1 2018-07-19 Martin Sebor <msebor@redhat.com>
3 PR tree-optimization/84047
4 PR tree-optimization/83776
5 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
6 (check_array_bounds): Call it.
8 2018-07-19 Martin Sebor <msebor@redhat.com>
10 * align.h (align_flags): Use member initialization.
12 2018-07-19 David Malcolm <dmalcolm@redhat.com>
14 * Makefile.in (OBJS): Add optinfo.o.
15 * coretypes.h (class symtab_node): New forward decl.
16 (struct cgraph_node): New forward decl.
17 (class varpool_node): New forward decl.
18 * dump-context.h: New file.
19 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
21 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
22 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
23 (set_alt_dump_file): Likewise.
24 (dump_context::~dump_context): New dtor.
25 (dump_gimple_stmt): Move implementation to...
26 (dump_context::dump_gimple_stmt): ...this new member function.
27 Add the stmt to any pending optinfo, creating one if need be.
28 (dump_gimple_stmt_loc): Move implementation to...
29 (dump_context::dump_gimple_stmt_loc): ...this new member function.
30 Start a new optinfo and add the stmt to it.
31 (dump_gimple_expr): Move implementation to...
32 (dump_context::dump_gimple_expr): ...this new member function.
33 Add the stmt to any pending optinfo, creating one if need be.
34 (dump_gimple_expr_loc): Move implementation to...
35 (dump_context::dump_gimple_expr_loc): ...this new member function.
36 Start a new optinfo and add the stmt to it.
37 (dump_generic_expr): Move implementation to...
38 (dump_context::dump_generic_expr): ...this new member function.
39 Add the tree to any pending optinfo, creating one if need be.
40 (dump_generic_expr_loc): Move implementation to...
41 (dump_context::dump_generic_expr_loc): ...this new member
42 function. Add the tree to any pending optinfo, creating one if
44 (dump_printf): Move implementation to...
45 (dump_context::dump_printf_va): ...this new member function. Add
46 the text to any pending optinfo, creating one if need be.
47 (dump_printf_loc): Move implementation to...
48 (dump_context::dump_printf_loc_va): ...this new member function.
49 Start a new optinfo and add the stmt to it.
50 (dump_dec): Move implementation to...
51 (dump_context::dump_dec): ...this new member function. Add the
52 value to any pending optinfo, creating one if need be.
53 (dump_context::dump_symtab_node): New member function.
54 (dump_context::get_scope_depth): New member function.
55 (dump_context::begin_scope): New member function.
56 (dump_context::end_scope): New member function.
57 (dump_context::ensure_pending_optinfo): New member function.
58 (dump_context::begin_next_optinfo): New member function.
59 (dump_context::end_any_optinfo): New member function.
60 (dump_context::s_current): New global.
61 (dump_context::s_default): New global.
62 (dump_scope_depth): Delete global.
63 (dumpfile_ensure_any_optinfo_are_flushed): New function.
64 (dump_symtab_node): New function.
65 (get_dump_scope_depth): Reimplement in terms of dump_context.
66 (dump_begin_scope): Likewise.
67 (dump_end_scope): Likewise.
68 (selftest::temp_dump_context::temp_dump_context): New ctor.
69 (selftest::temp_dump_context::~temp_dump_context): New dtor.
70 (selftest::verify_item): New function.
71 (ASSERT_IS_TEXT): New macro.
72 (ASSERT_IS_TREE): New macro.
73 (ASSERT_IS_GIMPLE): New macro.
74 (selftest::test_capture_of_dump_calls): New test.
75 (selftest::dumpfile_c_tests): Call it.
76 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
77 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
78 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
80 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
81 (dump_node, dump_bb): Move these unrelated decls.
82 (class dump_manager): Add leading comment.
83 * optinfo.cc: New file.
84 * optinfo.h: New file.
86 2018-07-19 Michael Collison <michael.collison@arm.com>
87 Richard Henderson <rth@redhat.com>
89 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
90 (subti): Handle op1 zero.
91 (subvti4, usub4ti4): New.
92 (*sub<GPI>3_compare1_imm): New.
93 (sub<GPI>3_carryinCV): New.
94 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
95 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
97 2018-07-19 Michael Collison <michael.collison@arm.com>
98 Richard Henderson <rth@redhat.com>
100 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
101 (addti3): Create simpler code if low part is already known to be 0.
102 (addvti4, uaddvti4): New.
103 (*add<GPI>3_compareC_cconly_imm): New.
104 (*add<GPI>3_compareC_cconly): New.
105 (*add<GPI>3_compareC_imm): New.
106 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
107 handle constants within this pattern..
108 (*add<GPI>3_compareV_cconly_imm): New.
109 (*add<GPI>3_compareV_cconly): New.
110 (*add<GPI>3_compareV_imm): New.
111 (add<GPI>3_compareV): New.
112 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
113 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
114 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
115 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
116 with 'comparison' operator.
117 (*add<GPI>3_compareV_cconly_imm): Ditto.
118 (*add<GPI>3_compareV_cconly): Ditto.
119 (*add<GPI>3_compareV_imm): Ditto.
120 (add<GPI>3_compareV): Ditto.
121 (add<mode>3_carryinC): Ditto.
122 (*add<mode>3_carryinC_zero): Ditto.
123 (*add<mode>3_carryinC): Ditto.
124 (add<mode>3_carryinV): Ditto.
125 (*add<mode>3_carryinV_zero): Ditto.
126 (*add<mode>3_carryinV): Ditto.
128 2018-07-19 Michael Collison <michael.collison@arm.com>
129 Richard Henderson <rth@redhat.com>
131 * config/aarch64/aarch64-modes.def (CC_V): New.
132 * config/aarch64/aarch64-protos.h
133 (aarch64_addti_scratch_regs): Declare
134 (aarch64_subvti_scratch_regs): Declare.
135 (aarch64_expand_subvti): Declare.
136 (aarch64_gen_unlikely_cbranch): Declare
137 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
138 for signed overflow using CC_Vmode.
139 (aarch64_get_condition_code_1): Handle CC_Vmode.
140 (aarch64_gen_unlikely_cbranch): New function.
141 (aarch64_addti_scratch_regs): New function.
142 (aarch64_subvti_scratch_regs): New function.
143 (aarch64_expand_subvti): New function.
145 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
147 * config/aarch64/aarch64-option-extensions.def: New entry for profile
149 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
150 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
153 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
156 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
157 address check not strict.
159 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
161 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
163 (store_pair_lanes<mode>): Likewise.
164 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
165 enum value 'ADDR_QUERY_LDP_STP_N'.
166 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
167 (aarch64_print_address_internal): Add declaration.
168 (aarch64_print_ldpstp_address): Remove.
169 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
170 (aarch64_print_operand): Change printing of 'y'.
171 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
172 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
173 'true' rather than '1'.
174 * gcc/config/aarch64/constraints.md (Uml): Likewise.
175 (Uml): Rename to Umn.
178 2018-07-19 Richard Biener <rguenther@suse.de>
180 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
182 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
183 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
184 (vn_reference_hasher): Likewise.
185 (struct vn_tables_s): Remove obstack and alloc-pool members.
186 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
187 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
188 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
189 (vn_reference_insert_pieces): Likewise.
190 (alloc_vn_nary_op_noinit): Adjust.
191 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
193 (shared_lookup_phiargs): Remove.
194 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
195 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
196 (visit_reference_op_call): Likewise.
197 (copy_nary, copy_phi, copy_reference): Remove.
198 (process_scc): Rewind the obstack when iterating. Do not
199 copy the elements to valid_info but just move them from one
200 hashtable to the other.
201 (allocate_vn_table): Adjust.
202 (free_vn_table): Likewise.
203 (init_scc_vn): Likewise.
204 (free_scc_vn): Likewise.
206 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
209 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
210 indirect_return as function type attribute.
211 (ix86_attribute_table): Change indirect_return to function
213 * doc/extend.texi: Update indirect_return attribute.
215 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
217 * wide-int.h (widest2_int): New.
218 * gimple-fold.c (arith_overflowed_p): Use it.
219 * tree.h (widest2_int_cst): New.
220 * tree-vrp.c (wide_int_binop_overflow): Rename from
222 Rewrite to work on trees.
223 (extract_range_from_multiplicative_op_1): Abstract code to...
224 (wide_int_range_min_max): ...here.
225 (wide_int_range_cross_product): ...and here.
226 (extract_range_from_binary_expr_1): Abstract overflow code to...
227 (wide_int_range_mult_wrapping): ...here.
228 * tree-vrp.h (wide_int_range_cross_product): New.
229 (wide_int_range_mult_wrapping): New.
231 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
232 Julia Koval <julia.koval@intel.com>
234 * config/i386/x86-tune-costs.h (skylake_memcpy,
235 skylake_memset): Replace rep_prefix with unrolling for size 512.
237 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
240 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle comparision with EQ_EXPR
243 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
245 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
246 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
247 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
248 previously known as "PowerPC AltiVec Built-in Functions". Move
249 some material to new subsubsections "PowerPC AltiVec Built-in
250 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
252 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
253 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
254 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
255 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
257 2018-07-18 Richard Biener <rguenther@suse.de>
259 PR tree-optimization/86557
260 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
263 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
265 * config/s390/s390.c (s390_function_profiler): Generate CFI.
267 2018-07-17 Jeff Law <law@redhat.com>
269 * config/arm/arm.c (get_label_padding): Update for recent
270 changes to label_to_alignment.
272 PR tree-optimization/86010
273 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
275 * config/mips/mips.c (vr4130_align_insns): Update for recent
276 changes to label_to_alignment.
278 * config/frv/frv.c (frv_label_align): Update for recent changes
279 to label_to_alignment.
281 * config/nios2/nios2.c (nios2_label_align): Update for recent
282 changes which dropped ALIGN_LABELS_LOG.
284 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
286 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
287 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
289 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
291 * config/arc/arc.c (arc_label_align): Use align_labels instead of
292 deprecated align_labels_log.
294 2018-07-17 Richard Biener <rguenther@suse.de>
297 * dwarf2out.c (init_sections_and_labels): Always generate
298 a debug_line_str_section for early LTO debug.
299 (dwarf2out_finish): Reset debug_line_str_hash output early.
300 Bump counter for extra dwarf5 .debug_loc labels to not conflict
302 (dwarf2out_early_finish): Output debug_line_str.
304 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
306 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
307 index register on z196 or later.
309 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
311 * config/s390/s390.c (s390_default_align): Set default function
313 (s390_override_options_after_change): Call s390_default align.
314 (s390_option_override_internal): Call s390_default_align.
315 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
317 2018-07-17 Jakub Jelinek <jakub@redhat.com>
320 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
321 to _looptemp_ clauses, other than the first two.
323 2018-07-17 Martin Liska <mliska@suse.cz>
325 * opts.c: Do not enable OPT_falign_* for -Os.
327 2018-07-17 Martin Liska <mliska@suse.cz>
329 * align.h (MAX_CODE_ALIGN): New.
330 (MAX_CODE_ALIGN_VALUE): New.
331 * common/config/i386/i386-common.c (ix86_handle_option):
332 (MAX_CODE_ALIGN): Moved to align.h.
333 * final.c (MAX_CODE_ALIGN): Likewise.
334 * opts.c (parse_and_check_align_values):
335 (MAX_CODE_ALIGN): Likewise.
336 (MAX_CODE_ALIGN_VALUE): Likewise.
338 2018-07-17 Martin Liska <mliska@suse.cz>
340 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
341 in order to fulfil coding style.
342 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
343 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
344 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
345 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
346 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
347 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
348 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
349 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
350 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
352 2018-07-17 Martin Liska <mliska@suse.cz>
355 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions directly.
356 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
357 align_flags of label_to_alignment.
358 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into align_flags
360 * config/m68k/m68k.c: Do not use removed align_labels_value and
362 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
363 (LOOP_ALIGN): Likewise.
364 (LABEL_ALIGN): Likewise.
365 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
366 Remove not used macro.
367 (rs6000_loop_align): Change return type to align_flags.
368 (rs6000_loop_align_max_skip): Remove.
369 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
370 Change return type to align_flags.
371 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
372 Remove not used macro.
373 (rs6000_loop_align): Change return type to align_flags.
374 (rs6000_loop_align_max_skip): Remove.
375 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
376 * config/rx/rx-protos.h (rx_align_for_label): Make it
378 * config/rx/rx.c (rx_align_for_label): Change return type
380 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
382 into align_flags class.
383 (LABEL_ALIGN): Likewise.
384 (LOOP_ALIGN): Likewise.
385 * config/s390/s390.c (s390_label_align): Use align_flags
387 (s390_asm_output_function_label): Likewise.
388 * config/sh/sh.c (sh_override_options_after_change):
389 Use align_flags class directly without macros.
390 (find_barrier): Likewise.
391 (barrier_align): Likewise.
392 (sh_loop_align): Likewise.
393 * config/spu/spu.c (spu_option_override):
394 Use align_flags_tuple::get_value instead of removed macros.
395 (spu_sched_init): Likewise.
396 * config/spu/spu.h (GTY): Likewise.
397 * config/visium/visium.c (visium_option_override):
398 Set "8" as default secondary alignment.
399 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
400 in order to guarantee secondary alignment of 8.
401 * coretypes.h: Include align.h header file.
402 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
403 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
404 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
405 * doc/tm.texi.in: Likewise.
406 * final.c (struct label_alignment): Remove not used structure.
407 (LABEL_ALIGN): Change type to align_flags.
408 (LOOP_ALIGN): Likewise.
409 (JUMP_ALIGN): Likewise.
410 (default_loop_align_max_skip): Remove.
411 (default_label_align_max_skip): Likewise.
412 (default_jump_align_max_skip): Likewise.
413 (default_label_align_after_barrier_max_skip):
414 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
415 (LABEL_TO_MAX_SKIP): Remove.
416 (label_to_alignment): Return align_flags type instead of integer.
417 (label_to_max_skip): Remove.
418 (align_fuzz): Use align_flags type.
419 (compute_alignments): Use align_flags type and use align_flags::max
420 to combine multiple alignments.
421 (grow_label_align): Grow vec instead of C array.
422 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
423 (shorten_branches): Use align_flags type and use align_flags::max
424 to combine multiple alignments.
425 (final_scan_insn_1): Remove usage of secondary alignment that comes
426 from label alignment, but instead use proper secondary alignment
427 which is computed in grow_label_align.
428 * flags.h (struct align_flags_tuple): Move to align.h.
429 (struct align_flags): Likewise.
430 (state_align_loops): Rename to align_loops.
431 (state_align_jumps): Rename to align_jumps.
432 (state_align_labels): Rename to align_labels.
433 (state_align_functions): Rename to align_functions.
434 (align_loops_log): Remove.
435 (align_jumps_log): Remove.
436 (align_labels_log): Remove.
437 (align_functions_log): Remove.
438 (align_loops_max_skip): Remove.
439 (align_jumps_max_skip): Remove.
440 (align_labels_max_skip): Remove.
441 (align_functions_max_skip): Remove.
442 (align_loops_value): Remove.
443 (align_jumps_value): Remove.
444 (align_labels_value): Remove.
445 (align_functions_value): Remove.
446 * output.h (label_to_alignment): Change return type to align_flags.
447 (label_to_max_skip): Remove.
448 * target.def: Remove loop_align_max_skip, label_align_max_skip,
449 jump_align_max_skip macros.
450 * targhooks.h (default_loop_align_max_skip): Remove.
451 (default_label_align_max_skip): Likewise.
452 (default_jump_align_max_skip): Likewise.
453 (default_label_align_after_barrier_max_skip): Remove.
454 * toplev.c (read_log_maxskip): Use ::normalize function.
455 (parse_N_M): Remove not used argument and also call ::normalize.
456 (parse_alignment_opts): Do not pass unused arguments.
457 * varasm.c (assemble_start_function): Use directly align_functions
458 instead of removed macros.
459 * system.h: Do not poison removed macros.
461 2018-07-17 Jakub Jelinek <jakub@redhat.com>
464 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
465 and cond temporaries don't have reference type if iterator has
466 pointer type. For init use &for_pre_body instead of pre_p if
467 for_pre_body is non-empty.
469 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
471 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
472 double-double modes to SFmode directly directly.
473 (trunc<mode>sf2_fprs): Delete.
475 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
477 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
478 for conversions between IFmode and the decimal floating point modes.
479 (init_float128_ieee): Use the correct names for conversions between
480 KFmode and the decimal floating point modes.
482 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
484 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
485 for the conversions between TDmode and IFmode.
486 (init_float128_ieee): Use more correct names for the conversions
487 between TDmode and KFmode.
489 2018-07-16 Jakub Jelinek <jakub@redhat.com>
491 PR tree-optimization/86526
492 * builtins.c (expand_builtin_memcmp): Formatting fixes.
493 (inline_expand_builtin_string_cmp): Likewise.
494 (inline_string_cmp): Likewise. Use c_readstr instead of
495 builtin_memcpy_read_str. Add unit_mode temporary.
497 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
500 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
501 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
503 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
505 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
506 Alphabetize prototypes of built-in functions, separating out
507 built-in functions that are listed in this section but should be
510 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
513 * expmed.c (emit_store_flag): Do not emit setcc followed by a
514 conditional move when trapping comparison was split to a
515 non-trapping one (and vice versa).
517 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
519 * config/s390/s390.c (s390_function_profiler): Generate nops
520 instead of profiler call sequences.
521 * config/s390/s390.opt: Add the new option.
523 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
525 * config/s390/s390.c (s390_function_profiler): Generate
526 __mcount_loc section.
527 * config/s390/s390.opt: Add the new option.
529 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
531 * common.opt: Add the new warning.
532 * config/s390/s390.c (s390_function_profiler): Emit "brasl
533 %r0,__fentry__" when -mfentry is specified.
534 (s390_option_override_internal): Disallow -mfentry for 31-bit
536 * config/s390/s390.opt: Add the new option.
538 2018-07-16 Richard Biener <rguenther@suse.de>
541 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
542 for function-local FUNCTION_DECL and RESULT_DECL immediately.
544 2018-07-16 Martin Liska <mliska@suse.cz>
547 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
550 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
552 * config/arc/arcHS.md: Update ARCHS scheduling rules.
554 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
556 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
558 * config/arc/arc-cpus.def (hs4x): New cpu.
560 * config/arc/arc-tables.opt: Regenerate.
561 * config/arc/arc.c (arc_sched_issue_rate): New function.
562 (TARGET_SCHED_ISSUE_RATE): Define.
563 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
564 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
566 (attr tune): Add ARCHS4x tune values.
567 (attr tune_dspmpy): Define.
568 (*tst): Correct instruction type.
569 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
570 * config/arc/arcHS4x.md: New file.
571 * config/arc/fpu.md: Update instruction type attributes.
572 * config/arc/t-multilib: Regenerate.
574 2018-07-16 Tom de Vries <tdevries@suse.de>
577 * var-tracking.c (vt_initialize): Fix pre_dec handling.
579 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
581 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
584 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
586 PR tree-optimization/86514
587 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
588 conversion to a boolean type from a type with greater precision.
590 2018-07-16 Tom de Vries <tdevries@suse.de>
592 * var-tracking.c (vt_initialize): Print adjusted insn slim if
593 dump_flags request TDF_SLIM.
595 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
597 * fold-const.c (int_const_binop_1): Abstract...
598 (wide_int_binop): ...wide int code here.
599 (poly_int_binop): ...poly int code here.
600 Abstract the rest of int_const_binop_1 into int_const_binop.
601 * fold-const.h (wide_int_binop): New.
602 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
603 Remove useless PLUS/MINUS_EXPR case.
604 (zero_nonzero_bits_from_vr): Move wide int code...
605 (zero_nonzero_bits_from_bounds): ...here.
606 (extract_range_from_binary_expr_1): Move mask optimization code...
607 (range_easy_mask_min_max): ...here.
608 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
609 (range_easy_mask_min_max): New.
611 2018-07-15 Jeff Law <law@redhat.com>
614 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
617 2018-07-14 Jim Wilson <jimw@sifive.com>
619 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
621 2018-07-14 Paul Koning <ni1d@arrl.net>
623 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
625 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
627 * lto-streamer-out.c (copy_function_or_variable): Dump info about
630 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
631 Steve Munroe <munroesj52@gmail.com>
633 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
634 (_mm_andnot_si128): Likewise.
635 (_mm_or_si128): Likewise.
636 (_mm_xor_si128): Likewise.
638 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
641 * builtins.c (expand_builtin_memcmp): Inline the calls first
642 when result_eq is false.
643 (expand_builtin_strcmp): Inline the calls first.
644 (expand_builtin_strncmp): Likewise.
645 (inline_string_cmp): New routine. Expand a string compare
646 call by using a sequence of char comparison.
647 (inline_expand_builtin_string_cmp): New routine. Inline expansion
648 a call to str(n)cmp/memcmp.
649 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
651 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
653 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
655 * config/arm/driver-arm.c: Include arm-native.h.
656 (host_detect_local_cpu): Use auto-generated data tables.
657 (vendors, arm_cpu_table): Delete. Move part information to ...
658 * config/arm/arm-cpus.in: ... here.
659 * config/arm/parsecpu.awk (gen_native): New function.
660 (vendor, part): New CPU fields.
661 (END): Add support for building the native CPU detection tables.
662 * config/arm/t-arm (arm-native.h): Add build rule.
663 (driver-arm.o): Add dependency on arm-native.h.
665 2018-07-13 Richard Biener <rguenther@suse.de>
668 * match.pd (addr1 - addr2): Allow either of the operand to
671 2018-07-13 Tom de Vries <tdevries@suse.de>
673 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
676 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
678 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
679 arrays instead of numbered variables.
681 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
683 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
684 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
686 2018-07-13 Richard Biener <rguenther@suse.de>
689 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
690 instead of get_context_die.
692 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
693 Richard Biener <rguenther@suse.de>
696 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
697 that the loop latch destination where phi is defined.
699 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
701 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
702 (riscv_expand_epilogue): Add assertion to check interrupt mode.
703 (riscv_set_current_function): Extract getting interrupt type to new
705 (riscv_get_interrupt_type): New function.
706 (riscv_merge_decl_attributes): New function, checking interrupt type is
708 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
710 2018-07-12 Paul Koning <ni1d@arrl.net>
712 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
715 2018-07-12 Paul Koning <ni1d@arrl.net>
717 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
718 zero reference, add doloop_end instead.
719 * doc/md.texi (decrement_and_branch_until_zero): Remove.
720 (Looping patterns): Remove decrement_and_branch_until_zero. Add
721 detail for doloop_end.
723 2018-07-12 Martin Sebor <msebor@redhat.com>
726 * attribs.c (decl_attributes): Reject conflicting attributes before
727 calling attribute handlers.
729 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
731 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
733 (gcc::dump_manager::get_dump_file_name): likewise.
734 (dump_begin): Likewise.
735 * dumpfile.h (dump_begin): Update prototype.
736 (gcc::dump_manager::get_dump_file_name,
737 gcc::dump_manager::get_dump_file_name): Update prototype.
739 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
741 * internal-fn.h (vectorizable_internal_fn_p): New function.
742 * tree-vect-slp.c (compatible_calls_p): Likewise.
743 (vect_build_slp_tree_1): Remove nops argument. Handle calls
744 to internal functions.
745 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
747 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
749 * fold-const.h (inverse_conditions_p): Declare.
750 * fold-const.c (inverse_conditions_p): New function.
751 * match.pd: Use inverse_conditions_p. Add folds of view_converts
752 that test the inverse condition of a conditional internal function.
753 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
754 * internal-fn.c (internal_fn_mask_index): Handle conditional
756 (vectorized_internal_fn_supported_p): New function.
757 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
758 (any_pred_load_store): Replace with...
759 (need_to_predicate): ...this new variable.
760 (redundant_ssa_names): New variable.
761 (ifcvt_can_use_mask_load_store): Move initial checks to...
762 (ifcvt_can_predicate): ...this new function. Handle tree codes
763 for which a conditional internal function exists.
764 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
765 instead of ifcvt_can_use_mask_load_store. Update after variable
767 (predicate_load_or_store): New function, split out from
768 predicate_mem_writes.
769 (check_redundant_cond_expr): New function.
770 (value_available_p): Likewise.
771 (predicate_rhs_code): Likewise.
772 (predicate_mem_writes): Rename to...
773 (predicate_statements): ...this. Use predicate_load_or_store
774 and predicate_rhs_code.
775 (combine_blocks, tree_if_conversion): Update after above name changes.
776 (ifcvt_local_dce): Handle redundant_ssa_names.
777 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
778 general conditional functions.
779 * tree-vect-stmts.c (vectorizable_call): Likewise.
781 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
782 Alan Hayward <alan.hayward@arm.com>
783 David Sherwood <david.sherwood@arm.com>
785 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
786 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
787 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
788 plus and minus and convert them into IFN_COND_FMA-based sequences.
789 (convert_mult_to_fma): Handle conditional plus and minus.
791 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
793 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
794 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
795 (cond_fnms_optab): New optabs.
796 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
798 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
799 * internal-fn.h (get_conditional_internal_fn): Declare.
800 (get_unconditional_internal_fn): Likewise.
801 * internal-fn.c (cond_ternary_direct): New macro.
802 (expand_cond_ternary_optab_fn): Likewise.
803 (direct_cond_ternary_optab_supported_p): Likewise.
804 (FOR_EACH_COND_FN_PAIR): Likewise.
805 (get_conditional_internal_fn): New function.
806 (get_unconditional_internal_fn): Likewise.
807 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
808 (gimple_match_op::gimple_match_op): Add a new overload for 5
810 (gimple_match_op::set_op): Likewise.
811 (gimple_resimplify5): Declare.
812 * genmatch.c (decision_tree::gen): Generate simplifications for
814 * gimple-match-head.c (gimple_simplify): Define an overload for
815 5 operands. Handle calls with 5 arguments in the top-level overload.
816 (convert_conditional_op): Handle conversions from unconditional
817 internal functions to conditional ones.
818 (gimple_resimplify5): New function.
819 (build_call_internal): Pass a fifth operand.
820 (maybe_push_res_to_seq): Likewise.
821 (try_conditional_simplification): Try converting conditional
822 internal functions to unconditional internal functions.
823 Handle 3-operand unconditional forms.
824 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
825 Define ternary equivalents of the current rules for binary conditional
827 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
829 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
830 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
831 (optab): Handle them.
832 (SVE_COND_FP_TERNARY): New int iterator.
833 (sve_fmla_op, sve_fmad_op): New int attributes.
834 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
835 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
836 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
838 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
840 * target.def (preferred_else_value): New target hook.
841 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
842 * doc/tm.texi: Regenerate.
843 * targhooks.h (default_preferred_else_value): Declare.
844 * targhooks.c (default_preferred_else_value): New function.
845 * internal-fn.h (conditional_internal_fn_code): Declare.
846 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
847 (get_conditional_internal_fn): Use it.
848 (conditional_internal_fn_code): New function.
849 * gimple-match.h (gimple_match_cond): New struct.
850 (gimple_match_op): Add a cond member function.
851 (gimple_match_op::gimple_match_op): Update all forms to take a
853 * genmatch.c (expr::gen_transform): Use the same condition as res_op
854 for the suboperation, but don't specify a particular else_value.
855 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
856 (visit_nary_op, visit_reference_op_load): Pass
857 gimple_match_cond::UNCOND to the gimple_match_op constructor.
858 * gimple-match-head.c: Include tree-eh.h
859 (convert_conditional_op): New function.
860 (maybe_resimplify_conditional_op): Likewise.
861 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
862 (gimple_resimplify2): Likewise.
863 (gimple_resimplify3): Likewise.
864 (gimple_resimplify4): Likewise.
865 (maybe_push_res_to_seq): Return null for conditional operations.
866 (try_conditional_simplification): New function.
867 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
869 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
871 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
873 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
875 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
877 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
879 (hash_tree): Do not hash DECL_FCONTEXT
880 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
881 Do not stream DECL_FCONTEXT.
882 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
883 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
885 2018-07-12 Richard Biener <rguenther@suse.de>
888 * dwarf2out.c (gen_block_die): Only output blocks when they have
889 at least one !DECL_IGNORED_P variable.
891 2018-07-12 Richard Biener <rguenther@suse.de>
894 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
895 Remove -mieee-fp handling.
897 2018-07-12 Richard Biener <rguenther@suse.de>
899 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
900 left-over from last patch.
902 2018-07-12 Jakub Jelinek <jakub@redhat.com>
904 PR tree-optimization/86492
905 * gimple-ssa-store-merging.c
906 (imm_store_chain_info::coalesce_immediate_stores): Call
907 check_no_overlap even for the merge_overlapping case. Formatting fix.
909 2018-07-12 Richard Biener <rguenther@suse.de>
912 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
913 move possibly trapping operations into the conditional.
915 2018-07-12 Richard Biener <rguenther@suse.de>
917 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
918 (vn_lookup_simplify_result): Remove recursion limit applied
920 (vn_nary_build_or_lookup_1): Adjust.
921 (try_to_simplify): Likewise.
922 * gimple-match-head.c (gimple_resimplify1): Instead apply one
924 (gimple_resimplify2): Likewise.
925 (gimple_resimplify3): Likewise.
926 (gimple_resimplify4): Likewise.
928 2018-07-11 Jakub Jelinek <jakub@redhat.com>
930 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
931 Use __mmask64 type instead of __mmask8 for __M argument.
932 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
933 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
935 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
936 __mmask16 for __M argument.
937 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
938 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
939 to __mmask16 instead of __mmask8.
940 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
941 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
942 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
943 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
944 instead of __mmask16 for __U argument.
945 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
946 __mmask16 instead of __mmask8 for __U argument.
947 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
949 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
951 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
953 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
955 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
957 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
959 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
961 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
963 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
965 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
967 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
968 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
969 return type as well as __M argument type and all casts from __mmask8
971 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
972 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
973 return type as well as __M argument type and all casts from __mmask8
975 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
976 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
977 return type as well as __M argument type and all casts from __mmask8
979 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
980 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
981 return type as well as __M argument type and all casts from __mmask8
983 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
984 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
987 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
989 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
990 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
993 2018-07-11 Paul Koning <ni1d@arrl.net>
995 * doc/md.texi (define_subst): Document how multiple occurrences of
996 the same argument in the replacement pattern are handled.
998 2018-07-11 Paul Koning <ni1d@arrl.net>
1000 * doc/extend.texi (Common Variable Attributes): Move "mode" into
1002 (Common Type Attributes): Add "mode" attribute.
1004 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
1006 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
1007 stream DECL_ORIGINAL_TYPE.
1008 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
1009 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
1010 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
1011 Do not walk original type.
1012 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
1014 (write_ts_decl_non_common_tree_pointers): Do not stream
1016 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
1017 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
1019 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
1021 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
1022 thread twice from the same starting edge.
1024 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
1026 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
1027 * gimple.c (gimple_call_nonnull_result_p): ...here...
1028 (gimple_call_nonnull_arg): ...and here.
1029 * gimple.h (gimple_call_nonnull_result_p): New.
1030 (gimple_call_nonnull_arg): New.
1032 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
1034 * config/arm/arm-cpus.in: Move information from fpu field of each
1035 cpu definition to the isa field.
1036 * config/arm/parsecpu.awk (fpu): Delete match rule.
1037 (gen_comm_data): Don't add bits from the CPU's FPU entry.
1039 2018-07-11 Richard Biener <rguenther@suse.de>
1042 * dwarf2out.c (init_sections_and_labels): Use
1043 output_asm_line_debug_info consistently.
1044 (dwarf2out_early_finish): Likewise.
1045 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
1048 2018-07-11 Richard Biener <rguenther@suse.de>
1050 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
1051 Rework father_bb setting in a way to avoid propagating constants
1052 multiple times on a loop body.
1054 2018-07-10 Mark Wielaard <mark@klomp.org>
1057 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
1059 2018-07-10 Richard Biener <rguenther@suse.de>
1061 * hash-map.h (hash_map::iterator::operator*): Return
1062 references to key and value.
1064 2018-07-10 Jakub Jelinek <jakub@redhat.com>
1067 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
1068 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
1069 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
1070 (gimplify_omp_for): For composite loops, move outer
1071 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
1072 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
1073 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
1074 TREE_LIST for both the original class iterator and the "last" helper
1075 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
1076 loop, remember has_decl_expr from outer composite loops for the
1077 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
1079 2018-07-09 Martin Sebor <msebor@redhat.com>
1083 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
1084 accessing implicitly initialized array elements.
1085 * expr.c (string_constant): Handle string initializers of
1086 character arrays within aggregates.
1087 * gimple-fold.c (fold_array_ctor_reference): Add argument.
1088 Store element offset. As a special case, handle zero size.
1089 (fold_nonarray_ctor_reference): Same.
1090 (fold_ctor_reference): Add argument. Store subobject offset.
1091 * gimple-fold.h (fold_ctor_reference): Add argument.
1093 2018-07-09 Paul Koning <ni1d@arrl.net>
1095 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
1096 (pdp11_insn_cost): New function.
1097 (pdp11_md_asm_adjust): New function.
1098 (TARGET_INVALID_WITHIN_DOLOOP): Define.
1099 (pdp11_rtx_costs): Update to match machine better.
1100 (output_addr_const_pdp11): Correct format mismatch warnings.
1101 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
1102 * config/pdp11/pdp11.md: General change to add base_cost and/or
1103 length attributes for use by new pdp11_insn_cost function.
1104 (MIN_BRANCH): Correct definition.
1106 (doloop_end): Use standard pattern name for looping pattern.
1107 (doloop_end_nocc): New.
1108 (movsf): Add another constraint alternative.
1109 (zero_extendqihi2): Add constraint alternatives for not in place
1111 (zero_extendhisi2): Remove.
1112 (shift patterns): Add CC handling variants.
1116 (define_peephole2): New peephole to recognize mov that sets CC for
1119 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1121 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
1122 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
1125 2018-07-09 Jakub Jelinek <jakub@redhat.com>
1128 * real.c (real_nextafter): Return true if result is denormal.
1130 2018-07-09 Martin Liska <mliska@suse.cz>
1132 * common.opt: Add back wrongly removed attribute.
1134 2018-07-09 Richard Biener <rguenther@suse.de>
1137 * dwarf2out.c (gen_block_die): For an early generated DIE
1138 always output high/low PC attributes.
1140 2018-07-09 Tom de Vries <tdevries@suse.de>
1142 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
1143 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
1144 onto VAR_DECL with abstract origin.
1146 2018-07-07 Jim Wilson <jimw@sifive.com>
1148 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
1150 2018-07-07 Tom de Vries <tdevries@suse.de>
1152 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
1154 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
1156 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
1158 (combine_bound): Use wide-int overflow calculation instead of
1160 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
1162 * fold-const.c (int_const_binop_2): Same.
1163 (extract_muldiv_1): Same.
1164 (fold_div_compare): Same.
1165 (fold_abs_const): Same.
1167 * poly-int.h (add): Same.
1171 * predict.c (predict_iv_comparison): Same.
1172 * profile-count.c (slow_safe_scale_64bit): Same.
1173 * simplify-rtx.c (simplify_const_binary_operation): Same.
1174 * tree-chrec.c (tree_fold_binomial): Same.
1175 * tree-data-ref.c (split_constant_offset_1): Same.
1176 * tree-if-conv.c (idx_within_array_bound): Same.
1177 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
1178 * tree-ssa-phiopt.c (minmax_replacement): Same.
1179 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
1180 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
1181 * vr-values.c (vr_values::adjust_range_with_scev): Same.
1182 * wide-int.cc (wi::add_large): Same.
1183 (wi::mul_internal): Same.
1184 (wi::sub_large): Same.
1185 (wi::divmod_internal): Same.
1186 * wide-int.h: Change overflow type to overflow_type for neg, add,
1187 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
1188 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
1189 mul_internal, divmod_internal.
1190 (overflow_type): New enum.
1191 (accumulate_overflow): New.
1193 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
1195 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
1196 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
1198 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
1200 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
1201 argument is checked for zero before entering loop, avoid checking again.
1203 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
1205 * gimplify.h (generic_expr_could_trap_p): Set as global function.
1206 * gimplify.h (generic_expr_could_trap_p): Likwise.
1207 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
1209 2018-07-06 Jakub Jelinek <jakub@redhat.com>
1211 PR tree-optimization/86401
1212 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
1213 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
1214 (fold_bit_and_mask): ... here. New helper function for match.pd.
1215 * fold-const.h (fold_bit_and_mask): Declare.
1216 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
1218 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
1221 * target.def (translate_mode_attribute): New hook.
1222 * targhooks.h (default_translate_mode_attribute): Declare.
1223 * targhooks.c (default_translate_mode_attribute): New function.
1224 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
1225 * doc/tm.texi: Regenerate.
1226 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
1227 (rs6000_translate_mode_attribute): New function.
1229 2018-07-06 Paul Koning <ni1d@arrl.net>
1231 * doc/md.texi (define_split): Document DONE and FAIL.
1232 (define_peephole2): Ditto.
1234 2018-07-05 Jeff Law <law@redhat.com>
1236 PR tree-optimization/86010
1237 * tree-ssa-dse.c (compute_trims): More aggressively trim at
1238 both the head and tail of mem* and str* calls.
1240 2018-07-05 Jim Wilson <jimw@sifive.com>
1242 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
1244 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
1246 * config/aarch64/aarch64-simd.md: correct flags text for
1247 MIN_EXPR replacement.
1249 2018-07-05 James Clarke <jrtc27@jrtc27.com>
1251 * configure: Regenerated.
1253 2018-07-05 Carl Love <cel@us.ibm.com>
1255 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
1256 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
1257 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
1258 VSX_BUILTIN_DOUBLEL_V4SF.
1260 2018-07-05 Martin Sebor <msebor@redhat.com>
1263 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
1264 than its domain to compute its the upper bound of a char array.
1266 2018-07-05 Nathan Sidwell <nathan@acm.org>
1268 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
1269 * doc/cpp.texi: Update comment.
1270 * doc/tm.texi: Rebuilt.
1271 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
1272 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
1273 * doc/extend.texi (Backwards Compatibility): Clarify it is system
1274 headers affected by extern "C".
1275 * system.h: Poison NO_IMPLICIT_EXTERN_C.
1276 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
1277 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
1278 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
1279 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
1280 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
1281 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
1282 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
1283 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
1284 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
1285 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
1286 config/sparc/sp64-elf.h, config/spu/spu.h,
1287 config/stormy16/stormy16.h, config/v850/v850.h,
1288 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
1289 define NO_IMPLICIT_EXTERN_C.
1290 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
1292 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1295 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1296 instead of GET_MODE_SIZE when comparing Units.
1298 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1301 * rtlanal.c (set_noop_p): Constrain on mode change,
1302 include hard-reg-set.h
1304 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1306 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
1308 2018-07-05 Jakub Jelinek <jakub@redhat.com>
1311 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
1314 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
1316 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
1318 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
1321 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
1323 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
1325 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
1328 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
1330 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
1332 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
1333 neg_*_op* variables.
1335 2018-07-04 Martin Liska <mliska@suse.cz>
1337 * tree-switch-conversion.c: Define
1338 max_ratio_for_speed and max_ratio_for_size constants.
1340 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
1341 Martin Liska <mliska@suse.cz>
1346 * common.opt: Rename align options with 'str_' prefix.
1347 * common/config/i386/i386-common.c (set_malign_value): New
1349 (ix86_handle_option): Use it to set -falign-* options/
1350 * config/aarch64/aarch64-protos.h (struct tune_params): Change
1351 type from int to string.
1352 * config/aarch64/aarch64.c: Update default values from int
1354 * config/alpha/alpha.c (alpha_override_options_after_change):
1356 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
1357 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1358 max skip conditionally.
1359 * config/i386/freebsd.h (SUBALIGN_LOG): New.
1360 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1361 max skip conditionally.
1362 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1363 max skip conditionally.
1364 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
1365 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1366 max skip conditionally.
1367 * config/i386/i386.c (struct ptt): Change type from int to
1369 (ix86_default_align): Set default values.
1370 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
1371 max skip conditionally.
1372 * config/i386/iamcu.h (SUBALIGN_LOG): New.
1373 (ASM_OUTPUT_MAX_SKIP_ALIGN):
1374 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
1375 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1376 max skip conditionally.
1377 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
1378 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
1379 * config/i386/x86-64.h (SUBALIGN_LOG): New.
1380 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1381 max skip conditionally.
1382 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1383 * config/ia64/ia64.c (ia64_option_override): Set default values
1384 for alignment options.
1385 * config/m68k/m68k.c: Handle new str_align_* options.
1386 * config/mips/mips.c (mips_set_compression_mode): Change
1388 (mips_option_override): Set default values for options.
1389 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
1391 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1393 * config/rx/rx.c (rx_option_override): Likewise.
1394 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
1395 (LABEL_ALIGN): Use align_labels_log.
1396 (LOOP_ALIGN): Use align_loops_align.
1397 * config/s390/s390.c (s390_asm_output_function_label): Use new
1399 * config/sh/sh.c (sh_override_options_after_change):
1400 Change type of constants.
1401 * config/spu/spu.c (spu_sched_init): Likewise.
1402 * config/sparc/sparc.c (sparc_option_override): Set default
1404 * config/visium/visium.c (visium_option_override): Likewise.
1405 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
1406 emit p2align format with last argument if it's not needed.
1407 * doc/invoke.texi: Document extended format of -falign-*.
1408 * final.c: Use align_labels alignment.
1409 * flags.h (struct target_flag_state): Change type to use
1411 (struct align_flags_tuple): New.
1412 (struct align_flags): Likewise.
1413 (align_loops_log): Redefine macro to use new types.
1414 (align_loops_max_skip): Redefine macro to use new types.
1415 (align_jumps_log): Redefine macro to use new types.
1416 (align_jumps_max_skip): Redefine macro to use new types.
1417 (align_labels_log): Redefine macro to use new types.
1418 (align_labels_max_skip): Redefine macro to use new types.
1419 (align_functions_log): Redefine macro to use new types.
1420 (align_loops): Redefine macro to use new types.
1421 (align_jumps): Redefine macro to use new types.
1422 (align_labels): Redefine macro to use new types.
1423 (align_functions): Redefine macro to use new types.
1424 (align_functions_max_skip): Redefine macro to use new types.
1425 (align_loops_value): New macro.
1426 (align_jumps_value): New macro.
1427 (align_labels_value): New macro.
1428 (align_functions_value): New macro.
1429 * function.c (invoke_set_current_function_hook): Propagate
1430 alignment values from flags to global variables default in
1432 * ipa-icf.c (sem_function::equals_wpa): Use
1433 cl_optimization_option_eq instead of memcmp.
1434 * lto-streamer.h (cl_optimization_stream_out): Support streaming
1436 (cl_optimization_stream_in): Likewise.
1437 * optc-save-gen.awk: Support strings in cl_optimization.
1438 * opth-gen.awk: Likewise.
1439 * opts.c (finish_options): Remove error checking of invalid
1441 (MAX_CODE_ALIGN): Remove.
1442 (MAX_CODE_ALIGN_VALUE): Likewise.
1443 (parse_and_check_align_values): New function.
1444 (check_alignment_argument): Likewise.
1445 (common_handle_option): Use check_alignment_argument.
1446 * opts.h (parse_and_check_align_values): Declare.
1447 * toplev.c (init_alignments): Remove.
1448 (read_log_maxskip): New.
1449 (parse_N_M): Likewise.
1450 (parse_alignment_opts): Likewise.
1451 (backend_init_target): Remove usage of init_alignments.
1452 * toplev.h (parse_alignment_opts): Declare.
1453 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
1455 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
1456 * tree.c (cl_option_hasher::equal): New.
1457 * varasm.c: Use new global macros.
1459 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
1461 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
1462 Use a simpler align directive also if MAXSKIP = ALIGN-1.
1463 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1464 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1465 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1466 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1467 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
1468 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
1469 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
1470 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1471 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1472 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1473 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1475 2018-07-04 Martin Liska <mliska@suse.cz>
1476 Jonathan Wakely <jwakely@redhat.com>
1478 * coverage.c: Use correct type.
1479 * doc/invoke.texi: Language correction.
1481 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
1484 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
1485 ENDBRANCH for non-tail call which may return via indirect branch.
1486 * doc/extend.texi: Document indirect_return attribute.
1488 2018-07-03 Martin Sebor <msebor@redhat.com>
1490 PR tree-optimization/86274
1491 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
1493 (format_floating): Correct handling of infinities and NaNs.
1495 2018-07-03 Martin Sebor <msebor@redhat.com>
1497 * print-tree.c (print_real_cst): New function.
1498 (print_node_brief): Call it.
1499 (print_node): Ditto.
1501 2018-07-03 Jeff Law <law@redhat.com>
1503 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
1504 into a single pattern.
1506 * config/h8300/h8300.md (ors code_iterator): New.
1507 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
1508 a single pattern and single splitter.
1509 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
1510 (iorqi3_1, xorqi3_1): Likewise.
1511 (iorqi3, xorqi3 expanders): Similarly.
1513 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
1514 (movmd_internal) into a single pattern using the P mode iterator.
1515 (movmd splitters): Similarly.
1516 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
1517 (movsd splitters): Similarly.
1519 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
1520 ADDB, ADDW and ADDL into a single ADD attribute which selects the
1521 right table based on the size of the operand.
1522 * config/h8300/h8300.md (length_table): Corresponding changes. All
1523 references to "addb", "addw" and "addl" changed to "add".
1524 (btst patterns): Merge two variants into a single pattern.
1525 (tstqi, tsthi): Likewise.
1526 (addhi3_incdec, addsi3_incdec): Likewise.
1527 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
1528 (mulhi3, mulsi3): Likewise.
1529 (udivhi3, udivsi3): Likewise.
1530 (divhi3, divsi3): Likewise.
1531 (andorqi3, andorhi3, andorsi3): Likewise.
1533 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
1536 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
1537 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
1539 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1541 PR tree-optimization/85694
1542 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
1543 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
1545 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
1546 (<u>avg<mode>3_ceil): New patterns.
1548 2018-07-03 David Malcolm <dmalcolm@redhat.com>
1550 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
1551 scan-tree-dump directive.
1552 * gcc.dg/vect/slp-perm-2.c: Likewise.
1553 * gcc.dg/vect/slp-perm-3.c: Likewise.
1554 * gcc.dg/vect/slp-perm-5.c: Likewise.
1555 * gcc.dg/vect/slp-perm-6.c: Likewise.
1556 * gcc.dg/vect/slp-perm-7.c: Likewise.
1557 * gcc.dg/vect/slp-perm-8.c: Likewise.
1559 2018-07-03 Marek Polacek <polacek@redhat.com>
1562 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
1564 2018-07-03 Richard Biener <rguenther@suse.de>
1567 * tree-ssa-structalias.c (find_func_clobbers): Properly
1568 handle indirect calls.
1570 2018-07-03 Jeff Law <law@redhat.com>
1572 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
1573 (shifts): New code iterator.
1574 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
1575 expander. Fix HImode handling on H8/SX.
1576 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
1577 (subqi3, subhi3, subsi3 expanders): Likewise.
1578 (andqi3, andhi3, andsi3 expanders): Likewise.
1579 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
1580 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
1581 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
1582 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
1583 (zero_extendqihi2, zero_extendqisi2): Likewise.
1584 (extendqihi2, extendqisi2): Likewise.
1585 (rotlqi3, rotlhi3, rotlsi3): Likewise.
1586 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
1587 (rotlqi3_1, rotlhi3_1): Likewise.
1588 (logicalhi3_sn, logicalsi3_sn): Likewise.
1589 (logicalhi3, logicalsi3): Likewise.
1591 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1593 * tree-vect-patterns.c (vect_recog_rotate_pattern)
1594 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1595 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
1596 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
1597 type to append_pattern_def_seq instead of creating a stmt_vec_info
1599 (build_mask_conversion): Likewise. Remove vinfo argument.
1600 (vect_add_conversion_to_patterm): Likewise, renaming to...
1601 (vect_add_conversion_to_pattern): ...this.
1602 (vect_recog_mask_conversion_pattern): Update call to
1603 build_mask_conversion. Pass the vector type to
1604 append_pattern_def_seq here too.
1605 (vect_recog_gather_scatter_pattern): Update call to
1606 vect_add_conversion_to_pattern.
1608 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1610 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
1611 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1612 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
1613 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
1614 STMT_VINFO_PATTERN_DEF_SEQ to null here.
1615 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
1616 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
1617 append_pattern_def_seq instead of new_pattern_def_seq.
1618 (vect_recog_divmod_pattern): Do both of the above.
1619 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
1622 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1624 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
1625 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
1626 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
1627 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
1628 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
1629 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1630 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
1631 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1632 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1633 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
1634 parameter with a single stmt_vec_info.
1635 (vect_recog_func_ptr): Likewise.
1636 (vect_recog_gather_scatter_pattern): Likewise, folding in...
1637 (vect_try_gather_scatter_pattern): ...this.
1638 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
1639 the stmt_vec_info of the statement to be matched. Don't clear
1640 STMT_VINFO_RELATED_STMT.
1641 (vect_pattern_recog): Update call accordingly.
1643 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1645 PR tree-optimization/85694
1646 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
1647 (uavgM3_ceil): Document new optabs.
1648 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
1649 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
1651 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
1652 (savg_ceil_optab): New optabs.
1653 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
1654 (vect_vect_recog_func_ptrs): Add it.
1655 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
1656 constant directly from the associated lhs.
1658 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1660 * tree-vect-patterns.c (vect_split_statement): New function.
1661 (vect_convert_input): Use it to try to split an existing cast.
1663 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1665 * poly-int.h (print_hex): New function.
1666 * dumpfile.h (dump_dec, dump_hex): Declare.
1667 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
1668 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
1669 min_input_precision, operation_precision and operation_sign.
1670 * tree-vect-patterns.c (vect_get_range_info): New function.
1671 (vect_same_loop_or_bb_p, vect_single_imm_use)
1672 (vect_operation_fits_smaller_type): Delete.
1673 (vect_look_through_possible_promotion): Add an optional
1674 single_use_p parameter.
1675 (vect_recog_over_widening_pattern): Rewrite to use new
1676 stmt_vec_info infomration. Handle one operation at a time.
1677 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
1678 (vect_truncatable_operation_p, vect_set_operation_type)
1679 (vect_set_min_input_precision): New functions.
1680 (vect_determine_min_output_precision_1): Likewise.
1681 (vect_determine_min_output_precision): Likewise.
1682 (vect_determine_precisions_from_range): Likewise.
1683 (vect_determine_precisions_from_users): Likewise.
1684 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
1685 (vect_vect_recog_func_ptrs): Put over_widening first.
1687 (vect_pattern_recog): Call vect_determine_precisions.
1689 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1691 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
1692 statements that have been replaced by further pattern statements.
1693 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
1695 2018-07-03 Richard Biener <rguenther@suse.de>
1697 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
1698 always set *dt. Dump vectype in vectype overload.
1699 * dumpfile.h (dump_gimple_expr): New function.
1700 (dump_gimple_expr_loc): Likewise.
1701 * dumpfile.c (dump_gimple_expr): New function.
1702 (dump_gimple_expr_loc): Likewise.
1704 2018-07-02 Jeff Law <law@redhat.com>
1706 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
1707 the H8/300, H8/300H and H8/S variants into a single pattern.
1708 (movhi_h8300, movqi_h8300hs): Similarly.
1709 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
1710 (QHI mode iterator): New.
1712 * config/h8300/h8300.md: Remove trailing whitespace.
1714 2018-07-02 Jim Wilson <jimw@sifive.com>
1716 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
1717 instead of emit_insn for interrupt returns.
1718 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
1719 (riscv_sret, riscv_uret): Likewise.
1721 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1723 * pretty-print.c (selftest::test_pp_format): Move save and restore
1724 of quotes to class auto_fix_quotes, and add an instance.
1725 * selftest.c: Include "intl.h".
1726 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
1727 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
1728 * selftest.h (selftest::auto_fix_quotes): New class.
1730 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
1732 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
1733 (aarch64_sve_prepare_conditional_op): Remove.
1734 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
1735 Allow aarch64_simd_reg_or_zero as select operand; remove
1736 the aarch64_sve_prepare_conditional_op call.
1737 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
1738 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1739 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
1740 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
1741 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
1742 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
1743 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
1744 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
1745 and a splitters to match all of the *_any patterns.
1746 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
1748 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
1749 (SVE_COND_FP_BINARY_REV): Remove.
1750 (sve_int_op_rev, sve_fp_op_rev): New.
1751 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
1752 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
1753 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
1754 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
1755 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
1756 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
1757 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
1758 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
1759 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
1761 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
1762 Remove match_dup 1 from the inner unspec.
1763 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1765 * config/aarch64/aarch64.md (movprfx): New attr.
1766 (length): Default movprfx to 8.
1767 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
1768 (*madd<SVE_I>, *msub<SVE_I): Likewise.
1769 (*<su>mul<SVE_I>3_highpart): Likewise.
1770 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
1771 (*v<ASHIFT><SVE_I>3): Likewise.
1772 (*<su><MAXMIN><SVE_I>3): Likewise.
1773 (*<su><MAXMIN><SVE_F>3): Likewise.
1774 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
1775 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
1776 (*div<SVE_F>4): Likewise.
1778 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
1780 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
1783 2018-07-02 Richard Biener <rguenther@suse.de>
1785 PR tree-optimization/86363
1786 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
1787 memset argument refers to a non-variable address.
1789 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1791 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
1792 {PLUS,MINUS}_EXPR code to...
1793 (adjust_symbolic_bound): ...here,
1794 (combine_bound): ...here,
1795 (set_value_range_with_overflow): ...and here.
1797 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1799 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
1801 (extract_range_from_abs_expr): ...here.
1803 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
1805 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
1806 -fno-omit-frame-pointer when not optimizing.
1808 2018-07-02 Martin Liska <mliska@suse.cz>
1811 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
1812 (propagate_nothrow): Likewise.
1814 2018-07-02 Martin Liska <mliska@suse.cz>
1817 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
1819 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1821 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
1822 function in r262149, changing "loc" param from source_location to
1823 const dump_location_t &.
1824 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
1825 declaration, as above.
1827 2018-07-01 Paul Koning <ni1d@arrl.net>
1829 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1830 -munit-asm, -mgnu-asm, -mdec-asm.
1831 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
1832 (pdp11_output_labelref): New.
1833 (pdp11_output_def): New.
1834 (pdp11_output_addr_vec_elt): New.
1835 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
1836 %# and %@ format codes.
1837 (pdp11_option_override): New.
1838 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
1839 (pdp11_output_ident): New.
1840 (pdp11_asm_named_section): New.
1841 (pdp11_asm_init_sections): New.
1842 (pdp11_file_start): New.
1843 (pdp11_file_end): New.
1844 (output_ascii): Use .ascii/.asciz for -mdec-asm.
1845 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
1846 %o, like %c but octal.
1847 (pdp11_option_override): New.
1848 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
1850 (DATA_SECTION_ASM_OP): Ditto.
1851 (READONLY_DATA_SECTION_ASM_OP): New.
1852 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
1853 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
1854 (ASM_OUTPUT_LABELREF): Ditto.
1855 (ASM_OUTPUT_DEF): Ditto.
1856 (ASM_OUTPUT_EXTERNAL): New.
1857 (ASM_OUTPUT_SOURCE_FILENAME): New.
1858 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
1859 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
1860 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
1861 %# and %@ format codes.
1862 * config/pdp11/pdp11.opt (mgnu-asm): New.
1863 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
1864 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
1865 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
1867 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
1869 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
1870 dereferencing path[] beyond its length.
1872 (debug_all_paths): New.
1873 (rewire_first_differing_edge): New.
1874 (adjust_paths_after_duplication): New.
1875 (duplicate_thread_path): Call adjust_paths_after_duplication.
1877 (thread_through_all_blocks): Add new argument to
1878 duplicate_thread_path.
1880 2018-06-30 Jim Wilson <jimw@sifive.com>
1882 * config/riscv/predicates.md (p2m1_shift_operand): New.
1883 (high_mask_shift_operand): New.
1884 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
1885 pattern using p2m1_shift_operand.
1886 (lshsi3_zero_extend_3+2): New combiner pattern using
1887 high_mask_shift_operand.
1889 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1891 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
1893 (vect_recog_rotate_pattern): ...here.
1894 (vect_convert_input): Try to insert casts of invariants in the
1896 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
1897 preheader to be empty.
1899 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1901 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
1902 vector type. If given, install it in the new statement's
1904 (vect_element_precision): New function.
1905 (vect_unpromoted_value): New struct.
1906 (vect_unpromoted_value::vect_unpromoted_value): New function.
1907 (vect_unpromoted_value::set_op): Likewise.
1908 (vect_look_through_possible_promotion): Likewise.
1909 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
1910 (vect_widened_op_tree, vect_convert_input): Likewise.
1911 (vect_convert_inputs, vect_convert_output): Likewise.
1912 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
1913 to handle the optional cast of the multiplication result and
1914 vect_widened_op_tree to detect the widened multiplication itself.
1915 Do not require the input and output of promotion casts to have
1916 the same sign, but base the signedness of the operation on the
1917 input rather than the result. If the pattern includes two
1918 promotions, check that those promotions have the same sign.
1919 Do not restrict the MULT_EXPR handling to a double-width result;
1920 handle quadruple-width results and wider. Use vect_convert_inputs
1921 to convert the inputs to the common type.
1922 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
1923 to handle the optional cast of the ABS result. Also allow a sign
1924 change or a sign extension between the ABS and MINUS.
1925 Use vect_widened_op_tree to detect the widened subtraction and use
1926 vect_convert_inputs to convert the inputs to the common type.
1927 (vect_handle_widen_op_by_const): Delete.
1928 (vect_recog_widen_op_pattern): New function.
1929 (vect_recog_widen_mult_pattern): Use it.
1930 (vect_recog_widen_shift_pattern): Likewise.
1931 (vect_recog_widen_sum_pattern): Use
1932 vect_look_through_possible_promotion to handle the promoted
1935 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1937 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
1938 the containing gimple_seq *.
1939 * gimple-iterator.h (gsi_for_stmt): Declare it.
1940 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
1941 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
1942 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1943 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1944 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
1946 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
1948 (vect_mark_pattern_stmts): ...here. Handle cases in which the
1949 statement being replaced is part of an existing pattern
1950 definition sequence, inserting the new pattern statements before
1952 (vect_pattern_recog_1): Don't return a bool. If the statement
1953 is already part of a pattern, instead apply pattern matching
1954 to the pattern definition statements. Don't clear the
1955 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
1956 (vect_pattern_recog): Don't break after the first match;
1957 continue processing the pattern definition statements instead.
1958 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
1960 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1962 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
1963 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1964 (vect_recog_widen_sum_pattern): Use it.
1966 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1968 * tree-vect-loop.c (vectorizable_reduction): Assert that the
1969 phi is not a pattern statement and has not been replaced by
1970 a pattern statement.
1971 * tree-vect-patterns.c (type_conversion_p): Don't check
1972 STMT_VINFO_IN_PATTERN_P.
1973 (vect_recog_vector_vector_shift_pattern): Likewise.
1974 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
1975 the pattern statement rather than the original statement; check
1976 directly for a WIDEN_MULT_EXPR here.
1977 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
1978 vect_is_simple_use to return the pattern statement rather
1979 than the original statement; use is_pattern_stmt_p to check
1980 for such a pattern statement.
1981 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
1982 to return the pattern statement rather than the original statement;
1983 don't do the same transformation here.
1984 (vect_is_simple_use): If the defining statement has been replaced
1985 by a pattern statement, return the pattern statement instead.
1986 Remove the corresponding (local) transformation from the vectype
1989 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1991 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
1992 end and default to null.
1993 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1994 (vectorizable_reduction): Update calls accordingly, dropping the
1995 gimple ** argument if the passed-back statement isn't needed.
1996 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
1997 (vect_recog_rotate_pattern): Likewise.
1998 (vect_recog_mask_conversion_pattern): Likewise.
1999 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2000 (vect_mask_constant_operand_p): Likewise.
2001 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
2002 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
2003 (get_group_load_store_type, get_load_store_type): Likewise.
2004 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
2005 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
2006 (vectorizable_conversion, vectorizable_assignment): Likewise.
2007 (vectorizable_shift, vectorizable_operation): Likewise.
2008 (vectorizable_store, vect_is_simple_cond): Likewise.
2009 (vectorizable_condition, vectorizable_comparison): Likewise.
2010 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
2011 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
2012 and move it to the end. Cope with null def_stmt_outs.
2014 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2016 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
2018 2018-06-29 Jeff Law <law@redhat.com>
2020 * config/v850/v850.c (v850_legitimate_address_p): Handle large
2021 displacements for TARGET_V850E2V3 and newer.
2022 (TARGET_LRA_P): Remove. Defaults to LRA now.
2023 * config/v850/v850.md (sign23byte_load): Remove.
2024 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
2025 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
2027 2018-06-29 Martin Liska <mliska@suse.cz>
2030 * coverage.c (coverage_init): Mangle full path name.
2031 * doc/invoke.texi: Document the change.
2032 * gcov-io.c (mangle_path): New.
2033 * gcov-io.h (mangle_path): Likewise.
2034 * gcov.c (mangle_name): Use mangle_path for path mangling.
2036 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2038 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
2039 if starting source register is not even.
2041 2018-06-29 Martin Liska <mliska@suse.cz>
2043 PR tree-optimization/86263
2044 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
2045 Make edge redirection.
2047 2018-06-29 David Malcolm <dmalcolm@redhat.com>
2049 * dumpfile.c (dump_loc): Add indentation based on scope depth.
2050 (dump_scope_depth): New variable.
2051 (get_dump_scope_depth): New function.
2052 (dump_begin_scope): New function.
2053 (dump_end_scope): New function.
2054 * dumpfile.h (get_dump_scope_depth): New declaration.
2055 (dump_begin_scope): New declaration.
2056 (dump_end_scope): New declaration.
2057 (class auto_dump_scope): New class.
2058 (AUTO_DUMP_SCOPE): New macro.
2059 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
2062 2018-06-29 Richard Biener <rguenther@suse.de>
2064 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
2065 compute_all_dependences succeeds.
2066 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
2067 exceed --param loop-max-datarefs-for-datadeps.
2069 2018-06-29 Jakub Jelinek <jakub@redhat.com>
2071 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
2073 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
2076 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
2077 alternative 0 in preferred_for_speed attribute.
2079 2018-06-28 Paul Koning <ni1d@arrl.net>
2081 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
2082 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
2083 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
2084 * config/pdp11/pdp11.md: Correct "length" attribute calculation
2085 for shift insn patterns.
2087 2018-06-28 David Malcolm <dmalcolm@redhat.com>
2089 * cgraph.c (cgraph_node::get_body): Replace assignments to
2090 "dump_file" with calls to set_dump_file.
2091 * dumpfile.c (alt_dump_file): Make static, and group with...
2092 (alt_flags): ...this definition.
2093 (dumps_are_enabled): New variable.
2094 (refresh_dumps_are_enabled): New function.
2095 (set_dump_file): New function.
2096 (set_alt_dump_file): New function.
2097 (gcc::dump_manager::dump_start): Replace assignments to
2098 "dump_file" and "alt_dump_file" with calls to set_dump_file and
2100 (gcc::dump_manager::dump_finish): Likewise.
2101 * dumpfile.h (alt_dump_file): Delete decl.
2102 (dumps_are_enabled): New variable decl.
2103 (set_dump_file): New function decl.
2104 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
2106 * tree-nested.c (lower_nested_functions): Replace assignments to
2107 "dump_file" with calls to set_dump_file.
2109 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
2111 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
2112 goto_locus of each outgoing edge of each basic block.
2114 2018-06-28 Richard Biener <rguenther@suse.de>
2116 * dwarf2out.c (decl_scope_table): Remove.
2117 (push_decl_scope): Likewise.
2118 (pop_decl_scope): Likewise.
2119 (gen_type_die_for_member): Do not call push/pop_decl_scope.
2120 (gen_struct_or_union_type_die): Likewise.
2121 (gen_tagged_type_die): Likewise.
2122 (dwarf2out_init): Do not initialize decl_scope_table.
2123 (dwarf2out_c_finalize): Do not free it.
2125 2018-06-28 Richard Biener <rguenther@suse.de>
2127 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
2128 deciding whether to not re-use a DIE.
2130 2018-06-28 Richard Biener <rguenther@suse.de>
2132 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
2133 DW_AT_abstract_origin attribute.
2135 2018-06-28 Martin Liska <mliska@suse.cz>
2137 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2138 Use newly introduced constants.
2139 * tree-switch-conversion.h (struct jump_table_cluster):
2140 Define max_ratio_for_size and max_ratio_for_speed.
2142 2018-06-28 Martin Liska <mliska@suse.cz>
2144 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2145 Add new checking assert to catch invalid state.
2146 (jump_table_cluster::can_be_handled): Handle single case
2148 (jump_table_cluster::is_beneficial): Bail out for such case.
2149 (bit_test_cluster::find_bit_tests):
2150 Add new checking assert to catch invalid state.
2151 (bit_test_cluster::can_be_handled): Handle single case
2153 (bit_test_cluster::is_beneficial): Bail out for such case.
2154 (switch_decision_tree::analyze_switch_statement):
2157 2018-06-28 Martin Liska <mliska@suse.cz>
2159 * common.opt: Introduce -completion option.
2160 * gcc.c (driver_handle_option): Handle it.
2161 (driver::main): Print completions if completion
2163 * opt-suggestions.c (option_proposer::get_completions):
2165 (option_proposer::suggest_completion): Likewise.
2166 (option_proposer::find_param_completions): Likewise.
2167 (verify_autocompletions): Likewise.
2168 (test_completion_valid_options): Likewise.
2169 (test_completion_valid_params): Likewise.
2170 (in_completion_p): Likewise.
2171 (empty_completion_p): Likewise.
2172 (test_completion_partial_match): Likewise.
2173 (test_completion_garbage): Likewise.
2174 (opt_proposer_c_tests): Likewise.
2175 * opt-suggestions.h: Declare new functions.
2176 * opts.c (common_handle_option): Handle OPT__completion_.
2177 * selftest-run-tests.c (selftest::run_tests): Add
2178 opt_proposer_c_tests.
2179 * selftest.c (assert_str_startswith): New.
2180 * selftest.h (assert_str_startswith): Likewise.
2181 (opt_proposer_c_tests): New.
2182 (ASSERT_STR_STARTSWITH): Likewise.
2184 2018-06-28 Martin Liska <mliska@suse.cz>
2186 * Makefile.in: Add opt-suggestions.o.
2187 * gcc-main.c: Include opt-suggestions.h.
2188 * gcc.c (driver::driver): Likewise.
2189 (driver::~driver): Remove m_option_suggestions.
2190 (driver::build_option_suggestions): Moved to option_proposer.
2191 (driver::suggest_option): Likewise.
2192 (driver::handle_unrecognized_options): Use option_proposer.
2193 * gcc.h (class driver): Add new memver m_option_proposer.
2194 * opt-suggestions.c: New file.
2195 * opt-suggestions.h: New file.
2197 2018-06-28 Martin Liska <mliska@suse.cz>
2199 * vec.h (class auto_string_vec): New (moved from auto_argvec).
2200 (auto_string_vec::~auto_string_vec): Likewise.
2202 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
2204 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
2205 prevent_decl_creation_for_types fields up and add reset_location field.
2206 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
2207 statement if id->reset_location is true.
2208 (copy_edges_for_bb): Do not set goto_locus on the new edges if
2209 id->reset_location is true.
2210 (copy_phis_for_bb): Force input_location on the arguments if
2211 id->reset_location is true.
2212 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
2213 is set on the function to be inlined.
2215 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
2217 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
2219 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
2221 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
2222 registers for Pmode.
2223 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
2224 hard registers for the clobbered pseudo.
2226 2018-06-27 Paul Koning <ni1d@arrl.net>
2228 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
2229 mutually exclusive options.
2230 * config/pdp11/constraints.md (h): New constraint.
2231 (O): Update definition to match shift code generation.
2232 (D): New constraint.
2233 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
2235 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
2237 (output_jump): Change arguments.
2238 (pdp11_fixed_cc_regs): New function.
2239 (pdp11_cc_mode): Ditto.
2240 (pdp11_expand_shift): Ditto.
2241 (pdp11_assemble_shift): Ditto.
2242 (pdp11_small_shift): Ditto.
2243 (pdp11_branch_cost): Remove.
2244 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
2246 (pdp11_register_move_cost): Update for CC registers.
2247 (pdp11_rtx_costs): Add case for LSHIFTRT.
