtl-optimization/85431
14661 * dse.c (record_store): Ignore zero width stores.
14664 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
14665 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
14666 __builtin_stack_restore rather than after it.
14667 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
14668 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
14669 argument instead of virtual_dynamic_stack_rtx.
14671 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
14673 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
14675 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14676 Add note to error message to explain internal mapping of overloaded
14677 built-in function name to non-overloaded built-in function name.
14678 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
14681 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14684 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
14685 where the inputs overlap with the output.
14687 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14689 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
14690 (=v, v) alternative and explicit "memory" attribute.
14691 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
14692 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14694 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
14695 "sselog1" type instead of "sselog".
14696 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
14697 "sselog". Remove explicit "memory" attribute.
14698 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
14699 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14701 (vec_extract_hi_v32hi): Merge all alternatives into one, use
14702 "sselog1" type instead of "sselog". Remove explicit "memory"
14704 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
14705 use "sselog1" type instead of "sselog". Remove explicit "memory"
14707 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
14708 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14710 (vec_extract_hi_v64qi): Merge all alternatives into one, use
14711 "sselog1" type instead of "sselog". Remove explicit "memory"
14713 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
14714 use "sselog1" type instead of "sselog". Remove explicit "memory"
14718 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
14720 PR middle-end/85414
14721 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
14724 2018-04-17 Martin Jambor <mjambor@suse.cz>
14727 * ipa-cp.c (create_specialized_node): Call
14728 expand_all_artificial_thunks if necessary.
14730 2018-04-17 Martin Liska <mliska@suse.cz>
14733 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
14734 in message, remote space in between '_G' and '('.
14736 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14739 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
14740 avx512f_vmcmp<mode>3<round_saeonly_name>,
14741 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
14742 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
14743 avx512f_rndscale<mode><round_saeonly_name>,
14744 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
14745 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
14746 Use %<iptr>2 instead of %2 for -masm=intel.
14747 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
14748 avx512f_vcvttss2usi<round_saeonly_name>,
14749 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
14751 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
14752 avx512f_vcvttsd2usi<round_saeonly_name>,
14753 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
14754 Use %q1 instead of %1 for -masm=intel.
14755 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
14756 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
14757 of %3 for -masm=intel.
14758 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
14760 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
14762 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
14764 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
14765 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
14767 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
14769 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
14770 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
14771 %g1 and one with %0 and %1.
14772 (avx512er_vmrcp28<mode><round_saeonly_name>,
14773 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
14774 %1 for -masm=intel.
14775 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
14776 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
14777 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
14778 of %0 and %{%4%} for -masm=intel.
14779 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
14780 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
14781 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
14782 order of %0 and %{%5%}%{z%} for -masm=intel.
14784 2018-04-17 Jan Hubicka <jh@suse.cz>
14787 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
14789 2018-04-17 Martin Liska <mliska@suse.cz>
14792 * multiple_target.c (create_dispatcher_calls): Set apostrophes
14793 for target_clone error message. Make default implementation
14794 clone to be a local declaration.
14795 (separate_attrs): Add new argument and check for an empty
14797 (expand_target_clones): Handle it.
14798 (ipa_target_clone): Make redirection just for target_clones
14801 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
14802 Tom de Vries <tom@codesourcery.com>
14804 PR middle-end/84955
14805 * omp-expand.c (expand_oacc_for): Add dummy false branch for
14806 tiled basic blocks without omp continue statements.
14808 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
14811 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
14812 vec_extract expression as having side effects to make sure it gets
14815 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
14818 * config/i386/i386.c (get_builtin_code_for_version): Check
14821 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
14824 * config.gcc: Support "skylake".
14825 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14827 * config/i386/i386.c (m_SKYLAKE): Define.
14828 (processor_target_table): Add "skylake".
14829 (ix86_option_override_internal): Add "skylake".
14830 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
14831 PROCESSOR_CANNONLAKE.
14832 (get_builtin_code_for_version): Fix priority for
14833 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
14834 PROCESSOR_SKYLAKE-AVX512.
14835 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
14836 (processor_type): Add PROCESSOR_SKYLAKE.
14838 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
14839 Jason Merrill <jason@redhat.com>
14842 * convert.c (convert_to_integer_1): Use direct recursion for
14843 enumeral types and types with a precision less than the number
14844 of bits in their mode.
14846 2018-04-16 Julia Koval <julia.koval@intel.com>
14849 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
14850 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
14852 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
14855 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
14856 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
14857 and -mno-direct-move.
14859 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
14862 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
14863 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
14864 Ensure negative shifts result in {0}.
14866 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
14868 PR rtl-optimization/79916
14869 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
14870 regs (if any) to define how to gnerate SD moves when LRA is in
14873 2018-04-13 Jakub Jelinek <jakub@redhat.com>
14875 PR rtl-optimization/85393
14876 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
14877 * except.c (expand_dw2_landing_pad_for_region): Make static.
14878 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
14879 a label and unconditional jump to old_bb, rather than
14880 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
14883 PR rtl-optimization/85376
14884 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
14885 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
14886 instead of a specific value.
14888 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
14889 Bin Cheng <bin.cheng@arm.com>
14891 PR tree-optimization/82965
14892 PR tree-optimization/83991
14893 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
14894 by_profile_only parameter.
14895 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
14896 information if the loop was predicted to iterate too many times.
14897 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
14899 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
14902 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
14905 2018-04-13 Martin Liska <mliska@suse.cz>
14906 Jakub Jelinek <jakub@redhat.com>
14908 PR middle-end/81657
14909 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
14910 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
14911 * builtins.c (expand_builtin_memory_copy_args): Use
14912 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
14913 handle dest_addr == pc_rtx.
14915 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
14918 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
14919 asked to not generate direct moves.
14920 (fix_trunc<mode>si2_stfiwx): Similar.
14921 (fix_trunc<mode>si2_internal): Similar.
14923 2018-04-12 Jakub Jelinek <jakub@redhat.com>
14926 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
14927 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
14928 lookup if dest in some wider mode is known to be const0_rtx and
14929 if so, record permanent equivalence for it to be ZERO_EXTEND of
14930 the narrower mode destination.
14932 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
14934 * lto-streamer-out.c (output_function): Revert 259346.
14935 * omp-expand.c (expand_oacc_for): Likewise.
14937 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
14939 PR rtl-optimization/85354
14940 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
14941 * sel-sched.c (sel_global_init): ... here.
14943 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
14946 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
14947 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
14948 mode for PE-COFF targets.
14949 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
14950 (i386_pe_asm_lto_end): Likewise.
14951 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
14952 (TARGET_ASM_LTO_END): Likewise.
14953 * config/i386/winnt.c (saved_debug_info_level): New static variable.
14954 (i386_pe_asm_lto_start): New function.
14955 (i386_pe_asm_lto_end): Likewise.
14957 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
14958 Richard Biener <rguenther@suse.de>
14960 PR middle-end/84955
14961 * lto-streamer-out.c (output_function): Fix CFG loop state before
14963 * omp-expand.c (expand_oacc_for): Handle calls to internal
14964 functions like regular functions.
14966 2018-04-12 Richard Biener <rguenther@suse.de>
14969 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
14970 for the early LTO debug to properly generate references to it
14971 during DIE emission. Do not re-use that for the skeleton for
14973 (dwarf2out_early_finish): Likewise.
14975 2018-04-12 Jakub Jelinek <jakub@redhat.com>
14978 * config/i386/sse.md
14979 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
14980 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
14981 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
14982 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
14983 and output is a reg, avoid creating invalid lowpart subreg, but
14984 instead split into a 512-bit move. Don't split if not AVX512VL,
14985 input is xmm16+ reg and output is a mem.
14986 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
14987 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
14988 xmm16+ reg and output is a mem.
14990 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14992 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
14993 also for flag_dwarf2_cfi_asm.
14995 2018-04-12 Jakub Jelinek <jakub@redhat.com>
14997 PR rtl-optimization/85342
14998 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
14999 a bool scalar var inside of the loop instead. Don't try to update
15000 recog_data.operand after failed apply_change_group.
15002 2018-04-12 Tom de Vries <tom@codesourcery.com>
15005 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
15006 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
15007 array with flexible array member as array without given dimension.
15008 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
15009 argument for undefined param to true.
15011 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
15014 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15015 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
15016 from PowerPC section.
15017 * config/rs6000/sysv4.opt (mcall-): Improve help text.
15018 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
15019 help text that is too long.
15020 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
15021 help text that is too long.
15022 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
15023 help text that is too long.
15025 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
15027 * config/alpha/alpha.md (stack_probe_internal): Rename
15028 from "probe_stack". Update all callers.
15030 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15032 PR rtl-optimization/84566
15033 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
15034 sched_macro_fuse_insns.
15036 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15039 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
15040 (compute_block_dependences): ... from here.
15042 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15044 PR tree-optimization/85331
15045 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
15046 from int to HOST_WIDE_INT.
15048 2018-04-11 Martin Jambor <mjambor@suse.cz>
15051 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
15052 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
15053 not the same as the source val.
15054 (cgraph_edge_brings_value_p): New parameter.
15055 (gather_edges_for_value): Pass destination value to
15056 cgraph_edge_brings_value_p.
15057 (perhaps_add_new_callers): Likewise.
15058 (get_info_about_necessary_edges): Likewise and exclude values brought
15059 only by self-recursive edges.
15060 (create_specialized_node): Redirect only clones of self-calling edges.
15061 (+self_recursive_pass_through_p): New function.
15062 (find_more_scalar_values_for_callers_subset): Use it.
15063 (find_aggregate_values_for_callers_subset): Likewise.
15064 (known_aggs_to_agg_replacement_list): Removed.
15065 (decide_whether_version_node): Re-calculate known constants for all
15066 remaining context clones.
15068 2018-04-11 Richard Biener <rguenther@suse.de>
15071 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
15072 from early DWARF output.
15073 (dwarf2out_early_finish): Output line info unconditionally into
15074 early DWARF and add reference to it.
15076 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15079 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
15080 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
15081 other than V2DFmode using iptr mode attribute.
15082 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
15084 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15086 PR rtl-optimization/84659
15087 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
15089 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15092 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
15094 (output_loc_list): Pass address of a dummy size variable even in the
15095 locview handling loop.
15096 (index_location_lists): Add comment on why skip_loc_list_entry can't
15099 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
15102 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
15105 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
15108 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15109 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
15110 and -mstring-compare-inline-limit.
15112 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15115 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
15116 for stack clash protection in a register whenever we need it to be in
15119 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15121 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
15122 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
15124 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15127 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
15129 (mlong-double-): Ditto.
15130 * config/rs6000/sysv4.opt (msdata=): Ditto.
15131 (mtls-size=): Ditto.
15133 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15135 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
15136 erroneous entries for
15137 "vector int vec_ldl (int, long int *)", and
15138 "vector unsigned int vec_ldl (int, unsigned long int *)".
15139 Add comments and entries for
15140 "vector bool char vec_ldl (int, bool char *)",
15141 "vector bool short vec_ldl (int, bool short *)",
15142 "vector bool int vec_ldl (int, bool int *)",
15143 "vector bool long long vec_ldl (int, bool long long *)",
15144 "vector pixel vec_ldl (int, pixel *)",
15145 "vector long long vec_ldl (int, long long *)",
15146 "vector unsigned long long vec_ldl (int, unsigned long long *)".
15147 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
15148 type tree bool_long_long_type_node and correct definition of
15149 bool_V2DI_type_node to make reference to this new type tree.
15150 (rs6000_mangle_type): Replace erroneous reference to
15151 bool_long_type_node with bool_long_long_type_node.
15152 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
15153 comments to emphasize sign distinctions for char and int types and
15154 replace RS6000_BTI_bool_long constant with
15155 RS6000_BTI_bool_long_long constant. Also add comment to restrict
15156 use of RS6000_BTI_pixel.
15157 (bool_long_type_node): Remove this macro definition.
15158 (bool_long_long_type_node): New macro definition
15160 2018-04-10 Jakub Jelinek <jakub@redhat.com>
15162 PR rtl-optimization/85300
15163 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
15164 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
15165 simplify_unary_operation fails.
15167 2018-04-10 Martin Liska <mliska@suse.cz>
15169 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
15170 cgraph_edge and ipa_ref.
15172 2018-04-10 Jakub Jelinek <jakub@redhat.com>
15176 * config/i386/sse.md
15177 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
15178 computation of the VEC_MERGE selector from mask.
15179 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
15180 Fix decoding of the VEC_MERGE selector into mask.
15182 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
15184 PR tree-optimization/85286
15185 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
15187 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
15189 * final.c (final_1): Set insn_last_address as well as
15190 insn_current_address.
15192 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15195 * explow.c (emit_stack_probe): Call validize_mem on memory location
15196 before passing it to gen_probe_stack. Create address operand and
15197 legitimize it for the probe_stack_address case.
15199 2018-04-09 Jan Hubicka <jh@suse.cz>
15202 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
15203 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
15204 * tree.c (free_lang_data_in_type): Fix handling of binfos;
15206 (free_lang_data): Rebuild type inheritance graph.
15208 2018-04-09 Martin Sebor <msebor@redhat.com>
15210 * invoke.texi (-finline-small-functions): Mention other optimization
15212 (-findirect-inlining, -fpartial-inlining): Same.
15213 (-finline-functions-called-once): Same.
15214 (-freorder-blocks-and-partition): Same.
15216 2018-04-09 Jan Hubicka <jh@suse.cz>
15219 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
15220 jumps; choose last target that matches the criteria (i.e.
15221 no partition changes for non-crossing jumps).
15222 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
15223 support for redirecting crossing jumps to non-crossing.
15225 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
15227 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
15228 also for naked functions.
15230 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
15232 * config/arc/arc.md (add_shift): New pattern.
15233 (add_shift2): Likewise.
15234 (sub_shift): Likewise.
15235 (sub_shift_cmp0_noout): Likewise.
15236 (compare_si_ashiftsi): Likewise.
15237 (xbfu_cmp0_noout): New combine pattern.
15238 (xbfu_cmp0"): Likewise.
15239 (movsi_set_cc_insn): Place the predicable variant first.
15240 (commutative_binary_cmp0_noout): Remove clobber.
15241 (commutative_binary_cmp0): New pattern.
15242 (noncommutative_binary_cmp0): Likewise.
15243 (noncommutative_binary_cmp0_noout): Likewise.
15244 (noncommutative_binary_comparison_result_used): Removed.
15245 (rsub_cmp0): New pattern.
15246 (rsub_cmp0_noout): Likewise.
15247 (extzvsi): Changed, keep only meaningful variants.
15248 (SQH, SEZ): New iterators.
15249 (SQH_postfix): New mode attribute.
15250 (SEZ_prefix): New code attribute.
15251 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
15252 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
15253 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
15254 of numerical value.
15255 (noncommutative_operator): Check the availability of barrel
15258 2018-04-09 Richard Biener <rguenther@suse.de>
15260 PR tree-optimization/85284
15261 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
15262 Only use the niter constraining form of simple_iv when the exit
15263 is always executed.
15265 2018-04-09 Tom de Vries <tom@codesourcery.com>
15268 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
15269 (define_expand "*memory_barrier"): New define_expand.
15270 (define_insn "memory_barrier"): New insn.
15272 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15274 PR rtl-optimization/80463
15275 PR rtl-optimization/83972
15276 PR rtl-optimization/83480
15278 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
15279 correct producer for the insn.
15280 (tidy_control_flow): Fixup seqnos in case of debug insns.
15282 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15284 PR rtl-optimization/83913
15286 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
15287 different sched-times when merging exprs.
15289 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15291 PR rtl-optimization/83962
15293 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
15294 tidy_fallthru_edge and tidy_control_flow.
15296 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15298 PR rtl-optimization/83530
15300 * sel-sched.c (force_next_insn): New global variable.
15301 (remove_insn_for_debug): When force_next_insn is true, also leave only
15302 next insn in the ready list.
15303 (sel_sched_region): When the region wasn't scheduled, make another pass
15304 over it with force_next_insn set to 1.
15306 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
15308 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
15310 * config/nds32/constants.md (unspec_volatile_element): Add enum values
15311 for interrupt control.
15312 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15313 functions for interrupt control.
15314 * config/nds32/nds32-intrinsic.md: Likewise.
15315 * config/nds32/nds32_intrinsic.h: Likewise.
15316 * config/nds32/nds32.h (nds32_builtins): Likewise.
15318 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
15320 * config/nds32/nds32.c (nds32_init_machine_status,
15321 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
15322 strict_aligned_p field.
15323 (nds32_expand_to_rtl_hook): New function.
15324 (TARGET_EXPAND_TO_RTL_HOOK): Define.
15325 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
15327 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15328 Chung-Ju Wu <jasonwucj@gmail.com>
15330 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
15331 * config/nds32/nds32-n7.md: New file.
15332 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
15333 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
15335 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
15336 * config/nds32/nds32.md (pipeline_model): Add n7.
15337 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
15338 * config/nds32/pipelines.md: Include n7 settings.
15340 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15341 Chung-Ju Wu <jasonwucj@gmail.com>
15343 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
15344 * config/nds32/nds32-e8.md: New file.
15345 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
15346 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
15348 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
15349 * config/nds32/nds32.md (pipeline_model): Add e8.
15350 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
15351 * config/nds32/pipelines.md: Include e8 settings.
15353 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15354 Chung-Ju Wu <jasonwucj@gmail.com>
15356 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
15357 * config/nds32/nds32-n8.md: New file.
15358 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
15359 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
15361 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
15362 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
15363 * config/nds32/nds32.md (pipeline_model): Add n8.
15364 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
15365 * config/nds32/pipelines.md: Include n8 settings.
15367 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15368 Chung-Ju Wu <jasonwucj@gmail.com>
15370 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
15371 * config/nds32/nds32-n9-2r1w.md: New file.
15372 * config/nds32/nds32-n9-3r2w.md: New file.
15373 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
15374 nds32_register_ports): New or modify for cpu n9.
15375 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
15377 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
15378 * config/nds32/nds32-utils.c: New file.
15379 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
15380 TARGET_MUL_SLOW): Define.
15381 * config/nds32/nds32.md (pipeline_model): New attribute.
15382 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
15383 New options that support cpu n9.
15384 * config/nds32/pipelines.md: Include n9 settings.
15385 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
15387 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
15389 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
15390 information if necessary.
15391 (output_cond_branch_compare_zero): Likewise.
15392 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
15393 (nds32_target_alignment): Refine for alignment.
15394 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
15395 (FUNCTION_BOUNDARY): Modify.
15396 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
15398 * config/nds32/nds32.opt (malways-align, malign-functions): New.
15400 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
15402 * config/nds32/constants.md (unspec_volatile_element): Add values for
15403 TLB operation and data prefetch.
15404 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15405 functions for TLB operation and data prefetch.
15406 * config/nds32/nds32-intrinsic.md: Likewise.
15407 * config/nds32/nds32_intrinsic.h: Likewise.
15408 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
15409 (nds32_print_operand): Likewise.
15410 * config/nds32/nds32.h (nds32_builtins): Likewise.
15412 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
15413 Andrew Pinski <pinsika@gcc.gnu.org>
15415 PR middle-end/82976
15416 * match.pd: Use constant_boolean_node of correct type instead of
15417 boolean_true_node or boolean_false_node for simplifying
15418 pointer comparisons to zero.
15420 2018-04-07 Jakub Jelinek <jakub@redhat.com>
15422 PR tree-optimization/80021
15423 * tree.c (verify_type_variant): Make error call in verify_variant_match
15424 translatable and remove final full stop.
15426 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15428 * config/nds32/constants.md (unspec_volatile_element): Add
15429 UNSPEC_VOLATILE_EH_RETURN.
15430 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
15431 nds32_output_stack_pop): Support dwarf exception handling process.
15432 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
15433 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
15434 exception handling process.
15435 (nds32_compute_stack_frame): Likewise.
15436 (nds32_return_addr_rtx): Likewise.
15437 (nds32_initial_elimination_offset): Likewise.
15438 (nds32_expand_prologue): Likewise.
15439 (nds32_expand_epilogue): Likewise.
15440 (nds32_dynamic_chain_address): New function.
15441 * config/nds32/nds32.h (machine_function): Add fields for dwarf
15442 exception handling.
15443 (DYNAMIC_CHAIN_ADDRESS): Define.
15444 (EH_RETURN_DATA_REGNO): Define.
15445 (EH_RETURN_STACKADJ_RTX): Define.
15446 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
15447 patterns for dwarf exception handling.
15449 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15451 * config/nds32/nds32.h: Clean up obsolete macros.
15453 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15455 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15456 Add enum values for particular instructions.
15457 * config/nds32/nds32-intrinsic.c: Implementation of expanding
15458 particular intrinsic functions.
15459 * config/nds32/nds32-intrinsic.md: Likewise.
15460 * config/nds32/nds32_intrinsic.h: Likewise.
15461 * config/nds32/nds32.h (nds32_builtins): Likewise.
15462 * config/nds32/nds32.md (type): Add pbsad and pbsada.
15463 (btst, ave): New patterns for particular instructions.
15465 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15467 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15468 Add enum values for atomic load/store and memory sync.
15469 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
15471 * config/nds32/nds32-intrinsic.md: Likewise.
15472 * config/nds32/nds32_intrinsic.h: Likewise.
15473 * config/nds32/nds32.h (nds32_builtins): Likewise.
15475 2018-04-07 Jakub Jelinek <jakub@redhat.com>
15477 PR tree-optimization/85257
15478 * fold-const.c (native_encode_vector): If not all elts could fit
15479 and off is -1, return 0 rather than offset.
15480 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
15481 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
15482 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
15483 adjust buffer in native_interpret_expr call.
15485 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15487 * config/nds32/constants.md (unspec_volatile_element): Add cache
15488 control enum values.
15489 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
15490 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
15491 * config/nds32/nds32.c (nds32_cctl_names): New.
15492 (nds32_print_operand): Handle cache control register names.
15493 * config/nds32/nds32.h (nds32_builtins): New enum values.
15494 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
15496 * config/nds32/nds32.md (type): Add mmu.
15497 * config/nds32/pipelines.md (simple_insn): Add mmu.
15499 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15501 * config/nds32/nds32.md (type): Remove call.
15502 * config/nds32/pipelines.md (simple_insn): Likewise.
15504 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15506 * config/nds32/constants.md (unspec_volatile_element): Add
15507 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
15508 UNSPEC_VOLATILE_FMFCFG.
15509 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
15510 description for fmfcfg and fmfcsr.
15511 (bdesc_1arg): Add fmtcsr.
15512 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
15513 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
15514 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
15515 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
15516 unspec_fmfcfg): New patterns.
15517 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
15518 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
15519 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
15520 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
15521 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
15522 __nds32__fmfcfg): Define.
15524 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15526 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
15527 intrinsic register names.
15528 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
15529 intrinsic register enum values and macros.
15531 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15533 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
15534 for load/store addressing form.
15535 (nds32_print_operand_address): Likewise.
15537 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
15540 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
15541 based on LABEL_REF. Remove useless assertion.
15542 (pic_address_needs_scratch): Fix formatting.
15543 (sparc_legitimize_pic_address): Minor tweaks.
15544 (sparc_delegitimize_address): Adjust assertion accordingly.
15545 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
15546 into symbolic_operand.
15547 (movsi_high_pic_label_ref): Likewise.
15548 (movsi_lo_sum_pic_label_ref): Likewise.
15549 (movdi_pic_label_ref): Likewise.
15550 (movdi_high_pic_label_ref): Likewise.
15551 (movdi_lo_sum_pic_label_ref): Likewise.
15553 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
15555 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
15556 custom LIB_SPEC setup.
15558 2018-04-06 Ruslan Bukin <br@bsdpad.com>
15559 Kito Cheng <kito.cheng@gmail.com>
15561 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
15562 * config/riscv/freebsd.h: New.
15564 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
15566 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
15567 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
15570 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
15571 Kito Cheng <kito.cheng@gmail.com>
15573 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
15574 nds32_output_call, nds32_symbol_binds_local_p): New functions.
15575 * config/nds32/nds32-protos.h (nds32_output_call,
15576 nds32_output_return): Declare.
15577 * config/nds32/nds32.md: Refine all the call and return patterns.
15579 2018-04-06 Jakub Jelinek <jakub@redhat.com>
15582 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
15583 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
15585 PR rtl-optimization/84872
15586 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
15587 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
15588 EDGE_CROSSING edge.
15590 2018-04-06 Tamar Christina <tamar.christina@arm.com>
15592 * expr.c (copy_blkmode_to_reg): Revert 254862.
15593 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
15595 2018-04-06 Richard Biener <rguenther@suse.de>
15597 PR middle-end/85244
15598 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
15599 after seeing a component reference with an adjacent field. Treat
15600 refs to arrays at struct end of external decls similar to
15601 refs to unconstrained commons.
15603 2018-04-06 Jakub Jelinek <jakub@redhat.com>
15606 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
15607 look through SAVE_EXPRs with non-side-effects argument. Adjust
15609 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
15612 2018-04-06 Richard Biener <rguenther@suse.de>
15614 PR middle-end/85180
15615 * alias.c (find_base_term): New wrapper around find_base_term
15616 unwinding CSELIB_VAL_PTR changes.
15617 (find_base_term): Do not restore CSELIB_VAL_PTR during the
15620 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15622 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
15624 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
15625 constant definitions.
15626 ("nop"): lr 0,0 -> nopr r0
15627 ("nop_lr0", "nop_lr1"): New insn definitions.
15629 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
15631 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
15632 NDS32_V3PUSH_AVAILABLE_P macro.
15634 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
15635 Chung-Ju Wu <jasonwucj@gmail.com>
15637 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
15638 (nds32*-*-*): Add float and fpu_config into supported_defaults.
15639 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
15640 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
15641 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
15642 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
15643 * config/nds32/constraints.md: New constraints and checking for hard
15644 float configuration.
15645 * config/nds32/iterators.md: New mode iterator and attribute for hard
15646 float configuration.
15647 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
15649 * config/nds32/nds32-fpu.md: New file.
15650 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
15651 deal with hard float code generation.
15652 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
15654 (abi_type, float_reg_number): New enum type.
15655 * config/nds32/nds32-predicates.c: New predicates for hard float.
15656 * config/nds32/nds32-protos.h: Declare functions for hard float.
15657 * config/nds32/nds32.c: Implementation for hard float configuration.
15658 * config/nds32/nds32.h: Definitions for hard float configuration.
15659 * config/nds32/nds32.md: Include hard float machine description and
15660 modify patterns for hard float configuration.
15661 * config/nds32/nds32.opt: New options for hard float configuration.
15662 * config/nds32/predicates.md: New predicates for hard float
15665 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
15667 * common/config/nds32/nds32-common.c
15668 (nds32_option_optimization_table): Enable -mreleax-hint by default.
15670 2018-04-05 Jakub Jelinek <jakub@redhat.com>
15672 PR middle-end/85195
15673 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
15674 CONSTRUCTOR_ELT (ctor, ...)->value.
15676 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
15679 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
15681 2018-04-05 Tom de Vries <tom@codesourcery.com>
15684 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
15687 2018-04-05 Shiva Chen <shiva0217@gmail.com>
15688 Kito Cheng <kito.cheng@gmail.com>
15690 * config/nds32/constraints.md (U33): Fine-tune checking condition.
15691 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
15692 * config/nds32/nds32.h (nds32_16bit_address_type): Add
15693 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
15695 2018-04-05 Shiva Chen <shiva0217@gmail.com>
15696 Kito Cheng <kito.cheng@gmail.com>
15698 * config/nds32/constraints.md (Ufe): New memory constraint.
15699 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
15700 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
15701 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
15703 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
15704 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
15706 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15708 * config/nds32/nds32.md: Use optimize_size in the condition for
15709 alu-shift instructions.
15711 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15713 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
15715 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15717 * config/nds32/nds32.md (negsi2): Refine pattern.
15719 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
15720 Chung-Ju Wu <jasonwucj@gmail.com>
15722 * config/nds32/iterators.md (shift_rotate): New code iterator.
15723 (shift): New code attribute.
15724 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
15725 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
15726 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
15727 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
15728 bit-wise operations.
15729 (andsi3, *andsi3): Ditto.
15730 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
15731 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
15732 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
15733 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
15734 nds32_ior_operand, nds32_xor_operand): New predicates.
15736 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15738 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
15739 (addsi3, subsi3): ... this.
15741 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15743 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
15745 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15747 * config/nds32/nds32.md: Adjust indention.
15749 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
15751 * config/nds32/nds32.md (feature): New attribute.
15753 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15755 * config/nds32/nds32.md (subtype): New attribute.
15757 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
15760 * config/arm/arm-builtins.c (arm_expand_builtin): Change
15761 expansion to perform a bitwise AND of the argument followed by a
15762 boolean negation of the result.
15764 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
15766 PR rtl-optimization/84878
15767 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
15768 the basic block. Assert the use reference is not artificial and that
15769 it has an associated insn.
15771 2018-04-04 Michael Matz <matz@suse.de>
15773 * builtins.c (compute_objsize): Pass correct operand
15774 to array_at_struct_end_p.
15776 2018-04-04 Richard Biener <rguenther@suse.de>
15779 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
15780 from contexts for DINFO_LEVEL_TERSE and below.
15782 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15784 * config/nds32/nds32-doubleword.md (move_<mode>): Require
15785 resiter_operand condition.
15786 * config/nds32/nds32.md (*move<mode>): Ditto.
15788 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15789 Monk Chiang <sh.chiang04@gmail.com>
15791 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
15793 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15795 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
15797 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15798 Kito Cheng <kito.cheng@gmail.com>
15800 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
15801 nds32_cond_code_str, output_cond_branch,
15802 output_cond_branch_compare_zero, nds32_expand_cbranch,
15803 nds32_expand_cstore, nds32_expand_movcc,
15804 nds32_output_cbranchsi4_equality_zero,
15805 nds32_output_cbranchsi4_equality_reg,
15806 nds32_output_cbranchsi4_equality_reg_or_const_int,
15807 nds32_output_cbranchsi4_greater_less_zero: New functions.
15808 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
15809 nds32_expand_cstore, nds32_expand_movcc,
15810 nds32_output_cbranchsi4_equality_zero,
15811 nds32_output_cbranchsi4_equality_reg,
15812 nds32_output_cbranchsi4_equality_reg_or_const_int,
15813 nds32_output_cbranchsi4_greater_less_zero): Declare.
15814 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
15815 nds32_rimm11s_operand): New predicates.
15816 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
15817 * config/nds32/nds32.md: Rewrite all the branch and conditional move
15820 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15822 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
15823 * config/nds32/nds32.md: Ditto.
15824 * config/nds32/pipelines.md: Ditto.
15826 2018-04-04 Richard Biener <rguenther@suse.de>
15828 PR tree-optimization/85168
15829 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
15830 propagating abnormals.
15832 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15834 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
15836 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15837 Kito Cheng <kito.cheng@gmail.com>
15839 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
15840 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
15841 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
15842 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
15843 * config/nds32/nds32.md (sibcall_internal): New.
15844 (sibcall_register): Remove.
15845 (sibcall_immediate): Remove.
15846 (sibcall_value_internal): New.
15847 (sibcall_value_register): Remove.
15848 (sibcall_value_immediate): Remove.
15849 * config/nds32/predicates.md (nds32_general_register_operand): New.
15850 (nds32_call_address_operand): New.
15852 2018-04-03 Jakub Jelinek <jakub@redhat.com>
15854 PR rtl-optimization/85167
15855 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
15856 bb_defs if *split_p, instead preinitialize it to NULL.
15858 PR tree-optimization/85156
15859 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
15860 evaluating the argument multiple times.
15862 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
15864 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
15866 (_mm_cvtpd_ps): Likewise.
15867 (_mm_cvttpd_epi32): Likewise.
15868 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
15869 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
15870 vector, pixel, and bool following altivec.h include.
15872 2018-04-03 Martin Sebor <msebor@redhat.com>
15874 * doc/extend.texi (Common Function Attributes): Clarify.
15875 (const attribute): Likewise.
15876 (pure attribute): Likewise.
15878 2018-04-03 Jakub Jelinek <jakub@redhat.com>
15881 * config/i386/i386.c (ix86_expand_vector_set): Use
15882 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
15884 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
15886 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
15887 instructions when changing rounding bits to preserve precision bits
15888 in the x87 control word.
15890 2018-04-03 Martin Liska <mliska@suse.cz>
15892 PR tree-optimization/82491
15893 * rtl.h (strip_offset_and_add): Replace += suboffset with
15894 poly_uint64 () + suboffset.
15896 2018-03-29 Martin Liska <mliska@suse.cz>
15897 Martin Jambor <mjambor@suse.cz>
15900 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
15901 param_type is not an integral or pointer type.
15903 2018-04-03 Richard Biener <rguenther@suse.de>
15905 * sese.h (recompute_all_dominators): Remove.
15907 2018-04-02 Martin Sebor <msebor@redhat.com>
15909 * doc/invoke.texi (-Wrestrict): Fix typos.
15911 2018-04-02 Jim Wilson <jimw@sifive.com>
15913 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
15914 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
15915 (<optab>di3, <optab>si3_extend): Likewise.
15916 (<optab>si3_mask, <optab>si3_mask_1): New.
15917 (<optab>di3_mask, <optab>di3_mask_1): New.
15918 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
15919 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
15920 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
15922 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
15924 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
15927 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
15929 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
15930 (nds32_canonicalize_comparison): New function.
15932 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
15933 Kito Cheng <kito.cheng@gmail.com>
15934 Kuan-Lin Chen <kuanlinchentw@gmail.com>
15936 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
15937 * config/nds32/constants.md (unspec_volatile_element): Add
15938 UNSPEC_VOLATILE_RELAX_GROUP.
15939 * config/nds32/nds32-relax-opt.c: New file.
15940 * config/nds32/nds32-predicates.c
15941 (nds32_symbol_load_store_p): New function.
15942 * config/nds32/nds32-protos.h
15943 (nds32_symbol_load_store_p): Declare function.
15944 (make_pass_nds32_relax_opt): Declare new rtl pass function.
15945 * config/nds32/nds32.c
15946 (nds32_register_pass): New function to register pass.
15947 (nds32_register_passes): New function to register passes.
15948 * config/nds32/nds32.md (relax_group): New pattern.
15949 * config/nds32/nds32.opt (mrelax-hint): New option.
15950 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
15952 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
15954 * config/nds32/t-nds32: Modify files dependency.
15956 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
15958 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
15959 (PROFILE_HOOK): Define its implementation.
15961 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
15963 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
15964 type and 32-bit size.
15966 2018-04-01 Jakub Jelinek <jakub@redhat.com>
15968 PR middle-end/85090
15969 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
15970 (V_128_256): New mode iterator.
15971 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
15972 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
15973 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
15975 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
15976 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
15978 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
15981 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
15982 NaN inputs correctly.
15984 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
15987 * config/rs6000/vsx.md (??r): New mode attribute.
15988 (*vsx_mov<mode>_64bit): Use it.
15989 (*vsx_mov<mode>_32bit): Likewise.
15991 2018-03-30 Martin Sebor <msebor@redhat.com>
15993 PR tree-optimization/84818
15994 * builtins.c (check_access): Use warning_n.
15996 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15999 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
16001 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
16004 2018-03-30 Julia Koval <julia.koval@intel.com>
16007 * x86-tune.def (movx, partial_reg_dependency): Enable for
16010 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
16012 PR inline-asm/84985
16013 * lra-constraints.c (process_alt_operands): Move setting
16014 this_alternative_matches below.
16016 2018-03-29 Martin Liska <mliska@suse.cz>
16019 * doc/invoke.texi: Document how LTO works with debug info.
16020 Describe auto-load support of binutils. Mention 'x86-64'
16021 as valid option value of -march option.
16023 2018-03-29 Jakub Jelinek <jakub@redhat.com>
16025 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
16028 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
16029 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
16030 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
16033 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
16036 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
16037 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
16038 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
16039 for RS6000_BTM_POWERPC64.
16040 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
16041 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
16042 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
16045 (DIVDEU): Likewise.
16047 2018-03-28 Carl Love <cel@us.ibm.com>
16050 2017-09-27 Carl Love <cel@us.ibm.com>
16052 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
16053 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
16054 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
16057 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16059 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
16060 instead of __vector bool.
16061 (_mm_max_pu8): Likewise.
16062 (_mm_min_pi16): Likewise.
16064 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
16067 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
16068 (DIVWEUO): Likewise.
16069 (DIVDEO): Likewise.
16070 (DIVDEUO): Likewise.
16071 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
16072 DIVWEUO and DIVDEUO.
16073 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
16074 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
16075 (div_extend): Likewise.
16076 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
16078 (__builtin_divweuo): Likewise.
16079 (__builtin_divdeo): Likewise.
16080 (__builtin_divdeuo): Likewise.
16082 2018-03-28 Jakub Jelinek <jakub@redhat.com>
16085 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
16086 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
16088 PR tree-optimization/82004
16089 * gimple-match-head.c (optimize_pow_to_exp): New function.
16090 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
16091 Don't fold to exp if optimize_pow_to_exp is false.
16093 2018-03-28 Martin Liska <mliska@suse.cz>
16096 * calls.c (initialize_argument_information): Fix trailing space.
16097 * common.opt: Fix typo and provide better explanation for
16098 -fsanitize-coverage option.
16099 * config/i386/i386.opt: Fix typo.
16101 2018-03-28 Jakub Jelinek <jakub@redhat.com>
16102 Martin Liska <mliska@suse.cz>
16105 * gimplify.c (asan_poison_variable): Don't do the check for
16106 gimplify_omp_ctxp here.
16107 (gimplify_decl_expr): Do it here.
16108 (gimplify_target_expr): Likewise.
16110 2018-03-28 Martin Liska <mliska@suse.cz>
16113 * config/i386/i386.c (ix86_function_arg_advance): Do not call
16114 chkp_type_bounds_count if MPX is not enabled.
16116 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
16118 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
16120 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
16123 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
16124 function to create the function decl for complex long double
16125 multiply and divide for -mabi=ieeelongdouble.
16126 (init_float128_ieee): Call it.
16128 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
16131 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
16132 -fcf-protection=branch -mibt.
16133 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
16135 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16138 * config/arm/arm.c (arm_valid_symbolic_address): Handle
16139 arm_word_relocations.
16141 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
16144 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
16145 extern array declarations.
16147 2018-03-27 Richard Biener <rguenther@suse.de>
16149 PR middle-end/84067
16150 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
16151 explicit single_use checks.
16153 2018-03-27 Richard Biener <rguenther@suse.de>
16155 PR tree-optimization/85082
16156 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
16159 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16161 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
16162 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
16163 Turn on fasynchronous-unwind-tables and funwind-tables.
16165 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
16168 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
16169 (*bmi_blsr_<mode>_ccz): Ditto.
16171 2018-03-26 Tom de Vries <tom@codesourcery.com>
16173 PR tree-optimization/85063
16174 * omp-general.c (offloading_function_p): New function. Factor out
16176 * omp-offload.c (pass_omp_target_link::gate): ... here.
16177 * omp-general.h (offloading_function_p): Declare.
16178 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
16179 with attribute omp declare target for offloading functions.
16181 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
16183 PR tree-optimization/84005
16184 * tree-data-ref.h (get_base_for_alignment): Declare.
16185 * tree-data-ref.c (get_base_for_alignment_1): New function.
16186 (get_base_for_alignment): Likewise.
16187 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
16188 get_base_for_alignment to find a suitable base object, instead
16189 of always using drb->base_address.
16191 2018-03-23 Jakub Jelinek <jakub@redhat.com>
16193 PR inline-asm/85022
16194 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
16195 known size by default.
16197 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
16199 PR inline-asm/85030
16200 * lra-constraints.c (process_alt_operands): Don't match BLKmode
16201 and non BLKmode operands.
16203 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16206 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
16207 Clean up attributes.
16209 2018-03-23 Richard Biener <rguenther@suse.de>
16212 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
16213 we are going to emit early debug for LTO.
16215 2018-03-23 Jakub Jelinek <jakub@redhat.com>
16217 PR inline-asm/85034
16218 * function.c (match_asm_constraints_1): Don't optimize if input
16219 doesn't satisfy general_operand predicate for output's mode.
16221 PR inline-asm/85022
16222 * alias.c (write_dependence_p): Don't require for x_canonicalized
16223 non-VOIDmode if x has VOIDmode.
16226 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
16227 just don't try to optimize it rather than assert it never happens.
16229 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
16231 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
16232 macro expansions for definition of ST_INTERNAL_<mode> and
16233 LD_INTERNAL_<mode> builtins.
16234 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
16236 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
16238 (altivec_expand_st_builtin): Likewise.
16239 (altivec_expand_builtin): Remove calls to deleted functions.
16240 (rs6000_address_for_altivec): Delete this function.
16241 * config/rs6000/vector.md: Remove expands for
16242 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
16244 2018-03-22 Sudakshina Das <sudi.das@arm.com>
16247 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
16248 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
16249 re-computing once computed.
16250 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
16251 (arm_init_machine_status): Initialize
16252 machine->static_chain_stack_bytes.
16254 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
16257 * doc/extend.texi: Add four new prototypes for vec_ld.
16258 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
16259 definitions for more logical presentation.
16260 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
16261 entries for V1TI variants of __builtin_altivec_ld builtin.
16262 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
16263 handling of V1TI variant of LVX icode pattern.
16264 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
16265 (rs6000_gimple_fold_builtin): Likewise.
16266 (altivec_init_builtins): Add code to define
16267 __builtin_altivec_lvx_v1ti function.
16269 2018-03-22 Jakub Jelinek <jakub@redhat.com>
16271 PR inline-asm/84941
16272 * function.c (match_asm_constraints_1): Don't do the optimization
16273 if input isn't a REG, SUBREG, MEM or constant.
16275 2018-03-22 Tom de Vries <tom@codesourcery.com>
16277 PR tree-optimization/84956
16278 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
16279 bb_has_abnormal_pred.
16281 2018-03-22 Jakub Jelinek <jakub@redhat.com>
16284 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
16285 DECL_INITIAL (decl) to decl at the end.
16286 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
16287 adjust the comment.
16289 2018-03-21 Joseph Myers <joseph@codesourcery.com>
16291 * doc/extend.texi (__builtin_tgmath): Document when complex
16292 integer types are treated as _Complex _Float64.
16294 2018-03-21 Tom de Vries <tom@codesourcery.com>
16296 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
16298 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16300 PR tree-optimization/84960
16301 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
16302 if it is ENTRY block, move them into single succ of ENTRY in that case.
16304 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
16306 PR tree-optimization/84811
16307 * poly-int.h (poly_span_traits): Remove the T3 parameter and
16308 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
16309 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
16310 (known_subrange_p): Update accordingly. Cast each value involved
16311 in the size comparison, rather than casting the result of the
16314 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16316 PR tree-optimization/84982
16317 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
16318 by flipping the least significant bit rather than all bits from
16319 bitpos to bitpos + bitsize - 1.
16321 2018-03-21 Nathan Sidwell <nathan@acm.org>
16323 * doc/extend.texi (Deprecated Features): Remove mention of
16324 long-deleted deprecations.
16326 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16329 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
16330 * configure: Regenerate.
16332 2018-03-21 Tom de Vries <tom@codesourcery.com>
16334 PR tree-optimization/83126
16335 * tree-parloops.c (num_phis): New function.
16336 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
16338 2018-03-21 Nathan Sidwell <nathan@acm.org>
16340 * doc/extend.texi (Deprecated Features): Update deprecated flags,
16341 mention anon-struct/union members and trailing attributes.
16343 2018-03-21 Bin Cheng <bin.cheng@arm.com>
16345 PR tree-optimization/84969
16346 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
16347 builtin memset partitions if they set different rhs values.
16349 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16351 PR rtl-optimization/84989
16352 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
16353 VEC_DUPLICATE with scalar result mode.
16355 2018-03-21 Martin Liska <mliska@suse.cz>
16358 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
16359 not intended return statement.
16361 2018-03-21 Martin Liska <mliska@suse.cz>
16364 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
16365 (chkp_find_bound_slots_1): Limit number of iterations.
16367 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
16370 * Minor grammar fixes for x86 options.
16372 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16375 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
16376 holding REG_CFA_RESTORE notes, instead turn them into a USE.
16378 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
16381 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
16382 (altivec_lvx_<mode>_1op): Likewise.
16383 (altivec_stvx_<mode>_2op): Likewise.
16384 (altivec_stvx_<mode>_1op): Likewise.
16385 (altivec_lvx_<VM2:mode>): New define_expand.
16386 (altivec_stvx_<VM2:mode>): Likewise.
16387 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
16388 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
16389 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
16390 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
16391 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
16392 (rs6000_gen_lvx): Likewise.
16393 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
16394 (altivec_expand_stv_builtin): Likewise.
16395 (altivec_expand_builtin): Likewise.
16396 * config/rs6000/vector.md: Likewise.
16398 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16401 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
16404 2018-03-20 Richard Biener <rguenther@suse.de>
16407 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
16408 sign-conversions as zero, fall back to standard scalar_stmt
16411 2018-03-20 Martin Liska <mliska@suse.cz>
16414 * predict.c (rebuild_frequencies): Handle case when we have
16415 PROFILE_ABSENT, but flag_guess_branch_prob is false.
16417 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16420 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
16421 flag_section_anchors.
16422 * varasm.c (use_blocks_for_decl_p): Remove hack for
16423 dw2_force_const_mem.
16426 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
16428 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
16429 be created, use lowpart_subreg of operands[0] rather than operands[0]
16431 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
16432 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
16433 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
16434 and n constraint instead of aarch64_shift_imm_di and Usd.
16435 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
16436 (*aarch64_<optab>_reg_minus<mode>3): ... this.
16438 2018-03-20 Sudakshina Das <sudi.das@arm.com>
16441 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
16442 to favor GPR over NEON registers.
16443 (<shift>di3_neon): Likewise.
16445 2018-03-20 Tom de Vries <tom@codesourcery.com>
16448 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
16449 (nvptx_process_pars): Emit bar.sync asap and alap.
16451 2018-03-20 Tom de Vries <tom@codesourcery.com>
16454 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
16455 seen_label if seen_label is already set.
16457 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16460 * config/i386/i386.c (fold_builtin_cpu): For features above 31
16461 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
16462 Use 1U instead of 1. Formatting fixes.
16465 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
16466 instead of TREE_TYPE (s1) for the return value.
16468 2018-03-19 Jakub Jelinek <jakub@redhat.com>
16470 PR tree-optimization/84946
16471 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
16472 bitsize + bitsize in poly_uint64 rather than poly_int64.
16475 * dwarf2asm.c: Include fold-const.c.
16476 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
16477 of decl rather than decl itself.
16479 PR rtl-optimization/84643
16480 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
16482 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
16485 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
16486 calling assemble_variable.
16488 2018-03-19 Sudakshina Das <sudi.das@arm.com>
16491 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
16492 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
16494 2018-03-19 Jim Wilson <jimw@sifive.com>
16497 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
16498 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
16499 (riscv_first_stack_step): Likewise.
16500 (riscv_option_override): Use STACK_BOUNDARY instead of
16501 MIN_STACK_BOUNDARY.
16502 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
16503 MIN_STACK_BOUNDARY.
16504 (BIGGEST_ALIGNMENT): Set to 128.
16505 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
16506 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
16509 2018-03-19 Richard Biener <rguenther@suse.de>
16511 PR tree-optimization/84933
16512 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
16513 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
16515 2018-03-19 Richard Biener <rguenther@suse.de>
16517 PR tree-optimization/84859
16518 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
16519 (cond_if_else_store_replacement): Perform sinking operation on
16520 single-store BBs regardless of MAX_STORES_TO_SINK setting.
16521 Generalize what a BB with a single eligible store is.
16523 2018-03-19 Richard Biener <rguenther@suse.de>
16525 PR tree-optimization/84929
16526 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
16527 chrec_is_positive against non-chrec arg.
16529 2018-03-19 Tamar Christina <tamar.christina@arm.com>
16532 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
16534 2018-03-18 Martin Liska <mliska@suse.cz>
16536 PR rtl-optimization/84635
16537 * regrename.c (build_def_use): Use matches_mode only when
16540 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
16542 PR tree-optimization/84913
16543 * tree-vect-loop.c (vectorizable_reduction): Don't try to
16544 vectorize chains of COND_EXPRs.
16546 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
16548 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
16550 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
16552 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
16554 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
16556 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
16558 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
16559 Kito Cheng <kito.cheng@gmail.com>
16561 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
16562 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
16563 (nds32_adjust_reg_alloc_order): New function.
16564 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
16566 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
16568 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
16569 nds32_print_operand, nds32_print_operand_address): Use
16570 HOST_WIDE_INT_PRINT_DEC instead.
16572 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
16574 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
16576 2018-03-17 Jakub Jelinek <jakub@redhat.com>
16579 * config/i386/i386.c (initial_ix86_tune_features,
16580 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
16581 unsigned long long.
16582 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
16583 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
16584 rather than 1u << ix86_tune. Formatting fix.
16585 (ix86_option_override_internal): Change ix86_arch_mask from
16586 unsigned int to unsigned HOST_WIDE_INT, initialize to
16587 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
16588 (ix86_function_specific_restore): Likewise.
16590 2018-03-16 Jakub Jelinek <jakub@redhat.com>
16593 * postreload.c (reload_combine_recognize_pattern): Perform
16594 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
16595 truncate_int_for_mode the result for the destination's mode.
16598 * hsa-gen.c (mem_type_for_type): Fix comment typo.
16599 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
16601 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
16604 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
16607 * lra-assigns.c (lra_split_hard_reg_for): Don't use
16608 regno_allocno_class_array and sorted_pseudos.
16609 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
16610 insns where regno is used.
16612 2018-03-16 Martin Liska <mliska@suse.cz>
16615 * multiple_target.c (create_dispatcher_calls): Redirect
16616 reference in the symbol table.
16618 2018-03-16 Martin Liska <mliska@suse.cz>
16621 * multiple_target.c (create_dispatcher_calls): Redirect also
16624 2018-03-16 Jakub Jelinek <jakub@redhat.com>
16628 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
16629 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
16630 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
16632 2018-03-16 Julia Koval <julia.koval@intel.com>
16634 * doc/invoke.texi (Skylake Server): Add CLWB.
16635 Cannonlake): Remove CLWB.
16637 2018-03-16 Jakub Jelinek <jakub@redhat.com>
16639 PR tree-optimization/84841
16640 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
16642 (FLOAT_ONE_CONST_TYPE): Define.
16643 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
16644 (sort_by_operand_rank): Put entries with higher constant_type last
16645 rather than first to match comments.
16647 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
16649 * config/nios2/nios2.md (movsi_internal): Fix thinko in
16652 2018-03-15 Jakub Jelinek <jakub@redhat.com>
16655 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
16656 check and use address of target always.
16658 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
16661 * config/i386/i386.c (indirect_thunk_needed): Update comments.
16662 (indirect_thunk_bnd_needed): Likewise.
16663 (indirect_thunks_used): Likewise.
16664 (indirect_thunks_bnd_used): Likewise.
16665 (indirect_return_needed): New.
16666 (indirect_return_bnd_needed): Likewise.
16667 (output_indirect_thunk_function): Add a bool argument for
16669 (output_indirect_thunk_function): Don't generate alias for
16670 function return thunk.
16671 (ix86_code_end): Call output_indirect_thunk_function to generate
16672 function return thunks.
16673 (ix86_output_function_return): Set indirect_return_bnd_needed
16674 and indirect_return_needed instead of indirect_thunk_bnd_needed
16675 and indirect_thunk_needed.
16677 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
16679 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
16680 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
16681 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
16683 2018-03-15 David Malcolm <dmalcolm@redhat.com>
16684 Paul Hua <paul.hua.gm@gmail.com>
16687 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
16689 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
16691 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
16692 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
16693 resp. SFmode cases.
16695 2018-03-15 Tamar Christina <tamar.christina@arm.com>
16698 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
16699 instead of GET_MODE_SIZE when comparing Units.
16701 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
16704 * varasm.c (hash_section): Return an unchangeble hash value
16705 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
16706 Return !aarch64_can_use_per_function_literal_pools_p ().
16708 2018-03-15 Jakub Jelinek <jakub@redhat.com>
16711 * optabs.c (emit_conditional_move): Pass address of cmode's copy
16712 rather than address of cmode as last argument to prepare_cmp_insn.
16714 2018-03-15 Julia Koval <julia.koval@intel.com>
16716 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
16717 F_AVX512VNNI, F_AVX512BITALG): New.
16719 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
16722 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
16723 insn for floating-point loads and stores.
16725 2018-03-14 Carl Love <cel@us.ibm.com>
16727 * config/rs6000/rs6000-c.c: Add macro definitions for
16728 ALTIVEC_BUILTIN_VEC_PERMXOR.
16729 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
16730 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
16731 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
16732 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
16734 * config/doc/extend.texi: Add prototypes for vec_permxor.
16736 2018-03-14 David Malcolm <dmalcolm@redhat.com>
16739 * diagnostic-show-locus.c (class layout_point): Convert m_line
16740 from int to linenum_type.
16741 (line_span::comparator): Use linenum "compare" function when
16742 comparing line numbers.
16743 (test_line_span): New function.
16744 (layout_range::contains_point): Convert param "row" from int to
16746 (layout_range::intersects_line_p): Likewise.
16747 (layout::will_show_line_p): Likewise.
16748 (layout::print_source_line): Likewise.
16749 (layout::should_print_annotation_line_p): Likewise.
16750 (layout::print_annotation_line): Likewise.
16751 (layout::print_leading_fixits): Likewise.
16752 (layout::annotation_line_showed_range_p): Likewise.
16753 (struct line_corrections): Likewise for field m_row.
16754 (line_corrections::line_corrections): Likewise for param "row".
16755 (layout::print_trailing_fixits): Likewise.
16756 (layout::get_state_at_point): Likewise.
16757 (layout::get_x_bound_for_row): Likewise.
16758 (layout::print_line): Likewise.
16759 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
16760 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
16761 * input.c (selftest::test_linenum_comparisons): New function.
16762 (selftest::input_c_tests): Call it.
16763 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
16764 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
16765 * selftest.h (ASSERT_GT): New macro.
16766 (ASSERT_GT_AT): New macro.
16767 (ASSERT_LT): New macro.
16768 (ASSERT_LT_AT): New macro.
16770 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
16772 PR rtl-optimization/84780
16773 * combine.c (distribute_links): Don't make a link based on pc_rtx.
16775 2018-03-14 Martin Liska <mliska@suse.cz>
16777 * tree.c (record_node_allocation_statistics): Use
16778 get_stats_node_kind.
16779 (get_stats_node_kind): New function extracted from
16780 record_node_allocation_statistics.
16781 (free_node): Use get_stats_node_kind.
16783 2018-03-14 Richard Biener <rguenther@suse.de>
16785 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
16786 that the value-set of ANTIC_IN doesn't grow.
16789 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
16791 (BB_VISITED_WITH_VISITED_SUCCS): New define.
16792 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
16794 2018-03-14 Julia Koval <julia.koval@intel.com>
16796 * config.gcc (icelake-client, icelake-server): New.
16798 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
16799 (initial_ix86_arch_features): Ditto.
16800 (PTA_SKYLAKE): Add SGX.
16801 (PTA_ICELAKE): Remove.
16802 (PTA_ICELAKE_CLIENT): New.
16803 (PTA_ICELAKE_SERVER): New.
16804 (ix86_option_override_internal): Split up icelake on icelake client and
16806 (get_builtin_code_for_version): Ditto.
16807 (fold_builtin_cpu): Ditto.
16808 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
16809 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
16810 * config/i386/i386.h (processor_type): Ditto.
16811 * doc/invoke.texi: Ditto.
16813 2018-03-14 Jakub Jelinek <jakub@redhat.com>
16816 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
16817 INTEGER_CST offset, add it together with bitpos / 8 and
16818 sign extend based on POINTER_SIZE.
16822 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
16825 * config/i386/constraints.md (Yc): New register constraint.
16826 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
16827 Use Yc constraint for alternative 2 of operand 0. Remove
16828 preferred_for_speed attribute.
16830 2018-03-14 Richard Biener <rguenther@suse.de>
16832 PR tree-optimization/84830
16833 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
16834 with the old one to avoid oscillations.
16836 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
16839 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
16841 (assign_by_spills): Return a flag of reload assignment failure.
16842 Do not process the reload assignment failures. Do not spill other
16843 reload pseudos if they has the same reg class. Update n if
16845 (lra_assign): Add a return arg. Set up from the result of
16846 assign_by_spills call.
16847 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
16848 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
16849 usage_insns if it is not NULL.
16850 (spill_hard_reg_in_range): New function.
16851 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
16852 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
16853 function prototypes.
16854 (lra_assign): Change prototype.
16855 * lra.c (lra): Add code to deal with fails by splitting hard reg
16858 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
16860 * config/riscv/riscv.opt (mrelax): New option.
16861 * config/riscv/riscv.c (riscv_file_start): Emit ".option
16862 "norelax" when riscv_mrelax is disabled.
16863 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
16865 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16868 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
16869 reassociation for int modes.
16871 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
16873 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
16874 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
16876 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
16877 * config/aarch64/aarch64-sve.md
16878 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
16879 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
16880 (*extend<mode><Vwide>2): Rename to...
16881 (aarch64_sve_extend<mode><Vwide>2): ...this.
16882 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
16883 renaming the old pattern to...
16884 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
16886 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
16887 define_expand, renaming the old pattern to...
16888 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
16889 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
16890 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
16891 account when deciding which SVE instruction the optab should use.
16892 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
16894 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
16896 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
16897 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
16898 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
16899 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
16900 (tlsdesc_small_<mode>): Turn a define_expand and use
16901 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
16902 (tlsdesc_small_advsimd_<mode>): ...this.
16903 (tlsdesc_small_sve_<mode>): New pattern.
16905 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
16907 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
16908 (UNSPEC_UMUL_HIGHPART): New constants.
16909 (MUL_HIGHPART): New int iteraor.
16910 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
16911 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
16913 (*<su>mul<mode>3_highpart): New define_insn.
16915 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
16918 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
16921 2018-03-13 Martin Liska <mliska@suse.cz>
16924 * (sem_item_optimizer::sem_item_optimizer): Initialize new
16926 (sem_item_optimizer::~sem_item_optimizer): Release it.
16927 (sem_item_optimizer::merge_classes): Register variable aliases.
16928 (sem_item_optimizer::fixup_pt_set): New function.
16929 (sem_item_optimizer::fixup_points_to_sets): Likewise.
16930 * ipa-icf.h: Declare new variables and functions.
16932 2018-03-13 Jakub Jelinek <jakub@redhat.com>
16934 PR middle-end/84834
16935 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
16936 integer_pow2p@2 and test integer_pow2p in condition.
16937 (A < 0 ? C : 0): Similarly for @1.
16939 PR middle-end/84831
16940 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
16941 characters starting at p contain '\0' character, don't look beyond
16945 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
16946 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
16949 * reg-stack.c (change_stack): Change update_end var from int to
16950 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
16951 also call set_block_for_insn on the newly added insns and rescan.
16954 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
16955 on the last operand.
16958 * tree.c (stabilize_reference_1): Return save_expr (e) for
16959 STATEMENT_LIST even if it doesn't have side-effects.
16961 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
16963 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
16965 2018-03-12 Renlin Li <renlin.li@arm.com>
16967 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
16968 aarch64_output_scalar_simd_mov_immediate.
16970 2018-03-12 Martin Sebor <msebor@redhat.com>
16972 PR tree-optimization/83456
16973 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
16974 for perfectly overlapping calls to memcpy.
16975 (gimple_fold_builtin_memory_chk): Same.
16976 (gimple_fold_builtin_strcpy): Handle no-warning.
16977 (gimple_fold_builtin_stxcpy_chk): Same.
16978 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
16980 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
16982 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
16983 parameter. Use it for SFmode.
16984 (rs6000_function_arg_advance_1): Adjust.
16985 (rs6000_function_arg): Adjust.
16986 (rs6000_gimplify_va_arg): Pass false for that new parameter.
16988 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
16990 PR rtl-optimization/84169
16991 PR rtl-optimization/84780
16992 * combine.c (can_combine_p): Check for a 2-insn combination whether
16993 the destination register is used between the two insns, too.
16995 2018-03-12 Richard Biener <rguenther@suse.de>
16997 PR tree-optimization/84803
16998 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
16999 for refs DR analysis didn't process.
17001 2018-03-12 Richard Biener <rguenther@suse.de>
17003 PR tree-optimization/84777
17004 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
17005 force-vectorize loops ignore whether we are optimizing for size.
17007 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
17009 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
17010 (TARGET_MD_ASM_ADJUST): Define.
17012 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
17013 Kito Cheng <kito.cheng@gmail.com>
17014 Chung-Ju Wu <jasonwucj@gmail.com>
17016 * config/nds32/nds32.c (nds32_compute_stack_frame,
17017 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
17018 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
17019 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
17020 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
17021 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
17022 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
17023 * config/nds32/nds32.md (prologue, epilogue): Use macro
17024 NDS32_V3PUSH_AVAILABLE_P to do checking.
17026 2018-03-11 Jakub Jelinek <jakub@redhat.com>
17029 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
17030 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
17031 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
17032 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
17033 addition of most attributes on !orig_type_die or the attribute not
17034 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
17036 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17037 Chung-Ju Wu <jasonwucj@gmail.com>
17039 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
17040 __NDS32_VH__ macro.
17041 * config/nds32/nds32.opt (mvh): New option.
17043 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17044 Chung-Ju Wu <jasonwucj@gmail.com>
17046 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
17048 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
17049 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
17052 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17053 Chung-Ju Wu <jasonwucj@gmail.com>
17055 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
17057 * config/nds32/nds32-multiple.md (strlensi): New pattern.
17058 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
17060 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
17061 Kito Cheng <kito.cheng@gmail.com>
17062 Chung-Ju Wu <jasonwucj@gmail.com>
17064 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
17065 UNSPEC_FFMISM and UNSPEC_FLMISM.
17066 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
17067 for ffb, ffmism and flmism.
17068 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
17069 (unspec_ffmism): Ditto.
17070 (unspec_flmism): Ditto.
17071 (nds32_expand_builtin_impl): Check if string extension is available.
17072 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
17073 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
17075 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
17078 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
17081 * lra-assigns.c (assign_by_spills): Return a flag of reload
17082 assignment failure. Do not process the reload assignment
17083 failures. Do not spill other reload pseudos if they has the same
17085 (lra_assign): Add a return arg. Set up from the result of
17086 assign_by_spills call.
17087 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17088 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17089 usage_insns if it is not NULL.
17090 (spill_hard_reg_in_range): New function.
17091 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17092 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17093 function prototypes.
17094 (lra_assign): Change prototype.
17095 * lra.c (lra): Add code to deal with fails by splitting hard reg
17098 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
17101 * config/i386/i386.opt: Replace Enforcment with Enforcement.
17103 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
17106 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
17107 (dw_val_node): Add val_symbolic_view.
17108 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
17109 (symview_upper_bound): New.
17110 (new_line_info_table): Initialize symviews_since_reset.
17111 (dwarf2out_source_line): Count symviews_since_reset and set
17112 symview_upper_bound.
17113 (dw_val_equal_p): Handle symview.
17114 (add_AT_symview): New.
17115 (print_dw_val): Handle symview.
17116 (attr_checksum, attr_checksum_ordered): Likewise.
17117 (same_dw_val_p, size_of_die): Likewise.
17118 (value_format, output_die): Likewise.
17119 (add_high_low_attributes): Use add_AT_symview for entry_view.
17120 (dwarf2out_finish): Reset symview_upper_bound, clear
17123 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
17126 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
17127 Add strict argument and use it.
17128 (rs6000_split_multireg_move): Update for new strict argument.
17129 (mem_operand_gpr): Disallow all non-offsettable addresses.
17130 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
17132 2018-03-09 Jakub Jelinek <jakub@redhat.com>
17135 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
17136 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
17137 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
17140 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
17141 decl, use remap_type if we want to use the type.
17143 2018-03-09 Martin Sebor <msebor@redhat.com>
17145 PR tree-optimization/84526
17146 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
17148 (builtin_access::generic_overlap): Be prepared to handle non-array
17151 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
17153 PR rtl-optimization/84682
17154 * lra-constraints.c (process_address_1): Check is_address flag
17155 for address constraints.
17156 (process_alt_operands): Likewise.
17157 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
17158 preprocess_constraints.
17159 * recog.h (preprocess_constraints): Add oploc parameter.
17161 * recog.c (preprocess_constraints): Test address_operand for
17162 CT_ADDRESS constraints.
17164 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
17167 * lra-assigns.c (assign_by_spills): Return a flag of reload
17168 assignment failure. Do not process the reload assignment
17169 failures. Do not spill other reload pseudos if they has the same
17171 (lra_assign): Add a return arg. Set up from the result of
17172 assign_by_spills call.
17173 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17174 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17175 usage_insns if it is not NULL.
17176 (spill_hard_reg_in_range): New function.
17177 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17178 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17179 function prototypes.
17180 (lra_assign): Change prototype.
17181 * lra.c (lra): Add code to deal with fails by splitting hard reg
17184 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17187 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
17188 Accept complain bool parameter. Only emit errors if it is true.
17189 (arm_parse_cpu_option_name): Likewise.
17190 (arm_target_thumb_only): Adjust callers of the above.
17191 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
17192 prototype to take a default true bool parameter.
17193 (arm_parse_arch_option_name): Likewise.
17195 2018-03-09 David Malcolm <dmalcolm@redhat.com>
17196 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
17200 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
17201 * configure: Regenerate.
17202 * configure.ac ("linker --version-script option"): New.
17203 ("linker soname option"): New.
17205 2018-03-09 Richard Biener <rguenther@suse.de>
17207 PR tree-optimization/84775
17208 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
17209 immediate uses of predicate stmts and mark them modified.
17212 PR tree-optimization/84178
17213 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17215 (version_loop_for_if_conversion): Delay update_ssa call.
17216 (tree_if_conversion): Delay update_ssa until after predicate
17219 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
17222 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
17223 when the function accesses prior frames.
17225 2018-03-08 Jakub Jelinek <jakub@redhat.com>
17228 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
17229 gen_llsym, otherwise call maybe_gen_llsym.
17231 PR inline-asm/84742
17232 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
17233 has ',' character inside of it.
17235 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17238 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
17239 as clobbering CC_REGNUM.
17241 2018-03-08 Richard Biener <rguenther@suse.de>
17243 PR middle-end/84552
17244 * tree-scalar-evolution.c: Include tree-into-ssa.h.
17245 (follow_copies_to_constant): Do not follow SSA names registered
17248 2018-03-08 Richard Biener <rguenther@suse.de>
17250 PR tree-optimization/84178
17251 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17253 (version_loop_for_if_conversion): Delay update_ssa call.
17254 (tree_if_conversion): Delay update_ssa until after predicate
17257 2018-03-08 David Malcolm <dmalcolm@redhat.com>
17259 PR tree-optimization/84178
17260 * tree-if-conv.c (release_bb_predicate): Remove the
17261 the assertion that the stmts have NULL use_ops.
17262 Discard the statements, asserting that they haven't
17263 yet been added to a BB.
17265 2018-03-08 Richard Biener <rguenther@suse.de>
17267 PR tree-optimization/84746
17268 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
17269 (phi_translate): Pass in destination ANTIC_OUT set.
17270 (phi_translate_1): Likewise. For a simplified result lookup
17271 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
17272 (phi_translate_set): Adjust.
17273 (do_pre_regular_insertion): Likewise.
17274 (do_pre_partial_partial_insertion): Likewise.
17276 2018-03-08 Martin Liska <mliska@suse.cz>
17278 PR gcov-profile/84735
17279 * doc/gcov.texi: Document usage of profile files.
17280 * gcov-io.h: Document changes in the format.
17282 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
17286 * dwarf2out.c (struct dw_line_info_table): Update comments for
17288 (FORCE_RESET_NEXT_VIEW): New.
17289 (FORCE_RESETTING_VIEW_P): New.
17290 (RESETTING_VIEW_P): Check for -1 too.
17291 (ZERO_VIEW_P): Likewise.
17292 (new_line_info_table): Force-reset next view.
17293 (dwarf2out_begin_function): Likewise.
17294 (dwarf2out_source_line): Simplify zero_view_p initialization.
17295 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
17296 view directly. Omit view when omitting .loc at line 0.
17298 2018-03-08 Jakub Jelinek <jakub@redhat.com>
17300 PR tree-optimization/84740
17301 * tree-switch-conversion.c (process_switch): Call build_constructors
17302 only if info.phi_count is non-zero.
17304 PR tree-optimization/84739
17305 * tree-tailcall.c (find_tail_calls): Check call arguments against
17306 DECL_ARGUMENTS (current_function_decl) rather than
17307 DECL_ARGUMENTS (func) when checking for tail recursion.
17309 2018-03-07 Jakub Jelinek <jakub@redhat.com>
17311 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
17312 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
17313 Volker Reichelt's entry and add entries for people that perform
17314 GCC fuzzy testing and report numerous bugs.
17316 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
17319 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
17320 readonly data in sdata, if that is disabled.
17321 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
17322 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
17323 -mreadonly-in-sdata option.
17325 2018-03-07 Martin Sebor <msebor@redhat.com>
17327 PR tree-optimization/84468
17328 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
17329 basic block when looking for nul assignment.
17331 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
17334 * except.h (output_function_exception_table): Adjust prototype.
17335 * except.c (output_function_exception_table): Remove FNNAME parameter
17336 and add SECTION parameter. Ouput one part of the table at a time.
17337 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
17338 the first part of the exception table and emit unwind directives.
17339 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
17340 (i386_pe_seh_cold_init): Likewise.
17341 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
17342 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
17343 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
17344 (ix86_output_call_insn): Emit a nop in one more case for SEH.
17345 * config/i386/winnt.c: Include except.h.
17346 (struct seh_frame_state): Add reg_offset, after_prologue and
17347 in_cold_section fields.
17348 (i386_pe_seh_end_prologue): Set seh->after_prologue.
17349 (i386_pe_seh_cold_init): New function.
17350 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
17351 to seh->in_cold_section.
17352 (seh_emit_push): Record the offset of the push.
17353 (seh_emit_save): Record the offet of the save.
17354 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
17355 Test seh->after_prologue to disregard the epilogue.
17356 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
17357 (i386_pe_end_cold_function): New function.
17359 2018-03-07 Jakub Jelinek <jakub@redhat.com>
17362 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
17363 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
17366 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
17368 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
17369 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
17371 PR middle-end/84723
17372 * multiple_target.c: Include tree-inline.h and intl.h.
17373 (expand_target_clones): Diagnose and fail if node->definition and
17374 !tree_versionable_function_p (node->decl).
17376 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
17378 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
17380 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
17381 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
17382 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
17384 2018-03-06 Jakub Jelinek <jakub@redhat.com>
17387 * combine.c (try_combine): Use reg_or_subregno instead of handling
17388 just paradoxical SUBREGs and REGs.
17390 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
17392 * config/arc/arc.c (arc_finalize_pic): Remove function.
17393 (arc_must_save_register): We use single base PIC register, remove
17394 checks to save/restore the PIC register.
17395 (arc_expand_prologue): Likewise.
17396 * config/arc/arc-protos.h (arc_set_default_type_attributes):
17398 (arc_verify_short): Likewise.
17399 (arc_attr_type): Likewise.
17400 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
17401 (walk_stores): Likewise.
17402 (arc_address_cost): Make it static.
17403 (arc_verify_short): Likewise.
17404 (branch_dest): Likewise.
17405 (arc_attr_type): Likewise.
17406 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
17407 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
17408 (arc_final_prescan_insn): Remove inserting the nops due to
17409 hardware hazards. It is done in reorg step.
17410 (insn_length_variant_t): Remove.
17411 (insn_length_parameters_t): Likewise.
17412 (arc_insn_length_parameters): Likewise.
17413 (arc_get_insn_variants): Likewise.
17414 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
17416 2018-03-06 Jakub Jelinek <jakub@redhat.com>
17418 PR inline-asm/84683
17419 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
17422 PR tree-optimization/84687
17423 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
17425 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
17427 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17429 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
17430 Rename to ppc_speculation_barrier.
17431 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
17432 __builtin_ppc_speculation_barrier.
17434 2018-03-05 Jakub Jelinek <jakub@redhat.com>
17437 * combine.c (combine_simplify_rtx): Don't try to simplify if
17438 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
17441 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
17443 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
17444 to 32 bytes when compiling for POWER9.
17446 2018-03-05 Jakub Jelinek <jakub@redhat.com>
17449 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
17450 regparm >= 3 with no arg reg available also for calls with
17451 flag_force_indirect_call. Pass decl to ix86_function_regparm.
17454 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
17456 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
17458 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
17461 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
17463 2018-03-05 Richard Biener <rguenther@suse.de>
17465 PR tree-optimization/84486
17466 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
17467 When inserting a __builtin_assume_aligned call set the LHS
17468 SSA name alignment info accordingly.
17470 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
17472 PR tree-optimization/84114
17473 * config/aarch64/aarch64.c (aarch64_reassociation_width)
17474 Avoid reassociation of FLOAT_MODE addition.
17476 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
17478 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
17479 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
17480 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
17481 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
17482 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
17483 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
17484 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
17486 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
17487 __builtin_ia32_wbinvd): New builtins.
17488 (SPECIAL_ARGS2): New.
17489 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
17490 (SPECIAL_ARGS2): New.
17491 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
17492 (ix86_valid_target_attribute_inner_p): Ditto.
17493 (ix86_init_mmx_sse_builtins): Add special_args2.
17494 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
17495 TARGET_WBNOINVD_P): New.
17496 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
17497 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
17498 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
17499 * config/i386/immintrin.h (_wbinvd): New intrinsic.
17500 * config/i386/pconfigintrin.h: New file.
17501 * config/i386/wbnoinvdintrin.h: Ditto.
17502 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
17504 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
17506 2018-03-05 Richard Biener <rguenther@suse.de>
17508 PR tree-optimization/84670
17509 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
17511 (BB_VISITED_WITH_VISITED_SUCCS): New define.
17512 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
17513 (compute_antic_aux): Only assert the number of values in ANTIC_IN
17514 doesn't grow if all successors (recursively) were visited at least
17517 2018-03-05 Richard Biener <rguenther@suse.de>
17519 PR tree-optimization/84650
17520 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
17521 if executed in the loop pipeline.
17523 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
17525 * doc/configfiles.texi (Configuration Files): Move info about
17526 conditionalizing $target-protos.h to...
17527 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
17528 differs from $target-protos.h.
17530 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
17531 Chung-Ju Wu <jasonwucj@gmail.com>
17533 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
17534 * config/nds32/nds32-multiple.md (setmemsi): Define.
17535 * config/nds32/nds32-memory-manipulation.c
17536 (nds32_gen_dup_4_byte_to_word_value): New.
17537 (emit_setmem_word_loop): New.
17538 (emit_setmem_byte_loop): New.
17539 (nds32_expand_setmem_loop): New.
17540 (nds32_expand_setmem_loop_v3m): New.
17541 (nds32_expand_setmem_unroll): New.
17542 (nds32_expand_setmem): New.
17544 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
17545 Chung-Ju Wu <jasonwucj@gmail.com>
17547 * config/nds32/nds32-memory-manipulation.c
17548 (nds32_emit_load_store): New.
17549 (nds32_emit_post_inc_load_store): New.
17550 (nds32_emit_mem_move): New.
17551 (nds32_emit_mem_move_block): New.
17552 (nds32_expand_movmemsi_loop_unknown_size): New.
17553 (nds32_expand_movmemsi_loop_known_size): New.
17554 (nds32_expand_movmemsi_loop): New.
17555 (nds32_expand_movmemsi_unroll): New.
17556 (nds32_expand_movmemqi): Rename ...
17557 (nds32_expand_movmemsi): ... to this.
17558 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
17559 (movmemsi): ... to this.
17560 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
17561 (nds32_expand_movmemsi): ... to this.
17563 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
17564 Monk Chiang <sh.chiang04@gmail.com>
17565 Chung-Ju Wu <jasonwucj@gmail.com>
17567 * config/nds32/nds32-protos.h
17568 (nds32_expand_load_multiple): New arguments.
17569 (nds32_expand_store_multiple): Ditto.
17570 (nds32_valid_multiple_load_store): Rename ...
17571 (nds32_valid_multiple_load_store_p): ... to this.
17572 * config/nds32/nds32-memory-manipulation.c
17573 (nds32_expand_load_multiple): Refine implementation.
17574 (nds32_expand_store_multiple): Ditto.
17575 * config/nds32/nds32-multiple.md
17576 (load_multiple): Update nds32_expand_load_multiple interface.
17577 (store_multiple): Update nds32_expand_store_multiple interface.
17578 * config/nds32/nds32-predicates.c
17579 (nds32_valid_multiple_load_store): Rename ...
17580 (nds32_valid_multiple_load_store_p): ... to this and refine
17582 * config/nds32/predicates.md
17583 (nds32_load_multiple_and_update_address_operation): New predicate.
17584 (nds32_store_multiple_and_update_address_operation): New predicate.
17586 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
17587 Chung-Ju Wu <jasonwucj@gmail.com>
17589 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
17590 (combo): New attribute.
17591 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
17593 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
17595 * config/nds32/nds32.opt: Change -mcmodel= default value.
17597 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
17598 Monk Chiang <sh.chiang04@gmail.com>
17599 Chung-Ju Wu <jasonwucj@gmail.com>
17601 * config/nds32/constants.md (unspec_element): New enum.
17602 * config/nds32/constraints.md (Umw): New constraint.
17603 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
17604 * config/nds32/nds32-intrinsic.md: Likewise.
17605 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
17606 (nds32_valid_smw_lwm_base_p): New.
17607 (nds32_output_smw_single_word): New.
17608 (nds32_output_lmw_single_word): New.
17609 (nds32_expand_unaligned_load): New.
17610 (nds32_expand_unaligned_store): New.
17611 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
17612 (nds32_output_smw_single_word): Declare.
17613 (nds32_output_lmw_single_word): Declare.
17614 (nds32_expand_unaligned_load): Declare.
17615 (nds32_expand_unaligned_store): Declare.
17616 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
17617 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
17618 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
17619 NDS32_BUILTIN_UASTORE_DW.
17620 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
17623 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
17624 Kito Cheng <kito.cheng@gmail.com>
17625 Chung-Ju Wu <jasonwucj@gmail.com>
17627 * config/nds32/nds32-intrinsic.c
17628 (nds32_expand_builtin_null_ftype_reg): Delete.
17629 (nds32_expand_builtin_reg_ftype_imm): Ditto.
17630 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
17631 (nds32_read_argument): New.
17632 (nds32_legitimize_target): Ditto.
17633 (nds32_legitimize_argument): Ditto.
17634 (nds32_check_constant_argument): Ditto.
17635 (nds32_expand_unop_builtin): Ditto.
17636 (nds32_expand_unopimm_builtin): Ditto.
17637 (nds32_expand_binop_builtin): Ditto.
17638 (nds32_builtin_decl_impl): Ditto.
17639 (builtin_description): Ditto.
17640 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
17641 (nds32_init_builtins_impl): Ditto.
17642 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
17643 (nds32_builtin_decl): New.
17644 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
17645 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
17647 2018-03-02 Jeff Law <law@redhat.com>
17649 * reorg.c (stop_search_p): Handle DEBUG_INSN.
17650 (redundant_insn, fill_simple_delay_slots): Likewise.
17651 (fill_slots_from_thread): Likewise.
17652 * resource.c (mark_referenced_resources): Likewise.
17653 (mark_set_resources, find_dead_or_set_registers): Likewise.
17655 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17657 * substring-locations.h (format_warning_va): Formatting fix for
17658 ATTRIBUTE_GCC_DIAG.
17659 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
17661 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
17662 * substring-locations.c: Include intl.h.
17663 (format_warning_va): Turned into small wrapper around
17664 format_warning_n_va, renamed to ...
17665 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
17666 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
17668 (format_warning_at_substring_n): New function.
17669 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
17670 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
17671 format_warning_at_substring with just a shorter name instead of
17672 const function pointer.
17673 (fmtwarn_n): New function.
17674 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
17675 appropriate, get rid of all the fmtstr temporaries, move conditionals
17676 with G_() wrapped string literals directly into fmtwarn arguments,
17677 cast dir.len to (int), formatting fixes.
17679 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
17681 * doc/invoke.texi: Remove "Cilk Plus" references.
17683 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17684 Richard Biener <rguenther@suse.de>
17687 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
17688 for error or warning attributes if CALL_FROM_THUNK_P is set.
17691 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17694 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
17695 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
17698 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
17699 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
17701 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
17702 instead of -1U in last predictors element's probability member.
17704 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
17707 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
17708 arguments if they are comparable.
17710 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
17712 PR tree-optimization/84634
17713 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
17714 masks and masked_loop_p with a single loop_masks, making sure it's
17715 null for bb vectorization.
17717 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
17719 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
17720 (vect_analyze_data_ref_access): Use loop->safe_len rather than
17721 loop->force_vectorize to check whether there is no alias.