2248 (pdp11_output_jump): Add CCNZ mode conditional branches.
2249 (notice_update_cc_on_set): Remove.
2250 (pdp11_cc_mode): New function.
2251 (simple_memory_operand): Correct pre/post decrement case.
2252 (no_side_effect_operand): New function.
2253 (pdp11_regno_reg_class): Add CC_REGS class.
2254 (pdp11_fixed_cc_regs): New function.
2255 (pdp11_small_shift): New function.
2256 (pdp11_expand_shift): New function to expand shift insns.
2257 (pdp11_assemble_shift): New function to output shifts.
2258 (pdp11_branch_cost): Remove.
2259 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
2260 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
2261 (WCHAR_TYPE): Ditto.
2262 (PTRDIFF_TYPE): Ditto.
2263 (ADJUST_INSN_LENGTH): New macro.
2264 (FIXED_REGISTERS): Add CC registers.
2265 (CALL_USED_REGISTERS): Ditto.
2267 (REG_CLASS_NAMES): Ditto.
2268 (REG_CLASS_CONTENTS): Ditto.
2269 (SELECT_CC_MODE): Use new function.
2270 (TARGET_FLAGS_REGNUM): New macro.
2271 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
2272 (cc0_reg_rtx): Remove.
2273 (CC_STATUS_MDEP): Remove.
2274 (CC_STATUS_MDEFP_INIT): Remove.
2275 (CC_IN_FPU): Remove.
2276 (NOTICE_UPDATE_CC): Remove.
2277 (REGISTER_NAMES): Add CC registers.
2278 (BRANCH_COST): Change to constant 1.
2279 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
2281 * config/pdp11/pdp11.opt (mbcopy): Remove.
2282 (mbcopy-builtin): Remove.
2283 (mbranch-cheap): Remove.
2284 (mbranch-expensive): Remove.
2285 * config/pdp11/predicates.md (expand_shift_operand): Update to
2286 match shift code generation.
2287 (ccnz_operator): New predicate.
2288 * doc/invoke.texi (PDP-11 Options): Remove deleted options
2289 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
2290 Remove non-existent option -mabshi, -mno-abshi. Document mutually
2292 * doc/md.texi (PDP-11): Document new D and h constraints. Update
2293 description of O constraint.
2295 2018-06-27 Jeff Law <law@redhat.com>
2296 Austin Law <austinklaw@gmail.com>
2298 * config/v850/v850.md (addsi3_set_flags): New pattern.
2299 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
2300 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
2301 (zero_extendhisi2_v850_set_flags): Likewise.
2302 (zero_extendqisi2_v850_set_flags): Likewise.
2303 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
2304 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
2305 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
2307 * config/v850/v850-protos.h (notice_update_cc): Remove.
2308 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
2309 (v850_print_operand): Handle 'D' and "d".
2310 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
2311 Add handling of arithmetic/logical operations compared against zero.
2312 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
2313 Do not look at v850_compare_op, instead get mode from last argument.
2314 (v850_gen_compare): Remove
2315 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
2316 after reload for prologue insns.
2317 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
2319 (construct_save_jarl): Likewise.
2320 (TARGET_FLAGS_REGNUM): Define.
2321 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
2322 (NOTICE_UPDATE_CC): Remove.
2323 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
2324 than cc0. Conditionalize on reload_completed.
2325 (cmpsi_insn, setfcc_insn): Likewise.
2326 (tst1 splitter): Turn into define_and_split which sets the flags
2328 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
2329 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
2330 (cstoresf4, cstoredf4): Clobber the flags.
2331 (cmpsi, cmpsf, cmpdf): Remove expanders.
2332 (setf_insn): Remove pattern.
2333 (addsi3): Turn into define_and_split which clobbers the flags after
2334 reload and a suitable pattern (addsi3_clobber_flags) for use after
2336 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
2337 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
2338 (ashrsi3, ashrsi3_v850e2): Likewise.
2339 (bins): Clobber the flags.
2340 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
2341 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
2342 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
2343 (call_value_internal_short, call_value_internal_long): Likewise.
2344 (callt_save_interrupt, callt_return_interrupt): Likewise.
2345 (save_interrupt, return_interrupt): Likewise.
2346 (callt_save_all_interrupt, save_all_interrupt): Likewise.
2347 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
2348 (restore_all_interrupt, _restore_all_interrupt): Likewise.
2349 (All FP comparisons): Only allow after reload has completed.
2351 (divh, divhu): Tweak output template.
2352 (branch_z_normal, branch_z_invert): Remove
2353 (branch_nz_normal, branch_nz_invert): Likewise.
2354 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
2356 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
2357 * config/v850/v850.c (notice_update_cc): Remove.
2358 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
2359 (CC_NO_CARRY): Likewise.
2360 (NOTICE_UPDATE_CC): Define to nothing.
2361 * config/v850/v850.md: Remove block comment on cc0 handling
2362 Remove "cc" attribute from all patterns. Remove cc_status handling
2363 from all patterns. Minor formatting fixes.
2365 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2367 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
2368 (cortex-a76.cortex-a55): Likewise.
2369 * config/aarch64/aarch64-tune.md: Regenerate.
2370 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
2371 cortex-a76.cortex-a55.
2373 2018-06-27 Jeff Law <law@redhat.com>
2375 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
2376 (MULTILIB_DIRNAMES): Similarly.
2378 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
2380 * gimple.h (gimple_return_retbnd): Delete.
2381 (gimple_return_set_retbnd): Likewise.
2382 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
2383 gimple_return_set_retbnd.
2384 * gimple-pretty-print.c (dump_gimple_return): Remove call to
2385 gimple_return_retbnd and adjust.
2386 * tree-inline.h (struct copy_body_data): Remove retbnd field.
2387 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
2388 Explicitly return NULL in a couple more cases. Move assertion
2389 on debug statements and remove unreachable code.
2390 (reset_debug_binding): Do not test id->retbnd.
2391 (expand_call_inline): Do not set it.
2393 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2395 * configure.ac: Add --disable-gcov option.
2396 * configure: Regenerate.
2397 * Makefile.in: Honour @enable_gcov@.
2398 * doc/install.texi: Document --disable-gcov.
2400 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2402 * config/arm/arm-cpus.in (cortex-a76): New entry.
2403 (cortex-a76.cortex-a55): Likewise.
2404 * config/arm/arm-tables.opt: Regenerate.
2405 * config/arm/arm-tune.md: Likewise.
2406 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
2407 * doc/invoke.texi (ARM Options): Document cortex-a76 and
2408 cortex-a76.cortex-a55.
2410 2018-06-27 Tamar Christina <tamar.christina@arm.com>
2413 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
2415 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
2417 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
2419 (EPILOGUE_USES): Likewise.
2421 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
2423 * tree-inline.c (remap_location): New function extracted from...
2424 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
2425 (copy_phis_for_bb): ...here. Call remap_location.
2426 (copy_cfg_body): Adjust call to copy_edges_for_bb.
2428 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
2430 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
2431 unaligned vsx for 16B memset.
2433 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
2436 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
2437 ieee128_float_type_node to long_double_type_node unless
2438 TARGET_LONG_DOUBLE_128 is set.
2440 2018-06-26 David Malcolm <dmalcolm@redhat.com>
2442 * cfgloop.c (get_loop_location): Convert return type from
2443 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
2444 by implicit construction from rtx_insn *, and using
2445 dump_user_location_t::from_function_decl for the fallback case.
2446 * cfgloop.h (get_loop_location): Convert return type from
2447 location_t to dump_user_location_t.
2448 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
2449 dump_printf_loc to pass in a dump_location_t rather than a
2450 location_t, via the gimple stmt.
2451 * coverage.c (get_coverage_counts): Update calls to
2452 dump_printf_loc to pass in dump_location_t rather than a
2454 * doc/optinfo.texi (Dump types): Convert example of
2455 dump_printf_loc from taking "locus" to taking "insn". Update
2456 description of the "_loc" calls to cover dump_location_t.
2457 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
2459 (dump_user_location_t::dump_user_location_t): New constructors,
2460 from gimple *stmt and rtx_insn *.
2461 (dump_user_location_t::from_function_decl): New function.
2462 (dump_loc): Make static.
2463 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
2464 const dump_location_t &.
2465 (dump_generic_expr_loc): Delete.
2466 (dump_printf_loc): Convert param "loc" from location_t to
2467 const dump_location_t &.
2468 (selftest::test_impl_location): New function.
2469 (selftest::dumpfile_c_tests): New function.
2470 * dumpfile.h: Include "profile-count.h".
2471 (class dump_user_location_t): New class.
2472 (struct dump_impl_location_t): New struct.
2473 (class dump_location_t): New class.
2474 (dump_printf_loc): Convert 2nd param from source_location to
2475 const dump_location_t &.
2476 (dump_generic_expr_loc): Delete.
2477 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
2478 const dump_location_t &.
2479 * gimple-fold.c (fold_gimple_assign): Update call to
2480 dump_printf_loc to pass in a dump_location_t rather than a
2481 location_t, via the gimple stmt.
2482 (gimple_fold_call): Likewise.
2483 * gimple-loop-interchange.cc
2484 (loop_cand::analyze_iloop_reduction_var): Update for change to
2485 check_reduction_path.
2486 (tree_loop_interchange::interchange): Update for change to
2488 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
2489 change in return-type of find_loop_location.
2490 (graphite_regenerate_ast_isl): Likewise.
2491 * graphite-optimize-isl.c (optimize_isl): Likewise.
2492 * graphite.c (graphite_transform_loops): Likewise.
2493 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
2494 pass in a dump_location_t rather than a location_t, via the
2496 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2497 * ipa.c (walk_polymorphic_call_targets): Likewise.
2498 * loop-unroll.c (report_unroll): Convert "locus" param from
2499 location_t to dump_location_t.
2500 (decide_unrolling): Update for change to get_loop_location's
2502 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
2503 location_t to dump_user_location_t.
2504 (grid_find_single_omp_among_assignments_1): Updates calls to
2505 dump_printf_loc to pass in a dump_location_t rather than a
2506 location_t, via the gimple stmt.
2507 (grid_parallel_clauses_gridifiable): Convert "tloc" from
2508 location_t to dump_location_t. Updates calls to dump_printf_loc
2509 to pass in a dump_location_t rather than a location_t, via the
2511 (grid_inner_loop_gridifiable_p): Likewise.
2512 (grid_dist_follows_simple_pattern): Likewise.
2513 (grid_gfor_follows_tiling_pattern): Likewise.
2514 (grid_target_follows_gridifiable_pattern): Likewise.
2515 (grid_attempt_target_gridification): Convert initialization
2516 of local "grid" from memset to zero-initialization; FIXME: does
2517 this require C++11? Update call to dump_printf_loc to pass in a
2518 optinfo_location rather than a location_t, via the gimple stmt.
2519 * profile.c (read_profile_edge_counts): Updates call to
2520 dump_printf_loc to pass in a dump_location_t rather than a
2522 (compute_branch_probabilities): Likewise.
2523 * selftest-run-tests.c (selftest::run_tests): Call
2525 * selftest.h (dumpfile_c_tests): New decl.
2526 * tree-loop-distribution.c (pass_loop_distribution::execute):
2527 Update for change in return type of find_loop_location.
2528 * tree-parloops.c (parallelize_loops): Likewise.
2529 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
2530 "locus" from location_t to dump_user_location_t.
2531 (canonicalize_loop_induction_variables): Likewise.
2532 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
2533 for change in return type of find_loop_location.
2534 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
2535 to dump_printf_loc to pass in a dump_location_t rather than a
2536 location_t, via the stmt.
2537 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
2539 * tree-vect-loop-manip.c (find_loop_location): Convert return
2540 type from source_location to dump_user_location_t.
2541 (vect_do_peeling): Update for above change.
2542 (vect_loop_versioning): Update for change in type of
2544 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
2545 from location_t to dump_user_location_t.
2546 (vect_estimate_min_profitable_iters): Update for change in type
2548 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
2549 location_t to dump_location_t.
2550 (vect_slp_bb): Update for change in type of vect_location.
2551 * tree-vectorizer.c (vect_location): Convert from source_location
2552 to dump_user_location_t.
2553 (try_vectorize_loop_1): Update for change in vect_location's type.
2554 (vectorize_loops): Likewise.
2555 (increase_alignment): Likewise.
2556 * tree-vectorizer.h (vect_location): Convert from source_location
2557 to dump_user_location_t.
2558 (find_loop_location): Convert return type from source_location to
2559 dump_user_location_t.
2560 (check_reduction_path): Convert 1st param from location_t to
2561 dump_user_location_t.
2562 * value-prof.c (check_counter): Update call to dump_printf_loc to
2563 pass in a dump_user_location_t rather than a location_t; update
2564 call to error_at for change in type of "locus".
2565 (check_ic_target): Update call to dump_printf_loc to
2566 pass in a dump_user_location_t rather than a location_t, via the
2569 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
2571 * config/s390/s390.h (enum processor_flags): Do not use
2572 default tune parameter when -march was specified.
2574 2018-06-26 Jakub Jelinek <jakub@redhat.com>
2577 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
2578 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
2581 2018-06-26 Richard Biener <rguenther@suse.de>
2583 PR tree-optimization/86287
2585 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
2586 (vect_analyze_loop): Initialize n_stmts.
2588 2018-06-26 Richard Biener <rguenther@suse.de>
2591 * fold-const.c (fold_convertible_p): Pointer extension
2594 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
2597 * dwarf2out.c (loc_list_has_views): Adjust comments.
2598 (dw_loc_list): Split single cross-partition range with
2601 2018-06-25 Jeff Law <law@redhat.com>
2603 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
2604 on -mbig-switch by default.
2606 * config/v850/predicates.md (const_float_1_operand): Fix match_code
2608 (const_float_0_operand): Remove unused predicate.
2609 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
2610 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
2611 (recipsf2): New expander. Original pattern now called
2613 (recipdf2, recipdf2_insn): Similarly.
2614 (rsqrtsf2, rsqrtsf2_insn): Similarly
2615 (rsqrtdf2, rsqrtdf2_insn): Similarly
2617 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
2619 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
2620 Simplify logic for FreeBSD (twice).
2622 2018-06-25 Martin Sebor <msebor@redhat.com>
2624 PR tree-optimization/86204
2625 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
2626 a strnlen result if it's less than the length of the string.
2628 2018-06-25 Martin Sebor <msebor@redhat.com>
2630 PR tree-optimization/85700
2631 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
2632 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
2633 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
2635 2018-06-25 Martin Sebor <msebor@redhat.com>
2637 * doc/extend.texi (Zero-length arrays): Update and clarify.
2639 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
2641 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
2642 added IEEE/IBM long double multilib support on PowerPC little
2643 endian Linux systems.
2644 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
2645 (MULTILIB_DEFAULTS): Likewise.
2646 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2648 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
2649 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
2650 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
2652 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
2655 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
2656 (REORDER_45): Likewise.
2658 2018-06-25 Jeff Law <law@redhat.com>
2660 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
2661 dividend to 32 bits. Adjust length.
2662 (udivmodhi4): Cleanup output template. Fix length.
2664 2018-06-25 Carl Love <cel@us.ibm.com>
2666 * config/rs6000/vsx.md: Change word selector to prefered location.
2668 2018-06-25 Richard Biener <rguenther@suse.de>
2670 PR tree-optimization/86304
2671 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
2672 epilogue-if-converted loops as well.
2674 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
2676 * lto-section-out.c (lto_begin_section): Do not print section
2677 name for noaddr and unnumbered dumps.
2679 2018-06-25 Richard Biener <rguenther@suse.de>
2681 * tree-vectorizer.h (struct vec_info_shared): New structure
2682 with parts split out from struct vec_info and loop_nest from
2683 struct _loop_vec_info.
2684 (struct vec_info): Adjust accordingly.
2685 (struct _loop_vec_info): Likewise.
2686 (LOOP_VINFO_LOOP_NEST): Adjust.
2687 (LOOP_VINFO_DATAREFS): Likewise.
2688 (LOOP_VINFO_DDRS): Likewise.
2689 (struct _bb_vec_info): Likewise.
2690 (BB_VINFO_DATAREFS): Likewise.
2691 (BB_VINFO_DDRS): Likewise.
2692 (struct _stmt_vec_info): Add dr_aux member.
2693 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
2694 (DR_MISALIGNMENT_UNINITIALIZED): New.
2695 (set_dr_misalignment): Adjust.
2696 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
2697 (vect_analyze_loop): Adjust prototype.
2698 (vect_analyze_loop_form): Likewise.
2699 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2700 Compute dependences lazily.
2701 (vect_record_base_alignments): Use shared datarefs/ddrs.
2702 (vect_verify_datarefs_alignment): Likewise.
2703 (vect_analyze_data_refs_alignment): Likewise.
2704 (vect_analyze_data_ref_accesses): Likewise.
2705 (vect_analyze_data_refs): Likewise.
2706 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
2707 constructor parameter for shared part.
2708 (vect_analyze_loop_form): Pass in shared part and adjust.
2709 (vect_analyze_loop_2): Pass in storage for the number of
2710 stmts. Move loop nest finding to the caller. Compute
2712 (vect_analyze_loop): Pass in shared part.
2713 (vect_transform_loop): Verify shared datarefs are unchanged.
2714 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
2715 constructor parameter for shared part.
2716 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
2717 (vect_slp_bb): Verify shared datarefs are unchanged before
2719 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
2721 (new_stmt_vec_info): Initialize DR_AUX misalignment to
2722 DR_MISALIGNMENT_UNINITIALIZED.
2723 * tree-vectorizer.c (vec_info::vec_info): Add constructor
2724 parameter for shared part.
2725 (vec_info::~vec_info): Adjust.
2726 (vec_info_shared::vec_info_shared): New.
2727 (vec_info_shared::~vec_info_shared): Likewise.
2728 (vec_info_shared::save_datarefs): Likewise.
2729 (vec_info_shared::check_datarefs): Likewise.
2730 (try_vectorize_loop_1): Construct shared part live for analyses
2731 of a single loop for multiple vector sizes.
2732 * tree-parloops.c (gather_scalar_reductions): Adjust.
2734 2018-06-25 Richard Biener <rguenther@suse.de>
2736 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
2737 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
2738 (vect_analyze_data_refs): Remove similar code from here and
2739 simplify accordingly.
2741 2018-06-25 Richard Biener <rguenther@suse.de>
2743 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
2744 for reverse storage order accesses rather than asserting
2745 they cannot happen here.
2747 2018-06-25 Tom de Vries <tdevries@suse.de>
2750 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
2751 Use data16 instead of .byte for insn prefix.
2753 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
2756 * parser.c (make_char_string_pack): Pass this literal chars
2757 through cpp_interpret_string.
2758 (cp_parser_userdef_numeric_literal): Check the result of
2759 make_char_string_pack.
2761 2018-06-24 Maya Rashish <coypu@sdf.org>
2763 * ginclude/stddef.h: Simplify conditions around avoiding
2764 re-definition of __size_t.
2766 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2768 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
2769 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
2771 2018-06-22 Maya Rashish <coypu@sdf.org>
2773 * doc/invoke.texi (mno-fancy-math-387): Update for changes
2774 made to OpenBSD and NetBSD through the years.
2776 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2778 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2779 behavior of vec_pack (vector double, vector double) to match
2780 behavior of vec_float2 (vector double, vector double).
2782 2018-06-22 Olivier Hainque <hainque@adacore.com>
2784 * gimplify.c (gimplify_function_tree): Prevent creation
2785 of a trampoline for the address of the current function
2786 passed to entry/exit instrumentation hooks.
2788 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
2791 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
2794 2018-06-22 Martin Liska <mliska@suse.cz>
2796 PR tree-optimization/86263
2797 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2798 Bail out if is_enabled is false.
2799 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
2801 (jump_table_cluster::is_enabled): New function.
2803 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2805 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
2806 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
2807 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
2808 (lto_input_ts_binfo_tree_pointers): Likewise.
2809 * tree-streamer-out.c (streamer_write_tree_bitfields,
2810 write_ts_binfo_tree_pointers): Likewise.
2811 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
2813 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2815 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
2817 2018-06-22 Martin Liska <mliska@suse.cz>
2819 * symbol-summary.h (get): Make it pure and inline move
2820 functionality from ::get function.
2821 (get): Remove and inline into ::get and ::get_create.
2822 (get_create): Move code from ::get function.
2824 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2827 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
2828 -x assembler-with-cpp.
2830 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2832 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
2833 _FILE_OFFSET_BITS=64 for C++.
2835 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
2837 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
2838 conversion insn that shows up when pr85657-3.c is compiled using
2839 IEEE 128-bit long double.
2840 (neg<mode>2_internal): Use the correct mode to check whether the
2841 mode is IBM extended.
2842 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
2843 multiply and divide external functions from being created more
2846 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2848 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
2850 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
2851 the edge can be forwarded.
2852 (cfg_layout_merge_blocks): Likewise.
2854 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2856 * except.c (finish_eh_generation): Commit edge insertions only after
2857 the EH edges have been redirected from post-landing to landing pads.
2859 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2861 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
2862 create_tmp_var_for to create the FRAME decl.
2863 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
2865 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2867 * tree-inline.c (copy_edges_for_bb): Minor tweak.
2868 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
2869 debug statement when resetting its value.
2870 (expand_call_inline): Copy the locus of the call onto the assignment
2871 of the return value, if any. Use local variable in more cases.
2873 2018-06-21 Martin Liska <mliska@suse.cz>
2875 * ipa-pure-const.c (propagate_nothrow): Use
2876 funct_state_summaries->get.
2877 (dump_malloc_lattice): Likewise.
2878 (propagate_malloc): Likewise.
2880 2018-06-21 Richard Biener <rguenther@suse.de>
2882 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
2883 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
2884 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
2886 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
2887 BLOCK_ABSTRACT_ORIGIN unconditionally.
2889 2018-06-21 David Malcolm <dmalcolm@redhat.com>
2891 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
2893 * ipa-reference.c (ipa_reference_c_finalize): Delete
2894 ipa_ref_opt_sum_summaries and set it to NULL.
2896 2018-06-21 Tom de Vries <tdevries@suse.de>
2898 PR tree-optimization/85859
2899 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
2900 test with comment from bb_no_side_effects_p.
2902 2018-06-21 Richard Biener <rguenther@suse.de>
2904 PR tree-optimization/86232
2905 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
2906 max for constant niter.
2908 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2910 * config/aarch64/aarch64-simd.md
2911 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
2913 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2915 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
2916 Make opernads of the unspec commutative.
2918 2018-06-21 Richard Biener <rguenther@suse.de>
2920 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
2921 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
2922 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
2923 (vect_analyze_data_ref_dependence): Re-order checks to deal with
2925 (vect_record_base_alignments): Do not record base alignment
2926 for gathers or scatters.
2927 (vect_compute_data_ref_alignment): Drop return value that is always
2928 true. Bail out early for gathers or scatters.
2929 (vect_enhance_data_refs_alignment): Bail out early for gathers
2931 (vect_find_same_alignment_drs): Likewise.
2932 (vect_analyze_data_refs_alignment): Remove dead code.
2933 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2934 (vect_analyze_data_refs): For possible gathers or scatters do
2935 not create an alternate DR, just check their possible validity
2936 and mark them. Adjust DECL_NONALIASED handling to not rely
2938 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
2939 update inits of gathers or scatters.
2940 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
2941 Also copy gather/scatter flag to pattern vinfo.
2943 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
2945 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2946 behavior of vec_packsu (vector unsigned long long, vector unsigned
2947 long long) to match behavior of vec_packs with same signature.
2949 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
2950 Thomas Schwinge <thomas@codesourcery.com>
2951 Cesar Philippidis <cesar@codesourcery.com>
2953 * gimplify.c (gimplify_scan_omp_clauses): Add support for
2954 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2955 (gimplify_adjust_omp_clauses): Likewise.
2956 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
2957 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
2958 (gimplify_omp_target_update): Update handling of acc update and
2960 * omp-low.c (install_var_field): Remove unused parameter
2961 base_pointers_restrict.
2962 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
2963 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
2965 (omp_target_base_pointers_restrict_p): Delete.
2966 (scan_omp_target): Update call to scan_sharing_clauses.
2967 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
2969 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
2970 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2971 (convert_local_omp_clauses): Likewise.
2972 * tree-pretty-print.c (dump_omp_clause): Likewise.
2973 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
2975 (omp_clause_code_name): Likewise.
2977 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2980 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
2983 PR tree-optimization/86231
2984 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
2985 anti-range don't overwrite *vr0min before using it to compute *vr0max.
2987 2018-06-20 Tom de Vries <tdevries@suse.de>
2989 PR tree-optimization/86097
2990 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
2991 iv type if signedness of iv type is not the same as that of *nit.
2993 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2995 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
2996 EDGE_EH edges, verify they are all EDGE_EH.
2998 2018-06-20 Maya Rashish <coypu@sdf.org>
3000 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
3002 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3004 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
3005 * config/aarch64/aarch64.c (xgene1_tunings): Add
3006 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
3007 (aarch64_mode_valid_for_sched_fusion_p):
3008 Allow 16-byte modes.
3009 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
3010 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
3012 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
3014 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
3015 * config/aarch64/iterators.md (VQ2): New mode iterator.
3017 2018-06-20 Martin Liska <mliska@suse.cz>
3019 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
3020 Change default ratio from 10 to 8.
3022 2018-06-20 Martin Liska <mliska@suse.cz>
3024 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3026 (bit_test_cluster::find_bit_tests): Likewise.
3027 (switch_decision_tree::analyze_switch_statement): Find clusters.
3028 * tree-switch-conversion.h (struct jump_table_cluster): Document
3031 2018-06-20 Martin Liska <mliska@suse.cz>
3033 * tree-switch-conversion.c (switch_conversion::collect):
3034 Record m_uniq property.
3035 (switch_conversion::expand): Bail out for special conditions.
3036 (group_cluster::~group_cluster): New.
3037 (group_cluster::group_cluster): Likewise.
3038 (group_cluster::dump): Likewise.
3039 (jump_table_cluster::emit): New.
3040 (switch_decision_tree::fix_phi_operands_for_edges): New.
3041 (struct case_node): Remove struct.
3042 (jump_table_cluster::can_be_handled): New.
3043 (case_values_threshold): Moved to header.
3044 (reset_out_edges_aux): Likewise.
3045 (jump_table_cluster::is_beneficial): New.
3046 (bit_test_cluster::can_be_handled): Likewise.
3047 (add_case_node): Remove.
3048 (bit_test_cluster::is_beneficial): New.
3049 (case_bit_test::cmp): New.
3050 (bit_test_cluster::emit): New.
3051 (expand_switch_as_decision_tree_p): Remove.
3052 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
3053 (fix_phi_operands_for_edge): Likewise.
3054 (switch_decision_tree::analyze_switch_statement): New.
3055 (compute_cases_per_edge): Move ...
3056 (switch_decision_tree::compute_cases_per_edge): ... here.
3057 (try_switch_expansion): Likewise.
3058 (switch_decision_tree::try_switch_expansion): Likewise.
3059 (record_phi_operand_mapping): Likewise.
3060 (switch_decision_tree::record_phi_operand_mapping): Likewise.
3061 (emit_case_decision_tree): Likewise.
3062 (switch_decision_tree::emit): Likewise.
3063 (balance_case_nodes): Likewise.
3064 (switch_decision_tree::balance_case_nodes): Likewise.
3065 (dump_case_nodes): Likewise.
3066 (switch_decision_tree::dump_case_nodes): Likewise.
3067 (emit_jump): Likewise.
3068 (switch_decision_tree::emit_jump): Likewise.
3069 (emit_cmp_and_jump_insns): Likewise.
3070 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
3071 (emit_case_nodes): Likewise.
3072 (switch_decision_tree::emit_case_nodes): Likewise.
3073 (conditional_probability): Remove.
3074 * tree-switch-conversion.h (enum cluster_type): New.
3076 (struct cluster): Likewise.
3077 (cluster::cluster): Likewise.
3078 (struct simple_cluster): Likewise.
3079 (simple_cluster::simple_cluster): Likewise.
3080 (struct group_cluster): Likewise.
3081 (struct jump_table_cluster): Likewise.
3082 (struct bit_test_cluster): Likewise.
3083 (struct min_cluster_item): Likewise.
3084 (struct case_tree_node): Likewise.
3085 (case_tree_node::case_tree_node): Likewise.
3086 (jump_table_cluster::case_values_threshold): Likewise.
3087 (struct case_bit_test): Likewise.
3088 (struct switch_decision_tree): Likewise.
3089 (struct switch_conversion): Likewise.
3090 (switch_decision_tree::reset_out_edges_aux): Likewise.
3092 2018-06-20 Martin Liska <mliska@suse.cz>
3094 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
3095 (hoist_edge_and_branch_if_true): Likewise.
3096 (expand_switch_using_bit_tests_p): Likewise.
3097 (struct case_bit_test): Likewise.
3098 (case_bit_test_cmp): Likewise.
3099 (emit_case_bit_tests): Likewise.
3100 (switch_conversion::switch_conversion): New class.
3101 (struct switch_conv_info): Remove old struct.
3102 (collect_switch_conv_info): More to ...
3103 (switch_conversion::collect): ... this.
3104 (check_range): Likewise.
3105 (switch_conversion::check_range): Likewise.
3106 (check_all_empty_except_final): Likewise.
3107 (switch_conversion::check_all_empty_except_final): Likewise.
3108 (check_final_bb): Likewise.
3109 (switch_conversion::check_final_bb): Likewise.
3110 (create_temp_arrays): Likewise.
3111 (switch_conversion::create_temp_arrays): Likewise.
3112 (free_temp_arrays): Likewise.
3113 (gather_default_values): Likewise.
3114 (switch_conversion::gather_default_values): Likewise.
3115 (build_constructors): Likewise.
3116 (switch_conversion::build_constructors): Likewise.
3117 (constructor_contains_same_values_p): Likewise.
3118 (switch_conversion::contains_same_values_p): Likewise.
3119 (array_value_type): Likewise.
3120 (switch_conversion::array_value_type): Likewise.
3121 (build_one_array): Likewise.
3122 (switch_conversion::build_one_array): Likewise.
3123 (build_arrays): Likewise.
3124 (switch_conversion::build_arrays): Likewise.
3125 (gen_def_assigns): Likewise.
3126 (switch_conversion::gen_def_assigns): Likewise.
3127 (prune_bbs): Likewise.
3128 (switch_conversion::prune_bbs): Likewise.
3129 (fix_phi_nodes): Likewise.
3130 (switch_conversion::fix_phi_nodes): Likewise.
3131 (gen_inbound_check): Likewise.
3132 (switch_conversion::gen_inbound_check): Likewise.
3133 (process_switch): Use the newly created class.
3134 (switch_conversion::expand): New.
3135 (switch_conversion::~switch_conversion): New.
3136 * tree-switch-conversion.h: New file.
3138 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3140 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
3141 tree-vect-patterns.c.
3142 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
3143 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
3144 (vect_recog_sad_pattern): Likewise.
3145 (vect_recog_widen_sum_pattern): Likewise.
3146 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
3147 (vect_recog_widen_shift_pattern): Remove the type_in argument.
3148 (vect_recog_rotate_pattern): Likewise.
3149 (vect_recog_mult_pattern): Likewise.
3150 (vect_recog_vector_vector_shift_pattern): Likewise.
3151 (vect_recog_divmod_pattern): Likewise.
3152 (vect_recog_mixed_size_cond_pattern): Likewise.
3153 (vect_recog_bool_pattern): Likewise.
3154 (vect_recog_mask_conversion_pattern): Likewise.
3155 (vect_try_gather_scatter_pattern): Likewise.
3156 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
3157 (vect_recog_over_widening_pattern): Likewise.
3158 (vect_recog_gather_scatter_pattern): Likewise.
3159 (vect_recog_func_ptr): Move from tree-vectorizer.h
3160 (vect_vect_recog_func_ptrs): Move further down the file.
3161 (vect_recog_func): Likewise. Remove the third argument.
3162 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
3163 (vect_pattern_recog_1): Expect the pattern function to do any
3164 necessary target tests. Also expect it to provide a vector type.
3165 Remove the type_in handling.
3167 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3169 * tree-vect-patterns.c (vect_pattern_detected): New function.
3170 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
3171 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
3172 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
3173 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
3174 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
3175 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
3176 (vect_recog_mask_conversion_pattern)
3177 (vect_try_gather_scatter_pattern): Likewise.
3179 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3181 * tree-vect-patterns.c (vect_get_internal_def): New function.
3182 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
3183 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
3184 (search_type_for_mask_1): Use it.
3186 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3188 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
3189 redundant WIDEN_SUM_EXPR handling.
3190 (vect_recog_sad_pattern): Likewise.
3192 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3194 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
3195 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
3196 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
3197 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
3198 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
3200 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3202 * tree-vect-stmts.c (vectorizable_call): Make sure that we
3203 use the stmt_vec_info of the original bb statement for the
3204 new zero assignment, even if the call is part of a pattern.
3206 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3208 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
3209 that the sequence is attached to the original statement rather
3210 than the pattern statement.
3211 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
3212 PATTERN_DEF_SEQ from the original statement rather than
3213 the main pattern statement.
3214 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
3215 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
3216 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
3218 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3220 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
3221 definition statements before the early exit for statements that aren't
3223 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
3225 (vect_transform_loop): ...here. Process pattern definition
3226 statements without first checking whether the main pattern
3227 statement is live or relevant.
3229 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
3231 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
3232 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
3234 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
3236 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
3237 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
3238 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
3239 (expand_block_compare): Change select_block_compare_mode call.
3240 (expand_strncmp_align_check): Use new functions, fix comment.
3241 (emit_final_str_compare_gpr): New function.
3242 (expand_strn_compare): Refactor and clean up code.
3243 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
3245 2018-06-19 Tony Reix <tony.reix@atos.com>
3246 Damien Bergamini <damien.bergamini@atos.com>
3247 David Edelsohn <dje.gcc@gmail.com>
3249 * collect2.c (static_obj): New variable.
3250 (static_libs): New variable.
3251 (is_in_list): Uncomment declaration.
3252 (main): Track AIX libraries linked statically.
3253 (is_in_list): Uncomment definition.
3254 (scan_prog_file): Don't add AIX shared libraries initializer
3255 to constructor list if linking statically.
3257 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
3259 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
3261 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
3263 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
3265 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
3268 2018-06-19 Martin Liska <mliska@suse.cz>
3270 * config/i386/i386.c (ix86_can_inline_p): Do not use
3271 ipa_fn_summaries::get_create.
3272 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
3274 (devirtualization_time_bonus): Likewise.
3275 (ipcp_propagate_stage): Likewise.
3276 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3277 (edge_set_predicate): Likewise.
3278 (evaluate_conditions_for_known_args): Likewise.
3279 (evaluate_properties_for_edge): Likewise.
3280 (ipa_call_summary::reset): Tranform to ...
3281 (ipa_call_summary::~ipa_call_summary): ... this.
3282 (ipa_fn_summary::reset): Transform to ...
3283 (ipa_fn_summary::~ipa_fn_summary): ... this.
3284 (ipa_fn_summary_t::remove): Rename to ...
3285 (ipa_fn_summary_t::remove_callees): ... this.
3286 (ipa_fn_summary_t::duplicate): Use placement new
3287 instead of memory copy.
3288 (ipa_call_summary_t::duplicate): Likewise.
3289 (ipa_call_summary_t::remove): Remove.
3290 (dump_ipa_call_summary): Change get_create to get.
3291 (ipa_dump_fn_summary): Dump only when summary exists.
3292 (analyze_function_body): Use symbol_summary::get instead
3294 (compute_fn_summary): Likewise.
3295 (estimate_edge_devirt_benefit): Likewise.
3296 (estimate_edge_size_and_time): Likewise.
3297 (inline_update_callee_summaries): Likewise.
3298 (remap_edge_change_prob): Likewise.
3299 (remap_edge_summaries): Likewise.
3300 (ipa_merge_fn_summary_after_inlining): Likewise.
3301 (write_ipa_call_summary): Likewise.
3302 (ipa_fn_summary_write): Likewise.
3303 (ipa_free_fn_summary): Likewise.
3304 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
3305 (struct ipa_call_summary): Likewise.
3306 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
3308 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
3309 (estimate_size_after_inlining): Likewise.
3310 (estimate_growth): Likewise.
3311 (growth_likely_positive): Likewise.
3312 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3313 (inline_call): Likewise.
3314 * ipa-inline.c (caller_growth_limits): Likewise.
3315 (can_inline_edge_p): Likewise.
3316 (can_inline_edge_by_limits_p): Likewise.
3317 (compute_uninlined_call_time): Likewise.
3318 (compute_inlined_call_time): Likewise.
3319 (want_inline_small_function_p): Likewise.
3320 (edge_badness): Likewise.
3321 (update_caller_keys): Likewise.
3322 (update_callee_keys): Likewise.
3323 (inline_small_functions): Likewise.
3324 (inline_to_all_callers_1): Likewise.
3325 (dump_overall_stats): Likewise.
3326 (early_inline_small_functions): Likewise.
3327 (early_inliner): Likewise.
3328 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3329 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3330 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3331 * ipa-split.c (execute_split_functions): Likewise.
3332 * symbol-summary.h: Likewise.
3333 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3335 2018-06-19 Richard Biener <rguenther@suse.de>
3337 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
3338 (vectorize_loops): ... here. Fix dbgcnt handling.
3339 (try_vectorize_loop): Wrap try_vectorize_loop_1.
3341 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
3344 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
3345 ieee128 argument takes up only one (vector) register, not two (floating
3348 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
3350 * gimplify.c (gimplify_init_constructor): Really never clear for an
3351 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
3353 2018-06-19 Richard Biener <rguenther@suse.de>
3355 PR tree-optimization/86179
3356 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
3357 after failed recognition.
3359 2018-06-18 Martin Sebor <msebor@redhat.com>
3362 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
3363 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
3364 Handle integer subtraction.
3365 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
3366 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
3368 2018-06-18 David Malcolm <dmalcolm@redhat.com>
3370 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
3371 param from rtx to rtx_insn *.
3372 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
3374 (frv_ifcvt_modify_insn): Likwise.
3375 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
3376 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
3377 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
3378 as_a <rtx_insn *> cast to local "unprotected_region" once
3379 it's been established that it's not NULL or pc_rtx.
3380 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
3381 param "sethi" from rtx to rtx_insn *.
3382 (nds32_group_float_insns): Likewise for param "insn".
3383 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
3385 (vax_output_int_subtract): Likewise.
3386 * config/vax/vax.c (vax_output_int_add): Likewise for param
3388 (vax_output_int_subtract): Likewise.
3389 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
3390 (emit_pattern_after): Likewise for param "after".
3391 (emit_insn_after): Likewise.
3392 (emit_jump_insn_after): Likewise.
3393 (emit_call_insn_after): Likewise.
3394 (emit_debug_insn_after): Likewise.
3395 (emit_pattern_before): Likewise for param "before".
3396 (emit_insn_before): Likewise.
3397 (emit_jump_insn_before): Likewise.
3398 * final.c (get_insn_template): Likewise for param "insn", removing
3400 * output.h (get_insn_template): Likewise for 2nd param.
3401 * rtl.h (emit_insn_before): Likewise.
3402 (emit_jump_insn_before): Likewise.
3403 (emit_debug_insn_before_noloc): Likewise.
3404 (emit_insn_after): Likewise.
3405 (emit_jump_insn_after): Likewise.
3406 (emit_call_insn_after): Likewise.
3407 (emit_debug_insn_after): Likewise.
3408 (set_insn_deleted): Likewise for param.
3410 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
3413 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
3414 floating point modes, so that IFmode is numerically greater than
3415 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
3416 to declare the ordering. This prevents IFmode from being
3417 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
3418 machine. Include rs6000-modes.h to share the fractional values
3419 between genmodes* and the rest of the compiler.
3423 * config/rs6000/rs6000-modes.h: New file.
3424 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
3425 meaning of rs6000_long_double_size so that 126..128 selects an
3426 appropriate 128-bit floating point type.
3427 (rs6000_option_override_internal): Likewise.
3428 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
3429 (TARGET_LONG_DOUBLE_128): Change the meaning of
3430 rs6000_long_double_size so that 126..128 selects an appropriate
3431 128-bit floating point type.
3432 (LONG_DOUBLE_TYPE_SIZE): Update comment.
3433 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
3434 source and destination to match the standard usage.
3435 (truncifkf2): Likewise.
3436 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
3437 ISA 2.07 to use an explicit clobber, instead of passing in a
3439 (copysign<mode>3_soft): Likewise.
3441 2018-06-18 David Malcolm <dmalcolm@redhat.com>
3443 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
3444 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
3445 (vect_slp_analyze_instance_dependence): Likewise.
3446 (vect_enhance_data_refs_alignment): Likewise.
3447 (vect_analyze_data_refs_alignment): Likewise.
3448 (vect_slp_analyze_and_verify_instance_alignment
3449 (vect_analyze_data_ref_accesses): Likewise.
3450 (vect_prune_runtime_alias_test_list): Likewise.
3451 (vect_analyze_data_refs): Likewise.
3452 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
3453 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
3454 (vect_analyze_scalar_cycles_1): Likewise.
3455 (vect_get_loop_niters): Likewise.
3456 (vect_analyze_loop_form_1): Likewise.
3457 (vect_update_vf_for_slp): Likewise.
3458 (vect_analyze_loop_operations): Likewise.
3459 (vect_analyze_loop): Likewise.
3460 (vectorizable_induction): Likewise.
3461 (vect_transform_loop): Likewise.
3462 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
3463 * tree-vect-slp.c (vect_analyze_slp): Likewise.
3464 (vect_make_slp_decision): Likewise.
3465 (vect_detect_hybrid_slp): Likewise.
3466 (vect_slp_analyze_operations): Likewise.
3467 (vect_slp_bb): Likewise.
3468 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
3469 (vectorizable_bswap): Likewise.
3470 (vectorizable_call): Likewise.
3471 (vectorizable_simd_clone_call): Likewise.
3472 (vectorizable_conversion): Likewise.
3473 (vectorizable_assignment): Likewise.
3474 (vectorizable_shift): Likewise.
3475 (vectorizable_operation): Likewise.
3476 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
3478 2018-06-18 Martin Sebor <msebor@redhat.com>
3480 PR tree-optimization/81384
3481 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
3482 * builtins.c (expand_builtin_strnlen): New function.
3483 (expand_builtin): Call it.
3484 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
3485 * builtins.def (BUILT_IN_STRNLEN): New.
3486 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
3487 Warn for bounds in excess of maximum object size.
3488 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
3489 single-value ranges. Handle strnlen.
3490 (handle_builtin_strlen): Handle strnlen.
3491 (strlen_check_and_optimize_stmt): Same.
3492 * doc/extend.texi (Other Builtins): Document strnlen.
3494 2018-06-18 Maya Rashish <coypu@sdf.org>
3496 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
3497 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
3498 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
3500 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
3502 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
3504 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3506 * tree.c (escaped_string::escape): Replace cast to char * by
3507 const_cast<char *> (unescaped).
3509 2018-06-18 Nick Clifton <nickc@redhat.com>
3512 * tree.c (escaped_string): New class. Converts an unescaped
3513 string into its escaped equivalent.
3514 (warn_deprecated_use): Use the new class to convert the
3515 deprecation message, if present.
3516 (test_escaped_strings): New self test.
3517 (test_c_tests): Add test_escaped_strings.
3518 * doc/extend.texi (deprecated): Add a note that the
3519 deprecation message is affected by the -fmessage-length
3520 option, and that control characters will be escaped.
3521 (#pragma GCC error): Document this pragma.
3522 (#pragma GCC warning): Likewise.
3523 * doc/invoke.texi (-fmessage-length): Document this option's
3524 effect on the #warning and #error preprocessor directives and
3525 the deprecated attribute.
3527 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
3529 * tree.c (decl_value_expr_lookup): Revert latest change.
3530 (decl_value_expr_insert): Likewise.
3532 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
3534 * gimplify.c (nonlocal_vlas): Delete.
3535 (nonlocal_vla_vars): Likewise.
3536 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
3538 (gimplify_body): Do not create and destroy nonlocal_vlas.
3539 * tree-nested.c: Include diagnostic.h.
3540 (use_pointer_in_frame): Tweak.
3541 (lookup_field_for_decl): Add assertion and declare the transformation.
3542 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
3543 internal error when the reference is in a wrong context. Do not
3544 create a debug decl by default.
3545 (note_nonlocal_block_vlas): Delete.
3546 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
3547 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
3548 create a debug decl by default.
3549 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
3550 call is in a wrong context.
3551 (fixup_vla_decls): New function.
3552 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
3553 debug variables were created.
3554 * tree.c (decl_value_expr_lookup): Add checking assertion.
3555 (decl_value_expr_insert): Likewise.
3557 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
3560 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
3561 * tree-scalar-evolution.c (interpret_expr): Likewise.
3562 (expression_expensive_p): Likewise.
3563 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
3564 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
3565 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
3566 (ssa_defined_by_minus_one_stmt_p): New.
3568 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
3571 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
3572 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
3573 * dojump.c (do_jump): Likewise.
3574 * expr.c (expand_expr_real_2): Check operand type's sign.
3575 * fold-const.c (const_unop): Handle ABSU_EXPR.
3576 (fold_abs_const): Likewise.
3577 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
3578 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
3579 (strip_sign_op_1): Likesise.
3580 * match.pd: Add new pattern to generate ABSU_EXPR.
3581 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
3582 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3583 * tree-eh.c (operation_could_trap_helper_p): Likewise.
3584 * tree-inline.c (estimate_operator_cost): Likewise.
3585 * tree-pretty-print.c (dump_generic_node): Likewise.
3586 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
3587 * tree.def (ABSU_EXPR): New.
3589 2018-06-16 Jakub Jelinek <jakub@redhat.com>
3592 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
3593 documented as preserved for backward compatibility only.
3594 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
3596 PR rtl-optimization/86108
3597 * bb-reorder.c (create_forwarder_block): Renamed to ...
3598 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
3599 jump from new landing pad to the second part.
3600 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
3603 2018-06-15 Jakub Jelinek <jakub@redhat.com>
3606 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
3607 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
3608 Only call store_expr for halves if the mode is the same.
3611 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
3612 Fix up comment formatting.
3614 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
3616 * typed-splay-tree.h (typed_splay_tree::remove): New function.
3617 (typed_splay_tree::closure,
3618 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
3619 (typed_splay_tree::typed_splay_tree,
3620 typed_splay_tree::operator =): Declared private.
3621 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
3622 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
3623 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
3624 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
3625 typed_splay_tree::splay_tree_splay,
3626 typed_splay_tree::splay_tree_foreach_helper,
3627 typed_splay_tree::splay_tree_insert,
3628 typed_splay_tree::splay_tree_remove,
3629 typed_splay_tree::splay_tree_lookup,
3630 typed_splay_tree::splay_tree_predecessor,
3631 typed_splay_tree::splay_tree_successor,
3632 typed_splay_tree::splay_tree_min,
3633 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
3634 (typed_splay_tree::root, typed_splay_tree::comp,
3635 typed_splay_tree::delete_key,
3636 typed_splay_tree::delete_value): New data members.
3637 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
3638 typed_splay_tree::remove.
3640 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
3642 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
3643 -mginv and -mno-ginv to the assembler.
3644 * config/mips/mips.opt (-mcrc): New option.
3646 * doc/invoke.text (-mcrc): Document.
3649 2018-06-15 Nick Clifton <nickc@redhat.com>
3652 * tree.c (escaped_string): New class. Converts an unescaped
3653 string into its escaped equivalent.
3654 (warn_deprecated_use): Use the new class to convert the
3655 deprecation message, if present.
3656 (test_escaped_strings): New self test.
3657 (test_c_tests): Add test_escaped_strings.
3658 * doc/extend.texi (deprecated): Add a note that the
3659 deprecation message is affected by the -fmessage-length
3660 option, and that control characters will be escaped.
3661 (#pragma GCC error): Document this pragma.
3662 (#pragma GCC warning): Likewise.
3663 * doc/invoke.texi (-fmessage-length): Document this option's
3664 effect on the #warning and #error preprocessor directives and
3665 the deprecated attribute.
3667 2018-06-15 Richard Biener <rguenther@suse.de>
3669 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
3670 here, also noting vector size used.
3671 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
3672 size used in MSG_OPTIMIZED_LOCATIONS dump.
3673 (pass_slp_vectorize::execute): Adjust.
3675 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
3678 * config/arc/arc.c (arc_return_address_register): Fix
3681 2018-06-15 Richard Biener <rguenther@suse.de>
3684 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
3685 leave useless conversion stripping to force_gimple_operand_gsi.
3686 (gimplify_build2): Likewise.
3687 (gimplify_build1): Likewise.
3689 2018-06-15 Richard Biener <rguenther@suse.de>
3692 * tree-cfg.c (move_stmt_op): unshare invariant addresses
3693 before adjusting their block.
3695 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
3697 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
3698 multilibs for *-*-rtems*.
3699 * config/riscv/t-rtems: New file.
3701 2018-06-14 Jakub Jelinek <jakub@redhat.com>
3704 * match.pd ((A +- CST1) +- CST2): Punt if last resort
3705 unsigned_type_for returns NULL.
3708 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
3709 subregs of multi-word pseudos unless the float mode has word size.
3711 2018-06-14 Richard Biener <rguenther@suse.de>
3714 * tree-vect-generic.c (build_word_mode_vector_type): Remove
3715 duplicate and harmful type_hash_canon.
3716 * tree.c (type_hash_canon): Assert we didn't find ourselves.
3718 2018-06-14 Richard Biener <rguenther@suse.de>
3721 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
3724 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
3726 * config/rtems.h (STDINT_LONG32): Define.
3728 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
3729 Prachi Godbole <prachi.godbole@imgtec.com>
3731 * config/mips/mips-cpus.def: Define P6600.
3732 * config/mips/mips-tables.opt: Regenerate.
3733 * config/mips/mips.c (mips_ucbranch_type): New enum.
3734 (mips_rtx_cost_data): Add support for P6600.
3735 (mips_issue_rate): Likewise.
3736 (mips_multipass_dfa_lookahead): Likewise.
3737 (mips_avoid_hazard): Likewise.
3738 (mips_reorg_process_insns): Likewise.
3739 (mips_classify_branch_p6600): New function.
3740 * config/mips/mips.h (TUNE_P6600): New define.
3741 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
3742 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
3743 * config/mips/mips.md: Include p6600.md.
3744 (processor): Add p6600.
3745 * config/mips/p6600.md: New file.
3746 * doc/invoke.texi: Add p6600 to supported architectures.
3748 2018-06-13 Martin Sebor <msebor@redhat.com>
3750 PR tree-optimization/86114
3751 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
3753 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
3755 2018-06-13 Richard Biener <rguenther@suse.de>
3757 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
3758 Properly set vector type of the intermediate stmt.
3759 * tree-vect-stmts.c (vectorizable_operation): The destination
3760 var always has vectype_out type.
3762 2018-06-13 Jeff Law <law@redhat.com>
3764 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
3765 integer 0 for argument to print_rtl_with_bb.
3766 (rl78_reorg): Likewise.
3768 2018-06-13 David Malcolm <dmalcolm@redhat.com>
3770 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
3771 from rtx to rtx_insn *.
3772 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
3774 (add_sched_insns_for_speculation): Likewise for local "target",
3775 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3776 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
3777 from rtx_insn ** to rtx_code_label **.
3778 (reorg_emit_nops): Likewise.
3779 (c6x_reorg): Likewise for local "call_labels".
3780 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
3782 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
3783 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
3784 the loops over LABEL_REFS.
3785 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
3787 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3788 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
3789 (split_branches): Strengthen local "olabel" from rtx to
3790 rtx_insn *, adding a safe_as_a cast.
3791 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
3793 (add_insn_after): Likewise for first two params.
3794 (add_insn_before): Likewise.
3795 (remove_insn): Likewise for param.
3796 (emit_pattern_before_noloc): Likewise for second and third params.
3797 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
3798 (emit_call_insn_before_noloc): Likewise.
3799 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
3801 (emit_barrier_before): Likewise.
3802 (emit_label_before): Strengthen "label" param from "rtx" to
3803 "rtx_code_label *". Strengthen "before" param from "rtx" to
3805 (emit_insn_after_1): Strengthen "after" param from "rtx" to
3807 (emit_pattern_after_noloc): Likewise.
3808 (emit_insn_after_noloc): Likewise.
3809 (emit_jump_insn_after_noloc): Likewise.
3810 (emit_call_insn_after_noloc): Likewise.
3811 (emit_debug_insn_after_noloc): Likewise.
3812 (emit_barrier_after): Likewise.
3813 (emit_label_after): Likewise for both params.
3814 (emit_pattern_after_setloc): Likewise for "after" param. Convert
3815 "loc" param from "int" to "location_t".
3816 (emit_insn_after_setloc): Likewise.
3817 (emit_jump_insn_after_setloc): Likewise.
3818 (emit_call_insn_after_setloc): Likewise.
3819 (emit_debug_insn_after_setloc): Likewise.
3820 (emit_pattern_before_setloc): Likewise for "before" param. Convert
3821 "loc" param from "int" to "location_t".
3822 (emit_pattern_before): Convert NULL_RTX to NULL.
3823 (emit_insn_before_setloc): Convert "loc" param from "int" to
3825 (emit_jump_insn_before_setloc): Likewise.
3826 (emit_call_insn_before_setloc): Likewise.
3827 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
3828 rtx_insn *. Convert "loc" param from "int" to "location_t".
3829 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
3830 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
3831 Convert 3rd param from "int" to "location_t".
3832 (emit_barrier_before, emit_barrier_after, next_real_insn):
3833 Strengthen param from rtx to rtx_insn *.
3834 (emit_label_before): Strengthen 1st param from "rtx" to
3835 "rtx_code_label *". Strengthen 2nd param from "rtx" to
3837 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
3838 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
3839 Strengthen 2nd param from "rtx" to "rtx_insn *".
3840 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
3841 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
3842 Likewise. Convert 3rd param from "int" to "location_t".
3843 (emit_label_after): Strengthen 1st param from "rtx" to
3845 (next_real_insn, remove_insn): Strengthen param from "rtx" to
3847 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
3848 from "rtx" to "rtx_insn *".
3850 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
3852 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
3853 bodies streamed in with -Q.
3854 * dumpfile.c (dump_files): Add lto-stream-out dump file.
3855 * dumpfile.h (tree_dump_index): Add lto_stream_out.
3856 * gimple-streamer-out.c: Include gimple-pretty-print.h
3857 (output_bb): Dump stmts streamed.
3858 * lto-section-out.c: Include print-tree.h
3859 (lto_begin_section): Dump sections created.
3860 (lto_output_decl_index): Dump decl encoded.
3861 * lto-streamer-out.c: Include print-tree.h
3862 (create_output_block): Dump output block created.
3863 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
3864 (output_function): Dump function output.
3865 (output_constructor): Dump constructor streamed.
3866 (write_global_stream): Output indexes encoded.
3867 (produce_asm_for_decls): Dump streams encoded.
3868 * lto-streamer.c (streamer_dump_file): New global var.
3869 * lto-streamer.h (streamer_dump_file): Declare.
3870 * passes.c (ipa_write_summaries): Initialize streamer dump.
3871 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
3874 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
3877 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
3878 offsets for register save directives. Emit a second batch of save
3879 directives, if need be, when the function accesses prior frames.
3881 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3883 * config/arc/fpu.md (fmasf4): Force operand to register.
3884 (fnmasf4): Likewise.
3886 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3888 * config/arc/arc-protos.h (arc_pad_return): Remove.
3889 * config/arc/arc.c (machine_function): Remove force_short_suffix
3891 (arc_print_operand): Adjust printing of '&'.
3892 (arc_verify_short): Remove conditional printing of short suffix.
3893 (arc_final_prescan_insn): Remove reference to size_reason.
3894 (pad_return): New function.
3895 (arc_reorg): Call pad_return.
3896 (arc_pad_return): Remove.
3897 (arc_init_machine_status): Remove reference to force_short_suffix.
3898 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
3899 (attr length): When attribute iscompact is true force to 2
3900 regardless; in the case of maybe check if we want to force the
3901 instruction to have 4 bytes length.
3902 (nopv): Change it to generate 4 byte long nop as well.
3903 (blockage): New pattern.
3904 (simple_return): Remove call to arc_pad_return.
3905 (p_return_i): Likewise.
3907 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3909 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
3911 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3913 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
3916 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3918 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
3919 for ARC700 and ARCv2.
3921 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
3924 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
3925 operands[2] instead of operands[1].
3928 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3930 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
3931 case, check whether the outer register overlaps an unallocatable
3932 register, not just whether it fits the required class.
3934 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3936 * poly-int.h (can_div_trunc_p): Add new overload in which all values
3938 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
3939 (memrefs_conflict_p): Likewise.
3940 (init_alias_analysis): Likewise.
3941 * cfgexpand.c (expand_debug_expr): Likewise.
3942 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
3943 * cse.c (fold_rtx): Likewise.
3944 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
3945 * expr.c (emit_block_move_hints): Likewise.
3946 (clear_storage_hints, push_block, emit_push_insn): Likewise.
3947 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
3948 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
3949 (emit_group_store): Likewise.
3950 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
3951 to read the PRE/POST_MODIFY increment.
3952 * calls.c (store_one_arg): Use strip_offset.
3953 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
3955 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
3957 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
3958 (simplify_binary_operation_1): Extend CONST_INT handling to
3960 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
3961 than a HOST_WIDE_INT.
3962 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
3964 (adjust_mems, add_stores): Update accodingly.
3965 (vt_canonicalize_addr): Track polynomial offsets.
3966 (emit_note_insn_var_location): Likewise.
3967 (vt_add_function_parameter): Likewise.
3968 (vt_initialize): Likewise.
3970 2018-06-12 Jeff Law <law@redhat.com>
3972 * config.gcc (alpha*-*-freebsd*): Remove.
3973 * config/alpha/freebsd.h: Remove.
3975 2018-06-12 David Malcolm <dmalcolm@redhat.com>
3978 * spellcheck-tree.c (levenshtein_distance): Rename to...
3979 (get_edit_distance): ...this, and update for underlying renaming.
3980 * spellcheck-tree.h (levenshtein_distance): Rename to...
3981 (get_edit_distance): ...this.
3982 * spellcheck.c (levenshtein_distance): Rename to...
3983 (get_edit_distance): ...this. Convert from Levenshtein distance
3984 to Damerau-Levenshtein distance by supporting transpositions of
3985 adjacent characters. Rename "v1" to "v_next" and "v0" to
3987 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
3988 (selftest::test_edit_distance_unit_test_oneway): ...this, and
3989 update for underlying renaming.
3990 (selftest::levenshtein_distance_unit_test): Rename to...
3991 (selftest::test_get_edit_distance_unit): ...this, and update for
3992 underlying renaming.
3993 (selftest::test_find_closest_string): Add example from PR 69968
3994 where transposition helps
3995 (selftest::test_metric_conditions): Update for renaming.
3996 (selftest::test_metric_conditions): Likewise.
3997 (selftest::spellcheck_c_tests): Likewise.
3998 * spellcheck.h (levenshtein_distance): Rename both overloads to...
3999 (get_edit_distance): ...this.
4000 (best_match::consider): Update for renaming.
4002 2018-06-12 Martin Sebor <msebor@redhat.com>
4004 PR tree-optimization/85259
4005 * builtins.c (compute_objsize): Handle constant offsets.
4006 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
4007 true iff a warning has been issued.
4008 * gimple.h (gimple_nonartificial_location): New function.
4009 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
4010 gimple_nonartificial_location and handle -Wno-system-headers.
4011 (handle_builtin_stxncpy): Same.
4013 2018-06-12 Martin Sebor <msebor@redhat.com>
4016 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
4018 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
4020 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
4021 BUILTIN_VEC_XST entries for pointer to double and long long.
4023 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
4026 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
4028 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
4031 2018-06-12 Martin Liska <mliska@suse.cz>
4033 * doc/options.texi: Document IntegerRange.
4035 2018-06-12 Martin Liska <mliska@suse.cz>
4037 * config/i386/i386.opt: Make MPX-related options as Deprecated.
4038 * opt-functions.awk: Handle Deprecated flag.
4039 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
4041 (read_cmdline_option): Report warning for a deprecated option.
4042 * opts.h (struct cl_option): Add new field cl_deprecated.
4043 (CL_ERR_DEPRECATED): New.
4045 2018-06-12 Martin Liska <mliska@suse.cz>
4047 * doc/options.texi: Document Deprecated option flag.
4049 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4051 * config/arc/arc-arch.h (arc_extras): New enum.
4052 (arc_cpu_t):Add field extra.
4053 (arc_cpu_types): Consider the extras.
4054 * config/arc/arc-cpus.def: Add extras info.
4055 * config/arc/arc-opts.h (processor_type): Consider extra field.
4056 * config/arc/arc.c (arc_override_options): Handle extra field.
4058 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4060 * config/arc/arc-arch.h: Update ARC_OPTX macro.
4061 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
4063 * config/arc/arc.c (arc_init): Update pic warning.
4064 (irq_range): Update irq range parsing warnings.
4065 (arc_override_options): Update various warning messages.
4066 (arc_handle_aux_attribute): Likewise.
4068 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
4070 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
4072 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4074 * doc/sourcebuild.texi: Document usage of line number 0 in verify
4075 compiler messages directives.
4077 2018-06-12 Matthew Fortune <mfortune@gmail.com>
4079 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
4080 * config/mips/mips-tables.opt: Regenerate.
4081 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
4083 * doc/invoke.texi: Document -march=i6500.
4085 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
4087 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
4088 (i6400_gpmul): Add cpu_unit.
4089 (i6400_gpdiv): Likewise.
4090 (i6400_msa_add_d): Update reservations.
4091 (i6400_msa_int_add) Likewise.
4092 (i6400_msa_short_logic3) Likewise.
4093 (i6400_msa_short_logic2) Likewise.
4094 (i6400_msa_short_logic) Likewise.
4095 (i6400_msa_move) Likewise.
4096 (i6400_msa_cmp) Likewise.
4097 (i6400_msa_short_float2) Likewise.
4098 (i6400_msa_div_d) Likewise.
4099 (i6400_msa_long_logic1) Likewise.
4100 (i6400_msa_long_logic2) Likewise.
4101 (i6400_msa_mult) Likewise.
4102 (i6400_msa_long_float2) Likewise.
4103 (i6400_msa_long_float4) Likewise.
4104 (i6400_msa_long_float5) Likewise.
4105 (i6400_msa_long_float8) Likewise.
4106 (i6400_fpu_fadd): Include frint type.
4107 (i6400_fpu_store): New define_insn_reservation.
4108 (i6400_fpu_load): Likewise.
4109 (i6400_fpu_move): Likewise.
4110 (i6400_fpu_fcmp): Likewise.
4111 (i6400_fpu_fmadd): Likewise.
4112 (i6400_int_mult): Include imul3nc type and update reservation.
4113 (i6400_int_div): Include idiv3 type and update reservation.
4114 (i6400_int_load): Update to check type not move_type.
4115 (i6400_int_store): Likewise.
4116 (i6400_int_prefetch): Set zero latency.
4118 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
4120 * gcc.c: Document new %@{...} sequence.
4121 (LINK_COMMAND_SPEC): Use it for the -L switches.
4122 (cpp_unique_options): Use it for the -I switches.
4123 (at_file_argbuf): New global variable.
4124 (in_at_file): Likewise.
4125 (alloc_args): Create at_file_argbuf.
4126 (clear_args): Truncate at_file_argbuf.
4127 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
4128 (open_at_file): New function.
4129 (close_at_file): Likewise.
4130 (create_at_file): Delete.
4131 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
4134 (validate_switches_from_spec): Deal with %@{...} sequence.
4135 (validate_switches): Likewise.
4136 (driver::finalize): Call clear_args.
4138 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
4140 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
4142 2018-06-11 Martin Sebor <msebor@redhat.com>
4144 * doc/invoke.texi (-Wall): List -Wc++17-compat.
4145 (Wno-class-memaccess): Add @opindex.
4146 (Wno-templates, Wno-multiple-inheritance): Same.
4147 (Wno-virtual-inheritance, Wno-namespaces): Same.
4148 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
4149 (Wno-format-overflow, Wno-format-truncation): Same.
4150 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
4151 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
4152 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
4153 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
4154 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
4155 (Wno-misspelled-isr): Same.
4157 2018-06-11 Martin Sebor <msebor@redhat.com>
4159 * PR tree-optimization/86083
4160 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
4162 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
4164 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
4166 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
4169 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
4170 on the correct operand.
4171 (*movdi_internal64): Ditto.
4173 2018-06-11 Martin Liska <mliska@suse.cz>
4175 PR tree-optimization/86089
4176 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
4178 2018-06-11 Julia Koval <julia.koval@intel.com>
4180 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
4181 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
4182 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
4184 2018-06-11 Olivier Hainque <hainque@adacore.com>
4186 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
4187 for Ada with strict dwarf2.
4189 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
4192 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
4195 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
4197 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
4199 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
4201 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
4204 2018-06-08 Martin Liska <mliska@suse.cz>
4206 * tree-cfg.h (debug_function): Fix argument type to match
4209 2018-06-08 Martin Liska <mliska@suse.cz>
4211 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
4212 Remove usage of MPX-related (and removed) fields.
4213 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
4215 2018-06-08 David Malcolm <dmalcolm@redhat.com>
4217 * cfg.c (debug): Use TDF_NONE rather than 0.
4218 * cfghooks.c (debug): Likewise.
4219 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
4220 (struct dump_option_value_info): Convert to...
4221 (struct kv_pair): ...this template type.
4222 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
4224 (optinfo_verbosity_options): Likewise.
4225 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
4227 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
4228 than int for "optgroup_flags" param.
4229 (dump_generic_expr_loc): Use dump_flags_t rather than int for
4231 (dump_dec): Likewise.
4232 (dump_finish): Use TDF_NONE rather than 0.
4233 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
4234 rather than int for "optgroup_flags" param. Use TDF_NONE rather
4235 than 0. Update for change to option_ptr.
4236 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
4237 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
4238 0. Update for changes to optinfo_verbosity_options and
4240 (opt_info_switch_p): Convert optgroup_flags from int to
4242 (dump_basic_block): Use dump_flags_t rather than int
4243 for "dump_kind" param.
4244 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
4245 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
4246 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
4247 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
4248 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
4249 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
4250 TDF_NONE): Convert from macros to...
4251 (enum dump_flag): ...this new enum.
4252 (dump_flags_t): Update to use enum.
4253 (operator|, operator&, operator~, operator|=, operator&=):
4254 Implement for dump_flags_t.
4255 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
4256 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
4257 Convert from macros to...
4258 (enum optgroup_flag): ...this new enum.
4259 (optgroup_flags_t): New typedef.
4260 (operator|, operator|=): Implement for optgroup_flags_t.
4261 (struct dump_file_info): Convert field "alt_flags" to
4262 dump_flags_t. Convert field "optgroup_flags" to
4264 (dump_basic_block): Use dump_flags_t rather than int for param.
4265 (dump_generic_expr_loc): Likewise.
4266 (dump_dec): Likewise.
4267 (dump_register): Convert param "optgroup_flags" to
4269 (opt_info_enable_passes): Likewise.
4270 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
4272 * gimple-pretty-print.c (debug): Likewise.
4273 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
4274 (merged_store_group::apply_stores): Likewise.
4275 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
4276 * gimple.c (verify_gimple_pp): Likewise.
4277 * graphite-poly.c (print_pbb_body): Likewise.
4278 * passes.c (pass_manager::register_one_dump_file): Convert
4279 local "optgroup_flags" to optgroup_flags_t.
4280 * print-tree.c (print_node): Use TDF_NONE rather than 0.
4282 (debug_body): Likewise.
4283 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
4284 to optgroup_flags_t.
4285 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
4287 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
4288 (convert_mult_to_fma): Likewise.
4289 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
4290 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
4291 * tree-vect-data-refs.c (dump_lower_bound): Convert param
4292 "dump_kind" to dump_flags_t.
4294 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
4296 * config/rs6000/rs6000.c (min, max): Delete.
4298 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
4300 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
4301 -mabi=spe and -mabi=no-spe.
4303 2018-06-08 Martin Liska <mliska@suse.cz>
4305 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
4306 where we expect an existing summary.
4308 2018-06-08 Martin Liska <mliska@suse.cz>
4310 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
4311 * ipa-inline.h (estimate_edge_growth): Likewise.