17723 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17726 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
17728 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
17730 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
17731 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
17732 instead of a loop around prev_real_insn.
17733 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
17736 PR inline-asm/84625
17737 * config/i386/i386.c (ix86_print_operand): Use conditional
17738 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
17741 2018-03-02 Richard Biener <rguenther@suse.de>
17743 PR tree-optimization/84427
17744 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
17745 (bitmap_set_subtract_values): Rewrite to handle multiple
17748 (prune_clobbered_mems): Likewise.
17749 (phi_translate): Take edge instead of pred/phiblock.
17750 (phi_translate_1): Likewise.
17751 (phi_translate_set): Likewise. Insert all translated
17752 exprs for a value into the set, keeping possibly multiple
17753 expressions per value.
17754 (compute_antic_aux): Adjust for phi_translate changes.
17755 When intersecting union the expressions and prune those
17756 not in the final value set, keeping possibly multiple
17757 expressions per value. Do not use value-insertion
17758 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
17759 all expressions. Add verification that the value-sets
17760 only shrink during iteration.
17761 (compute_partial_antic_aux): Adjust for the phi_translate changes.
17762 (do_pre_regular_insertion): Likewise.
17763 (do_pre_partial_partial_insertion): Likewise.
17765 2018-03-02 Richard Biener <rguenther@suse.de>
17768 * config/darwin.c (saved_debug_info_level): New static global.
17769 (darwin_asm_lto_start): Disable debug info generation for LTO out.
17770 (darwin_asm_lto_end): Restore debug info generation settings.
17772 2018-03-01 Martin Liska <mliska@suse.cz>
17775 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
17776 volatile arguments.
17778 2018-03-01 Richard Biener <rguenther@suse.de>
17781 * dwarf2out.c (gen_variable_die): Properly handle late VLA
17782 type annotation with LTO when debug was disabled at compile-time.
17784 2018-03-01 Matthew Fortune <mfortune@gmail.com>
17786 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
17788 (mips_final_postscan_insn): Likewise.
17790 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
17792 PR rtl-optimization/84528
17793 * alias.c (init_alias_target): Add commentary.
17794 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
17795 a unique base value if the frame pointer is not eliminated
17796 to the stack pointer.
17798 2018-03-01 Tom de Vries <tom@codesourcery.com>
17800 PR rtl-optimization/83327
17801 * lra-int.h (hard_regs_spilled_into): Declare.
17802 * lra.c (hard_regs_spilled_into): Define.
17803 (init_reg_info): Init hard_regs_spilled_into.
17804 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
17805 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
17806 (process_bb_lives): Handle hard_regs_spilled_into.
17807 (lra_create_live_ranges_1): Before doing liveness propagation, clear
17808 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
17810 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
17812 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
17813 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
17814 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
17815 * config/rs6000/aix72.h: New file.
17817 2018-02-28 Jakub Jelinek <jakub@redhat.com>
17819 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
17820 instead of warning_at with conditional singular and plural messages
17824 * stor-layout.c (update_alignment_for_field): For
17825 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
17826 && !DECL_PACKED (field), do the alignment update, just use
17827 only desired_align instead of MAX (type_align, desired_align)
17829 (place_field): Don't do known_align < desired_align handling
17830 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
17831 is non-NULL, instead do it after rli->prev_field handling and
17832 only if not within a bitfield word. For DECL_PACKED (field)
17833 use type_align of BITS_PER_UNIT.
17835 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
17837 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
17838 superfluous parentheses and trailing spaces.
17840 2018-02-28 Richard Biener <rguenther@suse.de>
17842 PR tree-optimization/84584
17843 * graphite-scop-detection.c (scop_detection::add_scop): Discard
17844 SCoPs with fake exit edge.
17846 2018-02-28 Martin Liska <mliska@suse.cz>
17849 * timevar.c (timer::print): Fix format to properly print 100%
17852 2018-02-28 Richard Biener <rguenther@suse.de>
17854 PR middle-end/84607
17855 * genmatch.c (capture_info::walk_match): Do not mark
17856 captured expressions without operands as expr_p given
17857 they act more like predicates and should be subject to
17858 "lost tail" side-effect preserving.
17860 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
17862 PR rtl-optimization/81611
17863 * auto-inc-dec.c (attempt_change): Move dead note from
17864 mem_insn if it's the next use of regno
17865 (find_address): Take address use of reg holding
17866 non-incremented value. Add parm to limit search to the named
17868 (merge_in_block): Attempt to use a mem insn that is the next
17869 use of the original regno.
17871 2018-02-27 Martin Sebor <msebor@redhat.com>
17874 * doc/invoke.texi (-Wmissing-attributes): New option.
17875 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
17877 2018-02-27 Martin Sebor <msebor@redhat.com>
17879 PR translation/84207
17880 * diagnostic-core.h (warning_n, error_n, inform_n): Change
17881 n argument to unsigned HOST_WIDE_INT.
17882 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
17883 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
17884 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
17885 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
17887 2018-02-27 Richard Biener <rguenther@suse.de>
17889 PR tree-optimization/84512
17890 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
17891 Do not use the estimate returned from record_stmt_cost for
17892 the scalar iteration cost but sum properly using add_stmt_cost.
17894 2018-02-27 Richard Biener <rguenther@suse.de>
17896 PR tree-optimization/84466
17897 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
17898 Adjust last change to less strictly validate use operands.
17900 2018-02-27 Martin Liska <mliska@suse.cz>
17902 PR gcov-profile/84548
17903 * gcov.c (process_file): Allow partial overlap and consider it
17904 also as group functions.
17905 (output_lines): Properly calculate range of lines for a group.
17907 2018-02-27 Martin Liska <mliska@suse.cz>
17909 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
17910 'ggc' suffixes. Change first column width.
17911 (timer::print): Fix formatting of the column.
17913 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
17915 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
17916 preserve inline entry blocks for the sake of debug inline
17917 entry point markers alone.
17918 (remove_unused_locals): Suggest in comments a better place to
17919 force the preservation of inline entry blocks that are
17920 otherwise unused, but do not preserve them.
17922 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
17924 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
17926 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
17929 * config/i386/constraints.md (Bs): Replace
17930 ix86_indirect_branch_register with
17931 TARGET_INDIRECT_BRANCH_REGISTER.
17933 * config/i386/i386.md (indirect_jump): Likewise.
17934 (tablejump): Likewise.
17935 (*sibcall_memory): Likewise.
17936 (*sibcall_value_memory): Likewise.
17937 Peepholes of indirect call and jump via memory: Likewise.
17938 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
17939 (*sibcall_value_GOT_32): Likewise.
17940 * config/i386/predicates.md (indirect_branch_operand): Likewise.
17941 (GOT_memory_operand): Likewise.
17942 (call_insn_operand): Likewise.
17943 (sibcall_insn_operand): Likewise.
17944 (GOT32_symbol_operand): Likewise.
17945 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
17947 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
17949 PR rtl-optimization/83496
17950 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
17951 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
17952 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
17953 redundant insn, if any.
17954 (relax_delay_slots): Likewise.
17955 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
17957 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
17959 PR tree-optimization/83965
17960 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
17961 that grouped statements are part of a reduction chain. Return
17962 true if the statement is not marked as a reduction itself but
17963 is part of a group.
17964 (vect_recog_dot_prod_pattern): Don't check whether the statement
17965 is part of a group here.
17966 (vect_recog_sad_pattern): Likewise.
17967 (vect_recog_widen_sum_pattern): Likewise.
17969 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
17972 * final.c (rest_of_clean_state): Also look for calls inside sequences.
17974 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
17977 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
17979 (ix86_output_indirect_function_return): New prototype.
17980 (ix86_split_simple_return_pop_internal): Likewise.
17981 * config/i386/i386.c (indirect_return_via_cx): New.
17982 (indirect_return_via_cx_bnd): Likewise.
17983 (indirect_thunk_name): Handle return va CX_REG.
17984 (output_indirect_thunk_function): Create alias for
17985 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
17986 (ix86_output_indirect_jmp): Remove the bool argument.
17987 (ix86_output_indirect_function_return): New function.
17988 (ix86_split_simple_return_pop_internal): Likewise.
17989 * config/i386/i386.md (*indirect_jump): Don't pass false
17990 to ix86_output_indirect_jmp.
17991 (*tablejump_1): Likewise.
17992 (simple_return_pop_internal): Change it to define_insn_and_split.
17993 Call ix86_split_simple_return_pop_internal to split it for
17994 -mfunction-return=.
17995 (simple_return_indirect_internal): Call
17996 ix86_output_indirect_function_return instead of
17997 ix86_output_indirect_jmp.
17999 2018-02-26 Jakub Jelinek <jakub@redhat.com>
18002 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
18003 memset and value initialization afterwards.
18005 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
18007 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
18009 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
18012 * common/config/aarch64/aarch64-common.c
18013 (aarch_option_optimization_table[]): Switch
18014 off fomit-frame-pointer
18016 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
18017 Chung-Ju Wu <jasonwucj@gmail.com>
18019 * config/nds32/nds32-multiple.md (load_multiple): Disallow
18021 (store_multiple): Ditto.
18023 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
18025 * config.gcc: Add --with-cpu support for nds32 target.
18026 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
18027 * config/nds32/nds32.opt: Add -mcpu= option.
18029 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
18031 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
18032 isel=yes): Warn for these deprecated options.
18034 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
18036 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
18037 ISA_2_5_MASKS_EMBEDDED.
18039 2018-02-23 Jakub Jelinek <jakub@redhat.com>
18041 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
18042 p->max as pointers rather than using iterative_hash_expr.
18044 2018-02-23 Carl Love <cel@us.ibm.com>
18046 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
18047 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
18049 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
18050 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
18051 P8V_BUILTIN_VEC_VUNSIGNED2.
18053 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
18056 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
18057 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
18059 * lra-constraints.c (curr_insn_transform): Set up
18060 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
18062 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
18063 * lra-eliminations.c (spill_pseudos): Ditto.
18064 (process_insn_for_elimination): Ditto.
18065 * lra-lives.c (reg_early_clobber_p): Use the new macros.
18066 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
18067 LRA_NON_CLOBBERED_ALT.
18069 2018-02-22 Martin Sebor <msebor@redhat.com>
18071 PR tree-optimization/84480
18072 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
18073 to maybe_diag_stxncpy_trunc. Call it.
18074 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
18075 from gimple_fold_builtin_strcpy. Print inlining stack.
18076 (handle_builtin_stxncpy): Print inlining stack.
18077 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
18079 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
18082 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
18083 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
18084 and -fcheck-pointer-bounds are used together.
18085 (indirect_thunk_prefix): New enum.
18086 (indirect_thunk_need_prefix): New function.
18087 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
18088 "_nt" instead of "_bnd" for NOTRACK prefix.
18089 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
18090 (output_indirect_thunk_function): Likewise.
18092 (ix86_code_end): Update output_indirect_thunk_function calls.
18093 (ix86_output_indirect_branch_via_reg): Replace
18094 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
18095 (ix86_output_indirect_branch_via_push): Likewise.
18096 (ix86_output_function_return): Likewise.
18097 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
18098 incompatible with -fcf-protection=branch and
18099 -fcheck-pointer-bounds.
18101 2018-02-22 Steve Ellcey <sellcey@cavium.com>
18104 * config/aarch64/aarch64.c (aarch64_print_address_internal):
18105 Change gcc_assert call to output_operand_lossage.
18107 2018-02-22 Steve Ellcey <sellcey@cavium.com>
18109 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
18111 2018-02-22 DJ Delorie <dj@redhat.com>
18112 Sebastian Perta <sebastian.perta@renesas.com>
18113 Oleg Endo <olegendo@gcc.gnu.org>
18115 * config/rx/rx.c (rx_rtx_costs): New function.
18116 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
18118 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
18120 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
18122 2018-02-22 Martin Liska <mliska@suse.cz>
18125 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
18126 Add "native" as a possible value.
18128 2018-02-22 Martin Liska <mliska@suse.cz>
18131 * config/i386/i386.c (ix86_option_override_internal):
18132 Add "native" as a possible value for -march and -mtune.
18134 2018-02-22 Jakub Jelinek <jakub@redhat.com>
18137 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
18138 to all type variants.
18140 PR tree-optimization/84503
18141 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
18142 width as info->bitpos + info->bitsize - start.
18143 (merged_store_group::merge_overlapping): Simplify width computation.
18144 (check_no_overlap): New function.
18145 (imm_store_chain_info::try_coalesce_bswap): Compute expected
18146 start + width and last_order of the group, fail if check_no_overlap
18148 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
18149 to group if check_no_overlap fails.
18151 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
18153 * config/rs6000/altivec.md: Delete contraint arguments to
18154 define_expand, define_split, and define_peephole2, and in
18155 define_insn_and_split if always unused.
18156 * config/rs6000/darwin.md: Ditto.
18157 * config/rs6000/dfp.md: Ditto.
18158 * config/rs6000/rs6000.md: Ditto.
18159 * config/rs6000/sync.md: Ditto.
18160 * config/rs6000/vector.md: Ditto.
18161 * config/rs6000/vsx.md: Ditto.
18163 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
18165 * config/rs6000/altivec.md: Write output control strings as braced
18166 blocks instead of double-quoted strings.
18167 * config/rs6000/darwin.md: Ditto.
18168 * config/rs6000/rs6000.md: Ditto.
18169 * config/rs6000/vector.md: Ditto.
18170 * config/rs6000/vsx.md: Ditto.
18172 2018-02-21 Jason Merrill <jason@redhat.com>
18174 PR c++/84314 - ICE with templates and fastcall attribute.
18175 * attribs.c (build_type_attribute_qual_variant): Remove assert.
18177 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
18179 * ipa-cp.c (determine_versionability): Fix comment typos.
18181 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
18184 * ipa-cp.c (determine_versionability): Do not version functions caling
18187 2018-02-21 Martin Liska <mliska@suse.cz>
18190 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
18191 Add "native" as a possible value.
18192 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
18193 the macro when native cpu detection is available.
18195 2018-02-21 Martin Liska <mliska@suse.cz>
18198 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
18199 Add "native" as a possible value.
18200 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
18201 when native cpu detection is available.
18203 2018-02-21 Jakub Jelinek <jakub@redhat.com>
18204 Martin Sebor <msebor@redhat.com>
18206 PR tree-optimization/84478
18207 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
18209 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
18210 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
18211 support which is conservatively correct, for 2 only stay conservative
18212 for maxlen. Formatting and comment capitalization fixes. Add STRICT
18213 argument to the 2 argument get_range_strlen, adjust 6 arg
18214 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
18216 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
18217 (gimple_fold_builtin_strlen): Pass true as last argument to
18220 2018-02-20 Martin Sebor <msebor@redhat.com>
18222 PR middle-end/84095
18223 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
18224 (builtin_memref::set_base_and_offset): Same. Handle inner references.
18225 (builtin_memref::builtin_memref): Factor out parts into
18226 set_base_and_offset and call it.
18228 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
18230 PR middle-end/84406
18231 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
18232 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
18233 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
18234 search at the associated MODE_INT.
18236 2018-02-20 Jeff Law <law@redhat.com>
18238 PR middle-end/82123
18239 PR tree-optimization/81592
18240 PR middle-end/79257
18241 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
18242 for range data rather than using global data.
18243 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
18244 range data rather than using global data.
18245 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
18246 pass it to children as needed.
18247 (struct directive::fmtresult): Similarly.
18248 (struct directive::set_width): Similarly.
18249 (struct directive::set_precision): Similarly.
18250 (format_integer, format_directive, parse_directive): Similarly.
18251 (format_none): Accept unnamed vr_values parameter.
18252 (format_percent, format_floating, format_character): Similarly.
18253 (format_string, format_plain): Similarly.
18254 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
18255 the EVRP range analyzer for range data rather than using global data.
18256 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
18257 gimple-ssa-evrp-analyze.h
18258 (class sprintf_dom_walker): Add after_dom_children member function.
18259 Add evrp_range_analyzer member.
18260 (sprintf_dom_walker::before_dom_children): Call into the EVRP
18261 range analyzer as needed.
18262 (sprintf_dom_walker::after_dom_children): New member function.
18263 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
18265 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
18266 (evrp_range_analyzer::pop_to_marker): Likewise.
18268 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
18270 PR tree-optimization/84419
18271 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
18272 with the required type if its current type is compatible but
18275 2018-02-20 Jakub Jelinek <jakub@redhat.com>
18277 PR middle-end/82004
18278 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
18279 after vectorization.
18281 2018-02-20 Martin Liska <mliska@suse.cz>
18284 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
18285 possible values if we don't have a hint.
18287 2018-02-20 Martin Liska <mliska@suse.cz>
18291 * final.c (shorten_branches): Build align_tab array with one
18293 * opts.c (finish_options): Add alignment option limit check.
18294 (MAX_CODE_ALIGN): Likewise.
18295 (MAX_CODE_ALIGN_VALUE): Likewise.
18296 * doc/invoke.texi: Document maximum allowed option value for
18297 all -falign-* options.
18299 2018-02-19 Jakub Jelinek <jakub@redhat.com>
18302 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
18303 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
18304 * var-tracking.c (emit_note_insn_var_location): Remove all references
18305 to NOTE_INSN_CALL_ARG_LOCATION.
18306 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
18307 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
18308 Use copy_rtx_if_shared.
18309 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
18310 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
18311 (dwarf2out_var_location): Remove handling of
18312 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
18314 * final.c (final_scan_insn): Remove all references to
18315 NOTE_INSN_CALL_ARG_LOCATION.
18316 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
18317 before dumping final insns.
18318 * except.c (emit_note_eh_region_end): Remove all references to
18319 NOTE_INSN_CALL_ARG_LOCATION.
18320 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
18321 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
18322 * config/arc/arc.c (hwloop_optimize): Likewise.
18323 * config/arm/arm.c (create_fix_barrier): Likewise.
18324 * config/s390/s390.c (s390_chunkify_start): Likewise.
18325 * config/sh/sh.c (find_barrier): Likewise.
18326 * config/i386/i386.c (rest_of_insert_endbranch,
18327 ix86_seh_fixup_eh_fallthru): Likewise.
18328 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
18329 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
18330 * config/frv/frv.c (frv_function_prologue): Likewise.
18331 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
18333 (note_outside_basic_block_p): Remove all references to
18334 NOTE_INSN_CALL_ARG_LOCATION.
18335 * gengtype.c (adjust_field_rtx_def): Likewise.
18336 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
18338 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
18339 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
18342 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
18345 PR tree-optimization/84452
18346 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
18347 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
18350 2018-02-19 Martin Liska <mliska@suse.cz>
18353 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
18355 2018-02-19 Martin Liska <mliska@suse.cz>
18356 Richard Sandiford <richard.sandiford@linaro.org>
18358 PR tree-optimization/82491
18359 * gimple-fold.c (get_base_constructor): Make earlier bail out
18362 2018-02-19 Carl Love <cel@us.ibm.com>
18364 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
18365 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
18367 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
18368 P8V_BUILTIN_VEC_NEG.
18370 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
18372 * config/rl78/rl78.md (movdf): New define expand.
18374 2018-02-19 Martin Liska <mliska@suse.cz>
18377 * doc/invoke.texi: Fix typo.
18378 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
18380 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
18382 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
18383 handle rs6000_single_float and rs6000_double_float specially for
18386 2018-02-16 Jeff Law <law@redhat.com>
18388 * config/rx/rx.c (add_pop_cfi_notes): New function.;
18389 (pop_regs): Use it.
18391 2018-02-16 Jakub Jelinek <jakub@redhat.com>
18394 * ipa-inline.c (inline_small_functions): Fix a typo.
18396 2018-02-16 Nathan Sidwell <nathan@acm.org>
18398 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
18400 2018-02-16 Carl Love <cel@us.ibm.com>
18402 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
18403 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
18404 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
18405 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
18406 expansion to P8V_BUILTIN_VEC_FLOAT2.
18408 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
18410 PR rtl-optimization/70023
18411 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
18412 src_regno into account.
18414 2018-02-16 Carl Love <cel@us.ibm.com>
18416 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
18417 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
18418 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
18419 * config/rs6000/rs6000.c: Remove case statements for
18420 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
18421 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
18422 and P9V_BUILTIN_VEC_VINSERT4B.
18423 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
18424 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
18425 * config/rs6000/vsx.md:
18426 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
18429 2018-02-16 Carl Love <cel@us.ibm.com>
18431 * config/rs6000/altivec.h: Add builtin names vec_extract4b
18433 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
18435 * config/rs6000/rs6000-c.c: Add the definitions for
18436 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
18437 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
18438 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
18439 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
18440 definition for insert4b and define insn *insert3b_internal.
18441 * doc/extend.texi: Add documentation for vec_extract4b.
18443 2018-02-16 Nathan Sidwell <nathan@acm.org>
18445 * doc/extend.texi (Backwards Compatibility): Mention friend
18446 injection. Note for-scope is deprecated.
18447 * doc/invoke.texi (-ffriend-injection): Deprecate.
18449 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
18451 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
18452 that moved to I2, also allow destinations that are a paradoxical
18453 subreg (instead of a normal reg).
18455 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
18458 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
18461 2018-02-16 Richard Biener <rguenther@suse.de>
18463 PR tree-optimization/84037
18464 PR tree-optimization/84016
18466 * config/i386/i386.c (ix86_builtin_vectorization_cost):
18467 Adjust vec_construct for the fact we need additional higher latency
18468 128bit inserts for AVX256 and AVX512 vector builds.
18469 (ix86_add_stmt_cost): Scale vector construction cost for
18472 2018-02-16 Richard Biener <rguenther@suse.de>
18474 PR tree-optimization/84417
18475 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
18476 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
18477 (non_rewritable_lvalue_p): Likewise, use poly-ints.
18479 2018-02-16 Martin Liska <mliska@suse.cz>
18482 * internal-fn.def (ASAN_CHECK): Set proper flags.
18483 (ASAN_MARK): Likewise.
18485 2018-02-16 Julia Koval <julia.koval@intel.com>
18487 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
18488 from PTA_CANNONLAKE.
18490 2018-02-16 Jakub Jelinek <jakub@redhat.com>
18493 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
18494 Use ++iter rather than iter++ for std::list iterators.
18495 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
18496 defer deleting them until all nodes in the forest are processed. Do
18497 free even leaf nodes. Change to_process into auto_vec.
18500 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
18501 * vec.h (vec_default_construct): Use memset instead of placement new
18502 if BROKEN_VALUE_INITIALIZATION is defined.
18503 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
18504 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
18507 PR rtl-optimization/83723
18508 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
18509 * lra.c (lra_substitute_pseudo): Likewise. If true, use
18510 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
18512 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
18514 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
18516 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
18518 PR rtl-optimization/81443
18519 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
18520 from inner REGs to paradoxical SUBREGs.
18522 2018-02-16 Richard Biener <rguenther@suse.de>
18524 PR tree-optimization/84399
18525 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
18526 For operands we can analyze at their definition make sure we can
18527 analyze them at each use as well.
18529 2018-02-16 Richard Biener <rguenther@suse.de>
18531 PR tree-optimization/84190
18532 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
18533 volatile accesses if the decl isn't volatile.
18535 2018-02-15 Jason Merrill <jason@redhat.com>
18537 PR c++/84314 - ICE with templates and fastcall attribute.
18538 * attribs.c (build_type_attribute_qual_variant): Don't clobber
18539 TYPE_CANONICAL on an existing type.
18541 2018-02-15 Jakub Jelinek <jakub@redhat.com>
18543 PR tree-optimization/84383
18544 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
18545 dstoff nor call operand_equal_p if dstbase is NULL.
18547 PR tree-optimization/84334
18548 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
18549 also a CONSTANT_CLASS_P, punt.
18551 2018-02-14 Jim Wilson <jimw@sifive.com>
18553 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
18554 first SMALL_OPERAND check. New local min_second_step. Move assert
18555 to where locals are set. Add TARGET_RVC support.
18556 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
18558 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
18560 * doc/invoke.texi: Correct -Wformat-overflow code sample.
18562 2018-02-14 Martin Sebor <msebor@redhat.com>
18564 PR tree-optimization/83698
18565 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
18566 arrays constrain the offset range to their bounds.
18567 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
18568 (builtin_access::overlap): Avoid setting the size of overlap if it's
18570 (maybe_diag_overlap): Also consider arrays when deciding what values
18571 of offsets to include in diagnostics.
18573 2018-02-14 Martin Sebor <msebor@redhat.com>
18576 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
18577 that correspond to the kind of a declaration.
18579 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
18582 * config/pa/pa.md: Load address of PIC label using the linkage table
18583 if the label is nonlocal.
18585 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
18587 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
18588 warning message if user requests -maltivec=be.
18589 * doc/invoke.texi: Document deprecation of -maltivec=be.
18591 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
18594 * config/rs6000/rs6000-c.c: Update definitions for
18595 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
18596 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
18598 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
18601 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
18602 add _get_ssp intrinsics. Remove argument from
18603 __builtin_ia32_rdssp[d|q].
18604 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
18605 * config/i386/i386-builtin.def: Remove argument from
18606 __builtin_ia32_rdssp[d|q].
18607 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
18608 ix86_expand_special_args_builtin for _rdssp[d|q].
18609 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
18610 Clear register before usage.
18611 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
18612 Add documentation for new _get_ssp and _inc_ssp intrinsics.
18614 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
18616 PR tree-optimization/84357
18617 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
18618 operand 1 of an ARRAY_REF too.
18620 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
18623 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
18624 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
18626 (set_of_reg): New struct.
18627 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
18628 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
18629 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
18631 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
18632 Split into bitclr, bitset, bitinvert patterns if appropriate.
18633 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
18634 use rx_fuse_in_memory_bitop.
18635 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
18636 to named insn, correct maximum insn length.
18638 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
18641 * machmode.def: Define a complex mode for PARTIAL_INT.
18642 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
18644 * doc/rtl.texi: Document CSPImode.
18645 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
18647 (msp430_hard_regno_nregs_with_padding): Likewise.
18649 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
18652 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
18654 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
18656 PR rtl-optimization/84169
18657 * combine.c (try_combine): New variable split_i2i3. Set it to true if
18658 we generated a parallel as new i3 and we split that to new i2 and i3
18659 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
18660 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
18661 those to i2, not i1. Partially rewrite this scan code.
18663 2018-02-13 Jakub Jelinek <jakub@redhat.com>
18666 * stor-layout.c (place_field): For variable length fields, adjust
18667 offset_align afterwards not just based on the field's alignment,
18668 but also on the size.
18670 PR middle-end/84309
18671 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
18672 of exps and logs in the use_exp2 case.
18674 2018-02-13 Jeff Law <law@redhat.com>
18676 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
18677 entry for "vector".
18679 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
18682 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
18686 * common.opt (gas-loc-support, gas-locview-support): New.
18687 (ginline-points, ginternal-reset-location-views): New.
18688 * doc/invoke.texi: Document them. Use @itemx where intended.
18689 (gvariable-location-views): Adjust.
18690 * target.def (reset_location_view): New.
18691 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
18692 (TARGET_RESET_LOCATION_VIEW): New.
18693 * doc/tm.texi: Rebuilt.
18694 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
18695 (dwarf2out_default_as_locview_support): New.
18696 (output_asm_line_debug_info): Use option variables.
18697 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
18698 (output_loc_list): Likewise.
18699 (add_high_low_attributes): Check option variables.
18700 Don't output entry view attribute in strict mode.
18701 (gen_inlined_subroutine_die): Check option variables.
18702 (dwarf2out_inline_entry): Likewise.
18703 (init_sections_and_labels): Likewise.
18704 (dwarf2out_early_finish): Likewise.
18705 (maybe_reset_location_view): New, from...
18706 (dwarf2out_var_location): ... here. Call it.
18707 * debug.h (dwarf2out_default_as_loc_support): Declare.
18708 (dwarf2out_default_as_locview_support): Declare.
18709 * hooks.c (hook_int_rtx_insn_0): New.
18710 * hooks.h (hook_int_rtx_insn_0): Declare.
18711 * toplev.c (process_options): Take -gas-loc-support and
18712 -gas-locview-support from dwarf2out. Enable
18713 -gvariable-location-views by default only with locview
18714 assembler support. Enable -ginternal-reset-location-views by
18715 default only if the target defines the corresponding hook.
18716 Enable -ginline-points by default if location views are
18717 enabled; force it disabled if statement frontiers are
18719 * tree-inline.c (expand_call_inline): Check option variables.
18720 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
18722 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
18724 PR tree-optimization/84321
18725 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
18726 handling. Also check whether the anti-range contains any values
18727 that satisfy the mask; switch to a VR_RANGE if not.
18729 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
18732 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
18734 2018-02-13 Martin Jambor <mjambor@suse.cz>
18737 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
18738 of call statements, also set location of a load to a temporary.
18740 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
18742 * config/rl78/rl78.c (add_vector_labels): New function.
18743 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
18744 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
18745 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
18746 which checks that no arguments are passed.
18747 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
18748 * doc/extend.texi: Documentation for the new attribute.
18750 2018-02-13 Andreas Schwab <schwab@suse.de>
18752 * config/riscv/linux.h (CPP_SPEC): Define.
18754 2018-02-13 Jakub Jelinek <jakub@redhat.com>
18757 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
18758 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
18759 OPTION_MASK_ISA_AES as first argument to def_builtin_const
18760 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
18761 instead of OPTION_MASK_ISA_PCLMUL as first argument to
18762 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
18763 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
18764 temporarily for AES and PCLMUL builtins.
18766 PR tree-optimization/84339
18767 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
18768 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
18771 PR middle-end/84309
18772 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
18773 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
18774 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
18776 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
18778 * omp-simd-clone.h: New file.
18779 * omp-simd-clone.c: Include omp-simd-clone.h.
18780 (expand_simd_clones): No longer static.
18781 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
18782 cgraph.h and omp-simd-clone.h.
18783 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
18784 (vect_recog_widen_shift_pattern): Formatting fix.
18785 (vect_pattern_recog_1): Don't check optab for calls.
18788 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
18789 operands[2] into a REG before using gen_lowpart on it.
18791 2018-02-12 Jeff Law <law@redhat.com>
18794 * config/sh/sh.c (find_barrier): Consider a sibling call
18797 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
18798 successfully back substituting a reg.
18800 2018-02-12 Richard Biener <rguenther@suse.de>
18802 PR tree-optimization/84037
18803 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
18804 parameter, move visited init to caller.
18805 (vect_slp_analyze_operations): Separate cost from validity
18806 check, initialize visited once for all instances.
18807 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
18809 * tree-vect-stmts.c (vect_model_simple_cost): Make early
18811 (vect_model_promotion_demotion_cost): Likewise.
18812 (vectorizable_bswap): Guard cost modeling with !slp_node
18813 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
18815 (vectorizable_call): Likewise.
18816 (vectorizable_conversion): Likewise.
18817 (vectorizable_assignment): Likewise.
18818 (vectorizable_shift): Likewise.
18819 (vectorizable_operation): Likewise.
18820 (vectorizable_store): Likewise.
18821 (vectorizable_load): Likewise.
18822 (vectorizable_condition): Likewise.
18823 (vectorizable_comparison): Likewise.
18825 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
18828 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
18829 (ASAN_MARK): Fix fnspec to account for return value, change pointer
18830 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
18832 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
18834 PR middle-end/83665
18835 * params.def (inline-min-speedup): Increase from 8 to 15.
18836 (max-inline-insns-auto): Decrease from 40 to 30.
18837 * ipa-split.c (consider_split): Add some buffer for function to
18838 be considered inlining candidate.
18839 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
18842 2018-02-12 Richard Biener <rguenther@suse.de>
18844 PR tree-optimization/84037
18845 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
18846 matched stmts if we cannot swap the non-matched ones.
18848 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
18850 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
18851 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
18852 _mm_maskz_scalef_round_ss): New intrinsics.
18853 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
18854 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
18855 __builtin_ia32_scalefss_round): Remove.
18856 (__builtin_ia32_scalefsd_mask_round,
18857 __builtin_ia32_scalefss_mask_round): New intrinsics.
18858 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
18859 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
18860 ((match_operand:VF_128 2 "<round_nimm_predicate>"
18861 "<round_constraint>")): Changed to ...
18862 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
18863 "<round_scalar_constraint>")): ... this.
18864 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
18865 %0, %1, %2<round_op3>}"): Changed to ...
18866 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
18867 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
18868 %2<round_scalar_mask_op3>}"): ... this.
18869 * config/i386/subst.md (round_scalar_nimm_predicate): New.
18871 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
18873 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
18874 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
18875 (_mm_maskz_sqrt_round_ss): New intrinsics.
18876 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
18877 (__builtin_ia32_sqrtsd_mask_round)
18878 (__builtin_ia32_sqrtss_mask_round): New builtins.
18879 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
18880 (__builtin_ia32_sqrtss_round): Remove.
18881 (__builtin_ia32_sqrtsd_mask_round)
18882 (__builtin_ia32_sqrtss_mask_round): New builtins.
18883 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
18884 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
18885 ((match_operand:VF_128 1 "vector_operand"
18886 "xBm,<round_constraint>")): Changed to ...
18887 ((match_operand:VF_128 1 "vector_operand"
18888 "xBm,<round_scalar_constraint>")): ... this.
18889 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
18890 %0, %2, %<iptr>1<round_op3>}): Changed to ...
18891 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
18892 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
18893 %<iptr>1<round_scalar_mask_op3>}): ... this.
18894 ((set_attr "prefix" "<round_prefix>")): Changed to ...
18895 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
18897 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
18900 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
18901 Cast vec_cmpeq result to correct type.
18902 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
18903 Cast vec_cmpgt result to correct type.
18905 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
18907 * final.c (final_scan_insn_1): Renamed from...
18908 (final_scan_insn): ... this. New wrapper, to recover
18909 seen from the outermost call in recursive ones.
18910 * config/sparc/sparc.c (output_return): Drop seen from call.
18911 (output_sibcall): Likewise.
18912 * config/visium/visium.c (output_branch): Likewise.
18914 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
18916 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
18919 2018-02-10 Alan Modra <amodra@gmail.com>
18922 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
18923 Specify LR as an input.
18925 2018-02-10 Jakub Jelinek <jakub@redhat.com>
18928 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
18929 remove_member_access_dummy_vars): New functions.
18930 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
18931 lower_omp_1, execute_lower_omp): Use them.
18933 PR rtl-optimization/84308
18934 * shrink-wrap.c (spread_components): Release todo vector.
18936 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
18938 PR rtl-optimization/57193
18939 * ira-color.c (struct allocno_color_data): Add member
18940 conflict_allocno_hard_prefs.
18941 (update_conflict_allocno_hard_prefs): New.
18942 (bucket_allocno_compare_func): Add a preference based on
18943 conflict_allocno_hard_prefs.
18944 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
18945 (color_allocnos): Remove a dead code. Initiate
18946 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
18948 2018-02-09 Jakub Jelinek <jakub@redhat.com>
18951 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
18952 constraint from =wa to wa. Avoid a subreg on the output operand,
18953 instead use a pseudo and subreg it in a move.
18954 (p9_xxbrd_<mode>): Changed to ...
18955 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
18956 (p9_xxbrd_v2df): New expander.
18957 (p9_xxbrw_<mode>): Changed to ...
18958 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
18959 (p9_xxbrw_v4sf): New expander.
18961 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
18963 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
18965 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
18968 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
18969 multiply in 32-bit mode.
18970 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
18971 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
18974 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
18976 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
18977 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
18978 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
18979 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
18981 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
18984 * dwarf2out.c (is_trivial_indirect_ref): New function.
18985 (dwarf2out_late_global_decl): Do not generate a location
18986 attribute for variables that have a non-trivial DECL_VALUE_EXPR
18987 and that are not defined in the current unit.
18989 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
18991 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
18992 instead of a libcall for UNORDERED.
18994 2018-02-09 Tamar Christina <tamar.christina@arm.com>
18997 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
18998 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
19000 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19003 * config/s390/s390.c (s390_set_current_function): Invoke
19004 s390_indirect_branch_settings also if fndecl didn't change.
19006 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
19008 * config/rs6000/rs6000.md (blockage): Set length to zero.
19010 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
19012 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
19014 2018-02-09 Jakub Jelinek <jakub@redhat.com>
19017 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
19018 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
19022 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
19023 PARALLEL incoming that failed vt_get_decl_and_offset check.
19025 PR middle-end/84237
19026 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
19027 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
19029 (get_variable_section): For decls in named .bss* sections pass true as
19030 second argument to bss_initializer_p.
19032 2018-02-09 Marek Polacek <polacek@redhat.com>
19033 Jakub Jelinek <jakub@redhat.com>
19036 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
19037 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
19038 Sync some changes from cxx_fold_indirect_ref.
19040 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
19042 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
19044 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
19045 (BLOCK_INLINE_ENTRY_LABEL): New.
19046 (dwarf2out_var_location): Disregard inline entry markers.
19047 (inline_entry_data): New struct.
19048 (inline_entry_data_hasher): New hashtable type.
19049 (inline_entry_data_hasher::hash): New.
19050 (inline_entry_data_hasher::equal): New.
19051 (inline_entry_data_table): New variable.
19052 (add_high_low_attributes): Add DW_AT_entry_pc and
19053 DW_AT_GNU_entry_view attributes if a pending entry is found
19054 in inline_entry_data_table. Add old entry_pc attribute only
19055 if debug nonbinding markers are disabled.
19056 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
19057 markers are enabled.
19058 (block_within_block_p, dwarf2out_inline_entry): New.
19059 (dwarf2out_finish): Check that no entries remained in
19060 inline_entry_data_table.
19061 * final.c (reemit_insn_block_notes): Handle inline entry notes.
19062 (final_scan_insn, notice_source_line): Likewise.
19063 (rest_of_clean_state): Skip inline entry markers.
19064 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
19066 * gimple.c (gimple_build_debug_inline_entry): New.
19067 * gimple.h (enum gimple_debug_subcode): Add
19068 GIMPLE_DEBUG_INLINE_ENTRY.
19069 (gimple_build_debug_inline_entry): Declare.
19070 (gimple_debug_inline_entry_p): New.
19071 (gimple_debug_nonbind_marker_p): Adjust.
19072 * insn-notes.def (INLINE_ENTRY): New.
19073 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
19074 inline entry marker notes.
19075 (print_insn): Likewise.
19076 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
19077 (INSN_DEBUG_MARKER_KIND): Likewise.
19078 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
19079 * tree-inline.c (expand_call_inline): Build and insert
19080 debug_inline_entry stmt.
19081 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
19082 inline entry blocks early, if nonbind markers are enabled.
19083 (dump_scope_block): Dump fragment info.
19084 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
19085 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
19086 (gimple_build_debug_inline_entry): New.
19087 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
19088 Enable/disable inline entry points too.
19089 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
19090 (DEBUG_INSN): Describe inline entry markers.
19092 * common.opt (gvariable-location-views): New.
19093 (gvariable-location-views=incompat5): New.
19094 * config.in: Rebuilt.
19095 * configure: Rebuilt.
19096 * configure.ac: Test assembler for view support.
19097 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
19098 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
19099 * dwarf2out.c (var_loc_view): New typedef.
19100 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
19101 (dwarf2out_locviews_in_attribute): New.
19102 (dwarf2out_locviews_in_loclist): New.
19103 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
19104 (enum dw_line_info_opcode): Add LI_adv_address.
19105 (struct dw_line_info_table): Add view.
19106 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
19107 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
19108 (zero_view_p): New variable.
19109 (ZERO_VIEW_P): New macro.
19110 (output_asm_line_debug_info): New.
19111 (struct var_loc_node): Add view.
19112 (add_AT_view_list, AT_loc_list): New.
19113 (add_var_loc_to_decl): Add view param. Test it against last.
19114 (new_loc_list): Add view params. Record them.
19115 (AT_loc_list_ptr): Handle loc and view lists.
19116 (view_list_to_loc_list_val_node): New.
19117 (print_dw_val): Handle dw_val_class_view_list.
19118 (size_of_die): Likewise.
19119 (value_format): Likewise.
19120 (loc_list_has_views): New.
19121 (gen_llsym): Set vl_symbol too.
19122 (maybe_gen_llsym, skip_loc_list_entry): New.
19123 (dwarf2out_maybe_output_loclist_view_pair): New.
19124 (output_loc_list): Output view list or entries too.
19125 (output_view_list_offset): New.
19126 (output_die): Handle dw_val_class_view_list.
19127 (output_dwarf_version): New.
19128 (output_compilation_unit_header): Use it.
19129 (output_skeleton_debug_sections): Likewise.
19130 (output_rnglists, output_line_info): Likewise.
19131 (output_pubnames, output_aranges): Update version comments.
19132 (output_one_line_info_table): Output view numbers in asm comments.
19133 (dw_loc_list): Determine current endview, pass it to new_loc_list.
19134 Call maybe_gen_llsym.
19135 (loc_list_from_tree_1): Adjust.
19136 (add_AT_location_description): Create view list attribute if
19137 needed, check it's absent otherwise.
19138 (convert_cfa_to_fb_loc_list): Adjust.
19139 (maybe_emit_file): Call output_asm_line_debug_info for test.
19140 (dwarf2out_var_location): Reset views as needed. Precompute
19141 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
19142 attribute. Set view.
19143 (new_line_info_table): Reset next view.
19144 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
19145 (dwarf2out_source_line): Likewise. Output view resets and labels to
19146 the assembler, or select appropriate line info opcodes.
19147 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
19148 (optimize_string_length): Catch it. Adjust.
19149 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
19150 dw_val_class_view_list, and remove it if no longer needed.
19151 (hash_loc_list): Hash view numbers.
19152 (loc_list_hasher::equal): Compare them.
19153 (optimize_location_lists): Check whether a view list symbol is
19154 needed, and whether the locview attribute is present, and
19155 whether they match. Remove the locview attribute if no longer
19157 (index_location_lists): Call skip_loc_list_entry for test.
19158 (dwarf2out_finish): Call output_asm_line_debug_info for test.
19159 Use output_dwarf_version.
19160 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
19161 (struct dw_val_node): Add val_view_list.
19162 * final.c (SEEN_NEXT_VIEW): New.
19163 (set_next_view_needed): New.
19164 (clear_next_view_needed): New.
19165 (maybe_output_next_view): New.
19166 (final_start_function): Rename to...
19167 (final_start_function_1): ... this. Take pointer to FIRST,
19168 add SEEN parameter. Emit param bindings in the initial view.
19169 (final_start_function): Reintroduce SEEN-less interface.
19170 (final): Rename to...
19171 (final_1): ... this. Take SEEN parameter. Output final pending
19172 next view at the end.
19173 (final): Reintroduce seen-less interface.
19174 (final_scan_insn): Output pending next view before switching
19175 sections or ending a block. Mark the next view as needed when
19176 outputting variable locations. Notify debug backend of section
19177 changes, and of location view changes.
19178 (rest_of_handle_final): Adjust.
19179 * toplev.c (process_options): Autodetect value for debug variable
19180 location views option. Warn on incompat5 without -gdwarf-5.
19181 * doc/invoke.texi (gvariable-location-views): New.
19182 (gvariable-location-views=incompat5): New.
19183 (gno-variable-location-views): New.
19185 2018-02-08 David Malcolm <dmalcolm@redhat.com>
19187 PR tree-optimization/84136
19188 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
19189 that the result of find_edge is non-NULL.
19191 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
19194 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
19195 storing integer register in SImode. Fix cost of 256 and 512
19196 byte aligned SSE register store.
19198 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
19200 * config/i386/i386.c (ix86_multiplication_cost): Fix
19201 multiplication cost for TARGET_AVX512DQ.
19203 2018-02-08 Marek Polacek <polacek@redhat.com>
19205 PR tree-optimization/84238
19206 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
19209 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19211 PR tree-optimization/84265
19212 * tree-vect-stmts.c (vectorizable_store): Don't treat
19213 VMAT_CONTIGUOUS accesses as grouped.
19214 (vectorizable_load): Likewise.
19216 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19218 PR tree-optimization/81635
19219 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
19220 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
19221 (test_round_for_mask): New functions.
19222 (wide_int_cc_tests): Call test_round_for_mask.
19223 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
19224 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
19225 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
19226 range returned by get_range_info.
19228 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
19231 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
19232 * symtab.c: Include builtins.h
19233 (symtab_node::output_to_lto_symbol_table_p): Move here
19234 from lto-streamer-out.c:output_symbol_p.
19235 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
19236 (output_symbol_p): Move all logic to symtab.c
19237 (produce_symtab): Update.
19239 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19241 * config/s390/s390-opts.h (enum indirect_branch): Define.
19242 * config/s390/s390-protos.h (s390_return_addr_from_memory)
19243 (s390_indirect_branch_via_thunk)
19244 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
19245 (enum s390_indirect_branch_type): Define.
19246 * config/s390/s390.c (struct s390_frame_layout, struct
19247 machine_function): Remove.
19248 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
19249 (indirect_branch_table_label_no, indirect_branch_table_name):
19251 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
19252 (enum s390_indirect_branch_option): Define.
19253 (s390_return_addr_from_memory): New function.
19254 (s390_handle_string_attribute): New function.
19255 (s390_attribute_table): Add new attribute handler.
19256 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
19257 (s390_indirect_branch_via_thunk): New function.
19258 (s390_indirect_branch_via_inline_thunk): New function.
19259 (s390_function_ok_for_sibcall): When jumping via thunk disallow
19260 sibling call optimization for non z10 compiles.
19261 (s390_emit_call): Force indirect branch target to be a single
19262 register. Add r1 clobber for non-z10 compiles.
19263 (s390_emit_epilogue): Emit return jump via return_use expander.
19264 (s390_reorg): Handle JUMP_INSNs as execute targets.
19265 (s390_option_override_internal): Perform validity checks for the
19266 new command line options.
19267 (s390_indirect_branch_attrvalue): New function.
19268 (s390_indirect_branch_settings): New function.
19269 (s390_set_current_function): Invoke s390_indirect_branch_settings.
19270 (s390_output_indirect_thunk_function): New function.
19271 (s390_code_end): Implement target hook.
19272 (s390_case_values_threshold): Implement target hook.
19273 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
19275 * config/s390/s390.h (struct s390_frame_layout)
19276 (struct machine_function): Move here from s390.c.
19277 (TARGET_INDIRECT_BRANCH_NOBP_RET)
19278 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
19279 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
19280 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
19281 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
19282 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
19283 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
19284 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
19285 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
19286 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
19287 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
19288 (mnemonic attribute): Add values which aren't recognized
19290 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
19291 pattern for branch conversion. Fix mnemonic attribute.
19292 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
19293 indirect branch via thunk if requested.
19294 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
19295 ("*indirect_jump"): Disable for branch conversion using out of
19297 ("indirect_jump_via_thunk<mode>_z10")
19298 ("indirect_jump_via_thunk<mode>")
19299 ("indirect_jump_via_inlinethunk<mode>_z10")
19300 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
19301 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
19302 ("casesi_jump_via_inlinethunk<mode>_z10")
19303 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
19304 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
19305 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
19306 ("*indirect2_jump"): Disable for branch conversion.
19307 ("casesi_jump"): Turn into expander and expand patterns for branch
19309 ("return_use"): New expander.
19310 ("*return"): Emit return via thunk and rename it to ...
19311 ("*return<mode>"): ... this one.
19312 * config/s390/s390.opt: Add new options and and enum for the
19315 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19317 * lra-constraints.c (match_reload): Unconditionally use
19318 gen_lowpart_SUBREG, rather than selecting between that
19319 and equivalent gen_rtx_SUBREG code.
19321 2018-02-08 Richard Biener <rguenther@suse.de>
19323 PR tree-optimization/84233
19324 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
19325 changed flag instead of boguously re-using phi_inserted.
19327 2018-02-08 Martin Jambor <mjambor@suse.cz>
19329 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
19330 static local variables.
19332 2018-02-08 Richard Biener <rguenther@suse.de>
19334 PR tree-optimization/84278
19335 * tree-vect-stmts.c (vectorizable_store): When looking for
19336 smaller vector types to perform grouped strided loads/stores
19337 make sure the mode is supported by the target.
19338 (vectorizable_load): Likewise.
19340 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
19342 * config/aarch64/aarch64.c (aarch64_components_for_bb):
19343 Increase LDP/STP opportunities by adding adjacent callee-saves.
19345 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
19347 PR rtl-optimization/84068
19348 PR rtl-optimization/83459
19349 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
19351 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
19353 PR tree-optimization/84224
19354 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
19355 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
19356 non-zero arguments.
19358 2018-02-07 Iain Sandoe <iain@codesourcery.com>
19361 * config/rs6000/altivec.md (*restore_world): Remove LR use.
19362 * config/rs6000/predicates.md (restore_world_operation): Adjust op
19363 count, remove one USE.
19365 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
19367 * doc/install.texi (Configuration): Document the
19368 --with-long-double-format={ibm,ieee} PowerPC configuration
19372 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
19373 Convert from define_expand to be define_insn_and_split. Rework
19374 float/double/_Float128 conversions to QI/HI/SImode to work with
19375 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
19376 conversions to QI/HImode types did a store and then a load to
19377 truncate the value. For conversions to VSX registers, don't split
19378 the insn, instead emit the code directly. Use the code iterator
19379 any_fix to combine signed and unsigned conversions.
19380 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
19381 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
19382 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
19383 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
19384 (fix_<mode>di2_hw): Likewise.
19385 (fixuns_<mode>di2_hw): Likewise.
19386 (fix_<mode>si2_hw): Likewise.
19387 (fixuns_<mode>si2_hw): Likewise.
19388 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
19389 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
19390 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
19391 fix<uns>_trunc<SFDF:mode>si2_p8.
19392 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
19393 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
19394 (fix<uns>_<mode>_mem): Likewise.
19395 (fctiw<u>z_<mode>_mem): Likewise.
19396 (fix<uns>_<mode>_mem): Likewise.
19397 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
19398 the register allocator from doing a direct move to the GPRs to do
19399 a store, and instead use the ISA 3.0 store byte/half-word from
19400 vector register instruction. For IEEE 128-bit floating point,
19401 also optimize stores of 32-bit ints.
19402 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
19404 2018-02-07 Alan Hayward <alan.hayward@arm.com>
19406 * genextract.c (push_pathstr_operand): New function to support
19408 (walk_rtx): Call push_pathstr_operand.
19409 (print_path): Support [a-zA-Z].
19411 2018-02-07 Richard Biener <rguenther@suse.de>
19413 PR tree-optimization/84037
19414 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
19415 (cse_and_gimplify_to_preheader): Declare.
19416 (vect_get_place_in_interleaving_chain): Likewise.
19417 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19419 (_loop_vec_info::~_loop_vec_info): Delete it.
19420 (cse_and_gimplify_to_preheader): New function.
19421 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
19422 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
19423 (vectorizable_load): Likewise. For grouped stores always base
19424 the IV on the first element.
19425 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
19426 condition before gimplifying.
19428 2018-02-07 Jakub Jelinek <jakub@redhat.com>
19430 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
19431 *DIV_EXPR and *MOD_EXPR.
19433 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
19436 * config/i386/i386.c (ix86_option_override_internal): Mask out
19437 the CF_SET bit when checking -fcf-protection.
19439 2018-02-07 Tom de Vries <tom@codesourcery.com>
19442 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
19445 2018-02-07 Richard Biener <rguenther@suse.de>
19447 PR tree-optimization/84204
19448 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
19451 PR tree-optimization/84205
19452 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
19453 special-case isl_ast_op_zdiv_r.
19455 PR tree-optimization/84223
19456 * graphite-scop-detection.c (gather_bbs::before_dom_children):
19457 Only add conditions from within the region.
19458 (gather_bbs::after_dom_children): Adjust.
19460 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
19463 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
19464 * config/avr/avr.md: Only post-reload split REG-REG moves if
19465 either register is GENERAL_REG_P.
19467 2018-02-07 Jakub Jelinek <jakub@redhat.com>
19469 PR tree-optimization/84235
19470 * tree-ssa-scopedtables.c
19471 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
19472 if the subtraction is performed in floating point type where NaNs are
19473 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
19474 build 1. Formatting fix.
19476 2018-02-06 Jakub Jelinek <jakub@redhat.com>
19479 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
19480 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
19481 and skip it regardless of bb boundaries. Use CALL_P macro,
19482 don't test INSN_P (insn) together with CALL_P or JUMP_P check
19483 unnecessarily, formatting fix.
19485 2018-02-06 Michael Collison <michael.collison@arm.com>
19487 * config/arm/thumb2.md:
19488 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
19489 (*thumb_mov_notscc): Ditto.
19491 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
19494 * config/rs6000/rs6000.md (su code attribute): Use "u" for
19495 unsigned_fix, not "s".
19497 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19499 * configure.ac (gcc_fn_eh_frame_ro): New function.
19500 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
19501 correct .eh_frame permissions.
19502 * configure: Regenerate.
19504 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
19506 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
19507 irrelevant options.
19509 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19511 * config/rs6000/rs6000.c (rs6000_option_override_internal):
19512 Display warning message for -mno-speculate-indirect-jumps.
19514 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
19516 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
19518 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
19520 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
19522 PR tree-optimization/84225
19523 * tree-eh.c (find_trapping_overflow): Only call
19524 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
19526 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
19529 * config/i386/i386.c: Reimplement the check of possible options
19530 -mibt/-mshstk conbination. Change error messages.
19531 * doc/invoke.texi: Fix a typo: remove extra '='.
19533 2018-02-06 Marek Polacek <polacek@redhat.com>
19535 PR tree-optimization/84228
19536 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
19538 2018-02-06 Tamar Christina <tamar.christina@arm.com>
19541 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
19542 emitted arch directives.
19543 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
19544 __ARM_FEATURE_COPROC before changing architectures.
19546 2018-02-06 Richard Biener <rguenther@suse.de>
19548 * config/i386/i386.c (print_reg): Fix typo.
19549 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
19551 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
19553 * configure: Regenerate.
19555 2018-02-05 Martin Sebor <msebor@redhat.com>
19557 PR tree-optimization/83369
19558 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
19561 2018-02-05 Martin Liska <mliska@suse.cz>
19563 * doc/invoke.texi: Cherry-pick upstream r323995.
19565 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
19567 * ira.c (ira_init_register_move_cost): Adjust comment.
19569 2018-02-05 Martin Liska <mliska@suse.cz>
19571 PR gcov-profile/84137
19572 * doc/gcov.texi: Fix typo in documentation.
19574 2018-02-05 Martin Liska <mliska@suse.cz>
19576 PR gcov-profile/83879
19577 * doc/gcov.texi: Document necessity of --dynamic-list-data when
19578 using dlopen functionality.
19580 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
19582 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
19583 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
19584 _mm_maskz_range_ss, _mm_mask_range_round_ss,
19585 _mm_maskz_range_round_ss): New intrinsics.
19586 (__builtin_ia32_rangesd128_round)
19587 (__builtin_ia32_rangess128_round): Remove.
19588 (__builtin_ia32_rangesd128_mask_round,
19589 __builtin_ia32_rangess128_mask_round): New builtins.
19590 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
19591 __builtin_ia32_rangess128_round): Remove.
19592 (__builtin_ia32_rangesd128_mask_round,
19593 __builtin_ia32_rangess128_mask_round): New builtins.
19594 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
19595 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
19596 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
19597 "<round_saeonly_constraint>")): Changed to ...
19598 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
19599 "<round_saeonly_scalar_constraint>")): ... this.
19600 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
19601 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
19602 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
19603 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
19604 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
19606 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
19608 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
19610 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
19611 Remove all values except native, 8540 and 8548.
19613 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
19615 * config/i386/i386.c (ix86_output_function_return): Pass
19616 INVALID_REGNUM, instead of -1, as invalid register number to
19617 indirect_thunk_name and output_indirect_thunk.
19619 2018-02-02 Julia Koval <julia.koval@intel.com>
19621 * config.gcc: Add -march=icelake.
19622 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
19623 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
19624 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
19625 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
19626 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
19627 (processor_target_table): Add icelake.
19628 (ix86_option_override_internal): Handle new PTAs.
19629 (get_builtin_code_for_version): Handle icelake.
19630 (M_INTEL_COREI7_ICELAKE): New.
19631 (fold_builtin_cpu): Handle icelake.
19632 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
19633 * doc/invoke.texi: Add -march=icelake.
19635 2018-02-02 Julia Koval <julia.koval@intel.com>
19637 * config/i386/i386.c (ix86_option_override_internal): Change flags type
19638 to wide_int_bitmask.
19639 * wide-int-bitmask.h: New.
19641 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
19644 * config/i386/i386.md: Replace Pmode with word_mode in
19645 builtin_setjmp_setup and builtin_longjmp to support x32.
19647 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
19651 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
19653 (rs6000_supported_cpu_names): New static variable.
19654 (linux_cpu_translation_table): Likewise.
19655 (elf_platform) <cpu>: Define new static variable and use it.
19656 Translate kernel AT_PLATFORM name to canonical name if needed.
19657 Error if platform name is unknown.
19659 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
19662 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
19664 2018-02-01 Jeff Law <law@redhat.com>
19667 * config/i386/i386.c (release_scratch_register_on_entry): Add new
19668 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
19669 the scratch if RELEASE_VIA_POP is false.
19670 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
19671 If we have to save a temporary register, decrement SIZE appropriately.
19672 Pass new arguments to release_scratch_register_on_entry.
19673 (ix86_adjust_stack_and_probe): Likewise.
19674 (ix86_emit_probe_stack_range): Pass new arguments to
19675 release_scratch_register_on_entry.
19677 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
19679 PR rtl-optimization/84157
19680 * combine.c (change_zero_ext): Use REG_P predicate in
19681 front of HARD_REGISTER_P predicate.
19683 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
19685 * config/avr/avr.c (avr_option_override): Move disabling of
19686 -fdelete-null-pointer-checks to...
19687 * common/config/avr/avr-common.c (avr_option_optimization_table):
19690 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19692 PR tree-optimization/81635
19693 * tree-data-ref.c (split_constant_offset_1): For types that
19694 wrap on overflow, try to use range info to prove that wrapping
19697 2018-02-01 Renlin Li <renlin.li@arm.com>
19700 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
19701 TAILCALL_ADDR_REGS.
19702 (aarch64_register_move_cost): Likewise.
19703 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
19704 TAILCALL_ADDR_REGS.
19705 (REG_CLASS_NAMES): Likewise.
19706 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
19707 TAILCALL_ADDR_REGS. Remove IP registers.
19708 * config/aarch64/aarch64.md (Ucs): Update register constraint.
19710 2018-02-01 Richard Biener <rguenther@suse.de>
19712 * domwalk.h (dom_walker::dom_walker): Add additional constructor
19713 for specifying RPO order and allow NULL for that.
19714 * domwalk.c (dom_walker::dom_walker): Likewise.
19715 (dom_walker::walk): Handle NULL RPO order.
19716 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
19718 (rewrite_update_dom_walker): Likewise.
19719 (mark_def_dom_walker): Likewise.
19721 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19723 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
19724 (aarch64_maybe_expand_sve_subreg_move): Declare.
19725 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
19726 * config/aarch64/predicates.md (aarch64_any_register_operand): New
19728 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
19729 that are semantically a reverse operation.
19730 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
19731 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
19732 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
19734 (aarch64_can_change_mode_class): For big-endian, forbid changes
19735 between two SVE modes if they have different element sizes.
19737 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19739 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
19740 the TImode handling for big-endian targets.
19742 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19744 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
19745 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
19747 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
19748 Remove BSWAP handing for big-endian targets and use the form of
19749 LD1RQ appropariate for the mode.
19751 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19753 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
19754 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
19755 duplicated element.
19757 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19760 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
19761 check for operands that need to go through aarch64_sve_reload_be.
19763 2018-02-01 Jakub Jelinek <jakub@redhat.com>
19765 PR tree-optimization/81661
19766 PR tree-optimization/84117
19767 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
19768 * tree-eh.c: Include gimplify.h.
19769 (find_trapping_overflow, replace_trapping_overflow,
19770 rewrite_to_non_trapping_overflow): New functions.
19771 * tree-vect-loop.c: Include tree-eh.h.
19772 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
19773 * tree-data-ref.c: Include tree-eh.h.
19774 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
19776 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
19778 PR rtl-optimization/84123
19779 * combine.c (change_zero_ext): Check if hard register satisfies
19780 can_change_dest_mode before calling gen_lowpart_SUBREG.
19782 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
19785 * ira.c (ira_init_register_move_cost): Remove assert.
19787 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19789 PR rtl-optimization/84071
19790 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
19791 * doc/tm.texi: Regenerate.
19793 2018-01-31 Richard Biener <rguenther@suse.de>
19795 PR tree-optimization/84132
19796 * tree-data-ref.c (analyze_miv_subscript): Properly
19797 check whether evolution_function_is_affine_multivariate_p
19798 before calling gcd_of_steps_may_divide_p.
19800 2018-01-31 Julia Koval <julia.koval@intel.com>
19803 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
19804 * config/i386/i386.md (rdpid_rex64) New.
19805 (rdpid): Make 32bit only.
19807 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
19810 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
19811 an IDENTIFIER_NODE for FUNCTION_TYPE's.
19813 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19816 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
19818 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
19820 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19822 PR rtl-optimization/84071
19823 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
19824 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
19826 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
19828 * config/arc/arc.c (arc_handle_aux_attribute): New function.
19829 (arc_attribute_table): Add 'aux' attribute.
19830 (arc_in_small_data_p): Consider aux like variables.
19831 (arc_is_aux_reg_p): New function.
19832 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
19833 (arc_get_aux_arg): New function.
19834 (prepare_move_operands): Handle aux-register access.
19835 (arc_handle_aux_attribute): New function.
19836 * doc/extend.texi (ARC Variable attributes): Add subsection.
19838 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
19840 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
19841 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
19842 (arc_attribute_table): Add 'uncached' attribute.
19843 (arc_print_operand): Print '.di' flag for uncached memory
19845 (arc_in_small_data_p): Do not consider for small data the uncached
19847 (arc_is_uncached_mem_p): New function.
19848 * config/arc/predicates.md (compact_store_memory_operand): Check
19849 for uncached memory accesses.
19850 (nonvol_nonimm_operand): Likewise.
19851 * doc/extend.texi (ARC Type Attribute): New subsection.
19853 2018-01-31 Jakub Jelinek <jakub@redhat.com>
19856 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
19857 falign-loops=): Add Optimization flag.
19859 2018-01-30 Jeff Law <law@redhat.com>
19862 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
19863 INT_REGISTERS_SAVED. Check it prior to calling
19864 get_scratch_register_on_entry.
19865 (ix86_adjust_stack_and_probe): Similarly.
19866 (ix86_emit_probe_stack_range): Similarly.
19867 (ix86_expand_prologue): Corresponding changes.
19869 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19872 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
19873 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
19875 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
19878 * lra-constraints.c (curr_insn_transform): Process AND in the
19881 2018-01-30 Jakub Jelinek <jakub@redhat.com>
19883 PR rtl-optimization/83986
19884 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
19885 dependence against last_pending_memory_flush in addition to
19886 pending_jump_insns.
19888 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
19890 PR tree-optimization/81611
19891 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
19894 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19897 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
19900 2018-01-30 Richard Biener <rguenther@suse.de>
19901 Jakub Jelinek <jakub@redhat.com>
19903 PR tree-optimization/84111
19904 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
19905 inner loops added during recursion, as they don't have up-to-date
19908 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
19911 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
19912 (can_inline_edge_by_limits_p): ... here.
19913 (can_early_inline_edge_p, check_callers,
19914 update_caller_keys, update_callee_keys, recursive_inlining,
19915 add_new_edges_to_heap, speculation_useful_p,
19916 inline_small_functions,
19917 inline_small_functions, flatten_function,
19918 inline_to_all_callers_1): Update.
19920 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
19922 * profile-count.c (profile_count::combine_with_ipa_count): Handle
19925 2018-01-30 Richard Biener <rguenther@suse.de>
19927 PR tree-optimization/83008
19928 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
19929 invariant and constant vector uses in stmts when they need
19930 more than one stmt.
19932 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19935 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
19936 * configure: Regenerate.
19938 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
19940 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
19942 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
19943 Use gen_rtx_REG rather than gen_lowpart.
19945 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
19947 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
19948 rather than 0 when creating partial subregs.
19950 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
19952 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
19955 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
19958 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
19959 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
19960 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
19961 flags. This restores the settings used before the 2017-07-24.
19962 Turning off pre increment/decrement/modify allows IVOPTS to
19963 optimize DF/SF loops where the index is an int.
19965 2018-01-29 Richard Biener <rguenther@suse.de>
19966 Kelvin Nilsen <kelvin@gcc.gnu.org>
19969 * tree-vect-stmts.c (vectorizable_call): Don't call
19970 targetm.vectorize_builtin_md_vectorized_function if callee is
19973 2018-01-22 Carl Love <cel@us.ibm.com>
19975 * doc/extend.tex: Fix typo in second arg in
19976 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
19978 2018-01-29 Richard Biener <rguenther@suse.de>
19980 PR tree-optimization/84086
19981 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
19982 (flush_ssaname_freelist): When SSA names were released reset
19983 the SCEV hash table.
19985 2018-01-29 Richard Biener <rguenther@suse.de>
19987 PR tree-optimization/84057
19988 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
19989 removed paths when removing edges.
19991 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
19993 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
19994 -mfunction-return=@var{choice}.
19996 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
19998 PR diagnostic/84034
19999 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
20000 Handle CR like TAB.
20001 (layout::print_source_line): Likewise.
20002 (test_get_line_width_without_trailing_whitespace): Add test cases.
20004 2018-01-27 Jakub Jelinek <jakub@redhat.com>
20006 PR middle-end/84040
20007 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
20010 2018-01-26 Jim Wilson <jimw@sifive.com>
20012 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
20014 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
20017 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20019 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
20020 and CMP + SUB-immediate -> SUBS.
20022 2018-01-26 Martin Sebor <msebor@redhat.com>
20024 PR tree-optimization/83896
20025 * tree-ssa-strlen.c (get_string_len): Rename...
20026 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
20027 Avoid assuming length is constant.
20028 (handle_char_store): Use HOST_WIDE_INT for string length.
20030 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
20033 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
20034 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
20036 2018-01-26 Richard Biener <rguenther@suse.de>
20038 PR rtl-optimization/84003
20039 * dse.c (record_store): Only record redundant stores when
20040 the earlier store aliases at least all accesses the later one does.
20042 2018-01-26 Jakub Jelinek <jakub@redhat.com>
20044 PR rtl-optimization/83985
20045 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
20046 REG_CFA_RESTORE insns.
20047 (delete_unmarked_insns): Don't ignore separate shrink wrapping
20048 REG_CFA_RESTORE insns here.
20051 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
20052 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
20054 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20056 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
20057 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
20058 (arc_init): Likewise.
20059 (arc_override_options): Likewise.
20060 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
20062 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
20064 * config/arc/arc.h (TARGET_DBNZ): Define.
20065 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
20066 properly set the tune attribute.
20067 (dbnz): Use TARGET_DBNZ guard.
20068 * config/arc/arc.opt (mtune): Add core3 option.
20070 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20072 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
20073 recognize new pic like addresses.
20074 (arc_delegitimize_address): Clean up.
20076 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20078 * config/arc/arc-arches.def: Option mrf16 valid for all
20080 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
20081 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
20082 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
20083 * config/arc/arc-tables.opt: Regenerate.
20084 * config/arc/arc.c (arc_conditional_register_usage): Handle
20085 reduced register file case.
20086 (arc_file_start): Set must have build attributes.
20087 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
20088 mrf16 option value.
20089 * config/arc/arc.opt (mrf16): Add new option.
20090 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
20091 * config/arc/genmultilib.awk: Handle new mrf16 option.
20092 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
20093 * config/arc/t-multilib: Regenerate.
20094 * doc/invoke.texi (ARC Options): Document mrf16 option.
20096 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20098 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
20099 * config/arc/arc.c (arc_handle_secure_attribute): New function.
20100 (arc_attribute_table): Add 'secure_call' attribute.
20101 (arc_print_operand): Print secure call operand.
20102 (arc_function_ok_for_sibcall): Don't optimize tail calls when
20104 (arc_is_secure_call_p): New function. * config/arc/arc.md
20105 (call_i): Add support for sjli instruction.
20106 (call_value_i): Likewise.
20107 * config/arc/constraints.md (Csc): New constraint.
20109 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20110 John Eric Martin <John.Martin@emmicro-us.com>
20112 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
20113 * config/arc/arc.c (_arc_jli_section): New struct.
20114 (arc_jli_section): New type.
20115 (rc_jli_sections): New static variable.
20116 (arc_handle_jli_attribute): New function.
20117 (arc_attribute_table): Add jli_always and jli_fixed attribute.
20118 (arc_file_end): New function.
20119 (TARGET_ASM_FILE_END): Define.
20120 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
20121 (arc_add_jli_section): New function.
20122 (jli_call_scan): Likewise.
20123 (arc_reorg): Call jli_call_scan.
20124 (arc_output_addsi): Remove 'S' from printing asm operand.
20125 (arc_is_jli_call_p): New function.
20126 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
20128 (movhi_insn): Likewise.
20129 (movsi_insn): Likewise.
20130 (movsi_set_cc_insn): Likewise.
20131 (loadqi_update): Likewise.
20132 (load_zeroextendqisi_update): Likewise.
20133 (load_signextendqisi_update): Likewise.
20134 (loadhi_update): Likewise.
20135 (load_zeroextendhisi_update): Likewise.
20136 (load_signextendhisi_update): Likewise.
20137 (loadsi_update): Likewise.
20138 (loadsf_update): Likewise.
20139 (movsicc_insn): Likewise.
20140 (bset_insn): Likewise.
20141 (bxor_insn): Likewise.
20142 (bclr_insn): Likewise.
20143 (bmsk_insn): Likewise.
20144 (bicsi3_insn): Likewise.
20145 (cmpsi_cc_c_insn): Likewise.
20146 (movsi_ne): Likewise.
20147 (movsi_cond_exec): Likewise.
20148 (clrsbsi2): Likewise.
20149 (norm_f): Likewise.
20159 (call_i): Remove 'S' asm letter, add jli instruction.
20160 (call_value_i): Likewise.
20161 * config/arc/arc.op (mjli-always): New option.
20162 * config/arc/constraints.md (Cji): New constraint.
20163 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
20165 (subsf3_fpx): Likewise.
20166 (mulsf3_fpx): Likewise.
20167 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
20169 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
20170 function attrbutes.
20171 * doc/invoke.texi (ARC): Document mjli-always option.
20173 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
20175 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
20176 avoid addition with 0 and use incw and decw where possible.
20178 2018-01-26 Richard Biener <rguenther@suse.de>
20180 PR tree-optimization/81082
20181 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
20182 association if it requires casting to unsigned.
20183 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
20184 from fold_plusminus_mult_expr to catch important cases late when
20185 range info is available.
20187 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20189 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
20190 * configure.ac (hidden_linkonce): New test.
20191 * configure: Regenerate.
20192 * config.in: Regenerate.
20194 2018-01-26 Julia Koval <julia.koval@intel.com>
20196 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
20197 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
20198 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
20199 _mm_mask_bitshuffle_epi64_mask): Fix type.
20200 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
20201 USI_FTYPE_V4DI_V4DI_USI): Remove.
20202 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
20203 __builtin_ia32_vpshufbitqmb256_mask,
20204 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
20205 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
20206 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
20208 2018-01-26 Alan Modra <amodra@gmail.com>
20211 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
20212 UNSPEC_VBPERMQ. Sort other unspecs.
20214 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
20216 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
20218 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
20220 PR middle-end/83055
20221 * predict.c (drop_profile): Do not push/pop cfun; update also
20223 (handle_missing_profiles): Fix logic looking for zero profiles.
20225 2018-01-25 Jakub Jelinek <jakub@redhat.com>
20227 PR middle-end/83977
20228 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
20229 on functions with #pragma omp declare simd or functions with simd
20231 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
20232 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
20233 Remove trailing \n from warning_at calls.
20235 2018-01-25 Tom de Vries <tom@codesourcery.com>
20238 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20239 for neutered workers.
20241 2018-01-24 Joseph Myers <joseph@codesourcery.com>
20244 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
20245 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
20247 2018-01-24 Jeff Law <law@redhat.com>
20250 * i386.c (get_probe_interval): Move to earlier point.
20251 (ix86_compute_frame_layout): If -fstack-clash-protection and
20252 the frame is larger than the probe interval, then use pushes
20253 to save registers rather than reg->mem moves.
20254 (ix86_expand_prologue): Remove conditional for int_registers_saved
20257 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
20260 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
20261 min/max for never referenced object.
20263 2018-01-24 Jakub Jelinek <jakub@redhat.com>
20265 PR middle-end/83977
20266 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
20268 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
20269 attributes from DECL_ATTRIBUTES (decl) without affecting
20270 DECL_ATTRIBUTES (current_function_decl).
20271 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
20272 functions with non-NULL DECL_ABSTRACT_ORIGIN.
20274 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
20276 PR tree-optimization/83979
20277 * fold-const.c (fold_comparison): Use constant_boolean_node
20278 instead of boolean_{true,false}_node.
20280 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
20282 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
20285 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20287 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20288 Simplify the clause that sets the length attribute.
20289 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20290 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
20291 clause that sets the length attribute.
20292 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20294 2018-01-24 Tom de Vries <tom@codesourcery.com>
20297 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
20298 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
20299 Add strict parameter.
20300 (prevent_branch_around_nothing): Insert dummy insn between branch to
20301 label and label with no ptx insn inbetween.
20302 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
20304 2018-01-24 Tom de Vries <tom@codesourcery.com>
20307 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20308 for neutered threads in warp.
20309 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
20311 2018-01-24 Richard Biener <rguenther@suse.de>
20313 PR tree-optimization/83176
20314 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
20317 2018-01-24 Richard Biener <rguenther@suse.de>
20319 PR tree-optimization/82819
20320 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
20321 code generating pluses that are no-ops in the target precision.
20323 2018-01-24 Richard Biener <rguenther@suse.de>
20325 PR middle-end/84000
20326 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
20328 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20330 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
20331 to merge probabilities.
20332 * predict.c (probably_never_executed): Also mark as cold functions
20333 with global 0 profile and guessed local profile.
20334 * profile-count.c (profile_probability::combine_with_count): New
20336 * profile-count.h (profile_probability::operator*,
20337 profile_probability::operator*=, profile_probability::operator/,
20338 profile_probability::operator/=): Reduce precision to adjusted
20339 and set value to guessed on contradictory divisions.
20340 (profile_probability::combine_with_freq): Remove.
20341 (profile_probability::combine_wiht_count): Declare.
20342 (profile_count::force_nonzero):: Set to adjusted.
20343 (profile_count::probability_in):: Set quality to adjusted.
20344 * tree-ssa-tail-merge.c (replace_block_by): Use
20345 combine_with_count.
20347 2018-01-23 Andrew Waterman <andrew@sifive.com>
20348 Jim Wilson <jimw@sifive.com>
20350 * config/riscv/riscv.c (riscv_stack_boundary): New.
20351 (riscv_option_override): Set riscv_stack_boundary. Handle
20352 riscv_preferred_stack_boundary_arg.
20353 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
20354 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
20355 (STACK_BOUNDARY): Set to riscv_stack_boundary.
20356 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
20357 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
20358 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
20360 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
20363 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
20364 of struct ix86_frame.
20365 (ix86_expand_epilogue): Likewise. Add a local variable for
20366 the reg_save_offset field in struct ix86_frame.
20368 2018-01-23 Bin Cheng <bin.cheng@arm.com>
20370 PR tree-optimization/82604
20371 * tree-loop-distribution.c (enum partition_kind): New enum item
20372 PKIND_PARTIAL_MEMSET.
20373 (partition_builtin_p): Support above new enum item.
20374 (generate_code_for_partition): Ditto.
20375 (compute_access_range): Differentiate cases that equality can be
20376 proven at all loops, the innermost loops or no loops.
20377 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
20378 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
20379 (finalize_partitions, distribute_loop): Don't fuse partition of
20380 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
20381 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
20382 parloop is enabled.
20384 2018-01-23 Martin Liska <mliska@suse.cz>
20386 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
20387 order to ignore the predictor.
20388 (PRED_POLYMORPHIC_CALL): Likewise.
20389 (PRED_RECURSIVE_CALL): Likewise.
20391 2018-01-23 Martin Liska <mliska@suse.cz>
20393 * tree-profile.c (tree_profiling): Print function header to
20394 aware reader which function we are working on.
20395 * value-prof.c (gimple_find_values_to_profile): Do not print
20396 not interesting value histograms.
20398 2018-01-23 Martin Liska <mliska@suse.cz>
20400 * profile-count.h (enum profile_quality): Add
20401 profile_uninitialized as the first value. Do not number values
20402 as they are zero based.
20403 (profile_count::verify): Update sanity check.
20404 (profile_probability::verify): Likewise.
20406 2018-01-23 Nathan Sidwell <nathan@acm.org>
20408 * doc/invoke.texi (ffor-scope): Deprecate.
20410 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20412 PR tree-optimization/83510
20413 * domwalk.c (set_all_edges_as_executable): New function.
20414 (dom_walker::dom_walker): Convert bool param
20415 "skip_unreachable_blocks" to enum reachability. Move setup of
20416 edge flags to set_all_edges_as_executable and only do it when
20417 reachability is REACHABLE_BLOCKS.
20418 * domwalk.h (enum dom_walker::reachability): New enum.
20419 (dom_walker::dom_walker): Convert bool param
20420 "skip_unreachable_blocks" to enum reachability.
20421 (set_all_edges_as_executable): New decl.
20422 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
20423 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
20425 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
20426 but converting true to REACHABLE_BLOCKS.
20427 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
20429 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
20430 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
20431 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
20433 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
20434 if check_all_array_refs will be called.
20436 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20438 * tree.c (selftest::test_location_wrappers): Add more test
20441 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20443 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
20444 (selftest::test_bit_in_range): Likewise.
20446 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
20449 * doc/sourcebuild.texi (vect_float): Say that the selector
20450 only describes the situation when -funsafe-math-optimizations is on.
20451 (vect_float_strict): Document.
20453 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
20455 PR tree-optimization/83965
20456 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
20457 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
20458 instead of checking only for a reduction.
20459 (vect_recog_widen_sum_pattern): Likewise.
20461 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20463 * predict.c (probably_never_executed): Only use precise profile info.
20464 (compute_function_frequency): Skip after inlining hack since we now
20465 have quality checking.
20467 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20469 * profile-count.h (profile_probability::very_unlikely,
20470 profile_probability::unlikely, profile_probability::even): Set
20471 precision to guessed.
20473 2018-01-23 Richard Biener <rguenther@suse.de>
20475 PR tree-optimization/83963
20476 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
20477 Properly terminate dominator walk when crossing the exit edge not
20478 when visiting its source block.
20480 2018-01-23 Jakub Jelinek <jakub@redhat.com>
20483 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
20484 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
20486 2018-01-22 Jakub Jelinek <jakub@redhat.com>
20488 PR tree-optimization/83957
20489 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
20490 semicolon after for body surrounded by braces.
20492 PR tree-optimization/83081
20493 * profile-count.h (profile_probability::split): New method.
20494 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
20495 Use profile_probability::split.
20496 (do_compare_rtx_and_jump): Fix adjustment of probabilities
20497 when splitting a single conditional jump into 2.
20499 2018-01-22 David Malcolm <dmalcolm@redhat.com>
20501 PR tree-optimization/69452
20502 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
20505 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20507 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
20508 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
20509 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
20511 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20513 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
20515 * config/rl78/rl78.md (movdi): New define_expand.
20516 * config/rl78/rl78.c (rl78_split_movdi): New function.
20518 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
20521 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
20523 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
20524 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
20525 128-bit to produce an UNSPEC move to get the double word with the
20526 signbit and then a shift directly to do signbit.
20527 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
20528 implementation with a new version that just does either a direct
20529 move or a regular move. Move memory interface to separate insns.
20530 Move insns so they are next to the expander.
20531 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
20532 with signbit move. Split big and little endian case.
20533 (signbit<mode>2_dm_mem_le): Likewise.
20534 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
20535 (signbit<mode>2_dm2): Likewise.
20537 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20539 * config/rl78/rl78.md (anddi3): New define_expand.
20541 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20543 * config/rl78/rl78.md (umindi3): New define_expand.
20545 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20547 * config/rl78/rl78.md (smindi3): New define_expand.
20549 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20551 * config/rl78/rl78.md (smaxdi3): New define_expand.
20553 2018-01-22 Carl Love <cel@us.ibm.com>
20555 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
20556 LVX_V1TI): Add macro expansion.
20557 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
20558 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
20559 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
20560 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
20561 Change check to determine if the instruction is a byte reversing
20562 entry. Fix typo in comment.
20563 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
20564 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
20565 Add def_builtin calls for new builtins.
20566 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
20567 Add define_insn expansion.
20569 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20571 * config/rl78/rl78.md (umaxdi3): New define_expand.
20573 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20575 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
20576 for non-QImode registers.