4313 2018-06-08 Martin Liska <mliska@suse.cz>
4315 * cgraph.c (function_version_hasher::hash): Use
4316 cgraph_node::get_uid ().
4317 (function_version_hasher::equal):
4318 * cgraph.h (cgraph_node::get_uid): New method.
4319 * ipa-inline.c (update_caller_keys): Use
4320 cgraph_node::get_uid ().
4321 (update_callee_keys): Likewise.
4322 * ipa-utils.c (searchc): Likewise.
4323 (ipa_reduced_postorder): Likewise.
4324 * lto-cgraph.c (input_node): Likewise.
4325 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
4326 * symbol-summary.h (symtab_insertion): Likewise.
4327 (symtab_removal): Likewise.
4328 (symtab_duplication): Likewise.
4329 * tree-pretty-print.c (dump_function_header): Likewise.
4330 * tree-sra.c (convert_callers_for_node): Likewise.
4332 2018-06-08 Martin Liska <mliska@suse.cz>
4334 * cgraph.c (symbol_table::create_edge): Always assign a new
4336 (symbol_table::free_edge): Do not recycle numbers.
4337 * cgraph.h (cgraph_edge::get): New method.
4338 * symbol-summary.h (symtab_removal): Use it.
4339 (symtab_duplication): Likewise.
4340 (call_summary::hashable_uid): Remove.
4342 2018-06-08 Martin Liska <mliska@suse.cz>
4344 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
4345 (initialize_growth_caches): Remove.
4346 (free_growth_caches): Likewise.
4347 (do_estimate_edge_time): Use edge_growth_cache.
4348 (do_estimate_edge_size): Likewise.
4349 (do_estimate_edge_hints): Likewise.
4350 * ipa-inline.c (reset_edge_caches): Likewise.
4351 (recursive_inlining): Likewise.
4352 (inline_small_functions): Likewise.
4353 * ipa-inline.h (initialize_growth_caches): Remove.
4354 (estimate_edge_size): Likewise.
4355 (estimate_edge_time): Likewise.
4356 (estimate_edge_hints): Likewise.
4357 (reset_edge_growth_cache): Likewise.
4358 * symbol-summary.h (call_summary::remove): New method.
4360 2018-06-08 Martin Liska <mliska@suse.cz>
4362 * ipa-cp.c (class edge_clone_summary): New summary.
4363 (grow_edge_clone_vectors): Remove.
4364 (ipcp_edge_duplication_hook): Remove.
4365 (class edge_clone_summary_t): New call_summary class.
4366 (ipcp_edge_removal_hook): Remove.
4367 (edge_clone_summary_t::duplicate): New function.
4368 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
4369 (create_specialized_node): Likewise.
4370 (ipcp_driver): Initialize edge_clone_summaries and do not
4373 2018-06-08 Martin Liska <mliska@suse.cz>
4375 * symbol-summary.h (get): New function.
4376 (call_summary::m_initialize_when_cloning): New class member.
4378 2018-06-08 Martin Liska <mliska@suse.cz>
4380 * cgraph.c (cgraph_node::remove): Do not recycle uid.
4381 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
4382 (symbol_table::allocate_cgraph_symbol): Do not set uid.
4383 * passes.c (uid_hash_t): Record removed_nodes by their uids.
4384 (remove_cgraph_node_from_order): Use the removed_nodes set.
4385 (do_per_function_toporder): Likwise.
4386 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
4387 instead of summary_uid.
4388 (symtab_removal): Likewise.
4389 (symtab_duplication): Likewise.
4391 2018-06-08 Martin Liska <mliska@suse.cz>
4393 * ipa-cp.c (ipcp_store_bits_results): Use
4394 ipcp_transformation_sum.
4395 (ipcp_store_vr_results): Likewise.
4396 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
4398 (ipcp_transformation_initialize): ... this.
4399 (ipa_set_node_agg_value_chain):
4400 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
4401 (write_ipcp_transformation_info): Likewise.
4402 (read_ipcp_transformation_info): Likewise.
4403 (ipcp_update_bits): Likewise.
4404 (ipcp_update_vr): Likewise.
4405 (ipcp_transform_function): Likewise.
4406 * ipa-prop.h: Rename ipcp_transformation_summary to
4407 ipcp_transformation.
4408 (class ipcp_transformation_t): New function summary.
4409 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
4410 (ipa_get_agg_replacements_for_node): Likewise.
4412 2018-06-08 Martin Liska <mliska@suse.cz>
4414 * ipa-pure-const.c (struct funct_state_d): Do it class instead
4416 (class funct_state_summary_t): New function_summary class.
4417 (has_function_state): Remove.
4418 (get_function_state): Likewise.
4419 (set_function_state): Likewise.
4420 (add_new_function): Likewise.
4421 (funct_state_summary_t::insert): New function.
4422 (duplicate_node_data): Remove.
4423 (remove_node_data): Remove.
4424 (funct_state_summary_t::duplicate): New function.
4425 (register_hooks): Create new funct_state_summaries.
4426 (pure_const_generate_summary): Use it.
4427 (pure_const_write_summary): Likewise.
4428 (pure_const_read_summary): Likewise.
4429 (propagate_pure_const): Likewise.
4430 (propagate_nothrow): Likewise.
4431 (dump_malloc_lattice): Likewise.
4432 (propagate_malloc): Likewise.
4433 (execute): Do not register hooks, just remove summary
4435 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
4438 2018-06-08 Martin Liska <mliska@suse.cz>
4440 * ipa-reference.c (remove_node_data): Remove.
4441 (duplicate_node_data): Likewise.
4442 (class ipa_ref_var_info_summary_t): New class.
4443 (class ipa_ref_opt_summary_t): Likewise.
4444 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
4445 (get_reference_optimization_summary): Use
4446 ipa_ref_opt_sum_summaries.
4447 (set_reference_vars_info): Remove.
4448 (set_reference_optimization_summary): Likewise.
4449 (ipa_init): Create summaries.
4450 (init_function_info): Use function summary.
4451 (ipa_ref_opt_summary_t::duplicate): New function.
4452 (ipa_ref_opt_summary_t::remove): New function.
4453 (get_read_write_all_from_node): Fix GNU coding style.
4454 (propagate): Use function summary.
4455 (write_node_summary_p): Fix GNU coding style.
4456 (stream_out_bitmap): Likewise.
4457 (ipa_reference_read_optimization_summary): Use function summary.
4458 (ipa_reference_c_finalize): Do not release hooks.
4460 2018-06-08 Martin Liska <mliska@suse.cz>
4462 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
4463 (analyze_function_body): Extract multiple calls of get_create.
4464 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
4465 * ipa-inline.c (recursive_inlining): Use ::get method.
4466 * ipa-inline.h (estimate_edge_growth): Likewise.
4468 2018-06-08 Martin Liska <mliska@suse.cz>
4470 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
4472 (hsa_function_summary::hsa_function_summary): Use the new enum
4474 (hsa_gpu_implementation_p): Use hsa_summaries::get.
4475 * hsa-gen.c (hsa_get_host_function): Likewise.
4476 (get_brig_function_name): Likewise.
4477 * ipa-hsa.c (process_hsa_functions): Likewise.
4478 (ipa_hsa_write_summary): Likewise.
4479 * symbol-summary.h (symtab_duplication): Use ::get function/
4480 (get): New function.
4482 2018-06-08 Martin Liska <mliska@suse.cz>
4484 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
4486 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
4487 (hsa_register_kernel): Likewise.
4488 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
4489 * hsa-gen.c (hsa_get_host_function): Likewise.
4490 (get_brig_function_name): Likewise.
4491 (generate_hsa): Likewise.
4492 (pass_gen_hsail::execute): Likewise.
4493 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4494 (devirtualization_time_bonus): Likewise.
4495 (ipcp_propagate_stage): Likewise.
4496 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
4497 (edge_set_predicate): Likewise.
4498 (evaluate_conditions_for_known_args): Likewise.
4499 (evaluate_properties_for_edge): Likewise.
4500 (ipa_fn_summary::reset): Likewise.
4501 (ipa_fn_summary_t::duplicate): Likewise.
4502 (dump_ipa_call_summary): Likewise.
4503 (ipa_dump_fn_summary): Likewise.
4504 (analyze_function_body): Likewise.
4505 (compute_fn_summary): Likewise.
4506 (estimate_edge_devirt_benefit): Likewise.
4507 (estimate_edge_size_and_time): Likewise.
4508 (estimate_calls_size_and_time): Likewise.
4509 (estimate_node_size_and_time): Likewise.
4510 (inline_update_callee_summaries): Likewise.
4511 (remap_edge_change_prob): Likewise.
4512 (remap_edge_summaries): Likewise.
4513 (ipa_merge_fn_summary_after_inlining): Likewise.
4514 (ipa_update_overall_fn_summary): Likewise.
4515 (read_ipa_call_summary): Likewise.
4516 (inline_read_section): Likewise.
4517 (write_ipa_call_summary): Likewise.
4518 (ipa_fn_summary_write): Likewise.
4519 (ipa_free_fn_summary): Likewise.
4520 * ipa-hsa.c (process_hsa_functions): Likewise.
4521 (ipa_hsa_write_summary): Likewise.
4522 (ipa_hsa_read_section): Likewise.
4523 * ipa-icf.c (sem_function::merge): Likewise.
4524 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
4525 (do_estimate_edge_time): Likewise.
4526 (estimate_size_after_inlining): Likewise.
4527 (estimate_growth): Likewise.
4528 (growth_likely_positive): Likewise.
4529 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
4530 (inline_call): Likewise.
4531 * ipa-inline.c (caller_growth_limits): Likewise.
4532 (can_inline_edge_p): Likewise.
4533 (can_inline_edge_by_limits_p): Likewise.
4534 (compute_uninlined_call_time): Likewise.
4535 (compute_inlined_call_time): Likewise.
4536 (want_inline_small_function_p): Likewise.
4537 (edge_badness): Likewise.
4538 (update_caller_keys): Likewise.
4539 (update_callee_keys): Likewise.
4540 (recursive_inlining): Likewise.
4541 (inline_small_functions): Likewise.
4542 (inline_to_all_callers_1): Likewise.
4543 (dump_overall_stats): Likewise.
4544 (early_inline_small_functions): Likewise.
4545 (early_inliner): Likewise.
4546 * ipa-inline.h (estimate_edge_growth): Likewise.
4547 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4548 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4549 * ipa-prop.h (IPA_NODE_REF): Likewise.
4550 (IPA_EDGE_REF): Likewise.
4551 * ipa-pure-const.c (malloc_candidate_p): Likewise.
4552 (propagate_malloc): Likewise.
4553 * ipa-split.c (execute_split_functions): Likewise.
4554 * symbol-summary.h: Rename get to get_create.
4556 (get_create): Likewise.
4557 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
4559 2018-06-08 Martin Liska <mliska@suse.cz>
4561 * symbol-summary.h (release): Move definition out of class
4563 (symtab_removal): Likewise.
4564 (symtab_duplication): Likewise.
4566 2018-06-08 Martin Liska <mliska@suse.cz>
4568 * symbol-summary.h (function_summary): Move constructor
4569 implementation out of class declaration.
4570 (release): Likewise.
4571 (symtab_insertion): Likewise.
4572 (symtab_removal): Likewise.
4573 (symtab_duplication): Likewise.
4576 2018-06-08 Martin Liska <mliska@suse.cz>
4578 * Makefile.in: Remove support for MPX (macros, related functions,
4579 fields in cgraph_node, ...).
4580 * builtin-types.def (BT_BND): Likewise.
4581 (BT_FN_BND_CONST_PTR): Likewise.
4582 (BT_FN_CONST_PTR_BND): Likewise.
4583 (BT_FN_VOID_PTR_BND): Likewise.
4584 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
4585 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
4586 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
4587 (expand_builtin_mempcpy_with_bounds): Likewise.
4588 (expand_builtin_memset_with_bounds): Likewise.
4589 (expand_builtin_memset_args): Likewise.
4590 (std_expand_builtin_va_start): Likewise.
4591 (expand_builtin): Likewise.
4592 (expand_builtin_with_bounds): Likewise.
4593 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
4594 (DEF_LIB_BUILTIN_CHKP): Likewise.
4595 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
4596 (DEF_CHKP_BUILTIN): Likewise.
4597 (BUILT_IN_MEMCPY): Likewise.
4598 (BUILT_IN_MEMMOVE): Likewise.
4599 (BUILT_IN_MEMPCPY): Likewise.
4600 (BUILT_IN_MEMSET): Likewise.
4601 (BUILT_IN_STPCPY): Likewise.
4602 (BUILT_IN_STRCAT): Likewise.
4603 (BUILT_IN_STRCHR): Likewise.
4604 (BUILT_IN_STRCPY): Likewise.
4605 (BUILT_IN_STRLEN): Likewise.
4606 (BUILT_IN_MEMCPY_CHK): Likewise.
4607 (BUILT_IN_MEMMOVE_CHK): Likewise.
4608 (BUILT_IN_MEMPCPY_CHK): Likewise.
4609 (BUILT_IN_MEMSET_CHK): Likewise.
4610 (BUILT_IN_STPCPY_CHK): Likewise.
4611 (BUILT_IN_STRCAT_CHK): Likewise.
4612 (BUILT_IN_STRCPY_CHK): Likewise.
4613 * calls.c (store_bounds): Likewise.
4614 (emit_call_1): Likewise.
4615 (special_function_p): Likewise.
4616 (maybe_warn_nonstring_arg): Likewise.
4617 (initialize_argument_information): Likewise.
4618 (finalize_must_preallocate): Likewise.
4619 (compute_argument_addresses): Likewise.
4620 (expand_call): Likewise.
4621 * cfgexpand.c (expand_call_stmt): Likewise.
4622 (expand_return): Likewise.
4623 (expand_gimple_stmt_1): Likewise.
4624 (pass_expand::execute): Likewise.
4625 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
4626 (cgraph_node::remove): Likewise.
4627 (cgraph_node::dump): Likewise.
4628 (cgraph_node::verify_node): Likewise.
4629 * cgraph.h (chkp_function_instrumented_p): Likewise.
4630 (symtab_node::get_alias_target): Likewise.
4631 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
4632 (cgraph_local_p): Likewise.
4633 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
4634 (cgraph_edge::rebuild_references): Likewise.
4635 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
4636 (walk_polymorphic_call_targets): Likewise.
4637 (cgraph_node::expand_thunk): Likewise.
4638 (symbol_table::output_weakrefs): Likewise.
4639 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
4640 (ix86_handle_option): Likewise.
4641 * config/i386/constraints.md: Likewise.
4642 * config/i386/i386-builtin-types.def (BND): Likewise.
4646 * config/i386/i386-builtin.def (BDESC_END): Likewise.
4647 (BDESC_FIRST): Likewise.
4649 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
4650 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
4651 * config/i386/i386.c (enum reg_class): Likewise.
4652 (ix86_target_string): Likewise.
4653 (ix86_option_override_internal): Likewise.
4654 (ix86_conditional_register_usage): Likewise.
4655 (ix86_valid_target_attribute_inner_p): Likewise.
4656 (ix86_set_indirect_branch_type): Likewise.
4657 (ix86_set_current_function): Likewise.
4658 (ix86_function_arg_regno_p): Likewise.
4659 (init_cumulative_args): Likewise.
4660 (ix86_function_arg_advance): Likewise.
4661 (ix86_function_arg): Likewise.
4662 (ix86_pass_by_reference): Likewise.
4663 (ix86_function_value_regno_p): Likewise.
4664 (ix86_function_value_1): Likewise.
4665 (ix86_function_value_bounds): Likewise.
4666 (ix86_return_in_memory): Likewise.
4667 (ix86_setup_incoming_vararg_bounds): Likewise.
4668 (ix86_va_start): Likewise.
4669 (indirect_thunk_need_prefix): Likewise.
4670 (print_reg): Likewise.
4671 (ix86_print_operand): Likewise.
4672 (ix86_expand_call): Likewise.
4673 (ix86_output_function_return): Likewise.
4674 (reg_encoded_number): Likewise.
4675 (BDESC_VERIFYS): Likewise.
4676 (ix86_init_mpx_builtins): Likewise.
4677 (ix86_init_builtins): Likewise.
4678 (ix86_emit_cmove): Likewise.
4679 (ix86_emit_move_max): Likewise.
4680 (ix86_expand_builtin): Likewise.
4681 (ix86_builtin_mpx_function): Likewise.
4682 (ix86_get_arg_address_for_bt): Likewise.
4683 (ix86_load_bounds): Likewise.
4684 (ix86_store_bounds): Likewise.
4685 (ix86_load_returned_bounds): Likewise.
4686 (ix86_store_returned_bounds): Likewise.
4687 (ix86_class_likely_spilled_p): Likewise.
4688 (ix86_hard_regno_mode_ok): Likewise.
4689 (x86_order_regs_for_local_alloc): Likewise.
4690 (ix86_mitigate_rop): Likewise.
4691 (ix86_bnd_prefixed_insn_p): Likewise.
4692 (ix86_mpx_bound_mode): Likewise.
4693 (ix86_make_bounds_constant): Likewise.
4694 (ix86_initialize_bounds): Likewise.
4695 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
4696 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
4697 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
4698 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
4699 (TARGET_CHKP_BOUND_MODE): Likewise.
4700 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
4701 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
4702 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
4703 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
4704 * config/i386/i386.h (TARGET_MPX): Likewise.
4705 (TARGET_MPX_P): Likewise.
4706 (VALID_BND_REG_MODE): Likewise.
4707 (FIRST_BND_REG): Likewise.
4708 (LAST_BND_REG): Likewise.
4709 (enum reg_class): Likewise.
4710 (BND_REG_P): Likewise.
4711 (BND_REGNO_P): Likewise.
4712 (BNDmode): Likewise.
4713 (ADJUST_INSN_LENGTH): Likewise.
4714 * config/i386/i386.md: Likewise.
4715 * config/i386/i386.opt: Likewise.
4716 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
4717 (defined): Likewise.
4718 (LINK_MPX): Likewise.
4719 (MPX_SPEC): Likewise.
4720 (LIBMPX_SPEC): Likewise.
4721 (LIBMPXWRAPPERS_SPEC): Likewise.
4722 (CHKP_SPEC): Likewise.
4723 * config/i386/predicates.md: Likewise.
4724 * dbxout.c (dbxout_type): Likewise.
4725 * doc/extend.texi: Likewise.
4726 * doc/invoke.texi: Likewise.
4727 * doc/md.texi: Likewise.
4728 * doc/tm.texi: Likewise.
4729 * doc/tm.texi.in: Likewise.
4730 * dwarf2out.c (is_base_type): Likewise.
4731 (gen_formal_types_die): Likewise.
4732 (gen_subprogram_die): Likewise.
4733 (gen_type_die_with_usage): Likewise.
4734 (gen_decl_die): Likewise.
4735 (dwarf2out_late_global_decl): Likewise.
4736 * expr.c (expand_assignment): Likewise.
4737 (emit_storent_insn): Likewise.
4738 (store_expr_with_bounds): Likewise.
4739 (store_expr): Likewise.
4740 (expand_expr_real_1): Likewise.
4741 * expr.h (store_expr_with_bounds): Likewise.
4742 * function.c (use_register_for_decl): Likewise.
4743 (struct bounds_parm_data): Likewise.
4744 (assign_parms_augmented_arg_list): Likewise.
4745 (assign_parm_find_entry_rtl): Likewise.
4746 (assign_parm_is_stack_parm): Likewise.
4747 (assign_parm_load_bounds): Likewise.
4748 (assign_bounds): Likewise.
4749 (assign_parms): Likewise.
4750 (expand_function_start): Likewise.
4751 * gcc.c (CHKP_SPEC): Likewise.
4752 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
4753 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
4754 (wrestrict_dom_walker::check_call): Likewise.
4755 * gimple.c (gimple_build_call_from_tree): Likewise.
4756 * gimple.h (enum gf_mask): Likewise.
4757 (gimple_call_with_bounds_p): Likewise.
4758 (gimple_call_set_with_bounds): Likewise.
4759 * gimplify.c (gimplify_init_constructor): Likewise.
4760 * ipa-cp.c (initialize_node_lattices): Likewise.
4761 (propagate_constants_across_call): Likewise.
4762 (find_more_scalar_values_for_callers_subset): Likewise.
4763 * ipa-hsa.c (process_hsa_functions): Likewise.
4764 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
4765 * ipa-icf.c (sem_function::merge): Likewise.
4766 * ipa-inline.c (early_inliner): Likewise.
4767 * ipa-pure-const.c (warn_function_noreturn): Likewise.
4768 (warn_function_cold): Likewise.
4769 (propagate_pure_const): Likewise.
4770 * ipa-ref.h (enum GTY): Likewise.
4771 * ipa-split.c (find_retbnd): Likewise.
4772 (consider_split): Likewise.
4773 (split_function): Likewise.
4774 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
4775 * ipa.c (walk_polymorphic_call_targets): Likewise.
4776 (symbol_table::remove_unreachable_nodes): Likewise.
4777 (process_references): Likewise.
4778 (cgraph_build_static_cdtor_1): Likewise.
4779 * lto-cgraph.c (lto_output_node): Likewise.
4780 (output_refs): Likewise.
4781 (compute_ltrans_boundary): Likewise.
4782 (input_overwrite_node): Likewise.
4783 (input_node): Likewise.
4784 (input_cgraph_1): Likewise.
4785 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
4786 * passes.c (pass_manager::execute_early_local_passes): Likewise.
4787 (class pass_chkp_instrumentation_passes): Likewise.
4788 (make_pass_chkp_instrumentation_passes): Likewise.
4789 * passes.def: Likewise.
4790 * rtl.h (struct GTY): Likewise.
4791 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
4792 * stor-layout.c (layout_type): Likewise.
4793 * symtab.c: Likewise.
4794 * target.def: Likewise.
4795 * targhooks.c (default_chkp_bound_type): Likewise.
4796 (default_chkp_bound_mode): Likewise.
4797 (default_builtin_chkp_function): Likewise.
4798 (default_chkp_function_value_bounds): Likewise.
4799 (default_chkp_make_bounds_constant): Likewise.
4800 (default_chkp_initialize_bounds): Likewise.
4801 * targhooks.h (default_chkp_bound_type): Likewise.
4802 (default_chkp_bound_mode): Likewise.
4803 (default_builtin_chkp_function): Likewise.
4804 (default_chkp_function_value_bounds): Likewise.
4805 (default_chkp_make_bounds_constant): Likewise.
4806 (default_chkp_initialize_bounds): Likewise.
4807 * toplev.c (compile_file): Likewise.
4808 (process_options): Likewise.
4809 * tree-core.h (DEF_BUILTIN): Likewise.
4810 (DEF_BUILTIN_CHKP): Likewise.
4811 * tree-inline.c (declare_return_variable): Likewise.
4812 (remap_gimple_stmt): Likewise.
4813 (copy_bb): Likewise.
4814 (initialize_inlined_parameters): Likewise.
4815 (expand_call_inline): Likewise.
4816 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
4817 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
4818 (make_pass_ipa_chkp_produce_thunks): Likewise.
4819 (make_pass_chkp): Likewise.
4820 (make_pass_chkp_opt): Likewise.
4821 (make_pass_chkp_instrumentation_passes): Likewise.
4822 * tree-pretty-print.c (dump_generic_node): Likewise.
4823 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
4824 * tree-ssa-dce.c (propagate_necessity): Likewise.
4825 (eliminate_unnecessary_stmts): Likewise.
4826 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
4827 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
4828 * tree-ssa-sccvn.h: Likewise.
4829 * tree-ssa-strlen.c (get_string_length): Likewise.
4830 (valid_builtin_call): Likewise.
4831 (adjust_last_stmt): Likewise.
4832 (handle_builtin_strchr): Likewise.
4833 (handle_builtin_strcpy): Likewise.
4834 (handle_builtin_stxncpy): Likewise.
4835 (handle_builtin_memcpy): Likewise.
4836 (handle_builtin_strcat): Likewise.
4837 (strlen_check_and_optimize_stmt): Likewise.
4838 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
4839 * tree-streamer-in.c: Likewise.
4840 * tree-streamer.c (record_common_node): Likewise.
4841 * tree.c (tree_code_size): Likewise.
4842 (wide_int_to_tree_1): Likewise.
4843 (type_contains_placeholder_1): Likewise.
4844 (build_common_tree_nodes): Likewise.
4845 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
4846 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
4847 (POINTER_BOUNDS_P): Likewise.
4848 (BOUNDED_TYPE_P): Likewise.
4849 (BOUNDED_P): Likewise.
4850 (CALL_WITH_BOUNDS_P): Likewise.
4851 (pointer_bounds_type_node): Likewise.
4852 * value-prof.c (gimple_ic): Likewise.
4853 * var-tracking.c (vt_add_function_parameters): Likewise.
4854 * varasm.c (make_decl_rtl): Likewise.
4855 (assemble_start_function): Likewise.
4856 (output_constant): Likewise.
4857 (maybe_assemble_visibility): Likewise.
4858 * varpool.c (ctor_for_folding): Likewise.
4859 * chkp-builtins.def: Remove.
4860 * ipa-chkp.c: Remove.
4861 * ipa-chkp.h: Remove.
4862 * rtl-chkp.c: Remove.
4863 * rtl-chkp.h: Remove.
4864 * tree-chkp-opt.c: Remove.
4865 * tree-chkp.c: Remove.
4866 * tree-chkp.h: Remove.
4868 2018-06-07 Carl Love <cel@us.ibm.com>
4870 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
4871 vextract_fp_from_shortl): Add BE support.
4873 2018-06-07 Paul Koning <ni1d@arrl.net>
4875 * compare-elim.c (try_merge_compare): Don't merge compare if
4876 address contains a side effect.
4877 (try_eliminate_compare): Likewise.
4879 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
4881 * config.gcc: Support "tremont".
4882 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
4883 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4885 * config/i386/i386.c (m_TREMONT): Define.
4886 (processor_target_table): Add "tremont".
4887 (PTA_TREMONT): Define.
4888 (ix86_lea_outperforms): Add TARGET_TREMONT.
4889 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
4890 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
4891 and M_INTEL_GOLDMONT_PLUS.
4892 (fold_builtin_cpu): Add "tremont".
4893 (ix86_add_stmt_cost): Add TARGET_TREMONT.
4894 (ix86_option_override_internal): Add "tremont".
4895 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
4896 (processor_type): Add PROCESSOR_TREMONT.
4897 * config/i386/x86-tune.def: Add m_TREMONT.
4898 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
4900 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4902 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
4903 symbol defined for msp430i* devices to be lower case.
4905 gcc/testsuite/gcc.target/msp430/
4906 * msp430i-device-symbol.c: New test.
4907 * msp430f-device-symbol.c: New test.
4908 * msp430.h: New test header file.
4910 2018-06-07 Richard Biener <rguenther@suse.de>
4912 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
4913 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
4914 Properly wrap signed arithmetic if overflow wraps.
4916 2018-06-07 Jakub Jelinek <jakub@redhat.com>
4918 PR tree-optimization/69615
4919 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
4920 of a cast from a same precision integral SSA_NAME in a bb dominated
4921 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
4922 cast to utype if rhs2 has already a compatible type.
4924 2018-06-07 Richard Biener <rguenther@suse.de>
4926 PR tree-optimization/85935
4927 * graphite-scop-detection.c (find_params_in_bb): Analyze
4928 condition operands with respect to the correct loop. Assert
4929 the analysis doesn't fail.
4931 2018-06-04 Carl Love <cel@us.ibm.com>
4933 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
4934 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
4935 as it is slightly cheaper.
4936 (first_match_or_eos_index_<mode>):
4937 Calculate index using natural element order.
4938 (first_match_index_<mode>):
4939 Calculate index using natural element order.
4940 (first_match_or_eos_index_<mode>):
4941 Calculate index using natural order.
4942 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
4943 for BE and LE modes.
4944 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
4945 P9V_BUILTIN_VCLZLSBB_V16QI.
4946 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
4949 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4951 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
4952 indentation and line wrap for many prototypes. Add missing
4953 @smallexample directives around block of prototypes for vec_xl and
4956 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
4958 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
4959 track if we pass or return IEEE 128-bit floating point.
4960 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
4961 C++ mangling that is compatible with GCC 8.1.
4962 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
4963 (init_cumulative_args): Note if we pass or return IEEE 128-bit
4964 floating point types.
4965 (rs6000_function_arg_advance_1): Likewise.
4966 (rs6000_mangle_type): Optionally generate mangled names that match
4967 what GCC 8.1 generated for IEEE 128-bit floating point types.
4968 (rs6000_globalize_decl_name): If we have an external function that
4969 passes or returns IEEE 128-bit types, generate a weak reference
4970 from the mangled name used in GCC 8.1 to the current mangled
4972 (rs6000_init_builtins): Make __ibm128 use the long double type if
4973 long double is IBM extended double. Make __float128 use the long
4974 double type if long double is IEEE 128-bit.
4977 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
4978 macro for __ibm128 built-in functions.
4979 (PACK_IF): Add __ibm128 pack/unpack functions.
4980 (UNPACK_IF): Likewise.
4981 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
4982 enable long double built-in functions if long double is IEEE
4983 128-bit floating point.
4984 (rs6000_invalid_builtin): Update long double built-in function
4986 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
4987 functions, adjust the built-in function to use the long double
4988 built-in function if __ibm128 and long double are the same type.
4989 * doc/extend.texi (PowerPC builtins): Update documention for
4990 __builtin_{,un}pack_longdouble. Add documentation for
4991 __builtin_{,un}pack_ibm128.
4993 2018-06-06 Jim Wilson <jimw@sifive.com>
4995 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
4996 (struct machine_function): New field interrupt_mode.
4997 (riscv_handle_type_attribute): New function. Add forward declaration.
4998 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
4999 (riscv_expand_epilogue): Check interrupt_mode field.
5000 (riscv_set_current_function): Check interrupt attribute args and
5001 set interrupt_mode field.
5002 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
5003 (riscv_sret, riscv_uret): New.
5004 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
5005 new arguments to interrupt attribute.
5007 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
5010 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
5011 Don't handle -mcpu=power8 if -mpower9-vector is also used.
5013 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
5015 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
5016 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
5017 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5018 several redundant entries.
5020 2018-06-06 David Malcolm <dmalcolm@redhat.com>
5022 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
5023 type from "rtx" to "rtx_insn *".
5024 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
5025 for local "call_insn", removing cast.
5026 (ix86_expand_call): Likewise, introducing a "call_insn" local.
5028 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
5030 PR tree-optimization/86066
5031 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
5032 for BIT_INSERT_EXPR stores.
5034 2018-06-06 Richard Biener <rguenther@suse.de>
5036 PR tree-optimization/86062
5037 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5038 component refs ontop
5039 of to be offsetted base.
5041 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5043 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
5044 to be static and remove check on interrupt attribute name.
5046 gcc/testsuite/gcc.target/msp430/
5047 * function-attributes-4.c: New test.
5048 * static-interrupts.c: New test.
5050 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
5052 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
5053 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
5055 2018-06-05 Steve Ellcey <sellcey@cavium.com>
5058 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
5060 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
5061 Remove second argument, change how error is called.
5062 (aarch64_layout_arg): Remove second argument from
5063 aarch64_err_no_fpadvsimd call.
5064 (aarch64_init_cumulative_args): Ditto.
5065 (aarch64_gimplify_va_arg_expr): Ditto.
5066 * config/aarch64/aarch64.md (mov<mode>): Ditto.
5068 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
5070 * config/i386/i386.md (simple_return_indirect_internal): New expander.
5071 (*simple_return_indirect_internal<mode>): Rename from
5072 simple_return_indirect_internal. Use W mode iterator.
5073 (rstorssp): New expander.
5074 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
5075 (clrssbsy): New expander.
5076 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
5078 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
5080 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
5082 (cmse_check_pointed_object): Likewise.
5084 2018-06-05 Martin Liska <mliska@suse.cz>
5086 PR gcov-profile/47618
5087 * doc/invoke.texi: Document how -fprofile-dir format
5090 2018-06-05 Richard Biener <rguenther@suse.de>
5092 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
5093 removal pretend DOM info isn't available so we do not update
5094 it and only remove edges, not dominated blocks. Actually free
5095 DOM info in case we removed something. Remove unreachable blocks.
5096 (mfb_keep_latches): Work with either DOM info or marked backedges.
5097 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
5098 first. Mark backedges if DOM info isn't available.
5099 (Re-)compute DOM info after cleanup_control_flow_pre.
5101 2018-06-05 Richard Biener <rguenther@suse.de>
5103 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
5104 (locus_discrim_hasher::hash): Adjust.
5105 (locus_discrim_hasher::equal): Likewise.
5106 (next_discriminator_for_locus): Work on line directly.
5107 (same_line_p): Pass in expanded locus1 as well.
5108 (assign_discriminators): Avoid redundant location expansions.
5110 2018-06-05 Richard Biener <rguenther@suse.de>
5112 PR tree-optimization/86046
5113 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
5114 if required after clearing TREE_ADDRESSABLE.
5116 2018-06-05 Richard Biener <rguenther@suse.de>
5118 PR tree-optimization/86047
5119 * tree-ssa-loop.c (for_each_index): Glob handling of all
5120 decls and constants and really handle all of them.
5122 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5125 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
5126 qualifier_void_pointer and qualifier_const_void_pointer.
5127 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
5128 (arm_init_builtins): Handle the above.
5129 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
5130 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
5131 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
5134 2018-06-05 Martin Liska <mliska@suse.cz>
5136 * auto-profile.c (read_autofdo_file): Do not use
5137 gcov_ctr_summary struct.
5138 (afdo_callsite_hot_enough_for_early_inline): Likewise.
5139 * coverage.c (struct counts_entry): Likewise.
5140 (read_counts_file): Read just single summary entry.
5141 (get_coverage_counts): Use gcov_summary struct.
5142 * coverage.h (get_coverage_counts): Likewise.
5143 * gcov-dump.c (dump_working_sets): Likewise.
5144 (tag_summary): Dump just single summary.
5145 * gcov-io.c (gcov_write_summary): Write just histogram
5147 (gcov_read_summary): Read just single summary.
5148 (compute_working_sets): Use gcov_summary struct.
5149 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
5150 of GCOV_COUNTERS_SUMMABLE.
5151 (GCOV_COUNTERS_SUMMABLE): Remove.
5152 (GCOV_FIRST_VALUE_COUNTER): Replace with
5153 GCOV_COUNTER_V_INTERVAL.
5154 (struct gcov_ctr_summary): Remove.
5155 (struct gcov_summary): Directly use fields of former
5157 (compute_working_sets): Use gcov_summary struct.
5158 * gcov.c (read_count_file): Do not use ctrs fields.
5159 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
5161 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
5163 * profile.c: Likewise.
5164 * profile.h: Likewise.
5166 2018-06-05 Martin Liska <mliska@suse.cz>
5168 PR gcov-profile/84846
5169 * gcov.c (output_lines): Print working directory only
5170 in intermediate format.
5172 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
5174 * config/s390/s390-builtin-types.def: Add void function type.
5175 * config/s390/s390-builtins.def: Use the function type for the
5178 2018-06-04 Jim Wilson <jimw@sifive.com>
5180 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
5182 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
5183 and maybe_eh_return. Change regno to unsigned int. Use new args to
5184 handle EH_RETURN_DATA_REGNO registers properly.
5185 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
5186 (riscv_expand_epilogue): Update comment. Change argument name and
5187 type. Update code to use new name and type. Pass new args to
5188 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
5190 * config/riscv/riscv.md (NORMAL_RETURN): New.
5191 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
5192 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
5193 (eh_return): Call gen_eh_return_internal and emit barrier.
5194 (eh_return_internal): Call riscv_expand_epilogue.
5196 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
5198 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
5199 bit_insertion field and declare can_be_merged_into method.
5200 (merged_store_group::can_be_merged_into): New method.
5201 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
5202 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
5203 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
5205 2018-06-04 Richard Biener <rguenther@suse.de>
5207 PR tree-optimization/85955
5208 * builtins.c (fold_builtin_sincos): Convert pointers to
5209 destination to appropriate type before dereferencing.
5211 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
5213 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
5215 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
5217 * expr.c (expand_expr_real_1): Force the operand into memory if
5218 its TYPE_MODE is BLKmode and if there is no integer mode for
5219 the number of bits being extracted.
5221 2018-06-04 Jakub Jelinek <jakub@redhat.com>
5225 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5226 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
5228 2018-06-04 Richard Biener <rguenther@suse.de>
5230 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
5231 (cleanup_tree_cfg_noloop): ... single caller. Do
5232 start_recording_case_labels later.
5234 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
5236 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
5237 to _IMMINTRIN_H_INCLUDED.
5238 * config/i386/pconfigintrin.h: Ditto.
5239 * config/i386/waitpkgintrin.h: Ditto.
5240 * config/i386/immintrin.h: Add includes for sgxintrin.h,
5241 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
5242 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
5243 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
5244 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
5245 waitpkgintrin.h and cldemoteintrin.h.
5247 2018-06-04 Richard Biener <rguenther@suse.de>
5249 PR tree-optimization/86038
5250 * tracer.c (find_best_successor): Check probability for
5251 being initialized, bail out if not.
5253 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
5256 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
5257 of bits to ignore when comparing architectures.
5259 2018-06-04 Jakub Jelinek <jakub@redhat.com>
5261 PR tree-optimization/69615
5262 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
5263 maximum or minimum of the type, try to merge it also as if
5264 range1 is + [-, x - 1] or + [x + 1, -].
5267 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
5269 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
5271 PR tree-optimization/86034
5272 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
5273 the unsigned bitfield type in a bit insertion sequence if it does not
5274 have a larger precision than the bitfield size.
5275 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
5277 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
5279 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
5281 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
5283 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
5284 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
5285 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
5286 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
5288 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
5290 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
5291 Disable -fdelete-null-pointer-checks for ELF toolchain.
5293 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
5294 Kito Cheng <kito.cheng@gmail.com>
5296 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
5297 (nds32le-*-*, nds32be-*-*): Integrate checking process.
5298 (nds32*-*-*): Add glibc and uclibc conditions.
5299 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
5300 (TARGET_EXCEPT_UNWIND_INFO): Define.
5301 * config/nds32/elf.h: New file.
5302 * config/nds32/linux.h: New file.
5303 * config/nds32/nds32-elf.opt: New file.
5304 * config/nds32/nds32-linux.opt: New file.
5305 * config/nds32/nds32-fp-as-gp.c
5306 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
5307 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
5309 (nds32_asm_file_end): Ditto.
5310 (nds32_print_operand): Ditto.
5311 (nds32_insert_attributes): Ditto.
5312 (nds32_init_libfuncs): New function.
5313 (TARGET_HAVE_TLS): Define.
5314 (TARGET_INIT_LIBFUNCS): Define.
5315 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
5317 (TARGET_ELF): Apply different mcmodel setting.
5318 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
5319 been migrated into elf.h and linux.h files.
5320 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
5321 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
5322 (mcmodel): The content has been migrated into nds32-elf.opt and
5323 nds32-linux.opt files.
5324 * config/nds32/t-elf: New file.
5325 * config/nds32/t-linux: New file.
5327 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
5328 Shiva Chen <shiva0217@gmail.com>
5330 * config/nds32/constants.md (unspec_volatile_element): Add
5331 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
5332 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
5334 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
5335 (make_pass_nds32_fp_as_gp): Declare.
5336 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
5338 (nds32_asm_function_end_prologue): Remove unused asm output.
5339 (nds32_asm_function_begin_epilogue): Remove unused asm output.
5340 (nds32_asm_file_start): Output necessary fp_as_gp information.
5341 (nds32_option_override): Adjust register usage.
5342 (nds32_expand_prologue): Consider fp_as_gp situation.
5343 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
5344 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
5347 (simple_return): Ditto.
5348 (omit_fp_begin): Output special directive for fp_as_gp.
5349 (omit_fp_end): Output special directive for fp_as_gp.
5350 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
5351 mforbid-fp-as-gp): New options.
5353 2018-06-01 Mark Wielaard <mark@klomp.org>
5355 * dwarf2out.c (dwarf2out_finish): Remove generation of
5356 DW_AT_loclists_base.
5358 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
5360 * gimple-ssa-store-merging.c: Include gimple-fold.h.
5361 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
5362 (struct merged_store_group): Add bit_insertion field.
5363 (dump_char_array): Use standard hexadecimal format.
5364 (merged_store_group::merged_store_group): Set bit_insertion to false.
5365 (merged_store_group::apply_stores): Use optimal buffer size. Deal
5366 with BIT_INSERT_EXPR stores. Move up code updating the mask and
5367 also print the mask in the dump file.
5368 (pass_store_merging::gate): Minor tweak.
5369 (imm_store_chain_info::coalesce_immediate): Fix wrong association
5370 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
5371 stores with INTEGER_CST stores.
5372 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
5373 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
5374 and use it throughout. Generate bit insertion sequences if need be.
5375 (pass_store_merging::process_store): Remove redundant condition.
5376 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
5378 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
5380 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
5381 the 128-bit floating point types. Fix function comment.
5383 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5385 * config/aarch64/aarch64-simd.md
5386 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
5388 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
5391 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
5393 PR tree-optimization/85989
5394 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
5396 (backprop::intersect_uses): Check it when deciding whether this
5397 is a backedge reference.
5398 (backprop::process_block): Add each phi to m_visited_phis
5399 after visiting it, then clear it at the end.
5401 2018-06-01 Richard Biener <rguenther@suse.de>
5403 * tree-vectorizer.h (vect_dr_stmt): New function.
5404 (vect_get_load_cost): Adjust.
5405 (vect_get_store_cost): Likewise.
5406 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
5407 Use vect_dr_stmt instead of DR_SMTT.
5408 (vect_record_base_alignments): Likewise.
5409 (vect_calculate_target_alignment): Likewise.
5410 (vect_compute_data_ref_alignment): Likewise and make static.
5411 (vect_update_misalignment_for_peel): Likewise.
5412 (vect_verify_datarefs_alignment): Likewise.
5413 (vector_alignment_reachable_p): Likewise.
5414 (vect_get_data_access_cost): Likewise. Pass down
5415 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
5416 (vect_get_peeling_costs_all_drs): Likewise.
5417 (vect_peeling_hash_get_lowest_cost): Likewise.
5418 (vect_enhance_data_refs_alignment): Likewise.
5419 (vect_find_same_alignment_drs): Likewise.
5420 (vect_analyze_data_refs_alignment): Likewise.
5421 (vect_analyze_group_access_1): Likewise.
5422 (vect_analyze_group_access): Likewise.
5423 (vect_analyze_data_ref_access): Likewise.
5424 (vect_analyze_data_ref_accesses): Likewise.
5425 (vect_vfa_segment_size): Likewise.
5426 (vect_small_gap_p): Likewise.
5427 (vectorizable_with_step_bound_p): Likewise.
5428 (vect_prune_runtime_alias_test_list): Likewise.
5429 (vect_analyze_data_refs): Likewise.
5430 (vect_supportable_dr_alignment): Likewise.
5431 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
5432 (vect_gen_prolog_loop_niters): Likewise.
5433 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5434 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
5436 (vect_recog_mask_conversion_pattern): Likewise.
5437 (vect_try_gather_scatter_pattern): Likewise.
5438 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
5439 to vect_get_store_cost.
5440 (vect_get_store_cost): Get stmt_info instead of DR.
5441 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
5442 (vect_get_load_cost): Get stmt_info instead of DR.
5444 2018-06-01 Richard Biener <rguenther@suse.de>
5447 * gimple-fold.c (var_decl_component_p): Also allow offsetted
5448 vars wrapped in MEM_REFs.
5450 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
5452 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
5453 Fix subreg tests so that we only return a choice between
5454 GENERAL_REGS and FP_REGS if the original classes included both.
5456 2018-06-01 Richard Biener <rguenther@suse.de>
5459 * tree-ssa-structalias.c (get_function_part_constraint):
5460 Handle NULL fi->decl.
5461 (find_func_aliases_for_call): Properly handle indirect
5462 fi from direct call.
5463 (find_func_clobbers): Likewise.
5464 (ipa_pta_execute): Likewise.
5465 (create_variable_info_for): For functions that are ifunc_resolver
5466 resolve to a varinfo that contains the result of the resolver call.
5467 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
5470 2018-05-31 Michael Collison <michael.collison@arm.com>
5472 * config/aarch64/aarch64.md:
5473 (*fix_to_zero_extenddfdi2): New pattern.
5474 * gcc.target/aarch64/fix_extend1.c: New testcase.
5476 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
5480 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
5481 and BUILT_IN_STRNCMP_EQ.
5482 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
5483 BUILT_IN_STRNCMP_EQ.
5484 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
5485 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
5486 (gimple_fold_builtin): Likewise.
5487 * tree-ssa-strlen.c (compute_string_length): New function.
5488 (determine_min_obsize): New function.
5489 (handle_builtin_string_cmp): New function to handle calls to
5490 string compare functions.
5491 (strlen_optimize_stmt): Add handling to builtin string compare
5493 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
5494 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
5495 * tree.c (build_common_builtin_nodes): Add new defines of
5496 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
5498 2018-05-31 Jakub Jelinek <jakub@redhat.com>
5501 * bb-reorder.c (pass_partition_blocks::gate): Return false for
5502 functions with naked attribute.
5504 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
5506 * config/i386/sse.md (avx_vec_concat<mode>):
5507 Substitute concat_tg_mode mode attribute with xtg_mode.
5508 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
5509 (concat_tg_mode): Remove mode attribute.
5511 2018-05-31 Martin Sebor <msebor@redhat.com>
5514 * calls.c (alloc_max_size): Correct a logic error/typo.
5515 Treat excessive arguments as infinite. Warn for invalid arguments.
5516 * doc/invoke.texi (-Walloc-size-larger-than): Update.
5518 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
5521 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
5522 and movx for Haswell.
5524 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
5525 Cesar Philippidis <cesar@codesourcery.com>
5528 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
5529 when emitting error on private/firstprivate reductions.
5530 * omp-low.c (lower_omp_target): Avoid reference-type processing
5531 on pointers for firstprivate clause.
5533 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
5535 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
5538 * config/aarch64/aarch64-simd.md
5539 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
5540 (aarch64_ld1_x3_<mode>): Likewise
5541 (aarch64_st1x2<VALLDIF:mode>): Likewise
5542 (aarch64_st1_x2_<mode>): Likewise
5543 (aarch64_st1x3<VALLDIF:mode>): Likewise
5544 (aarch64_st1_x3_<mode>): Likewise
5545 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
5546 (vld1_s8_x3): Likewise.
5547 (vld1_u16_x3): Likewise.
5548 (vld1_s16_x3): Likewise.
5549 (vld1_u32_x3): Likewise.
5550 (vld1_s32_x3): Likewise.
5551 (vld1_u64_x3): Likewise.
5552 (vld1_s64_x3): Likewise.
5553 (vld1_f16_x3): Likewise.
5554 (vld1_f32_x3): Likewise.
5555 (vld1_f64_x3): Likewise.
5556 (vld1_p8_x3): Likewise.
5557 (vld1_p16_x3): Likewise.
5558 (vld1_p64_x3): Likewise.
5559 (vld1q_u8_x3): Likewise.
5560 (vld1q_s8_x3): Likewise.
5561 (vld1q_u16_x3): Likewise.
5562 (vld1q_s16_x3): Likewise.
5563 (vld1q_u32_x3): Likewise.
5564 (vld1q_s32_x3): Likewise.
5565 (vld1q_u64_x3): Likewise.
5566 (vld1q_s64_x3): Likewise.
5567 (vld1q_f16_x3): Likewise.
5568 (vld1q_f32_x3): Likewise.
5569 (vld1q_f64_x3): Likewise.
5570 (vld1q_p8_x3): Likewise.
5571 (vld1q_p16_x3): Likewise.
5572 (vld1q_p64_x3): Likewise.
5573 (vst1_s64_x2): Likewise.
5574 (vst1_u64_x2): Likewise.
5575 (vst1_f64_x2): Likewise.
5576 (vst1_s8_x2): Likewise.
5577 (vst1_p8_x2): Likewise.
5578 (vst1_s16_x2): Likewise.
5579 (vst1_p16_x2): Likewise.
5580 (vst1_s32_x2): Likewise.
5581 (vst1_u8_x2): Likewise.
5582 (vst1_u16_x2): Likewise.
5583 (vst1_u32_x2): Likewise.
5584 (vst1_f16_x2): Likewise.
5585 (vst1_f32_x2): Likewise.
5586 (vst1_p64_x2): Likewise.
5587 (vst1q_s8_x2): Likewise.
5588 (vst1q_p8_x2): Likewise.
5589 (vst1q_s16_x2): Likewise.
5590 (vst1q_p16_x2): Likewise.
5591 (vst1q_s32_x2): Likewise.
5592 (vst1q_s64_x2): Likewise.
5593 (vst1q_u8_x2): Likewise.
5594 (vst1q_u16_x2): Likewise.
5595 (vst1q_u32_x2): Likewise.
5596 (vst1q_u64_x2): Likewise.
5597 (vst1q_f16_x2): Likewise.
5598 (vst1q_f32_x2): Likewise.
5599 (vst1q_f64_x2): Likewise.
5600 (vst1q_p64_x2): Likewise.
5601 (vst1_s64_x3): Likewise.
5602 (vst1_u64_x3): Likewise.
5603 (vst1_f64_x3): Likewise.
5604 (vst1_s8_x3): Likewise.
5605 (vst1_p8_x3): Likewise.
5606 (vst1_s16_x3): Likewise.
5607 (vst1_p16_x3): Likewise.
5608 (vst1_s32_x3): Likewise.
5609 (vst1_u8_x3): Likewise.
5610 (vst1_u16_x3): Likewise.
5611 (vst1_u32_x3): Likewise.
5612 (vst1_f16_x3): Likewise.
5613 (vst1_f32_x3): Likewise.
5614 (vst1_p64_x3): Likewise.
5615 (vst1q_s8_x3): Likewise.
5616 (vst1q_p8_x3): Likewise.
5617 (vst1q_s16_x3): Likewise.
5618 (vst1q_p16_x3): Likewise.
5619 (vst1q_s32_x3): Likewise.
5620 (vst1q_s64_x3): Likewise.
5621 (vst1q_u8_x3): Likewise.
5622 (vst1q_u16_x3): Likewise.
5623 (vst1q_u32_x3): Likewise.
5624 (vst1q_u64_x3): Likewise.
5625 (vst1q_f16_x3): Likewise.
5626 (vst1q_f32_x3): Likewise.
5627 (vst1q_f64_x3): Likewise.
5628 (vst1q_p64_x3): Likewise.
5630 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5632 * config/msp430/msp430.c (msp430_output_labelref): Prepend
5633 user_label_prefix to name.
5635 * tree-core.h: Update comment about the format of NAME string
5636 passed to handler in attribute_spec.
5638 * config/msp430/msp430.md: Remove erroneous subreg expression from
5639 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
5640 zero_extend{q,h}isi2.
5642 2018-05-30 Borislav Petkov <bp@suse.de>
5644 * doc/extend.texi: Document some architecture specific
5645 constraints and sort entries.
5647 2018-05-30 Martin Sebor <msebor@redhat.com>
5650 * builtins.c (expand_builtin_stpcpy_1): New function.
5651 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
5652 only if the former succeeds.
5654 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
5656 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
5659 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5661 * doc/invoke.texi (-flinker-output): Document
5663 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5665 * passes.c (ipa_write_summaries): Only modify statements if body
5667 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
5668 incrementally linking.
5669 (ipa_passes): Likewise.
5670 * lto-cgraph.c (lto_output_node): When incrementally linking do not
5671 pass down resolution info.
5672 * common.opt (flag_incremental_link): Update info.
5673 * gcc.c (plugin specs): Turn flinker-output=* to
5674 -plugin-opt=-linker-output-known
5675 * toplev.c (compile_file): Also cut compilation when doing incremental
5677 * flag-types. (enum lto_partition_model): Add
5678 LTO_LINKER_OUTPUT_NOLTOREL.
5679 (invoke.texi): Add -flinker-output docs.
5680 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
5681 link same way as WPA; do not stream in dead initializers.
5683 * dwarf2out.c (dwarf2out_die_ref_for_decl,
5684 darf2out_register_external_decl): Support incremental link.
5686 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5688 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
5690 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5692 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
5693 it down to simple_object_copy_lto_debug_sections.
5694 (run_gcc): Determine incremental LTO link time and configure
5695 lto1 into non-wpa mode, disable renaming of debug sections.
5697 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
5699 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
5700 descriptions of various incorrectly documented functions.
5702 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
5705 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5706 address check not strict.
5708 2018-05-30 Richard Biener <rguenther@suse.de>
5710 PR tree-optimization/85964
5711 * tracer.c (better_p): Drop initialized count check, we only
5712 call the function with initialized counts now.
5713 (find_best_successor): Do find a best edge if one
5714 has uninitialized count.
5715 (find_best_predecessor): Likewise. Do BB frequency check only
5716 if count is initialized.
5718 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
5720 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
5721 (aarch64_ldrstr_offset_compare): New.
5722 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
5723 load/store orderings.
5724 (aarch64_gen_adjusted_ldpstp): Likewise.
5726 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
5728 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
5729 Check for subset of GENERAL_REGS and FP_REGS.
5730 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
5733 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
5735 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
5736 and wi::to_poly_offset. Add the current offset and then check
5737 whether the sum fits, rather than using an unchecked addition of
5738 a checked term. Check for a shwi rather than a uhwi.
5739 * expr.c (get_bit_range): Use tree_to_poly_uint64.
5740 (store_constructor): Use poly_int_tree_p.
5741 (expand_expr_real_1): Likewise.
5742 * function.c (assign_temp): Likewise.
5743 * fold-const.c (const_binop): Use poly_int_tree_p and
5745 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
5747 * ipa-icf-gimple.c (func_checker::compare_operand): Use
5748 to_poly_offset for MEM offsets.
5749 * ipa-icf.c (sem_variable::equals): Likewise.
5750 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
5751 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
5752 wi::to_poly_offset for BIT_FIELD_REF offsets.
5753 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
5755 * var-tracking.c (emit_note_insn_var_location): Use
5756 tree_to_poly_uint64.
5758 2018-05-29 Jim Wilson <jimw@sifive.com>
5760 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
5762 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
5765 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
5766 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
5768 (sse4_1_round<mode>2): Use nonimmediate_operand
5769 for operand 1 predicate.
5771 2018-05-29 Martin Sebor <msebor@redhat.com>
5772 Richard Biener <rguenther@suse.de>
5775 * calls.c (get_size_range): Call determine_value_range instead
5776 of get_value_range..
5777 * tree-vrp.h (determine_value_range): Declared new function.
5778 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
5780 2018-05-29 Richard Biener <rguenther@suse.de>
5782 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
5783 sure to use non-pattern stmts for get_earlier_stmt arguments.
5784 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
5785 called on pattern stmts.
5786 (get_later_stmt): Likewise.
5788 2018-05-29 Martin Liska <mliska@suse.cz>
5790 PR gcov-profile/85759
5791 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
5794 2018-05-29 Jakub Jelinek <jakub@redhat.com>
5796 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
5798 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
5802 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5803 VEC_PACK_FLOAT_EXPR): New tree codes.
5804 * tree-pretty-print.c (op_code_prio): Handle
5805 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5806 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5807 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5808 * tree-inline.c (estimate_operator_cost): Likewise.
5809 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
5810 * fold-const.c (const_binop): Likewise.
5811 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
5812 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5813 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
5814 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
5815 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5816 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5817 * expr.c (expand_expr_real_2): Likewise.
5818 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
5819 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
5820 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
5822 * optabs.c (expand_widen_pattern_expr): For
5823 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
5824 sign from result type rather than operand's type.
5825 (expand_binop_directly): For vec_packu_float_optab and
5826 vec_packs_float_optab allow result type to be different from operand's
5828 * optabs-tree.c (optab_for_tree_code): Handle
5829 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5830 VEC_PACK_FLOAT_EXPR. Formatting fixes.
5831 * tree-vect-generic.c (expand_vector_operations_1): Handle
5832 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5833 VEC_PACK_FLOAT_EXPR.
5834 * tree-vect-stmts.c (supportable_widening_operation): Handle
5836 (supportable_narrowing_operation): Handle FLOAT_EXPR.
5837 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
5838 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
5839 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
5840 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
5842 (vec_pack<floatprefix>_float_<mode>): New expander.
5843 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
5845 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
5846 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
5847 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
5848 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
5849 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
5851 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
5852 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
5853 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5854 VEC_PACK_FLOAT_EXPR): Document.
5856 2018-05-29 Richard Biener <rguenther@suse.de>
5858 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
5860 (stmt_vec_info_vec): Make pointer.
5861 (init_stmt_vec_info_vec): Remove.
5862 (free_stmt_vec_info_vec): Likewise.
5863 (set_stmt_vec_info_vec): New function.
5864 (free_stmt_vec_infos): Likewise.
5865 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
5866 (set_vinfo_for_stmt): Likewise.
5867 (get_earlier_stmt): Likewise.
5868 (get_later_stmt): Likewise.
5869 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
5870 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
5871 (vec_info::~vec_info): Free stmt_vec_infos.
5872 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
5873 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
5874 (pass_slp_vectorize::execute): Likewise.
5875 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
5876 (free_stmt_vec_info_vec): Likewise.
5877 (set_stmt_vec_info_vec): New function.
5878 (free_stmt_vec_infos): Likewise.
5879 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
5880 the global stmt_vec_info_vec.
5881 * tree-parloops.c (gather_scalar_reductions): Use
5882 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
5885 2018-05-29 Richard Biener <rguenther@suse.de>
5887 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
5889 2018-05-29 Martin Liska <mliska@suse.cz>
5890 David Malcolm <dmalcolm@redhat.com>
5892 * vec.c (test_reverse): New.
5893 (vec_c_tests): Add new test.
5894 * vec.h (vl_ptr>::reverse): New function.
5896 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
5898 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
5900 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
5903 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
5905 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
5907 2018-05-28 Richard Biener <rguenther@suse.de>
5909 PR tree-optimization/85933
5910 * tree-vect-data-refs.c (vect_record_base_alignments): Only
5911 look at stmts marked as vectorizable.
5913 2018-05-28 Richard Biener <rguenther@suse.de>
5915 PR tree-optimization/85934
5916 * tree-vect-generic.c (expand_vector_operations_1): Hoist
5917 vector boolean check before scalar optimization.
5919 2018-05-28 Jakub Jelinek <jakub@redhat.com>
5921 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
5924 2018-05-28 Mark Wielaard <mark@klomp.org>
5926 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
5927 if it is an expression containing a minus sign.
5929 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
5931 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
5933 2018-05-27 Paul Koning <ni1d@arrl.net>
5935 * config/pdp11/pdp11.md (truncsihi2): Remove.
5937 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
5938 Chung-Ju Wu <jasonwucj@gmail.com>
5940 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
5942 (unaligned_store_dw): Ditto.
5943 * config/nds32/nds32-memory-manipulation.c
5944 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
5945 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
5946 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
5947 (emit_setmem_word_loop): Rename to ...
5948 (emit_setmem_doubleword_loop): ... this.
5949 (nds32_gen_dup_4_byte_to_word_value): New function.
5950 (nds32_gen_dup_8_byte_to_double_word_value): New function.
5951 (nds32_expand_setmem_loop): Refine implementation.
5952 (nds32_expand_setmem_loop_v3m): Ditto.
5953 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
5956 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5958 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
5960 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5962 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
5963 (nds32_init_machine_status): Initialize machine->attr_naked_p and
5964 machine->attr_no_prologue_p.
5965 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
5966 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
5967 (nds32_expand_epilogue): Consider attr_naked_p.
5968 (nds32_expand_epilogue_v3pop): Likewise.
5969 (nds32_can_use_return_insn): Likewise.
5970 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
5971 attr_no_prologue_p fields.
5972 * config/nds32/nds32.opt (mret-in-naked-func): New option.
5974 2018-05-27 Jakub Jelinek <jakub@redhat.com>
5977 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
5979 * config/i386/sse.md
5980 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
5982 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
5984 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
5986 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
5988 (*<floatsuffix>floatv2div2sf2): Rename to ...
5989 (*float<floatunssuffix>v2div2sf2): ... this.
5990 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
5991 (float<floatunssuffix>v2div2sf2_mask): ... this.
5992 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
5993 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
5994 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
5996 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
5998 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
6000 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
6002 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
6004 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
6006 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
6007 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
6008 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
6009 gen_ufix_truncv8dfv8si2.
6010 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
6011 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
6012 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
6013 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
6014 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
6015 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
6016 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
6017 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
6019 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
6023 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
6025 2018-05-25 Jim Wilson <jimw@sifive.com>
6027 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
6028 * config/riscv/riscv.c (struct machine_function): Add
6029 interrupt_handler_p and attribute_checked_p fields.
6030 (riscv_attribute_table): Add interrupt.
6031 (riscv_interrupt_type_p): New.
6032 (riscv_save_reg_p): Save extra regs for interrupt handler.
6033 (riscv_use_save_libcall): Return false for interrupt handler.
6034 (riscv_first_stack_step): Add forward declaration.
6035 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
6036 for interrupt handler with large frame. Use it for saved reg list.
6037 (riscv_expand_prologue): Move flag_stack_usage_info support to
6038 eliminate duplication.
6039 (riscv_expand_epilogue): Generate mret for interrupt handler.
6040 (riscv_epilogue_uses): New.
6041 (riscv_can_use_return_insn): Return false for interrupt handler.
6042 (riscv_function_ok_for_sibcall): Likewise.
6043 (riscv_set_current_function): Add interrupt handler support.
6044 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
6045 * config/riscv/riscv.md (UNSPECV_MRET): New.
6047 (riscv_frflags, riscv_fsflags): Use tab after opcode.
6049 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
6051 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
6053 PR tree-optimization/85712
6054 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
6055 this candidate has already been replaced in-situ by a copy.
6057 2018-05-25 Jason Merrill <jason@redhat.com>
6059 PR c++/80485 - inline function non-zero address.
6060 * symtab.c (nonzero_address): Check DECL_COMDAT.
6062 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
6065 * config/alpha/alpha.md (ashlsi3): New insn pattern.
6066 (*ashlsi_se): Rename from *ashldi_se. Define as sign
6067 extension of SImode operation. Use const123_operand predicate.
6068 (*saddsi_1): Remove.
6069 (*saddl_se_1): Ditto.
6071 (*ssubl_se_1): Ditto.
6072 * config/alpha/predicates.md (const123_operand): New predicate.
6073 * config/alpha/constraints.md (P): Use IN_RANGE.
6075 2018-05-25 Richard Biener <rguenther@suse.de>
6077 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
6079 (ref_maybe_used_by_stmt_p): Likewise.
6080 (stmt_may_clobber_ref_p): Likewise.
6081 (stmt_may_clobber_ref_p_1): Likewise.
6082 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
6084 (ref_maybe_used_by_stmt_p): Likewise.
6085 (stmt_may_clobber_ref_p): Likewise.
6086 (stmt_may_clobber_ref_p_1): Likewise.
6087 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
6088 the alias oracle to disambiguate DRs with stmts DR analysis
6090 (vect_analyze_data_refs): Do not give up on not analyzable
6091 DRs for BB vectorization. Remove code truncating the dataref
6094 2018-05-25 Jakub Jelinek <jakub@redhat.com>
6097 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
6098 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
6099 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
6101 2018-05-25 Richard Biener <rguenther@suse.de>
6103 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
6104 function, combining stmt data ref gathering and fatal analysis
6106 (vect_analyze_data_refs): Remove now redudnant code and simplify.
6107 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
6108 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
6109 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
6110 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
6112 2018-05-25 Bin Cheng <bin.cheng@arm.com>
6114 PR tree-optimization/85720
6115 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
6116 SCC if all partitions are builtins.
6117 (version_loop_by_alias_check): New parameter. Generate cancelable
6118 runtime alias check if all partitions are builtins.
6119 (distribute_loop): Update call to above function.
6121 2018-05-25 Bin Cheng <bin.cheng@arm.com>
6123 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
6124 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
6125 (parm_default_def_partition_arg): Ditto.
6126 (set_parm_default_def_partition): Ditto.
6127 (get_parm_default_def_partitions): Ditto and make it static.
6128 (get_undefined_value_partitions): Ditto and make it static.
6129 (remove_ssa_form): Refactor call to init_var_map here.
6130 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
6131 computation for loop region.
6132 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
6133 (register_default_def): Delete.
6134 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
6135 (parm_default_def_partition_arg): Ditto.
6136 (set_parm_default_def_partition): Ditto.
6137 (get_parm_default_def_partitions): Ditto and make it static.
6138 (get_undefined_value_partitions): Ditto and make it static.
6139 (coalesce_with_default, coalesce_with_default): Update comment.
6140 (create_coalesce_list_for_region): New func factored out from
6141 create_outofssa_var_map.
6142 (populate_coalesce_list_for_outofssa): New func factored out from
6143 create_outofssa_var_map and coalesce_ssa_name.
6144 (create_outofssa_var_map): Delete.
6145 (coalesce_ssa_name): Refactor to support live range computation.
6146 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
6147 (get_parm_default_def_partitions): Delete.
6148 (get_undefined_value_partitions): Ditto.
6149 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
6150 computation for loop region.
6151 (new_tree_live_info, loe_visit_block): Ditto.
6152 (live_worklist, set_var_live_on_entry): Ditto.
6153 (calculate_live_on_exit, verify_live_on_entry): Ditto.
6154 * tree-ssa-live.h (struct _var_map): New fields.
6155 (init_var_map): Change decl.
6156 (region_contains_p): New.
6158 2018-05-25 Bin Cheng <bin.cheng@arm.com>
6160 * tree-ssa-live.h (live_merge_and_clear): Delete.
6162 2018-05-25 Richard Biener <rguenther@suse.de>
6165 * tree-dump.c (dequeue_and_dump): Remove access to removed
6166 operand 2 of a SWITCH_EXPR.
6168 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6170 * doc/sourcebuild.texi (vect_double_cond_arith): Include
6171 multiplication and division.
6172 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
6173 (cond_udiv@var{m}, cond_umod@var{m}): Document.
6174 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
6175 (cond_udiv_optab, cond_umod_optab): New optabs.
6176 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
6177 (IFN_COND_RDIV): New internal functions.
6178 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
6179 TRUNC_MOD_EXPR and RDIV_EXPR.
6180 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
6181 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
6183 (SVE_INT_BINARY): Include mult.
6184 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
6185 (optab, sve_int_op): Handle mult.
6186 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
6188 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
6189 for SVE_INT_BINARY_SD.
6191 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6193 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
6194 (optab, sve_int_op): Handle div and udiv.
6195 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
6196 for SVE_INT_BINARY_SD.
6197 (*<optab><mode>3): New insn for the same.
6199 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6201 * tree-vect-patterns.c: Include predict.h.
6202 (vect_recog_divmod_pattern): Restrict check for division support
6203 to when optimizing for size.
6205 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6207 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
6208 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
6209 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
6210 (gimple_match_op::set_op): Likewise.
6211 (gimple_resimplify4): Declare.
6212 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
6213 (expr::gen_transform): Likewise.
6214 (decision_tree::gen): Generate a simplification routine for 4 operands.
6215 * gimple-match-head.c (gimple_simplify): Add an overload for
6216 4 operands. In the top-level function, handle up to 4 call
6217 arguments and call gimple_resimplify4.
6218 (gimple_resimplify4): New function.
6219 (build_call_internal): Pass a fourth operand.
6220 (maybe_push_to_seq): Likewise.
6221 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
6222 Fold VEC_COND_EXPRs of an operation and a default value into
6223 an IFN_COND_* function if possible.
6224 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
6226 (SVE_COND_FP_BINARY): Include them.
6227 (optab, sve_fp_op): Handle them.
6228 (SVE_INT_BINARY_REV): New code iterator.
6229 (SVE_COND_FP_BINARY_REV): New int iterator.
6230 (commutative): New int attribute.
6231 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
6233 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
6235 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
6236 (*cond_<optab><mode>): New patterns for reversed operands.
6238 2018-05-25 Richard Biener <rguenther@suse.de>
6240 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
6241 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
6242 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
6243 (STMT_VINFO_GROUPED_ACCESS): Adjust.
6244 * tree-vect-data-refs.c (everywhere): Adjust users.
6245 * tree-vect-loop.c (everywhere): Likewise.
6246 * tree-vect-slp.c (everywhere): Likewise.
6247 * tree-vect-stmts.c (everywhere): Likewise.