20578 2018-01-22 Richard Biener <rguenther@suse.de>
20580 PR tree-optimization/83963
20581 * graphite-scop-detection.c (scop_detection::get_sese): Delay
20582 including the loop exit block.
20583 (scop_detection::merge_sese): Likewise.
20584 (scop_detection::add_scop): Do it here instead.
20586 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20588 * doc/sourcebuild.texi (arm_softfloat): Document.
20590 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
20593 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
20594 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
20595 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
20597 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20598 David Edelsohn <dje.gcc@gmail.com>
20601 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20602 Change "crset eq" to "crset 2".
20603 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20604 (*call_indirect_aix<mode>_nospec): Likewise.
20605 (*call_value_indirect_aix<mode>_nospec): Likewise.
20606 (*call_indirect_elfv2<mode>_nospec): Likewise.
20607 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
20608 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
20609 change assembly output from . to $.
20610 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20611 (indirect_jump<mode>_nospec): Change assembly output from . to $.
20612 (*tablejump<mode>_internal1_nospec): Likewise.
20614 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
20617 * config/sh/sh_optimize_sett_clrt.cc:
20618 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
20620 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
20622 PR tree-optimization/83940
20623 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
20624 offset_dt to vect_constant_def rather than vect_unknown_def_type.
20625 (vect_check_load_store_mask): Add a mask_dt_out parameter and
20626 use it to pass back the definition type.
20627 (vect_check_store_rhs): Likewise rhs_dt_out.
20628 (vect_build_gather_load_calls): Add a mask_dt argument and use
20629 it instead of a call to vect_is_simple_use.
20630 (vectorizable_store): Update calls to vect_check_load_store_mask
20631 and vect_check_store_rhs. Use the dt returned by the latter instead
20632 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
20633 instead of calls to vect_is_simple_use. Pass the scalar rather
20634 than the vector operand to vect_is_simple_use when handling
20635 second and subsequent copies of an rhs value.
20636 (vectorizable_load): Update calls to vect_check_load_store_mask
20637 and vect_build_gather_load_calls. Use the cached mask_dt and
20638 gs_info.offset_dt instead of calls to vect_is_simple_use.
20640 2018-01-20 Jakub Jelinek <jakub@redhat.com>
20642 PR middle-end/83945
20643 * tree-emutls.c: Include gimplify.h.
20644 (lower_emutls_2): New function.
20645 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
20646 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
20647 it before further processing.
20650 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
20651 UINTVAL (trueop1) instead of INTVAL (op1).
20653 2018-01-19 Jakub Jelinek <jakub@redhat.com>
20657 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
20658 INCOMING_FRAME_SP_OFFSET if not defined.
20659 (scan_trace): Add ENTRY argument. If true and
20660 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
20661 emit a note to adjust the CFA offset.
20662 (create_cfi_notes): Adjust scan_trace callers.
20663 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
20664 INCOMING_FRAME_SP_OFFSET in the CIE.
20665 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
20666 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
20668 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
20669 * doc/tm.texi: Regenerated.
20671 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
20673 PR rtl-optimization/83147
20674 * lra-constraints.c (remove_inheritance_pseudos): Use
20675 lra_substitute_pseudo_within_insn.
20677 2018-01-19 Tom de Vries <tom@codesourcery.com>
20678 Cesar Philippidis <cesar@codesourcery.com>
20681 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
20683 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
20686 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
20687 spaces for function labels.
20689 2018-01-19 Martin Liska <mliska@suse.cz>
20691 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
20692 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
20693 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
20694 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
20695 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
20696 (PRED_CONST_RETURN): Change from 69 to 65.
20697 (PRED_NULL_RETURN): Change from 91 to 71.
20698 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
20699 (PRED_LOOP_GUARD): Change from 66 to 73.
20701 2018-01-19 Martin Liska <mliska@suse.cz>
20703 * predict.c (predict_insn_def): Add new assert.
20704 (struct branch_predictor): Change type to signed integer.
20705 (test_prediction_value_range): Amend test to cover
20706 PROB_UNINITIALIZED.
20707 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
20708 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
20709 (PRED_LOOP_ITERATIONS_MAX): Likewise.
20710 (PRED_LOOP_IV_COMPARE): Likewise.
20711 * predict.h (PROB_UNINITIALIZED): Define new constant.
20713 2018-01-19 Martin Liska <mliska@suse.cz>
20715 * predict.c (dump_prediction): Add new format for
20716 analyze_brprob.py script which is enabled with -details
20718 * profile-count.h (precise_p): New function.
20720 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
20722 PR tree-optimization/83922
20723 * tree-vect-loop.c (vect_verify_full_masking): Return false if
20724 there are no statements that need masking.
20725 (vect_active_double_reduction_p): New function.
20726 (vect_analyze_loop_operations): Use it when handling phis that
20727 are not in the loop header.
20729 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
20731 PR tree-optimization/83914
20732 * tree-vect-loop.c (vectorizable_induction): Don't convert
20733 init_expr or apply the peeling adjustment for inductions
20734 that are nested within the vectorized loop.
20736 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20738 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
20741 2018-01-18 Jakub Jelinek <jakub@redhat.com>
20745 * function.h (gimplify_parameters): Add gimple_seq * argument.
20746 * function.c: Include gimple.h and options.h.
20747 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
20748 for the added local temporaries if needed.
20749 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
20750 if there are any parameter cleanups, wrap whole body into a
20751 try/finally with the cleanups.
20753 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
20756 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
20757 Use GET_MODE_CLASS for scalar floating point.
20759 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
20763 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
20764 Fix call of call_cgraph_insertion_hooks.
20766 2018-01-18 Martin Sebor <msebor@redhat.com>
20768 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
20770 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
20773 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
20776 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
20779 * common.opt: (-ffile-prefix-map): New option.
20780 * opts.c (common_handle_option): Defer it.
20781 * opts-global.c (handle_common_deferred_options): Handle it.
20782 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
20783 * file-prefix-map.h: New file.
20784 (remap_debug_filename, add_debug_prefix_map): ...here.
20785 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
20786 * final.c (debug_prefix_map, add_debug_prefix_map
20787 remap_debug_filename): Move to...
20788 * file-prefix-map.c: New file.
20789 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
20790 generalize, get rid of alloca(), use strrchr() instead of strchr().
20791 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
20792 Implement in terms of add_prefix_map().
20793 (remap_macro_filename, remap_debug_filename): Implement in term of
20795 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
20796 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
20797 * dbxout.c: Include file-prefix-map.h.
20798 * varasm.c: Likewise.
20799 * vmsdbgout.c: Likewise.
20800 * xcoffout.c: Likewise.
20801 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
20802 * doc/cppopts.texi (-fmacro-prefix-map): Document.
20803 * doc/invoke.texi (-ffile-prefix-map): Document.
20804 (-fdebug-prefix-map): Update description.
20806 2018-01-18 Martin Liska <mliska@suse.cz>
20808 * config/i386/i386.c (indirect_thunk_name): Document that also
20810 (output_indirect_thunk): Document why both instructions
20811 (pause and lfence) are generated.
20813 2018-01-18 Richard Biener <rguenther@suse.de>
20815 PR tree-optimization/83887
20816 * graphite-scop-detection.c
20817 (scop_detection::get_nearest_dom_with_single_entry): Remove.
20818 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
20819 (scop_detection::merge_sese): Re-implement with a flood-fill
20820 algorithm that properly finds a SESE region if it exists.
20822 2018-01-18 Jakub Jelinek <jakub@redhat.com>
20825 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
20826 pointer_diff optimizations use view_convert instead of convert.
20828 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20830 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20831 Generate different code for -mno-speculate-indirect-jumps.
20832 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20833 (*call_indirect_aix<mode>): Disable for
20834 -mno-speculate-indirect-jumps.
20835 (*call_indirect_aix<mode>_nospec): New define_insn.
20836 (*call_value_indirect_aix<mode>): Disable for
20837 -mno-speculate-indirect-jumps.
20838 (*call_value_indirect_aix<mode>_nospec): New define_insn.
20839 (*sibcall_nonlocal_sysv<mode>): Generate different code for
20840 -mno-speculate-indirect-jumps.
20841 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20843 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
20845 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
20846 long double type, set the flags for noting the default long double
20847 type, even if we don't pass or return a long double type.
20849 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
20852 * ipa-inline.c (flatten_function): Do not overwrite final inlining
20855 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
20857 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
20858 support for merge[hl].
20859 (fold_mergehl_helper): New helper function.
20860 (tree-vector-builder.h): New #include for tree_vector_builder usage.
20861 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
20862 (altivec_vmrglw_direct): Add xxmrglw insn.
20864 2018-01-17 Andrew Waterman <andrew@sifive.com>
20866 * config/riscv/riscv.c (riscv_conditional_register_usage): If
20867 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
20869 2018-01-17 David Malcolm <dmalcolm@redhat.com>
20872 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
20873 call the lto_location_cache before reading the
20874 DECL_SOURCE_LOCATION of the types.
20876 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
20877 Richard Sandiford <richard.sandiford@linaro.org>
20879 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
20880 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
20881 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
20882 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
20883 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
20885 * config/aarch64/constraints.md (aarch64_movti_operand):
20887 * config/aarch64/predicates.md (Uti): Add new constraint.
20889 2018-01-17 Carl Love <cel@us.ibm.com>
20891 * config/rs6000/vsx.md (define_expand xl_len_r,
20892 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
20893 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
20895 (define_expand, define_insn): Move the shift left from the
20896 define_insn to the define_expand for lxvl and stxvl instructions.
20897 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
20898 and XL_LEN_R definitions to PURE.
20900 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
20902 * config/i386/i386.c (indirect_thunk_name): Declare regno
20903 as unsigned int. Compare regno with INVALID_REGNUM.
20904 (output_indirect_thunk): Ditto.
20905 (output_indirect_thunk_function): Ditto.
20906 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
20907 in the call to output_indirect_thunk_function.
20909 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
20911 PR middle-end/83884
20912 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
20913 rather than the size of inner_type to determine the stack slot size
20914 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
20916 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
20919 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
20922 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
20924 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
20925 endian Linux systems to optionally enable multilibs for selecting
20926 the long double type if the user configured an explicit type.
20927 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
20928 have no long double multilibs if not defined.
20929 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
20930 warn if the user used -mabi={ieee,ibm}longdouble and we built
20931 multilibs for long double.
20932 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
20933 appropriate multilib option.
20934 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
20936 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
20937 for building long double multilibs.
20938 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
20940 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
20942 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
20945 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
20947 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
20950 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
20953 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
20956 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
20958 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
20959 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
20961 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
20963 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
20964 different rtl trees depending on TARGET_64BIT.
20965 (rs6000_gen_lvx): Likewise.
20967 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
20969 * config/visium/visium.md (nop): Tweak comment.
20970 (hazard_nop): Likewise.
20972 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20974 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
20975 -mspeculate-indirect-jumps.
20976 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
20977 for -mno-speculate-indirect-jumps.
20978 (*call_indirect_elfv2<mode>_nospec): New define_insn.
20979 (*call_value_indirect_elfv2<mode>): Disable for
20980 -mno-speculate-indirect-jumps.
20981 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
20982 (indirect_jump): Emit different RTL for
20983 -mno-speculate-indirect-jumps.
20984 (*indirect_jump<mode>): Disable for
20985 -mno-speculate-indirect-jumps.
20986 (*indirect_jump<mode>_nospec): New define_insn.
20987 (tablejump): Emit different RTL for
20988 -mno-speculate-indirect-jumps.
20989 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
20990 (tablejumpsi_nospec): New define_expand.
20991 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
20992 (tablejumpdi_nospec): New define_expand.
20993 (*tablejump<mode>_internal1): Disable for
20994 -mno-speculate-indirect-jumps.
20995 (*tablejump<mode>_internal1_nospec): New define_insn.
20996 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
20999 2018-01-16 Artyom Skrobov tyomitch@gmail.com
21001 * caller-save.c (insert_save): Drop unnecessary parameter. All
21004 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21005 Richard Biener <rguenth@suse.de>
21008 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
21009 return early, inline manually is_gimple_sizepos. Make sure if we
21010 call gimplify_expr we don't end up with a gimple constant.
21011 * tree.c (variably_modified_type_p): Don't return true for
21012 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
21013 * gimplify.h (is_gimple_sizepos): Remove.
21015 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21017 PR tree-optimization/83857
21018 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
21019 vectorizable_live_operation for pure SLP statements.
21020 (vectorizable_live_operation): Handle PHIs.
21022 2018-01-16 Richard Biener <rguenther@suse.de>
21024 PR tree-optimization/83867
21025 * tree-vect-stmts.c (vect_transform_stmt): Precompute
21026 nested_in_vect_loop_p since the scalar stmt may get invalidated.
21028 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21031 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
21032 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
21033 If off is not INTEGER_CST, issue a may not be aligned warning
21034 rather than isn't aligned. Use isn%'t rather than isn't.
21035 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
21037 <case MULT_EXPR>: Improve the case when bottom and one of the
21038 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
21039 operand, in that case check if the other operand is multiple of
21040 bottom divided by the INTEGER_CST operand.
21042 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21045 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
21046 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
21047 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
21048 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
21049 * config/pa/pa.c (pa_function_arg_advance): Likewise.
21050 (pa_function_arg, pa_arg_partial_bytes): Likewise.
21051 (pa_function_arg_size): New function.
21053 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21055 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
21056 in a separate statement.
21058 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21060 PR tree-optimization/83847
21061 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
21062 group gathers and scatters.
21064 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21066 PR rtl-optimization/86620
21067 * params.def (max-sched-ready-insns): Bump minimum value to 1.
21069 PR rtl-optimization/83213
21070 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
21071 to last if both are JUMP_INSNs.
21073 PR tree-optimization/83843
21074 * gimple-ssa-store-merging.c
21075 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
21076 store_immediate_info for bswap/nop orig_stores.
21078 2018-01-15 Andrew Waterman <andrew@sifive.com>
21080 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
21082 <UDIV>: Increase cost if !TARGET_DIV.
21084 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
21086 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
21087 (define_attr "cr_logical_3op"): New.
21088 (cceq_ior_compare): Adjust.
21089 (cceq_ior_compare_complement): Adjust.
21090 (*cceq_rev_compare): Adjust.
21091 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
21092 (is_cracked_insn): Adjust.
21093 (insn_must_be_first_in_group): Adjust.
21094 * config/rs6000/40x.md: Adjust.
21095 * config/rs6000/440.md: Adjust.
21096 * config/rs6000/476.md: Adjust.
21097 * config/rs6000/601.md: Adjust.
21098 * config/rs6000/603.md: Adjust.
21099 * config/rs6000/6xx.md: Adjust.
21100 * config/rs6000/7450.md: Adjust.
21101 * config/rs6000/7xx.md: Adjust.
21102 * config/rs6000/8540.md: Adjust.
21103 * config/rs6000/cell.md: Adjust.
21104 * config/rs6000/e300c2c3.md: Adjust.
21105 * config/rs6000/e500mc.md: Adjust.
21106 * config/rs6000/e500mc64.md: Adjust.
21107 * config/rs6000/e5500.md: Adjust.
21108 * config/rs6000/e6500.md: Adjust.
21109 * config/rs6000/mpc.md: Adjust.
21110 * config/rs6000/power4.md: Adjust.
21111 * config/rs6000/power5.md: Adjust.
21112 * config/rs6000/power6.md: Adjust.
21113 * config/rs6000/power7.md: Adjust.
21114 * config/rs6000/power8.md: Adjust.
21115 * config/rs6000/power9.md: Adjust.
21116 * config/rs6000/rs64.md: Adjust.
21117 * config/rs6000/titan.md: Adjust.
21119 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21121 * config/i386/predicates.md (indirect_branch_operand): Rewrite
21122 ix86_indirect_branch_register logic.
21124 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21126 * config/i386/constraints.md (Bs): Update
21127 ix86_indirect_branch_register check. Don't check
21128 ix86_indirect_branch_register with GOT_memory_operand.
21130 * config/i386/predicates.md (GOT_memory_operand): Don't check
21131 ix86_indirect_branch_register here.
21132 (GOT32_symbol_operand): Likewise.
21134 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21136 * config/i386/predicates.md (constant_call_address_operand):
21137 Rewrite ix86_indirect_branch_register logic.
21138 (sibcall_insn_operand): Likewise.
21140 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21142 * config/i386/constraints.md (Bs): Replace
21143 ix86_indirect_branch_thunk_register with
21144 ix86_indirect_branch_register.
21146 * config/i386/i386.md (indirect_jump): Likewise.
21147 (tablejump): Likewise.
21148 (*sibcall_memory): Likewise.
21149 (*sibcall_value_memory): Likewise.
21150 Peepholes of indirect call and jump via memory: Likewise.
21151 * config/i386/i386.opt: Likewise.
21152 * config/i386/predicates.md (indirect_branch_operand): Likewise.
21153 (GOT_memory_operand): Likewise.
21154 (call_insn_operand): Likewise.
21155 (sibcall_insn_operand): Likewise.
21156 (GOT32_symbol_operand): Likewise.
21158 2018-01-15 Jakub Jelinek <jakub@redhat.com>
21160 PR middle-end/83837
21161 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
21162 type rather than type addr's type points to.
21163 (expand_omp_atomic_mutex): Likewise.
21164 (expand_omp_atomic): Likewise.
21166 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21169 * config/i386/i386.c (output_indirect_thunk_function): Use
21170 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
21171 for __x86_return_thunk.
21173 2018-01-15 Richard Biener <rguenther@suse.de>
21175 PR middle-end/83850
21176 * expmed.c (extract_bit_field_1): Fix typo.
21178 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21181 * config/arm/iterators.md (VF): New mode iterator.
21182 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
21183 Remove integer-related logic from pattern.
21184 (neon_vabd<mode>_3): Likewise.
21186 2018-01-15 Jakub Jelinek <jakub@redhat.com>
21188 PR middle-end/82694
21189 * common.opt (fstrict-overflow): No longer an alias.
21190 (fwrapv-pointer): New option.
21191 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
21192 also for pointer types based on flag_wrapv_pointer.
21193 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
21194 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
21195 opts->x_flag_wrapv got set.
21196 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
21197 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
21198 POINTER_TYPE_OVERFLOW_UNDEFINED.
21199 * match.pd: Likewise in address comparison pattern.
21200 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
21202 2018-01-15 Richard Biener <rguenther@suse.de>
21205 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
21206 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
21207 Reset type names to their identifier if their TYPE_DECL doesn't
21208 have linkage (and thus is used for ODR and devirt).
21209 (save_debug_info_for_decl): Remove.
21210 (save_debug_info_for_type): Likewise.
21211 (add_tree_to_fld_list): Adjust.
21212 * tree-pretty-print.c (dump_generic_node): Make dumping of
21213 type names more robust.
21215 2018-01-15 Richard Biener <rguenther@suse.de>
21217 * BASE-VER: Bump to 8.0.1.
21219 2018-01-14 Martin Sebor <msebor@redhat.com>
21222 * builtins.c (check_access): Avoid warning when the no-warning bit
21225 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
21227 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
21228 * ira-color (allocno_hard_regs_compare): Likewise.
21230 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
21233 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
21234 Use .pushsection/.popsection.
21236 2018-01-14 Martin Sebor <msebor@redhat.com>
21239 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
21241 2018-01-14 Jakub Jelinek <jakub@redhat.com>
21243 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
21244 entry from extra_headers.
21245 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
21246 extra_headers, make the list bitwise identical to the i?86-*-* one.
21248 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21250 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
21251 -mcmodel=large with -mindirect-branch=thunk,
21252 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
21253 -mfunction-return=thunk-extern.
21254 * doc/invoke.texi: Document -mcmodel=large is incompatible with
21255 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
21256 -mfunction-return=thunk and -mfunction-return=thunk-extern.
21258 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21260 * config/i386/i386.c (print_reg): Print the name of the full
21261 integer register without '%'.
21262 (ix86_print_operand): Handle 'V'.
21263 * doc/extend.texi: Document 'V' modifier.
21265 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21267 * config/i386/constraints.md (Bs): Disallow memory operand for
21268 -mindirect-branch-register.
21270 * config/i386/predicates.md (indirect_branch_operand): Likewise.
21271 (GOT_memory_operand): Likewise.
21272 (call_insn_operand): Likewise.
21273 (sibcall_insn_operand): Likewise.
21274 (GOT32_symbol_operand): Likewise.
21275 * config/i386/i386.md (indirect_jump): Call convert_memory_address
21276 for -mindirect-branch-register.
21277 (tablejump): Likewise.
21278 (*sibcall_memory): Likewise.
21279 (*sibcall_value_memory): Likewise.
21280 Disallow peepholes of indirect call and jump via memory for
21281 -mindirect-branch-register.
21282 (*call_pop): Replace m with Bw.
21283 (*call_value_pop): Likewise.
21284 (*sibcall_pop_memory): Replace m with Bs.
21285 * config/i386/i386.opt (mindirect-branch-register): New option.
21286 * doc/invoke.texi: Document -mindirect-branch-register option.
21288 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21290 * config/i386/i386-protos.h (ix86_output_function_return): New.
21291 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
21292 set function_return_type.
21293 (indirect_thunk_name): Add ret_p to indicate thunk for function
21295 (output_indirect_thunk_function): Pass false to
21296 indirect_thunk_name.
21297 (ix86_output_indirect_branch_via_reg): Likewise.
21298 (ix86_output_indirect_branch_via_push): Likewise.
21299 (output_indirect_thunk_function): Create alias for function
21300 return thunk if regno < 0.
21301 (ix86_output_function_return): New function.
21302 (ix86_handle_fndecl_attribute): Handle function_return.
21303 (ix86_attribute_table): Add function_return.
21304 * config/i386/i386.h (machine_function): Add
21305 function_return_type.
21306 * config/i386/i386.md (simple_return_internal): Use
21307 ix86_output_function_return.
21308 (simple_return_internal_long): Likewise.
21309 * config/i386/i386.opt (mfunction-return=): New option.
21310 (indirect_branch): Mention -mfunction-return=.
21311 * doc/extend.texi: Document function_return function attribute.
21312 * doc/invoke.texi: Document -mfunction-return= option.
21314 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21316 * config/i386/i386-opts.h (indirect_branch): New.
21317 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
21318 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
21319 with local indirect jump when converting indirect call and jump.
21320 (ix86_set_indirect_branch_type): New.
21321 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
21322 (indirectlabelno): New.
21323 (indirect_thunk_needed): Likewise.
21324 (indirect_thunk_bnd_needed): Likewise.
21325 (indirect_thunks_used): Likewise.
21326 (indirect_thunks_bnd_used): Likewise.
21327 (INDIRECT_LABEL): Likewise.
21328 (indirect_thunk_name): Likewise.
21329 (output_indirect_thunk): Likewise.
21330 (output_indirect_thunk_function): Likewise.
21331 (ix86_output_indirect_branch_via_reg): Likewise.
21332 (ix86_output_indirect_branch_via_push): Likewise.
21333 (ix86_output_indirect_branch): Likewise.
21334 (ix86_output_indirect_jmp): Likewise.
21335 (ix86_code_end): Call output_indirect_thunk_function if needed.
21336 (ix86_output_call_insn): Call ix86_output_indirect_branch if
21338 (ix86_handle_fndecl_attribute): Handle indirect_branch.
21339 (ix86_attribute_table): Add indirect_branch.
21340 * config/i386/i386.h (machine_function): Add indirect_branch_type
21341 and has_local_indirect_jump.
21342 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
21344 (tablejump): Likewise.
21345 (*indirect_jump): Use ix86_output_indirect_jmp.
21346 (*tablejump_1): Likewise.
21347 (simple_return_indirect_internal): Likewise.
21348 * config/i386/i386.opt (mindirect-branch=): New option.
21349 (indirect_branch): New.
21352 (thunk-inline): Likewise.
21353 (thunk-extern): Likewise.
21354 * doc/extend.texi: Document indirect_branch function attribute.
21355 * doc/invoke.texi: Document -mindirect-branch= option.
21357 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
21360 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
21362 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
21364 * ipa-inline.c (want_inline_small_function_p): Return false if
21365 inlining has already failed with CIF_FINAL_ERROR.
21366 (update_caller_keys): Call want_inline_small_function_p before
21368 (update_callee_keys): Likewise.
21370 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
21372 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
21374 (rs6000_quadword_masked_address_p): Likewise.
21375 (quad_aligned_load_p): Likewise.
21376 (quad_aligned_store_p): Likewise.
21377 (const_load_sequence_p): Add comment to describe the outer-most loop.
21378 (mimic_memory_attributes_and_flags): New function.
21379 (rs6000_gen_stvx): Likewise.
21380 (replace_swapped_aligned_store): Likewise.
21381 (rs6000_gen_lvx): Likewise.
21382 (replace_swapped_aligned_load): Likewise.
21383 (replace_swapped_load_constant): Capitalize argument name in
21384 comment describing this function.
21385 (rs6000_analyze_swaps): Add a third pass to search for vector loads
21386 and stores that access quad-word aligned addresses and replace
21387 with stvx or lvx instructions when appropriate.
21388 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
21389 New function prototype.
21390 (rs6000_quadword_masked_address_p): Likewise.
21391 (rs6000_gen_lvx): Likewise.
21392 (rs6000_gen_stvx): Likewise.
21393 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
21394 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
21395 when memory address is aligned.
21396 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
21397 this split to select lvx instruction when memory address is aligned.
21398 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
21399 instruction when memory address is aligned.
21400 (*vsx_le_perm_load_v16qi): Likewise.
21401 (four unnamed splitters): Modify to select the stvx instruction
21402 when memory is aligned.
21404 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
21406 * predict.c (determine_unlikely_bbs): Handle correctly BBs
21407 which appears in the queue multiple times.
21409 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21410 Alan Hayward <alan.hayward@arm.com>
21411 David Sherwood <david.sherwood@arm.com>
21413 * tree-vectorizer.h (vec_lower_bound): New structure.
21414 (_loop_vec_info): Add check_nonzero and lower_bounds.
21415 (LOOP_VINFO_CHECK_NONZERO): New macro.
21416 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
21417 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
21418 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
21419 fields. Make seg_len the distance travelled, not including the
21421 (dr_direction_indicator): Declare.
21422 (dr_zero_step_indicator): Likewise.
21423 (dr_known_forward_stride_p): Likewise.
21424 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
21426 (runtime_alias_check_p): Allow runtime alias checks with
21428 (operator ==): Compare access_size and align.
21429 (prune_runtime_alias_test_list): Rework for new distinction between
21430 the access_size and seg_len.
21431 (create_intersect_range_checks_index): Likewise. Cope with polynomial
21433 (get_segment_min_max): New function.
21434 (create_intersect_range_checks): Use it.
21435 (dr_step_indicator): New function.
21436 (dr_direction_indicator): Likewise.
21437 (dr_zero_step_indicator): Likewise.
21438 (dr_known_forward_stride_p): Likewise.
21439 * tree-loop-distribution.c (data_ref_segment_size): Return
21440 DR_STEP * (niters - 1).
21441 (compute_alias_check_pairs): Update call to the dr_with_seg_len
21443 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
21444 (vect_preserves_scalar_order_p): New function, split out from...
21445 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
21446 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
21447 (vect_vfa_access_size): New function.
21448 (vect_vfa_align): Likewise.
21449 (vect_compile_time_alias): Take access_size_a and access_b arguments.
21450 (dump_lower_bound): New function.
21451 (vect_check_lower_bound): Likewise.
21452 (vect_small_gap_p): Likewise.
21453 (vectorizable_with_step_bound_p): Likewise.
21454 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
21455 depencies if the vectorization factor is 1. Convert the checks
21456 for nonzero steps into checks on the bounds of DR_STEP. Try using
21457 a bunds check for variable steps if the minimum required step is
21458 relatively small. Update calls to the dr_with_seg_len
21459 constructor and to vect_compile_time_alias.
21460 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
21462 (vect_loop_versioning): Call it.
21463 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
21465 (vect_estimate_min_profitable_iters): Account for any bounds checks.
21467 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21468 Alan Hayward <alan.hayward@arm.com>
21469 David Sherwood <david.sherwood@arm.com>
21471 * doc/sourcebuild.texi (vect_scatter_store): Document.
21472 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
21474 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
21476 * genopinit.c (main): Add supports_vec_scatter_store and
21477 supports_vec_scatter_store_cached to target_optabs.
21478 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
21479 IFN_MASK_SCATTER_STORE.
21480 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
21482 * internal-fn.h (internal_store_fn_p): Declare.
21483 (internal_fn_stored_value_index): Likewise.
21484 * internal-fn.c (scatter_store_direct): New macro.
21485 (expand_scatter_store_optab_fn): New function.
21486 (direct_scatter_store_optab_supported_p): New macro.
21487 (internal_store_fn_p): New function.
21488 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
21489 IFN_MASK_SCATTER_STORE.
21490 (internal_fn_mask_index): Likewise.
21491 (internal_fn_stored_value_index): New function.
21492 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
21493 for scatter stores.
21494 * optabs-query.h (supports_vec_scatter_store_p): Declare.
21495 * optabs-query.c (supports_vec_scatter_store_p): New function.
21496 * tree-vectorizer.h (vect_get_store_rhs): Declare.
21497 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
21498 true for scatter stores.
21499 (vect_gather_scatter_fn_p): Handle scatter stores too.
21500 (vect_check_gather_scatter): Consider using scatter stores if
21501 supports_vec_scatter_store_p.
21502 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
21503 scatter stores too.
21504 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
21505 internal_fn_stored_value_index.
21506 (check_load_store_masking): Handle scatter stores too.
21507 (vect_get_store_rhs): Make public.
21508 (vectorizable_call): Use internal_store_fn_p.
21509 (vectorizable_store): Handle scatter store internal functions.
21510 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
21511 when deciding whether the end of the group has been reached.
21512 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
21513 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
21514 (mask_scatter_store<mode>): New insns.
21516 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21517 Alan Hayward <alan.hayward@arm.com>
21518 David Sherwood <david.sherwood@arm.com>
21520 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
21521 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
21522 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
21524 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
21525 Use vect_truncate_gather_scatter_offset if we can't treat the
21526 operation as a normal gather load or scatter store.
21527 (get_group_load_store_type): Take the gather_scatter_info
21528 as argument. Try using a gather load or scatter store for
21529 single-element groups.
21530 (get_load_store_type): Update calls to get_group_load_store_type
21531 and vect_use_strided_gather_scatters_p.
21533 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21534 Alan Hayward <alan.hayward@arm.com>
21535 David Sherwood <david.sherwood@arm.com>
21537 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
21538 optional tree argument.
21539 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
21541 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
21542 but continue to use the current value as a fallback.
21543 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
21544 to compare the updates.
21545 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
21546 (get_load_store_type): Use it when handling a strided access.
21547 (vect_get_strided_load_store_ops): New function.
21548 (vect_get_data_ptr_increment): Likewise.
21549 (vectorizable_load): Handle strided gather loads. Always pass
21550 a step to vect_create_data_ref_ptr and bump_vector_ptr.
21552 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21553 Alan Hayward <alan.hayward@arm.com>
21554 David Sherwood <david.sherwood@arm.com>
21556 * doc/md.texi (gather_load@var{m}): Document.
21557 (mask_gather_load@var{m}): Likewise.
21558 * genopinit.c (main): Add supports_vec_gather_load and
21559 supports_vec_gather_load_cached to target_optabs.
21560 * optabs-tree.c (init_tree_optimization_optabs): Use
21561 ggc_cleared_alloc to allocate target_optabs.
21562 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
21563 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
21565 * internal-fn.h (internal_load_fn_p): Declare.
21566 (internal_gather_scatter_fn_p): Likewise.
21567 (internal_fn_mask_index): Likewise.
21568 (internal_gather_scatter_fn_supported_p): Likewise.
21569 * internal-fn.c (gather_load_direct): New macro.
21570 (expand_gather_load_optab_fn): New function.
21571 (direct_gather_load_optab_supported_p): New macro.
21572 (direct_internal_fn_optab): New function.
21573 (internal_load_fn_p): Likewise.
21574 (internal_gather_scatter_fn_p): Likewise.
21575 (internal_fn_mask_index): Likewise.
21576 (internal_gather_scatter_fn_supported_p): Likewise.
21577 * optabs-query.c (supports_at_least_one_mode_p): New function.
21578 (supports_vec_gather_load_p): Likewise.
21579 * optabs-query.h (supports_vec_gather_load_p): Declare.
21580 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
21581 and memory_type field.
21582 (NUM_PATTERNS): Bump to 15.
21583 * tree-vect-data-refs.c: Include internal-fn.h.
21584 (vect_gather_scatter_fn_p): New function.
21585 (vect_describe_gather_scatter_call): Likewise.
21586 (vect_check_gather_scatter): Try using internal functions for
21587 gather loads. Recognize existing calls to a gather load function.
21588 (vect_analyze_data_refs): Consider using gather loads if
21589 supports_vec_gather_load_p.
21590 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
21591 (vect_get_gather_scatter_offset_type): Likewise.
21592 (vect_convert_mask_for_vectype): Likewise.
21593 (vect_add_conversion_to_patterm): Likewise.
21594 (vect_try_gather_scatter_pattern): Likewise.
21595 (vect_recog_gather_scatter_pattern): New pattern recognizer.
21596 (vect_vect_recog_func_ptrs): Add it.
21597 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
21598 internal_fn_mask_index and internal_gather_scatter_fn_p.
21599 (check_load_store_masking): Take the gather_scatter_info as an
21600 argument and handle gather loads.
21601 (vect_get_gather_scatter_ops): New function.
21602 (vectorizable_call): Check internal_load_fn_p.
21603 (vectorizable_load): Likewise. Handle gather load internal
21605 (vectorizable_store): Update call to check_load_store_masking.
21606 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
21607 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
21608 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
21609 (aarch64_gather_scale_operand_d): New predicates.
21610 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
21611 (mask_gather_load<mode>): New insns.
21613 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21614 Alan Hayward <alan.hayward@arm.com>
21615 David Sherwood <david.sherwood@arm.com>
21617 * optabs.def (fold_left_plus_optab): New optab.
21618 * doc/md.texi (fold_left_plus_@var{m}): Document.
21619 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
21620 * internal-fn.c (fold_left_direct): Define.
21621 (expand_fold_left_optab_fn): Likewise.
21622 (direct_fold_left_optab_supported_p): Likewise.
21623 * fold-const-call.c (fold_const_fold_left): New function.
21624 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
21625 * tree-parloops.c (valid_reduction_p): New function.
21626 (gather_scalar_reductions): Use it.
21627 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
21628 (vect_finish_replace_stmt): Declare.
21629 * tree-vect-loop.c (fold_left_reduction_fn): New function.
21630 (needs_fold_left_reduction_p): New function, split out from...
21631 (vect_is_simple_reduction): ...here. Accept reductions that
21632 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
21633 (vect_force_simple_reduction): Also store the reduction type in
21634 the assignment's STMT_VINFO_REDUC_TYPE.
21635 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
21636 (merge_with_identity): New function.
21637 (vect_expand_fold_left): Likewise.
21638 (vectorize_fold_left_reduction): Likewise.
21639 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
21640 scalar phi in place for it. Check for target support and reject
21641 cases that would reassociate the operation. Defer the transform
21642 phase to vectorize_fold_left_reduction.
21643 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
21644 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
21645 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
21647 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21649 * tree-if-conv.c (predicate_mem_writes): Remove redundant
21650 call to ifc_temp_var.
21652 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21653 Alan Hayward <alan.hayward@arm.com>
21654 David Sherwood <david.sherwood@arm.com>
21656 * target.def (legitimize_address_displacement): Take the original
21657 offset as a poly_int.
21658 * targhooks.h (default_legitimize_address_displacement): Update
21660 * targhooks.c (default_legitimize_address_displacement): Likewise.
21661 * doc/tm.texi: Regenerate.
21662 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
21663 as an argument, moving assert of ad->disp == ad->disp_term to...
21664 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
21665 Try calling targetm.legitimize_address_displacement before expanding
21666 the address rather than afterwards, and adjust for the new interface.
21667 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
21668 Match the new hook interface. Handle SVE addresses.
21669 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
21670 new hook interface.
21672 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21674 * Makefile.in (OBJS): Add early-remat.o.
21675 * target.def (select_early_remat_modes): New hook.
21676 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
21677 * doc/tm.texi: Regenerate.
21678 * targhooks.h (default_select_early_remat_modes): Declare.
21679 * targhooks.c (default_select_early_remat_modes): New function.
21680 * timevar.def (TV_EARLY_REMAT): New timevar.
21681 * passes.def (pass_early_remat): New pass.
21682 * tree-pass.h (make_pass_early_remat): Declare.
21683 * early-remat.c: New file.
21684 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
21686 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
21688 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21689 Alan Hayward <alan.hayward@arm.com>
21690 David Sherwood <david.sherwood@arm.com>
21692 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
21693 vfm1 with a bound_epilog parameter.
21694 (vect_do_peeling): Update calls accordingly, and move the prologue
21695 call earlier in the function. Treat the base bound_epilog as 0 for
21696 fully-masked loops and retain vf - 1 for other loops. Add 1 to
21697 this base when peeling for gaps.
21698 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
21699 with fully-masked loops.
21700 (vect_estimate_min_profitable_iters): Handle the single peeled
21701 iteration in that case.
21703 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21704 Alan Hayward <alan.hayward@arm.com>
21705 David Sherwood <david.sherwood@arm.com>
21707 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
21708 single-element interleaving even if the size is not a power of 2.
21709 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
21710 accesses for single-element interleaving if the group size is
21713 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21714 Alan Hayward <alan.hayward@arm.com>
21715 David Sherwood <david.sherwood@arm.com>
21717 * doc/md.texi (fold_extract_last_@var{m}): Document.
21718 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
21719 * optabs.def (fold_extract_last_optab): New optab.
21720 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
21721 * internal-fn.c (fold_extract_direct): New macro.
21722 (expand_fold_extract_optab_fn): Likewise.
21723 (direct_fold_extract_optab_supported_p): Likewise.
21724 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
21725 * tree-vect-loop.c (vect_model_reduction_cost): Handle
21726 EXTRACT_LAST_REDUCTION.
21727 (get_initial_def_for_reduction): Do not create an initial vector
21728 for EXTRACT_LAST_REDUCTION reductions.
21729 (vectorizable_reduction): Leave the scalar phi in place for
21730 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
21731 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
21732 epilogue code for EXTRACT_LAST_REDUCTION and defer the
21733 transform phase to vectorizable_condition.
21734 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
21736 (vect_finish_stmt_generation): ...here.
21737 (vect_finish_replace_stmt): New function.
21738 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
21739 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
21741 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
21743 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21744 Alan Hayward <alan.hayward@arm.com>
21745 David Sherwood <david.sherwood@arm.com>
21747 * doc/md.texi (extract_last_@var{m}): Document.
21748 * optabs.def (extract_last_optab): New optab.
21749 * internal-fn.def (EXTRACT_LAST): New internal function.
21750 * internal-fn.c (cond_unary_direct): New macro.
21751 (expand_cond_unary_optab_fn): Likewise.
21752 (direct_cond_unary_optab_supported_p): Likewise.
21753 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
21754 loops using EXTRACT_LAST.
21755 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
21756 (extract_last_<mode>): ...this optab.
21757 (vec_extract<mode><Vel>): Update accordingly.
21759 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21760 Alan Hayward <alan.hayward@arm.com>
21761 David Sherwood <david.sherwood@arm.com>
21763 * target.def (empty_mask_is_expensive): New hook.
21764 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
21765 * doc/tm.texi: Regenerate.
21766 * targhooks.h (default_empty_mask_is_expensive): Declare.
21767 * targhooks.c (default_empty_mask_is_expensive): New function.
21768 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
21769 if the target says that empty masks are expensive.
21770 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
21772 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
21774 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21775 Alan Hayward <alan.hayward@arm.com>
21776 David Sherwood <david.sherwood@arm.com>
21778 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
21779 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
21780 (vect_use_loop_mask_for_alignment_p): New function.
21781 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
21782 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
21783 niters_skip argument. Make sure that the first niters_skip elements
21784 of the first iteration are inactive.
21785 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
21786 Update call to vect_set_loop_masks_directly.
21787 (get_misalign_in_elems): New function, split out from...
21788 (vect_gen_prolog_loop_niters): ...here.
21789 (vect_update_init_of_dr): Take a code argument that specifies whether
21790 the adjustment should be added or subtracted.
21791 (vect_update_init_of_drs): Likewise.
21792 (vect_prepare_for_masked_peels): New function.
21793 (vect_do_peeling): Skip prologue peeling if we're using a mask
21794 instead. Update call to vect_update_inits_of_drs.