6248 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
6250 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6252 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
6254 (gcc_cv_as_section_exclude): ... this.
6255 Try Solaris as #exclude syntax.
6256 * configure: Regenerate.
6257 * config.in: Regenerate.
6258 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
6260 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
6261 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
6263 * varasm.c (default_elf_asm_named_section): Don't check if
6264 HAVE_GAS_SECTION_EXCLUDE is defined.
6266 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6268 * doc/md.texi: Update the documentation of the cond_* optabs
6269 to mention the new final operand. Fix GET_MODE_NUNITS call.
6270 Describe the scalar case too.
6271 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
6272 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
6274 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
6275 (get_conditional_internal_fn): Update comment.
6276 * tree-vect-loop.c (vectorizable_reduction): Pass the original
6277 accumulator value as a final argument to conditional functions.
6278 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
6279 a define_expand and add an "else" operand. Assert for now that
6280 the else operand is equal to operand 2. Use SVE_INT_BINARY and
6281 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
6282 (*cond_<optab><mode>): New patterns.
6283 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
6284 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6285 (UNSPEC_COND_EOR): Delete.
6286 (optab): Remove associated mappings.
6287 (SVE_INT_BINARY): New code iterator.
6288 (sve_int_op): Remove int attribute and add "minus" to the code
6290 (SVE_COND_INT_OP): Delete.
6291 (SVE_COND_FP_OP): Rename to...
6292 (SVE_COND_FP_BINARY): ...this.
6294 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6296 * optabs.c (can_reuse_operands_p): New function.
6297 (maybe_legitimize_operands): Try to reuse the results for
6300 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
6302 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
6303 Add {q} suffix to insn mnemonic.
6305 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6307 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
6308 (msp430_warn_func_return): New.
6310 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
6312 * fold-const.c (tree_nonzero_bits): New function.
6313 * fold-const.h (tree_nonzero_bits): Likewise.
6314 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
6315 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
6317 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
6321 * varasm.c (assemble_alias): Check ifunc_resolver only on
6324 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
6327 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
6328 when memory input operand is handled.
6330 2018-05-24 Luis Machado <luis.machado@linaro.org>
6332 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
6334 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
6336 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
6338 * match.pd: Delay FMA folds until after vectorization.
6340 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
6343 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
6344 address check not strict.
6346 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
6348 * gimple-match.h (gimple_match_op): New class.
6349 (mprts_hook): Replace parameters with a gimple_match_op *.
6350 (maybe_build_generic_op): Likewise.
6351 (gimple_simplified_result_is_gimple_val): Replace parameters with
6352 a const gimple_match_op *.
6353 (gimple_simplify): Replace code_helper * and tree * parameters with
6354 a gimple_match_op * parameter.
6355 (gimple_resimplify1): Replace code_helper *, tree and tree *
6356 parameters with a gimple_match_op * parameter.
6357 (gimple_resimplify2): Likewise.
6358 (gimple_resimplify3): Likewise.
6359 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
6360 parameters with a gimple_match_op * parameter.
6361 * gimple-match-head.c (gimple_simplify): Change prototypes of
6362 auto-generated functions to take a gimple_match_op * instead of
6363 separate code_helper * and tree * parameters. Make the same
6364 change in the top-level overload and update calls to the
6365 gimple_resimplify routines. Update calls to the auto-generated
6366 functions and to maybe_push_res_to_seq in the publicly-facing
6367 operation-specific gimple_simplify overloads.
6368 (gimple_match_op::MAX_NUM_OPS): Define.
6369 (gimple_resimplify1): Replace rcode and ops with a single res_op
6370 parameter. Update call to gimple_simplify.
6371 (gimple_resimplify2): Likewise.
6372 (gimple_resimplify3): Likewise.
6373 (mprts_hook): Replace parameters with a gimple_match_op *.
6374 (maybe_build_generic_op): Likewise.
6375 (build_call_internal): Replace type, nargs and ops with
6376 a gimple_match_op *.
6377 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
6378 with a single gimple_match_op *. Update calls to mprts_hook,
6379 build_call_internal and gimple_simplified_result_is_gimple_val.
6380 Factor out code that is common to the tree_code and combined_fn cases.
6381 * genmatch.c (expr::gen_transform): Replace tem_code and
6382 tem_ops with a gimple_match_op called tem_op. Update calls
6383 to the gimple_resimplify functions and maybe_push_res_to_seq.
6384 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
6385 res_ops. Update call to the gimple_resimplify functions.
6386 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
6387 (decision_tree::gen): Make the functions take a gimple_match_op *
6388 called res_op instead of separate res_code and res_ops parameters.
6389 Update call accordingly.
6390 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
6391 and ops with a single res_op parameter. Update calls to
6392 maybe_build_generic_op and maybe_push_res_to_seq.
6393 (fold_stmt_1): Update calls to gimple_simplify and
6394 replace_stmt_with_simplification.
6395 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
6396 and gimple_simplified_result_is_gimple_val.
6397 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
6399 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
6400 with a gimple_match_op *.
6401 (vn_nary_build_or_lookup): Likewise. Update call to
6402 vn_nary_build_or_lookup_1.
6403 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
6404 gimple_match_op *. Update calls to the gimple_resimplify routines
6405 and to gimple_simplified_result_is_gimple_val.
6406 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
6407 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
6408 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
6409 (visit_nary_op): Likewise.
6410 (visit_reference_op_load): Likewise.
6412 2018-05-23 Luis Machado <luis.machado@linaro.org>
6414 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
6415 modifier for printing the step amount.
6417 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
6420 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
6423 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
6425 * doc/sourcebuild.texi (Endianness): New subsubsection.
6427 2018-05-23 Luis Machado <luis.machado@linaro.org>
6429 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6430 <prefetch_dynamic_strides>: New const bool field.
6431 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6432 prefetch_dynamic_strides.
6433 (exynosm1_prefetch_tune): Likewise.
6434 (thunderxt88_prefetch_tune): Likewise.
6435 (thunderx_prefetch_tune): Likewise.
6436 (thunderx2t99_prefetch_tune): Likewise.
6437 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
6439 (aarch64_override_options_internal): Update to set
6440 PARAM_PREFETCH_DYNAMIC_STRIDES.
6441 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6442 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6443 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6444 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6445 prefetch-dynamic-strides setting.
6447 2018-05-23 Luis Machado <luis.machado@linaro.org>
6449 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6450 <minimum_stride>: New const int field.
6451 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6452 minimum_stride field defaulting to -1.
6453 (exynosm1_prefetch_tune): Likewise.
6454 (thunderxt88_prefetch_tune): Likewise.
6455 (thunderx_prefetch_tune): Likewise.
6456 (thunderx2t99_prefetch_tune): Likewise.
6457 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
6458 <default_opt_level>: Set to 3.
6459 (aarch64_override_options_internal): Update to set
6460 PARAM_PREFETCH_MINIMUM_STRIDE.
6461 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6462 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6463 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6464 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6465 stride is constant and is below the minimum stride threshold.
6467 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6469 * config/arm/arm-cpus.in (mode26): Delete.
6470 (armv4): Delete mode26 reference.
6471 * config/arm/arm.c (arm_configure_build_target): Delete use of
6474 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
6476 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
6478 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
6479 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
6481 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
6482 (floatunsdidf2): Ditto.
6484 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
6486 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
6487 (fixuns_trunc<mode>si2_avx512f): Ditto.
6488 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
6489 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
6490 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
6492 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
6494 PR rtl-optimization/79985
6495 * df-scan.c (df_insn_refs_collect): Remove special case for
6496 global registers and asm statements.
6498 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
6500 * extend.texi (Global Register Variables): Rewrite the bullet list.
6501 Note that the register is available for allocation. Note that access
6502 via inline asm must use constraints. Add note about async-signal
6503 handlers. Remove paragraph about automagic register selection.
6505 2018-05-23 Richard Biener <rguenther@suse.de>
6507 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
6508 of fixed offset from memset VN.
6510 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
6512 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
6514 (alloc_cand_and_find_basis): Initialize first_interp field.
6515 (slsr_process_mul): Modify first_interp field.
6516 (slsr_process_add): Likewise.
6517 (slsr_process_cast): Modify first_interp field for each new
6519 (slsr_process_copy): Likewise.
6520 (dump_candidate): Dump first_interp field.
6521 (replace_mult_candidate): Process all interpretations, not just
6523 (replace_rhs_if_not_dup): Likewise.
6524 (replace_one_candidate): Likewise.
6526 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
6528 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
6530 (aarch64_needs_frame_chain): New function.
6531 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
6533 2018-05-23 Sudakshina Das <sudi.das@arm.com>
6536 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
6537 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
6538 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
6539 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
6540 as true for strict-align.
6541 (aarch64_can_inline_p): Perform checks even when callee has no
6542 attributes to check for strict alignment.
6543 * doc/extend.texi (AArch64 Function Attributes): Document
6545 * doc/invoke.texi: (AArch64 Options): Likewise.
6547 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
6549 PR tree-optimization/85853
6550 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
6551 the handling of the root of the node to...
6552 (vect_slp_analyze_node_operations_1): ...this new function,
6553 and run the whole thing with the child nodes' def types
6554 set according to their SLP node's def type.
6556 2018-05-23 Richard Biener <rguenther@suse.de>
6559 * tree-data-ref.c (create_runtime_alias_checks): Defer
6560 and ignore overflow warnings.
6562 2018-05-23 Yury Gribov <tetra2005@gmail.com>
6564 PR tree-optimization/85822
6565 * tree-vrp.c (is_masked_range_test): Fix handling of negative
6568 2018-05-23 Richard Biener <rguenther@suse.de>
6570 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
6571 memset constants via native_interpret_expr.
6573 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
6576 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
6578 (cgraph_node::create_alias): Likewise.
6579 (cgraph_node::get_availability): Check ifunc_resolver instead
6580 of looking up ifunc attribute.
6581 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
6582 * varasm.c (do_assemble_alias): Likewise.
6583 (assemble_alias): Likewise.
6584 (default_binds_local_p_3): Likewise.
6585 * cgraph.h (cgraph_node): Add ifunc_resolver.
6586 (cgraph_node::only_called_directly_or_aliased_p): Return false
6588 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
6590 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
6591 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
6592 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
6593 instead of looking up ifunc attribute.
6595 2018-05-22 Luis Machado <luis.machado@linaro.org>
6597 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
6599 2018-05-22 Martin Sebor <msebor@redhat.com>
6602 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
6603 only when expasion succeeds.
6604 (expand_builtin_strcmp): Same.
6605 (expand_builtin_strncmp): Same.
6607 2018-05-22 Martin Sebor <msebor@redhat.com>
6609 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
6611 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
6612 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6614 * config/aarch64/aarch64-ldpstp.md: Replace uses of
6615 aarch64_mem_pair_operand with memory_operand and delete operand swapping
6617 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
6618 Add check for legitimate_address.
6619 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
6620 (aarch64_swap_ldrstr_operands): New.
6621 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
6624 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
6625 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6627 * config/aarch64/aarch64.md: New patterns to generate stp
6629 (store_pair_sw, store_pair_dw): New patterns to generate stp for
6630 single words and double words.
6631 (load_pair_sw, load_pair_dw): Likewise.
6632 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
6634 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
6636 * config/aarch64/aarch64-ldpstp.md: Modify peephole
6637 for different mode ldpstp and add peephole for merged zero stores.
6639 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
6641 (aarch64_gen_store_pair): Rename calls to match new patterns.
6642 (aarch64_gen_load_pair): Rename calls to match new patterns.
6643 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
6644 (load_pair<DREG:mode><DREG2:mode>): ... This.
6645 (store_pair<mode>): Rename to...
6646 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
6647 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
6649 (V_INT_EQUIV): Handle SImode.
6650 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
6653 2018-05-22 Martin Sebor <msebor@redhat.com>
6656 * calls.c (maybe_warn_nonstring_arg): Use string length to set
6657 or ajust the presumed bound on an operation to avoid unnecessary
6660 2018-05-22 Martin Sebor <msebor@redhat.com>
6662 PR tree-optimization/85826
6663 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
6664 assuming that a DECL necesarily has a constant size.
6666 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
6669 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
6671 2018-05-22 Richard Biener <rguenther@suse.de>
6673 PR tree-optimization/85834
6674 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
6675 non-constant and non-zero memset arguments.
6677 2018-05-22 Martin Liska <mliska@suse.cz>
6680 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
6682 2018-05-22 Richard Biener <rguenther@suse.de>
6684 PR tree-optimization/85863
6685 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
6686 comparisons when vectype is specified.
6687 (vectorizable_condition): Do not specify vectype for
6688 vect_is_simple_cond when SLP vectorizing.
6690 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
6693 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
6694 define __ibm128 as long double.
6695 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
6696 as a distinct type when IEEE 128-bit support is enabled.
6697 (init_float128_ieee): Fix up conversions between IFmode and IEEE
6698 128-bit types to use the correct functions.
6699 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
6700 convert between 128-bit floating point types that have different
6701 modes but the same representation, instead of using gen_lowpart to
6703 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
6705 (IFKF_reg): New attributes to give the register constraints for
6707 (extend<mode>tf2_internal): New insns to mark an explicit
6708 conversion between 128-bit floating point types that have a
6709 different mode but share the same representation.
6711 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
6713 PR tree-optimization/85814
6714 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
6715 a null return from get_strinfo when unsharing the next
6716 strinfo in the chain.
6718 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6721 * varasm.c (weak_finish): Clean up weak_decls.
6723 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6725 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
6726 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
6727 UNSPEC_UADALP values.
6728 * config/aarch64/iterators.md (ABAL): New int iterator.
6731 (sur): Add mappings for the above.
6732 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
6734 (aarch64_<sur>abal<mode>_4): Likewise.
6735 (aarch64_<sur>adalp<mode>_3): Likewise.
6736 (<sur>sadv16qi): New define_expand.
6738 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
6740 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
6741 (*movdf_internal): Ditto.
6742 (*rcpsf2_sse): Ditto.
6743 (*rsqrtsf2_sse): Ditto.
6744 (*sqrt<mode>2_sse): Ditto.
6746 2018-05-21 Tamar Christina <tamar.christina@arm.com>
6748 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
6750 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
6751 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
6752 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
6753 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
6755 * config/aarch64/arm_neon.h: Likewise.
6756 * config/aarch64/iterators.md (VQ_I): New.
6758 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
6760 * config.gcc: Add arc/t-multilib-linux to tmake_file for
6762 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
6765 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6767 * config/nds32/constraints.md (S): New constraint.
6768 * config/nds32/nds32.md (call_internal): Use constraint S.
6769 (call_value_internal): Likewise.
6770 (sibcall_internal): Likewise.
6771 (sibcall_value_internal): Likewise.
6773 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6774 Chung-Ju Wu <jasonwucj@gmail.com>
6776 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
6779 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6780 Chung-Ju Wu <jasonwucj@gmail.com>
6782 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
6783 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
6784 (nds32_rtx_costs_impl): Simplify.
6785 (nds32_address_cost_impl): Simplify.
6786 (nds32_init_rtx_costs): New function.
6787 (nds32_rtx_costs_speed_prefer): Likewise.
6788 (nds32_rtx_costs_size_prefer): Likewise.
6789 (nds32_address_cost_speed_prefer): Likewise.
6790 (nds32_address_cost_speed_fwprop): Likewise.
6791 (nds32_address_cost_size_prefer): Likewise.
6792 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
6793 * config/nds32/nds32.c (nds32_option_override): Use
6794 nds32_init_rtx_costs function.
6796 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6798 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
6799 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
6800 (TARGET_PIPELINE_N8): Likewise.
6801 (TARGET_PIPELINE_N10): Likewise.
6802 (TARGET_PIPELINE_N13): Likewise.
6803 (TARGET_PIPELINE_GRAYWOLF): Likewise.
6805 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
6807 * config/nds32/nds32-fpu.md: Update copyright year.
6809 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6811 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
6813 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6815 * config/nds32/nds32.c
6816 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
6817 * config/nds32/nds32.opt (minline-asm-r15): New option.
6819 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6821 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
6823 * config/nds32/nds32.md (abssi2): New pattern.
6825 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
6827 * config/i386/i386.md (rex64namesuffix): New mode attribute.
6828 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
6829 Merge insn pattern from sse_cvtsi2ss<round_name> and
6830 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
6831 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
6832 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
6833 using SWI48 mode iterator.
6834 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
6835 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
6836 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
6837 pattern from sse_cvttss2si<round_saeonly_name>
6838 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
6839 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
6840 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
6841 using SWI48 mode iterator.
6842 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
6843 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
6844 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
6845 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
6846 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
6847 using SWI48 mode iterator.
6848 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
6849 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
6850 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
6851 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
6852 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
6853 SWI48 mode iterator.
6854 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
6855 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
6856 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
6857 pattern from sse_cvttsd2si<round_saeonly_name>
6858 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
6860 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6862 * config/nds32/nds32-md-auxiliary.c
6863 (nds32_valid_smw_lwm_base_p): Refine.
6864 (nds32_output_smw_single_word): Refine.
6865 (nds32_output_smw_double_word): New.
6866 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
6868 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6870 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
6871 (nds32_output_stack_pop): Refine.
6872 (nds32_expand_unaligned_load): Refine.
6873 (nds32_expand_unaligned_store): Refine.
6875 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6876 Chung-Ju Wu <jasonwucj@gmail.com>
6878 * config/nds32/constants.md: Add TP_REGNUM constant.
6879 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
6880 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
6882 * config/nds32/nds32-doubleword.md: Consider flag_pic.
6883 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
6884 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
6885 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
6886 and PIC code generation.
6887 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
6889 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
6891 * config/nds32/nds32.md: Support TLS and PIC.
6892 * config/nds32/nds32.c: Support TLS and PIC.
6893 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
6894 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
6897 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6899 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
6900 mode with E_ prefix.
6902 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6903 Chung-Ju Wu <jasonwucj@gmail.com>
6905 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
6906 * config/nds32/nds32-md-auxiliary.c
6907 (symbolic_reference_mentioned_p): New.
6908 (nds32_legitimize_ict_address): New.
6909 (nds32_expand_ict_move): New.
6910 (nds32_indirect_call_referenced_p): New.
6911 (nds32_symbol_binds_local_p): Delete.
6912 (nds32_long_call_p): Modify.
6913 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
6914 * config/nds32/nds32-protos.h
6915 (symbolic_reference_mentioned_p): Declare.
6916 (nds32_legitimize_ict_address): Declare.
6917 (nds32_expand_ict_move): Declare.
6918 (nds32_indirect_call_referenced_p): Declare.
6919 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
6920 (nds32_relax_group): Use nds32_ict_const_p as condition.
6921 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
6922 (nds32_asm_file_start): Output ict_model directive in asm code.
6923 (nds32_legitimate_address_p): Consider indirect call.
6924 (nds32_print_operand): Consider indirect call.
6925 (nds32_print_operand_address): Consider indirect call.
6926 (nds32_insert_attributes): Handle "indirect_call" attribute.
6927 (TARGET_LEGITIMATE_ADDRESS_P): Define.
6928 (TARGET_LEGITIMATE_CONSTANT_P): Define.
6929 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
6930 (TARGET_DELEGITIMIZE_ADDRESS): Define.
6931 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
6932 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
6933 (TARGET_ICT_MODEL_SMALL): Define.
6934 (TARGET_ICT_MODEL_LARGE): Define.
6935 * config/nds32/nds32.md (movsi): Consider ict model.
6936 (call, call_value): Consider ict model.
6937 (sibcall, sibcall_value): Consider ict model.
6938 * config/nds32/nds32.opt (mict-model): New option.
6939 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
6942 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
6943 Monk Chiang <sh.chiang04@gmail.com>
6944 Jim Wilson <jimw@sifive.com>
6946 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
6947 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
6948 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
6949 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
6950 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
6951 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
6952 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
6953 compute save_libcall_adjustment properly.
6954 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
6955 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
6956 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
6957 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
6958 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
6959 (ABI_SPEC): Handle mabi=ilp32e.
6960 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
6961 (RVE): Add RVE mask.
6962 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
6963 <-march>: Add rv32e as an example.
6965 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
6968 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
6969 (intra_create_variable_infos): Handle C++ constructors.
6971 2018-05-18 Martin Liska <mliska@suse.cz>
6973 * passes.def: Remove a redundant pass.
6975 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
6978 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
6980 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6982 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
6984 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
6986 (armv2, armv2a, armv3, armv3m): Delete architectures.
6987 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
6988 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
6989 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
6991 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
6992 (*mulsidi3adddi): Likewise.
6993 (mulsidi3): Likewise.
6994 (*mulsidi3_nov6): Likewise.
6995 (umulsidi3): Likewise.
6996 (umulsidi3_nov6): Likewise.
6997 (umaddsidi4): Likewise.
6998 (*umulsidi3adddi): Likewise.
6999 (smulsi3_highpart): Likewise.
7000 (*smulsi3_highpart_nov6): Likewise.
7001 (umulsi3_highpart): Likewise.
7002 (*umulsi3_highpart_nov6): Likewise.
7003 * config/arm/arm.h (arm_arch3m): Delete.
7004 * config/arm/arm.c (arm_arch3m): Delete.
7005 (arm_option_override_internal): Update armv3-related comment.
7006 (arm_configure_build_target): Delete use of isa_bit_mode32.
7007 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
7008 (arm_rtx_costs_internal): Delete check of arm_arch3m.
7009 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
7011 (mulusa3): Likewise.
7012 * config/arm/arm-protos.h (arm_arch3m): Delete.
7013 * config/arm/arm-tables.opt: Regenerate.
7014 * config/arm/arm-tune.md: Likewise.
7015 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
7016 deleted architectures.
7018 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7020 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
7021 (armv5t, armv5te): New features.
7022 (ARMv5, ARMv5e): Delete fgroups.
7023 (ARMv5t, ARMv5te): Adjust for above changes.
7025 (armv5, armv5e): Delete arches.
7026 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
7028 (*call_reg_arm): Likewise.
7029 (*call_value_reg_armv5): Likewise.
7030 (*call_value_reg_arm): Likewise.
7031 (*call_symbol): Likewise.
7032 (*call_value_symbol): Likewise.
7033 (*sibcall_insn): Likewise.
7034 (*sibcall_value_insn): Likewise.
7036 (prefetch): Likewise.
7037 (define_split and define_peephole2 dependent on arm_arch5):
7039 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
7041 (TARGET_ARM_QBIT): Likewise.
7042 (TARGET_DSP_MULTIPLY): Likewise.
7043 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
7044 (arm_arch5, arm_arch5e): Delete.
7045 (arm_arch5t, arm_arch5te): Declare.
7046 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
7047 (arm_arch5t): Declare.
7048 (arm_option_reconfigure_globals): Update for the above.
7049 (arm_options_perform_arch_sanity_checks): Update comment, replace
7050 use of arm_arch5 with arm_arch5t.
7051 (use_return_insn): Likewise.
7052 (arm_emit_call_insn): Likewise.
7053 (output_return_instruction): Likewise.
7054 (arm_final_prescan_insn): Likewise.
7055 (arm_coproc_builtin_available): Likewise.
7056 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
7057 arm_arch5e with arm_arch5t and arm_arch5te.
7058 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
7059 (arm_arch5t, arm_arch5te): Declare.
7060 * config/arm/arm-tables.opt: Regenerate.
7061 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
7062 * config/arm/t-multilib: Likewise.
7063 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
7064 instead of arm_arch5.
7065 (*call_reg_thumb1): Likewise.
7066 (*call_value_reg_thumb1_v5): Likewise.
7067 (*call_value_reg_thumb1): Likewise.
7068 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
7070 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
7072 2018-05-18 Martin Liska <mliska@suse.cz>
7074 PR gcov-profile/84846
7075 * doc/gcov.texi: Document -t option of gcov tool.
7077 2018-05-18 Martin Liska <mliska@suse.cz>
7079 PR gcov-profile/84846
7080 * gcov.c (print_usage): Add new -t option.
7081 (process_args): Handle the option.
7082 (generate_results): Use stdout as output when requested by
7085 2018-05-18 Martin Liska <mliska@suse.cz>
7087 PR gcov-profile/84846
7088 * coverage.c (coverage_init): Write PWD to .gcno file.
7089 * doc/gcov.texi: Document how working directory is printed.
7090 * gcov-dump.c (dump_gcov_file): Print PWD.
7091 * gcov.c (output_intermediate_file): Likewise.
7092 (read_graph_file): Read PWD string.
7093 (output_lines): Print PWD.
7095 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7098 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
7099 for retval and return false if all args to phi are zero.
7101 2018-05-18 Richard Biener <rguenther@suse.de>
7103 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
7105 (evrp_dom_walker::before_dom_children): Call it.
7107 2018-05-18 Richard Biener <rguenther@suse.de>
7109 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
7110 results when processing array refs with variable index.
7112 2018-05-18 Toon Moene <toon@moene.org>
7114 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
7115 directly after that of -floop-interchange. Indicate that both
7116 options are enabled by default when specifying -O3.
7118 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7120 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
7121 iterator. Delete separate integer-mode vec_set<mode> expander.
7122 (aarch64_simd_vec_setv2di): Delete.
7123 (vec_setv2di): Delete.
7124 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
7125 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
7126 the "w, r" alternative.
7128 2018-05-18 Martin Liska <mliska@suse.cz>
7130 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
7131 * tree-pass.h (make_pass_lower_switch_O0): New function.
7132 * tree-switch-conversion.c (node_has_low_bound): Remove.
7133 (node_has_high_bound): Likewise.
7134 (node_is_bounded): Likewise.
7135 (class pass_lower_switch): Make it a template type and create
7137 (pass_lower_switch::execute): Add template argument.
7138 (make_pass_lower_switch): New function.
7139 (make_pass_lower_switch_O0): New function.
7140 (do_jump_if_equal): Remove.
7141 (emit_case_nodes): Simplify to just handle all 3 cases and leave
7142 all the hard work to tree optimization passes.
7144 2018-05-18 Martin Liska <mliska@suse.cz>
7146 * dbgcnt.c (limit_low): Renamed from limit.
7147 (limit_high): New variable.
7148 (dbg_cnt_is_enabled): Check for upper limit.
7149 (dbg_cnt): Adjust dumping.
7150 (dbg_cnt_set_limit_by_index): Add new argument for high
7152 (dbg_cnt_set_limit_by_name): Likewise.
7153 (dbg_cnt_process_single_pair): Parse new format.
7154 (dbg_cnt_process_opt): Use strtok.
7155 (dbg_cnt_list_all_counters): Remove 'value' and add
7157 * doc/invoke.texi: Document changes.
7159 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
7161 * doc/sourcebuild.texi (scalar_all_fma): Document.
7162 * tree.def (FMA_EXPR): Delete.
7163 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
7164 * internal-fn.c (ternary_direct): New macro.
7165 (expand_ternary_optab_fn): Likewise.
7166 (direct_ternary_optab_supported_p): Likewise.
7167 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
7168 * builtins.c (fold_builtin_fma): Delete.
7169 (fold_builtin_3): Don't call it.
7170 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
7171 * expr.c (expand_expr_real_2): Likewise.
7172 * fold-const.c (operand_equal_p): Likewise.
7173 (fold_ternary_loc): Likewise.
7174 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
7175 * gimple.c (DEFTREECODE): Likewise.
7176 * gimplify.c (gimplify_expr): Likewise.
7177 * optabs-tree.c (optab_for_tree_code): Likewise.
7178 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
7179 * tree-eh.c (operation_could_trap_p): Likewise.
7180 (stmt_could_throw_1_p): Likewise.
7181 * tree-inline.c (estimate_operator_cost): Likewise.
7182 * tree-pretty-print.c (dump_generic_node): Likewise.
7183 (op_code_prio): Likewise.
7184 * tree-ssa-loop-im.c (stmt_cost): Likewise.
7185 * tree-ssa-operands.c (get_expr_operands): Likewise.
7186 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
7187 * fold-const-call.h (fold_fma): Delete.
7188 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
7189 CFN_FNMA and CFN_FNMS.
7191 * genmatch.c (combined_fn): New enum.
7192 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
7193 (commutative_op): New function.
7194 (commutate): Use it. Handle more than 2 operands.
7195 (dt_operand::gen_gimple_expr): Use commutative_op.
7196 (parser::parse_expr): Allow :c to be used with non-binary
7197 operators if the commutative operand is known.
7198 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
7199 CFN_FMS, CFN_FNMA and CFN_FNMS.
7200 (backprop::process_assign_use): Remove FMA_EXPR handling.
7201 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
7202 (gen_hsa_fma): New function.
7203 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
7204 IFN_FNMA and IFN_FNMS.
7205 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
7206 * gimple-fold.h (follow_all_ssa_edges): Declare.
7207 * gimple-fold.c (follow_all_ssa_edges): New function.
7208 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
7209 gimple_build interface and use follow_all_ssa_edges to fold the result.
7210 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
7211 instead of checking for optabs directly.
7212 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
7213 rather than FMA_EXPRs.
7214 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
7215 call to IFN_FMA instead of an FMA_EXPR.
7217 2018-05-17 Jim Wilson <jimw@sifive.com>
7219 * expr.c (do_tablejump): When converting index to Pmode, if we have a
7220 sign extended promoted subreg, and the range does not have the sign bit
7221 set, then do a sign extend.
7223 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
7224 test, check for sign extended subreg and/or constant operands, and
7225 do a sign extend in that case.
7227 2018-05-17 Steve Ellcey <sellcey@cavium.com>
7229 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
7230 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
7232 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
7233 Change logics_shift_reg to logics_shift_imm.
7234 (thunderx2t99_fp_loadpair_basic): Delete.
7235 (thunderx2t99_fp_storepair_basic): Delete.
7236 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
7237 (thunderx2t99_asimd_polynomial): Delete.
7238 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
7239 and neon_fp_mul_d_scalar_q.
7240 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
7241 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
7242 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
7243 (thunderx2t99_asimd_lut): Add missing tbl types.
7244 (thunderx2t99_asimd_ext): Delete.
7245 (thunderx2t99_asimd_load1_1_mult): Delete.
7246 (thunderx2t99_asimd_load1_2_mult): Delete.
7247 (thunderx2t99_asimd_load1_ldp): New.
7248 (thunderx2t99_asimd_load1): New.
7249 (thunderx2t99_asimd_load2): Add missing *load2* types.
7250 (thunderx2t99_asimd_load3): New.
7251 (thunderx2t99_asimd_load4): New.
7252 (thunderx2t99_asimd_store1_1_mult): Delete.
7253 (thunderx2t99_asimd_store1_2_mult): Delete.
7254 (thunderx2t99_asimd_store2_mult): Delete.
7255 (thunderx2t99_asimd_store2_onelane): Delete.
7256 (thunderx2t99_asimd_store_stp): New.
7257 (thunderx2t99_asimd_store1): New.
7258 (thunderx2t99_asimd_store2): New.
7259 (thunderx2t99_asimd_store3): New.
7260 (thunderx2t99_asimd_store4): New.
7262 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
7264 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
7265 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
7267 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
7268 Segher Boessenkool <segher@kernel.crashing.org>
7271 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
7274 2018-05-17 Richard Biener <rguenther@suse.de>
7276 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
7277 for pruning loop and prune defs feeding only already visited PHIs.
7279 2018-05-17 Richard Biener <rguenther@suse.de>
7281 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
7283 2018-05-17 Bin Cheng <bin.cheng@arm.com>
7284 Richard Biener <rguenther@suse.de>
7286 PR tree-optimization/85793
7287 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
7288 for VMAT_ELEMENTWISE.
7290 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
7292 * internal-fn.h (lookup_internal_fn): Declare
7293 * internal-fn.c (lookup_internal_fn): New function.
7294 * gimple.c (gimple_build_call_from_tree): Handle calls to
7296 * gimple-pretty-print.c (dump_gimple_call): Print "." before
7297 internal function names.
7298 * tree-pretty-print.c (dump_generic_node): Likewise.
7299 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
7301 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
7303 * gimple-fold.h (gimple_build): Make the function forms take
7304 combined_fn rather than built_in_function.
7305 (gimple_simplify): Likewise.
7306 * gimple-match-head.c (gimple_simplify): Likewise.
7307 * gimple-fold.c (gimple_build): Likewise.
7308 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
7309 rather than gimple_build_call_internal.
7310 (get_initial_defs_for_reduction): Likewise.
7311 (vect_create_epilog_for_reduction): Likewise.
7312 (vectorizable_live_operation): Likewise.
7314 2018-05-17 Martin Liska <mliska@suse.cz>
7316 * gimple-ssa-sprintf.c (format_directive): Do not use
7317 space in between 'G_' and '('.
7319 2018-05-17 Jakub Jelinek <jakub@redhat.com>
7322 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
7323 even if the mask is not all ones.
7326 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
7328 (ix86_gimple_fold_builtin): Likewise.
7331 * config/i386/i386.c: Include tree-vector-builder.h.
7332 (ix86_vector_shift_count): New function.
7333 (ix86_fold_builtin): Fold shift builtins by scalar count.
7334 (ix86_gimple_fold_builtin): Likewise.
7336 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
7337 _mm512_setzero): New intrinsics.
7339 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
7340 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7342 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
7343 code generation for cases where splatting a value is not useful.
7344 * simplify-rtx.c (simplify_ternary_operation): Simplify
7345 vec_merge across a vec_duplicate and a paradoxical subreg forming
7346 a vector mode to a vec_concat.
7348 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
7350 * config.gcc: Support "goldmont-plus".
7351 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
7353 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7354 PROCESSOR_GOLDMONT_PLUS.
7355 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
7356 (processor_target_table): Add "goldmont-plus".
7357 (PTA_GOLDMONT_PLUS): Define.
7358 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
7359 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
7360 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
7361 (fold_builtin_cpu): Add "goldmont-plus".
7362 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
7363 (ix86_option_override_internal): Add "goldmont-plus".
7364 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
7365 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
7366 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
7367 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
7369 2018-05-17 Richard Biener <rguenther@suse.de>
7371 PR tree-optimization/85757
7372 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
7373 remove defs that only feed that PHI from further processing.
7375 2018-05-16 Jim Wilson <jimw@sifive.com>
7377 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
7379 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
7381 2018-05-16 Mark Wielaard <mark@klomp.org>
7383 * dwarf2out.c (count_index_strings): New function.
7384 (output_indirect_strings): Call count_index_strings and generate
7385 header for dwarf_version >= 5.
7387 2018-05-16 Mark Wielaard <mark@klomp.org>
7389 * dwarf2out.c (dwarf_FORM): New function.
7390 (set_indirect_string): Use dwarf_FORM.
7391 (reset_indirect_string): Likewise.
7392 (size_of_die): Likewise.
7393 (value_format): Likewise.
7394 (output_die): Likewise.
7395 (add_skeleton_AT_string): Likewise.
7396 (output_macinfo_op): Likewise.
7397 (index_string): Likewise.
7398 (output_index_string_offset): Likewise.
7399 (output_index_string): Likewise.
7400 (count_index_strings): Likewise.
7402 2018-05-16 Carl Love <cel@us.ibm.com>
7404 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
7405 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
7407 2018-05-16 Martin Jambor <mjambor@suse.cz>
7409 * ipa-prop.c (ipa_free_all_edge_args): Remove.
7410 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
7412 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
7414 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
7415 (fnma<mode>4): Likewise.
7416 (fms<mode>4): Likewise.
7417 (fnms<mode>4): Likewise.
7418 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
7419 (aarch64_fnma<mode>4): Likewise.
7420 (aarch64_fms<mode>4): Likewise.
7421 (aarch64_fnms<mode>4): Likewise.
7422 (aarch64_fnmadd<mode>4): Likewise.
7424 2018-05-16 Jason Merrill <jason@redhat.com>
7426 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
7428 2018-05-16 Richard Biener <rguenther@suse.de>
7430 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
7431 (dump_stmt_cost): Declare.
7432 (add_stmt_cost): Dump cost we add.
7433 (add_stmt_costs): New function.
7434 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
7436 (vect_analyze_stmt): Adjust prototype.
7437 (vectorizable_condition): Likewise.
7438 (vectorizable_live_operation): Likewise.
7439 (vectorizable_reduction): Likewise.
7440 (vectorizable_induction): Likewise.
7441 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
7442 cost vector to pass to vectorizable_ and record afterwards.
7443 (vect_model_reduction_cost): Take cost vector argument and adjust.
7444 (vect_model_induction_cost): Likewise.
7445 (vectorizable_reduction): Likewise.
7446 (vectorizable_induction): Likewise.
7447 (vectorizable_live_operation): Likewise.
7448 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
7449 SLP_TREE_NUMBER_OF_VEC_STMTS.
7450 (vect_analyze_slp_cost_1): Remove.
7451 (vect_analyze_slp_cost): Likewise.
7452 (vect_slp_analyze_node_operations): Take visited args and
7453 a target cost vector. Avoid processing already visited stmt sets.
7454 (vect_slp_analyze_operations): Use a local cost vector to gather
7455 costs and register those of non-discarded instances.
7456 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
7457 (vect_schedule_slp_instance): Remove copying of
7458 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
7460 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
7461 adding cost. Record cost entry location.
7462 (vect_prologue_cost_for_slp_op): Function to compute cost of
7463 a constant or invariant generated for SLP vect in the prologue,
7464 split out from vect_analyze_slp_cost_1.
7465 (vect_model_simple_cost): Make static. Adjust for SLP costing.
7466 (vect_model_promotion_demotion_cost): Likewise.
7467 (vect_model_store_cost): Likewise, make static.
7468 (vect_model_load_cost): Likewise.
7469 (vectorizable_bswap): Add cost vector arg and adjust.
7470 (vectorizable_call): Likewise.
7471 (vectorizable_simd_clone_call): Likewise.
7472 (vectorizable_conversion): Likewise.
7473 (vectorizable_assignment): Likewise.
7474 (vectorizable_shift): Likewise.
7475 (vectorizable_operation): Likewise.
7476 (vectorizable_store): Likewise.
7477 (vectorizable_load): Likewise.
7478 (vectorizable_condition): Likewise.
7479 (vectorizable_comparison): Likewise.
7480 (can_vectorize_live_stmts): Likewise.
7481 (vect_analyze_stmt): Likewise.
7482 (vect_transform_stmt): Adjust calls to vectorizable_*.
7483 * tree-vectorizer.c: Include gimple-pretty-print.h.
7484 (dump_stmt_cost): New function.
7486 2018-05-16 Richard Biener <rguenther@suse.de>
7488 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
7489 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
7490 * tree-ssa-dse.c: Include tree-ssa-loop.h.
7491 (check_name): New callback.
7492 (dse_classify_store): Track cycles via a visited bitmap of PHI
7493 defs and simplify handling of in-loop and across loop dead stores
7494 and properly fail for loop-variant refs. Handle byte-tracking with
7495 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
7498 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
7500 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
7501 (vect_get_mask_type_for_stmt): Likewise.
7502 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
7504 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
7505 to determine the statement's vector type and the vector type that
7506 should be used for calculating nunits. Deal with cases in which
7507 the type has to be deferred.
7508 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
7509 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
7510 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
7511 (vect_determine_vf_for_stmt): New functions, split out from...
7512 (vect_determine_vectorization_factor): ...here.
7513 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
7514 (vect_get_mask_type_for_stmt): New functions, split out from
7515 vect_determine_vectorization_factor.
7517 2018-05-16 Richard Biener <rguenther@suse.de>
7519 * tree-cfg.c (verify_gimple_assign_ternary): Properly
7520 verify the [VEC_]COND_EXPR embedded comparison.
7522 2018-05-15 Martin Sebor <msebor@redhat.com>
7524 PR tree-optimization/85753
7525 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
7526 RECORD_TYPE in addition to ARRAY_TYPE.
7528 2018-05-15 Martin Sebor <msebor@redhat.com>
7531 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
7533 2018-05-15 Richard Biener <rguenther@suse.de>
7535 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
7536 add by_clobber_p one. Change algorithm to collect all defs
7537 representing uses we need to walk and try reducing them to
7538 a single one before failing.
7539 (dse_dom_walker::dse_optimize_stmt): Adjust.
7541 2018-05-13 Mark Wielaard <mark@klomp.org>
7543 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
7544 (size_of_loc_descr): Likewise.
7545 (output_loc_operands): Likewise.
7546 (output_loc_operands_raw): Likewise.
7547 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
7548 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
7549 (hash_loc_operands): Likewise.
7550 (compare_loc_operands): Likewise.
7552 2018-05-14 Mark Wielaard <mark@klomp.org>
7554 * dwarf2out.c (count_index_addrs): New function.
7555 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
7557 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7559 PR tree-optimization/83648
7560 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
7561 return value as malloc candidate.
7563 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7566 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
7567 param as true in call to suggest_attribute.
7569 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
7571 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
7572 -mreadonly-in-sdata.
7574 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7576 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
7578 (aarch64_crypto_aesd_fused): Likewise.
7580 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
7582 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
7583 (movsi_aarch64): Likewise.
7584 (load_pairsi): Likewise.
7585 (load_pairdi): Likewise.
7586 (store_pairsi): Likewise.
7587 (store_pairdi): Likewise.
7588 (load_pairsf): Likewise.
7589 (load_pairdf): Likewise.
7590 (store_pairsf): Likewise.
7591 (store_pairdf): Likewise.
7592 (zero_extend): Likewise.
7593 (trunc): Swap alternatives.
7594 (fcvt_target): Add '?' to prefer w over r.
7596 2018-05-14 Jakub Jelinek <jakub@redhat.com>
7599 * config/i386/i386.md: Disallow non-commutative arithmetics in
7600 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
7601 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
7602 in the peephole2 before it.
7604 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
7606 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
7607 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
7608 (ix86_handle_option): Handle -mcldemote.
7609 * config.gcc: New header.
7610 * config/i386/cldemoteintrin.h: New file.
7611 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
7612 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
7614 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7615 OPTION_MASK_ISA_CLDEMOTE.
7616 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
7617 (ix86_valid_target_attribute_inner_p): Ditto.
7618 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
7619 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
7620 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
7621 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
7622 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
7624 * config/i386/i386.opt: Add -mcldemote.
7625 * config/i386/x86intrin.h: New header.
7626 * doc/invoke.texi: Add -mcldemote.
7628 2018-05-14 Richard Biener <rguenther@suse.de>
7630 * doc/match-and-simplify.texi: Adjust :s documentation.
7632 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
7634 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
7635 intended memcpy size.
7636 (REORDER_45): Likewise.
7638 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
7640 * sort.cc: New file.
7641 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
7642 * vec.c (qsort_chk): Use gcc_qsort.
7643 * Makefile.in (OBJS-libcommon): Add sort.o.
7644 (build/sort.o): New target. Use it...
7645 (BUILD_RTL): ... here, and...
7646 (build/gencfn-macros): ... here, and...
7647 (build/genmatch): ... here.
7649 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7650 Chung-Ju Wu <jasonwucj@gmail.com>
7652 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
7653 * config/nds32/nds32-graywolf.md: New file.
7654 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
7655 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
7657 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
7658 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
7659 * config/nds32/nds32.md (pipeline_model): Add graywolf.
7660 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
7661 * config/nds32/pipelines.md: Include n15 settings.
7663 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7664 Chung-Ju Wu <jasonwucj@gmail.com>
7666 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
7667 * config/nds32/nds32-n13.md: New file.
7668 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
7669 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
7671 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
7672 * config/nds32/nds32.md (pipeline_model): Add n13.
7673 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
7674 * config/nds32/pipelines.md: Include n13 settings.
7676 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7677 Chung-Ju Wu <jasonwucj@gmail.com>
7679 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
7680 * config/nds32/nds32-n10.md: New file.
7681 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
7682 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
7684 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
7685 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
7686 * config/nds32/nds32.md (pipeline_model): Add n10.
7687 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
7688 * config/nds32/pipelines.md: Include n10 settings.
7690 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
7691 Kito Cheng <kito.cheng@gmail.com>
7692 Chung-Ju Wu <jasonwucj@gmail.com>
7694 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
7695 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7696 Add enum values for DSP extension instructions.
7697 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
7699 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
7700 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
7702 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
7703 * config/nds32/nds32-dspext.md: New file for DSP implementation.
7704 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
7705 * config/nds32/nds32-intrinsic.md: Likewise.
7706 * config/nds32/nds32_intrinsic.h: Likewise.
7707 * config/nds32/nds32-md-auxiliary.c: Likewise.
7708 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
7709 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
7710 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
7711 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
7712 * config/nds32/nds32-protos.h: New declarations for DSP extension.
7713 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
7714 TYPE_DMAC in switch statement.
7715 * config/nds32/nds32.c: New checking and implementation for DSP
7716 extension instructions.
7717 * config/nds32/nds32.h: Likewise.
7718 * config/nds32/nds32.md: Likewise.
7719 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
7720 * config/nds32/predicates.md: Implement new predicates for DSP
7723 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7725 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
7726 Reformat alternatives and attributes so it is easier to identify
7727 which constraints/attributes go with which instruction.
7728 (mov<mode>_hardfloat32, FMOVE64): Likewise.
7729 (mov<mode>_softfloat32, FMOVE64): Likewise.
7730 (mov<mode>_hardfloat64, FMOVE64): Likewise.
7731 (mov<mode>_softfloat64, FMOVE64): Likewise.
7733 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
7735 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7737 (Basic PowerPC Built-in Functions): The new name of the
7738 subsection previously known as "PowerPC Built-in Functions".
7739 (Basic PowerPC Built-in Functions Available on all Configurations):
7741 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
7742 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
7743 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
7744 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
7746 2018-05-11 Martin Jambor <mjambor@suse.cz>
7749 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
7752 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7755 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
7757 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
7759 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
7760 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
7761 (ix86_handle_option): Handle -mwaitpkg.
7762 * config.gcc: New header.
7763 * config/i386/cpuid.h (bit_WAITPKG): New bit.
7764 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
7765 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
7767 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7768 OPTION_MASK_ISA_WAITPKG.
7769 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
7770 (ix86_option_override_internal): Add PTA_WAITPKG.
7771 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
7772 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
7773 IX86_BUILTIN_TPAUSE.
7774 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
7775 __builtin_ia32_umwait and __builtin_ia32_tpause.
7776 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
7777 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
7778 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
7779 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
7780 UNSPECV_TPAUSE): New.
7781 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
7782 * config/i386/i386.opt: Add -mwaitpkg.
7783 * config/i386/waitpkgintrin.h: New file.
7784 * config/i386/x86intrin.h: New header.
7785 * doc/invoke.texi: Add -mwaitpkg.
7787 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7790 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
7792 (cortex-m0): Use armv6s-m isa.
7793 (cortex-m0plus): Likewise.
7794 (cortex-m1): Likewise.
7795 (cortex-m0.small-multiply): Likewise.
7796 (cortex-m0plus.small-multiply): Likewise.
7797 (cortex-m1.small-multiply): Likewise.
7799 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
7800 Jakub Jelinek <jakub@redhat.com>
7802 PR tree-optimization/85692
7803 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
7804 source permute as well.
7806 2018-05-11 Martin Liska <mliska@suse.cz>
7809 * doc/extend.texi: Document LLVM style format for no_sanitize
7812 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7814 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
7815 mode_supports_vsx_dform_quad to mode_supports_dq_form.
7816 (mode_supports_vsx_dform_quad): Likewise.
7817 (mode_supports_vmx_dform): Move these functions to be next to the
7818 other mode_supports functions.
7819 (mode_supports_dq_form): Likewise.
7820 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
7821 mode_supports_dq_form.
7822 (reg_offset_addressing_ok_p): Likewise.
7823 (offsettable_ok_by_alignment): Likewise.
7824 (rs6000_legitimate_offset_address_p): Likewise.
7825 (legitimate_lo_sum_address_p): Likewise.
7826 (rs6000_legitimize_address): Likewise.
7827 (rs6000_legitimize_reload_address): Likewise.
7828 (rs6000_secondary_reload_inner): Likewise.
7829 (rs6000_preferred_reload_class): Likewise.
7830 (rs6000_output_move_128bit): Likewise.
7832 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7834 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
7835 Generate SImode target register for null target.
7836 <case IX86_BUILTIN_XGETBV>: Ditto.
7837 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
7838 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
7840 2018-05-10 Carl Love <cel@us.ibm.com>
7842 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
7843 dcbtt and dcbtstt if operands[2] is 0.
7845 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7848 * config/i386/sse.md (usadv64qi): New expander.
7850 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
7852 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
7853 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
7854 -maltivec=be support.
7855 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
7856 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
7857 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
7858 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
7859 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
7860 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
7861 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
7862 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
7863 altivec_vsumsws): Adjust.
7864 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
7865 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
7866 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
7868 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
7869 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
7870 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
7871 (altivec_lve<VI_char>x): Delete expand.
7872 (*altivec_lve<VI_char>x_internal): Rename to...
7873 (altivec_lve<VI_char>x): ... this.
7874 (altivec_lvxl_<mode>): Delete expand.
7875 (*altivec_lvxl_<mode>_internal): Rename to ...
7876 (altivec_lvxl_<mode>): ... this.
7877 (altivec_stvxl_<mode>): Delete expand.
7878 (*altivec_stvxl_<mode>_internal): Rename to ...
7879 (altivec_stvxl_<mode>): ... this.
7880 (altivec_stve<VI_char>x): Delete expand.
7881 (*altivec_stve<VI_char>x_internal): Rename to ...
7882 (altivec_stve<VI_char>x): ... this.
7883 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
7884 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
7885 reduc_plus_scal_<mode>): Adjust.
7886 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
7888 (rs6000_cpu_cpp_builtins): Adjust.
7889 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
7890 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
7891 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
7892 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
7893 -maltivec=be support.
7894 (rs6000_split_vec_extract_var): Adjust.
7895 (rs6000_split_v4si_init): Adjust.
7896 (swap_selector_for_mode): Delete.
7897 (altivec_expand_lvx_be, altivec_expand_stvx_be,
7898 altivec_expand_stvex_be): Delete.
7899 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
7900 -maltivec=be support.
7901 (rs6000_gimple_fold_builtin): Ditto.
7902 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
7904 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
7905 (TARGET_DIRECT_MOVE_64BIT): Adjust.
7906 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
7907 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
7908 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
7909 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
7910 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
7911 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
7912 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
7913 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
7914 anonymous split): Adjust.
7915 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
7916 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
7918 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
7920 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
7921 when --with-gxx-include-dir is also specified.
7922 * configure: Regenerate.
7924 2018-05-09 Jim Wilson <jimw@sifive.com>
7927 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
7928 * config/riscv/t-withmultilib: New.
7929 * config/riscv/withmultilib.h: New.
7930 * doc/install.texi: Document RISC-V --with-multilib-list support.
7932 2018-05-09 Richard Biener <rguenther@suse.de>
7934 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
7936 (vect_bb_vectorization_profitable_p): Adjust. Compute
7937 actual scalar cost using the cost vector and the add_stmt_cost
7940 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7942 PR rtl-optimization/85645
7943 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
7944 in the REG_CFA_REGISTER note for LR, don't leave it empty.
7946 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7948 PR rtl-optimization/85645
7949 * shrink-wrap.c (spread_components): Return a boolean saying if
7950 anything was changed.
7951 (try_shrink_wrapping_separate): Iterate spread_components until
7952 nothing changes anymore.
7954 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7956 PR rtl-optimization/85645
7957 * regrename.c (build_def_use): Also kill the chains that include the
7958 destination of a REG_CFA_REGISTER note.
7960 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7962 PR rtl-optimization/85645
7963 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
7964 insn that has a REG_CFA_REGISTER note.
7966 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
7968 * cfgexpand.c (expand_clobber): New function.
7969 (expand_gimple_stmt_1): Use it.
7970 * tree-vect-stmts.c (vect_clobber_variable): New function,
7972 (vectorizable_simd_clone_call): ...here.
7973 (vectorizable_store): Emit a clobber either side of an
7974 IFN_STORE_LANES sequence.
7975 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
7977 2018-05-09 Tom de Vries <tom@codesourcery.com>
7980 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
7981 (define_insn "trap_if_false"): Add exit after trap.
7983 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
7985 PR rtl-optimization/85638
7986 * bb-reorder.c: Include common/common-target.h.
7987 (create_forwarder_block): New function extracted from...
7988 (fix_up_crossing_landing_pad): ...here. Rename into...
7989 (dw2_fix_up_crossing_landing_pad): ...this.
7990 (sjlj_fix_up_crossing_landing_pad): New function.
7991 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
7992 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
7993 from both partitions and exit the loop after one iteration.
7995 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7998 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8000 (Basic PowerPC Built-in Functions): The new name of the
8001 subsection previously known as "PowerPC Built-in Functions".
8002 (Basic PowerPC Built-in Functions Available on all Configurations):
8004 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
8006 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
8008 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
8010 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
8013 2018-05-08 Jim Wilson <jimw@sifive.com>
8015 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
8016 (LD_EMUL_SUFFIX): New.
8017 (LINK_SPEC): Use it.
8019 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
8021 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8023 (Basic PowerPC Built-in Functions): The new name of the
8024 subsection previously known as "PowerPC Built-in Functions".
8025 (Basic PowerPC Built-in Functions Available on all Configurations):
8027 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
8029 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
8031 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
8033 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
8036 2018-05-08 Jakub Jelinek <jakub@redhat.com>
8039 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
8040 after cmpelim optimization.
8042 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
8044 * config.gcc: Support "goldmont".
8045 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
8046 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8048 * config/i386/i386.c (m_GOLDMONT): Define.
8049 (processor_target_table): Add "goldmont".
8050 (PTA_GOLDMONT): Define.
8051 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
8052 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
8053 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
8054 (fold_builtin_cpu): Add "goldmont".
8055 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
8056 (ix86_option_override_internal): Add "goldmont".
8057 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
8058 (processor_type): Add PROCESSOR_GOLDMONT.
8059 * config/i386/i386.md: Add CPU "glm".
8060 * config/i386/glm.md: New file.
8061 * config/i386/x86-tune.def: Add m_GOLDMONT.
8062 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
8064 2018-05-08 Jakub Jelinek <jakub@redhat.com>
8067 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
8069 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
8070 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
8071 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
8074 * config/i386/i386.c (ix86_fold_builtin): Handle
8075 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
8078 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
8079 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
8081 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
8084 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
8085 (check_arch): Likewise.
8086 (check_fpu): Return the result rather than printing it.
8087 (end arch): Fix operator precedence.
8088 (end cpu): Likewise.
8089 (END): Print the result from check_fpu.
8091 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
8092 Alan Hayward <alan.hayward@arm.com>
8093 David Sherwood <david.sherwood@arm.com>
8095 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
8096 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
8097 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
8098 (*fcmuo<mode>_and): New patterns.
8100 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
8102 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
8103 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
8104 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
8105 (cmp_op, sve_imm_con): New code attributes.
8106 (SVE_COND_INT_CMP, imm_con): Delete.
8107 (cmp_op): Remove above unspecs from int attribute.
8108 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
8110 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
8111 comparison-specific unspecs.
8112 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
8113 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
8114 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
8115 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
8116 (*vec_fcm<cmp_op><mode>): Rename to...
8117 (*fcm<cmp_op><mode>): ...this and adjust likewise.
8118 (*vec_fcmuo<mode>): Rename to...
8119 (*fcmuo<mode>): ...this and adjust likewise.
8120 (*pred_fcm<cmp_op><mode>): New pattern.
8121 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
8122 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
8124 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
8126 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
8127 (aarch64_emit_sve_predicated_cond): New function.
8128 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
8129 (aarch64_emit_unspec_cond_or): Replace with...
8130 (aarch64_emit_sve_or_conds): ...this new function. Use
8131 aarch64_emit_sve_ptrue_op for the individual comparisons and
8132 aarch64_emit_binop to OR them together.
8133 (aarch64_emit_inverted_unspec_cond): Replace with...
8134 (aarch64_emit_sve_inverted_cond): ...this new function. Use
8135 aarch64_emit_sve_ptrue_op for the comparison and
8136 aarch64_emit_unop to invert the result.
8137 (aarch64_expand_sve_vec_cmp_float): Update after the above
8138 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
8140 2018-05-07 Nathan Sidwell <nathan@acm.org>
8142 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
8143 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
8144 (Backwards Compatibility): Likewise.
8146 2018-05-07 Luis Machado <luis.machado@linaro.org>
8150 2018-05-07 Luis Machado <luis.machado@linaro.org>
8152 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8153 <prefetch_dynamic_strides>: New const bool field.
8154 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8155 prefetch_dynamic_strides.
8156 (exynosm1_prefetch_tune): Likewise.
8157 (thunderxt88_prefetch_tune): Likewise.
8158 (thunderx_prefetch_tune): Likewise.
8159 (thunderx2t99_prefetch_tune): Likewise.
8160 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
8162 (aarch64_override_options_internal): Update to set
8163 PARAM_PREFETCH_DYNAMIC_STRIDES.
8164 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
8165 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
8166 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
8167 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
8168 prefetch-dynamic-strides setting.
8170 2018-05-07 Luis Machado <luis.machado@linaro.org>
8172 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8173 <minimum_stride>: New const int field.
8174 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8175 minimum_stride field.
8176 (exynosm1_prefetch_tune): Likewise.
8177 (thunderxt88_prefetch_tune): Likewise.
8178 (thunderx_prefetch_tune): Likewise.
8179 (thunderx2t99_prefetch_tune): Likewise.
8180 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
8181 (aarch64_override_options_internal): Update to set
8182 PARAM_PREFETCH_MINIMUM_STRIDE.
8183 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
8184 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
8185 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
8186 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
8187 stride is constant and is below the minimum stride threshold.
8189 2018-05-07 Luis Machado <luis.machado@linaro.org>
8191 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
8194 2018-05-07 Luis Machado <luis.machado@linaro.org>
8196 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8197 <prefetch_dynamic_strides>: New const bool field.
8198 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8199 prefetch_dynamic_strides.
8200 (exynosm1_prefetch_tune): Likewise.
8201 (thunderxt88_prefetch_tune): Likewise.
8202 (thunderx_prefetch_tune): Likewise.
8203 (thunderx2t99_prefetch_tune): Likewise.
8204 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
8206 (aarch64_override_options_internal): Update to set
8207 PARAM_PREFETCH_DYNAMIC_STRIDES.
8208 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
8209 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
8210 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
8211 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
8212 prefetch-dynamic-strides setting.
8214 2018-05-07 Luis Machado <luis.machado@linaro.org>
8216 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8217 <minimum_stride>: New const int field.
8218 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8219 minimum_stride field.
8220 (exynosm1_prefetch_tune): Likewise.
8221 (thunderxt88_prefetch_tune): Likewise.
8222 (thunderx_prefetch_tune): Likewise.
8223 (thunderx2t99_prefetch_tune): Likewise.
8224 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
8225 (aarch64_override_options_internal): Update to set
8226 PARAM_PREFETCH_MINIMUM_STRIDE.
8227 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
8228 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
8229 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
8230 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
8231 stride is constant and is below the minimum stride threshold.
8233 2018-05-06 Jakub Jelinek <jakub@redhat.com>
8236 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
8237 the type is addressable. Don't force op into register if it has
8240 2018-05-05 Roland McGrath <mcgrathr@google.com>
8243 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
8244 any section for which we don't know a specific type it should have,
8245 regardless of name. Previously this was done only for the exact
8246 names ".init_array", ".fini_array", and ".preinit_array".
8247 (default_elf_asm_named_section): Add comment about
8248 relationship with default_section_type_flags and SECTION_NOTYPE.
8249 (get_section): Don't consider it a type conflict if one side has
8250 SECTION_NOTYPE and the other doesn't, as long as neither has the
8251 SECTION_BSS et al used in the default_section_type_flags logic.
8253 2018-05-05 Tom de Vries <tom@codesourcery.com>
8256 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
8257 (workaround_barsyncs): New function.
8258 (nvptx_reorg): Use workaround_barsyncs.
8259 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
8260 (define_expand "nvptx_membar_cta"): New define_expand.
8261 (define_insn "*nvptx_membar_cta"): New insn.
8263 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
8265 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
8266 To improve optimization opportunities.
8267 * builtin-types.def: The new needed builtin types for the above.
8269 2018-05-04 Richard Biener <rguenther@suse.de>
8271 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
8272 * gimple-ssa-store-merging.c
8273 (imm_store_chain_info::output_merged_store): Remove redundant create,
8274 release split_store vector contents on failure.
8275 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
8276 scalar stmt vector on cache hit.
8278 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
8280 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
8282 * config.gcc (powerpc-xilinx-eabi*): Remove.
8283 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
8285 (fusion_addis_mem_combo_load): Ditto.
8286 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
8288 (rs6000_cpu_cpp_builtins): Ditto.
8289 * config/rs6000/rs6000-linux.c
8290 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
8291 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
8292 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
8294 (rs6000_setup_reg_addr_masks): Ditto.
8295 (rs6000_init_hard_regno_mode_ok): Ditto.
8296 (rs6000_option_override_internal): Ditto.
8297 (legitimate_lo_sum_address_p): Ditto.
8298 (rs6000_legitimize_address): Ditto.
8299 (rs6000_legitimize_reload_address): Ditto.
8300 (rs6000_legitimate_address_p): Ditto.
8301 (abi_v4_pass_in_fpr): Ditto.
8302 (setup_incoming_varargs): Ditto.
8303 (rs6000_gimplify_va_arg): Ditto.
8304 (rs6000_split_multireg_move): Ditto.
8305 (rs6000_savres_strategy): Ditto.
8306 (rs6000_emit_prologue_components): Ditto.
8307 (rs6000_emit_epilogue_components): Ditto.
8308 (rs6000_emit_prologue): Ditto.
8309 (rs6000_emit_epilogue): Ditto.
8310 (rs6000_elf_file_end): Ditto.
8311 (rs6000_function_value): Ditto.
8312 (rs6000_libcall_value): Ditto.
8313 * config/rs6000/rs6000.h: Ditto.
8314 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
8315 (TARGET_MINMAX): ... this. New.
8316 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
8317 * config/rs6000/rs6000.md: Remove Xilinx FP support.
8318 (*movsi_internal1_single): Delete.
8319 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
8320 mfpu=, mxilinx-fpu): Delete.
8321 * config/rs6000/singlefp.h: Delete.
8322 * config/rs6000/sysv4.h: Remove Xilinx FP support.
8323 * config/rs6000/t-rs6000: Ditto.
8324 * config/rs6000/t-xilinx: Delete.
8325 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
8326 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
8327 (VStype_simple): Delete.
8328 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
8329 * config/rs6000/xfpu.h: Delete.
8330 * config/rs6000/xfpu.md: Delete.
8331 * config/rs6000/xilinx.h: Delete.
8332 * config/rs6000/xilinx.opt: Delete.
8333 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
8334 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
8336 2018-05-04 Tom de Vries <tom@codesourcery.com>
8339 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
8342 2018-05-04 Richard Biener <rguenther@suse.de>
8345 * tree-complex.c (update_complex_assignment): We are always in SSA form.
8346 (expand_complex_div_wide): Likewise.
8347 (expand_complex_operations_1): Likewise.
8348 (expand_complex_libcall): Preserve EH info of the original stmt.
8349 (tree_lower_complex): Handle removed blocks.
8350 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
8351 on complex multiplication and division libcall builtins.
8353 2018-05-04 Richard Biener <rguenther@suse.de>
8356 * fold-const.c (negate_expr_p): Restrict negation of operand
8357 zero of a division to when we know that can happen without
8359 (fold_negate_expr_1): Likewise.
8361 2018-05-04 Jakub Jelinek <jakub@redhat.com>
8364 * real.h (real_nextafter): Declare.
8365 * real.c (real_nextafter): New function.
8366 * fold-const-call.c (fold_const_nextafter): New function.
8367 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
8368 CASE_CFN_NEXTTOWARD.
8369 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
8370 even when arg1_mode is different from arg0_mode.
8372 2018-05-03 Nathan Sidwell <nathan@acm.org>
8374 * doc/extend.texi (Deprecated Features): Remove
8376 (Backwards Compatibility): Likewise.
8377 * doc/invoke.texi (C++ Language Options): Likewise.
8378 (C++ Dialect Options): Likewise.
8380 2018-05-03 Jakub Jelinek <jakub@redhat.com>
8383 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
8384 _mm512_mask_mullox_epi64): New intrinsics.
8386 2018-05-03 Tom de Vries <tom@codesourcery.com>
8389 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8390 dump files): Add offload-tree.
8392 2018-05-03 Richard Biener <rguenther@suse.de>
8394 PR tree-optimization/85615
8395 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
8396 to loops not nested in BBs loop father to avoid creating multi-entry
8399 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8401 PR tree-optimization/70291
8402 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
8403 arguments. Change return type to tree. Emit libcall as a new
8404 statement rather than replacing existing one when inplace_p is true.
8405 (expand_complex_multiplication_components): New function.
8406 (expand_complex_multiplication): Expand floating-point complex
8407 multiplication using the above.
8408 (expand_complex_division): Rename inner_type parameter to type.
8409 Update expand_complex_libcall call-site.
8410 (expand_complex_operations_1): Update expand_complex_multiplication
8411 and expand_complex_division call-sites.
8413 2018-05-02 Jakub Jelinek <jakub@redhat.com>
8416 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
8417 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
8418 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
8419 the highest significant bit of the shift count mask is clear. In
8420 check whether and[sq]i3 is needed verify that all significant bits
8421 of the shift count other than the highest are set.
8423 2018-05-02 Tom de Vries <tom@codesourcery.com>
8426 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
8427 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
8428 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
8429 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
8430 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
8431 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
8432 __builtin_goacc_parlevel_size.
8434 2018-05-02 Richard Biener <rguenther@suse.de>
8436 PR tree-optimization/85597
8437 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
8438 do not use split vect_get_vec_defs call but call vect_get_slp_defs
8441 2018-05-02 Tom de Vries <tom@codesourcery.com>
8444 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8445 dump files): Add ltrans-tree.
8447 2018-05-02 Tom de Vries <tom@codesourcery.com>
8450 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8451 dump files): Add wpa-ipa.
8453 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
8455 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
8456 powerpc*-*-linux*paired* target.
8457 * config/rs6000/750cl.h: Delete.
8458 * config/rs6000/paired.h: Delete.
8459 * config/rs6000/paired.md: Delete.
8460 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
8462 * config/rs6000/rs6000-builtin.def: Remove paired float support.
8463 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
8464 comment. Remove paired float support.
8465 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
8466 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
8468 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
8469 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
8471 * config/rs6000/rs6000.c: Remove paired float support.
8472 (paired_expand_vector_init, paired_expand_vector_move,
8473 paired_emit_vector_compare, paired_emit_vector_cond_expr,
8474 (paired_expand_lv_builtin, paired_expand_stv_builtin,
8475 paired_expand_builtin, paired_expand_predicate_builtin,
8476 paired_init_builtins): Delete.
8477 * config/rs6000/rs6000.h: Remove paired float support.
8478 * config/rs6000/rs6000.md: Remove paired float support.
8479 (move_from_CR_ov_bit): Delete.
8480 * config/rs6000/rs6000.opt (mpaired): Delete.
8481 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
8482 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
8484 2018-05-02 Richard Biener <rguenther@suse.de>
8487 * gimplify.c (gimplify_save_expr): When in SSA form allow
8488 SAVE_EXPRs to compute to SSA vars.
8490 2018-05-02 Jakub Jelinek <jakub@redhat.com>
8493 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
8494 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
8495 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
8496 clobber operands[2], instead use a new pseudo. Formatting fixes.
8498 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
8500 PR tree-optimization/85586
8501 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
8502 exit early for statements in the same group if the accesses are
8505 2018-05-02 Tom de Vries <tom@codesourcery.com>
8508 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
8511 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
8513 PR tree-optimization/85143
8514 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
8516 2018-05-01 Tom de Vries <tom@codesourcery.com>
8519 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
8520 not found" error message.
8522 2018-05-01 Tom de Vries <tom@codesourcery.com>
8525 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
8526 * vec.c (test_ordered_remove_if): New function.
8527 (vec_c_tests): Call test_ordered_remove_if.
8528 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
8529 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
8530 * tree-vect-patterns.c (vect_pattern_recog_1): Use
8531 VEC_ORDERED_REMOVE_IF.
8533 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8535 PR tree-optimization/82665
8536 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
8537 pointer subtraction where arguments come from a memchr call.
8539 2018-05-01 Jakub Jelinek <jakub@redhat.com>
8541 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
8542 --push-state --as-needed and --pop-state instead of --as-needed and
8543 --no-as-needed if ld supports it.
8544 * configure: Regenerated.
8547 * doc/install.texi2html: Replace _002d with - and _002a with * in
8548 generated html files using sed.
8550 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8553 * gcc-rich-location.c (blank_line_before_p): New function.
8554 (use_new_line): New function.
8555 (gcc_rich_location::add_fixit_insert_formatted): New function.
8556 * gcc-rich-location.h
8557 (gcc_rich_location::add_fixit_insert_formatted): New function.
8559 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8561 * selftest.c (assert_streq): Rename "expected" and "actual" to
8562 "val1" and "val2". Extend NULL-handling to cover both inputs
8563 symmetrically, while still requiring both to be non-NULL for a pass.
8564 * selftest.h (assert_streq): Rename "expected" and "actual" to
8566 (ASSERT_EQ): Likewise.
8567 (ASSERT_EQ_AT): Likewise.
8568 (ASSERT_KNOWN_EQ): Likewise.
8569 (ASSERT_KNOWN_EQ_AT): Likewise.
8570 (ASSERT_NE): Likewise.
8571 (ASSERT_MAYBE_NE): Likewise.
8572 (ASSERT_MAYBE_NE_AT): Likewise.
8573 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
8574 the assertion to pass.
8575 (ASSERT_STREQ_AT): Likewise.
8577 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
8579 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
8580 interaction with -pie.
8582 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8584 * selftest.h: Fix alphabetization of per-source-file selftest
8587 2018-04-30 Jason Merrill <jason@redhat.com>
8589 PR c++/61982 - dead stores to destroyed objects.
8590 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
8593 2018-04-30 Jason Merrill <jason@redhat.com>
8595 * tree.c (build_clobber): New.
8596 * tree.h: Declare it.
8597 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
8599 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8601 * diagnostic-show-locus.c (layout::layout): Update for
8602 location_get_source_line returning a char_span.
8603 (struct char_span): Move to input.h.
8604 (struct correction): Update for fields in char_span becoming
8606 (struct source_line): Update for location_get_source_line
8607 returning a char_span.
8608 (layout::print_line): Likewise.
8609 * edit-context.c (edited_file::print_content): Likewise.
8610 (edited_file::print_diff_hunk): Likewise.
8611 (edited_file::print_run_of_changed_lines): Likewise.
8612 (edited_file::get_num_lines): Likewise.
8613 (edited_line::edited_line): Likewise.
8614 * final.c (asm_show_source): Likewise.
8615 * input.c (location_get_source_line): Convert return type
8616 from const char * to char_span, losing the final "line_len"
8618 (dump_location_info): Update for the above.
8619 (get_substring_ranges_for_loc): Likewise. Use a char_span
8620 when handling the literal within the line.
8621 (test_reading_source_line): Update for location_get_source_line
8622 returning a char_span.
8623 * input.h (class char_span): Move here from
8624 diagnostic-show-locus.c, converting from a struct to a class.
8625 Make data members private.
8626 (char_span::operator bool): New.
8627 (char_span::length): New.
8628 (char_span::get_buffer): New.
8629 (char_span::operator[]): New.
8630 (char_span::subspan): Make const.
8631 (char_span::xstrdup): New.
8632 (location_get_source_line): Convert return type from const char *
8633 to char_span, losing the final "line_size" param.
8635 2018-04-30 Jan Hubicka <jh@suse.cz>
8637 * lto-wrapper.c (ltrans_priorities): New static var.
8638 (cmp_priority): New.
8639 (run_gcc): Read priorities and if doing parallel build order
8640 the Makefile by them.
8642 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8644 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
8646 2018-04-30 Richard Biener <rguenther@suse.de>
8648 * tree-cfg.c (verify_address): Remove base argument, add
8649 flag whether to check TREE_ADDRESSABLE and do that.
8650 (verify_expr): Remove.
8651 (verify_types_in_gimple_reference): Add pieces from verify_expr.
8652 (verify_gimple_assign_single): Likewise.
8653 (verify_gimple_switch): Likewise.
8654 (verify_expr_location_1): Dereference tp once. Add (disabled)
8655 piece from verify_expr.
8656 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
8658 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8660 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
8662 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8664 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
8665 (small_data_pattern): Likewise.
8666 (arc_rewrite_small_data): Likewise.
8667 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
8668 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
8669 (get_symbol_alignment): New function.
8670 (legitimate_small_data_address_p): Likewise.
8671 (legitimate_scaled_address): Update, call
8672 legitimate_small_data_address_p.
8673 (output_sdata): New static variable.
8674 (arc_print_operand): Update how we handle small data operands.
8675 (arc_print_operand_address): Likewise.
8676 (arc_legitimate_address_p): Update, use
8677 legitimate_small_data_address_p.
8678 (arc_rewrite_small_data_p): Remove.
8679 (arc_rewrite_small_data_1): Likewise.
8680 (arc_rewrite_small_data): Likewise.
8681 (small_data_pattern): Likewise.
8682 (compact_sda_memory_operand): Update to use
8683 legitimate_small_data_address_p and get_symbol_alignment.
8684 (prepare_move_operands): Don't rewite sdata pattern.
8685 (prepare_extend_operands): Remove.
8686 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
8688 (zero_extendqisi2): Likewise.
8689 (zero_extendhisi2): Likewise.
8690 (extendqihi2): Likewise.
8691 (extendqisi2): Likewise.
8692 (extendhisi2): Likewise.
8696 * config/arc/constraints.md (Usd): Change it to memory constraint.
8698 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8700 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
8701 as source of std instructions.
8702 * config/arc/arc.md (movsi_insn): Update pattern predicate to
8703 allow 6-bit constants as source for store instructions.
8704 (movdi_insn): Update instruction pattern to allow 6-bit constants
8705 as source for store instructions.
8707 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
8709 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
8711 2018-04-30 Nathan Sidwell <nathan@acm.org>
8712 Sandra Loosemore <sandra@codesourcery.com>
8714 * dumpfile.c (dump_open): Allow '-' for stdout.
8715 * doc/invoke.texi (Developer Options): Document dump filename
8716 determination early. Document stdin/stdout selection.
8718 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
8720 Microblaze Target: PIC data text relative
8722 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
8723 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
8725 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
8726 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
8727 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
8728 New addressing mode for data-text relative position indepenedent code.
8729 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
8730 'ADDRESS_SYMBOLIC_TXT_REL'.
8731 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
8732 (microblaze_legitimate_pic_operand): Exclude function calls from
8733 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
8734 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
8736 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8737 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8738 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
8739 for 'address + offset'.
8740 (microblaze_expand_prologue): Add new function prologue call for
8742 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
8743 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
8744 table in case of TARGET_PIC_DATA_TEXT_REL.
8745 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
8746 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
8747 Add new macros 'UNSPEC_TEXT',
8748 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
8749 + exclude function calls from 'UNSPEC_PLT' in case of data text
8751 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8752 new target hook for generating address diff vector tables in case of
8754 * doc/tm.texi : Regenerate.
8755 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
8756 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
8757 of addr diff vector generation.
8758 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8759 target hook definition.
8760 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
8761 Add default function for generate_pic_addr_diff_vec -> flag_pic.
8762 * doc/invoke.texi (Add new pic option): Add new microblaze pic
8763 option for data text relative.
8765 2018-04-30 Richard Biener <rguenther@suse.de>
8767 * tree-chrec.h (evolution_function_is_constant_p): Remove
8769 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
8771 2018-04-30 Richard Biener <rguenther@suse.de>
8774 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
8776 2018-04-30 Richard Biener <rguenther@suse.de>
8778 PR tree-optimization/28364
8779 PR tree-optimization/85275
8780 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
8781 copying first exit test.
8783 2018-04-28 Mark Wielaard <mark@klomp.org>
8785 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
8787 (dwarf_AT): Handle DW_AT_addr_base.
8788 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
8790 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
8793 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
8794 (*ashl<dwi>3_doubleword_mask_1): Ditto.
8795 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
8796 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
8798 2018-04-28 Richard Biener <rguenther@suse.de>
8800 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
8801 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
8802 to reflect use. Only add interesting stmts.
8804 2018-04-27 Martin Jambor <mjambor@suse.cz>
8807 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
8808 the jump function allows for passing through aggregate values.
8810 2018-04-27 David Malcolm <dmalcolm@redhat.com>
8812 * input.h (in_system_header_at): Convert from macro to inline
8814 (from_macro_expansion_at): Likewise.
8815 (from_macro_definition_at): Likewise.
8817 2018-04-27 Jeff Law <law@redhat.com>
8819 * config.gcc: Mark tile* targets as deprecated/obsolete.
8821 2018-04-27 Richard Biener <rguenther@suse.de>
8823 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
8826 2018-04-27 Richard Biener <rguenther@suse.de>
8828 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
8830 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
8832 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
8833 with Yd constraint. Set "preferred_for_speed" attribute from
8834 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
8836 (*movdi_internal): Ditto.
8837 (movti_interunit splitters): Remove
8838 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
8839 (movdi_interunit splitters): Ditto.
8840 * config/i386/constraints.md (Ye): Remove.
8841 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
8843 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8846 * config/aarch64/constraints.md (Usg): Limit to 31.
8849 2018-04-27 Jakub Jelinek <jakub@redhat.com>
8851 PR tree-optimization/85529
8852 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
8853 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
8854 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
8855 zero extension or masking of the MSB bit.
8856 (optimize_range_tests): Add FIRST_BB argument, pass it through
8857 to optimize_range_tests_var_bound.
8858 (maybe_optimize_range_tests, reassociate_bb): Adjust
8859 optimize_range_tests callers.
8861 2018-04-26 Richard Biener <rguenther@suse.de>
8862 Jakub Jelinek <jakub@redhat.com>
8864 * cgraph.h (symbol_table): Just declare debug method here.
8865 * symtab.c (symbol_table::debug): Define.
8867 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
8869 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
8871 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
8873 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
8874 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
8875 (*movdi_internal): Substitute Yi and Yj constraint with x
8876 and Ym and Yn constraint with y constraint. Update "isa"
8877 attribute and set "preferred_for_speed" attribute from
8878 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
8879 (*movsi_internal): Ditto.
8880 (*movdf_internal): Ditto.
8881 (*movsf_internal): Ditto.
8882 (*zero_extendsidi2): Ditto.
8883 * config/i386/sse.md (vec_set<mode>_0): Ditto.
8884 (sse2_loadld): Ditto.
8885 (*vec_extract<ssevecmodelower>_0): Ditto.
8886 (*vec_extractv4si_0_zext_sse4): Ditto.
8887 (vec_concatv2di): Ditto.
8888 (*vec_dup<mode>): Ditto.
8889 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
8890 * config/i386/constraints.md (Yi): Remove.
8895 2018-04-26 Nathan Sidwell <nathan@acm.org>
8897 * dumpfile.c (dump_open): New.
8898 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
8899 (dump_finish): Detect stdio/stderr by value not name.
8901 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
8903 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
8905 2018-04-26 Tom de Vries <tom@codesourcery.com>
8908 * config/nvptx/nvptx.c (verify_neutering_jumps)
8909 (verify_neutering_labels): New function
8910 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
8912 2018-04-26 Tom de Vries <tom@codesourcery.com>
8915 * config/nvptx/nvptx.c (needs_neutering_p): New function.
8916 (nvptx_single): Use needs_neutering_p to skip over insns that do not
8919 2018-04-26 Richard Biener <rguenther@suse.de>
8920 Tom de Vries <tom@codesourcery.com>
8923 * lto-streamer-out.c (output_function): Fixup loops if required to match
8924 discovery done in the reader.
8926 2018-04-26 Richard Biener <rguenther@suse.de>
8928 PR tree-optimization/85116
8929 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
8930 have a loop exit from the single latch predecessor. Remove
8931 case of header with just condition.
8932 (ch_base::copy_headers): Exclude infinite loops from any
8934 (pass_ch::execute): Record exits.
8936 2018-04-26 Richard Biener <rguenther@suse.de>
8938 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
8939 prologue cost vector and pass it to vect_get_load_cost.
8940 (vect_get_peeling_costs_all_drs): Likewise.
8941 (vect_peeling_hash_get_lowest_cost): Likewise.
8942 (vect_enhance_data_refs_alignment): Likewise.
8944 2018-04-26 Richard Biener <rguenther@suse.de>
8947 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
8948 checking of integer<->pointer conversions.
8949 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
8950 sign-/zero-extending pointer types.
8951 (expand_omp_for_static_chunk): Likewise.
8953 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
8954 Jean Lee <xiaoyur347@gmail.com>
8956 * config/mips/mips.c (mips_asan_shadow_offset): New function.
8957 (TARGET_ASAN_SHADOW_OFFSET): Define.
8958 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
8959 true for -fsanitize=address.
8961 2018-04-25 Mark Wielaard <mark@klomp.org>
8963 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
8966 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8968 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
8969 than "alu", remove explicit "memory" and "imm_disp" attributes.
8970 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
8973 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
8974 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
8975 gen_lowpart_no_emit.
8977 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
8980 * config/i386/i386.c (ix86_expand_builtin): Change memory
8981 operand to XI, extend p0 to Pmode.
8982 * config/i386/i386.md: Change unspec volatile and operand
8983 1 mode to XI, change operand 0 mode to P.
8985 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8987 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
8988 GET_MODE_MASK before any checking.
8989 (nds32_can_use_bset_p): Likewise.
8990 (nds32_can_use_btgl_p): Likewise.
8992 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8994 * config/nds32/nds32-doubleword.md: New define_split pattern for
8995 illegal register number.
8997 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8999 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
9001 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9003 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
9005 2018-04-25 Richard Biener <rguenther@suse.de>
9007 * lto-streamer.h (LTO_major_version): Bump to 8.
9009 2018-04-25 Jakub Jelinek <jakub@redhat.com>
9011 * BASE-VER: Set to 9.0.0.
9013 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
9015 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
9016 in __abskf2 and __powikf2.
9018 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9021 * config/aarch64/constraints.md (Usg, Usj): New constraints.
9022 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
9023 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
9024 Use the above on operand 2. Reindent.
9025 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
9027 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
9030 * common/config/i386/i386-common.c (ix86_handle_option): Don't
9032 * config/i386/i386.opt (mcet): Removed.
9033 * doc/install.texi: Remove -mcet documentation.
9034 * doc/invoke.texi: Likewise.
9036 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
9039 * doc/install.texi: Remove -mcet from bootstrap-cet.
9041 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9044 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
9045 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
9049 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
9050 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
9051 containing a CONST_VECTOR.
9053 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
9055 * doc/install.texi: Update newlib dependency for nvptx.
9057 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9060 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
9061 instead of INTVAL when shifting x left.
9063 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
9065 PR tree-optimization/85478
9066 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
9067 vect_grouped_store_supported for single element vectors.
9069 2018-04-24 Richard Biener <rguenther@suse.de>
9072 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
9073 load cost increase to the case of non-constant step.
9075 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9078 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
9079 destination if any_malformed_asm.
9081 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
9084 * expr.c (store_field): In the bitfield case, if the value comes from
9085 a function call and is returned in registers by means of a PARALLEL,
9086 do not change the mode of the temporary unless BLKmode and VOIDmode.
9088 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
9090 PR rtl-optimization/85423
9091 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
9092 dependencies to debug insns when the previous insn is non-debug.
9094 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
9096 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
9097 enums into a single definition.
9098 (fls): Fix predicates and printing.
9101 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
9103 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
9104 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
9105 and short u6 immediate.
9106 (check_if_valid_sleep_operand): Remove.
9107 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
9109 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9111 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
9112 flag_always_save_lp condition.
9113 * config/nds32/nds32.opt (malways-save-lp): New option.
9115 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9117 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
9118 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
9119 * config/nds32/nds32.h
9120 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
9121 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
9123 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9125 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
9126 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
9128 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9129 Chung-Ju Wu <jasonwucj@gmail.com>
9131 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
9133 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
9134 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
9136 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9138 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
9140 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9142 * config/nds32/nds32-protos.h (nds32_data_alignment,
9143 nds32_local_alignment): Declare.
9144 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
9145 nds32_local_alignment): New functions.
9146 (TARGET_CONSTANT_ALIGNMENT): Define.
9147 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
9149 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9151 * config/nds32/nds32.c
9152 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
9153 (TARGET_MODES_TIEABLE_P): Likewise.
9155 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9157 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
9160 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
9161 Chung-Ju Wu <jasonwucj@gmail.com>
9163 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9164 for unaligned access.
9165 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9167 * config/nds32/nds32-intrinsic.md: Likewise.
9168 * config/nds32/nds32_intrinsic.h: Likewise.
9169 * config/nds32/nds32.h (nds32_builtins): Likewise.
9170 * config/nds32/nds32.opt (munaligned-access): New option.
9171 * config/nds32/nds32.c (nds32_asm_file_start): Display
9172 flag_unaligned_access status.
9174 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
9176 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
9177 -mno-relax is present.
9178 * config/riscv/linux.h (LINK_SPEC): Ditto.
9180 2018-04-20 Martin Sebor <msebor@redhat.com>
9183 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
9185 (gimple_fold_builtin_stxcpy_chk): Same.
9186 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
9188 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
9191 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
9192 __powikf2 when long double is IEEE 128-bit.
9194 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
9196 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
9197 step to make sure stack always aligned.
9199 2018-04-20 Carl Love <cel@us.ibm.com>
9202 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
9203 size check for arg0.
9205 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
9206 Tom de Vries <tom@codesourcery.com>
9209 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
9210 Emit insns for calls too.
9211 (nvptx_find_par): Always look for worker-level predecessor insn.
9212 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
9214 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
9215 (nvptx_process_pars): Propagate frames for calls.
9217 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
9220 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
9222 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
9223 (ix86_handle_option): Don't handle OPT_mibt.
9224 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
9226 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
9228 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
9229 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
9230 (ix86_target_macros): Define __CET__ with flag_cf_protection
9231 for -fcf-protection.
9232 * config/i386/i386.c (isa2_opts): Remove -mibt.
9233 * config/i386/i386.h (TARGET_IBT): Removed.
9234 (TARGET_IBT_P): Likewise.
9235 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
9236 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
9237 * config/i386/i386.opt (mcet): Update help message.
9240 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
9241 -mcet as an alias for -mshstk.
9243 2018-04-20 Richard Biener <rguenther@suse.de>
9246 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
9247 complexity by forcing a single use of the multiply operand.
9249 2018-04-20 Martin Jambor <mjambor@suse.cz>
9252 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
9253 recursion dependency to only apply to non-clones.
9255 2018-04-20 Martin Jambor <mjambor@suse.cz>
9258 * ipa-cp.c (create_specialized_node): Check that clones of
9259 self-recursive edges exist during IPA-CP.
9261 2018-04-19 Toon Moene <toon@moene.org>
9263 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
9266 2018-04-19 Jakub Jelinek <jakub@redhat.com>
9268 PR tree-optimization/85467
9269 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
9270 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
9271 VECTOR_CST element to type.
9273 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9276 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
9277 * config/i386/i386.md (builtin_setjmp_setup): Removed.
9278 (builtin_longjmp): Likewise.
9279 (save_stack_nonlocal): New pattern.
9280 (restore_stack_nonlocal): Likewise.
9282 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9285 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
9286 Replace ASM_OUTPUT_LABEL with fprintf.
9288 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9291 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
9292 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
9293 * config/i386/i386-c.c (ix86_target_macros_internal): Also
9294 define __IBT__ and __SHSTK__ for -fcf-protection.
9295 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
9297 (ix86_trampoline_init): Likewise.
9298 (x86_output_mi_thunk): Likewise.
9299 (ix86_notrack_prefixed_insn_p): Likewise.
9300 (ix86_option_override_internal): Don't disallow -fcf-protection.
9301 * config/i386/i386.md (rdssp<mode>): Also enable for
9303 (incssp<mode>): Likewise.
9304 (nop_endbr): Likewise.
9305 * config/i386/i386.opt (mcet): Change help message to built-in
9309 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
9310 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
9311 enable CET built-in functions.
9313 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
9315 * common/config/i386/i386-common.c
9316 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
9317 OPTION_MASK_ISA_MOVDIRI_UNSET,
9318 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
9319 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
9320 * config.gcc (movdirintrin.h): New header.
9321 * config/i386/cpuid.h (bit_MOVDIRI,
9322 bit_MOVDIR64B): New bits.
9323 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
9325 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
9326 (VOID, PVOID, PCVOID)): New function types.
9327 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
9328 __builtin_ia32_directstoreu_u64,
9329 __builtin_ia32_movdir64b): New builtins.
9330 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
9331 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
9333 (ix86_valid_target_attribute_inner_p): Ditto.
9334 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
9335 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
9336 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
9337 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
9338 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
9339 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
9340 (movdiri<mode>, movdir64b_<mode>): New.
9341 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
9342 * config/i386/immintrin.h: Include movdirintrin.h.
9343 * config/i386/movdirintrin.h: New file.
9344 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
9346 2018-04-19 Richard Biener <rguenther@suse.de>
9349 * cfg.c (clear_bb_flags): When loop state says we have
9350 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
9352 2018-04-19 Richard Biener <rguenther@suse.de>
9354 PR tree-optimization/84737
9355 * tree-vect-data-refs.c (vect_copy_ref_info): New function
9356 copying restrict info.
9357 (vect_setup_realignment): Use it.
9358 * tree-vectorizer.h (vect_copy_ref_info): Declare.
9359 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
9360 the first DR to all generated stores.
9361 (vectorizable_load): Likewise for loads.
9363 2018-04-19 Jakub Jelinek <jakub@redhat.com>
9365 PR tree-optimization/85446
9366 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
9367 the integral and pointer types to have the same precision.
9369 * doc/install.texi: Document --disable-cet being the default and
9372 2018-04-18 Martin Liska <mliska@suse.cz>
9374 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
9377 2018-04-18 Martin Liska <mliska@suse.cz>
9380 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9383 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9384 arguments if they are comparable.
9386 2018-04-18 Martin Liska <mliska@suse.cz>
9389 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9392 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9395 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
9398 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
9399 ENDBR after calling __morestack.
9401 2018-04-18 David Malcolm <dmalcolm@redhat.com>
9404 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
9405 by using gcc_base_ver to generate a gcc_driver_version, and use
9406 it when generating GCC_DRIVER_NAME.
9407 * configure: Regenerate.
9409 2018-04-18 Jakub Jelinek <jakub@redhat.com>
9412 * config.gcc: Obsolete powerpc*-*-*spe*.
9414 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9417 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
9418 (stabstr_D): Change type of unum from unsigned int to
9419 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
9422 2018-04-17 Jim Wilson <jimw@sifive.com>
9425 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
9426 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
9427 Set arg_pointer_offset after using pretend_args_size.
9429 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9431 PR rtl-optimization/85431
9432 * dse.c (record_store): Ignore zero width stores.
9435 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
9436 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
9437 __builtin_stack_restore rather than after it.
9438 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
9439 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
9440 argument instead of virtual_dynamic_stack_rtx.
9442 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
9444 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
9446 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
9447 Add note to error message to explain internal mapping of overloaded
9448 built-in function name to non-overloaded built-in function name.
9449 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
9452 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9455 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
9456 where the inputs overlap with the output.
9458 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9460 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
9461 (=v, v) alternative and explicit "memory" attribute.
9462 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
9463 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9465 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
9466 "sselog1" type instead of "sselog".
9467 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
9468 "sselog". Remove explicit "memory" attribute.
9469 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
9470 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9472 (vec_extract_hi_v32hi): Merge all alternatives into one, use
9473 "sselog1" type instead of "sselog". Remove explicit "memory"
9475 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
9476 use "sselog1" type instead of "sselog". Remove explicit "memory"
9478 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
9479 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9481 (vec_extract_hi_v64qi): Merge all alternatives into one, use
9482 "sselog1" type instead of "sselog". Remove explicit "memory"
9484 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
9485 use "sselog1" type instead of "sselog". Remove explicit "memory"
9489 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
9492 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
9495 2018-04-17 Martin Jambor <mjambor@suse.cz>
9498 * ipa-cp.c (create_specialized_node): Call
9499 expand_all_artificial_thunks if necessary.
9501 2018-04-17 Martin Liska <mliska@suse.cz>
9504 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
9505 in message, remote space in between '_G' and '('.
9507 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9510 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
9511 avx512f_vmcmp<mode>3<round_saeonly_name>,
9512 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
9513 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
9514 avx512f_rndscale<mode><round_saeonly_name>,
9515 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
9516 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
9517 Use %<iptr>2 instead of %2 for -masm=intel.
9518 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
9519 avx512f_vcvttss2usi<round_saeonly_name>,
9520 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
9522 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
9523 avx512f_vcvttsd2usi<round_saeonly_name>,
9524 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
9525 Use %q1 instead of %1 for -masm=intel.
9526 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
9527 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
9528 of %3 for -masm=intel.
9529 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
9531 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
9533 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
9535 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
9536 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
9538 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
9540 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
9541 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
9542 %g1 and one with %0 and %1.
9543 (avx512er_vmrcp28<mode><round_saeonly_name>,
9544 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
9546 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
9547 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
9548 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
9549 of %0 and %{%4%} for -masm=intel.
9550 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
9551 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
9552 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
9553 order of %0 and %{%5%}%{z%} for -masm=intel.
9555 2018-04-17 Jan Hubicka <jh@suse.cz>
9558 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
9560 2018-04-17 Martin Liska <mliska@suse.cz>
9563 * multiple_target.c (create_dispatcher_calls): Set apostrophes
9564 for target_clone error message. Make default implementation
9565 clone to be a local declaration.
9566 (separate_attrs): Add new argument and check for an empty
9568 (expand_target_clones): Handle it.
9569 (ipa_target_clone): Make redirection just for target_clones
9572 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
9573 Tom de Vries <tom@codesourcery.com>
9576 * omp-expand.c (expand_oacc_for): Add dummy false branch for
9577 tiled basic blocks without omp continue statements.
9579 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
9582 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
9583 vec_extract expression as having side effects to make sure it gets
9586 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
9589 * config/i386/i386.c (get_builtin_code_for_version): Check
9592 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
9595 * gcc/config.gcc: Support "skylake".
9596 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
9598 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
9599 (processor_target_table): Add "skylake".
9600 (ix86_option_override_internal): Add "skylake".
9601 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
9602 PROCESSOR_CANNONLAKE.
9603 (get_builtin_code_for_version): Fix priority for
9604 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
9605 PROCESSOR_SKYLAKE-AVX512.
9606 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
9607 (processor_type): Add PROCESSOR_SKYLAKE.
9609 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
9610 Jason Merrill <jason@redhat.com>
9613 * convert.c (convert_to_integer_1): Use direct recursion for
9614 enumeral types and types with a precision less than the number
9615 of bits in their mode.
9617 2018-04-16 Julia Koval <julia.koval@intel.com>
9620 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
9621 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
9623 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
9626 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
9627 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
9628 and -mno-direct-move.
9630 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
9633 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
9634 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
9635 Ensure negative shifts result in {0}.
9637 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
9639 PR rtl-optimization/79916
9640 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
9641 regs (if any) to define how to gnerate SD moves when LRA is in
9644 2018-04-13 Jakub Jelinek <jakub@redhat.com>
9646 PR rtl-optimization/85393
9647 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
9648 * except.c (expand_dw2_landing_pad_for_region): Make static.
9649 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
9650 a label and unconditional jump to old_bb, rather than
9651 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
9654 PR rtl-optimization/85376
9655 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
9656 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
9657 instead of a specific value.
9659 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
9660 Bin Cheng <bin.cheng@arm.com>
9662 PR tree-optimization/82965
9663 PR tree-optimization/83991
9664 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
9665 by_profile_only parameter.
9666 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
9667 information if the loop was predicted to iterate too many times.
9668 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
9670 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
9673 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
9676 2018-04-13 Martin Liska <mliska@suse.cz>
9677 Jakub Jelinek <jakub@redhat.com>
9680 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
9681 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
9682 * builtins.c (expand_builtin_memory_copy_args): Use
9683 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
9684 handle dest_addr == pc_rtx.
9686 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
9689 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
9690 asked to not generate direct moves.
9691 (fix_trunc<mode>si2_stfiwx): Similar.
9692 (fix_trunc<mode>si2_internal): Similar.
9694 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9697 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
9698 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
9699 lookup if dest in some wider mode is known to be const0_rtx and
9700 if so, record permanent equivalence for it to be ZERO_EXTEND of
9701 the narrower mode destination.
9703 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9705 * lto-streamer-out.c (output_function): Revert 259346.
9706 * omp-expand.c (expand_oacc_for): Likewise.
9708 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
9710 PR rtl-optimization/85354
9711 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
9712 * sel-sched.c (sel_global_init): ... here.
9714 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
9717 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
9718 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
9719 mode for PE-COFF targets.
9720 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
9721 (i386_pe_asm_lto_end): Likewise.
9722 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
9723 (TARGET_ASM_LTO_END): Likewise.
9724 * config/i386/winnt.c (saved_debug_info_level): New static variable.
9725 (i386_pe_asm_lto_start): New function.
9726 (i386_pe_asm_lto_end): Likewise.
9728 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9729 Richard Biener <rguenther@suse.de>
9732 * lto-streamer-out.c (output_function): Fix CFG loop state before
9734 * omp-expand.c (expand_oacc_for): Handle calls to internal
9735 functions like regular functions.
9737 2018-04-12 Richard Biener <rguenther@suse.de>
9740 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
9741 for the early LTO debug to properly generate references to it
9742 during DIE emission. Do not re-use that for the skeleton for
9744 (dwarf2out_early_finish): Likewise.
9746 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9749 * config/i386/sse.md
9750 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
9751 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
9752 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
9753 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
9754 and output is a reg, avoid creating invalid lowpart subreg, but
9755 instead split into a 512-bit move. Don't split if not AVX512VL,
9756 input is xmm16+ reg and output is a mem.
9757 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
9758 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
9759 xmm16+ reg and output is a mem.
9761 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9763 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
9764 also for flag_dwarf2_cfi_asm.
9766 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9768 PR rtl-optimization/85342
9769 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
9770 a bool scalar var inside of the loop instead. Don't try to update
9771 recog_data.operand after failed apply_change_group.
9773 2018-04-12 Tom de Vries <tom@codesourcery.com>
9776 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
9777 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
9778 array with flexible array member as array without given dimension.
9779 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
9780 argument for undefined param to true.
9782 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
9785 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9786 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
9787 from PowerPC section.
9788 * config/rs6000/sysv4.opt (mcall-): Improve help text.
9789 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
9790 help text that is too long.
9791 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
9792 help text that is too long.
9793 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
9794 help text that is too long.
9796 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
9798 * config/alpha/alpha.md (stack_probe_internal): Rename
9799 from "probe_stack". Update all callers.
9801 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9803 PR rtl-optimization/84566
9804 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
9805 sched_macro_fuse_insns.
9807 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9810 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
9811 (compute_block_dependences): ... from here.
9813 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9815 PR tree-optimization/85331
9816 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
9817 from int to HOST_WIDE_INT.
9819 2018-04-11 Martin Jambor <mjambor@suse.cz>
9822 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
9823 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
9824 not the same as the source val.
9825 (cgraph_edge_brings_value_p): New parameter.
9826 (gather_edges_for_value): Pass destination value to
9827 cgraph_edge_brings_value_p.
9828 (perhaps_add_new_callers): Likewise.
9829 (get_info_about_necessary_edges): Likewise and exclude values brought
9830 only by self-recursive edges.
9831 (create_specialized_node): Redirect only clones of self-calling edges.
9832 (+self_recursive_pass_through_p): New function.
9833 (find_more_scalar_values_for_callers_subset): Use it.
9834 (find_aggregate_values_for_callers_subset): Likewise.
9835 (known_aggs_to_agg_replacement_list): Removed.
9836 (decide_whether_version_node): Re-calculate known constants for all
9837 remaining context clones.
9839 2018-04-11 Richard Biener <rguenther@suse.de>
9842 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
9843 from early DWARF output.
9844 (dwarf2out_early_finish): Output line info unconditionally into
9845 early DWARF and add reference to it.
9847 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9850 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
9851 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
9852 other than V2DFmode using iptr mode attribute.
9853 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
9855 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9857 PR rtl-optimization/84659
9858 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
9860 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9863 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
9865 (output_loc_list): Pass address of a dummy size variable even in the
9866 locview handling loop.
9867 (index_location_lists): Add comment on why skip_loc_list_entry can't
9870 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
9873 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
9876 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
9879 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9880 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
9881 and -mstring-compare-inline-limit.
9883 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9886 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
9887 for stack clash protection in a register whenever we need it to be in
9890 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9892 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
9893 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
9895 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9898 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
9900 (mlong-double-): Ditto.
9901 * config/rs6000/sysv4.opt (msdata=): Ditto.
9902 (mtls-size=): Ditto.
9904 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9906 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
9907 erroneous entries for
9908 "vector int vec_ldl (int, long int *)", and
9909 "vector unsigned int vec_ldl (int, unsigned long int *)".
9910 Add comments and entries for
9911 "vector bool char vec_ldl (int, bool char *)",
9912 "vector bool short vec_ldl (int, bool short *)",
9913 "vector bool int vec_ldl (int, bool int *)",
9914 "vector bool long long vec_ldl (int, bool long long *)",
9915 "vector pixel vec_ldl (int, pixel *)",
9916 "vector long long vec_ldl (int, long long *)",
9917 "vector unsigned long long vec_ldl (int, unsigned long long *)".
9918 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
9919 type tree bool_long_long_type_node and correct definition of
9920 bool_V2DI_type_node to make reference to this new type tree.
9921 (rs6000_mangle_type): Replace erroneous reference to
9922 bool_long_type_node with bool_long_long_type_node.
9923 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
9924 comments to emphasize sign distinctions for char and int types and
9925 replace RS6000_BTI_bool_long constant with
9926 RS6000_BTI_bool_long_long constant. Also add comment to restrict
9927 use of RS6000_BTI_pixel.
9928 (bool_long_type_node): Remove this macro definition.
9929 (bool_long_long_type_node): New macro definition
9931 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9933 PR rtl-optimization/85300
9934 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
9935 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
9936 simplify_unary_operation fails.
9938 2018-04-10 Martin Liska <mliska@suse.cz>
9940 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
9941 cgraph_edge and ipa_ref.
9943 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9947 * config/i386/sse.md
9948 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
9949 computation of the VEC_MERGE selector from mask.
9950 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
9951 Fix decoding of the VEC_MERGE selector into mask.
9953 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9955 PR tree-optimization/85286
9956 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9958 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9960 * final.c (final_1): Set insn_last_address as well as
9961 insn_current_address.
9963 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9966 * explow.c (emit_stack_probe): Call validize_mem on memory location
9967 before passing it to gen_probe_stack. Create address operand and
9968 legitimize it for the probe_stack_address case.
9970 2018-04-09 Jan Hubicka <jh@suse.cz>
9973 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
9974 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
9975 * tree.c (free_lang_data_in_type): Fix handling of binfos;
9977 (free_lang_data): Rebuild type inheritance graph.
9979 2018-04-09 Martin Sebor <msebor@redhat.com>
9981 * invoke.texi (-finline-small-functions): Mention other optimization
9983 (-findirect-inlining, -fpartial-inlining): Same.
9984 (-finline-functions-called-once): Same.
9985 (-freorder-blocks-and-partition): Same.
9987 2018-04-09 Jan Hubicka <jh@suse.cz>
9990 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
9991 jumps; choose last target that matches the criteria (i.e.
9992 no partition changes for non-crossing jumps).
9993 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
9994 support for redirecting crossing jumps to non-crossing.
9996 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
9998 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
9999 also for naked functions.
10001 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
10003 * config/arc/arc.md (add_shift): New pattern.
10004 (add_shift2): Likewise.
10005 (sub_shift): Likewise.
10006 (sub_shift_cmp0_noout): Likewise.
10007 (compare_si_ashiftsi): Likewise.
10008 (xbfu_cmp0_noout): New combine pattern.
10009 (xbfu_cmp0"): Likewise.
10010 (movsi_set_cc_insn): Place the predicable variant first.
10011 (commutative_binary_cmp0_noout): Remove clobber.
10012 (commutative_binary_cmp0): New pattern.
10013 (noncommutative_binary_cmp0): Likewise.
10014 (noncommutative_binary_cmp0_noout): Likewise.
10015 (noncommutative_binary_comparison_result_used): Removed.
10016 (rsub_cmp0): New pattern.
10017 (rsub_cmp0_noout): Likewise.
10018 (extzvsi): Changed, keep only meaningful variants.
10019 (SQH, SEZ): New iterators.
10020 (SQH_postfix): New mode attribute.
10021 (SEZ_prefix): New code attribute.
10022 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
10023 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
10024 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
10025 of numerical value.
10026 (noncommutative_operator): Check the availability of barrel
10029 2018-04-09 Richard Biener <rguenther@suse.de>
10031 PR tree-optimization/85284
10032 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
10033 Only use the niter constraining form of simple_iv when the exit
10034 is always executed.
10036 2018-04-09 Tom de Vries <tom@codesourcery.com>
10039 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
10040 (define_expand "*memory_barrier"): New define_expand.
10041 (define_insn "memory_barrier"): New insn.
10043 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10045 PR rtl-optimization/80463
10046 PR rtl-optimization/83972
10047 PR rtl-optimization/83480
10049 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
10050 correct producer for the insn.
10051 (tidy_control_flow): Fixup seqnos in case of debug insns.
10053 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10055 PR rtl-optimization/83913
10057 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
10058 different sched-times when merging exprs.
10060 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10062 PR rtl-optimization/83962
10064 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
10065 tidy_fallthru_edge and tidy_control_flow.
10067 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10069 PR rtl-optimization/83530
10071 * sel-sched.c (force_next_insn): New global variable.
10072 (remove_insn_for_debug): When force_next_insn is true, also leave only
10073 next insn in the ready list.
10074 (sel_sched_region): When the region wasn't scheduled, make another pass
10075 over it with force_next_insn set to 1.
10077 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
10079 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
10081 * config/nds32/constants.md (unspec_volatile_element): Add enum values
10082 for interrupt control.
10083 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
10084 functions for interrupt control.
10085 * config/nds32/nds32-intrinsic.md: Likewise.
10086 * config/nds32/nds32_intrinsic.h: Likewise.
10087 * config/nds32/nds32.h (nds32_builtins): Likewise.
10089 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
10091 * config/nds32/nds32.c (nds32_init_machine_status,
10092 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
10093 strict_aligned_p field.
10094 (nds32_expand_to_rtl_hook): New function.
10095 (TARGET_EXPAND_TO_RTL_HOOK): Define.
10096 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
10098 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10099 Chung-Ju Wu <jasonwucj@gmail.com>
10101 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
10102 * config/nds32/nds32-n7.md: New file.
10103 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
10104 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
10106 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
10107 * config/nds32/nds32.md (pipeline_model): Add n7.
10108 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
10109 * config/nds32/pipelines.md: Include n7 settings.
10111 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10112 Chung-Ju Wu <jasonwucj@gmail.com>
10114 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
10115 * config/nds32/nds32-e8.md: New file.
10116 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
10117 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
10119 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
10120 * config/nds32/nds32.md (pipeline_model): Add e8.
10121 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
10122 * config/nds32/pipelines.md: Include e8 settings.
10124 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10125 Chung-Ju Wu <jasonwucj@gmail.com>
10127 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
10128 * config/nds32/nds32-n8.md: New file.
10129 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
10130 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
10132 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
10133 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
10134 * config/nds32/nds32.md (pipeline_model): Add n8.
10135 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
10136 * config/nds32/pipelines.md: Include n8 settings.
10138 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10139 Chung-Ju Wu <jasonwucj@gmail.com>
10141 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
10142 * config/nds32/nds32-n9-2r1w.md: New file.
10143 * config/nds32/nds32-n9-3r2w.md: New file.
10144 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
10145 nds32_register_ports): New or modify for cpu n9.
10146 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
10148 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
10149 * config/nds32/nds32-utils.c: New file.
10150 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
10151 TARGET_MUL_SLOW): Define.
10152 * config/nds32/nds32.md (pipeline_model): New attribute.
10153 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
10154 New options that support cpu n9.
10155 * config/nds32/pipelines.md: Include n9 settings.
10156 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
10158 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
10160 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
10161 information if necessary.
10162 (output_cond_branch_compare_zero): Likewise.
10163 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
10164 (nds32_target_alignment): Refine for alignment.
10165 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
10166 (FUNCTION_BOUNDARY): Modify.
10167 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
10169 * config/nds32/nds32.opt (malways-align, malign-functions): New.
10171 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
10173 * config/nds32/constants.md (unspec_volatile_element): Add values for
10174 TLB operation and data prefetch.
10175 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
10176 functions for TLB operation and data prefetch.
10177 * config/nds32/nds32-intrinsic.md: Likewise.
10178 * config/nds32/nds32_intrinsic.h: Likewise.
10179 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
10180 (nds32_print_operand): Likewise.
10181 * config/nds32/nds32.h (nds32_builtins): Likewise.
10183 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
10184 Andrew Pinski <pinsika@gcc.gnu.org>
10186 PR middle-end/82976
10187 * match.pd: Use constant_boolean_node of correct type instead of
10188 boolean_true_node or boolean_false_node for simplifying
10189 pointer comparisons to zero.
10191 2018-04-07 Jakub Jelinek <jakub@redhat.com>
10193 PR tree-optimization/80021
10194 * tree.c (verify_type_variant): Make error call in verify_variant_match
10195 translatable and remove final full stop.
10197 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10199 * config/nds32/constants.md (unspec_volatile_element): Add
10200 UNSPEC_VOLATILE_EH_RETURN.
10201 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
10202 nds32_output_stack_pop): Support dwarf exception handling process.
10203 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
10204 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
10205 exception handling process.
10206 (nds32_compute_stack_frame): Likewise.
10207 (nds32_return_addr_rtx): Likewise.
10208 (nds32_initial_elimination_offset): Likewise.
10209 (nds32_expand_prologue): Likewise.
10210 (nds32_expand_epilogue): Likewise.
10211 (nds32_dynamic_chain_address): New function.
10212 * config/nds32/nds32.h (machine_function): Add fields for dwarf
10213 exception handling.
10214 (DYNAMIC_CHAIN_ADDRESS): Define.
10215 (EH_RETURN_DATA_REGNO): Define.
10216 (EH_RETURN_STACKADJ_RTX): Define.
10217 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
10218 patterns for dwarf exception handling.
10220 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10222 * config/nds32/nds32.h: Clean up obsolete macros.
10224 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10226 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10227 Add enum values for particular instructions.
10228 * config/nds32/nds32-intrinsic.c: Implementation of expanding
10229 particular intrinsic functions.
10230 * config/nds32/nds32-intrinsic.md: Likewise.
10231 * config/nds32/nds32_intrinsic.h: Likewise.
10232 * config/nds32/nds32.h (nds32_builtins): Likewise.
10233 * config/nds32/nds32.md (type): Add pbsad and pbsada.
10234 (btst, ave): New patterns for particular instructions.
10236 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10238 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10239 Add enum values for atomic load/store and memory sync.
10240 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
10242 * config/nds32/nds32-intrinsic.md: Likewise.
10243 * config/nds32/nds32_intrinsic.h: Likewise.
10244 * config/nds32/nds32.h (nds32_builtins): Likewise.
10246 2018-04-07 Jakub Jelinek <jakub@redhat.com>
10248 PR tree-optimization/85257
10249 * fold-const.c (native_encode_vector): If not all elts could fit
10250 and off is -1, return 0 rather than offset.
10251 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
10252 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
10253 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
10254 adjust buffer in native_interpret_expr call.
10256 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10258 * config/nds32/constants.md (unspec_volatile_element): Add cache
10259 control enum values.
10260 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
10261 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
10262 * config/nds32/nds32.c (nds32_cctl_names): New.
10263 (nds32_print_operand): Handle cache control register names.
10264 * config/nds32/nds32.h (nds32_builtins): New enum values.
10265 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
10267 * config/nds32/nds32.md (type): Add mmu.
10268 * config/nds32/pipelines.md (simple_insn): Add mmu.
10270 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10272 * config/nds32/nds32.md (type): Remove call.
10273 * config/nds32/pipelines.md (simple_insn): Likewise.
10275 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10277 * config/nds32/constants.md (unspec_volatile_element): Add
10278 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
10279 UNSPEC_VOLATILE_FMFCFG.
10280 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
10281 description for fmfcfg and fmfcsr.
10282 (bdesc_1arg): Add fmtcsr.
10283 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
10284 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
10285 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
10286 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
10287 unspec_fmfcfg): New patterns.
10288 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
10289 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
10290 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
10291 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
10292 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
10293 __nds32__fmfcfg): Define.
10295 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10297 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
10298 intrinsic register names.
10299 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
10300 intrinsic register enum values and macros.
10302 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10304 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
10305 for load/store addressing form.
10306 (nds32_print_operand_address): Likewise.
10308 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
10311 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
10312 based on LABEL_REF. Remove useless assertion.
10313 (pic_address_needs_scratch): Fix formatting.
10314 (sparc_legitimize_pic_address): Minor tweaks.
10315 (sparc_delegitimize_address): Adjust assertion accordingly.
10316 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
10317 into symbolic_operand.
10318 (movsi_high_pic_label_ref): Likewise.
10319 (movsi_lo_sum_pic_label_ref): Likewise.
10320 (movdi_pic_label_ref): Likewise.
10321 (movdi_high_pic_label_ref): Likewise.
10322 (movdi_lo_sum_pic_label_ref): Likewise.
10324 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
10326 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
10327 custom LIB_SPEC setup.
10329 2018-04-06 Ruslan Bukin <br@bsdpad.com>
10330 Kito Cheng <kito.cheng@gmail.com>
10332 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
10333 * config/riscv/freebsd.h: New.
10335 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10337 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
10338 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
10341 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10342 Kito Cheng <kito.cheng@gmail.com>
10344 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
10345 nds32_output_call, nds32_symbol_binds_local_p): New functions.
10346 * config/nds32/nds32-protos.h (nds32_output_call,
10347 nds32_output_return): Declare.
10348 * config/nds32/nds32.md: Refine all the call and return patterns.
10350 2018-04-06 Jakub Jelinek <jakub@redhat.com>
10353 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
10354 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
10356 PR rtl-optimization/84872
10357 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
10358 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
10359 EDGE_CROSSING edge.
10361 2018-04-06 Tamar Christina <tamar.christina@arm.com>
10363 * expr.c (copy_blkmode_to_reg): Revert 254862.
10364 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
10366 2018-04-06 Richard Biener <rguenther@suse.de>
10368 PR middle-end/85244
10369 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
10370 after seeing a component reference with an adjacent field. Treat
10371 refs to arrays at struct end of external decls similar to
10372 refs to unconstrained commons.
10374 2018-04-06 Jakub Jelinek <jakub@redhat.com>
10377 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
10378 look through SAVE_EXPRs with non-side-effects argument. Adjust
10380 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
10383 2018-04-06 Richard Biener <rguenther@suse.de>
10385 PR middle-end/85180
10386 * alias.c (find_base_term): New wrapper around find_base_term
10387 unwinding CSELIB_VAL_PTR changes.
10388 (find_base_term): Do not restore CSELIB_VAL_PTR during the
10391 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10393 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
10395 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
10396 constant definitions.
10397 ("nop"): lr 0,0 -> nopr r0
10398 ("nop_lr0", "nop_lr1"): New insn definitions.
10400 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10402 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
10403 NDS32_V3PUSH_AVAILABLE_P macro.
10405 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
10406 Chung-Ju Wu <jasonwucj@gmail.com>
10408 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
10409 (nds32*-*-*): Add float and fpu_config into supported_defaults.
10410 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
10411 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
10412 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
10413 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
10414 * config/nds32/constraints.md: New constraints and checking for hard
10415 float configuration.
10416 * config/nds32/iterators.md: New mode iterator and attribute for hard
10417 float configuration.
10418 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
10420 * config/nds32/nds32-fpu.md: New file.
10421 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
10422 deal with hard float code generation.
10423 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
10425 (abi_type, float_reg_number): New enum type.
10426 * config/nds32/nds32-predicates.c: New predicates for hard float.
10427 * config/nds32/nds32-protos.h: Declare functions for hard float.
10428 * config/nds32/nds32.c: Implementation for hard float configuration.
10429 * config/nds32/nds32.h: Definitions for hard float configuration.
10430 * config/nds32/nds32.md: Include hard float machine description and
10431 modify patterns for hard float configuration.
10432 * config/nds32/nds32.opt: New options for hard float configuration.
10433 * config/nds32/predicates.md: New predicates for hard float
10436 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10438 * common/config/nds32/nds32-common.c
10439 (nds32_option_optimization_table): Enable -mreleax-hint by default.
10441 2018-04-05 Jakub Jelinek <jakub@redhat.com>
10443 PR middle-end/85195
10444 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
10445 CONSTRUCTOR_ELT (ctor, ...)->value.
10447 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
10450 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
10452 2018-04-05 Tom de Vries <tom@codesourcery.com>
10455 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
10458 2018-04-05 Shiva Chen <shiva0217@gmail.com>
10459 Kito Cheng <kito.cheng@gmail.com>
10461 * config/nds32/constraints.md (U33): Fine-tune checking condition.
10462 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
10463 * config/nds32/nds32.h (nds32_16bit_address_type): Add
10464 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
10466 2018-04-05 Shiva Chen <shiva0217@gmail.com>
10467 Kito Cheng <kito.cheng@gmail.com>
10469 * config/nds32/constraints.md (Ufe): New memory constraint.
10470 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
10471 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
10472 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
10474 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
10475 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
10477 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10479 * config/nds32/nds32.md: Use optimize_size in the condition for
10480 alu-shift instructions.
10482 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10484 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
10486 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10488 * config/nds32/nds32.md (negsi2): Refine pattern.
10490 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
10491 Chung-Ju Wu <jasonwucj@gmail.com>
10493 * config/nds32/iterators.md (shift_rotate): New code iterator.
10494 (shift): New code attribute.
10495 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
10496 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
10497 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
10498 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
10499 bit-wise operations.
10500 (andsi3, *andsi3): Ditto.
10501 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
10502 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
10503 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
10504 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
10505 nds32_ior_operand, nds32_xor_operand): New predicates.
10507 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10509 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
10510 (addsi3, subsi3): ... this.
10512 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10514 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
10516 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10518 * config/nds32/nds32.md: Adjust indention.
10520 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
10522 * config/nds32/nds32.md (feature): New attribute.
10524 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10526 * config/nds32/nds32.md (subtype): New attribute.
10528 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
10531 * config/arm/arm-builtins.c (arm_expand_builtin): Change
10532 expansion to perform a bitwise AND of the argument followed by a
10533 boolean negation of the result.
10535 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
10537 PR rtl-optimization/84878
10538 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
10539 the basic block. Assert the use reference is not artificial and that
10540 it has an associated insn.
10542 2018-04-04 Michael Matz <matz@suse.de>
10544 * builtins.c (compute_objsize): Pass correct operand
10545 to array_at_struct_end_p.
10547 2018-04-04 Richard Biener <rguenther@suse.de>
10550 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
10551 from contexts for DINFO_LEVEL_TERSE and below.
10553 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10555 * config/nds32/nds32-doubleword.md (move_<mode>): Require
10556 resiter_operand condition.
10557 * config/nds32/nds32.md (*move<mode>): Ditto.
10559 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10560 Monk Chiang <sh.chiang04@gmail.com>
10562 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
10564 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10566 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
10568 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10569 Kito Cheng <kito.cheng@gmail.com>
10571 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
10572 nds32_cond_code_str, output_cond_branch,
10573 output_cond_branch_compare_zero, nds32_expand_cbranch,
10574 nds32_expand_cstore, nds32_expand_movcc,
10575 nds32_output_cbranchsi4_equality_zero,
10576 nds32_output_cbranchsi4_equality_reg,
10577 nds32_output_cbranchsi4_equality_reg_or_const_int,
10578 nds32_output_cbranchsi4_greater_less_zero: New functions.
10579 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
10580 nds32_expand_cstore, nds32_expand_movcc,
10581 nds32_output_cbranchsi4_equality_zero,
10582 nds32_output_cbranchsi4_equality_reg,
10583 nds32_output_cbranchsi4_equality_reg_or_const_int,
10584 nds32_output_cbranchsi4_greater_less_zero): Declare.
10585 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
10586 nds32_rimm11s_operand): New predicates.
10587 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
10588 * config/nds32/nds32.md: Rewrite all the branch and conditional move
10591 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10593 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
10594 * config/nds32/nds32.md: Ditto.
10595 * config/nds32/pipelines.md: Ditto.
10597 2018-04-04 Richard Biener <rguenther@suse.de>
10599 PR tree-optimization/85168
10600 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
10601 propagating abnormals.
10603 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10605 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
10607 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10608 Kito Cheng <kito.cheng@gmail.com>
10610 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
10611 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
10612 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
10613 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
10614 * config/nds32/nds32.md (sibcall_internal): New.
10615 (sibcall_register): Remove.
10616 (sibcall_immediate): Remove.
10617 (sibcall_value_internal): New.
10618 (sibcall_value_register): Remove.
10619 (sibcall_value_immediate): Remove.
10620 * config/nds32/predicates.md (nds32_general_register_operand): New.
10621 (nds32_call_address_operand): New.
10623 2018-04-03 Jakub Jelinek <jakub@redhat.com>
10625 PR rtl-optimization/85167
10626 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
10627 bb_defs if *split_p, instead preinitialize it to NULL.
10629 PR tree-optimization/85156
10630 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
10631 evaluating the argument multiple times.
10633 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
10635 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
10637 (_mm_cvtpd_ps): Likewise.
10638 (_mm_cvttpd_epi32): Likewise.
10639 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
10640 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
10641 vector, pixel, and bool following altivec.h include.
10643 2018-04-03 Martin Sebor <msebor@redhat.com>
10645 * doc/extend.texi (Common Function Attributes): Clarify.
10646 (const attribute): Likewise.
10647 (pure attribute): Likewise.
10649 2018-04-03 Jakub Jelinek <jakub@redhat.com>
10652 * config/i386/i386.c (ix86_expand_vector_set): Use
10653 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
10655 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
10657 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
10658 instructions when changing rounding bits to preserve precision bits
10659 in the x87 control word.
10661 2018-04-03 Martin Liska <mliska@suse.cz>
10663 PR tree-optimization/82491
10664 * rtl.h (strip_offset_and_add): Replace += suboffset with
10665 poly_uint64 () + suboffset.
10667 2018-03-29 Martin Liska <mliska@suse.cz>
10668 Martin Jambor <mjambor@suse.cz>
10671 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
10672 param_type is not an integral or pointer type.
10674 2018-04-03 Richard Biener <rguenther@suse.de>
10676 * sese.h (recompute_all_dominators): Remove.
10678 2018-04-02 Martin Sebor <msebor@redhat.com>
10680 * doc/invoke.texi (-Wrestrict): Fix typos.
10682 2018-04-02 Jim Wilson <jimw@sifive.com>
10684 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
10685 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
10686 (<optab>di3, <optab>si3_extend): Likewise.
10687 (<optab>si3_mask, <optab>si3_mask_1): New.
10688 (<optab>di3_mask, <optab>di3_mask_1): New.
10689 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
10690 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
10691 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
10693 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
10695 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
10698 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
10700 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
10701 (nds32_canonicalize_comparison): New function.
10703 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10704 Kito Cheng <kito.cheng@gmail.com>
10705 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10707 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
10708 * config/nds32/constants.md (unspec_volatile_element): Add
10709 UNSPEC_VOLATILE_RELAX_GROUP.
10710 * config/nds32/nds32-relax-opt.c: New file.
10711 * config/nds32/nds32-predicates.c
10712 (nds32_symbol_load_store_p): New function.
10713 * config/nds32/nds32-protos.h
10714 (nds32_symbol_load_store_p): Declare function.
10715 (make_pass_nds32_relax_opt): Declare new rtl pass function.
10716 * config/nds32/nds32.c
10717 (nds32_register_pass): New function to register pass.
10718 (nds32_register_passes): New function to register passes.
10719 * config/nds32/nds32.md (relax_group): New pattern.
10720 * config/nds32/nds32.opt (mrelax-hint): New option.
10721 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
10723 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
10725 * config/nds32/t-nds32: Modify files dependency.
10727 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10729 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
10730 (PROFILE_HOOK): Define its implementation.
10732 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10734 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
10735 type and 32-bit size.
10737 2018-04-01 Jakub Jelinek <jakub@redhat.com>
10739 PR middle-end/85090
10740 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
10741 (V_128_256): New mode iterator.
10742 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
10743 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
10744 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
10746 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
10747 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
10749 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
10752 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
10753 NaN inputs correctly.
10755 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
10758 * config/rs6000/vsx.md (??r): New mode attribute.
10759 (*vsx_mov<mode>_64bit): Use it.
10760 (*vsx_mov<mode>_32bit): Likewise.
10762 2018-03-30 Martin Sebor <msebor@redhat.com>
10764 PR tree-optimization/84818
10765 * builtins.c (check_access): Use warning_n.
10767 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10770 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
10772 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
10775 2018-03-30 Julia Koval <julia.koval@intel.com>
10778 * x86-tune.def (movx, partial_reg_dependency): Enable for
10781 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
10783 PR inline-asm/84985
10784 * lra-constraints.c (process_alt_operands): Move setting
10785 this_alternative_matches below.
10787 2018-03-29 Martin Liska <mliska@suse.cz>
10790 * doc/invoke.texi: Document how LTO works with debug info.
10791 Describe auto-load support of binutils. Mention 'x86-64'
10792 as valid option value of -march option.
10794 2018-03-29 Jakub Jelinek <jakub@redhat.com>
10796 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
10799 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
10800 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
10801 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
10804 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10807 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
10808 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
10809 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
10810 for RS6000_BTM_POWERPC64.
10811 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
10812 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
10813 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
10816 (DIVDEU): Likewise.
10818 2018-03-28 Carl Love <cel@us.ibm.com>
10821 2017-09-27 Carl Love <cel@us.ibm.com>
10823 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
10824 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
10825 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
10828 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10830 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
10831 instead of __vector bool.
10832 (_mm_max_pu8): Likewise.
10833 (_mm_min_pi16): Likewise.
10835 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10838 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
10839 (DIVWEUO): Likewise.
10840 (DIVDEO): Likewise.
10841 (DIVDEUO): Likewise.
10842 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
10843 DIVWEUO and DIVDEUO.
10844 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
10845 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
10846 (div_extend): Likewise.
10847 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
10849 (__builtin_divweuo): Likewise.
10850 (__builtin_divdeo): Likewise.
10851 (__builtin_divdeuo): Likewise.
10853 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10856 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
10857 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
10859 PR tree-optimization/82004
10860 * gimple-match-head.c (optimize_pow_to_exp): New function.
10861 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
10862 Don't fold to exp if optimize_pow_to_exp is false.
10864 2018-03-28 Martin Liska <mliska@suse.cz>
10867 * calls.c (initialize_argument_information): Fix trailing space.
10868 * common.opt: Fix typo and provide better explanation for
10869 -fsanitize-coverage option.
10870 * config/i386/i386.opt: Fix typo.
10872 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10873 Martin Liska <mliska@suse.cz>
10876 * gimplify.c (asan_poison_variable): Don't do the check for
10877 gimplify_omp_ctxp here.
10878 (gimplify_decl_expr): Do it here.
10879 (gimplify_target_expr): Likewise.
10881 2018-03-28 Martin Liska <mliska@suse.cz>
10884 * config/i386/i386.c (ix86_function_arg_advance): Do not call
10885 chkp_type_bounds_count if MPX is not enabled.
10887 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
10889 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
10891 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
10894 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
10895 function to create the function decl for complex long double
10896 multiply and divide for -mabi=ieeelongdouble.
10897 (init_float128_ieee): Call it.
10899 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
10902 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
10903 -fcf-protection=branch -mibt.
10904 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
10906 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10909 * config/arm/arm.c (arm_valid_symbolic_address): Handle
10910 arm_word_relocations.
10912 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
10915 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
10916 extern array declarations.
10918 2018-03-27 Richard Biener <rguenther@suse.de>
10920 PR middle-end/84067
10921 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
10922 explicit single_use checks.
10924 2018-03-27 Richard Biener <rguenther@suse.de>
10926 PR tree-optimization/85082
10927 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
10930 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10932 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
10933 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
10934 Turn on fasynchronous-unwind-tables and funwind-tables.
10936 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
10939 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
10940 (*bmi_blsr_<mode>_ccz): Ditto.
10942 2018-03-26 Tom de Vries <tom@codesourcery.com>
10944 PR tree-optimization/85063
10945 * omp-general.c (offloading_function_p): New function. Factor out
10947 * omp-offload.c (pass_omp_target_link::gate): ... here.
10948 * omp-general.h (offloading_function_p): Declare.
10949 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
10950 with attribute omp declare target for offloading functions.
10952 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
10954 PR tree-optimization/84005
10955 * tree-data-ref.h (get_base_for_alignment): Declare.
10956 * tree-data-ref.c (get_base_for_alignment_1): New function.
10957 (get_base_for_alignment): Likewise.
10958 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
10959 get_base_for_alignment to find a suitable base object, instead
10960 of always using drb->base_address.
10962 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10964 PR inline-asm/85022
10965 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
10966 known size by default.
10968 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
10970 PR inline-asm/85030
10971 * lra-constraints.c (process_alt_operands): Don't match BLKmode
10972 and non BLKmode operands.
10974 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10977 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
10978 Clean up attributes.
10980 2018-03-23 Richard Biener <rguenther@suse.de>
10983 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
10984 we are going to emit early debug for LTO.
10986 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10988 PR inline-asm/85034
10989 * function.c (match_asm_constraints_1): Don't optimize if input
10990 doesn't satisfy general_operand predicate for output's mode.
10992 PR inline-asm/85022
10993 * alias.c (write_dependence_p): Don't require for x_canonicalized
10994 non-VOIDmode if x has VOIDmode.
10997 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
10998 just don't try to optimize it rather than assert it never happens.
11000 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
11002 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
11003 macro expansions for definition of ST_INTERNAL_<mode> and
11004 LD_INTERNAL_<mode> builtins.
11005 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
11007 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
11009 (altivec_expand_st_builtin): Likewise.
11010 (altivec_expand_builtin): Remove calls to deleted functions.
11011 (rs6000_address_for_altivec): Delete this function.
11012 * config/rs6000/vector.md: Remove expands for
11013 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
11015 2018-03-22 Sudakshina Das <sudi.das@arm.com>
11018 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
11019 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
11020 re-computing once computed.
11021 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
11022 (arm_init_machine_status): Initialize
11023 machine->static_chain_stack_bytes.
11025 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
11028 * doc/extend.texi: Add four new prototypes for vec_ld.
11029 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
11030 definitions for more logical presentation.
11031 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
11032 entries for V1TI variants of __builtin_altivec_ld builtin.
11033 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
11034 handling of V1TI variant of LVX icode pattern.
11035 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
11036 (rs6000_gimple_fold_builtin): Likewise.
11037 (altivec_init_builtins): Add code to define
11038 __builtin_altivec_lvx_v1ti function.
11040 2018-03-22 Jakub Jelinek <jakub@redhat.com>
11042 PR inline-asm/84941
11043 * function.c (match_asm_constraints_1): Don't do the optimization
11044 if input isn't a REG, SUBREG, MEM or constant.
11046 2018-03-22 Tom de Vries <tom@codesourcery.com>
11048 PR tree-optimization/84956
11049 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
11050 bb_has_abnormal_pred.
11052 2018-03-22 Jakub Jelinek <jakub@redhat.com>
11055 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
11056 DECL_INITIAL (decl) to decl at the end.
11057 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
11058 adjust the comment.
11060 2018-03-21 Joseph Myers <joseph@codesourcery.com>
11062 * doc/extend.texi (__builtin_tgmath): Document when complex
11063 integer types are treated as _Complex _Float64.
11065 2018-03-21 Tom de Vries <tom@codesourcery.com>
11067 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
11069 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11071 PR tree-optimization/84960
11072 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
11073 if it is ENTRY block, move them into single succ of ENTRY in that case.
11075 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
11077 PR tree-optimization/84811
11078 * poly-int.h (poly_span_traits): Remove the T3 parameter and
11079 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
11080 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
11081 (known_subrange_p): Update accordingly. Cast each value involved
11082 in the size comparison, rather than casting the result of the
11085 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11087 PR tree-optimization/84982
11088 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
11089 by flipping the least significant bit rather than all bits from
11090 bitpos to bitpos + bitsize - 1.
11092 2018-03-21 Nathan Sidwell <nathan@acm.org>
11094 * doc/extend.texi (Deprecated Features): Remove mention of
11095 long-deleted deprecations.
11097 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11100 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
11101 * configure: Regenerate.
11103 2018-03-21 Tom de Vries <tom@codesourcery.com>
11105 PR tree-optimization/83126
11106 * tree-parloops.c (num_phis): New function.
11107 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
11109 2018-03-21 Nathan Sidwell <nathan@acm.org>
11111 * doc/extend.texi (Deprecated Features): Update deprecated flags,
11112 mention anon-struct/union members and trailing attributes.
11114 2018-03-21 Bin Cheng <bin.cheng@arm.com>
11116 PR tree-optimization/84969
11117 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
11118 builtin memset partitions if they set different rhs values.
11120 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11122 PR rtl-optimization/84989
11123 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
11124 VEC_DUPLICATE with scalar result mode.
11126 2018-03-21 Martin Liska <mliska@suse.cz>
11129 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
11130 not intended return statement.
11132 2018-03-21 Martin Liska <mliska@suse.cz>
11135 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
11136 (chkp_find_bound_slots_1): Limit number of iterations.
11138 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
11141 * Minor grammar fixes for x86 options.
11143 2018-03-20 Jakub Jelinek <jakub@redhat.com>
11146 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
11147 holding REG_CFA_RESTORE notes, instead turn them into a USE.
11149 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
11152 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
11153 (altivec_lvx_<mode>_1op): Likewise.
11154 (altivec_stvx_<mode>_2op): Likewise.
11155 (altivec_stvx_<mode>_1op): Likewise.
11156 (altivec_lvx_<VM2:mode>): New define_expand.
11157 (altivec_stvx_<VM2:mode>): Likewise.
11158 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
11159 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
11160 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
11161 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
11162 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
11163 (rs6000_gen_lvx): Likewise.
11164 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
11165 (altivec_expand_stv_builtin): Likewise.
11166 (altivec_expand_builtin): Likewise.
11167 * config/rs6000/vector.md: Likewise.
11169 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11172 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
11175 2018-03-20 Richard Biener <rguenther@suse.de>
11178 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
11179 sign-conversions as zero, fall back to standard scalar_stmt
11182 2018-03-20 Martin Liska <mliska@suse.cz>
11185 * predict.c (rebuild_frequencies): Handle case when we have
11186 PROFILE_ABSENT, but flag_guess_branch_prob is false.
11188 2018-03-20 Jakub Jelinek <jakub@redhat.com>
11191 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
11192 flag_section_anchors.
11193 * varasm.c (use_blocks_for_decl_p): Remove hack for
11194 dw2_force_const_mem.
11197 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
11199 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
11200 be created, use lowpart_subreg of operands[0] rather than operands[0]
11202 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
11203 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
11204 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
11205 and n constraint instead of aarch64_shift_imm_di and Usd.
11206 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
11207 (*aarch64_<optab>_reg_minus<mode>3): ... this.
11209 2018-03-20 Sudakshina Das <sudi.das@arm.com>
11212 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
11213 to favor GPR over NEON registers.
11214 (<shift>di3_neon): Likewise.
11216 2018-03-20 Tom de Vries <tom@codesourcery.com>
11219 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
11220 (nvptx_process_pars): Emit bar.sync asap and alap.
11222 2018-03-20 Tom de Vries <tom@codesourcery.com>
11225 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
11226 seen_label if seen_label is already set.
11228 2018-03-20 Jakub Jelinek <jakub@redhat.com>
11231 * config/i386/i386.c (fold_builtin_cpu): For features above 31
11232 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
11233 Use 1U instead of 1. Formatting fixes.
11236 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
11237 instead of TREE_TYPE (s1) for the return value.
11239 2018-03-19 Jakub Jelinek <jakub@redhat.com>
11241 PR tree-optimization/84946
11242 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
11243 bitsize + bitsize in poly_uint64 rather than poly_int64.
11246 * dwarf2asm.c: Include fold-const.c.
11247 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
11248 of decl rather than decl itself.
11250 PR rtl-optimization/84643
11251 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
11253 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
11256 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
11257 calling assemble_variable.