21795 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21797 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
21798 alignment. Do not include the number of peeled iterations in
21799 the minimum threshold in that case.
21800 (vectorizable_induction): Adjust the start value down by
21801 LOOP_VINFO_MASK_SKIP_NITERS iterations.
21802 (vect_transform_loop): Call vect_prepare_for_masked_peels.
21803 Take the number of skipped iterations into account when calculating
21805 * tree-vect-stmts.c (vect_gen_while_not): New function.
21807 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21808 Alan Hayward <alan.hayward@arm.com>
21809 David Sherwood <david.sherwood@arm.com>
21811 * doc/sourcebuild.texi (vect_fully_masked): Document.
21812 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
21813 default value to 0.
21814 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
21816 (vect_analyze_loop_2): ...here. Don't check the vectorization
21817 factor against the number of loop iterations if the loop is
21820 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21821 Alan Hayward <alan.hayward@arm.com>
21822 David Sherwood <david.sherwood@arm.com>
21824 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
21825 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
21826 (dump_groups): Update accordingly.
21827 (iv_use::mem_type): New member variable.
21828 (address_p): New function.
21829 (record_use): Add a mem_type argument and initialize the new
21831 (record_group_use): Add a mem_type argument. Use address_p.
21832 Remove obsolete null checks of base_object. Update call to record_use.
21833 (find_interesting_uses_op): Update call to record_group_use.
21834 (find_interesting_uses_cond): Likewise.
21835 (find_interesting_uses_address): Likewise.
21836 (get_mem_type_for_internal_fn): New function.
21837 (find_address_like_use): Likewise.
21838 (find_interesting_uses_stmt): Try find_address_like_use before
21839 calling find_interesting_uses_op.
21840 (addr_offset_valid_p): Use the iv mem_type field as the type
21841 of the addressed memory.
21842 (add_autoinc_candidates): Likewise.
21843 (get_address_cost): Likewise.
21844 (split_small_address_groups_p): Use address_p.
21845 (split_address_groups): Likewise.
21846 (add_iv_candidate_for_use): Likewise.
21847 (autoinc_possible_for_pair): Likewise.
21848 (rewrite_groups): Likewise.
21849 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
21850 (determine_group_iv_cost): Update after split of USE_ADDRESS.
21851 (get_alias_ptr_type_for_ptr_address): New function.
21852 (rewrite_use_address): Rewrite address uses in calls that were
21853 identified by find_address_like_use.
21855 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21856 Alan Hayward <alan.hayward@arm.com>
21857 David Sherwood <david.sherwood@arm.com>
21859 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
21861 * gimple-expr.h (is_gimple_addressable: Likewise.
21862 * gimple-expr.c (is_gimple_address): Likewise.
21863 * internal-fn.c (expand_call_mem_ref): New function.
21864 (expand_mask_load_optab_fn): Use it.
21865 (expand_mask_store_optab_fn): Likewise.
21867 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21868 Alan Hayward <alan.hayward@arm.com>
21869 David Sherwood <david.sherwood@arm.com>
21871 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
21872 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
21873 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
21874 (cond_umax@var{mode}): Document.
21875 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
21876 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
21877 (cond_umin_optab, cond_umax_optab): New optabs.
21878 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
21879 (COND_IOR, COND_XOR): New internal functions.
21880 * internal-fn.h (get_conditional_internal_fn): Declare.
21881 * internal-fn.c (cond_binary_direct): New macro.
21882 (expand_cond_binary_optab_fn): Likewise.
21883 (direct_cond_binary_optab_supported_p): Likewise.
21884 (get_conditional_internal_fn): New function.
21885 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
21886 Cope with reduction statements that are vectorized as calls rather
21888 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
21889 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
21890 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
21891 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
21892 (UNSPEC_COND_EOR): New unspecs.
21893 (optab): Add mappings for them.
21894 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
21895 (sve_int_op, sve_fp_op): New int attributes.
21897 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21898 Alan Hayward <alan.hayward@arm.com>
21899 David Sherwood <david.sherwood@arm.com>
21901 * optabs.def (while_ult_optab): New optab.
21902 * doc/md.texi (while_ult@var{m}@var{n}): Document.
21903 * internal-fn.def (WHILE_ULT): New internal function.
21904 * internal-fn.h (direct_internal_fn_supported_p): New override
21905 that takes two types as argument.
21906 * internal-fn.c (while_direct): New macro.
21907 (expand_while_optab_fn): New function.
21908 (convert_optab_supported_p): Likewise.
21909 (direct_while_optab_supported_p): New macro.
21910 * wide-int.h (wi::udiv_ceil): New function.
21911 * tree-vectorizer.h (rgroup_masks): New structure.
21912 (vec_loop_masks): New typedef.
21913 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
21914 and fully_masked_p.
21915 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
21916 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
21917 (vect_max_vf): New function.
21918 (slpeel_make_loop_iterate_ntimes): Delete.
21919 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
21920 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
21921 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
21922 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
21923 internal-fn.h, stor-layout.h and optabs-query.h.
21924 (vect_set_loop_mask): New function.
21925 (add_preheader_seq): Likewise.
21926 (add_header_seq): Likewise.
21927 (interleave_supported_p): Likewise.
21928 (vect_maybe_permute_loop_masks): Likewise.
21929 (vect_set_loop_masks_directly): Likewise.
21930 (vect_set_loop_condition_masked): Likewise.
21931 (vect_set_loop_condition_unmasked): New function, split out from
21932 slpeel_make_loop_iterate_ntimes.
21933 (slpeel_make_loop_iterate_ntimes): Rename to..
21934 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
21935 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
21936 (vect_do_peeling): Update call accordingly.
21937 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
21939 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21940 mask_compare_type, can_fully_mask_p and fully_masked_p.
21941 (release_vec_loop_masks): New function.
21942 (_loop_vec_info): Use it to free the loop masks.
21943 (can_produce_all_loop_masks_p): New function.
21944 (vect_get_max_nscalars_per_iter): Likewise.
21945 (vect_verify_full_masking): Likewise.
21946 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
21947 retries, and free the mask rgroups before retrying. Check loop-wide
21948 reasons for disallowing fully-masked loops. Make the final decision
21949 about whether use a fully-masked loop or not.
21950 (vect_estimate_min_profitable_iters): Do not assume that peeling
21951 for the number of iterations will be needed for fully-masked loops.
21952 (vectorizable_reduction): Disable fully-masked loops.
21953 (vectorizable_live_operation): Likewise.
21954 (vect_halve_mask_nunits): New function.
21955 (vect_double_mask_nunits): Likewise.
21956 (vect_record_loop_mask): Likewise.
21957 (vect_get_loop_mask): Likewise.
21958 (vect_transform_loop): Handle the case in which the final loop
21959 iteration might handle a partial vector. Call vect_set_loop_condition
21960 instead of slpeel_make_loop_iterate_ntimes.
21961 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
21962 (check_load_store_masking): New function.
21963 (prepare_load_store_mask): Likewise.
21964 (vectorizable_store): Handle fully-masked loops.
21965 (vectorizable_load): Likewise.
21966 (supportable_widening_operation): Use vect_halve_mask_nunits for
21968 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
21969 (vect_gen_while): New function.
21970 * config/aarch64/aarch64.md (umax<mode>3): New expander.
21971 (aarch64_uqdec<mode>): New insn.
21973 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21974 Alan Hayward <alan.hayward@arm.com>
21975 David Sherwood <david.sherwood@arm.com>
21977 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
21978 (reduc_xor_scal_optab): New optabs.
21979 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
21980 (reduc_xor_scal_@var{m}): Document.
21981 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
21982 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
21983 internal functions.
21984 * fold-const-call.c (fold_const_call): Handle them.
21985 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
21986 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
21987 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
21988 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
21989 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
21990 (UNSPEC_XORV): New unspecs.
21991 (optab): Add entries for them.
21992 (BITWISEV): New int iterator.
21993 (bit_reduc_op): New int attributes.
21995 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21996 Alan Hayward <alan.hayward@arm.com>
21997 David Sherwood <david.sherwood@arm.com>
21999 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
22000 * internal-fn.def (VEC_SHL_INSERT): New internal function.
22001 * optabs.def (vec_shl_insert_optab): New optab.
22002 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
22003 (duplicate_and_interleave): Likewise.
22004 * tree-vect-loop.c: Include internal-fn.h.
22005 (neutral_op_for_slp_reduction): New function, split out from
22006 get_initial_defs_for_reduction.
22007 (get_initial_def_for_reduction): Handle option 2 for variable-length
22008 vectors by loading the neutral value into a vector and then shifting
22009 the initial value into element 0.
22010 (get_initial_defs_for_reduction): Replace the code argument with
22011 the neutral value calculated by neutral_op_for_slp_reduction.
22012 Use gimple_build_vector for constant-length vectors.
22013 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
22014 but the first group_size elements have a neutral value.
22015 Use duplicate_and_interleave otherwise.
22016 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
22017 Update call to get_initial_defs_for_reduction. Handle SLP
22018 reductions for variable-length vectors by creating one vector
22019 result for each scalar result, with the elements associated
22020 with other scalar results stubbed out with the neutral value.
22021 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
22022 Require IFN_VEC_SHL_INSERT for double reductions on
22023 variable-length vectors, or SLP reductions that have
22024 a neutral value. Require can_duplicate_and_interleave_p
22025 support for variable-length unchained SLP reductions if there
22026 is no neutral value, such as for MIN/MAX reductions. Also require
22027 the number of vector elements to be a multiple of the number of
22028 SLP statements when doing variable-length unchained SLP reductions.
22029 Update call to vect_create_epilog_for_reduction.
22030 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
22031 and remove initial values.
22032 (duplicate_and_interleave): Make public.
22033 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
22034 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
22036 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22037 Alan Hayward <alan.hayward@arm.com>
22038 David Sherwood <david.sherwood@arm.com>
22040 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
22041 (can_duplicate_and_interleave_p): New function.
22042 (vect_get_and_check_slp_defs): Take the vector of statements
22043 rather than just the current one. Remove excess parentheses.
22044 Restriction rejectinon of vect_constant_def and vect_external_def
22045 for variable-length vectors to boolean types, or types for which
22046 can_duplicate_and_interleave_p is false.
22047 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
22048 (duplicate_and_interleave): New function.
22049 (vect_get_constant_vectors): Use gimple_build_vector for
22050 constant-length vectors and suitable variable-length constant
22051 vectors. Use duplicate_and_interleave for other variable-length
22052 vectors. Don't defer the update when inserting new statements.
22054 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22055 Alan Hayward <alan.hayward@arm.com>
22056 David Sherwood <david.sherwood@arm.com>
22058 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
22059 min_profitable_iters doesn't go negative.
22061 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22062 Alan Hayward <alan.hayward@arm.com>
22063 David Sherwood <david.sherwood@arm.com>
22065 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
22066 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
22067 * optabs.def (vec_mask_load_lanes_optab): New optab.
22068 (vec_mask_store_lanes_optab): Likewise.
22069 * internal-fn.def (MASK_LOAD_LANES): New internal function.
22070 (MASK_STORE_LANES): Likewise.
22071 * internal-fn.c (mask_load_lanes_direct): New macro.
22072 (mask_store_lanes_direct): Likewise.
22073 (expand_mask_load_optab_fn): Handle masked operations.
22074 (expand_mask_load_lanes_optab_fn): New macro.
22075 (expand_mask_store_optab_fn): Handle masked operations.
22076 (expand_mask_store_lanes_optab_fn): New macro.
22077 (direct_mask_load_lanes_optab_supported_p): Likewise.
22078 (direct_mask_store_lanes_optab_supported_p): Likewise.
22079 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
22081 (vect_load_lanes_supported): Likewise.
22082 * tree-vect-data-refs.c (strip_conversion): New function.
22083 (can_group_stmts_p): Likewise.
22084 (vect_analyze_data_ref_accesses): Use it instead of checking
22085 for a pair of assignments.
22086 (vect_store_lanes_supported): Take a masked_p parameter.
22087 (vect_load_lanes_supported): Likewise.
22088 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
22089 vect_store_lanes_supported and vect_load_lanes_supported.
22090 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
22091 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
22092 parameter. Don't allow gaps for masked accesses.
22093 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
22094 and vect_load_lanes_supported.
22095 (get_load_store_type): Take a masked_p parameter and update
22096 call to get_group_load_store_type.
22097 (vectorizable_store): Update call to get_load_store_type.
22098 Handle IFN_MASK_STORE_LANES.
22099 (vectorizable_load): Update call to get_load_store_type.
22100 Handle IFN_MASK_LOAD_LANES.
22102 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22103 Alan Hayward <alan.hayward@arm.com>
22104 David Sherwood <david.sherwood@arm.com>
22106 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
22108 * config/aarch64/aarch64-protos.h
22109 (aarch64_sve_struct_memory_operand_p): Declare.
22110 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
22111 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
22112 (VPRED, vpred): Handle SVE structure modes.
22113 * config/aarch64/constraints.md (Utx): New constraint.
22114 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
22115 (aarch64_sve_struct_nonimmediate_operand): New predicates.
22116 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
22117 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
22118 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
22119 structure modes. Split into pieces after RA.
22120 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
22121 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
22123 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
22124 SVE structure modes.
22125 (aarch64_classify_address): Likewise.
22126 (sizetochar): Move earlier in file.
22127 (aarch64_print_operand): Handle SVE register lists.
22128 (aarch64_array_mode): New function.
22129 (aarch64_sve_struct_memory_operand_p): Likewise.
22130 (TARGET_ARRAY_MODE): Redefine.
22132 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22133 Alan Hayward <alan.hayward@arm.com>
22134 David Sherwood <david.sherwood@arm.com>
22136 * target.def (array_mode): New target hook.
22137 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
22138 * doc/tm.texi: Regenerate.
22139 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
22140 * hooks.c (hook_optmode_mode_uhwi_none): New function.
22141 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
22142 targetm.array_mode.
22143 * stor-layout.c (mode_for_array): Likewise. Support polynomial
22146 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22147 Alan Hayward <alan.hayward@arm.com>
22148 David Sherwood <david.sherwood@arm.com>
22150 * fold-const.c (fold_binary_loc): Check the argument types
22151 rather than the result type when testing for a vector operation.
22153 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22155 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
22156 * doc/tm.texi: Regenerate.
22158 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22159 Alan Hayward <alan.hayward@arm.com>
22160 David Sherwood <david.sherwood@arm.com>
22162 * doc/invoke.texi (-msve-vector-bits=): Document new option.
22163 (sve): Document new AArch64 extension.
22164 * doc/md.texi (w): Extend the description of the AArch64
22165 constraint to include SVE vectors.
22166 (Upl, Upa): Document new AArch64 predicate constraints.
22167 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
22169 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
22170 (msve-vector-bits=): New option.
22171 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
22172 SVE when these are disabled.
22173 (sve): New extension.
22174 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
22175 modes. Adjust their number of units based on aarch64_sve_vg.
22176 (MAX_BITSIZE_MODE_ANY_MODE): Define.
22177 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
22178 aarch64_addr_query_type.
22179 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
22180 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
22181 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
22182 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
22183 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
22184 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
22185 (aarch64_simd_imm_zero_p): Delete.
22186 (aarch64_check_zero_based_sve_index_immediate): Declare.
22187 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22188 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22189 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22190 (aarch64_sve_float_mul_immediate_p): Likewise.
22191 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22192 rather than an rtx.
22193 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
22194 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
22195 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
22196 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
22197 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
22198 (aarch64_regmode_natural_size): Likewise.
22199 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
22200 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
22202 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
22203 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
22204 for VG and the SVE predicate registers.
22205 (V_ALIASES): Add a "z"-prefixed alias.
22206 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
22207 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
22208 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
22209 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
22210 (REG_CLASS_NAMES): Add entries for them.
22211 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
22212 and the predicate registers.
22213 (aarch64_sve_vg): Declare.
22214 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
22215 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
22216 (REGMODE_NATURAL_SIZE): Define.
22217 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
22219 * config/aarch64/aarch64.c: Include cfgrtl.h.
22220 (simd_immediate_info): Add a constructor for series vectors,
22221 and an associated step field.
22222 (aarch64_sve_vg): New variable.
22223 (aarch64_dbx_register_number): Handle VG and the predicate registers.
22224 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
22225 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
22226 (VEC_ANY_DATA, VEC_STRUCT): New constants.
22227 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
22228 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
22229 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
22230 (aarch64_get_mask_mode): New functions.
22231 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
22232 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
22233 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
22234 predicate modes and predicate registers. Explicitly restrict
22235 GPRs to modes of 16 bytes or smaller. Only allow FP registers
22236 to store a vector mode if it is recognized by
22237 aarch64_classify_vector_mode.
22238 (aarch64_regmode_natural_size): New function.
22239 (aarch64_hard_regno_caller_save_mode): Return the original mode
22241 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
22242 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
22243 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
22244 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
22246 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
22247 does not overlap dest if the function is frame-related. Handle
22249 (aarch64_split_add_offset): New function.
22250 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
22251 them aarch64_add_offset.
22252 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
22253 and update call to aarch64_sub_sp.
22254 (aarch64_add_cfa_expression): New function.
22255 (aarch64_expand_prologue): Pass extra temporary registers to the
22256 functions above. Handle the case in which we need to emit new
22257 DW_CFA_expressions for registers that were originally saved
22258 relative to the stack pointer, but now have to be expressed
22259 relative to the frame pointer.
22260 (aarch64_output_mi_thunk): Pass extra temporary registers to the
22262 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
22263 IP0 and IP1 values for SVE frames.
22264 (aarch64_expand_vec_series): New function.
22265 (aarch64_expand_sve_widened_duplicate): Likewise.
22266 (aarch64_expand_sve_const_vector): Likewise.
22267 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
22268 Handle SVE constants. Use emit_move_insn to move a force_const_mem
22269 into the register, rather than emitting a SET directly.
22270 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
22271 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
22272 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
22273 (offset_9bit_signed_scaled_p): New functions.
22274 (aarch64_replicate_bitmask_imm): New function.
22275 (aarch64_bitmask_imm): Use it.
22276 (aarch64_cannot_force_const_mem): Reject expressions involving
22277 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
22278 (aarch64_classify_index): Handle SVE indices, by requiring
22279 a plain register index with a scale that matches the element size.
22280 (aarch64_classify_address): Handle SVE addresses. Assert that
22281 the mode of the address is VOIDmode or an integer mode.
22282 Update call to aarch64_classify_symbol.
22283 (aarch64_classify_symbolic_expression): Update call to
22284 aarch64_classify_symbol.
22285 (aarch64_const_vec_all_in_range_p): New function.
22286 (aarch64_print_vector_float_operand): Likewise.
22287 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
22288 "vN" for FP registers with SVE modes. Handle (const ...) vectors
22289 and the FP immediates 1.0 and 0.5.
22290 (aarch64_print_address_internal): Handle SVE addresses.
22291 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
22292 (aarch64_regno_regclass): Handle predicate registers.
22293 (aarch64_secondary_reload): Handle big-endian reloads of SVE
22295 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
22296 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
22297 (aarch64_convert_sve_vector_bits): New function.
22298 (aarch64_override_options): Use it to handle -msve-vector-bits=.
22299 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22300 rather than an rtx.
22301 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
22302 Handle SVE vector and predicate modes. Accept VL-based constants
22303 that need only one temporary register, and VL offsets that require
22304 no temporary registers.
22305 (aarch64_conditional_register_usage): Mark the predicate registers
22306 as fixed if SVE isn't available.
22307 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
22308 Return true for SVE vector and predicate modes.
22309 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
22310 rather than an unsigned int. Handle SVE modes.
22311 (aarch64_preferred_simd_mode): Update call accordingly. Handle
22313 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
22315 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22316 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22317 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22318 (aarch64_sve_float_mul_immediate_p): New functions.
22319 (aarch64_sve_valid_immediate): New function.
22320 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
22321 Explicitly reject structure modes. Check for INDEX constants.
22322 Handle PTRUE and PFALSE constants.
22323 (aarch64_check_zero_based_sve_index_immediate): New function.
22324 (aarch64_simd_imm_zero_p): Delete.
22325 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
22326 vector modes. Accept constants in the range of CNT[BHWD].
22327 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
22328 ask for an Advanced SIMD mode.
22329 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
22330 (aarch64_simd_vector_alignment): Handle SVE predicates.
22331 (aarch64_vectorize_preferred_vector_alignment): New function.
22332 (aarch64_simd_vector_alignment_reachable): Use it instead of
22334 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
22335 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
22337 (MAX_VECT_LEN): Delete.
22338 (expand_vec_perm_d): Add a vec_flags field.
22339 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
22340 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
22341 (aarch64_evpc_ext): Don't apply a big-endian lane correction
22343 (aarch64_evpc_rev): Rename to...
22344 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
22345 (aarch64_evpc_rev_global): New function.
22346 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
22347 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
22349 (aarch64_evpc_sve_tbl): New function.
22350 (aarch64_expand_vec_perm_const_1): Update after rename of
22351 aarch64_evpc_rev. Handle SVE permutes too, trying
22352 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
22353 than aarch64_evpc_tbl.
22354 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
22355 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
22356 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
22357 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
22358 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
22359 (aarch64_expand_sve_vcond): New functions.
22360 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
22361 of aarch64_vector_mode_p.
22362 (aarch64_dwarf_poly_indeterminate_value): New function.
22363 (aarch64_compute_pressure_classes): Likewise.
22364 (aarch64_can_change_mode_class): Likewise.
22365 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
22366 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
22367 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
22368 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
22369 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
22370 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
22371 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
22372 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
22374 (Dn, Dl, Dr): Accept const as well as const_vector.
22375 (Dz): Likewise. Compare against CONST0_RTX.
22376 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
22377 of "vector" where appropriate.
22378 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
22379 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
22380 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
22381 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
22382 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
22383 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
22384 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
22385 (v_int_equiv): Extend to SVE modes.
22386 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
22388 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
22389 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
22390 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
22391 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
22392 (SVE_COND_FP_CMP): New int iterators.
22393 (perm_hilo): Handle the new unpack unspecs.
22394 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
22396 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
22397 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
22398 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
22399 (aarch64_equality_operator, aarch64_constant_vector_operand)
22400 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
22401 (aarch64_sve_nonimmediate_operand): Likewise.
22402 (aarch64_sve_general_operand): Likewise.
22403 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
22404 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
22405 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
22406 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
22407 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
22408 (aarch64_sve_float_arith_immediate): Likewise.
22409 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
22410 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
22411 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
22412 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
22413 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
22414 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
22415 (aarch64_sve_float_arith_operand): Likewise.
22416 (aarch64_sve_float_arith_with_sub_operand): Likewise.
22417 (aarch64_sve_float_mul_operand): Likewise.
22418 (aarch64_sve_vec_perm_operand): Likewise.
22419 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
22420 (aarch64_mov_operand): Accept const_poly_int and const_vector.
22421 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
22422 as well as const_vector.
22423 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
22424 in file. Use CONST0_RTX and CONSTM1_RTX.
22425 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
22426 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
22427 Use aarch64_simd_imm_zero.
22428 * config/aarch64/aarch64-sve.md: New file.
22429 * config/aarch64/aarch64.md: Include it.
22430 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
22431 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
22432 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
22433 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
22434 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
22435 (sve): New attribute.
22436 (enabled): Disable instructions with the sve attribute unless
22438 (movqi, movhi): Pass CONST_POLY_INT operaneds through
22439 aarch64_expand_mov_immediate.
22440 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
22441 CNT[BHSD] immediates.
22442 (movti): Split CONST_POLY_INT moves into two halves.
22443 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
22444 Split additions that need a temporary here if the destination
22445 is the stack pointer.
22446 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
22447 (*add<mode>3_poly_1): New instruction.
22448 (set_clobber_cc): New expander.
22450 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22452 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
22453 parameter and use it instead of GET_MODE_SIZE (innermode). Use
22454 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
22455 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
22456 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
22457 Change innermode from fixed_mode_size to machine_mode.
22458 (simplify_subreg): Update call accordingly. Handle a constant-sized
22459 subreg of a variable-length CONST_VECTOR.
22461 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22462 Alan Hayward <alan.hayward@arm.com>
22463 David Sherwood <david.sherwood@arm.com>
22465 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
22466 (add_offset_to_base): New function, split out from...
22467 (create_mem_ref): ...here. When handling a scale other than 1,
22468 check first whether the address is valid without the offset.
22469 Add it into the base if so, leaving the index and scale as-is.
22471 2018-01-12 Jakub Jelinek <jakub@redhat.com>
22474 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
22475 fold_for_warn before checking if arg2 is INTEGER_CST.
22477 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
22479 * config/rs6000/predicates.md (load_multiple_operation): Delete.
22480 (store_multiple_operation): Delete.
22481 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
22482 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
22483 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
22484 guarded by TARGET_STRING.
22485 (rs6000_output_load_multiple): Delete.
22486 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
22487 OPTION_MASK_STRING / TARGET_STRING handling.
22488 (print_operand) <'N', 'O'>: Add comment that these are unused now.
22489 (const rs6000_opt_masks) <"string">: Change mask to 0.
22490 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
22491 (MASK_STRING): Delete.
22492 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
22494 (load_multiple): Delete.
22501 (store_multiple): Delete.
22508 (movmemsi_8reg): Delete.
22509 (corresponding unnamed define_insn): Delete.
22510 (movmemsi_6reg): Delete.
22511 (corresponding unnamed define_insn): Delete.
22512 (movmemsi_4reg): Delete.
22513 (corresponding unnamed define_insn): Delete.
22514 (movmemsi_2reg): Delete.
22515 (corresponding unnamed define_insn): Delete.
22516 (movmemsi_1reg): Delete.
22517 (corresponding unnamed define_insn): Delete.
22518 * config/rs6000/rs6000.opt (mno-string): New.
22519 (mstring): Replace by deprecation warning stub.
22520 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
22522 2018-01-12 Jakub Jelinek <jakub@redhat.com>
22524 * regrename.c (regrename_do_replace): If replacing the same
22525 reg multiple times, try to reuse last created gen_raw_REG.
22528 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
22529 main to workaround a bug in GDB.
22531 2018-01-12 Tom de Vries <tom@codesourcery.com>
22534 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
22536 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
22538 PR rtl-optimization/80481
22539 * ira-color.c (get_cap_member): New function.
22540 (allocnos_conflict_by_live_ranges_p): Use it.
22541 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
22542 (setup_slot_coalesced_allocno_live_ranges): Ditto.
22544 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
22547 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
22548 (*saddl_se_1): Ditto.
22549 (*ssubsi_1): Ditto.
22550 (*ssubl_se_1): Ditto.
22552 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
22554 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
22555 rather than wi::to_widest for DR_INITs.
22556 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
22557 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
22558 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
22560 (vect_analyze_group_access_1): Note that here.
22562 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
22564 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
22565 polynomial type sizes.
22567 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
22569 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
22570 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
22571 (gimple_add_tmp_var): Likewise.
22573 2018-01-12 Martin Liska <mliska@suse.cz>
22575 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
22576 (gimple_alloc_sizes): Likewise.
22577 (dump_gimple_statistics): Use PRIu64 in printf format.
22578 * gimple.h: Change uint64_t to int.
22580 2018-01-12 Martin Liska <mliska@suse.cz>
22582 * tree-core.h: Use uint64_t instead of int.
22583 * tree.c (tree_node_counts): Likewise.
22584 (tree_node_sizes): Likewise.
22585 (dump_tree_statistics): Use PRIu64 in printf format.
22587 2018-01-12 Martin Liska <mliska@suse.cz>
22589 * Makefile.in: As qsort_chk is implemented in vec.c, add
22590 vec.o to linkage of gencfn-macros.
22591 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
22592 passing the info to record_node_allocation_statistics.
22593 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
22595 * ggc-common.c (struct ggc_usage): Add operator== and use
22596 it in operator< and compare function.
22597 * mem-stats.h (struct mem_usage): Likewise.
22598 * vec.c (struct vec_usage): Remove operator< and compare
22599 function. Can be simply inherited.
22601 2018-01-12 Martin Jambor <mjambor@suse.cz>
22604 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
22605 * tree-ssa-math-opts.c: Include domwalk.h.
22606 (convert_mult_to_fma_1): New function.
22607 (fma_transformation_info): New type.
22608 (fma_deferring_state): Likewise.
22609 (cancel_fma_deferring): New function.
22610 (result_of_phi): Likewise.
22611 (last_fma_candidate_feeds_initial_phi): Likewise.
22612 (convert_mult_to_fma): Added deferring logic, split actual
22613 transformation to convert_mult_to_fma_1.
22614 (math_opts_dom_walker): New type.
22615 (math_opts_dom_walker::after_dom_children): New method, body moved
22616 here from pass_optimize_widening_mul::execute, added deferring logic
22618 (pass_optimize_widening_mul::execute): Moved most of code to
22619 math_opts_dom_walker::after_dom_children.
22620 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
22621 * config/i386/i386.c (ix86_option_override_internal): Added
22622 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
22624 2018-01-12 Richard Biener <rguenther@suse.de>
22627 * dwarf2out.c (gen_variable_die): Do not reset old_die for
22628 inline instance vars.
22630 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
22633 * config/rx/rx.c (rx_is_restricted_memory_address):
22634 Handle SUBREG case.
22636 2018-01-12 Richard Biener <rguenther@suse.de>
22638 PR tree-optimization/80846
22639 * target.def (split_reduction): New target hook.
22640 * targhooks.c (default_split_reduction): New function.
22641 * targhooks.h (default_split_reduction): Declare.
22642 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
22643 target requests first reduce vectors by combining low and high
22645 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
22646 (get_vectype_for_scalar_type_and_size): Export.
22647 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
22648 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
22649 * doc/tm.texi: Regenerate.
22650 * config/i386/i386.c (ix86_split_reduction): Implement
22651 TARGET_VECTORIZE_SPLIT_REDUCTION.
22653 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
22656 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
22657 in PIC mode except for TARGET_VXWORKS_RTP.
22658 * config/sparc/sparc.c: Include cfgrtl.h.
22659 (TARGET_INIT_PIC_REG): Define.
22660 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
22661 (sparc_pic_register_p): New predicate.
22662 (sparc_legitimate_address_p): Use it.
22663 (sparc_legitimize_pic_address): Likewise.
22664 (sparc_delegitimize_address): Likewise.
22665 (sparc_mode_dependent_address_p): Likewise.
22666 (gen_load_pcrel_sym): Remove 4th parameter.
22667 (load_got_register): Adjust call to above. Remove obsolete stuff.
22668 (sparc_expand_prologue): Do not call load_got_register here.
22669 (sparc_flat_expand_prologue): Likewise.
22670 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
22671 (sparc_use_pseudo_pic_reg): New function.
22672 (sparc_init_pic_reg): Likewise.
22673 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
22674 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
22676 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
22678 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
22679 Add item for branch_cost.
22681 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
22683 PR rtl-optimization/83565
22684 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
22685 not extend the result to a larger mode for rotate operations.
22686 (num_sign_bit_copies1): Likewise.
22688 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22691 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
22693 Use values-Xc.o for -pedantic.
22694 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
22696 2018-01-12 Martin Liska <mliska@suse.cz>
22699 * ipa-devirt.c (final_warning_record::grow_type_warnings):
22701 (possible_polymorphic_call_targets): Use it.
22702 (ipa_devirt): Likewise.
22704 2018-01-12 Martin Liska <mliska@suse.cz>
22706 * profile-count.h (enum profile_quality): Use 0 as invalid
22707 enum value of profile_quality.
22709 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
22711 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
22712 -mext-string options.
22714 2018-01-12 Richard Biener <rguenther@suse.de>
22716 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
22717 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
22718 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
22720 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
22722 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
22724 * configure.ac (--with-long-double-format): Add support for the
22725 configuration option to change the default long double format on
22727 * config.gcc (powerpc*-linux*-*): Likewise.
22728 * configure: Regenerate.
22729 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
22730 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
22731 used without modification.
22733 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22735 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
22736 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
22737 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
22738 MISC_BUILTIN_SPEC_BARRIER.
22739 (rs6000_init_builtins): Likewise.
22740 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
22742 (speculation_barrier): New define_insn.
22743 * doc/extend.texi: Document __builtin_speculation_barrier.
22745 2018-01-11 Jakub Jelinek <jakub@redhat.com>
22748 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
22749 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
22750 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
22752 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
22753 integral modes instead of "ss" and "sd".
22754 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
22755 vectors with 32-bit and 64-bit elements.
22756 (vecdupssescalarmodesuffix): New mode attribute.
22757 (vec_dup<mode>): Use it.
22759 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
22762 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
22763 frame if argument is passed on stack.
22765 2018-01-11 Jakub Jelinek <jakub@redhat.com>
22768 * ree.c (combine_reaching_defs): Optimize also
22769 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
22770 reg2=any_extend(exp); reg1=reg2;, formatting fix.
22772 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
22774 PR middle-end/83189
22775 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
22777 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
22779 PR middle-end/83718
22780 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
22781 after they are computed.
22783 2018-01-11 Bin Cheng <bin.cheng@arm.com>
22785 PR tree-optimization/83695
22786 * gimple-loop-linterchange.cc
22787 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
22788 reset cached scev information after interchange.
22789 (pass_linterchange::execute): Remove call to scev_reset_htab.
22791 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22793 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
22794 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
22795 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
22796 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
22797 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
22798 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
22799 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
22800 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
22801 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
22802 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
22803 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
22804 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
22805 (V_lane_reg): Likewise.
22806 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
22808 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
22809 (vfmal_lane_low<mode>_intrinsic,
22810 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
22811 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
22812 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
22813 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
22814 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
22815 vfmsl_lane_high<mode>_intrinsic): New define_insns.
22817 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22819 * config/arm/arm-cpus.in (fp16fml): New feature.
22820 (ALL_SIMD): Add fp16fml.
22821 (armv8.2-a): Add fp16fml as an option.
22822 (armv8.3-a): Likewise.
22823 (armv8.4-a): Add fp16fml as part of fp16.
22824 * config/arm/arm.h (TARGET_FP16FML): Define.
22825 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
22827 * config/arm/arm-modes.def (V2HF): Define.
22828 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
22829 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
22830 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
22831 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
22832 vfmsl_low, vfmsl_high): New set of builtins.
22833 * config/arm/iterators.md (PLUSMINUS): New code iterator.
22834 (vfml_op): New code attribute.
22835 (VFMLHALVES): New int iterator.
22836 (VFML, VFMLSEL): New mode attributes.
22837 (V_reg): Define mapping for V2HF.
22838 (V_hi, V_lo): New mode attributes.
22839 (VF_constraint): Likewise.
22840 (vfml_half, vfml_half_selector): New int attributes.
22841 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
22843 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
22844 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
22846 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
22847 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
22848 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
22849 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
22851 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
22852 Document new effective target and option set.
22854 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22856 * config/arm/arm-cpus.in (armv8_4): New feature.
22857 (ARMv8_4a): New fgroup.
22858 (armv8.4-a): New arch.
22859 * config/arm/arm-tables.opt: Regenerate.
22860 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
22861 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
22862 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
22863 Add matching rules for -march=armv8.4-a and extensions.
22864 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
22866 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
22869 * config/rx/rx.md (BW): New mode attribute.
22870 (sync_lock_test_and_setsi): Add mode suffix to insn output.
22872 2018-01-11 Richard Biener <rguenther@suse.de>
22874 PR tree-optimization/83435
22875 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
22876 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
22877 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
22879 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22880 Alan Hayward <alan.hayward@arm.com>
22881 David Sherwood <david.sherwood@arm.com>
22883 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
22885 (aarch64_classify_address): Initialize it. Track polynomial offsets.
22886 (aarch64_print_address_internal): Use it to check for a zero offset.
22888 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22889 Alan Hayward <alan.hayward@arm.com>
22890 David Sherwood <david.sherwood@arm.com>
22892 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
22893 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
22894 Return a poly_int64 rather than a HOST_WIDE_INT.
22895 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
22896 rather than a HOST_WIDE_INT.
22897 * config/aarch64/aarch64.h (aarch64_frame): Protect with
22898 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
22899 hard_fp_offset, frame_size, initial_adjust, callee_offset and
22900 final_offset from HOST_WIDE_INT to poly_int64.
22901 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
22902 to_constant when getting the number of units in an Advanced SIMD
22904 (aarch64_builtin_vectorized_function): Check for a constant number
22906 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
22908 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
22909 attribute instead of GET_MODE_NUNITS.
22910 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
22911 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
22912 GET_MODE_SIZE for fixed-size registers.
22913 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
22914 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
22915 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
22916 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
22917 (aarch64_print_operand, aarch64_print_address_internal)
22918 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
22919 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
22920 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
22921 Handle polynomial GET_MODE_SIZE.
22922 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
22923 wider than SImode without modification.
22924 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
22925 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
22926 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
22927 passing and returning SVE modes.
22928 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
22929 rather than GEN_INT.
22930 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
22931 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
22932 (aarch64_allocate_and_probe_stack_space): Likewise.
22933 (aarch64_layout_frame): Cope with polynomial offsets.
22934 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
22935 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
22936 polynomial offsets.
22937 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
22938 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
22939 poly_int64 rather than a HOST_WIDE_INT.
22940 (aarch64_get_separate_components, aarch64_process_components)
22941 (aarch64_expand_prologue, aarch64_expand_epilogue)
22942 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
22943 (aarch64_anchor_offset): New function, split out from...
22944 (aarch64_legitimize_address): ...here.
22945 (aarch64_builtin_vectorization_cost): Handle polynomial
22946 TYPE_VECTOR_SUBPARTS.
22947 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
22949 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
22950 number of elements from the PARALLEL rather than the mode.
22951 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
22952 rather than GET_MODE_BITSIZE.
22953 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
22954 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
22955 (aarch64_expand_vec_perm_const_1): Handle polynomial
22956 d->perm.length () and d->perm elements.
22957 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
22958 Apply to_constant to d->perm elements.
22959 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
22960 polynomial CONST_VECTOR_NUNITS.
22961 (aarch64_move_pointer): Take amount as a poly_int64 rather
22963 (aarch64_progress_pointer): Avoid temporary variable.
22964 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
22965 the mode attribute instead of GET_MODE.
22967 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22968 Alan Hayward <alan.hayward@arm.com>
22969 David Sherwood <david.sherwood@arm.com>
22971 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
22972 x exists before using it.
22973 (aarch64_add_constant_internal): Rename to...
22974 (aarch64_add_offset_1): ...this. Replace regnum with separate
22975 src and dest rtxes. Handle the case in which they're different,
22976 including when the offset is zero. Replace scratchreg with an rtx.
22977 Use 2 additions if there is no spare register into which we can
22978 move a 16-bit constant.
22979 (aarch64_add_constant): Delete.
22980 (aarch64_add_offset): Replace reg with separate src and dest
22981 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
22982 Use aarch64_add_offset_1.
22983 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
22984 an rtx rather than an int. Take the delta as a poly_int64
22985 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
22986 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
22987 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
22988 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
22989 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
22990 and aarch64_add_sp.
22991 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
22992 aarch64_add_constant.
22994 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22996 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22997 Use scalar_float_mode.
22999 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23001 * config/aarch64/aarch64-simd.md
23002 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
23003 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
23004 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
23005 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
23006 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
23007 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
23008 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
23009 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
23010 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
23011 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
23013 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23016 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
23017 targ_options->x_arm_arch_string is non NULL.
23019 2018-01-11 Tamar Christina <tamar.christina@arm.com>
23021 * config/aarch64/aarch64.h
23022 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
23024 2018-01-11 Sudakshina Das <sudi.das@arm.com>
23027 * expmed.c (emit_store_flag_force): Swap if const op0
23028 and change VOIDmode to mode of op0.
23030 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23032 PR rtl-optimization/83761
23033 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
23034 than bytes to mode_for_size.
23036 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23038 PR middle-end/83189
23039 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
23040 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
23043 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23045 PR middle-end/83575
23046 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
23047 when in layout mode.
23048 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
23049 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
23052 2018-01-10 Michael Collison <michael.collison@arm.com>
23054 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
23055 * config/aarch64/aarch64-option-extension.def: Add
23056 AARCH64_OPT_EXTENSION of 'fp16fml'.
23057 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23058 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
23059 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
23060 * config/aarch64/constraints.md (Ui7): New constraint.
23061 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
23062 (VFMLA_SEL_W): Ditto.
23065 (VFMLA16_LOW): New int iterator.
23066 (VFMLA16_HIGH): Ditto.
23067 (UNSPEC_FMLAL): New unspec.
23068 (UNSPEC_FMLSL): Ditto.
23069 (UNSPEC_FMLAL2): Ditto.
23070 (UNSPEC_FMLSL2): Ditto.
23071 (f16mac): New code attribute.
23072 * config/aarch64/aarch64-simd-builtins.def
23073 (aarch64_fmlal_lowv2sf): Ditto.
23074 (aarch64_fmlsl_lowv2sf): Ditto.
23075 (aarch64_fmlalq_lowv4sf): Ditto.
23076 (aarch64_fmlslq_lowv4sf): Ditto.
23077 (aarch64_fmlal_highv2sf): Ditto.
23078 (aarch64_fmlsl_highv2sf): Ditto.
23079 (aarch64_fmlalq_highv4sf): Ditto.
23080 (aarch64_fmlslq_highv4sf): Ditto.
23081 (aarch64_fmlal_lane_lowv2sf): Ditto.
23082 (aarch64_fmlsl_lane_lowv2sf): Ditto.
23083 (aarch64_fmlal_laneq_lowv2sf): Ditto.
23084 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
23085 (aarch64_fmlalq_lane_lowv4sf): Ditto.
23086 (aarch64_fmlsl_lane_lowv4sf): Ditto.
23087 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
23088 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
23089 (aarch64_fmlal_lane_highv2sf): Ditto.
23090 (aarch64_fmlsl_lane_highv2sf): Ditto.
23091 (aarch64_fmlal_laneq_highv2sf): Ditto.
23092 (aarch64_fmlsl_laneq_highv2sf): Ditto.
23093 (aarch64_fmlalq_lane_highv4sf): Ditto.
23094 (aarch64_fmlsl_lane_highv4sf): Ditto.
23095 (aarch64_fmlalq_laneq_highv4sf): Ditto.
23096 (aarch64_fmlsl_laneq_highv4sf): Ditto.
23097 * config/aarch64/aarch64-simd.md:
23098 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
23099 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23100 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
23101 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23102 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
23103 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
23104 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
23105 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
23106 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
23107 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
23108 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
23109 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
23110 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
23111 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
23112 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
23113 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
23114 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
23115 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
23116 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
23117 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
23118 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
23119 (vfmlsl_low_u32): Ditto.
23120 (vfmlalq_low_u32): Ditto.
23121 (vfmlslq_low_u32): Ditto.
23122 (vfmlal_high_u32): Ditto.
23123 (vfmlsl_high_u32): Ditto.
23124 (vfmlalq_high_u32): Ditto.
23125 (vfmlslq_high_u32): Ditto.
23126 (vfmlal_lane_low_u32): Ditto.
23127 (vfmlsl_lane_low_u32): Ditto.
23128 (vfmlal_laneq_low_u32): Ditto.
23129 (vfmlsl_laneq_low_u32): Ditto.
23130 (vfmlalq_lane_low_u32): Ditto.
23131 (vfmlslq_lane_low_u32): Ditto.
23132 (vfmlalq_laneq_low_u32): Ditto.
23133 (vfmlslq_laneq_low_u32): Ditto.
23134 (vfmlal_lane_high_u32): Ditto.
23135 (vfmlsl_lane_high_u32): Ditto.
23136 (vfmlal_laneq_high_u32): Ditto.
23137 (vfmlsl_laneq_high_u32): Ditto.
23138 (vfmlalq_lane_high_u32): Ditto.
23139 (vfmlslq_lane_high_u32): Ditto.
23140 (vfmlalq_laneq_high_u32): Ditto.
23141 (vfmlslq_laneq_high_u32): Ditto.
23142 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
23143 (AARCH64_FL_FOR_ARCH8_4): New.
23144 (AARCH64_ISA_F16FML): New ISA flag.
23145 (TARGET_F16FML): New feature flag for fp16fml.
23146 (doc/invoke.texi): Document new fp16fml option.
23148 2018-01-10 Michael Collison <michael.collison@arm.com>
23150 * config/aarch64/aarch64-builtins.c:
23151 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
23152 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23153 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
23154 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
23155 (AARCH64_ISA_SHA3): New ISA flag.
23156 (TARGET_SHA3): New feature flag for sha3.
23157 * config/aarch64/iterators.md (sha512_op): New int attribute.
23158 (CRYPTO_SHA512): New int iterator.
23159 (UNSPEC_SHA512H): New unspec.
23160 (UNSPEC_SHA512H2): Ditto.
23161 (UNSPEC_SHA512SU0): Ditto.
23162 (UNSPEC_SHA512SU1): Ditto.
23163 * config/aarch64/aarch64-simd-builtins.def
23164 (aarch64_crypto_sha512hqv2di): New builtin.
23165 (aarch64_crypto_sha512h2qv2di): Ditto.
23166 (aarch64_crypto_sha512su0qv2di): Ditto.
23167 (aarch64_crypto_sha512su1qv2di): Ditto.
23168 (aarch64_eor3qv8hi): Ditto.
23169 (aarch64_rax1qv2di): Ditto.
23170 (aarch64_xarqv2di): Ditto.
23171 (aarch64_bcaxqv8hi): Ditto.
23172 * config/aarch64/aarch64-simd.md:
23173 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
23174 (aarch64_crypto_sha512su0qv2di): Ditto.
23175 (aarch64_crypto_sha512su1qv2di): Ditto.
23176 (aarch64_eor3qv8hi): Ditto.
23177 (aarch64_rax1qv2di): Ditto.
23178 (aarch64_xarqv2di): Ditto.
23179 (aarch64_bcaxqv8hi): Ditto.
23180 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
23181 (vsha512h2q_u64): Ditto.
23182 (vsha512su0q_u64): Ditto.
23183 (vsha512su1q_u64): Ditto.
23184 (veor3q_u16): Ditto.
23185 (vrax1q_u64): Ditto.
23186 (vxarq_u64): Ditto.
23187 (vbcaxq_u16): Ditto.
23188 * config/arm/types.md (crypto_sha512): New type attribute.
23189 (crypto_sha3): Ditto.
23190 (doc/invoke.texi): Document new sha3 option.
23192 2018-01-10 Michael Collison <michael.collison@arm.com>
23194 * config/aarch64/aarch64-builtins.c:
23195 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
23196 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23197 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
23198 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
23199 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
23200 (AARCH64_ISA_SM4): New ISA flag.
23201 (TARGET_SM4): New feature flag for sm4.
23202 * config/aarch64/aarch64-simd-builtins.def
23203 (aarch64_sm3ss1qv4si): Ditto.
23204 (aarch64_sm3tt1aq4si): Ditto.
23205 (aarch64_sm3tt1bq4si): Ditto.
23206 (aarch64_sm3tt2aq4si): Ditto.
23207 (aarch64_sm3tt2bq4si): Ditto.
23208 (aarch64_sm3partw1qv4si): Ditto.
23209 (aarch64_sm3partw2qv4si): Ditto.
23210 (aarch64_sm4eqv4si): Ditto.
23211 (aarch64_sm4ekeyqv4si): Ditto.
23212 * config/aarch64/aarch64-simd.md:
23213 (aarch64_sm3ss1qv4si): Ditto.
23214 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
23215 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
23216 (aarch64_sm4eqv4si): Ditto.
23217 (aarch64_sm4ekeyqv4si): Ditto.
23218 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
23219 (sm3part_op): Ditto.
23220 (CRYPTO_SM3TT): Ditto.
23221 (CRYPTO_SM3PART): Ditto.
23222 (UNSPEC_SM3SS1): New unspec.
23223 (UNSPEC_SM3TT1A): Ditto.
23224 (UNSPEC_SM3TT1B): Ditto.
23225 (UNSPEC_SM3TT2A): Ditto.
23226 (UNSPEC_SM3TT2B): Ditto.
23227 (UNSPEC_SM3PARTW1): Ditto.
23228 (UNSPEC_SM3PARTW2): Ditto.
23229 (UNSPEC_SM4E): Ditto.
23230 (UNSPEC_SM4EKEY): Ditto.
23231 * config/aarch64/constraints.md (Ui2): New constraint.
23232 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
23233 * config/arm/types.md (crypto_sm3): New type attribute.
23234 (crypto_sm4): Ditto.
23235 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
23236 (vsm3tt1aq_u32): Ditto.
23237 (vsm3tt1bq_u32): Ditto.
23238 (vsm3tt2aq_u32): Ditto.
23239 (vsm3tt2bq_u32): Ditto.
23240 (vsm3partw1q_u32): Ditto.
23241 (vsm3partw2q_u32): Ditto.
23242 (vsm4eq_u32): Ditto.
23243 (vsm4ekeyq_u32): Ditto.
23244 (doc/invoke.texi): Document new sm4 option.
23246 2018-01-10 Michael Collison <michael.collison@arm.com>
23248 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
23249 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
23250 (AARCH64_FL_FOR_ARCH8_4): New.
23251 (AARCH64_FL_V8_4): New flag.
23252 (doc/invoke.texi): Document new armv8.4-a option.
23254 2018-01-10 Michael Collison <michael.collison@arm.com>
23256 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23257 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
23258 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
23259 * config/aarch64/aarch64-option-extension.def: Add
23260 AARCH64_OPT_EXTENSION of 'sha2'.
23261 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
23262 (crypto): Disable sha2 and aes if crypto disabled.
23263 (crypto): Enable aes and sha2 if enabled.
23264 (simd): Disable sha2 and aes if simd disabled.
23265 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
23267 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
23268 (TARGET_SHA2): New feature flag for sha2.
23269 (TARGET_AES): New feature flag for aes.
23270 * config/aarch64/aarch64-simd.md:
23271 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
23272 conditional on TARGET_AES.
23273 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
23274 (aarch64_crypto_sha1hsi): Make pattern conditional
23276 (aarch64_crypto_sha1hv4si): Ditto.
23277 (aarch64_be_crypto_sha1hv4si): Ditto.
23278 (aarch64_crypto_sha1su1v4si): Ditto.
23279 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
23280 (aarch64_crypto_sha1su0v4si): Ditto.
23281 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
23282 (aarch64_crypto_sha256su0v4si): Ditto.
23283 (aarch64_crypto_sha256su1v4si): Ditto.
23284 (doc/invoke.texi): Document new aes and sha2 options.
23286 2018-01-10 Martin Sebor <msebor@redhat.com>
23288 PR tree-optimization/83781
23289 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
23292 2018-01-11 Martin Sebor <msebor@gmail.com>
23293 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23295 PR tree-optimization/83501
23296 PR tree-optimization/81703
23298 * tree-ssa-strlen.c (get_string_cst): Rename...
23299 (get_string_len): ...to this. Handle global constants.
23300 (handle_char_store): Adjust.
23302 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
23303 Jim Wilson <jimw@sifive.com>
23305 * config/riscv/riscv-protos.h (riscv_output_return): New.
23306 * config/riscv/riscv.c (struct machine_function): New naked_p field.
23307 (riscv_attribute_table, riscv_output_return),
23308 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
23309 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
23310 (riscv_compute_frame_info): Only compute frame->mask if not a naked
23312 (riscv_expand_prologue): Add early return for naked function.
23313 (riscv_expand_epilogue): Likewise.
23314 (riscv_function_ok_for_sibcall): Return false for naked function.
23315 (riscv_set_current_function): New.
23316 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
23317 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
23318 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
23319 * doc/extend.texi (RISC-V Function Attributes): New.
23321 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
23323 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
23324 check for 128-bit long double before checking TCmode.
23325 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
23326 128-bit long doubles before checking TFmode or TCmode.
23327 (FLOAT128_IBM_P): Likewise.
23329 2018-01-10 Martin Sebor <msebor@redhat.com>
23331 PR tree-optimization/83671
23332 * builtins.c (c_strlen): Unconditionally return zero for the empty
23334 Use -Warray-bounds for warnings.
23335 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
23336 for non-constant array indices with COMPONENT_REF, arrays of
23337 arrays, and pointers to arrays.
23338 (gimple_fold_builtin_strlen): Determine and set length range for
23339 non-constant character arrays.
23341 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
23343 PR middle-end/81897
23344 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
23347 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
23349 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
23351 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
23354 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
23355 VECTOR_MEM_ALTIVEC_OR_VSX_P.
23356 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
23357 indexed_or_indirect_operand predicate.
23358 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
23359 (*vsx_le_perm_load_v8hi): Likewise.
23360 (*vsx_le_perm_load_v16qi): Likewise.
23361 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
23362 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
23363 (*vsx_le_perm_store_v8hi): Likewise.
23364 (*vsx_le_perm_store_v16qi): Likewise.
23365 (eight unnamed splitters): Likewise.
23367 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
23369 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
23370 * config/rs6000/emmintrin.h: Likewise.
23371 * config/rs6000/mmintrin.h: Likewise.
23372 * config/rs6000/xmmintrin.h: Likewise.
23374 2018-01-10 David Malcolm <dmalcolm@redhat.com>
23377 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
23379 * tree.c (tree_nop_conversion): Return true for location wrapper
23381 (maybe_wrap_with_location): New function.
23382 (selftest::check_strip_nops): New function.
23383 (selftest::test_location_wrappers): New function.
23384 (selftest::tree_c_tests): Call it.
23385 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
23386 (maybe_wrap_with_location): New decl.
23387 (EXPR_LOCATION_WRAPPER_P): New macro.
23388 (location_wrapper_p): New inline function.
23389 (tree_strip_any_location_wrapper): New inline function.
23391 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
23394 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
23395 stack_realign_offset for the largest alignment of stack slot
23397 (ix86_find_max_used_stack_alignment): New function.
23398 (ix86_finalize_stack_frame_flags): Use it. Set
23399 max_used_stack_alignment if we don't realign stack.
23400 * config/i386/i386.h (machine_function): Add
23401 max_used_stack_alignment.
23403 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
23405 * config/arm/arm.opt (-mbranch-cost): New option.
23406 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
23409 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
23412 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
23413 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
23415 2018-01-10 Richard Biener <rguenther@suse.de>
23418 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
23419 early out so it also covers the case where we have a non-NULL
23422 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
23424 PR tree-optimization/83753
23425 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
23426 for non-strided grouped accesses if the number of elements is 1.
23428 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23431 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
23432 * i386.h (TARGET_USE_GATHER): Define.
23433 * x86-tune.def (X86_TUNE_USE_GATHER): New.
23435 2018-01-10 Martin Liska <mliska@suse.cz>
23438 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
23439 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
23441 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
23442 CLEANUP_NO_PARTITIONING is not set.
23444 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
23446 * doc/rtl.texi: Remove documentation of (const ...) wrappers
23447 for vectors, as a partial revert of r254296.
23448 * rtl.h (const_vec_p): Delete.
23449 (const_vec_duplicate_p): Don't test for vector CONSTs.
23450 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
23451 * expmed.c (make_tree): Likewise.
23454 * common.md (E, F): Use CONSTANT_P instead of checking for
23456 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
23457 checking for CONST_VECTOR.
23459 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
23461 PR middle-end/83575
23462 * predict.c (force_edge_cold): Handle in more sane way edges
23463 with no prediction.
23465 2018-01-09 Carl Love <cel@us.ibm.com>
23467 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
23469 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
23470 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
23471 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
23472 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
23473 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
23474 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
23475 * config/rs6000/rs6000-protos.h: Add extern defition for
23476 rs6000_generate_float2_double_code.
23477 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
23479 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
23480 (float2_v2df): Add define_expand.
23482 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
23485 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
23486 op_mode in the force_to_mode call.
23488 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
23490 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
23491 instead of checking each element individually.
23492 (aarch64_evpc_uzp): Likewise.
23493 (aarch64_evpc_zip): Likewise.
23494 (aarch64_evpc_ext): Likewise.
23495 (aarch64_evpc_rev): Likewise.
23496 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
23497 instead of checking each element individually. Return true without
23499 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
23500 whether all selected elements come from the same input, instead of
23501 checking each element individually. Remove calls to gen_rtx_REG,
23502 start_sequence and end_sequence and instead assert that no rtl is
23505 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
23507 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
23508 order of HIGH and CONST checks.
23510 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
23512 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
23513 if the destination isn't an SSA_NAME.
23515 2018-01-09 Richard Biener <rguenther@suse.de>
23517 PR tree-optimization/83668
23518 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
23520 (canonicalize_loop_form): ... here, renamed from ...
23521 (canonicalize_loop_closed_ssa_form): ... this and amended to
23522 swap successor edges for loop exit blocks to make us use
23523 the RPO order we need for initial schedule generation.
23525 2018-01-09 Joseph Myers <joseph@codesourcery.com>
23527 PR tree-optimization/64811
23528 * match.pd: When optimizing comparisons with Inf, avoid
23529 introducing or losing exceptions from comparisons with NaN.
23531 2018-01-09 Martin Liska <mliska@suse.cz>
23534 * asan.c (shadow_mem_size): Add gcc_assert.
23536 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
23538 Don't save registers in main().
23541 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
23542 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
23543 * config/avr/avr.c (avr_set_current_function): Don't error if
23544 naked, OS_task or OS_main are specified at the same time.
23545 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
23547 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
23549 * common/config/avr/avr-common.c (avr_option_optimization_table):
23550 Switch on -mmain-is-OS_task for optimizing compilations.
23552 2018-01-09 Richard Biener <rguenther@suse.de>
23554 PR tree-optimization/83572
23555 * graphite.c: Include cfganal.h.
23556 (graphite_transform_loops): Connect infinite loops to exit
23557 and remove fake edges at the end.
23559 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
23561 * ipa-inline.c (edge_badness): Revert accidental checkin.
23563 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
23566 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
23567 symbols; not inline clones.
23569 2018-01-09 Jakub Jelinek <jakub@redhat.com>
23572 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
23573 hard registers. Formatting fixes.
23575 PR preprocessor/83722
23576 * gcc.c (try_generate_repro): Pass
23577 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
23578 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
23581 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
23582 Kito Cheng <kito.cheng@gmail.com>
23584 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
23585 (riscv_leaf_function_p): Delete.
23586 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
23588 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23590 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
23592 (do_ifelse): New function.
23593 (do_isel): New function.
23594 (do_sub3): New function.
23595 (do_add3): New function.
23596 (do_load_mask_compare): New function.
23597 (do_overlap_load_compare): New function.
23598 (expand_compare_loop): New function.
23599 (expand_block_compare): Call expand_compare_loop() when appropriate.
23600 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
23601 option description.
23602 (-mblock-compare-inline-loop-limit): New option.
23604 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23607 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
23608 Reverse order of second and third operands in first alternative.
23609 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
23610 of first and second elements in UNSPEC_VPERMR vector.
23611 (altivec_expand_vec_perm_le): Likewise.
23613 2018-01-08 Jeff Law <law@redhat.com>
23615 PR rtl-optimizatin/81308
23616 * tree-switch-conversion.c (cfg_altered): New file scoped static.
23617 (process_switch): If group_case_labels makes a change, then set
23619 (pass_convert_switch::execute): If a switch is converted, then
23620 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
23622 PR rtl-optimization/81308
23623 * recog.c (split_all_insns): Conditionally cleanup the CFG after
23626 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
23628 PR target/83663 - Revert r255946
23629 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
23630 generation for cases where splatting a value is not useful.
23631 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
23632 across a vec_duplicate and a paradoxical subreg forming a vector
23633 mode to a vec_concat.
23635 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23637 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
23638 -march=armv8.3-a variants.
23639 * config/arm/t-multilib: Likewise.
23640 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
23642 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23644 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
23646 (cceq_ior_compare_complement): Give it a name so I can use it, and
23647 change boolean_or_operator predicate to boolean_operator so it can
23648 be used to generate a crand.
23649 (eqne): New code iterator.
23650 (bd/bd_neg): New code_attrs.
23651 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
23652 a single define_insn.
23653 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
23654 decrement (bdnzt/bdnzf/bdzt/bdzf).
23655 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
23656 with the new names of the branch decrement patterns, and added the
23657 names of the branch decrement conditional patterns.
23659 2018-01-08 Richard Biener <rguenther@suse.de>
23661 PR tree-optimization/83563
23662 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
23665 2018-01-08 Richard Biener <rguenther@suse.de>
23667 PR middle-end/83713
23668 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
23670 2018-01-08 Richard Biener <rguenther@suse.de>
23672 PR tree-optimization/83685
23673 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
23674 references to abnormals.
23676 2018-01-08 Richard Biener <rguenther@suse.de>
23679 * dwarf2out.c (output_indirect_strings): Handle empty
23680 skeleton_debug_str_hash.
23681 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
23683 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
23685 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
23686 (emit_store_direct): Likewise.
23687 (arc_trampoline_adjust_address): Likewise.
23688 (arc_asm_trampoline_template): New function.
23689 (arc_initialize_trampoline): Use asm_trampoline_template.
23690 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
23691 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
23692 * config/arc/arc.md (flush_icache): Delete pattern.
23694 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
23696 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
23697 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
23700 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
23703 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
23704 by not USED_FOR_TARGET.
23705 (make_pass_resolve_sw_modes): Likewise.
23707 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
23709 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
23712 2018-01-08 Richard Biener <rguenther@suse.de>
23714 PR middle-end/83580
23715 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
23717 2018-01-08 Richard Biener <rguenther@suse.de>
23719 PR middle-end/83517
23720 * match.pd ((t * 2) / 2) -> t): Add missing :c.
23722 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
23724 PR middle-end/81897
23725 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
23726 basic blocks with a small number of successors.
23727 (convert_control_dep_chain_into_preds): Improve handling of
23729 (dump_predicates): Split apart into...
23730 (dump_pred_chain): ...here...
23731 (dump_pred_info): ...and here.
23732 (can_one_predicate_be_invalidated_p): Add debugging printfs.
23733 (can_chain_union_be_invalidated_p): Improve check for invalidation
23735 (uninit_uses_cannot_happen): Avoid unnecessary if
23736 convert_control_dep_chain_into_preds yielded nothing.
23738 2018-01-06 Martin Sebor <msebor@redhat.com>
23740 PR tree-optimization/83640
23741 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
23742 subtracting negative offset from size.
23743 (builtin_access::overlap): Adjust offset bounds of the access to fall
23744 within the size of the object if possible.
23746 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
23748 PR rtl-optimization/83699
23749 * expmed.c (extract_bit_field_1): Restrict the vector usage of
23750 extract_bit_field_as_subreg to cases in which the extracted
23751 value is also a vector.
23753 * lra-constraints.c (process_alt_operands): Test for the equivalence
23754 substitutions when detecting a possible reload cycle.
23756 2018-01-06 Jakub Jelinek <jakub@redhat.com>
23759 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
23760 by default if flag_selective_schedling{,2}. Formatting fixes.
23762 PR rtl-optimization/83682
23763 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
23764 if it has non-VECTOR_MODE element mode.
23765 (vec_duplicate_p): Likewise.
23767 PR middle-end/83694
23768 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
23769 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
23771 2018-01-05 Jakub Jelinek <jakub@redhat.com>
23774 * config/i386/i386-builtin.def
23775 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
23776 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
23777 Require also OPTION_MASK_ISA_AVX512F in addition to
23778 OPTION_MASK_ISA_GFNI.
23779 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
23780 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
23781 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
23782 to OPTION_MASK_ISA_GFNI.
23783 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
23784 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
23785 OPTION_MASK_ISA_AVX512BW.
23786 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
23787 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
23788 addition to OPTION_MASK_ISA_GFNI.
23789 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
23790 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
23791 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
23792 to OPTION_MASK_ISA_GFNI.
23793 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
23794 a requirement for all ISAs rather than any of them with a few
23796 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
23798 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
23799 bitmasks to be enabled with 3 exceptions, instead of requiring any
23800 enabled ISA with lots of exceptions.
23801 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
23802 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
23803 Change avx512bw in isa attribute to avx512f.
23804 * config/i386/sgxintrin.h: Add license boilerplate.
23805 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
23806 to __AVX512F__ and __AVX512VL to __AVX512VL__.
23807 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
23808 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
23810 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
23811 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
23812 temporarily sse2 rather than sse if not enabled already.
23815 * config/i386/sse.md (VI248_VLBW): Rename to ...
23816 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
23817 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
23818 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
23819 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
23820 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
23821 mode iterator instead of VI248_VLBW.
23823 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
23825 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
23826 (record_modified): Skip clobbers; add debug output.
23827 (param_change_prob): Use sreal frequencies.
23829 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
23831 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
23832 punt for user-aligned variables.
23834 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
23836 * tree-chrec.c (chrec_contains_symbols): Return true for
23839 2018-01-05 Sudakshina Das <sudi.das@arm.com>
23842 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
23843 of (x|y) == x for BICS pattern.
23845 2018-01-05 Jakub Jelinek <jakub@redhat.com>
23847 PR tree-optimization/83605
23848 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
23849 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
23852 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
23854 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
23855 * config/epiphany/rtems.h: New file.
23857 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23858 Uros Bizjak <ubizjak@gmail.com>
23861 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
23862 QIreg_operand instead of register_operand predicate.
23863 * config/i386/i386.c (ix86_rop_should_change_byte_p,
23864 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
23865 comments instead of -fmitigate[-_]rop.
23867 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
23870 * cgraphunit.c (symbol_table::compile): Switch to text_section
23871 before calling assembly_start debug hook.
23872 * run-rtl-passes.c (run_rtl_passes): Likewise.
23875 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23877 * tree-vrp.c (extract_range_from_binary_expr_1): Check
23878 range_int_cst_p rather than !symbolic_range_p before calling
23879 extract_range_from_multiplicative_op_1.
23881 2018-01-04 Jeff Law <law@redhat.com>
23883 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
23884 redundant test in assertion.
23886 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23888 * doc/rtl.texi: Document machine_mode wrapper classes.
23890 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23892 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
23893 using tree_to_uhwi.
23895 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23897 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
23898 the VEC_PERM_EXPR fold to fail.
23900 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23903 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
23904 to switched_sections.
23906 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23909 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
23910 test for d.testing.
23912 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
23915 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
23916 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
23918 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23921 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
23922 is BLKmode and bitpos not zero or mode change is needed.
23924 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23927 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
23930 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
23933 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
23934 instead of MULT rtx. Update all corresponding splitters.
23935 (*saddl_se): Ditto.
23936 (*ssub<modesuffix>): Ditto.
23937 (*ssubl_se): Ditto.
23938 (*cmp_sadd_di): Update split patterns.
23939 (*cmp_sadd_si): Ditto.
23940 (*cmp_sadd_sidi): Ditto.
23941 (*cmp_ssub_di): Ditto.
23942 (*cmp_ssub_si): Ditto.
23943 (*cmp_ssub_sidi): Ditto.
23944 * config/alpha/predicates.md (const23_operand): New predicate.
23945 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
23946 Look for ASHIFT, not MULT inner operand.
23947 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
23949 2018-01-04 Martin Liska <mliska@suse.cz>
23951 PR gcov-profile/83669
23952 * gcov.c (output_intermediate_file): Add version to intermediate
23954 * doc/gcov.texi: Document new field 'version' in intermediate
23955 file format. Fix location of '-k' option of gcov command.
23957 2018-01-04 Martin Liska <mliska@suse.cz>
23960 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
23962 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23964 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
23966 2018-01-03 Martin Sebor <msebor@redhat.com>
23968 PR tree-optimization/83655
23969 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
23970 checking calls with invalid arguments.
23972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23974 * tree-vect-stmts.c (vect_get_store_rhs): New function.
23975 (vectorizable_mask_load_store): Delete.
23976 (vectorizable_call): Return false for masked loads and stores.
23977 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
23978 instead of gimple_assign_rhs1.
23979 (vectorizable_load): Handle IFN_MASK_LOAD.
23980 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
23982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23984 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
23986 (vectorizable_mask_load_store): ...here.
23987 (vectorizable_load): ...and here.
23989 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23991 * tree-vect-stmts.c (vect_build_all_ones_mask)
23992 (vect_build_zero_merge_argument): New functions, split out from...
23993 (vectorizable_load): ...here.
23995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23997 * tree-vect-stmts.c (vect_check_store_rhs): New function,
23999 (vectorizable_mask_load_store): ...here.
24000 (vectorizable_store): ...and here.
24002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24004 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
24006 (vectorizable_mask_load_store): ...here.
24008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24010 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
24011 (vect_model_store_cost): Take a vec_load_store_type instead of a
24013 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
24014 (vect_model_store_cost): Take a vec_load_store_type instead of a
24016 (vectorizable_mask_load_store): Update accordingly.
24017 (vectorizable_store): Likewise.
24018 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
24020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24022 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
24023 IFN_MASK_LOAD calls here rather than...
24024 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
24026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24027 Alan Hayward <alan.hayward@arm.com>
24028 David Sherwood <david.sherwood@arm.com>
24030 * expmed.c (extract_bit_field_1): For vector extracts,
24031 fall back to extract_bit_field_as_subreg if vec_extract
24034 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24035 Alan Hayward <alan.hayward@arm.com>
24036 David Sherwood <david.sherwood@arm.com>
24038 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
24039 they are variable or constant sized.
24040 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
24041 slots for constant-sized data.
24043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24044 Alan Hayward <alan.hayward@arm.com>
24045 David Sherwood <david.sherwood@arm.com>
24047 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
24048 handling COND_EXPRs with boolean comparisons, try to find a better
24049 basis for the mask type than the boolean itself.
24051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24053 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
24054 is calculated and how it can be overridden.
24055 * genmodes.c (max_bitsize_mode_any_mode): New variable.
24056 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
24058 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
24061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24062 Alan Hayward <alan.hayward@arm.com>
24063 David Sherwood <david.sherwood@arm.com>
24065 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
24066 Remove the mode argument.
24067 (aarch64_simd_valid_immediate): Remove the mode and inverse
24069 * config/aarch64/iterators.md (bitsize): New iterator.
24070 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
24071 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
24072 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
24073 aarch64_simd_valid_immediate.
24074 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
24075 (aarch64_reg_or_bic_imm): Likewise.
24076 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
24077 with an insn_type enum and msl with a modifier_type enum.
24078 Replace element_width with a scalar_mode. Change the shift
24079 to unsigned int. Add constructors for scalar_float_mode and
24080 scalar_int_mode elements.
24081 (aarch64_vect_float_const_representable_p): Delete.
24082 (aarch64_can_const_movi_rtx_p)
24083 (aarch64_simd_scalar_immediate_valid_for_move)
24084 (aarch64_simd_make_constant): Update call to
24085 aarch64_simd_valid_immediate.
24086 (aarch64_advsimd_valid_immediate_hs): New function.
24087 (aarch64_advsimd_valid_immediate): Likewise.
24088 (aarch64_simd_valid_immediate): Remove mode and inverse
24089 arguments. Rewrite to use the above. Use const_vec_duplicate_p
24090 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
24091 and aarch64_float_const_representable_p on the result.
24092 (aarch64_output_simd_mov_immediate): Remove mode argument.
24093 Update call to aarch64_simd_valid_immediate and use of
24094 simd_immediate_info.
24095 (aarch64_output_scalar_simd_mov_immediate): Update call
24098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24099 Alan Hayward <alan.hayward@arm.com>
24100 David Sherwood <david.sherwood@arm.com>
24102 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
24103 (mode_nunits): Likewise CONST_MODE_NUNITS.
24104 * machmode.def (ADJUST_NUNITS): Document.
24105 * genmodes.c (mode_data::need_nunits_adj): New field.
24106 (blank_mode): Update accordingly.
24107 (adj_nunits): New variable.
24108 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
24110 (emit_mode_size_inline): Set need_bytesize_adj for all modes
24111 listed in adj_nunits.
24112 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
24113 listed in adj_nunits. Don't emit case statements for such modes.
24114 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
24115 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
24116 nothing if adj_nunits is nonnull.
24117 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
24118 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
24119 (emit_mode_fbit): Update use of print_maybe_const_decl.
24120 (emit_move_size): Likewise. Treat the array as non-const
24122 (emit_mode_adjustments): Handle adj_nunits.
24124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24126 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
24127 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
24128 (VECTOR_MODES): Use it.
24129 (make_vector_modes): Take the prefix as an argument.
24131 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24132 Alan Hayward <alan.hayward@arm.com>
24133 David Sherwood <david.sherwood@arm.com>
24135 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
24136 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
24137 for MODE_VECTOR_BOOL.
24138 * machmode.def (VECTOR_BOOL_MODE): Document.
24139 * genmodes.c (VECTOR_BOOL_MODE): New macro.
24140 (make_vector_bool_mode): New function.
24141 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
24143 * lto-streamer-in.c (lto_input_mode_table): Likewise.
24144 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
24146 * stor-layout.c (int_mode_for_mode): Likewise.
24147 * tree.c (build_vector_type_for_mode): Likewise.
24148 * varasm.c (output_constant_pool_2): Likewise.
24149 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
24150 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
24151 for MODE_VECTOR_BOOL.
24152 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
24153 of mode class checks.
24154 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
24155 instead of a list of mode class checks.
24156 (expand_vector_scalar_condition): Likewise.
24157 (type_for_widest_vector_mode): Handle BImode as an inner mode.
24159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24160 Alan Hayward <alan.hayward@arm.com>
24161 David Sherwood <david.sherwood@arm.com>
24163 * machmode.h (mode_size): Change from unsigned short to
24165 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
24166 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24167 or if measurement_type is not polynomial.
24168 (fixed_size_mode::includes_p): Check for constant-sized modes.
24169 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
24170 return a poly_uint16 rather than an unsigned short.
24171 (emit_mode_size): Change the type of mode_size from unsigned short
24172 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
24173 (emit_mode_adjustments): Cope with polynomial vector sizes.
24174 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24176 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24178 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
24179 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
24180 * caller-save.c (setup_save_areas): Likewise.
24181 (replace_reg_with_saved_mem): Likewise.
24182 * calls.c (emit_library_call_value_1): Likewise.
24183 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
24184 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
24185 (gen_lowpart_for_combine): Likewise.
24186 * convert.c (convert_to_integer_1): Likewise.
24187 * cse.c (equiv_constant, cse_insn): Likewise.
24188 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
24189 (cselib_subst_to_values): Likewise.
24190 * dce.c (word_dce_process_block): Likewise.
24191 * df-problems.c (df_word_lr_mark_ref): Likewise.
24192 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
24193 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
24194 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
24195 (rtl_for_decl_location): Likewise.
24196 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
24197 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
24198 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
24199 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
24200 (expand_expr_real_1): Likewise.
24201 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
24202 (pad_below): Likewise.
24203 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24204 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
24205 * ira.c (get_subreg_tracking_sizes): Likewise.
24206 * ira-build.c (ira_create_allocno_objects): Likewise.
24207 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
24208 (ira_sort_regnos_for_alter_reg): Likewise.
24209 * ira-costs.c (record_operand_costs): Likewise.
24210 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
24211 (resolve_simple_move): Likewise.
24212 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
24213 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
24214 (lra_constraints): Likewise.
24215 (CONST_POOL_OK_P): Reject variable-sized modes.
24216 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
24217 (add_pseudo_to_slot, lra_spill): Likewise.
24218 * omp-low.c (omp_clause_aligned_alignment): Likewise.
24219 * optabs-query.c (get_best_extraction_insn): Likewise.
24220 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24221 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
24222 (expand_mult_highpart, valid_multiword_target_p): Likewise.
24223 * recog.c (offsettable_address_addr_space_p): Likewise.
24224 * regcprop.c (maybe_mode_change): Likewise.
24225 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
24226 * regrename.c (build_def_use): Likewise.
24227 * regstat.c (dump_reg_info): Likewise.
24228 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
24229 (find_reloads, find_reloads_subreg_address): Likewise.
24230 * reload1.c (eliminate_regs_1): Likewise.
24231 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
24232 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
24233 (simplify_binary_operation_1, simplify_subreg): Likewise.
24234 * targhooks.c (default_function_arg_padding): Likewise.
24235 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
24236 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
24237 (verify_gimple_assign_ternary): Likewise.
24238 * tree-inline.c (estimate_move_cost): Likewise.
24239 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24240 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
24241 (get_address_cost_ainc): Likewise.
24242 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
24243 (vect_supportable_dr_alignment): Likewise.
24244 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24245 (vectorizable_reduction): Likewise.
24246 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
24247 (vectorizable_operation, vectorizable_load): Likewise.
24248 * tree.c (build_same_sized_truth_vector_type): Likewise.
24249 * valtrack.c (cleanup_auto_inc_dec): Likewise.
24250 * var-tracking.c (emit_note_insn_var_location): Likewise.
24251 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
24252 (ADDR_VEC_ALIGN): Likewise.
24254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24255 Alan Hayward <alan.hayward@arm.com>
24256 David Sherwood <david.sherwood@arm.com>
24258 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
24260 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24261 or if measurement_type is polynomial.
24262 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
24263 * combine.c (make_extraction): Likewise.
24264 * dse.c (find_shift_sequence): Likewise.
24265 * dwarf2out.c (mem_loc_descriptor): Likewise.
24266 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
24267 (extract_bit_field, extract_low_bits): Likewise.
24268 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
24269 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
24270 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
24271 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
24272 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24273 * reload.c (find_reloads): Likewise.
24274 * reload1.c (alter_reg): Likewise.
24275 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
24276 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
24277 * tree-if-conv.c (predicate_mem_writes): Likewise.
24278 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
24279 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
24280 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
24281 * valtrack.c (dead_debug_insert_temp): Likewise.
24282 * varasm.c (mergeable_constant_section): Likewise.
24283 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
24285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24286 Alan Hayward <alan.hayward@arm.com>
24287 David Sherwood <david.sherwood@arm.com>
24289 * expr.c (expand_assignment): Cope with polynomial mode sizes
24290 when assigning to a CONCAT.
24292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24293 Alan Hayward <alan.hayward@arm.com>
24294 David Sherwood <david.sherwood@arm.com>
24296 * machmode.h (mode_precision): Change from unsigned short to
24298 (mode_to_precision): Return a poly_uint16 rather than an unsigned
24300 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
24301 or if measurement_type is not polynomial.
24302 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
24303 in which the mode is already known to be a scalar_int_mode.
24304 * genmodes.c (emit_mode_precision): Change the type of mode_precision
24305 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
24307 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24308 for GET_MODE_PRECISION.
24309 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24310 for GET_MODE_PRECISION.
24311 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
24313 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
24314 (expand_field_assignment, make_extraction): Likewise.
24315 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
24316 (get_last_value): Likewise.
24317 * convert.c (convert_to_integer_1): Likewise.
24318 * cse.c (cse_insn): Likewise.
24319 * expr.c (expand_expr_real_1): Likewise.
24320 * lra-constraints.c (simplify_operand_subreg): Likewise.
24321 * optabs-query.c (can_atomic_load_p): Likewise.
24322 * optabs.c (expand_atomic_load): Likewise.
24323 (expand_atomic_store): Likewise.
24324 * ree.c (combine_reaching_defs): Likewise.
24325 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
24326 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
24327 * tree.h (type_has_mode_precision_p): Likewise.
24328 * ubsan.c (instrument_si_overflow): Likewise.
24330 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24331 Alan Hayward <alan.hayward@arm.com>
24332 David Sherwood <david.sherwood@arm.com>
24334 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
24335 polynomial numbers of units.
24336 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
24337 (valid_vector_subparts_p): New function.
24338 (build_vector_type): Remove temporary shim and take the number
24339 of units as a poly_uint64 rather than an int.
24340 (build_opaque_vector_type): Take the number of units as a
24341 poly_uint64 rather than an int.
24342 * tree.c (build_vector_from_ctor): Handle polynomial
24343 TYPE_VECTOR_SUBPARTS.
24344 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
24345 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
24346 (build_vector_from_val): If the number of units is variable,
24347 use build_vec_duplicate_cst for constant operands and
24348 VEC_DUPLICATE_EXPR otherwise.
24349 (make_vector_type): Remove temporary is_constant ().
24350 (build_vector_type, build_opaque_vector_type): Take the number of
24351 units as a poly_uint64 rather than an int.
24352 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
24354 * cfgexpand.c (expand_debug_expr): Likewise.
24355 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
24356 (store_constructor, expand_expr_real_1): Likewise.
24357 (const_scalar_mask_from_tree): Likewise.
24358 * fold-const-call.c (fold_const_reduction): Likewise.
24359 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
24360 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
24361 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
24362 (fold_relational_const): Likewise.
24363 (native_interpret_vector): Likewise. Change the size from an
24364 int to an unsigned int.
24365 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
24366 TYPE_VECTOR_SUBPARTS.
24367 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
24368 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
24369 duplicating a non-constant operand into a variable-length vector.
24370 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
24371 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
24372 * ipa-icf.c (sem_variable::equals): Likewise.
24373 * match.pd: Likewise.
24374 * omp-simd-clone.c (simd_clone_subparts): Likewise.
24375 * print-tree.c (print_node): Likewise.
24376 * stor-layout.c (layout_type): Likewise.
24377 * targhooks.c (default_builtin_vectorization_cost): Likewise.
24378 * tree-cfg.c (verify_gimple_comparison): Likewise.
24379 (verify_gimple_assign_binary): Likewise.