11259 2018-03-19 Sudakshina Das <sudi.das@arm.com>
11262 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
11263 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
11265 2018-03-19 Jim Wilson <jimw@sifive.com>
11268 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
11269 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
11270 (riscv_first_stack_step): Likewise.
11271 (riscv_option_override): Use STACK_BOUNDARY instead of
11272 MIN_STACK_BOUNDARY.
11273 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
11274 MIN_STACK_BOUNDARY.
11275 (BIGGEST_ALIGNMENT): Set to 128.
11276 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
11277 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
11280 2018-03-19 Richard Biener <rguenther@suse.de>
11282 PR tree-optimization/84933
11283 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
11284 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
11286 2018-03-19 Richard Biener <rguenther@suse.de>
11288 PR tree-optimization/84859
11289 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
11290 (cond_if_else_store_replacement): Perform sinking operation on
11291 single-store BBs regardless of MAX_STORES_TO_SINK setting.
11292 Generalize what a BB with a single eligible store is.
11294 2018-03-19 Richard Biener <rguenther@suse.de>
11296 PR tree-optimization/84929
11297 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
11298 chrec_is_positive against non-chrec arg.
11300 2018-03-19 Tamar Christina <tamar.christina@arm.com>
11303 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
11305 2018-03-18 Martin Liska <mliska@suse.cz>
11307 PR rtl-optimization/84635
11308 * regrename.c (build_def_use): Use matches_mode only when
11311 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
11313 PR tree-optimization/84913
11314 * tree-vect-loop.c (vectorizable_reduction): Don't try to
11315 vectorize chains of COND_EXPRs.
11317 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11319 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
11321 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11323 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
11325 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11327 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
11329 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
11330 Kito Cheng <kito.cheng@gmail.com>
11332 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
11333 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
11334 (nds32_adjust_reg_alloc_order): New function.
11335 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
11337 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
11339 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
11340 nds32_print_operand, nds32_print_operand_address): Use
11341 HOST_WIDE_INT_PRINT_DEC instead.
11343 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
11345 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
11347 2018-03-17 Jakub Jelinek <jakub@redhat.com>
11350 * config/i386/i386.c (initial_ix86_tune_features,
11351 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
11352 unsigned long long.
11353 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
11354 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
11355 rather than 1u << ix86_tune. Formatting fix.
11356 (ix86_option_override_internal): Change ix86_arch_mask from
11357 unsigned int to unsigned HOST_WIDE_INT, initialize to
11358 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
11359 (ix86_function_specific_restore): Likewise.
11361 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11364 * postreload.c (reload_combine_recognize_pattern): Perform
11365 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
11366 truncate_int_for_mode the result for the destination's mode.
11369 * hsa-gen.c (mem_type_for_type): Fix comment typo.
11370 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
11372 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11375 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
11378 * lra-assigns.c (lra_split_hard_reg_for): Don't use
11379 regno_allocno_class_array and sorted_pseudos.
11380 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
11381 insns where regno is used.
11383 2018-03-16 Martin Liska <mliska@suse.cz>
11386 * multiple_target.c (create_dispatcher_calls): Redirect
11387 reference in the symbol table.
11389 2018-03-16 Martin Liska <mliska@suse.cz>
11392 * multiple_target.c (create_dispatcher_calls): Redirect also
11395 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11399 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
11400 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
11401 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
11403 2018-03-16 Julia Koval <julia.koval@intel.com>
11405 * doc/invoke.texi (Skylake Server): Add CLWB.
11406 Cannonlake): Remove CLWB.
11408 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11410 PR tree-optimization/84841
11411 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
11413 (FLOAT_ONE_CONST_TYPE): Define.
11414 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
11415 (sort_by_operand_rank): Put entries with higher constant_type last
11416 rather than first to match comments.
11418 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
11420 * config/nios2/nios2.md (movsi_internal): Fix thinko in
11423 2018-03-15 Jakub Jelinek <jakub@redhat.com>
11426 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
11427 check and use address of target always.
11429 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
11432 * config/i386/i386.c (indirect_thunk_needed): Update comments.
11433 (indirect_thunk_bnd_needed): Likewise.
11434 (indirect_thunks_used): Likewise.
11435 (indirect_thunks_bnd_used): Likewise.
11436 (indirect_return_needed): New.
11437 (indirect_return_bnd_needed): Likewise.
11438 (output_indirect_thunk_function): Add a bool argument for
11440 (output_indirect_thunk_function): Don't generate alias for
11441 function return thunk.
11442 (ix86_code_end): Call output_indirect_thunk_function to generate
11443 function return thunks.
11444 (ix86_output_function_return): Set indirect_return_bnd_needed
11445 and indirect_return_needed instead of indirect_thunk_bnd_needed
11446 and indirect_thunk_needed.
11448 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
11450 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
11451 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
11452 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
11454 2018-03-15 David Malcolm <dmalcolm@redhat.com>
11455 Paul Hua <paul.hua.gm@gmail.com>
11458 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
11460 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
11462 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
11463 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
11464 resp. SFmode cases.
11466 2018-03-15 Tamar Christina <tamar.christina@arm.com>
11469 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
11470 instead of GET_MODE_SIZE when comparing Units.
11472 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
11475 * varasm.c (hash_section): Return an unchangeble hash value
11476 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
11477 Return !aarch64_can_use_per_function_literal_pools_p ().
11479 2018-03-15 Jakub Jelinek <jakub@redhat.com>
11482 * optabs.c (emit_conditional_move): Pass address of cmode's copy
11483 rather than address of cmode as last argument to prepare_cmp_insn.
11485 2018-03-15 Julia Koval <julia.koval@intel.com>
11487 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
11488 F_AVX512VNNI, F_AVX512BITALG): New.
11490 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
11493 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
11494 insn for floating-point loads and stores.
11496 2018-03-14 Carl Love <cel@us.ibm.com>
11498 * config/rs6000/rs6000-c.c: Add macro definitions for
11499 ALTIVEC_BUILTIN_VEC_PERMXOR.
11500 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
11501 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
11502 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
11503 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
11505 * config/doc/extend.texi: Add prototypes for vec_permxor.
11507 2018-03-14 David Malcolm <dmalcolm@redhat.com>
11510 * diagnostic-show-locus.c (class layout_point): Convert m_line
11511 from int to linenum_type.
11512 (line_span::comparator): Use linenum "compare" function when
11513 comparing line numbers.
11514 (test_line_span): New function.
11515 (layout_range::contains_point): Convert param "row" from int to
11517 (layout_range::intersects_line_p): Likewise.
11518 (layout::will_show_line_p): Likewise.
11519 (layout::print_source_line): Likewise.
11520 (layout::should_print_annotation_line_p): Likewise.
11521 (layout::print_annotation_line): Likewise.
11522 (layout::print_leading_fixits): Likewise.
11523 (layout::annotation_line_showed_range_p): Likewise.
11524 (struct line_corrections): Likewise for field m_row.
11525 (line_corrections::line_corrections): Likewise for param "row".
11526 (layout::print_trailing_fixits): Likewise.
11527 (layout::get_state_at_point): Likewise.
11528 (layout::get_x_bound_for_row): Likewise.
11529 (layout::print_line): Likewise.
11530 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
11531 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
11532 * input.c (selftest::test_linenum_comparisons): New function.
11533 (selftest::input_c_tests): Call it.
11534 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
11535 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
11536 * selftest.h (ASSERT_GT): New macro.
11537 (ASSERT_GT_AT): New macro.
11538 (ASSERT_LT): New macro.
11539 (ASSERT_LT_AT): New macro.
11541 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
11543 PR rtl-optimization/84780
11544 * combine.c (distribute_links): Don't make a link based on pc_rtx.
11546 2018-03-14 Martin Liska <mliska@suse.cz>
11548 * tree.c (record_node_allocation_statistics): Use
11549 get_stats_node_kind.
11550 (get_stats_node_kind): New function extracted from
11551 record_node_allocation_statistics.
11552 (free_node): Use get_stats_node_kind.
11554 2018-03-14 Richard Biener <rguenther@suse.de>
11556 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
11557 that the value-set of ANTIC_IN doesn't grow.
11560 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11562 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11563 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11565 2018-03-14 Julia Koval <julia.koval@intel.com>
11567 * config.gcc (icelake-client, icelake-server): New.
11569 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
11570 (initial_ix86_arch_features): Ditto.
11571 (PTA_SKYLAKE): Add SGX.
11572 (PTA_ICELAKE): Remove.
11573 (PTA_ICELAKE_CLIENT): New.
11574 (PTA_ICELAKE_SERVER): New.
11575 (ix86_option_override_internal): Split up icelake on icelake client and
11577 (get_builtin_code_for_version): Ditto.
11578 (fold_builtin_cpu): Ditto.
11579 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
11580 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
11581 * config/i386/i386.h (processor_type): Ditto.
11582 * doc/invoke.texi: Ditto.
11584 2018-03-14 Jakub Jelinek <jakub@redhat.com>
11587 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
11588 INTEGER_CST offset, add it together with bitpos / 8 and
11589 sign extend based on POINTER_SIZE.
11593 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
11596 * config/i386/constraints.md (Yc): New register constraint.
11597 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
11598 Use Yc constraint for alternative 2 of operand 0. Remove
11599 preferred_for_speed attribute.
11601 2018-03-14 Richard Biener <rguenther@suse.de>
11603 PR tree-optimization/84830
11604 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
11605 with the old one to avoid oscillations.
11607 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
11610 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
11612 (assign_by_spills): Return a flag of reload assignment failure.
11613 Do not process the reload assignment failures. Do not spill other
11614 reload pseudos if they has the same reg class. Update n if
11616 (lra_assign): Add a return arg. Set up from the result of
11617 assign_by_spills call.
11618 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11619 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11620 usage_insns if it is not NULL.
11621 (spill_hard_reg_in_range): New function.
11622 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11623 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11624 function prototypes.
11625 (lra_assign): Change prototype.
11626 * lra.c (lra): Add code to deal with fails by splitting hard reg
11629 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
11631 * config/riscv/riscv.opt (mrelax): New option.
11632 * config/riscv/riscv.c (riscv_file_start): Emit ".option
11633 "norelax" when riscv_mrelax is disabled.
11634 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
11636 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11639 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
11640 reassociation for int modes.
11642 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11644 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
11645 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
11647 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
11648 * config/aarch64/aarch64-sve.md
11649 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
11650 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
11651 (*extend<mode><Vwide>2): Rename to...
11652 (aarch64_sve_extend<mode><Vwide>2): ...this.
11653 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
11654 renaming the old pattern to...
11655 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
11657 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
11658 define_expand, renaming the old pattern to...
11659 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
11660 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
11661 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
11662 account when deciding which SVE instruction the optab should use.
11663 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
11665 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11667 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
11668 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
11669 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
11670 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
11671 (tlsdesc_small_<mode>): Turn a define_expand and use
11672 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
11673 (tlsdesc_small_advsimd_<mode>): ...this.
11674 (tlsdesc_small_sve_<mode>): New pattern.
11676 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11678 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
11679 (UNSPEC_UMUL_HIGHPART): New constants.
11680 (MUL_HIGHPART): New int iteraor.
11681 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
11682 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
11684 (*<su>mul<mode>3_highpart): New define_insn.
11686 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
11689 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
11692 2018-03-13 Martin Liska <mliska@suse.cz>
11695 * (sem_item_optimizer::sem_item_optimizer): Initialize new
11697 (sem_item_optimizer::~sem_item_optimizer): Release it.
11698 (sem_item_optimizer::merge_classes): Register variable aliases.
11699 (sem_item_optimizer::fixup_pt_set): New function.
11700 (sem_item_optimizer::fixup_points_to_sets): Likewise.
11701 * ipa-icf.h: Declare new variables and functions.
11703 2018-03-13 Jakub Jelinek <jakub@redhat.com>
11705 PR middle-end/84834
11706 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
11707 integer_pow2p@2 and test integer_pow2p in condition.
11708 (A < 0 ? C : 0): Similarly for @1.
11710 PR middle-end/84831
11711 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
11712 characters starting at p contain '\0' character, don't look beyond
11716 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
11717 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
11720 * reg-stack.c (change_stack): Change update_end var from int to
11721 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
11722 also call set_block_for_insn on the newly added insns and rescan.
11725 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
11726 on the last operand.
11729 * tree.c (stabilize_reference_1): Return save_expr (e) for
11730 STATEMENT_LIST even if it doesn't have side-effects.
11732 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
11734 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
11736 2018-03-12 Renlin Li <renlin.li@arm.com>
11738 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
11739 aarch64_output_scalar_simd_mov_immediate.
11741 2018-03-12 Martin Sebor <msebor@redhat.com>
11743 PR tree-optimization/83456
11744 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
11745 for perfectly overlapping calls to memcpy.
11746 (gimple_fold_builtin_memory_chk): Same.
11747 (gimple_fold_builtin_strcpy): Handle no-warning.
11748 (gimple_fold_builtin_stxcpy_chk): Same.
11749 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
11751 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11753 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
11754 parameter. Use it for SFmode.
11755 (rs6000_function_arg_advance_1): Adjust.
11756 (rs6000_function_arg): Adjust.
11757 (rs6000_gimplify_va_arg): Pass false for that new parameter.
11759 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11761 PR rtl-optimization/84169
11762 PR rtl-optimization/84780
11763 * combine.c (can_combine_p): Check for a 2-insn combination whether
11764 the destination register is used between the two insns, too.
11766 2018-03-12 Richard Biener <rguenther@suse.de>
11768 PR tree-optimization/84803
11769 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
11770 for refs DR analysis didn't process.
11772 2018-03-12 Richard Biener <rguenther@suse.de>
11774 PR tree-optimization/84777
11775 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
11776 force-vectorize loops ignore whether we are optimizing for size.
11778 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
11780 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
11781 (TARGET_MD_ASM_ADJUST): Define.
11783 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
11784 Kito Cheng <kito.cheng@gmail.com>
11785 Chung-Ju Wu <jasonwucj@gmail.com>
11787 * config/nds32/nds32.c (nds32_compute_stack_frame,
11788 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
11789 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
11790 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
11791 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
11792 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
11793 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
11794 * config/nds32/nds32.md (prologue, epilogue): Use macro
11795 NDS32_V3PUSH_AVAILABLE_P to do checking.
11797 2018-03-11 Jakub Jelinek <jakub@redhat.com>
11800 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
11801 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
11802 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
11803 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
11804 addition of most attributes on !orig_type_die or the attribute not
11805 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
11807 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11808 Chung-Ju Wu <jasonwucj@gmail.com>
11810 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
11811 __NDS32_VH__ macro.
11812 * config/nds32/nds32.opt (mvh): New option.
11814 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11815 Chung-Ju Wu <jasonwucj@gmail.com>
11817 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
11819 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
11820 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
11823 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11824 Chung-Ju Wu <jasonwucj@gmail.com>
11826 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
11828 * config/nds32/nds32-multiple.md (strlensi): New pattern.
11829 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
11831 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
11832 Kito Cheng <kito.cheng@gmail.com>
11833 Chung-Ju Wu <jasonwucj@gmail.com>
11835 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
11836 UNSPEC_FFMISM and UNSPEC_FLMISM.
11837 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
11838 for ffb, ffmism and flmism.
11839 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
11840 (unspec_ffmism): Ditto.
11841 (unspec_flmism): Ditto.
11842 (nds32_expand_builtin_impl): Check if string extension is available.
11843 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
11844 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
11846 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
11849 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11852 * lra-assigns.c (assign_by_spills): Return a flag of reload
11853 assignment failure. Do not process the reload assignment
11854 failures. Do not spill other reload pseudos if they has the same
11856 (lra_assign): Add a return arg. Set up from the result of
11857 assign_by_spills call.
11858 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11859 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11860 usage_insns if it is not NULL.
11861 (spill_hard_reg_in_range): New function.
11862 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11863 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11864 function prototypes.
11865 (lra_assign): Change prototype.
11866 * lra.c (lra): Add code to deal with fails by splitting hard reg
11869 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
11872 * config/i386/i386.opt: Replace Enforcment with Enforcement.
11874 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
11877 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
11878 (dw_val_node): Add val_symbolic_view.
11879 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
11880 (symview_upper_bound): New.
11881 (new_line_info_table): Initialize symviews_since_reset.
11882 (dwarf2out_source_line): Count symviews_since_reset and set
11883 symview_upper_bound.
11884 (dw_val_equal_p): Handle symview.
11885 (add_AT_symview): New.
11886 (print_dw_val): Handle symview.
11887 (attr_checksum, attr_checksum_ordered): Likewise.
11888 (same_dw_val_p, size_of_die): Likewise.
11889 (value_format, output_die): Likewise.
11890 (add_high_low_attributes): Use add_AT_symview for entry_view.
11891 (dwarf2out_finish): Reset symview_upper_bound, clear
11894 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
11897 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
11898 Add strict argument and use it.
11899 (rs6000_split_multireg_move): Update for new strict argument.
11900 (mem_operand_gpr): Disallow all non-offsettable addresses.
11901 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
11903 2018-03-09 Jakub Jelinek <jakub@redhat.com>
11906 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
11907 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
11908 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
11911 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
11912 decl, use remap_type if we want to use the type.
11914 2018-03-09 Martin Sebor <msebor@redhat.com>
11916 PR tree-optimization/84526
11917 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11919 (builtin_access::generic_overlap): Be prepared to handle non-array
11922 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
11924 PR rtl-optimization/84682
11925 * lra-constraints.c (process_address_1): Check is_address flag
11926 for address constraints.
11927 (process_alt_operands): Likewise.
11928 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
11929 preprocess_constraints.
11930 * recog.h (preprocess_constraints): Add oploc parameter.
11932 * recog.c (preprocess_constraints): Test address_operand for
11933 CT_ADDRESS constraints.
11935 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11938 * lra-assigns.c (assign_by_spills): Return a flag of reload
11939 assignment failure. Do not process the reload assignment
11940 failures. Do not spill other reload pseudos if they has the same
11942 (lra_assign): Add a return arg. Set up from the result of
11943 assign_by_spills call.
11944 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11945 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11946 usage_insns if it is not NULL.
11947 (spill_hard_reg_in_range): New function.
11948 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11949 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11950 function prototypes.
11951 (lra_assign): Change prototype.
11952 * lra.c (lra): Add code to deal with fails by splitting hard reg
11955 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11958 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
11959 Accept complain bool parameter. Only emit errors if it is true.
11960 (arm_parse_cpu_option_name): Likewise.
11961 (arm_target_thumb_only): Adjust callers of the above.
11962 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
11963 prototype to take a default true bool parameter.
11964 (arm_parse_arch_option_name): Likewise.
11966 2018-03-09 David Malcolm <dmalcolm@redhat.com>
11967 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
11971 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
11972 * configure: Regenerate.
11973 * configure.ac ("linker --version-script option"): New.
11974 ("linker soname option"): New.
11976 2018-03-09 Richard Biener <rguenther@suse.de>
11978 PR tree-optimization/84775
11979 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
11980 immediate uses of predicate stmts and mark them modified.
11983 PR tree-optimization/84178
11984 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11986 (version_loop_for_if_conversion): Delay update_ssa call.
11987 (tree_if_conversion): Delay update_ssa until after predicate
11990 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
11993 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
11994 when the function accesses prior frames.
11996 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11999 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
12000 gen_llsym, otherwise call maybe_gen_llsym.
12002 PR inline-asm/84742
12003 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
12004 has ',' character inside of it.
12006 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12009 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
12010 as clobbering CC_REGNUM.
12012 2018-03-08 Richard Biener <rguenther@suse.de>
12014 PR middle-end/84552
12015 * tree-scalar-evolution.c: Include tree-into-ssa.h.
12016 (follow_copies_to_constant): Do not follow SSA names registered
12019 2018-03-08 Richard Biener <rguenther@suse.de>
12021 PR tree-optimization/84178
12022 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
12024 (version_loop_for_if_conversion): Delay update_ssa call.
12025 (tree_if_conversion): Delay update_ssa until after predicate
12028 2018-03-08 David Malcolm <dmalcolm@redhat.com>
12030 PR tree-optimization/84178
12031 * tree-if-conv.c (release_bb_predicate): Remove the
12032 the assertion that the stmts have NULL use_ops.
12033 Discard the statements, asserting that they haven't
12034 yet been added to a BB.
12036 2018-03-08 Richard Biener <rguenther@suse.de>
12038 PR tree-optimization/84746
12039 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
12040 (phi_translate): Pass in destination ANTIC_OUT set.
12041 (phi_translate_1): Likewise. For a simplified result lookup
12042 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
12043 (phi_translate_set): Adjust.
12044 (do_pre_regular_insertion): Likewise.
12045 (do_pre_partial_partial_insertion): Likewise.
12047 2018-03-08 Martin Liska <mliska@suse.cz>
12049 PR gcov-profile/84735
12050 * doc/gcov.texi: Document usage of profile files.
12051 * gcov-io.h: Document changes in the format.
12053 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
12057 * dwarf2out.c (struct dw_line_info_table): Update comments for
12059 (FORCE_RESET_NEXT_VIEW): New.
12060 (FORCE_RESETTING_VIEW_P): New.
12061 (RESETTING_VIEW_P): Check for -1 too.
12062 (ZERO_VIEW_P): Likewise.
12063 (new_line_info_table): Force-reset next view.
12064 (dwarf2out_begin_function): Likewise.
12065 (dwarf2out_source_line): Simplify zero_view_p initialization.
12066 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
12067 view directly. Omit view when omitting .loc at line 0.
12069 2018-03-08 Jakub Jelinek <jakub@redhat.com>
12071 PR tree-optimization/84740
12072 * tree-switch-conversion.c (process_switch): Call build_constructors
12073 only if info.phi_count is non-zero.
12075 PR tree-optimization/84739
12076 * tree-tailcall.c (find_tail_calls): Check call arguments against
12077 DECL_ARGUMENTS (current_function_decl) rather than
12078 DECL_ARGUMENTS (func) when checking for tail recursion.
12080 2018-03-07 Jakub Jelinek <jakub@redhat.com>
12082 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
12083 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
12084 Volker Reichelt's entry and add entries for people that perform
12085 GCC fuzzy testing and report numerous bugs.
12087 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
12090 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
12091 readonly data in sdata, if that is disabled.
12092 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
12093 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
12094 -mreadonly-in-sdata option.
12096 2018-03-07 Martin Sebor <msebor@redhat.com>
12098 PR tree-optimization/84468
12099 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
12100 basic block when looking for nul assignment.
12102 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
12105 * except.h (output_function_exception_table): Adjust prototype.
12106 * except.c (output_function_exception_table): Remove FNNAME parameter
12107 and add SECTION parameter. Ouput one part of the table at a time.
12108 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
12109 the first part of the exception table and emit unwind directives.
12110 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
12111 (i386_pe_seh_cold_init): Likewise.
12112 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
12113 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
12114 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
12115 (ix86_output_call_insn): Emit a nop in one more case for SEH.
12116 * config/i386/winnt.c: Include except.h.
12117 (struct seh_frame_state): Add reg_offset, after_prologue and
12118 in_cold_section fields.
12119 (i386_pe_seh_end_prologue): Set seh->after_prologue.
12120 (i386_pe_seh_cold_init): New function.
12121 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
12122 to seh->in_cold_section.
12123 (seh_emit_push): Record the offset of the push.
12124 (seh_emit_save): Record the offet of the save.
12125 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
12126 Test seh->after_prologue to disregard the epilogue.
12127 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
12128 (i386_pe_end_cold_function): New function.
12130 2018-03-07 Jakub Jelinek <jakub@redhat.com>
12133 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
12134 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
12137 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
12139 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
12140 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
12142 PR middle-end/84723
12143 * multiple_target.c: Include tree-inline.h and intl.h.
12144 (expand_target_clones): Diagnose and fail if node->definition and
12145 !tree_versionable_function_p (node->decl).
12147 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
12149 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
12151 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
12152 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
12153 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
12155 2018-03-06 Jakub Jelinek <jakub@redhat.com>
12158 * combine.c (try_combine): Use reg_or_subregno instead of handling
12159 just paradoxical SUBREGs and REGs.
12161 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
12163 * config/arc/arc.c (arc_finalize_pic): Remove function.
12164 (arc_must_save_register): We use single base PIC register, remove
12165 checks to save/restore the PIC register.
12166 (arc_expand_prologue): Likewise.
12167 * config/arc/arc-protos.h (arc_set_default_type_attributes):
12169 (arc_verify_short): Likewise.
12170 (arc_attr_type): Likewise.
12171 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
12172 (walk_stores): Likewise.
12173 (arc_address_cost): Make it static.
12174 (arc_verify_short): Likewise.
12175 (branch_dest): Likewise.
12176 (arc_attr_type): Likewise.
12177 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
12178 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
12179 (arc_final_prescan_insn): Remove inserting the nops due to
12180 hardware hazards. It is done in reorg step.
12181 (insn_length_variant_t): Remove.
12182 (insn_length_parameters_t): Likewise.
12183 (arc_insn_length_parameters): Likewise.
12184 (arc_get_insn_variants): Likewise.
12185 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
12187 2018-03-06 Jakub Jelinek <jakub@redhat.com>
12189 PR inline-asm/84683
12190 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
12193 PR tree-optimization/84687
12194 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
12196 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
12198 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12200 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
12201 Rename to ppc_speculation_barrier.
12202 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
12203 __builtin_ppc_speculation_barrier.
12205 2018-03-05 Jakub Jelinek <jakub@redhat.com>
12208 * combine.c (combine_simplify_rtx): Don't try to simplify if
12209 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
12212 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
12214 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
12215 to 32 bytes when compiling for POWER9.
12217 2018-03-05 Jakub Jelinek <jakub@redhat.com>
12220 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
12221 regparm >= 3 with no arg reg available also for calls with
12222 flag_force_indirect_call. Pass decl to ix86_function_regparm.
12225 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
12227 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
12229 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
12232 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
12234 2018-03-05 Richard Biener <rguenther@suse.de>
12236 PR tree-optimization/84486
12237 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
12238 When inserting a __builtin_assume_aligned call set the LHS
12239 SSA name alignment info accordingly.
12241 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
12243 PR tree-optimization/84114
12244 * config/aarch64/aarch64.c (aarch64_reassociation_width)
12245 Avoid reassociation of FLOAT_MODE addition.
12247 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
12249 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
12250 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
12251 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
12252 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
12253 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
12254 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
12255 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
12257 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
12258 __builtin_ia32_wbinvd): New builtins.
12259 (SPECIAL_ARGS2): New.
12260 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
12261 (SPECIAL_ARGS2): New.
12262 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
12263 (ix86_valid_target_attribute_inner_p): Ditto.
12264 (ix86_init_mmx_sse_builtins): Add special_args2.
12265 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
12266 TARGET_WBNOINVD_P): New.
12267 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
12268 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
12269 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
12270 * config/i386/immintrin.h (_wbinvd): New intrinsic.
12271 * config/i386/pconfigintrin.h: New file.
12272 * config/i386/wbnoinvdintrin.h: Ditto.
12273 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
12275 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
12277 2018-03-05 Richard Biener <rguenther@suse.de>
12279 PR tree-optimization/84670
12280 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
12282 (BB_VISITED_WITH_VISITED_SUCCS): New define.
12283 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
12284 (compute_antic_aux): Only assert the number of values in ANTIC_IN
12285 doesn't grow if all successors (recursively) were visited at least
12288 2018-03-05 Richard Biener <rguenther@suse.de>
12290 PR tree-optimization/84650
12291 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
12292 if executed in the loop pipeline.
12294 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
12296 * doc/configfiles.texi (Configuration Files): Move info about
12297 conditionalizing $target-protos.h to...
12298 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
12299 differs from $target-protos.h.
12301 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
12302 Chung-Ju Wu <jasonwucj@gmail.com>
12304 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
12305 * config/nds32/nds32-multiple.md (setmemsi): Define.
12306 * config/nds32/nds32-memory-manipulation.c
12307 (nds32_gen_dup_4_byte_to_word_value): New.
12308 (emit_setmem_word_loop): New.
12309 (emit_setmem_byte_loop): New.
12310 (nds32_expand_setmem_loop): New.
12311 (nds32_expand_setmem_loop_v3m): New.
12312 (nds32_expand_setmem_unroll): New.
12313 (nds32_expand_setmem): New.
12315 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12316 Chung-Ju Wu <jasonwucj@gmail.com>
12318 * config/nds32/nds32-memory-manipulation.c
12319 (nds32_emit_load_store): New.
12320 (nds32_emit_post_inc_load_store): New.
12321 (nds32_emit_mem_move): New.
12322 (nds32_emit_mem_move_block): New.
12323 (nds32_expand_movmemsi_loop_unknown_size): New.
12324 (nds32_expand_movmemsi_loop_known_size): New.
12325 (nds32_expand_movmemsi_loop): New.
12326 (nds32_expand_movmemsi_unroll): New.
12327 (nds32_expand_movmemqi): Rename ...
12328 (nds32_expand_movmemsi): ... to this.
12329 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
12330 (movmemsi): ... to this.
12331 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
12332 (nds32_expand_movmemsi): ... to this.
12334 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12335 Monk Chiang <sh.chiang04@gmail.com>
12336 Chung-Ju Wu <jasonwucj@gmail.com>
12338 * config/nds32/nds32-protos.h
12339 (nds32_expand_load_multiple): New arguments.
12340 (nds32_expand_store_multiple): Ditto.
12341 (nds32_valid_multiple_load_store): Rename ...
12342 (nds32_valid_multiple_load_store_p): ... to this.
12343 * config/nds32/nds32-memory-manipulation.c
12344 (nds32_expand_load_multiple): Refine implementation.
12345 (nds32_expand_store_multiple): Ditto.
12346 * config/nds32/nds32-multiple.md
12347 (load_multiple): Update nds32_expand_load_multiple interface.
12348 (store_multiple): Update nds32_expand_store_multiple interface.
12349 * config/nds32/nds32-predicates.c
12350 (nds32_valid_multiple_load_store): Rename ...
12351 (nds32_valid_multiple_load_store_p): ... to this and refine
12353 * config/nds32/predicates.md
12354 (nds32_load_multiple_and_update_address_operation): New predicate.
12355 (nds32_store_multiple_and_update_address_operation): New predicate.
12357 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12358 Chung-Ju Wu <jasonwucj@gmail.com>
12360 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
12361 (combo): New attribute.
12362 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
12364 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
12366 * config/nds32/nds32.opt: Change -mcmodel= default value.
12368 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
12369 Monk Chiang <sh.chiang04@gmail.com>
12370 Chung-Ju Wu <jasonwucj@gmail.com>
12372 * config/nds32/constants.md (unspec_element): New enum.
12373 * config/nds32/constraints.md (Umw): New constraint.
12374 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
12375 * config/nds32/nds32-intrinsic.md: Likewise.
12376 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
12377 (nds32_valid_smw_lwm_base_p): New.
12378 (nds32_output_smw_single_word): New.
12379 (nds32_output_lmw_single_word): New.
12380 (nds32_expand_unaligned_load): New.
12381 (nds32_expand_unaligned_store): New.
12382 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
12383 (nds32_output_smw_single_word): Declare.
12384 (nds32_output_lmw_single_word): Declare.
12385 (nds32_expand_unaligned_load): Declare.
12386 (nds32_expand_unaligned_store): Declare.
12387 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
12388 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
12389 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
12390 NDS32_BUILTIN_UASTORE_DW.
12391 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
12394 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
12395 Kito Cheng <kito.cheng@gmail.com>
12396 Chung-Ju Wu <jasonwucj@gmail.com>
12398 * config/nds32/nds32-intrinsic.c
12399 (nds32_expand_builtin_null_ftype_reg): Delete.
12400 (nds32_expand_builtin_reg_ftype_imm): Ditto.
12401 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
12402 (nds32_read_argument): New.
12403 (nds32_legitimize_target): Ditto.
12404 (nds32_legitimize_argument): Ditto.
12405 (nds32_check_constant_argument): Ditto.
12406 (nds32_expand_unop_builtin): Ditto.
12407 (nds32_expand_unopimm_builtin): Ditto.
12408 (nds32_expand_binop_builtin): Ditto.
12409 (nds32_builtin_decl_impl): Ditto.
12410 (builtin_description): Ditto.
12411 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
12412 (nds32_init_builtins_impl): Ditto.
12413 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
12414 (nds32_builtin_decl): New.
12415 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
12416 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
12418 2018-03-02 Jeff Law <law@redhat.com>
12420 * reorg.c (stop_search_p): Handle DEBUG_INSN.
12421 (redundant_insn, fill_simple_delay_slots): Likewise.
12422 (fill_slots_from_thread): Likewise.
12423 * resource.c (mark_referenced_resources): Likewise.
12424 (mark_set_resources, find_dead_or_set_registers): Likewise.
12426 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12428 * substring-locations.h (format_warning_va): Formatting fix for
12429 ATTRIBUTE_GCC_DIAG.
12430 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
12432 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
12433 * substring-locations.c: Include intl.h.
12434 (format_warning_va): Turned into small wrapper around
12435 format_warning_n_va, renamed to ...
12436 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
12437 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
12439 (format_warning_at_substring_n): New function.
12440 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
12441 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
12442 format_warning_at_substring with just a shorter name instead of
12443 const function pointer.
12444 (fmtwarn_n): New function.
12445 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
12446 appropriate, get rid of all the fmtstr temporaries, move conditionals
12447 with G_() wrapped string literals directly into fmtwarn arguments,
12448 cast dir.len to (int), formatting fixes.
12450 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
12452 * doc/invoke.texi: Remove "Cilk Plus" references.
12454 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12455 Richard Biener <rguenther@suse.de>
12458 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
12459 for error or warning attributes if CALL_FROM_THUNK_P is set.
12462 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12465 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
12466 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
12469 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
12470 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
12472 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
12473 instead of -1U in last predictors element's probability member.
12475 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
12478 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
12479 arguments if they are comparable.
12481 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
12483 PR tree-optimization/84634
12484 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
12485 masks and masked_loop_p with a single loop_masks, making sure it's
12486 null for bb vectorization.
12488 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
12490 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
12491 (vect_analyze_data_ref_access): Use loop->safe_len rather than
12492 loop->force_vectorize to check whether there is no alias.
12494 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12497 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
12499 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
12501 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
12502 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
12503 instead of a loop around prev_real_insn.
12504 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
12507 PR inline-asm/84625
12508 * config/i386/i386.c (ix86_print_operand): Use conditional
12509 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
12512 2018-03-02 Richard Biener <rguenther@suse.de>
12514 PR tree-optimization/84427
12515 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
12516 (bitmap_set_subtract_values): Rewrite to handle multiple
12519 (prune_clobbered_mems): Likewise.
12520 (phi_translate): Take edge instead of pred/phiblock.
12521 (phi_translate_1): Likewise.
12522 (phi_translate_set): Likewise. Insert all translated
12523 exprs for a value into the set, keeping possibly multiple
12524 expressions per value.
12525 (compute_antic_aux): Adjust for phi_translate changes.
12526 When intersecting union the expressions and prune those
12527 not in the final value set, keeping possibly multiple
12528 expressions per value. Do not use value-insertion
12529 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
12530 all expressions. Add verification that the value-sets
12531 only shrink during iteration.
12532 (compute_partial_antic_aux): Adjust for the phi_translate changes.
12533 (do_pre_regular_insertion): Likewise.
12534 (do_pre_partial_partial_insertion): Likewise.
12536 2018-03-02 Richard Biener <rguenther@suse.de>
12539 * config/darwin.c (saved_debug_info_level): New static global.
12540 (darwin_asm_lto_start): Disable debug info generation for LTO out.
12541 (darwin_asm_lto_end): Restore debug info generation settings.
12543 2018-03-01 Martin Liska <mliska@suse.cz>
12546 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12547 volatile arguments.
12549 2018-03-01 Richard Biener <rguenther@suse.de>
12552 * dwarf2out.c (gen_variable_die): Properly handle late VLA
12553 type annotation with LTO when debug was disabled at compile-time.
12555 2018-03-01 Matthew Fortune <mfortune@gmail.com>
12557 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
12559 (mips_final_postscan_insn): Likewise.
12561 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
12563 PR rtl-optimization/84528
12564 * alias.c (init_alias_target): Add commentary.
12565 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
12566 a unique base value if the frame pointer is not eliminated
12567 to the stack pointer.
12569 2018-03-01 Tom de Vries <tom@codesourcery.com>
12571 PR rtl-optimization/83327
12572 * lra-int.h (hard_regs_spilled_into): Declare.
12573 * lra.c (hard_regs_spilled_into): Define.
12574 (init_reg_info): Init hard_regs_spilled_into.
12575 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
12576 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
12577 (process_bb_lives): Handle hard_regs_spilled_into.
12578 (lra_create_live_ranges_1): Before doing liveness propagation, clear
12579 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
12581 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
12583 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
12584 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
12585 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
12586 * config/rs6000/aix72.h: New file.
12588 2018-02-28 Jakub Jelinek <jakub@redhat.com>
12590 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
12591 instead of warning_at with conditional singular and plural messages
12595 * stor-layout.c (update_alignment_for_field): For
12596 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
12597 && !DECL_PACKED (field), do the alignment update, just use
12598 only desired_align instead of MAX (type_align, desired_align)
12600 (place_field): Don't do known_align < desired_align handling
12601 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
12602 is non-NULL, instead do it after rli->prev_field handling and
12603 only if not within a bitfield word. For DECL_PACKED (field)
12604 use type_align of BITS_PER_UNIT.
12606 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
12608 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
12609 superfluous parentheses and trailing spaces.
12611 2018-02-28 Richard Biener <rguenther@suse.de>
12613 PR tree-optimization/84584
12614 * graphite-scop-detection.c (scop_detection::add_scop): Discard
12615 SCoPs with fake exit edge.
12617 2018-02-28 Martin Liska <mliska@suse.cz>
12620 * timevar.c (timer::print): Fix format to properly print 100%
12623 2018-02-28 Richard Biener <rguenther@suse.de>
12625 PR middle-end/84607
12626 * genmatch.c (capture_info::walk_match): Do not mark
12627 captured expressions without operands as expr_p given
12628 they act more like predicates and should be subject to
12629 "lost tail" side-effect preserving.
12631 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
12633 PR rtl-optimization/81611
12634 * auto-inc-dec.c (attempt_change): Move dead note from
12635 mem_insn if it's the next use of regno
12636 (find_address): Take address use of reg holding
12637 non-incremented value. Add parm to limit search to the named
12639 (merge_in_block): Attempt to use a mem insn that is the next
12640 use of the original regno.
12642 2018-02-27 Martin Sebor <msebor@redhat.com>
12645 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
12646 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
12648 2018-02-27 Martin Sebor <msebor@redhat.com>
12650 PR translation/84207
12651 * diagnostic-core.h (warning_n, error_n, inform_n): Change
12652 n argument to unsigned HOST_WIDE_INT.
12653 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
12654 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
12655 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
12656 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
12658 2018-02-27 Richard Biener <rguenther@suse.de>
12660 PR tree-optimization/84512
12661 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
12662 Do not use the estimate returned from record_stmt_cost for
12663 the scalar iteration cost but sum properly using add_stmt_cost.
12665 2018-02-27 Richard Biener <rguenther@suse.de>
12667 PR tree-optimization/84466
12668 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12669 Adjust last change to less strictly validate use operands.
12671 2018-02-27 Martin Liska <mliska@suse.cz>
12673 PR gcov-profile/84548
12674 * gcov.c (process_file): Allow partial overlap and consider it
12675 also as group functions.
12676 (output_lines): Properly calculate range of lines for a group.
12678 2018-02-27 Martin Liska <mliska@suse.cz>
12680 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
12681 'ggc' suffixes. Change first column width.
12682 (timer::print): Fix formatting of the column.
12684 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
12686 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
12687 preserve inline entry blocks for the sake of debug inline
12688 entry point markers alone.
12689 (remove_unused_locals): Suggest in comments a better place to
12690 force the preservation of inline entry blocks that are
12691 otherwise unused, but do not preserve them.
12693 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12695 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
12697 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12700 * config/i386/constraints.md (Bs): Replace
12701 ix86_indirect_branch_register with
12702 TARGET_INDIRECT_BRANCH_REGISTER.
12704 * config/i386/i386.md (indirect_jump): Likewise.
12705 (tablejump): Likewise.
12706 (*sibcall_memory): Likewise.
12707 (*sibcall_value_memory): Likewise.
12708 Peepholes of indirect call and jump via memory: Likewise.
12709 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
12710 (*sibcall_value_GOT_32): Likewise.
12711 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12712 (GOT_memory_operand): Likewise.
12713 (call_insn_operand): Likewise.
12714 (sibcall_insn_operand): Likewise.
12715 (GOT32_symbol_operand): Likewise.
12716 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
12718 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12720 PR rtl-optimization/83496
12721 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
12722 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
12723 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
12724 redundant insn, if any.
12725 (relax_delay_slots): Likewise.
12726 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
12728 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
12730 PR tree-optimization/83965
12731 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
12732 that grouped statements are part of a reduction chain. Return
12733 true if the statement is not marked as a reduction itself but
12734 is part of a group.
12735 (vect_recog_dot_prod_pattern): Don't check whether the statement
12736 is part of a group here.
12737 (vect_recog_sad_pattern): Likewise.
12738 (vect_recog_widen_sum_pattern): Likewise.
12740 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12743 * final.c (rest_of_clean_state): Also look for calls inside sequences.
12745 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12748 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
12750 (ix86_output_indirect_function_return): New prototype.
12751 (ix86_split_simple_return_pop_internal): Likewise.
12752 * config/i386/i386.c (indirect_return_via_cx): New.
12753 (indirect_return_via_cx_bnd): Likewise.
12754 (indirect_thunk_name): Handle return va CX_REG.
12755 (output_indirect_thunk_function): Create alias for
12756 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
12757 (ix86_output_indirect_jmp): Remove the bool argument.
12758 (ix86_output_indirect_function_return): New function.
12759 (ix86_split_simple_return_pop_internal): Likewise.
12760 * config/i386/i386.md (*indirect_jump): Don't pass false
12761 to ix86_output_indirect_jmp.
12762 (*tablejump_1): Likewise.
12763 (simple_return_pop_internal): Change it to define_insn_and_split.
12764 Call ix86_split_simple_return_pop_internal to split it for
12765 -mfunction-return=.
12766 (simple_return_indirect_internal): Call
12767 ix86_output_indirect_function_return instead of
12768 ix86_output_indirect_jmp.
12770 2018-02-26 Jakub Jelinek <jakub@redhat.com>
12773 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
12774 memset and value initialization afterwards.
12776 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
12778 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
12780 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12783 * common/config/aarch64/aarch64-common.c
12784 (aarch_option_optimization_table[]): Switch
12785 off fomit-frame-pointer
12787 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12788 Chung-Ju Wu <jasonwucj@gmail.com>
12790 * config/nds32/nds32-multiple.md (load_multiple): Disallow
12792 (store_multiple): Ditto.
12794 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12796 * config.gcc: Add --with-cpu support for nds32 target.
12797 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
12798 * config/nds32/nds32.opt: Add -mcpu= option.
12800 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
12802 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
12803 isel=yes): Warn for these deprecated options.
12805 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
12807 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
12808 ISA_2_5_MASKS_EMBEDDED.
12810 2018-02-23 Jakub Jelinek <jakub@redhat.com>
12812 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
12813 p->max as pointers rather than using iterative_hash_expr.
12815 2018-02-23 Carl Love <cel@us.ibm.com>
12817 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
12818 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
12820 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
12821 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
12822 P8V_BUILTIN_VEC_VUNSIGNED2.
12824 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
12827 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
12828 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
12830 * lra-constraints.c (curr_insn_transform): Set up
12831 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
12833 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
12834 * lra-eliminations.c (spill_pseudos): Ditto.
12835 (process_insn_for_elimination): Ditto.
12836 * lra-lives.c (reg_early_clobber_p): Use the new macros.
12837 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
12838 LRA_NON_CLOBBERED_ALT.
12840 2018-02-22 Martin Sebor <msebor@redhat.com>
12842 PR tree-optimization/84480
12843 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
12844 to maybe_diag_stxncpy_trunc. Call it.
12845 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
12846 from gimple_fold_builtin_strcpy. Print inlining stack.
12847 (handle_builtin_stxncpy): Print inlining stack.
12848 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
12850 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
12853 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
12854 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
12855 and -fcheck-pointer-bounds are used together.
12856 (indirect_thunk_prefix): New enum.
12857 (indirect_thunk_need_prefix): New function.
12858 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
12859 "_nt" instead of "_bnd" for NOTRACK prefix.
12860 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
12861 (output_indirect_thunk_function): Likewise.
12863 (ix86_code_end): Update output_indirect_thunk_function calls.
12864 (ix86_output_indirect_branch_via_reg): Replace
12865 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
12866 (ix86_output_indirect_branch_via_push): Likewise.
12867 (ix86_output_function_return): Likewise.
12868 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
12869 incompatible with -fcf-protection=branch and
12870 -fcheck-pointer-bounds.
12872 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12875 * config/aarch64/aarch64.c (aarch64_print_address_internal):
12876 Change gcc_assert call to output_operand_lossage.
12878 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12880 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
12882 2018-02-22 DJ Delorie <dj@redhat.com>
12883 Sebastian Perta <sebastian.perta@renesas.com>
12884 Oleg Endo <olegendo@gcc.gnu.org>
12886 * config/rx/rx.c (rx_rtx_costs): New function.
12887 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
12889 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
12891 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
12893 2018-02-22 Martin Liska <mliska@suse.cz>
12896 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
12897 Add "native" as a possible value.
12899 2018-02-22 Martin Liska <mliska@suse.cz>
12902 * config/i386/i386.c (ix86_option_override_internal):
12903 Add "native" as a possible value for -march and -mtune.
12905 2018-02-22 Jakub Jelinek <jakub@redhat.com>
12908 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
12909 to all type variants.
12911 PR tree-optimization/84503
12912 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
12913 width as info->bitpos + info->bitsize - start.
12914 (merged_store_group::merge_overlapping): Simplify width computation.
12915 (check_no_overlap): New function.
12916 (imm_store_chain_info::try_coalesce_bswap): Compute expected
12917 start + width and last_order of the group, fail if check_no_overlap
12919 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
12920 to group if check_no_overlap fails.
12922 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12924 * config/rs6000/altivec.md: Delete contraint arguments to
12925 define_expand, define_split, and define_peephole2, and in
12926 define_insn_and_split if always unused.
12927 * config/rs6000/darwin.md: Ditto.
12928 * config/rs6000/dfp.md: Ditto.
12929 * config/rs6000/rs6000.md: Ditto.
12930 * config/rs6000/sync.md: Ditto.
12931 * config/rs6000/vector.md: Ditto.
12932 * config/rs6000/vsx.md: Ditto.
12934 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12936 * config/rs6000/altivec.md: Write output control strings as braced
12937 blocks instead of double-quoted strings.
12938 * config/rs6000/darwin.md: Ditto.
12939 * config/rs6000/rs6000.md: Ditto.
12940 * config/rs6000/vector.md: Ditto.
12941 * config/rs6000/vsx.md: Ditto.
12943 2018-02-21 Jason Merrill <jason@redhat.com>
12945 PR c++/84314 - ICE with templates and fastcall attribute.
12946 * attribs.c (build_type_attribute_qual_variant): Remove assert.
12948 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12950 * ipa-cp.c (determine_versionability): Fix comment typos.
12952 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12955 * ipa-cp.c (determine_versionability): Do not version functions caling
12958 2018-02-21 Martin Liska <mliska@suse.cz>
12961 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
12962 Add "native" as a possible value.
12963 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
12964 the macro when native cpu detection is available.
12966 2018-02-21 Martin Liska <mliska@suse.cz>
12969 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
12970 Add "native" as a possible value.
12971 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
12972 when native cpu detection is available.
12974 2018-02-21 Jakub Jelinek <jakub@redhat.com>
12975 Martin Sebor <msebor@redhat.com>
12977 PR tree-optimization/84478
12978 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
12980 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
12981 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
12982 support which is conservatively correct, for 2 only stay conservative
12983 for maxlen. Formatting and comment capitalization fixes. Add STRICT
12984 argument to the 2 argument get_range_strlen, adjust 6 arg
12985 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
12987 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
12988 (gimple_fold_builtin_strlen): Pass true as last argument to
12991 2018-02-20 Martin Sebor <msebor@redhat.com>
12993 PR middle-end/84095
12994 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
12995 (builtin_memref::set_base_and_offset): Same. Handle inner references.
12996 (builtin_memref::builtin_memref): Factor out parts into
12997 set_base_and_offset and call it.
12999 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
13001 PR middle-end/84406
13002 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
13003 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
13004 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
13005 search at the associated MODE_INT.
13007 2018-02-20 Jeff Law <law@redhat.com>
13009 PR middle-end/82123
13010 PR tree-optimization/81592
13011 PR middle-end/79257
13012 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
13013 for range data rather than using global data.
13014 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
13015 range data rather than using global data.
13016 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
13017 pass it to children as needed.
13018 (struct directive::fmtresult): Similarly.
13019 (struct directive::set_width): Similarly.
13020 (struct directive::set_precision): Similarly.
13021 (format_integer, format_directive, parse_directive): Similarly.
13022 (format_none): Accept unnamed vr_values parameter.
13023 (format_percent, format_floating, format_character): Similarly.
13024 (format_string, format_plain): Similarly.
13025 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
13026 the EVRP range analyzer for range data rather than using global data.
13027 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
13028 gimple-ssa-evrp-analyze.h
13029 (class sprintf_dom_walker): Add after_dom_children member function.
13030 Add evrp_range_analyzer member.
13031 (sprintf_dom_walker::before_dom_children): Call into the EVRP
13032 range analyzer as needed.
13033 (sprintf_dom_walker::after_dom_children): New member function.
13034 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
13036 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
13037 (evrp_range_analyzer::pop_to_marker): Likewise.
13039 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
13041 PR tree-optimization/84419
13042 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
13043 with the required type if its current type is compatible but
13046 2018-02-20 Jakub Jelinek <jakub@redhat.com>
13048 PR middle-end/82004
13049 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
13050 after vectorization.
13052 2018-02-20 Martin Liska <mliska@suse.cz>
13055 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
13056 possible values if we don't have a hint.
13058 2018-02-20 Martin Liska <mliska@suse.cz>
13062 * final.c (shorten_branches): Build align_tab array with one
13064 * opts.c (finish_options): Add alignment option limit check.
13065 (MAX_CODE_ALIGN): Likewise.
13066 (MAX_CODE_ALIGN_VALUE): Likewise.
13067 * doc/invoke.texi: Document maximum allowed option value for
13068 all -falign-* options.
13070 2018-02-19 Jakub Jelinek <jakub@redhat.com>
13073 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
13074 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
13075 * var-tracking.c (emit_note_insn_var_location): Remove all references
13076 to NOTE_INSN_CALL_ARG_LOCATION.
13077 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
13078 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
13079 Use copy_rtx_if_shared.
13080 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
13081 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
13082 (dwarf2out_var_location): Remove handling of
13083 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
13085 * final.c (final_scan_insn): Remove all references to
13086 NOTE_INSN_CALL_ARG_LOCATION.
13087 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
13088 before dumping final insns.
13089 * except.c (emit_note_eh_region_end): Remove all references to
13090 NOTE_INSN_CALL_ARG_LOCATION.
13091 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
13092 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
13093 * config/arc/arc.c (hwloop_optimize): Likewise.
13094 * config/arm/arm.c (create_fix_barrier): Likewise.
13095 * config/s390/s390.c (s390_chunkify_start): Likewise.
13096 * config/sh/sh.c (find_barrier): Likewise.
13097 * config/i386/i386.c (rest_of_insert_endbranch,
13098 ix86_seh_fixup_eh_fallthru): Likewise.
13099 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
13100 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
13101 * config/frv/frv.c (frv_function_prologue): Likewise.
13102 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
13104 (note_outside_basic_block_p): Remove all references to
13105 NOTE_INSN_CALL_ARG_LOCATION.
13106 * gengtype.c (adjust_field_rtx_def): Likewise.
13107 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
13109 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
13110 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
13113 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
13116 PR tree-optimization/84452
13117 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
13118 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
13121 2018-02-19 Martin Liska <mliska@suse.cz>
13124 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
13126 2018-02-19 Martin Liska <mliska@suse.cz>
13127 Richard Sandiford <richard.sandiford@linaro.org>
13129 PR tree-optimization/82491
13130 * gimple-fold.c (get_base_constructor): Make earlier bail out
13133 2018-02-19 Carl Love <cel@us.ibm.com>
13135 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
13136 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
13138 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
13139 P8V_BUILTIN_VEC_NEG.
13141 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
13143 * config/rl78/rl78.md (movdf): New define expand.
13145 2018-02-19 Martin Liska <mliska@suse.cz>
13148 * doc/invoke.texi: Fix typo.
13149 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
13151 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
13153 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
13154 handle rs6000_single_float and rs6000_double_float specially for
13157 2018-02-16 Jeff Law <law@redhat.com>
13159 * config/rx/rx.c (add_pop_cfi_notes): New function.;
13160 (pop_regs): Use it.
13162 2018-02-16 Jakub Jelinek <jakub@redhat.com>
13165 * ipa-inline.c (inline_small_functions): Fix a typo.
13167 2018-02-16 Nathan Sidwell <nathan@acm.org>
13169 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
13171 2018-02-16 Carl Love <cel@us.ibm.com>
13173 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
13174 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
13175 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
13176 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
13177 expansion to P8V_BUILTIN_VEC_FLOAT2.
13179 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
13181 PR rtl-optimization/70023
13182 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
13183 src_regno into account.
13185 2018-02-16 Carl Love <cel@us.ibm.com>
13187 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
13188 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
13189 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
13190 * config/rs6000/rs6000.c: Remove case statements for
13191 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
13192 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
13193 and P9V_BUILTIN_VEC_VINSERT4B.
13194 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
13195 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
13196 * config/rs6000/vsx.md:
13197 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
13200 2018-02-16 Carl Love <cel@us.ibm.com>
13202 * config/rs6000/altivec.h: Add builtin names vec_extract4b
13204 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
13206 * config/rs6000/rs6000-c.c: Add the definitions for
13207 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
13208 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
13209 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
13210 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
13211 definition for insert4b and define insn *insert3b_internal.
13212 * doc/extend.texi: Add documentation for vec_extract4b.
13214 2018-02-16 Nathan Sidwell <nathan@acm.org>
13216 * doc/extend.texi (Backwards Compatibility): Mention friend
13217 injection. Note for-scope is deprecated.
13218 * doc/invoke.texi (-ffriend-injection): Deprecate.
13220 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
13222 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
13223 that moved to I2, also allow destinations that are a paradoxical
13224 subreg (instead of a normal reg).
13226 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
13229 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
13232 2018-02-16 Richard Biener <rguenther@suse.de>
13234 PR tree-optimization/84037
13235 PR tree-optimization/84016
13237 * config/i386/i386.c (ix86_builtin_vectorization_cost):
13238 Adjust vec_construct for the fact we need additional higher latency
13239 128bit inserts for AVX256 and AVX512 vector builds.
13240 (ix86_add_stmt_cost): Scale vector construction cost for
13243 2018-02-16 Richard Biener <rguenther@suse.de>
13245 PR tree-optimization/84417
13246 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
13247 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
13248 (non_rewritable_lvalue_p): Likewise, use poly-ints.
13250 2018-02-16 Martin Liska <mliska@suse.cz>
13253 * internal-fn.def (ASAN_CHECK): Set proper flags.
13254 (ASAN_MARK): Likewise.
13256 2018-02-16 Julia Koval <julia.koval@intel.com>
13258 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
13259 from PTA_CANNONLAKE.
13261 2018-02-16 Jakub Jelinek <jakub@redhat.com>
13264 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
13265 Use ++iter rather than iter++ for std::list iterators.
13266 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
13267 defer deleting them until all nodes in the forest are processed. Do
13268 free even leaf nodes. Change to_process into auto_vec.
13271 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
13272 * vec.h (vec_default_construct): Use memset instead of placement new
13273 if BROKEN_VALUE_INITIALIZATION is defined.
13274 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
13275 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
13278 PR rtl-optimization/83723
13279 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
13280 * lra.c (lra_substitute_pseudo): Likewise. If true, use
13281 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
13283 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
13285 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
13287 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
13289 PR rtl-optimization/81443
13290 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
13291 from inner REGs to paradoxical SUBREGs.
13293 2018-02-16 Richard Biener <rguenther@suse.de>
13295 PR tree-optimization/84399
13296 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
13297 For operands we can analyze at their definition make sure we can
13298 analyze them at each use as well.
13300 2018-02-16 Richard Biener <rguenther@suse.de>
13302 PR tree-optimization/84190
13303 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
13304 volatile accesses if the decl isn't volatile.
13306 2018-02-15 Jason Merrill <jason@redhat.com>
13308 PR c++/84314 - ICE with templates and fastcall attribute.
13309 * attribs.c (build_type_attribute_qual_variant): Don't clobber
13310 TYPE_CANONICAL on an existing type.
13312 2018-02-15 Jakub Jelinek <jakub@redhat.com>
13314 PR tree-optimization/84383
13315 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
13316 dstoff nor call operand_equal_p if dstbase is NULL.
13318 PR tree-optimization/84334
13319 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
13320 also a CONSTANT_CLASS_P, punt.
13322 2018-02-14 Jim Wilson <jimw@sifive.com>
13324 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
13325 first SMALL_OPERAND check. New local min_second_step. Move assert
13326 to where locals are set. Add TARGET_RVC support.
13327 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
13329 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
13331 * doc/invoke.texi: Correct -Wformat-overflow code sample.
13333 2018-02-14 Martin Sebor <msebor@redhat.com>
13335 PR tree-optimization/83698
13336 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
13337 arrays constrain the offset range to their bounds.
13338 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
13339 (builtin_access::overlap): Avoid setting the size of overlap if it's
13341 (maybe_diag_overlap): Also consider arrays when deciding what values
13342 of offsets to include in diagnostics.
13344 2018-02-14 Martin Sebor <msebor@redhat.com>
13347 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
13348 that correspond to the kind of a declaration.
13350 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
13353 * config/pa/pa.md: Load address of PIC label using the linkage table
13354 if the label is nonlocal.
13356 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
13358 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
13359 warning message if user requests -maltivec=be.
13360 * doc/invoke.texi: Document deprecation of -maltivec=be.
13362 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
13365 * config/rs6000/rs6000-c.c: Update definitions for
13366 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
13367 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
13369 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13372 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
13373 add _get_ssp intrinsics. Remove argument from
13374 __builtin_ia32_rdssp[d|q].
13375 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
13376 * config/i386/i386-builtin.def: Remove argument from
13377 __builtin_ia32_rdssp[d|q].
13378 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
13379 ix86_expand_special_args_builtin for _rdssp[d|q].
13380 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
13381 Clear register before usage.
13382 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
13383 Add documentation for new _get_ssp and _inc_ssp intrinsics.
13385 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
13387 PR tree-optimization/84357
13388 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
13389 operand 1 of an ARRAY_REF too.
13391 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
13394 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
13395 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
13397 (set_of_reg): New struct.
13398 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
13399 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
13400 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
13402 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
13403 Split into bitclr, bitset, bitinvert patterns if appropriate.
13404 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
13405 use rx_fuse_in_memory_bitop.
13406 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
13407 to named insn, correct maximum insn length.
13409 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
13412 * machmode.def: Define a complex mode for PARTIAL_INT.
13413 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
13415 * doc/rtl.texi: Document CSPImode.
13416 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
13418 (msp430_hard_regno_nregs_with_padding): Likewise.
13420 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
13423 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
13425 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
13427 PR rtl-optimization/84169
13428 * combine.c (try_combine): New variable split_i2i3. Set it to true if
13429 we generated a parallel as new i3 and we split that to new i2 and i3
13430 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
13431 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
13432 those to i2, not i1. Partially rewrite this scan code.
13434 2018-02-13 Jakub Jelinek <jakub@redhat.com>
13437 * stor-layout.c (place_field): For variable length fields, adjust
13438 offset_align afterwards not just based on the field's alignment,
13439 but also on the size.
13441 PR middle-end/84309
13442 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
13443 of exps and logs in the use_exp2 case.
13445 2018-02-13 Jeff Law <law@redhat.com>
13447 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
13448 entry for "vector".
13450 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
13453 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
13457 * common.opt (gas-loc-support, gas-locview-support): New.
13458 (ginline-points, ginternal-reset-location-views): New.
13459 * doc/invoke.texi: Document them. Use @itemx where intended.
13460 (gvariable-location-views): Adjust.
13461 * target.def (reset_location_view): New.
13462 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
13463 (TARGET_RESET_LOCATION_VIEW): New.
13464 * doc/tm.texi: Rebuilt.
13465 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
13466 (dwarf2out_default_as_locview_support): New.
13467 (output_asm_line_debug_info): Use option variables.
13468 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
13469 (output_loc_list): Likewise.
13470 (add_high_low_attributes): Check option variables.
13471 Don't output entry view attribute in strict mode.
13472 (gen_inlined_subroutine_die): Check option variables.
13473 (dwarf2out_inline_entry): Likewise.
13474 (init_sections_and_labels): Likewise.
13475 (dwarf2out_early_finish): Likewise.
13476 (maybe_reset_location_view): New, from...
13477 (dwarf2out_var_location): ... here. Call it.
13478 * debug.h (dwarf2out_default_as_loc_support): Declare.
13479 (dwarf2out_default_as_locview_support): Declare.
13480 * hooks.c (hook_int_rtx_insn_0): New.
13481 * hooks.h (hook_int_rtx_insn_0): Declare.
13482 * toplev.c (process_options): Take -gas-loc-support and
13483 -gas-locview-support from dwarf2out. Enable
13484 -gvariable-location-views by default only with locview
13485 assembler support. Enable -ginternal-reset-location-views by
13486 default only if the target defines the corresponding hook.
13487 Enable -ginline-points by default if location views are
13488 enabled; force it disabled if statement frontiers are
13490 * tree-inline.c (expand_call_inline): Check option variables.
13491 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
13493 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
13495 PR tree-optimization/84321
13496 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
13497 handling. Also check whether the anti-range contains any values
13498 that satisfy the mask; switch to a VR_RANGE if not.
13500 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
13503 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
13505 2018-02-13 Martin Jambor <mjambor@suse.cz>
13508 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
13509 of call statements, also set location of a load to a temporary.
13511 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
13513 * config/rl78/rl78.c (add_vector_labels): New function.
13514 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
13515 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
13516 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
13517 which checks that no arguments are passed.
13518 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
13519 * doc/extend.texi: Documentation for the new attribute.
13521 2018-02-13 Andreas Schwab <schwab@suse.de>
13523 * config/riscv/linux.h (CPP_SPEC): Define.
13525 2018-02-13 Jakub Jelinek <jakub@redhat.com>
13528 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
13529 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
13530 OPTION_MASK_ISA_AES as first argument to def_builtin_const
13531 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
13532 instead of OPTION_MASK_ISA_PCLMUL as first argument to
13533 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
13534 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
13535 temporarily for AES and PCLMUL builtins.
13537 PR tree-optimization/84339
13538 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
13539 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
13542 PR middle-end/84309
13543 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
13544 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
13545 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
13547 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
13549 * omp-simd-clone.h: New file.
13550 * omp-simd-clone.c: Include omp-simd-clone.h.
13551 (expand_simd_clones): No longer static.
13552 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
13553 cgraph.h and omp-simd-clone.h.
13554 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
13555 (vect_recog_widen_shift_pattern): Formatting fix.
13556 (vect_pattern_recog_1): Don't check optab for calls.
13559 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
13560 operands[2] into a REG before using gen_lowpart on it.
13562 2018-02-12 Jeff Law <law@redhat.com>
13565 * config/sh/sh.c (find_barrier): Consider a sibling call
13568 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
13569 successfully back substituting a reg.
13571 2018-02-12 Richard Biener <rguenther@suse.de>
13573 PR tree-optimization/84037
13574 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
13575 parameter, move visited init to caller.
13576 (vect_slp_analyze_operations): Separate cost from validity
13577 check, initialize visited once for all instances.
13578 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
13580 * tree-vect-stmts.c (vect_model_simple_cost): Make early
13582 (vect_model_promotion_demotion_cost): Likewise.
13583 (vectorizable_bswap): Guard cost modeling with !slp_node
13584 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
13586 (vectorizable_call): Likewise.
13587 (vectorizable_conversion): Likewise.
13588 (vectorizable_assignment): Likewise.
13589 (vectorizable_shift): Likewise.
13590 (vectorizable_operation): Likewise.
13591 (vectorizable_store): Likewise.
13592 (vectorizable_load): Likewise.
13593 (vectorizable_condition): Likewise.
13594 (vectorizable_comparison): Likewise.
13596 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
13599 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
13600 (ASAN_MARK): Fix fnspec to account for return value, change pointer
13601 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
13603 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13605 PR middle-end/83665
13606 * params.def (inline-min-speedup): Increase from 8 to 15.
13607 (max-inline-insns-auto): Decrease from 40 to 30.
13608 * ipa-split.c (consider_split): Add some buffer for function to
13609 be considered inlining candidate.
13610 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
13613 2018-02-12 Richard Biener <rguenther@suse.de>
13615 PR tree-optimization/84037
13616 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
13617 matched stmts if we cannot swap the non-matched ones.
13619 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
13621 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
13622 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
13623 _mm_maskz_scalef_round_ss): New intrinsics.
13624 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
13625 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
13626 __builtin_ia32_scalefss_round): Remove.
13627 (__builtin_ia32_scalefsd_mask_round,
13628 __builtin_ia32_scalefss_mask_round): New intrinsics.
13629 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
13630 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
13631 ((match_operand:VF_128 2 "<round_nimm_predicate>"
13632 "<round_constraint>")): Changed to ...
13633 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
13634 "<round_scalar_constraint>")): ... this.
13635 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
13636 %0, %1, %2<round_op3>}"): Changed to ...
13637 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
13638 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
13639 %2<round_scalar_mask_op3>}"): ... this.
13640 * config/i386/subst.md (round_scalar_nimm_predicate): New.
13642 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
13644 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
13645 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
13646 (_mm_maskz_sqrt_round_ss): New intrinsics.
13647 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
13648 (__builtin_ia32_sqrtsd_mask_round)
13649 (__builtin_ia32_sqrtss_mask_round): New builtins.
13650 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
13651 (__builtin_ia32_sqrtss_round): Remove.
13652 (__builtin_ia32_sqrtsd_mask_round)
13653 (__builtin_ia32_sqrtss_mask_round): New builtins.
13654 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
13655 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
13656 ((match_operand:VF_128 1 "vector_operand"
13657 "xBm,<round_constraint>")): Changed to ...
13658 ((match_operand:VF_128 1 "vector_operand"
13659 "xBm,<round_scalar_constraint>")): ... this.
13660 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
13661 %0, %2, %<iptr>1<round_op3>}): Changed to ...
13662 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
13663 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
13664 %<iptr>1<round_scalar_mask_op3>}): ... this.
13665 ((set_attr "prefix" "<round_prefix>")): Changed to ...
13666 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
13668 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
13671 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
13672 Cast vec_cmpeq result to correct type.
13673 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
13674 Cast vec_cmpgt result to correct type.
13676 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
13678 * final.c (final_scan_insn_1): Renamed from...
13679 (final_scan_insn): ... this. New wrapper, to recover
13680 seen from the outermost call in recursive ones.
13681 * config/sparc/sparc.c (output_return): Drop seen from call.
13682 (output_sibcall): Likewise.
13683 * config/visium/visium.c (output_branch): Likewise.
13685 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
13687 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
13690 2018-02-10 Alan Modra <amodra@gmail.com>
13693 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
13694 Specify LR as an input.
13696 2018-02-10 Jakub Jelinek <jakub@redhat.com>
13699 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
13700 remove_member_access_dummy_vars): New functions.
13701 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
13702 lower_omp_1, execute_lower_omp): Use them.
13704 PR rtl-optimization/84308
13705 * shrink-wrap.c (spread_components): Release todo vector.
13707 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
13709 PR rtl-optimization/57193
13710 * ira-color.c (struct allocno_color_data): Add member
13711 conflict_allocno_hard_prefs.
13712 (update_conflict_allocno_hard_prefs): New.
13713 (bucket_allocno_compare_func): Add a preference based on
13714 conflict_allocno_hard_prefs.
13715 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
13716 (color_allocnos): Remove a dead code. Initiate
13717 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
13719 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13722 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
13723 constraint from =wa to wa. Avoid a subreg on the output operand,
13724 instead use a pseudo and subreg it in a move.
13725 (p9_xxbrd_<mode>): Changed to ...
13726 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
13727 (p9_xxbrd_v2df): New expander.
13728 (p9_xxbrw_<mode>): Changed to ...
13729 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
13730 (p9_xxbrw_v4sf): New expander.
13732 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13734 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
13736 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
13739 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
13740 multiply in 32-bit mode.
13741 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
13742 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
13745 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13747 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
13748 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
13749 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
13750 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
13752 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
13755 * dwarf2out.c (is_trivial_indirect_ref): New function.
13756 (dwarf2out_late_global_decl): Do not generate a location
13757 attribute for variables that have a non-trivial DECL_VALUE_EXPR
13758 and that are not defined in the current unit.
13760 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13762 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
13763 instead of a libcall for UNORDERED.
13765 2018-02-09 Tamar Christina <tamar.christina@arm.com>
13768 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
13769 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
13771 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13774 * config/s390/s390.c (s390_set_current_function): Invoke
13775 s390_indirect_branch_settings also if fndecl didn't change.
13777 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13779 * config/rs6000/rs6000.md (blockage): Set length to zero.
13781 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13783 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
13785 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13788 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
13789 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
13793 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
13794 PARALLEL incoming that failed vt_get_decl_and_offset check.
13796 PR middle-end/84237
13797 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
13798 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
13800 (get_variable_section): For decls in named .bss* sections pass true as
13801 second argument to bss_initializer_p.
13803 2018-02-09 Marek Polacek <polacek@redhat.com>
13804 Jakub Jelinek <jakub@redhat.com>
13807 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
13808 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
13809 Sync some changes from cxx_fold_indirect_ref.
13811 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13813 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
13815 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
13816 (BLOCK_INLINE_ENTRY_LABEL): New.
13817 (dwarf2out_var_location): Disregard inline entry markers.
13818 (inline_entry_data): New struct.
13819 (inline_entry_data_hasher): New hashtable type.
13820 (inline_entry_data_hasher::hash): New.
13821 (inline_entry_data_hasher::equal): New.
13822 (inline_entry_data_table): New variable.
13823 (add_high_low_attributes): Add DW_AT_entry_pc and
13824 DW_AT_GNU_entry_view attributes if a pending entry is found
13825 in inline_entry_data_table. Add old entry_pc attribute only
13826 if debug nonbinding markers are disabled.
13827 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
13828 markers are enabled.
13829 (block_within_block_p, dwarf2out_inline_entry): New.
13830 (dwarf2out_finish): Check that no entries remained in
13831 inline_entry_data_table.
13832 * final.c (reemit_insn_block_notes): Handle inline entry notes.
13833 (final_scan_insn, notice_source_line): Likewise.
13834 (rest_of_clean_state): Skip inline entry markers.
13835 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
13837 * gimple.c (gimple_build_debug_inline_entry): New.
13838 * gimple.h (enum gimple_debug_subcode): Add
13839 GIMPLE_DEBUG_INLINE_ENTRY.
13840 (gimple_build_debug_inline_entry): Declare.
13841 (gimple_debug_inline_entry_p): New.
13842 (gimple_debug_nonbind_marker_p): Adjust.
13843 * insn-notes.def (INLINE_ENTRY): New.
13844 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
13845 inline entry marker notes.
13846 (print_insn): Likewise.
13847 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
13848 (INSN_DEBUG_MARKER_KIND): Likewise.
13849 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
13850 * tree-inline.c (expand_call_inline): Build and insert
13851 debug_inline_entry stmt.
13852 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
13853 inline entry blocks early, if nonbind markers are enabled.
13854 (dump_scope_block): Dump fragment info.
13855 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
13856 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
13857 (gimple_build_debug_inline_entry): New.
13858 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
13859 Enable/disable inline entry points too.
13860 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
13861 (DEBUG_INSN): Describe inline entry markers.
13863 * common.opt (gvariable-location-views): New.
13864 (gvariable-location-views=incompat5): New.
13865 * config.in: Rebuilt.
13866 * configure: Rebuilt.
13867 * configure.ac: Test assembler for view support.
13868 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
13869 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
13870 * dwarf2out.c (var_loc_view): New typedef.
13871 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
13872 (dwarf2out_locviews_in_attribute): New.
13873 (dwarf2out_locviews_in_loclist): New.
13874 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
13875 (enum dw_line_info_opcode): Add LI_adv_address.
13876 (struct dw_line_info_table): Add view.
13877 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
13878 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
13879 (zero_view_p): New variable.
13880 (ZERO_VIEW_P): New macro.
13881 (output_asm_line_debug_info): New.
13882 (struct var_loc_node): Add view.
13883 (add_AT_view_list, AT_loc_list): New.
13884 (add_var_loc_to_decl): Add view param. Test it against last.
13885 (new_loc_list): Add view params. Record them.
13886 (AT_loc_list_ptr): Handle loc and view lists.
13887 (view_list_to_loc_list_val_node): New.
13888 (print_dw_val): Handle dw_val_class_view_list.
13889 (size_of_die): Likewise.
13890 (value_format): Likewise.
13891 (loc_list_has_views): New.
13892 (gen_llsym): Set vl_symbol too.
13893 (maybe_gen_llsym, skip_loc_list_entry): New.
13894 (dwarf2out_maybe_output_loclist_view_pair): New.
13895 (output_loc_list): Output view list or entries too.
13896 (output_view_list_offset): New.
13897 (output_die): Handle dw_val_class_view_list.
13898 (output_dwarf_version): New.
13899 (output_compilation_unit_header): Use it.
13900 (output_skeleton_debug_sections): Likewise.
13901 (output_rnglists, output_line_info): Likewise.
13902 (output_pubnames, output_aranges): Update version comments.
13903 (output_one_line_info_table): Output view numbers in asm comments.
13904 (dw_loc_list): Determine current endview, pass it to new_loc_list.
13905 Call maybe_gen_llsym.
13906 (loc_list_from_tree_1): Adjust.
13907 (add_AT_location_description): Create view list attribute if
13908 needed, check it's absent otherwise.
13909 (convert_cfa_to_fb_loc_list): Adjust.
13910 (maybe_emit_file): Call output_asm_line_debug_info for test.
13911 (dwarf2out_var_location): Reset views as needed. Precompute
13912 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
13913 attribute. Set view.
13914 (new_line_info_table): Reset next view.
13915 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
13916 (dwarf2out_source_line): Likewise. Output view resets and labels to
13917 the assembler, or select appropriate line info opcodes.
13918 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
13919 (optimize_string_length): Catch it. Adjust.
13920 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
13921 dw_val_class_view_list, and remove it if no longer needed.
13922 (hash_loc_list): Hash view numbers.
13923 (loc_list_hasher::equal): Compare them.
13924 (optimize_location_lists): Check whether a view list symbol is
13925 needed, and whether the locview attribute is present, and
13926 whether they match. Remove the locview attribute if no longer
13928 (index_location_lists): Call skip_loc_list_entry for test.
13929 (dwarf2out_finish): Call output_asm_line_debug_info for test.
13930 Use output_dwarf_version.
13931 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
13932 (struct dw_val_node): Add val_view_list.
13933 * final.c (SEEN_NEXT_VIEW): New.
13934 (set_next_view_needed): New.
13935 (clear_next_view_needed): New.
13936 (maybe_output_next_view): New.
13937 (final_start_function): Rename to...
13938 (final_start_function_1): ... this. Take pointer to FIRST,
13939 add SEEN parameter. Emit param bindings in the initial view.
13940 (final_start_function): Reintroduce SEEN-less interface.
13941 (final): Rename to...
13942 (final_1): ... this. Take SEEN parameter. Output final pending
13943 next view at the end.
13944 (final): Reintroduce seen-less interface.
13945 (final_scan_insn): Output pending next view before switching
13946 sections or ending a block. Mark the next view as needed when
13947 outputting variable locations. Notify debug backend of section
13948 changes, and of location view changes.
13949 (rest_of_handle_final): Adjust.
13950 * toplev.c (process_options): Autodetect value for debug variable
13951 location views option. Warn on incompat5 without -gdwarf-5.
13952 * doc/invoke.texi (gvariable-location-views): New.
13953 (gvariable-location-views=incompat5): New.
13954 (gno-variable-location-views): New.
13956 2018-02-08 David Malcolm <dmalcolm@redhat.com>
13958 PR tree-optimization/84136
13959 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
13960 that the result of find_edge is non-NULL.
13962 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13965 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
13966 storing integer register in SImode. Fix cost of 256 and 512
13967 byte aligned SSE register store.
13969 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13971 * config/i386/i386.c (ix86_multiplication_cost): Fix
13972 multiplication cost for TARGET_AVX512DQ.
13974 2018-02-08 Marek Polacek <polacek@redhat.com>
13976 PR tree-optimization/84238
13977 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
13980 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13982 PR tree-optimization/84265
13983 * tree-vect-stmts.c (vectorizable_store): Don't treat
13984 VMAT_CONTIGUOUS accesses as grouped.
13985 (vectorizable_load): Likewise.
13987 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13989 PR tree-optimization/81635
13990 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
13991 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
13992 (test_round_for_mask): New functions.
13993 (wide_int_cc_tests): Call test_round_for_mask.
13994 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
13995 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
13996 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
13997 range returned by get_range_info.
13999 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
14002 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
14003 * symtab.c: Include builtins.h
14004 (symtab_node::output_to_lto_symbol_table_p): Move here
14005 from lto-streamer-out.c:output_symbol_p.
14006 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
14007 (output_symbol_p): Move all logic to symtab.c
14008 (produce_symtab): Update.
14010 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14012 * config/s390/s390-opts.h (enum indirect_branch): Define.
14013 * config/s390/s390-protos.h (s390_return_addr_from_memory)
14014 (s390_indirect_branch_via_thunk)
14015 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
14016 (enum s390_indirect_branch_type): Define.
14017 * config/s390/s390.c (struct s390_frame_layout, struct
14018 machine_function): Remove.
14019 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
14020 (indirect_branch_table_label_no, indirect_branch_table_name):
14022 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
14023 (enum s390_indirect_branch_option): Define.
14024 (s390_return_addr_from_memory): New function.
14025 (s390_handle_string_attribute): New function.
14026 (s390_attribute_table): Add new attribute handler.
14027 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
14028 (s390_indirect_branch_via_thunk): New function.
14029 (s390_indirect_branch_via_inline_thunk): New function.
14030 (s390_function_ok_for_sibcall): When jumping via thunk disallow
14031 sibling call optimization for non z10 compiles.
14032 (s390_emit_call): Force indirect branch target to be a single
14033 register. Add r1 clobber for non-z10 compiles.
14034 (s390_emit_epilogue): Emit return jump via return_use expander.
14035 (s390_reorg): Handle JUMP_INSNs as execute targets.
14036 (s390_option_override_internal): Perform validity checks for the
14037 new command line options.
14038 (s390_indirect_branch_attrvalue): New function.
14039 (s390_indirect_branch_settings): New function.
14040 (s390_set_current_function): Invoke s390_indirect_branch_settings.
14041 (s390_output_indirect_thunk_function): New function.
14042 (s390_code_end): Implement target hook.
14043 (s390_case_values_threshold): Implement target hook.
14044 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
14046 * config/s390/s390.h (struct s390_frame_layout)
14047 (struct machine_function): Move here from s390.c.
14048 (TARGET_INDIRECT_BRANCH_NOBP_RET)
14049 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
14050 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
14051 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
14052 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
14053 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
14054 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
14055 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
14056 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
14057 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
14058 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
14059 (mnemonic attribute): Add values which aren't recognized
14061 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
14062 pattern for branch conversion. Fix mnemonic attribute.
14063 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
14064 indirect branch via thunk if requested.
14065 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
14066 ("*indirect_jump"): Disable for branch conversion using out of
14068 ("indirect_jump_via_thunk<mode>_z10")
14069 ("indirect_jump_via_thunk<mode>")
14070 ("indirect_jump_via_inlinethunk<mode>_z10")
14071 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
14072 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
14073 ("casesi_jump_via_inlinethunk<mode>_z10")
14074 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
14075 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
14076 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
14077 ("*indirect2_jump"): Disable for branch conversion.
14078 ("casesi_jump"): Turn into expander and expand patterns for branch
14080 ("return_use"): New expander.
14081 ("*return"): Emit return via thunk and rename it to ...
14082 ("*return<mode>"): ... this one.
14083 * config/s390/s390.opt: Add new options and and enum for the
14086 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14088 * lra-constraints.c (match_reload): Unconditionally use
14089 gen_lowpart_SUBREG, rather than selecting between that
14090 and equivalent gen_rtx_SUBREG code.
14092 2018-02-08 Richard Biener <rguenther@suse.de>
14094 PR tree-optimization/84233
14095 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
14096 changed flag instead of boguously re-using phi_inserted.
14098 2018-02-08 Martin Jambor <mjambor@suse.cz>
14100 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
14101 static local variables.
14103 2018-02-08 Richard Biener <rguenther@suse.de>
14105 PR tree-optimization/84278
14106 * tree-vect-stmts.c (vectorizable_store): When looking for
14107 smaller vector types to perform grouped strided loads/stores
14108 make sure the mode is supported by the target.
14109 (vectorizable_load): Likewise.
14111 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
14113 * config/aarch64/aarch64.c (aarch64_components_for_bb):
14114 Increase LDP/STP opportunities by adding adjacent callee-saves.
14116 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
14118 PR rtl-optimization/84068
14119 PR rtl-optimization/83459
14120 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
14122 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
14124 PR tree-optimization/84224
14125 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
14126 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
14127 non-zero arguments.
14129 2018-02-07 Iain Sandoe <iain@codesourcery.com>
14132 * config/rs6000/altivec.md (*restore_world): Remove LR use.
14133 * config/rs6000/predicates.md (restore_world_operation): Adjust op
14134 count, remove one USE.
14136 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
14138 * doc/install.texi (Configuration): Document the
14139 --with-long-double-format={ibm,ieee} PowerPC configuration
14143 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
14144 Convert from define_expand to be define_insn_and_split. Rework
14145 float/double/_Float128 conversions to QI/HI/SImode to work with
14146 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
14147 conversions to QI/HImode types did a store and then a load to
14148 truncate the value. For conversions to VSX registers, don't split
14149 the insn, instead emit the code directly. Use the code iterator
14150 any_fix to combine signed and unsigned conversions.
14151 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
14152 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
14153 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
14154 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
14155 (fix_<mode>di2_hw): Likewise.
14156 (fixuns_<mode>di2_hw): Likewise.
14157 (fix_<mode>si2_hw): Likewise.
14158 (fixuns_<mode>si2_hw): Likewise.
14159 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
14160 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
14161 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
14162 fix<uns>_trunc<SFDF:mode>si2_p8.
14163 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
14164 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
14165 (fix<uns>_<mode>_mem): Likewise.
14166 (fctiw<u>z_<mode>_mem): Likewise.
14167 (fix<uns>_<mode>_mem): Likewise.
14168 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
14169 the register allocator from doing a direct move to the GPRs to do
14170 a store, and instead use the ISA 3.0 store byte/half-word from
14171 vector register instruction. For IEEE 128-bit floating point,
14172 also optimize stores of 32-bit ints.
14173 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
14175 2018-02-07 Alan Hayward <alan.hayward@arm.com>
14177 * genextract.c (push_pathstr_operand): New function to support
14179 (walk_rtx): Call push_pathstr_operand.
14180 (print_path): Support [a-zA-Z].
14182 2018-02-07 Richard Biener <rguenther@suse.de>
14184 PR tree-optimization/84037
14185 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
14186 (cse_and_gimplify_to_preheader): Declare.
14187 (vect_get_place_in_interleaving_chain): Likewise.
14188 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14190 (_loop_vec_info::~_loop_vec_info): Delete it.
14191 (cse_and_gimplify_to_preheader): New function.
14192 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
14193 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
14194 (vectorizable_load): Likewise. For grouped stores always base
14195 the IV on the first element.
14196 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
14197 condition before gimplifying.
14199 2018-02-07 Jakub Jelinek <jakub@redhat.com>
14201 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
14202 *DIV_EXPR and *MOD_EXPR.
14204 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
14207 * config/i386/i386.c (ix86_option_override_internal): Mask out
14208 the CF_SET bit when checking -fcf-protection.
14210 2018-02-07 Tom de Vries <tom@codesourcery.com>
14213 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
14216 2018-02-07 Richard Biener <rguenther@suse.de>
14218 PR tree-optimization/84204
14219 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
14222 PR tree-optimization/84205
14223 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
14224 special-case isl_ast_op_zdiv_r.
14226 PR tree-optimization/84223
14227 * graphite-scop-detection.c (gather_bbs::before_dom_children):
14228 Only add conditions from within the region.
14229 (gather_bbs::after_dom_children): Adjust.
14231 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
14234 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
14235 * config/avr/avr.md: Only post-reload split REG-REG moves if
14236 either register is GENERAL_REG_P.
14238 2018-02-07 Jakub Jelinek <jakub@redhat.com>
14240 PR tree-optimization/84235
14241 * tree-ssa-scopedtables.c
14242 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
14243 if the subtraction is performed in floating point type where NaNs are
14244 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
14245 build 1. Formatting fix.
14247 2018-02-06 Jakub Jelinek <jakub@redhat.com>
14250 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
14251 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
14252 and skip it regardless of bb boundaries. Use CALL_P macro,
14253 don't test INSN_P (insn) together with CALL_P or JUMP_P check
14254 unnecessarily, formatting fix.
14256 2018-02-06 Michael Collison <michael.collison@arm.com>
14258 * config/arm/thumb2.md:
14259 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
14260 (*thumb_mov_notscc): Ditto.
14262 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
14265 * config/rs6000/rs6000.md (su code attribute): Use "u" for
14266 unsigned_fix, not "s".
14268 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14270 * configure.ac (gcc_fn_eh_frame_ro): New function.
14271 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
14272 correct .eh_frame permissions.
14273 * configure: Regenerate.
14275 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
14277 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
14278 irrelevant options.
14280 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14282 * config/rs6000/rs6000.c (rs6000_option_override_internal):
14283 Display warning message for -mno-speculate-indirect-jumps.
14285 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
14287 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
14289 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
14291 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
14293 PR tree-optimization/84225
14294 * tree-eh.c (find_trapping_overflow): Only call
14295 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
14297 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14300 * config/i386/i386.c: Reimplement the check of possible options
14301 -mibt/-mshstk conbination. Change error messages.
14302 * doc/invoke.texi: Fix a typo: remove extra '='.
14304 2018-02-06 Marek Polacek <polacek@redhat.com>
14306 PR tree-optimization/84228
14307 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
14309 2018-02-06 Tamar Christina <tamar.christina@arm.com>
14312 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
14313 emitted arch directives.
14314 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
14315 __ARM_FEATURE_COPROC before changing architectures.
14317 2018-02-06 Richard Biener <rguenther@suse.de>
14319 * config/i386/i386.c (print_reg): Fix typo.
14320 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
14322 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
14324 * configure: Regenerate.
14326 2018-02-05 Martin Sebor <msebor@redhat.com>
14328 PR tree-optimization/83369
14329 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
14332 2018-02-05 Martin Liska <mliska@suse.cz>
14334 * doc/invoke.texi: Cherry-pick upstream r323995.
14336 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
14338 * ira.c (ira_init_register_move_cost): Adjust comment.
14340 2018-02-05 Martin Liska <mliska@suse.cz>
14342 PR gcov-profile/84137
14343 * doc/gcov.texi: Fix typo in documentation.
14345 2018-02-05 Martin Liska <mliska@suse.cz>
14347 PR gcov-profile/83879
14348 * doc/gcov.texi: Document necessity of --dynamic-list-data when
14349 using dlopen functionality.
14351 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
14353 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
14354 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
14355 _mm_maskz_range_ss, _mm_mask_range_round_ss,
14356 _mm_maskz_range_round_ss): New intrinsics.
14357 (__builtin_ia32_rangesd128_round)
14358 (__builtin_ia32_rangess128_round): Remove.
14359 (__builtin_ia32_rangesd128_mask_round,
14360 __builtin_ia32_rangess128_mask_round): New builtins.
14361 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
14362 __builtin_ia32_rangess128_round): Remove.
14363 (__builtin_ia32_rangesd128_mask_round,
14364 __builtin_ia32_rangess128_mask_round): New builtins.
14365 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
14366 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
14367 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
14368 "<round_saeonly_constraint>")): Changed to ...
14369 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
14370 "<round_saeonly_scalar_constraint>")): ... this.
14371 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
14372 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
14373 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
14374 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
14375 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
14377 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
14379 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
14381 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
14382 Remove all values except native, 8540 and 8548.
14384 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
14386 * config/i386/i386.c (ix86_output_function_return): Pass
14387 INVALID_REGNUM, instead of -1, as invalid register number to
14388 indirect_thunk_name and output_indirect_thunk.
14390 2018-02-02 Julia Koval <julia.koval@intel.com>
14392 * config.gcc: Add -march=icelake.
14393 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
14394 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
14395 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
14396 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
14397 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
14398 (processor_target_table): Add icelake.
14399 (ix86_option_override_internal): Handle new PTAs.
14400 (get_builtin_code_for_version): Handle icelake.
14401 (M_INTEL_COREI7_ICELAKE): New.
14402 (fold_builtin_cpu): Handle icelake.
14403 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
14404 * doc/invoke.texi: Add -march=icelake.
14406 2018-02-02 Julia Koval <julia.koval@intel.com>
14408 * config/i386/i386.c (ix86_option_override_internal): Change flags type
14409 to wide_int_bitmask.
14410 * wide-int-bitmask.h: New.
14412 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14415 * config/i386/i386.md: Replace Pmode with word_mode in
14416 builtin_setjmp_setup and builtin_longjmp to support x32.
14418 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
14422 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
14424 (rs6000_supported_cpu_names): New static variable.
14425 (linux_cpu_translation_table): Likewise.
14426 (elf_platform) <cpu>: Define new static variable and use it.
14427 Translate kernel AT_PLATFORM name to canonical name if needed.
14428 Error if platform name is unknown.
14430 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
14433 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
14435 2018-02-01 Jeff Law <law@redhat.com>
14438 * config/i386/i386.c (release_scratch_register_on_entry): Add new
14439 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
14440 the scratch if RELEASE_VIA_POP is false.
14441 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
14442 If we have to save a temporary register, decrement SIZE appropriately.
14443 Pass new arguments to release_scratch_register_on_entry.
14444 (ix86_adjust_stack_and_probe): Likewise.
14445 (ix86_emit_probe_stack_range): Pass new arguments to
14446 release_scratch_register_on_entry.
14448 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
14450 PR rtl-optimization/84157
14451 * combine.c (change_zero_ext): Use REG_P predicate in
14452 front of HARD_REGISTER_P predicate.
14454 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
14456 * config/avr/avr.c (avr_option_override): Move disabling of
14457 -fdelete-null-pointer-checks to...
14458 * common/config/avr/avr-common.c (avr_option_optimization_table):
14461 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14463 PR tree-optimization/81635
14464 * tree-data-ref.c (split_constant_offset_1): For types that
14465 wrap on overflow, try to use range info to prove that wrapping
14468 2018-02-01 Renlin Li <renlin.li@arm.com>
14471 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
14472 TAILCALL_ADDR_REGS.
14473 (aarch64_register_move_cost): Likewise.
14474 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
14475 TAILCALL_ADDR_REGS.
14476 (REG_CLASS_NAMES): Likewise.
14477 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
14478 TAILCALL_ADDR_REGS. Remove IP registers.
14479 * config/aarch64/aarch64.md (Ucs): Update register constraint.
14481 2018-02-01 Richard Biener <rguenther@suse.de>
14483 * domwalk.h (dom_walker::dom_walker): Add additional constructor
14484 for specifying RPO order and allow NULL for that.
14485 * domwalk.c (dom_walker::dom_walker): Likewise.
14486 (dom_walker::walk): Handle NULL RPO order.
14487 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
14489 (rewrite_update_dom_walker): Likewise.
14490 (mark_def_dom_walker): Likewise.
14492 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14494 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
14495 (aarch64_maybe_expand_sve_subreg_move): Declare.
14496 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
14497 * config/aarch64/predicates.md (aarch64_any_register_operand): New
14499 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
14500 that are semantically a reverse operation.
14501 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
14502 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
14503 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
14505 (aarch64_can_change_mode_class): For big-endian, forbid changes
14506 between two SVE modes if they have different element sizes.
14508 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14510 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
14511 the TImode handling for big-endian targets.
14513 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14515 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
14516 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
14518 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
14519 Remove BSWAP handing for big-endian targets and use the form of
14520 LD1RQ appropariate for the mode.
14522 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14524 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
14525 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
14526 duplicated element.
14528 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14531 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
14532 check for operands that need to go through aarch64_sve_reload_be.
14534 2018-02-01 Jakub Jelinek <jakub@redhat.com>
14536 PR tree-optimization/81661
14537 PR tree-optimization/84117
14538 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
14539 * tree-eh.c: Include gimplify.h.
14540 (find_trapping_overflow, replace_trapping_overflow,
14541 rewrite_to_non_trapping_overflow): New functions.
14542 * tree-vect-loop.c: Include tree-eh.h.
14543 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
14544 * tree-data-ref.c: Include tree-eh.h.
14545 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
14547 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
14549 PR rtl-optimization/84123
14550 * combine.c (change_zero_ext): Check if hard register satisfies
14551 can_change_dest_mode before calling gen_lowpart_SUBREG.
14553 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
14556 * ira.c (ira_init_register_move_cost): Remove assert.
14558 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14560 PR rtl-optimization/84071
14561 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
14562 * doc/tm.texi: Regenerate.
14564 2018-01-31 Richard Biener <rguenther@suse.de>
14566 PR tree-optimization/84132
14567 * tree-data-ref.c (analyze_miv_subscript): Properly
14568 check whether evolution_function_is_affine_multivariate_p
14569 before calling gcd_of_steps_may_divide_p.
14571 2018-01-31 Julia Koval <julia.koval@intel.com>
14574 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
14575 * config/i386/i386.md (rdpid_rex64) New.
14576 (rdpid): Make 32bit only.
14578 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
14581 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
14582 an IDENTIFIER_NODE for FUNCTION_TYPE's.
14584 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14587 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14589 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14591 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14593 PR rtl-optimization/84071
14594 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
14595 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
14597 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
14599 * config/arc/arc.c (arc_handle_aux_attribute): New function.
14600 (arc_attribute_table): Add 'aux' attribute.
14601 (arc_in_small_data_p): Consider aux like variables.
14602 (arc_is_aux_reg_p): New function.
14603 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
14604 (arc_get_aux_arg): New function.
14605 (prepare_move_operands): Handle aux-register access.
14606 (arc_handle_aux_attribute): New function.
14607 * doc/extend.texi (ARC Variable attributes): Add subsection.
14609 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
14611 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
14612 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
14613 (arc_attribute_table): Add 'uncached' attribute.
14614 (arc_print_operand): Print '.di' flag for uncached memory
14616 (arc_in_small_data_p): Do not consider for small data the uncached
14618 (arc_is_uncached_mem_p): New function.
14619 * config/arc/predicates.md (compact_store_memory_operand): Check
14620 for uncached memory accesses.
14621 (nonvol_nonimm_operand): Likewise.
14622 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
14624 2018-01-31 Jakub Jelinek <jakub@redhat.com>
14627 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
14628 falign-loops=): Add Optimization flag.
14630 2018-01-30 Jeff Law <law@redhat.com>
14633 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
14634 INT_REGISTERS_SAVED. Check it prior to calling
14635 get_scratch_register_on_entry.
14636 (ix86_adjust_stack_and_probe): Similarly.
14637 (ix86_emit_probe_stack_range): Similarly.
14638 (ix86_expand_prologue): Corresponding changes.
14640 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14643 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
14644 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
14646 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
14649 * lra-constraints.c (curr_insn_transform): Process AND in the
14652 2018-01-30 Jakub Jelinek <jakub@redhat.com>
14654 PR rtl-optimization/83986
14655 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
14656 dependence against last_pending_memory_flush in addition to
14657 pending_jump_insns.
14659 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
14661 PR tree-optimization/81611
14662 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
14665 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14668 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
14671 2018-01-30 Richard Biener <rguenther@suse.de>
14672 Jakub Jelinek <jakub@redhat.com>
14674 PR tree-optimization/84111
14675 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
14676 inner loops added during recursion, as they don't have up-to-date
14679 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
14682 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
14683 (can_inline_edge_by_limits_p): ... here.
14684 (can_early_inline_edge_p, check_callers,
14685 update_caller_keys, update_callee_keys, recursive_inlining,
14686 add_new_edges_to_heap, speculation_useful_p,
14687 inline_small_functions,
14688 inline_small_functions, flatten_function,
14689 inline_to_all_callers_1): Update.
14691 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
14693 * profile-count.c (profile_count::combine_with_ipa_count): Handle
14696 2018-01-30 Richard Biener <rguenther@suse.de>
14698 PR tree-optimization/83008
14699 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
14700 invariant and constant vector uses in stmts when they need
14701 more than one stmt.
14703 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14706 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
14707 * configure: Regenerate.
14709 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14711 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
14713 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
14714 Use gen_rtx_REG rather than gen_lowpart.
14716 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14718 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
14719 rather than 0 when creating partial subregs.
14721 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14723 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
14726 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
14729 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
14730 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
14731 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
14732 flags. This restores the settings used before the 2017-07-24.
14733 Turning off pre increment/decrement/modify allows IVOPTS to
14734 optimize DF/SF loops where the index is an int.
14736 2018-01-29 Richard Biener <rguenther@suse.de>
14737 Kelvin Nilsen <kelvin@gcc.gnu.org>
14740 * tree-vect-stmts.c (vectorizable_call): Don't call
14741 targetm.vectorize_builtin_md_vectorized_function if callee is
14744 2018-01-22 Carl Love <cel@us.ibm.com>
14746 * doc/extend.tex: Fix typo in second arg in
14747 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
14749 2018-01-29 Richard Biener <rguenther@suse.de>
14751 PR tree-optimization/84086
14752 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
14753 (flush_ssaname_freelist): When SSA names were released reset
14754 the SCEV hash table.
14756 2018-01-29 Richard Biener <rguenther@suse.de>
14758 PR tree-optimization/84057
14759 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
14760 removed paths when removing edges.
14762 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
14764 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
14765 -mfunction-return=@var{choice}.
14767 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
14769 PR diagnostic/84034
14770 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
14771 Handle CR like TAB.
14772 (layout::print_source_line): Likewise.
14773 (test_get_line_width_without_trailing_whitespace): Add test cases.
14775 2018-01-27 Jakub Jelinek <jakub@redhat.com>
14777 PR middle-end/84040
14778 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
14781 2018-01-26 Jim Wilson <jimw@sifive.com>
14783 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
14785 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
14788 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14790 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
14791 and CMP + SUB-immediate -> SUBS.
14793 2018-01-26 Martin Sebor <msebor@redhat.com>
14795 PR tree-optimization/83896
14796 * tree-ssa-strlen.c (get_string_len): Rename...
14797 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
14798 Avoid assuming length is constant.
14799 (handle_char_store): Use HOST_WIDE_INT for string length.
14801 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
14804 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
14805 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
14807 2018-01-26 Richard Biener <rguenther@suse.de>
14809 PR rtl-optimization/84003
14810 * dse.c (record_store): Only record redundant stores when
14811 the earlier store aliases at least all accesses the later one does.
14813 2018-01-26 Jakub Jelinek <jakub@redhat.com>
14815 PR rtl-optimization/83985
14816 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
14817 REG_CFA_RESTORE insns.
14818 (delete_unmarked_insns): Don't ignore separate shrink wrapping
14819 REG_CFA_RESTORE insns here.
14822 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
14823 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
14825 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14827 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
14828 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
14829 (arc_init): Likewise.
14830 (arc_override_options): Likewise.
14831 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
14833 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
14835 * config/arc/arc.h (TARGET_DBNZ): Define.
14836 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
14837 properly set the tune attribute.
14838 (dbnz): Use TARGET_DBNZ guard.
14839 * config/arc/arc.opt (mtune): Add core3 option.
14841 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14843 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
14844 recognize new pic like addresses.
14845 (arc_delegitimize_address): Clean up.
14847 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14849 * config/arc/arc-arches.def: Option mrf16 valid for all
14851 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
14852 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
14853 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
14854 * config/arc/arc-tables.opt: Regenerate.
14855 * config/arc/arc.c (arc_conditional_register_usage): Handle
14856 reduced register file case.
14857 (arc_file_start): Set must have build attributes.
14858 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
14859 mrf16 option value.
14860 * config/arc/arc.opt (mrf16): Add new option.
14861 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
14862 * config/arc/genmultilib.awk: Handle new mrf16 option.
14863 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
14864 * config/arc/t-multilib: Regenerate.
14865 * doc/invoke.texi (ARC Options): Document mrf16 option.
14867 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14869 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
14870 * config/arc/arc.c (arc_handle_secure_attribute): New function.
14871 (arc_attribute_table): Add 'secure_call' attribute.
14872 (arc_print_operand): Print secure call operand.
14873 (arc_function_ok_for_sibcall): Don't optimize tail calls when
14875 (arc_is_secure_call_p): New function. * config/arc/arc.md
14876 (call_i): Add support for sjli instruction.
14877 (call_value_i): Likewise.
14878 * config/arc/constraints.md (Csc): New constraint.
14880 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14881 John Eric Martin <John.Martin@emmicro-us.com>
14883 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
14884 * config/arc/arc.c (_arc_jli_section): New struct.
14885 (arc_jli_section): New type.
14886 (rc_jli_sections): New static variable.
14887 (arc_handle_jli_attribute): New function.
14888 (arc_attribute_table): Add jli_always and jli_fixed attribute.
14889 (arc_file_end): New function.
14890 (TARGET_ASM_FILE_END): Define.
14891 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
14892 (arc_add_jli_section): New function.
14893 (jli_call_scan): Likewise.
14894 (arc_reorg): Call jli_call_scan.
14895 (arc_output_addsi): Remove 'S' from printing asm operand.
14896 (arc_is_jli_call_p): New function.
14897 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
14899 (movhi_insn): Likewise.
14900 (movsi_insn): Likewise.
14901 (movsi_set_cc_insn): Likewise.
14902 (loadqi_update): Likewise.
14903 (load_zeroextendqisi_update): Likewise.
14904 (load_signextendqisi_update): Likewise.
14905 (loadhi_update): Likewise.
14906 (load_zeroextendhisi_update): Likewise.
14907 (load_signextendhisi_update): Likewise.
14908 (loadsi_update): Likewise.
14909 (loadsf_update): Likewise.
14910 (movsicc_insn): Likewise.
14911 (bset_insn): Likewise.
14912 (bxor_insn): Likewise.
14913 (bclr_insn): Likewise.
14914 (bmsk_insn): Likewise.
14915 (bicsi3_insn): Likewise.
14916 (cmpsi_cc_c_insn): Likewise.
14917 (movsi_ne): Likewise.
14918 (movsi_cond_exec): Likewise.
14919 (clrsbsi2): Likewise.
14920 (norm_f): Likewise.
14930 (call_i): Remove 'S' asm letter, add jli instruction.
14931 (call_value_i): Likewise.
14932 * config/arc/arc.op (mjli-always): New option.
14933 * config/arc/constraints.md (Cji): New constraint.
14934 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
14936 (subsf3_fpx): Likewise.
14937 (mulsf3_fpx): Likewise.
14938 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
14940 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
14941 function attrbutes.
14942 * doc/invoke.texi (ARC): Document mjli-always option.
14944 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
14946 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
14947 avoid addition with 0 and use incw and decw where possible.
14949 2018-01-26 Richard Biener <rguenther@suse.de>
14951 PR tree-optimization/81082
14952 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
14953 association if it requires casting to unsigned.
14954 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
14955 from fold_plusminus_mult_expr to catch important cases late when
14956 range info is available.
14958 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14960 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
14961 * configure.ac (hidden_linkonce): New test.
14962 * configure: Regenerate.
14963 * config.in: Regenerate.
14965 2018-01-26 Julia Koval <julia.koval@intel.com>
14967 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
14968 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
14969 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
14970 _mm_mask_bitshuffle_epi64_mask): Fix type.
14971 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
14972 USI_FTYPE_V4DI_V4DI_USI): Remove.
14973 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
14974 __builtin_ia32_vpshufbitqmb256_mask,
14975 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
14976 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
14977 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
14979 2018-01-26 Alan Modra <amodra@gmail.com>
14982 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
14983 UNSPEC_VBPERMQ. Sort other unspecs.
14985 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
14987 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
14989 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
14991 PR middle-end/83055
14992 * predict.c (drop_profile): Do not push/pop cfun; update also
14994 (handle_missing_profiles): Fix logic looking for zero profiles.
14996 2018-01-25 Jakub Jelinek <jakub@redhat.com>
14998 PR middle-end/83977
14999 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
15000 on functions with #pragma omp declare simd or functions with simd
15002 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
15003 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
15004 Remove trailing \n from warning_at calls.
15006 2018-01-25 Tom de Vries <tom@codesourcery.com>
15009 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
15010 for neutered workers.
15012 2018-01-24 Joseph Myers <joseph@codesourcery.com>
15015 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
15016 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
15018 2018-01-24 Jeff Law <law@redhat.com>
15021 * i386.c (get_probe_interval): Move to earlier point.
15022 (ix86_compute_frame_layout): If -fstack-clash-protection and
15023 the frame is larger than the probe interval, then use pushes
15024 to save registers rather than reg->mem moves.
15025 (ix86_expand_prologue): Remove conditional for int_registers_saved
15028 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
15031 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
15032 min/max for never referenced object.
15034 2018-01-24 Jakub Jelinek <jakub@redhat.com>
15036 PR middle-end/83977
15037 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
15039 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
15040 attributes from DECL_ATTRIBUTES (decl) without affecting
15041 DECL_ATTRIBUTES (current_function_decl).
15042 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
15043 functions with non-NULL DECL_ABSTRACT_ORIGIN.
15045 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
15047 PR tree-optimization/83979
15048 * fold-const.c (fold_comparison): Use constant_boolean_node
15049 instead of boolean_{true,false}_node.
15051 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
15053 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
15056 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15058 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15059 Simplify the clause that sets the length attribute.
15060 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15061 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
15062 clause that sets the length attribute.
15063 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15065 2018-01-24 Tom de Vries <tom@codesourcery.com>
15068 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
15069 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
15070 Add strict parameter.
15071 (prevent_branch_around_nothing): Insert dummy insn between branch to
15072 label and label with no ptx insn inbetween.
15073 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
15075 2018-01-24 Tom de Vries <tom@codesourcery.com>
15078 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
15079 for neutered threads in warp.
15080 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
15082 2018-01-24 Richard Biener <rguenther@suse.de>
15084 PR tree-optimization/83176
15085 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
15088 2018-01-24 Richard Biener <rguenther@suse.de>
15090 PR tree-optimization/82819
15091 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
15092 code generating pluses that are no-ops in the target precision.
15094 2018-01-24 Richard Biener <rguenther@suse.de>
15096 PR middle-end/84000
15097 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
15099 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
15101 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
15102 to merge probabilities.
15103 * predict.c (probably_never_executed): Also mark as cold functions
15104 with global 0 profile and guessed local profile.
15105 * profile-count.c (profile_probability::combine_with_count): New
15107 * profile-count.h (profile_probability::operator*,
15108 profile_probability::operator*=, profile_probability::operator/,
15109 profile_probability::operator/=): Reduce precision to adjusted
15110 and set value to guessed on contradictory divisions.
15111 (profile_probability::combine_with_freq): Remove.
15112 (profile_probability::combine_wiht_count): Declare.
15113 (profile_count::force_nonzero):: Set to adjusted.
15114 (profile_count::probability_in):: Set quality to adjusted.
15115 * tree-ssa-tail-merge.c (replace_block_by): Use
15116 combine_with_count.
15118 2018-01-23 Andrew Waterman <andrew@sifive.com>
15119 Jim Wilson <jimw@sifive.com>
15121 * config/riscv/riscv.c (riscv_stack_boundary): New.
15122 (riscv_option_override): Set riscv_stack_boundary. Handle
15123 riscv_preferred_stack_boundary_arg.
15124 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
15125 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
15126 (STACK_BOUNDARY): Set to riscv_stack_boundary.
15127 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
15128 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
15129 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
15131 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
15134 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
15135 of struct ix86_frame.
15136 (ix86_expand_epilogue): Likewise. Add a local variable for
15137 the reg_save_offset field in struct ix86_frame.
15139 2018-01-23 Bin Cheng <bin.cheng@arm.com>
15141 PR tree-optimization/82604
15142 * tree-loop-distribution.c (enum partition_kind): New enum item
15143 PKIND_PARTIAL_MEMSET.
15144 (partition_builtin_p): Support above new enum item.
15145 (generate_code_for_partition): Ditto.
15146 (compute_access_range): Differentiate cases that equality can be
15147 proven at all loops, the innermost loops or no loops.
15148 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
15149 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
15150 (finalize_partitions, distribute_loop): Don't fuse partition of
15151 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
15152 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
15153 parloop is enabled.
15155 2018-01-23 Martin Liska <mliska@suse.cz>
15157 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
15158 order to ignore the predictor.
15159 (PRED_POLYMORPHIC_CALL): Likewise.
15160 (PRED_RECURSIVE_CALL): Likewise.
15162 2018-01-23 Martin Liska <mliska@suse.cz>
15164 * tree-profile.c (tree_profiling): Print function header to
15165 aware reader which function we are working on.
15166 * value-prof.c (gimple_find_values_to_profile): Do not print
15167 not interesting value histograms.
15169 2018-01-23 Martin Liska <mliska@suse.cz>
15171 * profile-count.h (enum profile_quality): Add
15172 profile_uninitialized as the first value. Do not number values
15173 as they are zero based.
15174 (profile_count::verify): Update sanity check.
15175 (profile_probability::verify): Likewise.
15177 2018-01-23 Nathan Sidwell <nathan@acm.org>
15179 * doc/invoke.texi (ffor-scope): Deprecate.
15181 2018-01-23 David Malcolm <dmalcolm@redhat.com>
15183 PR tree-optimization/83510
15184 * domwalk.c (set_all_edges_as_executable): New function.
15185 (dom_walker::dom_walker): Convert bool param
15186 "skip_unreachable_blocks" to enum reachability. Move setup of
15187 edge flags to set_all_edges_as_executable and only do it when
15188 reachability is REACHABLE_BLOCKS.
15189 * domwalk.h (enum dom_walker::reachability): New enum.
15190 (dom_walker::dom_walker): Convert bool param
15191 "skip_unreachable_blocks" to enum reachability.
15192 (set_all_edges_as_executable): New decl.
15193 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
15194 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
15196 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
15197 but converting true to REACHABLE_BLOCKS.
15198 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
15200 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
15201 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
15202 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
15204 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
15205 if check_all_array_refs will be called.
15207 2018-01-23 David Malcolm <dmalcolm@redhat.com>
15209 * tree.c (selftest::test_location_wrappers): Add more test
15212 2018-01-23 David Malcolm <dmalcolm@redhat.com>
15214 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
15215 (selftest::test_bit_in_range): Likewise.
15217 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
15220 * doc/sourcebuild.texi (vect_float): Say that the selector
15221 only describes the situation when -funsafe-math-optimizations is on.
15222 (vect_float_strict): Document.
15224 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
15226 PR tree-optimization/83965
15227 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
15228 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
15229 instead of checking only for a reduction.
15230 (vect_recog_widen_sum_pattern): Likewise.
15232 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
15234 * predict.c (probably_never_executed): Only use precise profile info.
15235 (compute_function_frequency): Skip after inlining hack since we now
15236 have quality checking.
15238 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
15240 * profile-count.h (profile_probability::very_unlikely,
15241 profile_probability::unlikely, profile_probability::even): Set
15242 precision to guessed.
15244 2018-01-23 Richard Biener <rguenther@suse.de>
15246 PR tree-optimization/83963
15247 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
15248 Properly terminate dominator walk when crossing the exit edge not
15249 when visiting its source block.
15251 2018-01-23 Jakub Jelinek <jakub@redhat.com>
15254 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
15255 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
15257 2018-01-22 Jakub Jelinek <jakub@redhat.com>
15259 PR tree-optimization/83957
15260 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
15261 semicolon after for body surrounded by braces.
15263 PR tree-optimization/83081
15264 * profile-count.h (profile_probability::split): New method.
15265 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
15266 Use profile_probability::split.
15267 (do_compare_rtx_and_jump): Fix adjustment of probabilities
15268 when splitting a single conditional jump into 2.
15270 2018-01-22 David Malcolm <dmalcolm@redhat.com>
15272 PR tree-optimization/69452
15273 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
15276 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15278 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
15279 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
15280 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
15282 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15284 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
15286 * config/rl78/rl78.md (movdi): New define_expand.
15287 * config/rl78/rl78.c (rl78_split_movdi): New function.
15289 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
15292 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
15294 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
15295 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
15296 128-bit to produce an UNSPEC move to get the double word with the
15297 signbit and then a shift directly to do signbit.
15298 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
15299 implementation with a new version that just does either a direct
15300 move or a regular move. Move memory interface to separate insns.
15301 Move insns so they are next to the expander.
15302 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
15303 with signbit move. Split big and little endian case.
15304 (signbit<mode>2_dm_mem_le): Likewise.
15305 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
15306 (signbit<mode>2_dm2): Likewise.
15308 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15310 * config/rl78/rl78.md (anddi3): New define_expand.
15312 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15314 * config/rl78/rl78.md (umindi3): New define_expand.
15316 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15318 * config/rl78/rl78.md (smindi3): New define_expand.
15320 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15322 * config/rl78/rl78.md (smaxdi3): New define_expand.
15324 2018-01-22 Carl Love <cel@us.ibm.com>
15326 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
15327 LVX_V1TI): Add macro expansion.
15328 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
15329 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
15330 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
15331 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
15332 Change check to determine if the instruction is a byte reversing
15333 entry. Fix typo in comment.
15334 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
15335 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
15336 Add def_builtin calls for new builtins.
15337 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
15338 Add define_insn expansion.
15340 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15342 * config/rl78/rl78.md (umaxdi3): New define_expand.
15344 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15346 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
15347 for non-QImode registers.
15349 2018-01-22 Richard Biener <rguenther@suse.de>
15351 PR tree-optimization/83963
15352 * graphite-scop-detection.c (scop_detection::get_sese): Delay
15353 including the loop exit block.
15354 (scop_detection::merge_sese): Likewise.
15355 (scop_detection::add_scop): Do it here instead.
15357 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15359 * doc/sourcebuild.texi (arm_softfloat): Document.
15361 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
15364 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
15365 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
15366 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
15368 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15369 David Edelsohn <dje.gcc@gmail.com>
15372 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15373 Change "crset eq" to "crset 2".
15374 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15375 (*call_indirect_aix<mode>_nospec): Likewise.
15376 (*call_value_indirect_aix<mode>_nospec): Likewise.
15377 (*call_indirect_elfv2<mode>_nospec): Likewise.
15378 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
15379 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
15380 change assembly output from . to $.
15381 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15382 (indirect_jump<mode>_nospec): Change assembly output from . to $.
15383 (*tablejump<mode>_internal1_nospec): Likewise.
15385 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
15388 * config/sh/sh_optimize_sett_clrt.cc:
15389 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
15391 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
15393 PR tree-optimization/83940
15394 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
15395 offset_dt to vect_constant_def rather than vect_unknown_def_type.
15396 (vect_check_load_store_mask): Add a mask_dt_out parameter and
15397 use it to pass back the definition type.
15398 (vect_check_store_rhs): Likewise rhs_dt_out.
15399 (vect_build_gather_load_calls): Add a mask_dt argument and use
15400 it instead of a call to vect_is_simple_use.
15401 (vectorizable_store): Update calls to vect_check_load_store_mask
15402 and vect_check_store_rhs. Use the dt returned by the latter instead
15403 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
15404 instead of calls to vect_is_simple_use. Pass the scalar rather
15405 than the vector operand to vect_is_simple_use when handling
15406 second and subsequent copies of an rhs value.
15407 (vectorizable_load): Update calls to vect_check_load_store_mask
15408 and vect_build_gather_load_calls. Use the cached mask_dt and
15409 gs_info.offset_dt instead of calls to vect_is_simple_use.
15411 2018-01-20 Jakub Jelinek <jakub@redhat.com>
15413 PR middle-end/83945
15414 * tree-emutls.c: Include gimplify.h.
15415 (lower_emutls_2): New function.
15416 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
15417 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
15418 it before further processing.
15421 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
15422 UINTVAL (trueop1) instead of INTVAL (op1).
15424 2018-01-19 Jakub Jelinek <jakub@redhat.com>
15428 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
15429 INCOMING_FRAME_SP_OFFSET if not defined.
15430 (scan_trace): Add ENTRY argument. If true and
15431 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
15432 emit a note to adjust the CFA offset.
15433 (create_cfi_notes): Adjust scan_trace callers.
15434 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
15435 INCOMING_FRAME_SP_OFFSET in the CIE.
15436 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
15437 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
15439 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
15440 * doc/tm.texi: Regenerated.
15442 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15444 PR rtl-optimization/83147
15445 * lra-constraints.c (remove_inheritance_pseudos): Use
15446 lra_substitute_pseudo_within_insn.
15448 2018-01-19 Tom de Vries <tom@codesourcery.com>
15449 Cesar Philippidis <cesar@codesourcery.com>
15452 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
15454 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
15457 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
15458 spaces for function labels.
15460 2018-01-19 Martin Liska <mliska@suse.cz>
15462 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
15463 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
15464 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
15465 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
15466 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
15467 (PRED_CONST_RETURN): Change from 69 to 65.
15468 (PRED_NULL_RETURN): Change from 91 to 71.
15469 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
15470 (PRED_LOOP_GUARD): Change from 66 to 73.
15472 2018-01-19 Martin Liska <mliska@suse.cz>
15474 * predict.c (predict_insn_def): Add new assert.
15475 (struct branch_predictor): Change type to signed integer.
15476 (test_prediction_value_range): Amend test to cover
15477 PROB_UNINITIALIZED.
15478 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
15479 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
15480 (PRED_LOOP_ITERATIONS_MAX): Likewise.
15481 (PRED_LOOP_IV_COMPARE): Likewise.
15482 * predict.h (PROB_UNINITIALIZED): Define new constant.
15484 2018-01-19 Martin Liska <mliska@suse.cz>
15486 * predict.c (dump_prediction): Add new format for
15487 analyze_brprob.py script which is enabled with -details
15489 * profile-count.h (precise_p): New function.
15491 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
15493 PR tree-optimization/83922
15494 * tree-vect-loop.c (vect_verify_full_masking): Return false if
15495 there are no statements that need masking.
15496 (vect_active_double_reduction_p): New function.
15497 (vect_analyze_loop_operations): Use it when handling phis that
15498 are not in the loop header.
15500 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
15502 PR tree-optimization/83914
15503 * tree-vect-loop.c (vectorizable_induction): Don't convert
15504 init_expr or apply the peeling adjustment for inductions
15505 that are nested within the vectorized loop.
15507 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15509 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
15512 2018-01-18 Jakub Jelinek <jakub@redhat.com>
15516 * function.h (gimplify_parameters): Add gimple_seq * argument.
15517 * function.c: Include gimple.h and options.h.
15518 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
15519 for the added local temporaries if needed.
15520 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
15521 if there are any parameter cleanups, wrap whole body into a
15522 try/finally with the cleanups.
15524 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
15527 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
15528 Use GET_MODE_CLASS for scalar floating point.
15530 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
15534 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
15535 Fix call of call_cgraph_insertion_hooks.
15537 2018-01-18 Martin Sebor <msebor@redhat.com>
15539 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
15541 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
15544 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
15547 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
15550 * common.opt: (-ffile-prefix-map): New option.
15551 * opts.c (common_handle_option): Defer it.
15552 * opts-global.c (handle_common_deferred_options): Handle it.
15553 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
15554 * file-prefix-map.h: New file.
15555 (remap_debug_filename, add_debug_prefix_map): ...here.
15556 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
15557 * final.c (debug_prefix_map, add_debug_prefix_map
15558 remap_debug_filename): Move to...
15559 * file-prefix-map.c: New file.
15560 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
15561 generalize, get rid of alloca(), use strrchr() instead of strchr().
15562 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
15563 Implement in terms of add_prefix_map().
15564 (remap_macro_filename, remap_debug_filename): Implement in term of
15566 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
15567 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
15568 * dbxout.c: Include file-prefix-map.h.
15569 * varasm.c: Likewise.
15570 * vmsdbgout.c: Likewise.
15571 * xcoffout.c: Likewise.
15572 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
15573 * doc/cppopts.texi (-fmacro-prefix-map): Document.
15574 * doc/invoke.texi (-ffile-prefix-map): Document.
15575 (-fdebug-prefix-map): Update description.
15577 2018-01-18 Martin Liska <mliska@suse.cz>
15579 * config/i386/i386.c (indirect_thunk_name): Document that also
15581 (output_indirect_thunk): Document why both instructions
15582 (pause and lfence) are generated.
15584 2018-01-18 Richard Biener <rguenther@suse.de>
15586 PR tree-optimization/83887
15587 * graphite-scop-detection.c
15588 (scop_detection::get_nearest_dom_with_single_entry): Remove.
15589 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
15590 (scop_detection::merge_sese): Re-implement with a flood-fill
15591 algorithm that properly finds a SESE region if it exists.
15593 2018-01-18 Jakub Jelinek <jakub@redhat.com>
15596 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
15597 pointer_diff optimizations use view_convert instead of convert.
15599 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15601 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15602 Generate different code for -mno-speculate-indirect-jumps.
15603 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15604 (*call_indirect_aix<mode>): Disable for
15605 -mno-speculate-indirect-jumps.
15606 (*call_indirect_aix<mode>_nospec): New define_insn.
15607 (*call_value_indirect_aix<mode>): Disable for
15608 -mno-speculate-indirect-jumps.
15609 (*call_value_indirect_aix<mode>_nospec): New define_insn.
15610 (*sibcall_nonlocal_sysv<mode>): Generate different code for
15611 -mno-speculate-indirect-jumps.
15612 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15614 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
15616 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
15617 long double type, set the flags for noting the default long double
15618 type, even if we don't pass or return a long double type.
15620 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
15623 * ipa-inline.c (flatten_function): Do not overwrite final inlining
15626 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
15628 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
15629 support for merge[hl].
15630 (fold_mergehl_helper): New helper function.
15631 (tree-vector-builder.h): New #include for tree_vector_builder usage.
15632 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
15633 (altivec_vmrglw_direct): Add xxmrglw insn.
15635 2018-01-17 Andrew Waterman <andrew@sifive.com>
15637 * config/riscv/riscv.c (riscv_conditional_register_usage): If
15638 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
15640 2018-01-17 David Malcolm <dmalcolm@redhat.com>
15643 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
15644 call the lto_location_cache before reading the
15645 DECL_SOURCE_LOCATION of the types.
15647 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
15648 Richard Sandiford <richard.sandiford@linaro.org>
15650 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
15651 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
15652 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
15653 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
15654 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
15656 * config/aarch64/constraints.md (aarch64_movti_operand):
15658 * config/aarch64/predicates.md (Uti): Add new constraint.
15660 2018-01-17 Carl Love <cel@us.ibm.com>
15662 * config/rs6000/vsx.md (define_expand xl_len_r,
15663 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
15664 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
15666 (define_expand, define_insn): Move the shift left from the
15667 define_insn to the define_expand for lxvl and stxvl instructions.
15668 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
15669 and XL_LEN_R definitions to PURE.
15671 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
15673 * config/i386/i386.c (indirect_thunk_name): Declare regno
15674 as unsigned int. Compare regno with INVALID_REGNUM.
15675 (output_indirect_thunk): Ditto.
15676 (output_indirect_thunk_function): Ditto.
15677 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
15678 in the call to output_indirect_thunk_function.
15680 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
15682 PR middle-end/83884
15683 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
15684 rather than the size of inner_type to determine the stack slot size
15685 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
15687 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
15690 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
15693 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
15695 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
15696 endian Linux systems to optionally enable multilibs for selecting
15697 the long double type if the user configured an explicit type.
15698 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
15699 have no long double multilibs if not defined.
15700 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
15701 warn if the user used -mabi={ieee,ibm}longdouble and we built
15702 multilibs for long double.
15703 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
15704 appropriate multilib option.
15705 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
15707 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
15708 for building long double multilibs.
15709 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
15711 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
15713 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
15716 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
15718 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
15721 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
15724 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
15727 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15729 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
15730 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
15732 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
15734 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
15735 different rtl trees depending on TARGET_64BIT.
15736 (rs6000_gen_lvx): Likewise.
15738 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15740 * config/visium/visium.md (nop): Tweak comment.
15741 (hazard_nop): Likewise.
15743 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15745 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
15746 -mspeculate-indirect-jumps.
15747 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
15748 for -mno-speculate-indirect-jumps.
15749 (*call_indirect_elfv2<mode>_nospec): New define_insn.
15750 (*call_value_indirect_elfv2<mode>): Disable for
15751 -mno-speculate-indirect-jumps.
15752 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
15753 (indirect_jump): Emit different RTL for
15754 -mno-speculate-indirect-jumps.
15755 (*indirect_jump<mode>): Disable for
15756 -mno-speculate-indirect-jumps.
15757 (*indirect_jump<mode>_nospec): New define_insn.
15758 (tablejump): Emit different RTL for
15759 -mno-speculate-indirect-jumps.
15760 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
15761 (tablejumpsi_nospec): New define_expand.
15762 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
15763 (tablejumpdi_nospec): New define_expand.
15764 (*tablejump<mode>_internal1): Disable for
15765 -mno-speculate-indirect-jumps.
15766 (*tablejump<mode>_internal1_nospec): New define_insn.
15767 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
15770 2018-01-16 Artyom Skrobov tyomitch@gmail.com
15772 * caller-save.c (insert_save): Drop unnecessary parameter. All
15775 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15776 Richard Biener <rguenth@suse.de>
15779 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
15780 return early, inline manually is_gimple_sizepos. Make sure if we
15781 call gimplify_expr we don't end up with a gimple constant.
15782 * tree.c (variably_modified_type_p): Don't return true for
15783 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
15784 * gimplify.h (is_gimple_sizepos): Remove.
15786 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15788 PR tree-optimization/83857
15789 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
15790 vectorizable_live_operation for pure SLP statements.
15791 (vectorizable_live_operation): Handle PHIs.
15793 2018-01-16 Richard Biener <rguenther@suse.de>
15795 PR tree-optimization/83867
15796 * tree-vect-stmts.c (vect_transform_stmt): Precompute
15797 nested_in_vect_loop_p since the scalar stmt may get invalidated.
15799 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15802 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
15803 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
15804 If off is not INTEGER_CST, issue a may not be aligned warning
15805 rather than isn't aligned. Use isn%'t rather than isn't.
15806 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
15808 <case MULT_EXPR>: Improve the case when bottom and one of the
15809 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
15810 operand, in that case check if the other operand is multiple of
15811 bottom divided by the INTEGER_CST operand.
15813 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15816 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
15817 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
15818 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
15819 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
15820 * config/pa/pa.c (pa_function_arg_advance): Likewise.
15821 (pa_function_arg, pa_arg_partial_bytes): Likewise.
15822 (pa_function_arg_size): New function.
15824 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15826 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
15827 in a separate statement.
15829 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15831 PR tree-optimization/83847
15832 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
15833 group gathers and scatters.
15835 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15837 PR rtl-optimization/86620
15838 * params.def (max-sched-ready-insns): Bump minimum value to 1.
15840 PR rtl-optimization/83213
15841 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
15842 to last if both are JUMP_INSNs.
15844 PR tree-optimization/83843
15845 * gimple-ssa-store-merging.c
15846 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
15847 store_immediate_info for bswap/nop orig_stores.
15849 2018-01-15 Andrew Waterman <andrew@sifive.com>
15851 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
15853 <UDIV>: Increase cost if !TARGET_DIV.
15855 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
15857 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
15858 (define_attr "cr_logical_3op"): New.
15859 (cceq_ior_compare): Adjust.
15860 (cceq_ior_compare_complement): Adjust.
15861 (*cceq_rev_compare): Adjust.
15862 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
15863 (is_cracked_insn): Adjust.
15864 (insn_must_be_first_in_group): Adjust.
15865 * config/rs6000/40x.md: Adjust.
15866 * config/rs6000/440.md: Adjust.
15867 * config/rs6000/476.md: Adjust.
15868 * config/rs6000/601.md: Adjust.
15869 * config/rs6000/603.md: Adjust.
15870 * config/rs6000/6xx.md: Adjust.
15871 * config/rs6000/7450.md: Adjust.
15872 * config/rs6000/7xx.md: Adjust.
15873 * config/rs6000/8540.md: Adjust.
15874 * config/rs6000/cell.md: Adjust.
15875 * config/rs6000/e300c2c3.md: Adjust.
15876 * config/rs6000/e500mc.md: Adjust.
15877 * config/rs6000/e500mc64.md: Adjust.
15878 * config/rs6000/e5500.md: Adjust.
15879 * config/rs6000/e6500.md: Adjust.
15880 * config/rs6000/mpc.md: Adjust.
15881 * config/rs6000/power4.md: Adjust.
15882 * config/rs6000/power5.md: Adjust.
15883 * config/rs6000/power6.md: Adjust.
15884 * config/rs6000/power7.md: Adjust.
15885 * config/rs6000/power8.md: Adjust.
15886 * config/rs6000/power9.md: Adjust.
15887 * config/rs6000/rs64.md: Adjust.
15888 * config/rs6000/titan.md: Adjust.
15890 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15892 * config/i386/predicates.md (indirect_branch_operand): Rewrite
15893 ix86_indirect_branch_register logic.
15895 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15897 * config/i386/constraints.md (Bs): Update
15898 ix86_indirect_branch_register check. Don't check
15899 ix86_indirect_branch_register with GOT_memory_operand.
15901 * config/i386/predicates.md (GOT_memory_operand): Don't check
15902 ix86_indirect_branch_register here.
15903 (GOT32_symbol_operand): Likewise.
15905 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15907 * config/i386/predicates.md (constant_call_address_operand):
15908 Rewrite ix86_indirect_branch_register logic.
15909 (sibcall_insn_operand): Likewise.
15911 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15913 * config/i386/constraints.md (Bs): Replace
15914 ix86_indirect_branch_thunk_register with
15915 ix86_indirect_branch_register.
15917 * config/i386/i386.md (indirect_jump): Likewise.
15918 (tablejump): Likewise.
15919 (*sibcall_memory): Likewise.
15920 (*sibcall_value_memory): Likewise.
15921 Peepholes of indirect call and jump via memory: Likewise.
15922 * config/i386/i386.opt: Likewise.
15923 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15924 (GOT_memory_operand): Likewise.
15925 (call_insn_operand): Likewise.
15926 (sibcall_insn_operand): Likewise.
15927 (GOT32_symbol_operand): Likewise.
15929 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15931 PR middle-end/83837
15932 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
15933 type rather than type addr's type points to.
15934 (expand_omp_atomic_mutex): Likewise.
15935 (expand_omp_atomic): Likewise.
15937 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15940 * config/i386/i386.c (output_indirect_thunk_function): Use
15941 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
15942 for __x86_return_thunk.
15944 2018-01-15 Richard Biener <rguenther@suse.de>
15946 PR middle-end/83850
15947 * expmed.c (extract_bit_field_1): Fix typo.
15949 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15952 * config/arm/iterators.md (VF): New mode iterator.
15953 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
15954 Remove integer-related logic from pattern.
15955 (neon_vabd<mode>_3): Likewise.
15957 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15959 PR middle-end/82694
15960 * common.opt (fstrict-overflow): No longer an alias.
15961 (fwrapv-pointer): New option.
15962 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
15963 also for pointer types based on flag_wrapv_pointer.
15964 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
15965 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
15966 opts->x_flag_wrapv got set.
15967 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
15968 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
15969 POINTER_TYPE_OVERFLOW_UNDEFINED.
15970 * match.pd: Likewise in address comparison pattern.
15971 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
15973 2018-01-15 Richard Biener <rguenther@suse.de>
15976 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
15977 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
15978 Reset type names to their identifier if their TYPE_DECL doesn't
15979 have linkage (and thus is used for ODR and devirt).
15980 (save_debug_info_for_decl): Remove.
15981 (save_debug_info_for_type): Likewise.
15982 (add_tree_to_fld_list): Adjust.
15983 * tree-pretty-print.c (dump_generic_node): Make dumping of
15984 type names more robust.
15986 2018-01-15 Richard Biener <rguenther@suse.de>
15988 * BASE-VER: Bump to 8.0.1.
15990 2018-01-14 Martin Sebor <msebor@redhat.com>
15993 * builtins.c (check_access): Avoid warning when the no-warning bit
15996 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
15998 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
15999 * ira-color (allocno_hard_regs_compare): Likewise.
16001 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
16004 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
16005 Use .pushsection/.popsection.
16007 2018-01-14 Martin Sebor <msebor@redhat.com>
16010 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
16012 2018-01-14 Jakub Jelinek <jakub@redhat.com>
16014 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
16015 entry from extra_headers.
16016 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
16017 extra_headers, make the list bitwise identical to the i?86-*-* one.
16019 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16021 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
16022 -mcmodel=large with -mindirect-branch=thunk,
16023 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
16024 -mfunction-return=thunk-extern.
16025 * doc/invoke.texi: Document -mcmodel=large is incompatible with
16026 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
16027 -mfunction-return=thunk and -mfunction-return=thunk-extern.
16029 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16031 * config/i386/i386.c (print_reg): Print the name of the full
16032 integer register without '%'.
16033 (ix86_print_operand): Handle 'V'.
16034 * doc/extend.texi: Document 'V' modifier.
16036 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16038 * config/i386/constraints.md (Bs): Disallow memory operand for
16039 -mindirect-branch-register.
16041 * config/i386/predicates.md (indirect_branch_operand): Likewise.
16042 (GOT_memory_operand): Likewise.
16043 (call_insn_operand): Likewise.
16044 (sibcall_insn_operand): Likewise.
16045 (GOT32_symbol_operand): Likewise.
16046 * config/i386/i386.md (indirect_jump): Call convert_memory_address
16047 for -mindirect-branch-register.
16048 (tablejump): Likewise.
16049 (*sibcall_memory): Likewise.
16050 (*sibcall_value_memory): Likewise.
16051 Disallow peepholes of indirect call and jump via memory for
16052 -mindirect-branch-register.
16053 (*call_pop): Replace m with Bw.
16054 (*call_value_pop): Likewise.
16055 (*sibcall_pop_memory): Replace m with Bs.
16056 * config/i386/i386.opt (mindirect-branch-register): New option.
16057 * doc/invoke.texi: Document -mindirect-branch-register option.
16059 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16061 * config/i386/i386-protos.h (ix86_output_function_return): New.
16062 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
16063 set function_return_type.
16064 (indirect_thunk_name): Add ret_p to indicate thunk for function
16066 (output_indirect_thunk_function): Pass false to
16067 indirect_thunk_name.
16068 (ix86_output_indirect_branch_via_reg): Likewise.
16069 (ix86_output_indirect_branch_via_push): Likewise.
16070 (output_indirect_thunk_function): Create alias for function
16071 return thunk if regno < 0.
16072 (ix86_output_function_return): New function.
16073 (ix86_handle_fndecl_attribute): Handle function_return.
16074 (ix86_attribute_table): Add function_return.
16075 * config/i386/i386.h (machine_function): Add
16076 function_return_type.
16077 * config/i386/i386.md (simple_return_internal): Use
16078 ix86_output_function_return.
16079 (simple_return_internal_long): Likewise.
16080 * config/i386/i386.opt (mfunction-return=): New option.
16081 (indirect_branch): Mention -mfunction-return=.
16082 * doc/extend.texi: Document function_return function attribute.
16083 * doc/invoke.texi: Document -mfunction-return= option.
16085 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16087 * config/i386/i386-opts.h (indirect_branch): New.
16088 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
16089 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
16090 with local indirect jump when converting indirect call and jump.
16091 (ix86_set_indirect_branch_type): New.
16092 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
16093 (indirectlabelno): New.
16094 (indirect_thunk_needed): Likewise.
16095 (indirect_thunk_bnd_needed): Likewise.
16096 (indirect_thunks_used): Likewise.
16097 (indirect_thunks_bnd_used): Likewise.
16098 (INDIRECT_LABEL): Likewise.
16099 (indirect_thunk_name): Likewise.