24380 (verify_gimple_assign_ternary): Likewise.
24381 (verify_gimple_assign_single): Likewise.
24382 * tree-pretty-print.c (dump_generic_node): Likewise.
24383 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24384 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
24385 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
24386 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
24387 (vect_shift_permute_load_chain): Likewise.
24388 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
24389 (expand_vector_condition, optimize_vector_constructor): Likewise.
24390 (lower_vec_perm, get_compute_type): Likewise.
24391 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24392 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
24393 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
24394 (vect_recog_mask_conversion_pattern): Likewise.
24395 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
24396 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
24397 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24398 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
24399 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
24400 (vectorizable_shift, vectorizable_operation, vectorizable_store)
24401 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
24402 (supportable_widening_operation): Likewise.
24403 (supportable_narrowing_operation): Likewise.
24404 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
24406 * varasm.c (output_constant): Likewise.
24408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24409 Alan Hayward <alan.hayward@arm.com>
24410 David Sherwood <david.sherwood@arm.com>
24412 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
24413 so that both the length == 3 and length != 3 cases set up their
24414 own permute vectors. Add comments explaining why we know the
24415 number of elements is constant.
24416 (vect_permute_load_chain): Likewise.
24418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24419 Alan Hayward <alan.hayward@arm.com>
24420 David Sherwood <david.sherwood@arm.com>
24422 * machmode.h (mode_nunits): Change from unsigned char to
24424 (ONLY_FIXED_SIZE_MODES): New macro.
24425 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
24426 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
24427 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
24429 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
24430 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
24431 or if measurement_type is not polynomial.
24432 * genmodes.c (ZERO_COEFFS): New macro.
24433 (emit_mode_nunits_inline): Make mode_nunits_inline return a
24435 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
24436 Use ZERO_COEFFS when emitting initializers.
24437 * data-streamer.h (bp_pack_poly_value): New function.
24438 (bp_unpack_poly_value): Likewise.
24439 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24440 for GET_MODE_NUNITS.
24441 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24442 for GET_MODE_NUNITS.
24443 * tree.c (make_vector_type): Remove temporary shim and make
24444 the real function take the number of units as a poly_uint64
24445 rather than an int.
24446 (build_vector_type_for_mode): Handle polynomial nunits.
24447 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
24448 * emit-rtl.c (const_vec_series_p_1): Likewise.
24449 (gen_rtx_CONST_VECTOR): Likewise.
24450 * fold-const.c (test_vec_duplicate_folding): Likewise.
24451 * genrecog.c (validate_pattern): Likewise.
24452 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
24453 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24454 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
24455 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
24456 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
24457 * rtlanal.c (subreg_get_info): Likewise.
24458 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
24459 (vect_grouped_load_supported): Likewise.
24460 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
24461 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
24462 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
24463 (simplify_const_unary_operation, simplify_binary_operation_1)
24464 (simplify_const_binary_operation, simplify_ternary_operation)
24465 (test_vector_ops_duplicate, test_vector_ops): Likewise.
24466 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
24467 instead of CONST_VECTOR_NUNITS.
24468 * varasm.c (output_constant_pool_2): Likewise.
24469 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
24470 explicit-encoded elements in the XVEC for variable-length vectors.
24472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24474 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
24476 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24477 Alan Hayward <alan.hayward@arm.com>
24478 David Sherwood <david.sherwood@arm.com>
24480 * coretypes.h (fixed_size_mode): Declare.
24481 (fixed_size_mode_pod): New typedef.
24482 * builtins.h (target_builtins::x_apply_args_mode)
24483 (target_builtins::x_apply_result_mode): Change type to
24484 fixed_size_mode_pod.
24485 * builtins.c (apply_args_size, apply_result_size, result_vector)
24486 (expand_builtin_apply_args_1, expand_builtin_apply)
24487 (expand_builtin_return): Update accordingly.
24489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24491 * cse.c (hash_rtx_cb): Hash only the encoded elements.
24492 * cselib.c (cselib_hash_rtx): Likewise.
24493 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
24494 CONST_VECTOR encoding.
24496 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24497 Jeff Law <law@redhat.com>
24500 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
24501 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
24502 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
24503 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
24506 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
24507 explicitly probe *sp in a noreturn function if there were any callee
24508 register saves or frame pointer is needed.
24510 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24513 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
24514 BLKmode for ternary, binary or unary expressions.
24517 * var-tracking.c (delete_vta_debug_insn): New inline function.
24518 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
24519 insns from get_insns () to NULL instead of each bb separately.
24520 Use delete_vta_debug_insn. No longer static.
24521 (vt_debug_insns_local, variable_tracking_main_1): Adjust
24522 delete_vta_debug_insns callers.
24523 * rtl.h (delete_vta_debug_insns): Declare.
24524 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
24525 instead of variable_tracking_main.
24527 2018-01-03 Martin Sebor <msebor@redhat.com>
24529 PR tree-optimization/83603
24530 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
24531 arguments past the endof the argument list in functions declared
24532 without a prototype.
24533 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
24534 Avoid checking when arguments are null.
24536 2018-01-03 Martin Sebor <msebor@redhat.com>
24539 * doc/extend.texi (attribute const): Fix a typo.
24540 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
24541 issuing -Wsuggest-attribute for void functions.
24543 2018-01-03 Martin Sebor <msebor@redhat.com>
24545 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
24546 offset_int::from instead of wide_int::to_shwi.
24547 (maybe_diag_overlap): Remove assertion.
24548 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
24549 * gimple-ssa-sprintf.c (format_directive): Same.
24550 (parse_directive): Same.
24551 (sprintf_dom_walker::compute_format_length): Same.
24552 (try_substitute_return_value): Same.
24554 2018-01-03 Jeff Law <law@redhat.com>
24556 PR middle-end/83654
24557 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
24558 non-constant residual for zero at runtime and avoid probing in
24559 that case. Reorganize code for trailing problem to mirror handling
24562 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24564 PR tree-optimization/83501
24565 * tree-ssa-strlen.c (get_string_cst): New.
24566 (handle_char_store): Call get_string_cst.
24568 2018-01-03 Martin Liska <mliska@suse.cz>
24570 PR tree-optimization/83593
24571 * tree-ssa-strlen.c: Include tree-cfg.h.
24572 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
24573 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
24574 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
24576 (strlen_dom_walker::before_dom_children): Call
24577 gimple_purge_dead_eh_edges. Dump tranformation with details
24579 (strlen_dom_walker::before_dom_children): Update call by adding
24580 new argument cleanup_eh.
24581 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
24583 2018-01-03 Martin Liska <mliska@suse.cz>
24586 * cif-code.def (VARIADIC_THUNK): New enum value.
24587 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
24590 2018-01-03 Jan Beulich <jbeulich@suse.com>
24592 * sse.md (mov<mode>_internal): Tighten condition for when to use
24593 vmovdqu<ssescalarsize> for TI and OI modes.
24595 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24597 Update copyright years.
24599 2018-01-03 Martin Liska <mliska@suse.cz>
24602 * ipa-visibility.c (function_and_variable_visibility): Skip
24603 functions with noipa attribure.
24605 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24607 * gcc.c (process_command): Update copyright notice dates.
24608 * gcov-dump.c (print_version): Ditto.
24609 * gcov.c (print_version): Ditto.
24610 * gcov-tool.c (print_version): Ditto.
24611 * gengtype.c (create_file): Ditto.
24612 * doc/cpp.texi: Bump @copying's copyright year.
24613 * doc/cppinternals.texi: Ditto.
24614 * doc/gcc.texi: Ditto.
24615 * doc/gccint.texi: Ditto.
24616 * doc/gcov.texi: Ditto.
24617 * doc/install.texi: Ditto.
24618 * doc/invoke.texi: Ditto.
24620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24622 * vector-builder.h (vector_builder::m_full_nelts): Change from
24623 unsigned int to poly_uint64.
24624 (vector_builder::full_nelts): Update prototype accordingly.
24625 (vector_builder::new_vector): Likewise.
24626 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
24627 (vector_builder::operator ==): Likewise.
24628 (vector_builder::finalize): Likewise.
24629 * int-vector-builder.h (int_vector_builder::int_vector_builder):
24630 Take the number of elements as a poly_uint64 rather than an
24632 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
24633 from unsigned int to poly_uint64.
24634 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
24635 (vec_perm_indices::new_vector): Likewise.
24636 (vec_perm_indices::length): Likewise.
24637 (vec_perm_indices::nelts_per_input): Likewise.
24638 (vec_perm_indices::input_nelts): Likewise.
24639 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
24640 number of elements per input as a poly_uint64 rather than an
24641 unsigned int. Use the original encoding for variable-length
24642 vectors, rather than clamping each individual element.
24643 For the second and subsequent elements in each pattern,
24644 clamp the step and base before clamping their sum.
24645 (vec_perm_indices::series_p): Handle polynomial element counts.
24646 (vec_perm_indices::all_in_range_p): Likewise.
24647 (vec_perm_indices_to_tree): Likewise.
24648 (vec_perm_indices_to_rtx): Likewise.
24649 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
24650 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
24651 (tree_vector_builder::new_binary_operation): Handle polynomial
24652 element counts. Return false if we need to know the number
24653 of elements at compile time.
24654 * fold-const.c (fold_vec_perm): Punt if the number of elements
24655 isn't known at compile time.
24657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24659 * vec-perm-indices.h (vec_perm_builder): Change element type
24660 from HOST_WIDE_INT to poly_int64.
24661 (vec_perm_indices::element_type): Update accordingly.
24662 (vec_perm_indices::clamp): Handle polynomial element_types.
24663 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
24664 (vec_perm_indices::all_in_range_p): Likewise.
24665 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
24667 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
24668 polynomial vec_perm_indices element types.
24669 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
24670 * fold-const.c (fold_vec_perm): Likewise.
24671 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
24672 * tree-vect-generic.c (lower_vec_perm): Likewise.
24673 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
24674 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
24675 element type to HOST_WIDE_INT.
24677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24678 Alan Hayward <alan.hayward@arm.com>
24679 David Sherwood <david.sherwood@arm.com>
24681 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
24682 rather than an int. Use plus_constant.
24683 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
24684 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
24686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24687 Alan Hayward <alan.hayward@arm.com>
24688 David Sherwood <david.sherwood@arm.com>
24690 * calls.c (emit_call_1, expand_call): Change struct_value_size from
24691 a HOST_WIDE_INT to a poly_int64.
24693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24694 Alan Hayward <alan.hayward@arm.com>
24695 David Sherwood <david.sherwood@arm.com>
24697 * calls.c (load_register_parameters): Cope with polynomial
24698 mode sizes. Require a constant size for BLKmode parameters
24699 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
24700 forces a parameter to be padded at the lsb end in order to
24701 fill a complete number of words, require the parameter size
24702 to be ordered wrt UNITS_PER_WORD.
24704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24705 Alan Hayward <alan.hayward@arm.com>
24706 David Sherwood <david.sherwood@arm.com>
24708 * reload1.c (spill_stack_slot_width): Change element type
24709 from unsigned int to poly_uint64_pod.
24710 (alter_reg): Treat mode sizes as polynomial.
24712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24713 Alan Hayward <alan.hayward@arm.com>
24714 David Sherwood <david.sherwood@arm.com>
24716 * reload.c (complex_word_subreg_p): New function.
24717 (reload_inner_reg_of_subreg, push_reload): Use it.
24719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24720 Alan Hayward <alan.hayward@arm.com>
24721 David Sherwood <david.sherwood@arm.com>
24723 * lra-constraints.c (process_alt_operands): Reject matched
24724 operands whose sizes aren't ordered.
24725 (match_reload): Refer to this check here.
24727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24728 Alan Hayward <alan.hayward@arm.com>
24729 David Sherwood <david.sherwood@arm.com>
24731 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
24732 that the mode size is in the set {1, 2, 4, 8, 16}.
24734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24735 Alan Hayward <alan.hayward@arm.com>
24736 David Sherwood <david.sherwood@arm.com>
24738 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
24739 Use plus_constant instead of gen_rtx_PLUS.
24741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24742 Alan Hayward <alan.hayward@arm.com>
24743 David Sherwood <david.sherwood@arm.com>
24745 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
24746 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
24747 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
24748 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
24749 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
24750 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
24751 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
24752 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
24753 * config/i386/i386.c (ix86_push_rounding): ...this new function.
24754 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
24756 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
24757 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
24758 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
24759 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
24760 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
24761 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
24762 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
24763 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
24764 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
24765 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
24767 * expr.c (emit_move_resolve_push): Treat the input and result
24768 of PUSH_ROUNDING as a poly_int64.
24769 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
24770 (emit_push_insn): Likewise.
24771 * lra-eliminations.c (mark_not_eliminable): Likewise.
24772 * recog.c (push_operand): Likewise.
24773 * reload1.c (elimination_effects): Likewise.
24774 * rtlanal.c (nonzero_bits1): Likewise.
24775 * calls.c (store_one_arg): Likewise. Require the padding to be
24776 known at compile time.
24778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24779 Alan Hayward <alan.hayward@arm.com>
24780 David Sherwood <david.sherwood@arm.com>
24782 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
24783 Use plus_constant instead of gen_rtx_PLUS.
24785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24786 Alan Hayward <alan.hayward@arm.com>
24787 David Sherwood <david.sherwood@arm.com>
24789 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
24790 rather than an int.
24792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24793 Alan Hayward <alan.hayward@arm.com>
24794 David Sherwood <david.sherwood@arm.com>
24796 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
24797 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
24798 via stack temporaries. Treat the mode size as polynomial too.
24800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24801 Alan Hayward <alan.hayward@arm.com>
24802 David Sherwood <david.sherwood@arm.com>
24804 * expr.c (expand_expr_real_2): When handling conversions involving
24805 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
24806 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
24807 as a poly_uint64 too.
24809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24810 Alan Hayward <alan.hayward@arm.com>
24811 David Sherwood <david.sherwood@arm.com>
24813 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
24815 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24816 Alan Hayward <alan.hayward@arm.com>
24817 David Sherwood <david.sherwood@arm.com>
24819 * combine.c (can_change_dest_mode): Handle polynomial
24820 REGMODE_NATURAL_SIZE.
24821 * expmed.c (store_bit_field_1): Likewise.
24822 * expr.c (store_constructor): Likewise.
24823 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
24824 and polynomial REGMODE_NATURAL_SIZE.
24825 (gen_lowpart_common): Likewise.
24826 * reginfo.c (record_subregs_of_mode): Likewise.
24827 * rtlanal.c (read_modify_subreg_p): Likewise.
24829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24830 Alan Hayward <alan.hayward@arm.com>
24831 David Sherwood <david.sherwood@arm.com>
24833 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
24834 numbers of elements.
24836 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24837 Alan Hayward <alan.hayward@arm.com>
24838 David Sherwood <david.sherwood@arm.com>
24840 * match.pd: Cope with polynomial numbers of vector elements.
24842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24843 Alan Hayward <alan.hayward@arm.com>
24844 David Sherwood <david.sherwood@arm.com>
24846 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
24847 in a POINTER_PLUS_EXPR.
24849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24850 Alan Hayward <alan.hayward@arm.com>
24851 David Sherwood <david.sherwood@arm.com>
24853 * omp-simd-clone.c (simd_clone_subparts): New function.
24854 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
24855 (ipa_simd_modify_function_body): Likewise.
24857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24858 Alan Hayward <alan.hayward@arm.com>
24859 David Sherwood <david.sherwood@arm.com>
24861 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
24862 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
24863 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
24864 (expand_vector_condition, vector_element): Likewise.
24865 (subparts_gt): New function.
24866 (get_compute_type): Use subparts_gt.
24867 (count_type_subparts): Delete.
24868 (expand_vector_operations_1): Use subparts_gt instead of
24869 count_type_subparts.
24871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24872 Alan Hayward <alan.hayward@arm.com>
24873 David Sherwood <david.sherwood@arm.com>
24875 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
24876 (vect_compile_time_alias): ...this new function. Do the calculation
24877 on poly_ints rather than trees.
24878 (vect_prune_runtime_alias_test_list): Update call accordingly.
24880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24881 Alan Hayward <alan.hayward@arm.com>
24882 David Sherwood <david.sherwood@arm.com>
24884 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
24886 (vect_schedule_slp_instance): Likewise.
24888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24889 Alan Hayward <alan.hayward@arm.com>
24890 David Sherwood <david.sherwood@arm.com>
24892 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
24893 constant and extern definitions for variable-length vectors.
24894 (vect_get_constant_vectors): Note that the number of units
24895 is known to be constant.
24897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24898 Alan Hayward <alan.hayward@arm.com>
24899 David Sherwood <david.sherwood@arm.com>
24901 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
24902 of units as polynomial. Choose between WIDE and NARROW based
24905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24906 Alan Hayward <alan.hayward@arm.com>
24907 David Sherwood <david.sherwood@arm.com>
24909 * tree-vect-stmts.c (simd_clone_subparts): New function.
24910 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
24912 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24913 Alan Hayward <alan.hayward@arm.com>
24914 David Sherwood <david.sherwood@arm.com>
24916 * tree-vect-stmts.c (vectorizable_call): Treat the number of
24917 vectors as polynomial. Use build_index_vector for
24918 IFN_GOMP_SIMD_LANE.
24920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24921 Alan Hayward <alan.hayward@arm.com>
24922 David Sherwood <david.sherwood@arm.com>
24924 * tree-vect-stmts.c (get_load_store_type): Treat the number of
24925 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
24926 for variable-length vectors.
24927 (vectorizable_mask_load_store): Treat the number of units as
24928 polynomial, asserting that it is constant if the condition has
24929 already been enforced.
24930 (vectorizable_store, vectorizable_load): Likewise.
24932 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24933 Alan Hayward <alan.hayward@arm.com>
24934 David Sherwood <david.sherwood@arm.com>
24936 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
24937 of units as polynomial. Punt if we can't tell at compile time
24938 which vector contains the final result.
24940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24941 Alan Hayward <alan.hayward@arm.com>
24942 David Sherwood <david.sherwood@arm.com>
24944 * tree-vect-loop.c (vectorizable_induction): Treat the number
24945 of units as polynomial. Punt on SLP inductions. Use an integer
24946 VEC_SERIES_EXPR for variable-length integer reductions. Use a
24947 cast of such a series for variable-length floating-point
24950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24951 Alan Hayward <alan.hayward@arm.com>
24952 David Sherwood <david.sherwood@arm.com>
24954 * tree.h (build_index_vector): Declare.
24955 * tree.c (build_index_vector): New function.
24956 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
24957 of units as polynomial, forcibly converting it to a constant if
24958 vectorizable_reduction has already enforced the condition.
24959 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
24960 to create a {1,2,3,...} vector.
24961 (vectorizable_reduction): Treat the number of units as polynomial.
24962 Choose vectype_in based on the largest scalar element size rather
24963 than the smallest number of units. Enforce the restrictions
24966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24967 Alan Hayward <alan.hayward@arm.com>
24968 David Sherwood <david.sherwood@arm.com>
24970 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
24971 number of units as polynomial.
24973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24974 Alan Hayward <alan.hayward@arm.com>
24975 David Sherwood <david.sherwood@arm.com>
24977 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
24978 * target.def (autovectorize_vector_sizes): Return the vector sizes
24979 by pointer, using vector_sizes rather than a bitmask.
24980 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
24981 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
24982 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
24984 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
24985 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
24986 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
24987 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
24988 * omp-general.c (omp_max_vf): Likewise.
24989 * omp-low.c (omp_clause_aligned_alignment): Likewise.
24990 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
24991 * tree-vect-loop.c (vect_analyze_loop): Likewise.
24992 * tree-vect-slp.c (vect_slp_bb): Likewise.
24993 * doc/tm.texi: Regenerate.
24994 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
24996 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
24997 the vector size as a poly_uint64 rather than an unsigned int.
24998 (current_vector_size): Change from an unsigned int to a poly_uint64.
24999 (get_vectype_for_scalar_type): Update accordingly.
25000 * tree.h (build_truth_vector_type): Take the size and number of
25001 units as a poly_uint64 rather than an unsigned int.
25002 (build_vector_type): Add a temporary overload that takes
25003 the number of units as a poly_uint64 rather than an unsigned int.
25004 * tree.c (make_vector_type): Likewise.
25005 (build_truth_vector_type): Take the number of units as a poly_uint64
25006 rather than an unsigned int.
25008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25009 Alan Hayward <alan.hayward@arm.com>
25010 David Sherwood <david.sherwood@arm.com>
25012 * target.def (get_mask_mode): Take the number of units and length
25013 as poly_uint64s rather than unsigned ints.
25014 * targhooks.h (default_get_mask_mode): Update accordingly.
25015 * targhooks.c (default_get_mask_mode): Likewise.
25016 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
25017 * doc/tm.texi: Regenerate.
25019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25020 Alan Hayward <alan.hayward@arm.com>
25021 David Sherwood <david.sherwood@arm.com>
25023 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
25024 * omp-general.c (omp_max_vf): Likewise.
25025 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
25026 (expand_omp_simd): Handle polynomial safelen.
25027 * omp-low.c (omplow_simd_context): Add a default constructor.
25028 (omplow_simd_context::max_vf): Change from int to poly_uint64.
25029 (lower_rec_simd_input_clauses): Update accordingly.
25030 (lower_rec_input_clauses): Likewise.
25032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25033 Alan Hayward <alan.hayward@arm.com>
25034 David Sherwood <david.sherwood@arm.com>
25036 * tree-vectorizer.h (vect_nunits_for_cost): New function.
25037 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
25038 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
25039 (vect_analyze_slp_cost): Likewise.
25040 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
25041 (vect_model_load_cost): Likewise.
25043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25044 Alan Hayward <alan.hayward@arm.com>
25045 David Sherwood <david.sherwood@arm.com>
25047 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
25048 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
25049 from an unsigned int * to a poly_uint64_pod *.
25050 (calculate_unrolling_factor): New function.
25051 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
25053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25054 Alan Hayward <alan.hayward@arm.com>
25055 David Sherwood <david.sherwood@arm.com>
25057 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
25058 from an unsigned int to a poly_uint64.
25059 (_loop_vec_info::slp_unrolling_factor): Likewise.
25060 (_loop_vec_info::vectorization_factor): Change from an int
25062 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
25063 (vect_get_num_vectors): New function.
25064 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
25065 (vect_get_num_copies): Use vect_get_num_vectors.
25066 (vect_analyze_data_ref_dependences): Change max_vf from an int *
25067 to an unsigned int *.
25068 (vect_analyze_data_refs): Change min_vf from an int * to a
25070 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25071 than an unsigned HOST_WIDE_INT.
25072 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
25073 (vect_analyze_data_ref_dependence): Change max_vf from an int *
25074 to an unsigned int *.
25075 (vect_analyze_data_ref_dependences): Likewise.
25076 (vect_compute_data_ref_alignment): Handle polynomial vf.
25077 (vect_enhance_data_refs_alignment): Likewise.
25078 (vect_prune_runtime_alias_test_list): Likewise.
25079 (vect_shift_permute_load_chain): Likewise.
25080 (vect_supportable_dr_alignment): Likewise.
25081 (dependence_distance_ge_vf): Take the vectorization factor as a
25082 poly_uint64 rather than an unsigned HOST_WIDE_INT.
25083 (vect_analyze_data_refs): Change min_vf from an int * to a
25085 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
25086 vfm1 as a poly_uint64 rather than an int. Make the same change
25087 for the returned bound_scalar.
25088 (vect_gen_vector_loop_niters): Handle polynomial vf.
25089 (vect_do_peeling): Likewise. Update call to
25090 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
25091 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
25093 * tree-vect-loop.c (vect_determine_vectorization_factor)
25094 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
25095 (vect_get_known_peeling_cost): Likewise.
25096 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
25097 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
25098 (vect_transform_loop): Likewise. Use the lowest possible VF when
25099 updating the upper bounds of the loop.
25100 (vect_min_worthwhile_factor): Make static. Return an unsigned int
25101 rather than an int.
25102 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
25103 polynomial unroll factors.
25104 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
25105 (vect_make_slp_decision): Likewise.
25106 (vect_supported_load_permutation_p): Likewise, and polynomial
25108 (vect_analyze_slp_cost): Handle polynomial vf.
25109 (vect_slp_analyze_node_operations): Likewise.
25110 (vect_slp_analyze_bb_1): Likewise.
25111 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25112 than an unsigned HOST_WIDE_INT.
25113 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
25114 (vectorizable_load): Handle polynomial vf.
25115 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
25117 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
25119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25120 Alan Hayward <alan.hayward@arm.com>
25121 David Sherwood <david.sherwood@arm.com>
25123 * match.pd: Handle bit operations involving three constants
25124 and try to fold one pair.
25126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25128 * tree-vect-loop-manip.c: Include gimple-fold.h.
25129 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
25130 niters_maybe_zero parameters. Handle other cases besides a step of 1.
25131 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
25132 Add a path that uses a step of VF instead of 1, but disable it
25134 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
25135 and niters_no_overflow parameters. Update calls to
25136 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
25137 Create a new SSA name if the latter choses to use a ste other
25138 than zero, and return it via niters_vector_mult_vf_var.
25139 * tree-vect-loop.c (vect_transform_loop): Update calls to
25140 vect_do_peeling, vect_gen_vector_loop_niters and
25141 slpeel_make_loop_iterate_ntimes.
25142 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
25143 (vect_gen_vector_loop_niters): Update declarations after above changes.
25145 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
25147 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
25148 128-bit round to integer instructions.
25149 (ceil<mode>2): Likewise.
25150 (btrunc<mode>2): Likewise.
25151 (round<mode>2): Likewise.
25153 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25155 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
25156 unaligned VSX load/store on P8/P9.
25157 (expand_block_clear): Allow the use of unaligned VSX
25158 load/store on P8/P9.
25160 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25162 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
25164 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
25165 swap associated with both a load and a store.
25167 2018-01-02 Andrew Waterman <andrew@sifive.com>
25169 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
25170 * config/riscv/riscv.md (clear_cache): Use it.
25172 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
25174 * web.c: Remove out-of-date comment.
25176 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25178 * expr.c (fixup_args_size_notes): Check that any existing
25179 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
25180 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
25181 (emit_single_push_insn): ...here.
25183 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25185 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
25186 (const_vector_encoded_nelts): New function.
25187 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
25188 (const_vector_int_elt, const_vector_elt): Declare.
25189 * emit-rtl.c (const_vector_int_elt_1): New function.
25190 (const_vector_elt): Likewise.
25191 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
25192 of CONST_VECTOR_ELT.
25194 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25196 * expr.c: Include rtx-vector-builder.h.
25197 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
25198 directly on the tree encoding.
25199 (const_vector_from_tree): Likewise.
25200 * optabs.c: Include rtx-vector-builder.h.
25201 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
25202 sequence of "u" values.
25203 * vec-perm-indices.c: Include rtx-vector-builder.h.
25204 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
25205 directly on the vec_perm_indices encoding.
25207 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25209 * doc/rtl.texi (const_vector): Describe new encoding scheme.
25210 * Makefile.in (OBJS): Add rtx-vector-builder.o.
25211 * rtx-vector-builder.h: New file.
25212 * rtx-vector-builder.c: Likewise.
25213 * rtl.h (rtx_def::u2): Add a const_vector field.
25214 (CONST_VECTOR_NPATTERNS): New macro.
25215 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
25216 (CONST_VECTOR_DUPLICATE_P): Likewise.
25217 (CONST_VECTOR_STEPPED_P): Likewise.
25218 (CONST_VECTOR_ENCODED_ELT): Likewise.
25219 (const_vec_duplicate_p): Check for a duplicated vector encoding.
25220 (unwrap_const_vec_duplicate): Likewise.
25221 (const_vec_series_p): Check for a non-duplicated vector encoding.
25222 Say that the function only returns true for integer vectors.
25223 * emit-rtl.c: Include rtx-vector-builder.h.
25224 (gen_const_vec_duplicate_1): Delete.
25225 (gen_const_vector): Call gen_const_vec_duplicate instead of
25226 gen_const_vec_duplicate_1.
25227 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
25228 (gen_const_vec_duplicate): Use rtx_vector_builder.
25229 (gen_const_vec_series): Likewise.
25230 (gen_rtx_CONST_VECTOR): Likewise.
25231 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
25232 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25233 Build a new vector rather than modifying a CONST_VECTOR in-place.
25234 (handle_special_swappables): Update call accordingly.
25235 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
25236 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25237 Build a new vector rather than modifying a CONST_VECTOR in-place.
25238 (handle_special_swappables): Update call accordingly.
25240 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25242 * simplify-rtx.c (simplify_const_binary_operation): Use
25243 CONST_VECTOR_ELT instead of XVECEXP.
25245 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25247 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
25248 the selector elements to be different from the data elements
25249 if the selector is a VECTOR_CST.
25250 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
25251 ssizetype for the selector.
25253 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25255 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
25256 before testing each element individually.
25257 * tree-vect-generic.c (lower_vec_perm): Likewise.
25259 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25261 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
25262 * selftest-run-tests.c (selftest::run_tests): Call it.
25263 * vector-builder.h (vector_builder::operator ==): New function.
25264 (vector_builder::operator !=): Likewise.
25265 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
25266 (vec_perm_indices::all_from_input_p): New function.
25267 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
25268 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
25269 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
25270 instead of reading the VECTOR_CST directly. Detect whether both
25271 vector inputs are the same before constructing the vec_perm_indices,
25272 and update the number of inputs argument accordingly. Use the
25273 utility functions added above. Only construct sel2 if we need to.
25275 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25277 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
25278 the broadcast of the low byte.
25279 (expand_mult_highpart): Use an explicit encoding for the permutes.
25280 * optabs-query.c (can_mult_highpart_p): Likewise.
25281 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25282 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25283 (vectorizable_bswap): Likewise.
25284 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
25285 explicit encoding for the power-of-2 permutes.
25286 (vect_permute_store_chain): Likewise.
25287 (vect_grouped_load_supported): Likewise.
25288 (vect_permute_load_chain): Likewise.
25290 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25292 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
25293 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
25294 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
25295 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
25296 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
25297 (vect_gen_perm_mask_any): Likewise.
25299 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25301 * int-vector-builder.h: New file.
25302 * vec-perm-indices.h: Include int-vector-builder.h.
25303 (vec_perm_indices): Redefine as an int_vector_builder.
25304 (auto_vec_perm_indices): Delete.
25305 (vec_perm_builder): Redefine as a stand-alone class.
25306 (vec_perm_indices::vec_perm_indices): New function.
25307 (vec_perm_indices::clamp): Likewise.
25308 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
25309 (vec_perm_indices::new_vector): New function.
25310 (vec_perm_indices::new_expanded_vector): Update for new
25311 vec_perm_indices class.
25312 (vec_perm_indices::rotate_inputs): New function.
25313 (vec_perm_indices::all_in_range_p): Operate directly on the
25314 encoded form, without computing elided elements.
25315 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
25316 encoding. Update for new vec_perm_indices class.
25317 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
25318 the given vec_perm_builder.
25319 (expand_vec_perm_var): Update vec_perm_builder constructor.
25320 (expand_mult_highpart): Use vec_perm_builder instead of
25321 auto_vec_perm_indices.
25322 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
25323 vec_perm_indices instead of auto_vec_perm_indices. Use a single
25324 or double series encoding as appropriate.
25325 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
25326 vec_perm_indices instead of auto_vec_perm_indices.
25327 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25328 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25329 (vect_permute_store_chain): Likewise.
25330 (vect_grouped_load_supported): Likewise.
25331 (vect_permute_load_chain): Likewise.
25332 (vect_shift_permute_load_chain): Likewise.
25333 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
25334 (vect_transform_slp_perm_load): Likewise.
25335 (vect_schedule_slp_instance): Likewise.
25336 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25337 (vectorizable_mask_load_store): Likewise.
25338 (vectorizable_bswap): Likewise.
25339 (vectorizable_store): Likewise.
25340 (vectorizable_load): Likewise.
25341 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
25342 vec_perm_indices instead of auto_vec_perm_indices. Use
25343 tree_to_vec_perm_builder to read the vector from a tree.
25344 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
25345 vec_perm_builder instead of a vec_perm_indices.
25346 (have_whole_vector_shift): Use vec_perm_builder and
25347 vec_perm_indices instead of auto_vec_perm_indices. Leave the
25348 truncation to calc_vec_perm_mask_for_shift.
25349 (vect_create_epilog_for_reduction): Likewise.
25350 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
25351 from auto_vec_perm_indices to vec_perm_indices.
25352 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
25353 instead of changing individual elements.
25354 (aarch64_vectorize_vec_perm_const): Use new_vector to install
25355 the vector in d.perm.
25356 * config/arm/arm.c (expand_vec_perm_d::perm): Change
25357 from auto_vec_perm_indices to vec_perm_indices.
25358 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
25359 instead of changing individual elements.
25360 (arm_vectorize_vec_perm_const): Use new_vector to install
25361 the vector in d.perm.
25362 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
25363 Update vec_perm_builder constructor.
25364 (rs6000_expand_interleave): Likewise.
25365 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
25366 (rs6000_expand_interleave): Likewise.
25368 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25370 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
25371 to qimode could truncate the indices.
25372 * optabs.c (expand_vec_perm_var): Likewise.
25374 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25376 * Makefile.in (OBJS): Add vec-perm-indices.o.
25377 * vec-perm-indices.h: New file.
25378 * vec-perm-indices.c: Likewise.
25379 * target.h (vec_perm_indices): Replace with a forward class
25381 (auto_vec_perm_indices): Move to vec-perm-indices.h.
25382 * optabs.h: Include vec-perm-indices.h.
25383 (expand_vec_perm): Delete.
25384 (selector_fits_mode_p, expand_vec_perm_var): Declare.
25385 (expand_vec_perm_const): Declare.
25386 * target.def (vec_perm_const_ok): Replace with...
25387 (vec_perm_const): ...this new hook.
25388 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
25389 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
25390 * doc/tm.texi: Regenerate.
25391 * optabs.def (vec_perm_const): Delete.
25392 * doc/md.texi (vec_perm_const): Likewise.
25393 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
25394 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
25395 expand_vec_perm for constant permutation vectors. Assert that
25396 the mode of variable permutation vectors is the integer equivalent
25397 of the mode that is being permuted.
25398 * optabs-query.h (selector_fits_mode_p): Declare.
25399 * optabs-query.c: Include vec-perm-indices.h.
25400 (selector_fits_mode_p): New function.
25401 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
25402 is defined, instead of checking whether the vec_perm_const_optab
25403 exists. Use targetm.vectorize.vec_perm_const instead of
25404 targetm.vectorize.vec_perm_const_ok. Check whether the indices
25405 fit in the vector mode before using a variable permute.
25406 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
25407 vec_perm_indices instead of an rtx.
25408 (expand_vec_perm): Replace with...
25409 (expand_vec_perm_const): ...this new function. Take the selector
25410 as a vec_perm_indices rather than an rtx. Also take the mode of
25411 the selector. Update call to shift_amt_for_vec_perm_mask.
25412 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
25413 Use vec_perm_indices::new_expanded_vector to expand the original
25414 selector into bytes. Check whether the indices fit in the vector
25415 mode before using a variable permute.
25416 (expand_vec_perm_var): Make global.
25417 (expand_mult_highpart): Use expand_vec_perm_const.
25418 * fold-const.c: Includes vec-perm-indices.h.
25419 * tree-ssa-forwprop.c: Likewise.
25420 * tree-vect-data-refs.c: Likewise.
25421 * tree-vect-generic.c: Likewise.
25422 * tree-vect-loop.c: Likewise.
25423 * tree-vect-slp.c: Likewise.
25424 * tree-vect-stmts.c: Likewise.
25425 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
25427 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
25428 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
25429 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
25430 (aarch64_vectorize_vec_perm_const): ...this new function.
25431 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25432 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25433 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
25434 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
25435 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25436 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25437 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
25439 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
25440 check for NEON modes.
25441 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
25442 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
25443 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
25444 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
25446 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
25447 the old VEC_PERM_CONST conditions.
25448 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
25449 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
25450 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
25451 (ia64_vectorize_vec_perm_const_ok): Merge into...
25452 (ia64_vectorize_vec_perm_const): ...this new function.
25453 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
25454 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
25455 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
25456 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
25457 * config/mips/mips.c (mips_expand_vec_perm_const)
25458 (mips_vectorize_vec_perm_const_ok): Merge into...
25459 (mips_vectorize_vec_perm_const): ...this new function.
25460 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
25461 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
25462 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
25463 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
25464 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
25465 (rs6000_expand_vec_perm_const): Delete.
25466 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
25468 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25469 (altivec_expand_vec_perm_const_le): Take each operand individually.
25470 Operate on constant selectors rather than rtxes.
25471 (altivec_expand_vec_perm_const): Likewise. Update call to
25472 altivec_expand_vec_perm_const_le.
25473 (rs6000_expand_vec_perm_const): Delete.
25474 (rs6000_vectorize_vec_perm_const_ok): Delete.
25475 (rs6000_vectorize_vec_perm_const): New function.
25476 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
25477 an element count and rtx array.
25478 (rs6000_expand_extract_even): Update call accordingly.
25479 (rs6000_expand_interleave): Likewise.
25480 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
25481 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
25482 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
25483 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
25484 (rs6000_expand_vec_perm_const): Delete.
25485 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25486 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25487 (altivec_expand_vec_perm_const_le): Take each operand individually.
25488 Operate on constant selectors rather than rtxes.
25489 (altivec_expand_vec_perm_const): Likewise. Update call to
25490 altivec_expand_vec_perm_const_le.
25491 (rs6000_expand_vec_perm_const): Delete.
25492 (rs6000_vectorize_vec_perm_const_ok): Delete.
25493 (rs6000_vectorize_vec_perm_const): New function. Remove stray
25494 reference to the SPE evmerge intructions.
25495 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
25496 an element count and rtx array.
25497 (rs6000_expand_extract_even): Update call accordingly.
25498 (rs6000_expand_interleave): Likewise.
25499 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
25500 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
25502 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25504 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25506 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
25507 vector mode and that that mode matches the mode of the data
25509 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
25510 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
25511 directly using expand_vec_perm_1 when forcing selectors into
25513 (expand_vec_perm_var): New function, split out from expand_vec_perm.
25515 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25517 * optabs-query.h (can_vec_perm_p): Delete.
25518 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
25519 * optabs-query.c (can_vec_perm_p): Split into...
25520 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
25521 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
25522 particular selector is valid.
25523 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25524 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25525 (vect_grouped_load_supported): Likewise.
25526 (vect_shift_permute_load_chain): Likewise.
25527 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
25528 (vect_transform_slp_perm_load): Likewise.
25529 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25530 (vectorizable_bswap): Likewise.
25531 (vect_gen_perm_mask_checked): Likewise.
25532 * fold-const.c (fold_ternary_loc): Likewise. Don't take
25533 implementations of variable permutation vectors into account
25534 when deciding which selector to use.
25535 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
25536 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
25537 with a false third argument.
25538 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
25539 to test whether the constant selector is valid and can_vec_perm_var_p
25540 to test whether a variable selector is valid.
25542 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25544 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
25545 * optabs-query.c (can_vec_perm_p): Likewise.
25546 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
25547 instead of vec_perm_indices.
25548 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
25549 (vect_gen_perm_mask_checked): Likewise,
25550 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
25551 (vect_gen_perm_mask_checked): Likewise,
25553 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25555 * optabs-query.h (qimode_for_vec_perm): Declare.
25556 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
25557 (qimode_for_vec_perm): ...this new function.
25558 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
25560 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25562 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
25563 does not have a conditional at the top.
25565 2018-01-02 Richard Biener <rguenther@suse.de>
25567 * ipa-inline.c (big_speedup_p): Fix expression.
25569 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
25572 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
25575 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
25579 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
25580 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
25581 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
25582 cond_taken_branch_cost 3->4.
25584 2018-01-01 Jakub Jelinek <jakub@redhat.com>
25586 PR tree-optimization/83581
25587 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
25588 TODO_cleanup_cfg if any changes have been made.
25590 PR middle-end/83608
25591 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
25592 convert_modes if target mode has the right side, but different mode
25595 PR middle-end/83609
25596 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
25597 last argument when extracting from CONCAT. If either from_real or
25598 from_imag is NULL, use expansion through memory. If result is not
25599 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
25600 the parts directly to inner mode, if even that fails, use expansion
25603 PR middle-end/83623
25604 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
25605 check for bswap in mode rather than HImode and use that in expand_unop
25608 Copyright (C) 2018 Free Software Foundation, Inc.
25610 Copying and distribution of this file, with or without modification,
25611 are permitted in any medium without royalty provided the copyright
25612 notice and this notice are preserved.