16100 (output_indirect_thunk): Likewise.
16101 (output_indirect_thunk_function): Likewise.
16102 (ix86_output_indirect_branch_via_reg): Likewise.
16103 (ix86_output_indirect_branch_via_push): Likewise.
16104 (ix86_output_indirect_branch): Likewise.
16105 (ix86_output_indirect_jmp): Likewise.
16106 (ix86_code_end): Call output_indirect_thunk_function if needed.
16107 (ix86_output_call_insn): Call ix86_output_indirect_branch if
16109 (ix86_handle_fndecl_attribute): Handle indirect_branch.
16110 (ix86_attribute_table): Add indirect_branch.
16111 * config/i386/i386.h (machine_function): Add indirect_branch_type
16112 and has_local_indirect_jump.
16113 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
16115 (tablejump): Likewise.
16116 (*indirect_jump): Use ix86_output_indirect_jmp.
16117 (*tablejump_1): Likewise.
16118 (simple_return_indirect_internal): Likewise.
16119 * config/i386/i386.opt (mindirect-branch=): New option.
16120 (indirect_branch): New.
16123 (thunk-inline): Likewise.
16124 (thunk-extern): Likewise.
16125 * doc/extend.texi: Document indirect_branch function attribute.
16126 * doc/invoke.texi: Document -mindirect-branch= option.
16128 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
16131 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
16133 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
16135 * ipa-inline.c (want_inline_small_function_p): Return false if
16136 inlining has already failed with CIF_FINAL_ERROR.
16137 (update_caller_keys): Call want_inline_small_function_p before
16139 (update_callee_keys): Likewise.
16141 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
16143 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
16145 (rs6000_quadword_masked_address_p): Likewise.
16146 (quad_aligned_load_p): Likewise.
16147 (quad_aligned_store_p): Likewise.
16148 (const_load_sequence_p): Add comment to describe the outer-most loop.
16149 (mimic_memory_attributes_and_flags): New function.
16150 (rs6000_gen_stvx): Likewise.
16151 (replace_swapped_aligned_store): Likewise.
16152 (rs6000_gen_lvx): Likewise.
16153 (replace_swapped_aligned_load): Likewise.
16154 (replace_swapped_load_constant): Capitalize argument name in
16155 comment describing this function.
16156 (rs6000_analyze_swaps): Add a third pass to search for vector loads
16157 and stores that access quad-word aligned addresses and replace
16158 with stvx or lvx instructions when appropriate.
16159 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
16160 New function prototype.
16161 (rs6000_quadword_masked_address_p): Likewise.
16162 (rs6000_gen_lvx): Likewise.
16163 (rs6000_gen_stvx): Likewise.
16164 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
16165 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
16166 when memory address is aligned.
16167 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
16168 this split to select lvx instruction when memory address is aligned.
16169 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
16170 instruction when memory address is aligned.
16171 (*vsx_le_perm_load_v16qi): Likewise.
16172 (four unnamed splitters): Modify to select the stvx instruction
16173 when memory is aligned.
16175 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
16177 * predict.c (determine_unlikely_bbs): Handle correctly BBs
16178 which appears in the queue multiple times.
16180 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16181 Alan Hayward <alan.hayward@arm.com>
16182 David Sherwood <david.sherwood@arm.com>
16184 * tree-vectorizer.h (vec_lower_bound): New structure.
16185 (_loop_vec_info): Add check_nonzero and lower_bounds.
16186 (LOOP_VINFO_CHECK_NONZERO): New macro.
16187 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
16188 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
16189 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
16190 fields. Make seg_len the distance travelled, not including the
16192 (dr_direction_indicator): Declare.
16193 (dr_zero_step_indicator): Likewise.
16194 (dr_known_forward_stride_p): Likewise.
16195 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
16197 (runtime_alias_check_p): Allow runtime alias checks with
16199 (operator ==): Compare access_size and align.
16200 (prune_runtime_alias_test_list): Rework for new distinction between
16201 the access_size and seg_len.
16202 (create_intersect_range_checks_index): Likewise. Cope with polynomial
16204 (get_segment_min_max): New function.
16205 (create_intersect_range_checks): Use it.
16206 (dr_step_indicator): New function.
16207 (dr_direction_indicator): Likewise.
16208 (dr_zero_step_indicator): Likewise.
16209 (dr_known_forward_stride_p): Likewise.
16210 * tree-loop-distribution.c (data_ref_segment_size): Return
16211 DR_STEP * (niters - 1).
16212 (compute_alias_check_pairs): Update call to the dr_with_seg_len
16214 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
16215 (vect_preserves_scalar_order_p): New function, split out from...
16216 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
16217 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
16218 (vect_vfa_access_size): New function.
16219 (vect_vfa_align): Likewise.
16220 (vect_compile_time_alias): Take access_size_a and access_b arguments.
16221 (dump_lower_bound): New function.
16222 (vect_check_lower_bound): Likewise.
16223 (vect_small_gap_p): Likewise.
16224 (vectorizable_with_step_bound_p): Likewise.
16225 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
16226 depencies if the vectorization factor is 1. Convert the checks
16227 for nonzero steps into checks on the bounds of DR_STEP. Try using
16228 a bunds check for variable steps if the minimum required step is
16229 relatively small. Update calls to the dr_with_seg_len
16230 constructor and to vect_compile_time_alias.
16231 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
16233 (vect_loop_versioning): Call it.
16234 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
16236 (vect_estimate_min_profitable_iters): Account for any bounds checks.
16238 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16239 Alan Hayward <alan.hayward@arm.com>
16240 David Sherwood <david.sherwood@arm.com>
16242 * doc/sourcebuild.texi (vect_scatter_store): Document.
16243 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
16245 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
16247 * genopinit.c (main): Add supports_vec_scatter_store and
16248 supports_vec_scatter_store_cached to target_optabs.
16249 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
16250 IFN_MASK_SCATTER_STORE.
16251 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
16253 * internal-fn.h (internal_store_fn_p): Declare.
16254 (internal_fn_stored_value_index): Likewise.
16255 * internal-fn.c (scatter_store_direct): New macro.
16256 (expand_scatter_store_optab_fn): New function.
16257 (direct_scatter_store_optab_supported_p): New macro.
16258 (internal_store_fn_p): New function.
16259 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
16260 IFN_MASK_SCATTER_STORE.
16261 (internal_fn_mask_index): Likewise.
16262 (internal_fn_stored_value_index): New function.
16263 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
16264 for scatter stores.
16265 * optabs-query.h (supports_vec_scatter_store_p): Declare.
16266 * optabs-query.c (supports_vec_scatter_store_p): New function.
16267 * tree-vectorizer.h (vect_get_store_rhs): Declare.
16268 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
16269 true for scatter stores.
16270 (vect_gather_scatter_fn_p): Handle scatter stores too.
16271 (vect_check_gather_scatter): Consider using scatter stores if
16272 supports_vec_scatter_store_p.
16273 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
16274 scatter stores too.
16275 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
16276 internal_fn_stored_value_index.
16277 (check_load_store_masking): Handle scatter stores too.
16278 (vect_get_store_rhs): Make public.
16279 (vectorizable_call): Use internal_store_fn_p.
16280 (vectorizable_store): Handle scatter store internal functions.
16281 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
16282 when deciding whether the end of the group has been reached.
16283 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
16284 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
16285 (mask_scatter_store<mode>): New insns.
16287 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16288 Alan Hayward <alan.hayward@arm.com>
16289 David Sherwood <david.sherwood@arm.com>
16291 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
16292 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
16293 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
16295 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
16296 Use vect_truncate_gather_scatter_offset if we can't treat the
16297 operation as a normal gather load or scatter store.
16298 (get_group_load_store_type): Take the gather_scatter_info
16299 as argument. Try using a gather load or scatter store for
16300 single-element groups.
16301 (get_load_store_type): Update calls to get_group_load_store_type
16302 and vect_use_strided_gather_scatters_p.
16304 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16305 Alan Hayward <alan.hayward@arm.com>
16306 David Sherwood <david.sherwood@arm.com>
16308 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
16309 optional tree argument.
16310 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
16312 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
16313 but continue to use the current value as a fallback.
16314 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
16315 to compare the updates.
16316 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
16317 (get_load_store_type): Use it when handling a strided access.
16318 (vect_get_strided_load_store_ops): New function.
16319 (vect_get_data_ptr_increment): Likewise.
16320 (vectorizable_load): Handle strided gather loads. Always pass
16321 a step to vect_create_data_ref_ptr and bump_vector_ptr.
16323 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16324 Alan Hayward <alan.hayward@arm.com>
16325 David Sherwood <david.sherwood@arm.com>
16327 * doc/md.texi (gather_load@var{m}): Document.
16328 (mask_gather_load@var{m}): Likewise.
16329 * genopinit.c (main): Add supports_vec_gather_load and
16330 supports_vec_gather_load_cached to target_optabs.
16331 * optabs-tree.c (init_tree_optimization_optabs): Use
16332 ggc_cleared_alloc to allocate target_optabs.
16333 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
16334 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
16336 * internal-fn.h (internal_load_fn_p): Declare.
16337 (internal_gather_scatter_fn_p): Likewise.
16338 (internal_fn_mask_index): Likewise.
16339 (internal_gather_scatter_fn_supported_p): Likewise.
16340 * internal-fn.c (gather_load_direct): New macro.
16341 (expand_gather_load_optab_fn): New function.
16342 (direct_gather_load_optab_supported_p): New macro.
16343 (direct_internal_fn_optab): New function.
16344 (internal_load_fn_p): Likewise.
16345 (internal_gather_scatter_fn_p): Likewise.
16346 (internal_fn_mask_index): Likewise.
16347 (internal_gather_scatter_fn_supported_p): Likewise.
16348 * optabs-query.c (supports_at_least_one_mode_p): New function.
16349 (supports_vec_gather_load_p): Likewise.
16350 * optabs-query.h (supports_vec_gather_load_p): Declare.
16351 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
16352 and memory_type field.
16353 (NUM_PATTERNS): Bump to 15.
16354 * tree-vect-data-refs.c: Include internal-fn.h.
16355 (vect_gather_scatter_fn_p): New function.
16356 (vect_describe_gather_scatter_call): Likewise.
16357 (vect_check_gather_scatter): Try using internal functions for
16358 gather loads. Recognize existing calls to a gather load function.
16359 (vect_analyze_data_refs): Consider using gather loads if
16360 supports_vec_gather_load_p.
16361 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
16362 (vect_get_gather_scatter_offset_type): Likewise.
16363 (vect_convert_mask_for_vectype): Likewise.
16364 (vect_add_conversion_to_patterm): Likewise.
16365 (vect_try_gather_scatter_pattern): Likewise.
16366 (vect_recog_gather_scatter_pattern): New pattern recognizer.
16367 (vect_vect_recog_func_ptrs): Add it.
16368 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
16369 internal_fn_mask_index and internal_gather_scatter_fn_p.
16370 (check_load_store_masking): Take the gather_scatter_info as an
16371 argument and handle gather loads.
16372 (vect_get_gather_scatter_ops): New function.
16373 (vectorizable_call): Check internal_load_fn_p.
16374 (vectorizable_load): Likewise. Handle gather load internal
16376 (vectorizable_store): Update call to check_load_store_masking.
16377 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
16378 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
16379 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
16380 (aarch64_gather_scale_operand_d): New predicates.
16381 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
16382 (mask_gather_load<mode>): New insns.
16384 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16385 Alan Hayward <alan.hayward@arm.com>
16386 David Sherwood <david.sherwood@arm.com>
16388 * optabs.def (fold_left_plus_optab): New optab.
16389 * doc/md.texi (fold_left_plus_@var{m}): Document.
16390 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
16391 * internal-fn.c (fold_left_direct): Define.
16392 (expand_fold_left_optab_fn): Likewise.
16393 (direct_fold_left_optab_supported_p): Likewise.
16394 * fold-const-call.c (fold_const_fold_left): New function.
16395 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
16396 * tree-parloops.c (valid_reduction_p): New function.
16397 (gather_scalar_reductions): Use it.
16398 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
16399 (vect_finish_replace_stmt): Declare.
16400 * tree-vect-loop.c (fold_left_reduction_fn): New function.
16401 (needs_fold_left_reduction_p): New function, split out from...
16402 (vect_is_simple_reduction): ...here. Accept reductions that
16403 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
16404 (vect_force_simple_reduction): Also store the reduction type in
16405 the assignment's STMT_VINFO_REDUC_TYPE.
16406 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
16407 (merge_with_identity): New function.
16408 (vect_expand_fold_left): Likewise.
16409 (vectorize_fold_left_reduction): Likewise.
16410 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
16411 scalar phi in place for it. Check for target support and reject
16412 cases that would reassociate the operation. Defer the transform
16413 phase to vectorize_fold_left_reduction.
16414 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
16415 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
16416 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
16418 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16420 * tree-if-conv.c (predicate_mem_writes): Remove redundant
16421 call to ifc_temp_var.
16423 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16424 Alan Hayward <alan.hayward@arm.com>
16425 David Sherwood <david.sherwood@arm.com>
16427 * target.def (legitimize_address_displacement): Take the original
16428 offset as a poly_int.
16429 * targhooks.h (default_legitimize_address_displacement): Update
16431 * targhooks.c (default_legitimize_address_displacement): Likewise.
16432 * doc/tm.texi: Regenerate.
16433 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
16434 as an argument, moving assert of ad->disp == ad->disp_term to...
16435 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
16436 Try calling targetm.legitimize_address_displacement before expanding
16437 the address rather than afterwards, and adjust for the new interface.
16438 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
16439 Match the new hook interface. Handle SVE addresses.
16440 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
16441 new hook interface.
16443 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16445 * Makefile.in (OBJS): Add early-remat.o.
16446 * target.def (select_early_remat_modes): New hook.
16447 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
16448 * doc/tm.texi: Regenerate.
16449 * targhooks.h (default_select_early_remat_modes): Declare.
16450 * targhooks.c (default_select_early_remat_modes): New function.
16451 * timevar.def (TV_EARLY_REMAT): New timevar.
16452 * passes.def (pass_early_remat): New pass.
16453 * tree-pass.h (make_pass_early_remat): Declare.
16454 * early-remat.c: New file.
16455 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
16457 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
16459 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16460 Alan Hayward <alan.hayward@arm.com>
16461 David Sherwood <david.sherwood@arm.com>
16463 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
16464 vfm1 with a bound_epilog parameter.
16465 (vect_do_peeling): Update calls accordingly, and move the prologue
16466 call earlier in the function. Treat the base bound_epilog as 0 for
16467 fully-masked loops and retain vf - 1 for other loops. Add 1 to
16468 this base when peeling for gaps.
16469 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
16470 with fully-masked loops.
16471 (vect_estimate_min_profitable_iters): Handle the single peeled
16472 iteration in that case.
16474 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16475 Alan Hayward <alan.hayward@arm.com>
16476 David Sherwood <david.sherwood@arm.com>
16478 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
16479 single-element interleaving even if the size is not a power of 2.
16480 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
16481 accesses for single-element interleaving if the group size is
16484 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16485 Alan Hayward <alan.hayward@arm.com>
16486 David Sherwood <david.sherwood@arm.com>
16488 * doc/md.texi (fold_extract_last_@var{m}): Document.
16489 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
16490 * optabs.def (fold_extract_last_optab): New optab.
16491 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
16492 * internal-fn.c (fold_extract_direct): New macro.
16493 (expand_fold_extract_optab_fn): Likewise.
16494 (direct_fold_extract_optab_supported_p): Likewise.
16495 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
16496 * tree-vect-loop.c (vect_model_reduction_cost): Handle
16497 EXTRACT_LAST_REDUCTION.
16498 (get_initial_def_for_reduction): Do not create an initial vector
16499 for EXTRACT_LAST_REDUCTION reductions.
16500 (vectorizable_reduction): Leave the scalar phi in place for
16501 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
16502 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
16503 epilogue code for EXTRACT_LAST_REDUCTION and defer the
16504 transform phase to vectorizable_condition.
16505 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
16507 (vect_finish_stmt_generation): ...here.
16508 (vect_finish_replace_stmt): New function.
16509 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
16510 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
16512 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
16514 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16515 Alan Hayward <alan.hayward@arm.com>
16516 David Sherwood <david.sherwood@arm.com>
16518 * doc/md.texi (extract_last_@var{m}): Document.
16519 * optabs.def (extract_last_optab): New optab.
16520 * internal-fn.def (EXTRACT_LAST): New internal function.
16521 * internal-fn.c (cond_unary_direct): New macro.
16522 (expand_cond_unary_optab_fn): Likewise.
16523 (direct_cond_unary_optab_supported_p): Likewise.
16524 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
16525 loops using EXTRACT_LAST.
16526 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
16527 (extract_last_<mode>): ...this optab.
16528 (vec_extract<mode><Vel>): Update accordingly.
16530 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16531 Alan Hayward <alan.hayward@arm.com>
16532 David Sherwood <david.sherwood@arm.com>
16534 * target.def (empty_mask_is_expensive): New hook.
16535 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
16536 * doc/tm.texi: Regenerate.
16537 * targhooks.h (default_empty_mask_is_expensive): Declare.
16538 * targhooks.c (default_empty_mask_is_expensive): New function.
16539 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
16540 if the target says that empty masks are expensive.
16541 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
16543 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
16545 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16546 Alan Hayward <alan.hayward@arm.com>
16547 David Sherwood <david.sherwood@arm.com>
16549 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
16550 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
16551 (vect_use_loop_mask_for_alignment_p): New function.
16552 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
16553 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
16554 niters_skip argument. Make sure that the first niters_skip elements
16555 of the first iteration are inactive.
16556 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
16557 Update call to vect_set_loop_masks_directly.
16558 (get_misalign_in_elems): New function, split out from...
16559 (vect_gen_prolog_loop_niters): ...here.
16560 (vect_update_init_of_dr): Take a code argument that specifies whether
16561 the adjustment should be added or subtracted.
16562 (vect_update_init_of_drs): Likewise.
16563 (vect_prepare_for_masked_peels): New function.
16564 (vect_do_peeling): Skip prologue peeling if we're using a mask
16565 instead. Update call to vect_update_inits_of_drs.
16566 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16568 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
16569 alignment. Do not include the number of peeled iterations in
16570 the minimum threshold in that case.
16571 (vectorizable_induction): Adjust the start value down by
16572 LOOP_VINFO_MASK_SKIP_NITERS iterations.
16573 (vect_transform_loop): Call vect_prepare_for_masked_peels.
16574 Take the number of skipped iterations into account when calculating
16576 * tree-vect-stmts.c (vect_gen_while_not): New function.
16578 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16579 Alan Hayward <alan.hayward@arm.com>
16580 David Sherwood <david.sherwood@arm.com>
16582 * doc/sourcebuild.texi (vect_fully_masked): Document.
16583 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
16584 default value to 0.
16585 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
16587 (vect_analyze_loop_2): ...here. Don't check the vectorization
16588 factor against the number of loop iterations if the loop is
16591 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16592 Alan Hayward <alan.hayward@arm.com>
16593 David Sherwood <david.sherwood@arm.com>
16595 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
16596 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
16597 (dump_groups): Update accordingly.
16598 (iv_use::mem_type): New member variable.
16599 (address_p): New function.
16600 (record_use): Add a mem_type argument and initialize the new
16602 (record_group_use): Add a mem_type argument. Use address_p.
16603 Remove obsolete null checks of base_object. Update call to record_use.
16604 (find_interesting_uses_op): Update call to record_group_use.
16605 (find_interesting_uses_cond): Likewise.
16606 (find_interesting_uses_address): Likewise.
16607 (get_mem_type_for_internal_fn): New function.
16608 (find_address_like_use): Likewise.
16609 (find_interesting_uses_stmt): Try find_address_like_use before
16610 calling find_interesting_uses_op.
16611 (addr_offset_valid_p): Use the iv mem_type field as the type
16612 of the addressed memory.
16613 (add_autoinc_candidates): Likewise.
16614 (get_address_cost): Likewise.
16615 (split_small_address_groups_p): Use address_p.
16616 (split_address_groups): Likewise.
16617 (add_iv_candidate_for_use): Likewise.
16618 (autoinc_possible_for_pair): Likewise.
16619 (rewrite_groups): Likewise.
16620 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
16621 (determine_group_iv_cost): Update after split of USE_ADDRESS.
16622 (get_alias_ptr_type_for_ptr_address): New function.
16623 (rewrite_use_address): Rewrite address uses in calls that were
16624 identified by find_address_like_use.
16626 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16627 Alan Hayward <alan.hayward@arm.com>
16628 David Sherwood <david.sherwood@arm.com>
16630 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
16632 * gimple-expr.h (is_gimple_addressable: Likewise.
16633 * gimple-expr.c (is_gimple_address): Likewise.
16634 * internal-fn.c (expand_call_mem_ref): New function.
16635 (expand_mask_load_optab_fn): Use it.
16636 (expand_mask_store_optab_fn): Likewise.
16638 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16639 Alan Hayward <alan.hayward@arm.com>
16640 David Sherwood <david.sherwood@arm.com>
16642 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
16643 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
16644 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
16645 (cond_umax@var{mode}): Document.
16646 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
16647 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
16648 (cond_umin_optab, cond_umax_optab): New optabs.
16649 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
16650 (COND_IOR, COND_XOR): New internal functions.
16651 * internal-fn.h (get_conditional_internal_fn): Declare.
16652 * internal-fn.c (cond_binary_direct): New macro.
16653 (expand_cond_binary_optab_fn): Likewise.
16654 (direct_cond_binary_optab_supported_p): Likewise.
16655 (get_conditional_internal_fn): New function.
16656 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
16657 Cope with reduction statements that are vectorized as calls rather
16659 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
16660 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
16661 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
16662 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
16663 (UNSPEC_COND_EOR): New unspecs.
16664 (optab): Add mappings for them.
16665 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
16666 (sve_int_op, sve_fp_op): New int attributes.
16668 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16669 Alan Hayward <alan.hayward@arm.com>
16670 David Sherwood <david.sherwood@arm.com>
16672 * optabs.def (while_ult_optab): New optab.
16673 * doc/md.texi (while_ult@var{m}@var{n}): Document.
16674 * internal-fn.def (WHILE_ULT): New internal function.
16675 * internal-fn.h (direct_internal_fn_supported_p): New override
16676 that takes two types as argument.
16677 * internal-fn.c (while_direct): New macro.
16678 (expand_while_optab_fn): New function.
16679 (convert_optab_supported_p): Likewise.
16680 (direct_while_optab_supported_p): New macro.
16681 * wide-int.h (wi::udiv_ceil): New function.
16682 * tree-vectorizer.h (rgroup_masks): New structure.
16683 (vec_loop_masks): New typedef.
16684 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
16685 and fully_masked_p.
16686 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
16687 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
16688 (vect_max_vf): New function.
16689 (slpeel_make_loop_iterate_ntimes): Delete.
16690 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
16691 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
16692 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
16693 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
16694 internal-fn.h, stor-layout.h and optabs-query.h.
16695 (vect_set_loop_mask): New function.
16696 (add_preheader_seq): Likewise.
16697 (add_header_seq): Likewise.
16698 (interleave_supported_p): Likewise.
16699 (vect_maybe_permute_loop_masks): Likewise.
16700 (vect_set_loop_masks_directly): Likewise.
16701 (vect_set_loop_condition_masked): Likewise.
16702 (vect_set_loop_condition_unmasked): New function, split out from
16703 slpeel_make_loop_iterate_ntimes.
16704 (slpeel_make_loop_iterate_ntimes): Rename to..
16705 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
16706 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
16707 (vect_do_peeling): Update call accordingly.
16708 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
16710 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16711 mask_compare_type, can_fully_mask_p and fully_masked_p.
16712 (release_vec_loop_masks): New function.
16713 (_loop_vec_info): Use it to free the loop masks.
16714 (can_produce_all_loop_masks_p): New function.
16715 (vect_get_max_nscalars_per_iter): Likewise.
16716 (vect_verify_full_masking): Likewise.
16717 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
16718 retries, and free the mask rgroups before retrying. Check loop-wide
16719 reasons for disallowing fully-masked loops. Make the final decision
16720 about whether use a fully-masked loop or not.
16721 (vect_estimate_min_profitable_iters): Do not assume that peeling
16722 for the number of iterations will be needed for fully-masked loops.
16723 (vectorizable_reduction): Disable fully-masked loops.
16724 (vectorizable_live_operation): Likewise.
16725 (vect_halve_mask_nunits): New function.
16726 (vect_double_mask_nunits): Likewise.
16727 (vect_record_loop_mask): Likewise.
16728 (vect_get_loop_mask): Likewise.
16729 (vect_transform_loop): Handle the case in which the final loop
16730 iteration might handle a partial vector. Call vect_set_loop_condition
16731 instead of slpeel_make_loop_iterate_ntimes.
16732 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
16733 (check_load_store_masking): New function.
16734 (prepare_load_store_mask): Likewise.
16735 (vectorizable_store): Handle fully-masked loops.
16736 (vectorizable_load): Likewise.
16737 (supportable_widening_operation): Use vect_halve_mask_nunits for
16739 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
16740 (vect_gen_while): New function.
16741 * config/aarch64/aarch64.md (umax<mode>3): New expander.
16742 (aarch64_uqdec<mode>): New insn.
16744 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16745 Alan Hayward <alan.hayward@arm.com>
16746 David Sherwood <david.sherwood@arm.com>
16748 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
16749 (reduc_xor_scal_optab): New optabs.
16750 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
16751 (reduc_xor_scal_@var{m}): Document.
16752 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
16753 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
16754 internal functions.
16755 * fold-const-call.c (fold_const_call): Handle them.
16756 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
16757 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
16758 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
16759 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
16760 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
16761 (UNSPEC_XORV): New unspecs.
16762 (optab): Add entries for them.
16763 (BITWISEV): New int iterator.
16764 (bit_reduc_op): New int attributes.
16766 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16767 Alan Hayward <alan.hayward@arm.com>
16768 David Sherwood <david.sherwood@arm.com>
16770 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
16771 * internal-fn.def (VEC_SHL_INSERT): New internal function.
16772 * optabs.def (vec_shl_insert_optab): New optab.
16773 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
16774 (duplicate_and_interleave): Likewise.
16775 * tree-vect-loop.c: Include internal-fn.h.
16776 (neutral_op_for_slp_reduction): New function, split out from
16777 get_initial_defs_for_reduction.
16778 (get_initial_def_for_reduction): Handle option 2 for variable-length
16779 vectors by loading the neutral value into a vector and then shifting
16780 the initial value into element 0.
16781 (get_initial_defs_for_reduction): Replace the code argument with
16782 the neutral value calculated by neutral_op_for_slp_reduction.
16783 Use gimple_build_vector for constant-length vectors.
16784 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
16785 but the first group_size elements have a neutral value.
16786 Use duplicate_and_interleave otherwise.
16787 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
16788 Update call to get_initial_defs_for_reduction. Handle SLP
16789 reductions for variable-length vectors by creating one vector
16790 result for each scalar result, with the elements associated
16791 with other scalar results stubbed out with the neutral value.
16792 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
16793 Require IFN_VEC_SHL_INSERT for double reductions on
16794 variable-length vectors, or SLP reductions that have
16795 a neutral value. Require can_duplicate_and_interleave_p
16796 support for variable-length unchained SLP reductions if there
16797 is no neutral value, such as for MIN/MAX reductions. Also require
16798 the number of vector elements to be a multiple of the number of
16799 SLP statements when doing variable-length unchained SLP reductions.
16800 Update call to vect_create_epilog_for_reduction.
16801 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
16802 and remove initial values.
16803 (duplicate_and_interleave): Make public.
16804 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
16805 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
16807 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16808 Alan Hayward <alan.hayward@arm.com>
16809 David Sherwood <david.sherwood@arm.com>
16811 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
16812 (can_duplicate_and_interleave_p): New function.
16813 (vect_get_and_check_slp_defs): Take the vector of statements
16814 rather than just the current one. Remove excess parentheses.
16815 Restriction rejectinon of vect_constant_def and vect_external_def
16816 for variable-length vectors to boolean types, or types for which
16817 can_duplicate_and_interleave_p is false.
16818 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
16819 (duplicate_and_interleave): New function.
16820 (vect_get_constant_vectors): Use gimple_build_vector for
16821 constant-length vectors and suitable variable-length constant
16822 vectors. Use duplicate_and_interleave for other variable-length
16823 vectors. Don't defer the update when inserting new statements.
16825 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16826 Alan Hayward <alan.hayward@arm.com>
16827 David Sherwood <david.sherwood@arm.com>
16829 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
16830 min_profitable_iters doesn't go negative.
16832 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16833 Alan Hayward <alan.hayward@arm.com>
16834 David Sherwood <david.sherwood@arm.com>
16836 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
16837 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
16838 * optabs.def (vec_mask_load_lanes_optab): New optab.
16839 (vec_mask_store_lanes_optab): Likewise.
16840 * internal-fn.def (MASK_LOAD_LANES): New internal function.
16841 (MASK_STORE_LANES): Likewise.
16842 * internal-fn.c (mask_load_lanes_direct): New macro.
16843 (mask_store_lanes_direct): Likewise.
16844 (expand_mask_load_optab_fn): Handle masked operations.
16845 (expand_mask_load_lanes_optab_fn): New macro.
16846 (expand_mask_store_optab_fn): Handle masked operations.
16847 (expand_mask_store_lanes_optab_fn): New macro.
16848 (direct_mask_load_lanes_optab_supported_p): Likewise.
16849 (direct_mask_store_lanes_optab_supported_p): Likewise.
16850 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
16852 (vect_load_lanes_supported): Likewise.
16853 * tree-vect-data-refs.c (strip_conversion): New function.
16854 (can_group_stmts_p): Likewise.
16855 (vect_analyze_data_ref_accesses): Use it instead of checking
16856 for a pair of assignments.
16857 (vect_store_lanes_supported): Take a masked_p parameter.
16858 (vect_load_lanes_supported): Likewise.
16859 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
16860 vect_store_lanes_supported and vect_load_lanes_supported.
16861 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
16862 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
16863 parameter. Don't allow gaps for masked accesses.
16864 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
16865 and vect_load_lanes_supported.
16866 (get_load_store_type): Take a masked_p parameter and update
16867 call to get_group_load_store_type.
16868 (vectorizable_store): Update call to get_load_store_type.
16869 Handle IFN_MASK_STORE_LANES.
16870 (vectorizable_load): Update call to get_load_store_type.
16871 Handle IFN_MASK_LOAD_LANES.
16873 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16874 Alan Hayward <alan.hayward@arm.com>
16875 David Sherwood <david.sherwood@arm.com>
16877 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
16879 * config/aarch64/aarch64-protos.h
16880 (aarch64_sve_struct_memory_operand_p): Declare.
16881 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
16882 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
16883 (VPRED, vpred): Handle SVE structure modes.
16884 * config/aarch64/constraints.md (Utx): New constraint.
16885 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
16886 (aarch64_sve_struct_nonimmediate_operand): New predicates.
16887 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
16888 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
16889 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
16890 structure modes. Split into pieces after RA.
16891 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
16892 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
16894 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
16895 SVE structure modes.
16896 (aarch64_classify_address): Likewise.
16897 (sizetochar): Move earlier in file.
16898 (aarch64_print_operand): Handle SVE register lists.
16899 (aarch64_array_mode): New function.
16900 (aarch64_sve_struct_memory_operand_p): Likewise.
16901 (TARGET_ARRAY_MODE): Redefine.
16903 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16904 Alan Hayward <alan.hayward@arm.com>
16905 David Sherwood <david.sherwood@arm.com>
16907 * target.def (array_mode): New target hook.
16908 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
16909 * doc/tm.texi: Regenerate.
16910 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
16911 * hooks.c (hook_optmode_mode_uhwi_none): New function.
16912 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
16913 targetm.array_mode.
16914 * stor-layout.c (mode_for_array): Likewise. Support polynomial
16917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16918 Alan Hayward <alan.hayward@arm.com>
16919 David Sherwood <david.sherwood@arm.com>
16921 * fold-const.c (fold_binary_loc): Check the argument types
16922 rather than the result type when testing for a vector operation.
16924 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16926 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
16927 * doc/tm.texi: Regenerate.
16929 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16930 Alan Hayward <alan.hayward@arm.com>
16931 David Sherwood <david.sherwood@arm.com>
16933 * doc/invoke.texi (-msve-vector-bits=): Document new option.
16934 (sve): Document new AArch64 extension.
16935 * doc/md.texi (w): Extend the description of the AArch64
16936 constraint to include SVE vectors.
16937 (Upl, Upa): Document new AArch64 predicate constraints.
16938 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
16940 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
16941 (msve-vector-bits=): New option.
16942 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
16943 SVE when these are disabled.
16944 (sve): New extension.
16945 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
16946 modes. Adjust their number of units based on aarch64_sve_vg.
16947 (MAX_BITSIZE_MODE_ANY_MODE): Define.
16948 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
16949 aarch64_addr_query_type.
16950 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
16951 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
16952 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
16953 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
16954 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
16955 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
16956 (aarch64_simd_imm_zero_p): Delete.
16957 (aarch64_check_zero_based_sve_index_immediate): Declare.
16958 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16959 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16960 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16961 (aarch64_sve_float_mul_immediate_p): Likewise.
16962 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16963 rather than an rtx.
16964 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
16965 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
16966 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
16967 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
16968 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
16969 (aarch64_regmode_natural_size): Likewise.
16970 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
16971 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
16973 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
16974 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
16975 for VG and the SVE predicate registers.
16976 (V_ALIASES): Add a "z"-prefixed alias.
16977 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
16978 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
16979 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
16980 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
16981 (REG_CLASS_NAMES): Add entries for them.
16982 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
16983 and the predicate registers.
16984 (aarch64_sve_vg): Declare.
16985 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
16986 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
16987 (REGMODE_NATURAL_SIZE): Define.
16988 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
16990 * config/aarch64/aarch64.c: Include cfgrtl.h.
16991 (simd_immediate_info): Add a constructor for series vectors,
16992 and an associated step field.
16993 (aarch64_sve_vg): New variable.
16994 (aarch64_dbx_register_number): Handle VG and the predicate registers.
16995 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
16996 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
16997 (VEC_ANY_DATA, VEC_STRUCT): New constants.
16998 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
16999 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
17000 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
17001 (aarch64_get_mask_mode): New functions.
17002 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
17003 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
17004 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
17005 predicate modes and predicate registers. Explicitly restrict
17006 GPRs to modes of 16 bytes or smaller. Only allow FP registers
17007 to store a vector mode if it is recognized by
17008 aarch64_classify_vector_mode.
17009 (aarch64_regmode_natural_size): New function.
17010 (aarch64_hard_regno_caller_save_mode): Return the original mode
17012 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
17013 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
17014 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
17015 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
17017 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
17018 does not overlap dest if the function is frame-related. Handle
17020 (aarch64_split_add_offset): New function.
17021 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
17022 them aarch64_add_offset.
17023 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
17024 and update call to aarch64_sub_sp.
17025 (aarch64_add_cfa_expression): New function.
17026 (aarch64_expand_prologue): Pass extra temporary registers to the
17027 functions above. Handle the case in which we need to emit new
17028 DW_CFA_expressions for registers that were originally saved
17029 relative to the stack pointer, but now have to be expressed
17030 relative to the frame pointer.
17031 (aarch64_output_mi_thunk): Pass extra temporary registers to the
17033 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
17034 IP0 and IP1 values for SVE frames.
17035 (aarch64_expand_vec_series): New function.
17036 (aarch64_expand_sve_widened_duplicate): Likewise.
17037 (aarch64_expand_sve_const_vector): Likewise.
17038 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
17039 Handle SVE constants. Use emit_move_insn to move a force_const_mem
17040 into the register, rather than emitting a SET directly.
17041 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
17042 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
17043 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
17044 (offset_9bit_signed_scaled_p): New functions.
17045 (aarch64_replicate_bitmask_imm): New function.
17046 (aarch64_bitmask_imm): Use it.
17047 (aarch64_cannot_force_const_mem): Reject expressions involving
17048 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
17049 (aarch64_classify_index): Handle SVE indices, by requiring
17050 a plain register index with a scale that matches the element size.
17051 (aarch64_classify_address): Handle SVE addresses. Assert that
17052 the mode of the address is VOIDmode or an integer mode.
17053 Update call to aarch64_classify_symbol.
17054 (aarch64_classify_symbolic_expression): Update call to
17055 aarch64_classify_symbol.
17056 (aarch64_const_vec_all_in_range_p): New function.
17057 (aarch64_print_vector_float_operand): Likewise.
17058 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
17059 "vN" for FP registers with SVE modes. Handle (const ...) vectors
17060 and the FP immediates 1.0 and 0.5.
17061 (aarch64_print_address_internal): Handle SVE addresses.
17062 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
17063 (aarch64_regno_regclass): Handle predicate registers.
17064 (aarch64_secondary_reload): Handle big-endian reloads of SVE
17066 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
17067 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
17068 (aarch64_convert_sve_vector_bits): New function.
17069 (aarch64_override_options): Use it to handle -msve-vector-bits=.
17070 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
17071 rather than an rtx.
17072 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
17073 Handle SVE vector and predicate modes. Accept VL-based constants
17074 that need only one temporary register, and VL offsets that require
17075 no temporary registers.
17076 (aarch64_conditional_register_usage): Mark the predicate registers
17077 as fixed if SVE isn't available.
17078 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
17079 Return true for SVE vector and predicate modes.
17080 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
17081 rather than an unsigned int. Handle SVE modes.
17082 (aarch64_preferred_simd_mode): Update call accordingly. Handle
17084 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
17086 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
17087 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
17088 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
17089 (aarch64_sve_float_mul_immediate_p): New functions.
17090 (aarch64_sve_valid_immediate): New function.
17091 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
17092 Explicitly reject structure modes. Check for INDEX constants.
17093 Handle PTRUE and PFALSE constants.
17094 (aarch64_check_zero_based_sve_index_immediate): New function.
17095 (aarch64_simd_imm_zero_p): Delete.
17096 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
17097 vector modes. Accept constants in the range of CNT[BHWD].
17098 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
17099 ask for an Advanced SIMD mode.
17100 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
17101 (aarch64_simd_vector_alignment): Handle SVE predicates.
17102 (aarch64_vectorize_preferred_vector_alignment): New function.
17103 (aarch64_simd_vector_alignment_reachable): Use it instead of
17105 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
17106 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
17108 (MAX_VECT_LEN): Delete.
17109 (expand_vec_perm_d): Add a vec_flags field.
17110 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
17111 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
17112 (aarch64_evpc_ext): Don't apply a big-endian lane correction
17114 (aarch64_evpc_rev): Rename to...
17115 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
17116 (aarch64_evpc_rev_global): New function.
17117 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
17118 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
17120 (aarch64_evpc_sve_tbl): New function.
17121 (aarch64_expand_vec_perm_const_1): Update after rename of
17122 aarch64_evpc_rev. Handle SVE permutes too, trying
17123 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
17124 than aarch64_evpc_tbl.
17125 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
17126 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
17127 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
17128 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
17129 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
17130 (aarch64_expand_sve_vcond): New functions.
17131 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
17132 of aarch64_vector_mode_p.
17133 (aarch64_dwarf_poly_indeterminate_value): New function.
17134 (aarch64_compute_pressure_classes): Likewise.
17135 (aarch64_can_change_mode_class): Likewise.
17136 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
17137 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
17138 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
17139 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
17140 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
17141 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
17142 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
17143 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
17145 (Dn, Dl, Dr): Accept const as well as const_vector.
17146 (Dz): Likewise. Compare against CONST0_RTX.
17147 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
17148 of "vector" where appropriate.
17149 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
17150 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
17151 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
17152 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
17153 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
17154 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
17155 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
17156 (v_int_equiv): Extend to SVE modes.
17157 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
17159 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
17160 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
17161 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
17162 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
17163 (SVE_COND_FP_CMP): New int iterators.
17164 (perm_hilo): Handle the new unpack unspecs.
17165 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
17167 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
17168 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
17169 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
17170 (aarch64_equality_operator, aarch64_constant_vector_operand)
17171 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
17172 (aarch64_sve_nonimmediate_operand): Likewise.
17173 (aarch64_sve_general_operand): Likewise.
17174 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
17175 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
17176 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
17177 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
17178 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
17179 (aarch64_sve_float_arith_immediate): Likewise.
17180 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
17181 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
17182 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
17183 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
17184 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
17185 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
17186 (aarch64_sve_float_arith_operand): Likewise.
17187 (aarch64_sve_float_arith_with_sub_operand): Likewise.
17188 (aarch64_sve_float_mul_operand): Likewise.
17189 (aarch64_sve_vec_perm_operand): Likewise.
17190 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
17191 (aarch64_mov_operand): Accept const_poly_int and const_vector.
17192 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
17193 as well as const_vector.
17194 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
17195 in file. Use CONST0_RTX and CONSTM1_RTX.
17196 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
17197 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
17198 Use aarch64_simd_imm_zero.
17199 * config/aarch64/aarch64-sve.md: New file.
17200 * config/aarch64/aarch64.md: Include it.
17201 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
17202 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
17203 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
17204 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
17205 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
17206 (sve): New attribute.
17207 (enabled): Disable instructions with the sve attribute unless
17209 (movqi, movhi): Pass CONST_POLY_INT operaneds through
17210 aarch64_expand_mov_immediate.
17211 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
17212 CNT[BHSD] immediates.
17213 (movti): Split CONST_POLY_INT moves into two halves.
17214 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
17215 Split additions that need a temporary here if the destination
17216 is the stack pointer.
17217 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
17218 (*add<mode>3_poly_1): New instruction.
17219 (set_clobber_cc): New expander.
17221 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17223 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
17224 parameter and use it instead of GET_MODE_SIZE (innermode). Use
17225 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
17226 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
17227 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
17228 Change innermode from fixed_mode_size to machine_mode.
17229 (simplify_subreg): Update call accordingly. Handle a constant-sized
17230 subreg of a variable-length CONST_VECTOR.
17232 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17233 Alan Hayward <alan.hayward@arm.com>
17234 David Sherwood <david.sherwood@arm.com>
17236 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
17237 (add_offset_to_base): New function, split out from...
17238 (create_mem_ref): ...here. When handling a scale other than 1,
17239 check first whether the address is valid without the offset.
17240 Add it into the base if so, leaving the index and scale as-is.
17242 2018-01-12 Jakub Jelinek <jakub@redhat.com>
17245 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
17246 fold_for_warn before checking if arg2 is INTEGER_CST.
17248 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
17250 * config/rs6000/predicates.md (load_multiple_operation): Delete.
17251 (store_multiple_operation): Delete.
17252 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
17253 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
17254 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
17255 guarded by TARGET_STRING.
17256 (rs6000_output_load_multiple): Delete.
17257 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
17258 OPTION_MASK_STRING / TARGET_STRING handling.
17259 (print_operand) <'N', 'O'>: Add comment that these are unused now.
17260 (const rs6000_opt_masks) <"string">: Change mask to 0.
17261 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
17262 (MASK_STRING): Delete.
17263 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
17265 (load_multiple): Delete.
17272 (store_multiple): Delete.
17279 (movmemsi_8reg): Delete.
17280 (corresponding unnamed define_insn): Delete.
17281 (movmemsi_6reg): Delete.
17282 (corresponding unnamed define_insn): Delete.
17283 (movmemsi_4reg): Delete.
17284 (corresponding unnamed define_insn): Delete.
17285 (movmemsi_2reg): Delete.
17286 (corresponding unnamed define_insn): Delete.
17287 (movmemsi_1reg): Delete.
17288 (corresponding unnamed define_insn): Delete.
17289 * config/rs6000/rs6000.opt (mno-string): New.
17290 (mstring): Replace by deprecation warning stub.
17291 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
17293 2018-01-12 Jakub Jelinek <jakub@redhat.com>
17295 * regrename.c (regrename_do_replace): If replacing the same
17296 reg multiple times, try to reuse last created gen_raw_REG.
17299 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
17300 main to workaround a bug in GDB.
17302 2018-01-12 Tom de Vries <tom@codesourcery.com>
17305 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
17307 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
17309 PR rtl-optimization/80481
17310 * ira-color.c (get_cap_member): New function.
17311 (allocnos_conflict_by_live_ranges_p): Use it.
17312 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
17313 (setup_slot_coalesced_allocno_live_ranges): Ditto.
17315 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
17318 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
17319 (*saddl_se_1): Ditto.
17320 (*ssubsi_1): Ditto.
17321 (*ssubl_se_1): Ditto.
17323 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17325 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
17326 rather than wi::to_widest for DR_INITs.
17327 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
17328 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
17329 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
17331 (vect_analyze_group_access_1): Note that here.
17333 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17335 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
17336 polynomial type sizes.
17338 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17340 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
17341 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
17342 (gimple_add_tmp_var): Likewise.
17344 2018-01-12 Martin Liska <mliska@suse.cz>
17346 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
17347 (gimple_alloc_sizes): Likewise.
17348 (dump_gimple_statistics): Use PRIu64 in printf format.
17349 * gimple.h: Change uint64_t to int.
17351 2018-01-12 Martin Liska <mliska@suse.cz>
17353 * tree-core.h: Use uint64_t instead of int.
17354 * tree.c (tree_node_counts): Likewise.
17355 (tree_node_sizes): Likewise.
17356 (dump_tree_statistics): Use PRIu64 in printf format.
17358 2018-01-12 Martin Liska <mliska@suse.cz>
17360 * Makefile.in: As qsort_chk is implemented in vec.c, add
17361 vec.o to linkage of gencfn-macros.
17362 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
17363 passing the info to record_node_allocation_statistics.
17364 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
17366 * ggc-common.c (struct ggc_usage): Add operator== and use
17367 it in operator< and compare function.
17368 * mem-stats.h (struct mem_usage): Likewise.
17369 * vec.c (struct vec_usage): Remove operator< and compare
17370 function. Can be simply inherited.
17372 2018-01-12 Martin Jambor <mjambor@suse.cz>
17375 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
17376 * tree-ssa-math-opts.c: Include domwalk.h.
17377 (convert_mult_to_fma_1): New function.
17378 (fma_transformation_info): New type.
17379 (fma_deferring_state): Likewise.
17380 (cancel_fma_deferring): New function.
17381 (result_of_phi): Likewise.
17382 (last_fma_candidate_feeds_initial_phi): Likewise.
17383 (convert_mult_to_fma): Added deferring logic, split actual
17384 transformation to convert_mult_to_fma_1.
17385 (math_opts_dom_walker): New type.
17386 (math_opts_dom_walker::after_dom_children): New method, body moved
17387 here from pass_optimize_widening_mul::execute, added deferring logic
17389 (pass_optimize_widening_mul::execute): Moved most of code to
17390 math_opts_dom_walker::after_dom_children.
17391 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
17392 * config/i386/i386.c (ix86_option_override_internal): Added
17393 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
17395 2018-01-12 Richard Biener <rguenther@suse.de>
17398 * dwarf2out.c (gen_variable_die): Do not reset old_die for
17399 inline instance vars.
17401 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
17404 * config/rx/rx.c (rx_is_restricted_memory_address):
17405 Handle SUBREG case.
17407 2018-01-12 Richard Biener <rguenther@suse.de>
17409 PR tree-optimization/80846
17410 * target.def (split_reduction): New target hook.
17411 * targhooks.c (default_split_reduction): New function.
17412 * targhooks.h (default_split_reduction): Declare.
17413 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
17414 target requests first reduce vectors by combining low and high
17416 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
17417 (get_vectype_for_scalar_type_and_size): Export.
17418 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
17419 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
17420 * doc/tm.texi: Regenerate.
17421 * config/i386/i386.c (ix86_split_reduction): Implement
17422 TARGET_VECTORIZE_SPLIT_REDUCTION.
17424 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17427 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
17428 in PIC mode except for TARGET_VXWORKS_RTP.
17429 * config/sparc/sparc.c: Include cfgrtl.h.
17430 (TARGET_INIT_PIC_REG): Define.
17431 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
17432 (sparc_pic_register_p): New predicate.
17433 (sparc_legitimate_address_p): Use it.
17434 (sparc_legitimize_pic_address): Likewise.
17435 (sparc_delegitimize_address): Likewise.
17436 (sparc_mode_dependent_address_p): Likewise.
17437 (gen_load_pcrel_sym): Remove 4th parameter.
17438 (load_got_register): Adjust call to above. Remove obsolete stuff.
17439 (sparc_expand_prologue): Do not call load_got_register here.
17440 (sparc_flat_expand_prologue): Likewise.
17441 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
17442 (sparc_use_pseudo_pic_reg): New function.
17443 (sparc_init_pic_reg): Likewise.
17444 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
17445 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
17447 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
17449 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
17450 Add item for branch_cost.
17452 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17454 PR rtl-optimization/83565
17455 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
17456 not extend the result to a larger mode for rotate operations.
17457 (num_sign_bit_copies1): Likewise.
17459 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17462 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
17464 Use values-Xc.o for -pedantic.
17465 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
17467 2018-01-12 Martin Liska <mliska@suse.cz>
17470 * ipa-devirt.c (final_warning_record::grow_type_warnings):
17472 (possible_polymorphic_call_targets): Use it.
17473 (ipa_devirt): Likewise.
17475 2018-01-12 Martin Liska <mliska@suse.cz>
17477 * profile-count.h (enum profile_quality): Use 0 as invalid
17478 enum value of profile_quality.
17480 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
17482 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
17483 -mext-string options.
17485 2018-01-12 Richard Biener <rguenther@suse.de>
17487 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
17488 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
17489 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
17491 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
17493 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
17495 * configure.ac (--with-long-double-format): Add support for the
17496 configuration option to change the default long double format on
17498 * config.gcc (powerpc*-linux*-*): Likewise.
17499 * configure: Regenerate.
17500 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
17501 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
17502 used without modification.
17504 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17506 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
17507 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
17508 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
17509 MISC_BUILTIN_SPEC_BARRIER.
17510 (rs6000_init_builtins): Likewise.
17511 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
17513 (speculation_barrier): New define_insn.
17514 * doc/extend.texi: Document __builtin_speculation_barrier.
17516 2018-01-11 Jakub Jelinek <jakub@redhat.com>
17519 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
17520 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
17521 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
17523 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
17524 integral modes instead of "ss" and "sd".
17525 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
17526 vectors with 32-bit and 64-bit elements.
17527 (vecdupssescalarmodesuffix): New mode attribute.
17528 (vec_dup<mode>): Use it.
17530 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
17533 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
17534 frame if argument is passed on stack.
17536 2018-01-11 Jakub Jelinek <jakub@redhat.com>
17539 * ree.c (combine_reaching_defs): Optimize also
17540 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
17541 reg2=any_extend(exp); reg1=reg2;, formatting fix.
17543 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
17545 PR middle-end/83189
17546 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
17548 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
17550 PR middle-end/83718
17551 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
17552 after they are computed.
17554 2018-01-11 Bin Cheng <bin.cheng@arm.com>
17556 PR tree-optimization/83695
17557 * gimple-loop-linterchange.cc
17558 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
17559 reset cached scev information after interchange.
17560 (pass_linterchange::execute): Remove call to scev_reset_htab.
17562 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17564 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
17565 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
17566 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
17567 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
17568 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
17569 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
17570 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
17571 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
17572 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
17573 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
17574 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
17575 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
17576 (V_lane_reg): Likewise.
17577 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
17579 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
17580 (vfmal_lane_low<mode>_intrinsic,
17581 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
17582 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
17583 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
17584 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
17585 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
17586 vfmsl_lane_high<mode>_intrinsic): New define_insns.
17588 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17590 * config/arm/arm-cpus.in (fp16fml): New feature.
17591 (ALL_SIMD): Add fp16fml.
17592 (armv8.2-a): Add fp16fml as an option.
17593 (armv8.3-a): Likewise.
17594 (armv8.4-a): Add fp16fml as part of fp16.
17595 * config/arm/arm.h (TARGET_FP16FML): Define.
17596 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
17598 * config/arm/arm-modes.def (V2HF): Define.
17599 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
17600 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
17601 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
17602 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
17603 vfmsl_low, vfmsl_high): New set of builtins.
17604 * config/arm/iterators.md (PLUSMINUS): New code iterator.
17605 (vfml_op): New code attribute.
17606 (VFMLHALVES): New int iterator.
17607 (VFML, VFMLSEL): New mode attributes.
17608 (V_reg): Define mapping for V2HF.
17609 (V_hi, V_lo): New mode attributes.
17610 (VF_constraint): Likewise.
17611 (vfml_half, vfml_half_selector): New int attributes.
17612 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
17614 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
17615 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
17617 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
17618 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
17619 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
17620 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
17622 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
17623 Document new effective target and option set.
17625 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17627 * config/arm/arm-cpus.in (armv8_4): New feature.
17628 (ARMv8_4a): New fgroup.
17629 (armv8.4-a): New arch.
17630 * config/arm/arm-tables.opt: Regenerate.
17631 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
17632 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
17633 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
17634 Add matching rules for -march=armv8.4-a and extensions.
17635 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
17637 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
17640 * config/rx/rx.md (BW): New mode attribute.
17641 (sync_lock_test_and_setsi): Add mode suffix to insn output.
17643 2018-01-11 Richard Biener <rguenther@suse.de>
17645 PR tree-optimization/83435
17646 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
17647 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
17648 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
17650 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17651 Alan Hayward <alan.hayward@arm.com>
17652 David Sherwood <david.sherwood@arm.com>
17654 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
17656 (aarch64_classify_address): Initialize it. Track polynomial offsets.
17657 (aarch64_print_address_internal): Use it to check for a zero offset.
17659 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17660 Alan Hayward <alan.hayward@arm.com>
17661 David Sherwood <david.sherwood@arm.com>
17663 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
17664 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
17665 Return a poly_int64 rather than a HOST_WIDE_INT.
17666 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
17667 rather than a HOST_WIDE_INT.
17668 * config/aarch64/aarch64.h (aarch64_frame): Protect with
17669 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
17670 hard_fp_offset, frame_size, initial_adjust, callee_offset and
17671 final_offset from HOST_WIDE_INT to poly_int64.
17672 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
17673 to_constant when getting the number of units in an Advanced SIMD
17675 (aarch64_builtin_vectorized_function): Check for a constant number
17677 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
17679 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
17680 attribute instead of GET_MODE_NUNITS.
17681 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
17682 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
17683 GET_MODE_SIZE for fixed-size registers.
17684 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
17685 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
17686 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
17687 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
17688 (aarch64_print_operand, aarch64_print_address_internal)
17689 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
17690 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
17691 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
17692 Handle polynomial GET_MODE_SIZE.
17693 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
17694 wider than SImode without modification.
17695 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
17696 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
17697 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
17698 passing and returning SVE modes.
17699 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
17700 rather than GEN_INT.
17701 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
17702 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
17703 (aarch64_allocate_and_probe_stack_space): Likewise.
17704 (aarch64_layout_frame): Cope with polynomial offsets.
17705 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
17706 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
17707 polynomial offsets.
17708 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
17709 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
17710 poly_int64 rather than a HOST_WIDE_INT.
17711 (aarch64_get_separate_components, aarch64_process_components)
17712 (aarch64_expand_prologue, aarch64_expand_epilogue)
17713 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
17714 (aarch64_anchor_offset): New function, split out from...
17715 (aarch64_legitimize_address): ...here.
17716 (aarch64_builtin_vectorization_cost): Handle polynomial
17717 TYPE_VECTOR_SUBPARTS.
17718 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
17720 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
17721 number of elements from the PARALLEL rather than the mode.
17722 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
17723 rather than GET_MODE_BITSIZE.
17724 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
17725 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
17726 (aarch64_expand_vec_perm_const_1): Handle polynomial
17727 d->perm.length () and d->perm elements.
17728 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
17729 Apply to_constant to d->perm elements.
17730 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
17731 polynomial CONST_VECTOR_NUNITS.
17732 (aarch64_move_pointer): Take amount as a poly_int64 rather
17734 (aarch64_progress_pointer): Avoid temporary variable.
17735 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
17736 the mode attribute instead of GET_MODE.
17738 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17739 Alan Hayward <alan.hayward@arm.com>
17740 David Sherwood <david.sherwood@arm.com>
17742 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
17743 x exists before using it.
17744 (aarch64_add_constant_internal): Rename to...
17745 (aarch64_add_offset_1): ...this. Replace regnum with separate
17746 src and dest rtxes. Handle the case in which they're different,
17747 including when the offset is zero. Replace scratchreg with an rtx.
17748 Use 2 additions if there is no spare register into which we can
17749 move a 16-bit constant.
17750 (aarch64_add_constant): Delete.
17751 (aarch64_add_offset): Replace reg with separate src and dest
17752 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
17753 Use aarch64_add_offset_1.
17754 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
17755 an rtx rather than an int. Take the delta as a poly_int64
17756 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
17757 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
17758 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
17759 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
17760 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
17761 and aarch64_add_sp.
17762 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
17763 aarch64_add_constant.
17765 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17767 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
17768 Use scalar_float_mode.
17770 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17772 * config/aarch64/aarch64-simd.md
17773 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
17774 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
17775 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
17776 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
17777 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
17778 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
17779 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
17780 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
17781 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
17782 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
17784 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17787 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
17788 targ_options->x_arm_arch_string is non NULL.
17790 2018-01-11 Tamar Christina <tamar.christina@arm.com>
17792 * config/aarch64/aarch64.h
17793 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
17795 2018-01-11 Sudakshina Das <sudi.das@arm.com>
17798 * expmed.c (emit_store_flag_force): Swap if const op0
17799 and change VOIDmode to mode of op0.
17801 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17803 PR rtl-optimization/83761
17804 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
17805 than bytes to mode_for_size.
17807 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17809 PR middle-end/83189
17810 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
17811 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
17814 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17816 PR middle-end/83575
17817 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
17818 when in layout mode.
17819 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
17820 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
17823 2018-01-10 Michael Collison <michael.collison@arm.com>
17825 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
17826 * config/aarch64/aarch64-option-extension.def: Add
17827 AARCH64_OPT_EXTENSION of 'fp16fml'.
17828 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17829 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
17830 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
17831 * config/aarch64/constraints.md (Ui7): New constraint.
17832 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
17833 (VFMLA_SEL_W): Ditto.
17836 (VFMLA16_LOW): New int iterator.
17837 (VFMLA16_HIGH): Ditto.
17838 (UNSPEC_FMLAL): New unspec.
17839 (UNSPEC_FMLSL): Ditto.
17840 (UNSPEC_FMLAL2): Ditto.
17841 (UNSPEC_FMLSL2): Ditto.
17842 (f16mac): New code attribute.
17843 * config/aarch64/aarch64-simd-builtins.def
17844 (aarch64_fmlal_lowv2sf): Ditto.
17845 (aarch64_fmlsl_lowv2sf): Ditto.
17846 (aarch64_fmlalq_lowv4sf): Ditto.
17847 (aarch64_fmlslq_lowv4sf): Ditto.
17848 (aarch64_fmlal_highv2sf): Ditto.
17849 (aarch64_fmlsl_highv2sf): Ditto.
17850 (aarch64_fmlalq_highv4sf): Ditto.
17851 (aarch64_fmlslq_highv4sf): Ditto.
17852 (aarch64_fmlal_lane_lowv2sf): Ditto.
17853 (aarch64_fmlsl_lane_lowv2sf): Ditto.
17854 (aarch64_fmlal_laneq_lowv2sf): Ditto.
17855 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
17856 (aarch64_fmlalq_lane_lowv4sf): Ditto.
17857 (aarch64_fmlsl_lane_lowv4sf): Ditto.
17858 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
17859 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
17860 (aarch64_fmlal_lane_highv2sf): Ditto.
17861 (aarch64_fmlsl_lane_highv2sf): Ditto.
17862 (aarch64_fmlal_laneq_highv2sf): Ditto.
17863 (aarch64_fmlsl_laneq_highv2sf): Ditto.
17864 (aarch64_fmlalq_lane_highv4sf): Ditto.
17865 (aarch64_fmlsl_lane_highv4sf): Ditto.
17866 (aarch64_fmlalq_laneq_highv4sf): Ditto.
17867 (aarch64_fmlsl_laneq_highv4sf): Ditto.
17868 * config/aarch64/aarch64-simd.md:
17869 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
17870 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17871 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
17872 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17873 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
17874 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
17875 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
17876 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
17877 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
17878 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
17879 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
17880 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
17881 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
17882 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
17883 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
17884 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
17885 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
17886 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
17887 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
17888 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
17889 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
17890 (vfmlsl_low_u32): Ditto.
17891 (vfmlalq_low_u32): Ditto.
17892 (vfmlslq_low_u32): Ditto.
17893 (vfmlal_high_u32): Ditto.
17894 (vfmlsl_high_u32): Ditto.
17895 (vfmlalq_high_u32): Ditto.
17896 (vfmlslq_high_u32): Ditto.
17897 (vfmlal_lane_low_u32): Ditto.
17898 (vfmlsl_lane_low_u32): Ditto.
17899 (vfmlal_laneq_low_u32): Ditto.
17900 (vfmlsl_laneq_low_u32): Ditto.
17901 (vfmlalq_lane_low_u32): Ditto.
17902 (vfmlslq_lane_low_u32): Ditto.
17903 (vfmlalq_laneq_low_u32): Ditto.
17904 (vfmlslq_laneq_low_u32): Ditto.
17905 (vfmlal_lane_high_u32): Ditto.
17906 (vfmlsl_lane_high_u32): Ditto.
17907 (vfmlal_laneq_high_u32): Ditto.
17908 (vfmlsl_laneq_high_u32): Ditto.
17909 (vfmlalq_lane_high_u32): Ditto.
17910 (vfmlslq_lane_high_u32): Ditto.
17911 (vfmlalq_laneq_high_u32): Ditto.
17912 (vfmlslq_laneq_high_u32): Ditto.
17913 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
17914 (AARCH64_FL_FOR_ARCH8_4): New.
17915 (AARCH64_ISA_F16FML): New ISA flag.
17916 (TARGET_F16FML): New feature flag for fp16fml.
17917 (doc/invoke.texi): Document new fp16fml option.
17919 2018-01-10 Michael Collison <michael.collison@arm.com>
17921 * config/aarch64/aarch64-builtins.c:
17922 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
17923 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17924 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
17925 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
17926 (AARCH64_ISA_SHA3): New ISA flag.
17927 (TARGET_SHA3): New feature flag for sha3.
17928 * config/aarch64/iterators.md (sha512_op): New int attribute.
17929 (CRYPTO_SHA512): New int iterator.
17930 (UNSPEC_SHA512H): New unspec.
17931 (UNSPEC_SHA512H2): Ditto.
17932 (UNSPEC_SHA512SU0): Ditto.
17933 (UNSPEC_SHA512SU1): Ditto.
17934 * config/aarch64/aarch64-simd-builtins.def
17935 (aarch64_crypto_sha512hqv2di): New builtin.
17936 (aarch64_crypto_sha512h2qv2di): Ditto.
17937 (aarch64_crypto_sha512su0qv2di): Ditto.
17938 (aarch64_crypto_sha512su1qv2di): Ditto.
17939 (aarch64_eor3qv8hi): Ditto.
17940 (aarch64_rax1qv2di): Ditto.
17941 (aarch64_xarqv2di): Ditto.
17942 (aarch64_bcaxqv8hi): Ditto.
17943 * config/aarch64/aarch64-simd.md:
17944 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
17945 (aarch64_crypto_sha512su0qv2di): Ditto.
17946 (aarch64_crypto_sha512su1qv2di): Ditto.
17947 (aarch64_eor3qv8hi): Ditto.
17948 (aarch64_rax1qv2di): Ditto.
17949 (aarch64_xarqv2di): Ditto.
17950 (aarch64_bcaxqv8hi): Ditto.
17951 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
17952 (vsha512h2q_u64): Ditto.
17953 (vsha512su0q_u64): Ditto.
17954 (vsha512su1q_u64): Ditto.
17955 (veor3q_u16): Ditto.
17956 (vrax1q_u64): Ditto.
17957 (vxarq_u64): Ditto.
17958 (vbcaxq_u16): Ditto.
17959 * config/arm/types.md (crypto_sha512): New type attribute.
17960 (crypto_sha3): Ditto.
17961 (doc/invoke.texi): Document new sha3 option.
17963 2018-01-10 Michael Collison <michael.collison@arm.com>
17965 * config/aarch64/aarch64-builtins.c:
17966 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
17967 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17968 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
17969 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
17970 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
17971 (AARCH64_ISA_SM4): New ISA flag.
17972 (TARGET_SM4): New feature flag for sm4.
17973 * config/aarch64/aarch64-simd-builtins.def
17974 (aarch64_sm3ss1qv4si): Ditto.
17975 (aarch64_sm3tt1aq4si): Ditto.
17976 (aarch64_sm3tt1bq4si): Ditto.
17977 (aarch64_sm3tt2aq4si): Ditto.
17978 (aarch64_sm3tt2bq4si): Ditto.
17979 (aarch64_sm3partw1qv4si): Ditto.
17980 (aarch64_sm3partw2qv4si): Ditto.
17981 (aarch64_sm4eqv4si): Ditto.
17982 (aarch64_sm4ekeyqv4si): Ditto.
17983 * config/aarch64/aarch64-simd.md:
17984 (aarch64_sm3ss1qv4si): Ditto.
17985 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
17986 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
17987 (aarch64_sm4eqv4si): Ditto.
17988 (aarch64_sm4ekeyqv4si): Ditto.
17989 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
17990 (sm3part_op): Ditto.
17991 (CRYPTO_SM3TT): Ditto.
17992 (CRYPTO_SM3PART): Ditto.
17993 (UNSPEC_SM3SS1): New unspec.
17994 (UNSPEC_SM3TT1A): Ditto.
17995 (UNSPEC_SM3TT1B): Ditto.
17996 (UNSPEC_SM3TT2A): Ditto.
17997 (UNSPEC_SM3TT2B): Ditto.
17998 (UNSPEC_SM3PARTW1): Ditto.
17999 (UNSPEC_SM3PARTW2): Ditto.
18000 (UNSPEC_SM4E): Ditto.
18001 (UNSPEC_SM4EKEY): Ditto.
18002 * config/aarch64/constraints.md (Ui2): New constraint.
18003 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
18004 * config/arm/types.md (crypto_sm3): New type attribute.
18005 (crypto_sm4): Ditto.
18006 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
18007 (vsm3tt1aq_u32): Ditto.
18008 (vsm3tt1bq_u32): Ditto.
18009 (vsm3tt2aq_u32): Ditto.
18010 (vsm3tt2bq_u32): Ditto.
18011 (vsm3partw1q_u32): Ditto.
18012 (vsm3partw2q_u32): Ditto.
18013 (vsm4eq_u32): Ditto.
18014 (vsm4ekeyq_u32): Ditto.
18015 (doc/invoke.texi): Document new sm4 option.
18017 2018-01-10 Michael Collison <michael.collison@arm.com>
18019 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
18020 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
18021 (AARCH64_FL_FOR_ARCH8_4): New.
18022 (AARCH64_FL_V8_4): New flag.
18023 (doc/invoke.texi): Document new armv8.4-a option.
18025 2018-01-10 Michael Collison <michael.collison@arm.com>
18027 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18028 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
18029 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
18030 * config/aarch64/aarch64-option-extension.def: Add
18031 AARCH64_OPT_EXTENSION of 'sha2'.
18032 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
18033 (crypto): Disable sha2 and aes if crypto disabled.
18034 (crypto): Enable aes and sha2 if enabled.
18035 (simd): Disable sha2 and aes if simd disabled.
18036 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
18038 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
18039 (TARGET_SHA2): New feature flag for sha2.
18040 (TARGET_AES): New feature flag for aes.
18041 * config/aarch64/aarch64-simd.md:
18042 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
18043 conditional on TARGET_AES.
18044 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
18045 (aarch64_crypto_sha1hsi): Make pattern conditional
18047 (aarch64_crypto_sha1hv4si): Ditto.
18048 (aarch64_be_crypto_sha1hv4si): Ditto.
18049 (aarch64_crypto_sha1su1v4si): Ditto.
18050 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
18051 (aarch64_crypto_sha1su0v4si): Ditto.
18052 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
18053 (aarch64_crypto_sha256su0v4si): Ditto.
18054 (aarch64_crypto_sha256su1v4si): Ditto.
18055 (doc/invoke.texi): Document new aes and sha2 options.
18057 2018-01-10 Martin Sebor <msebor@redhat.com>
18059 PR tree-optimization/83781
18060 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
18063 2018-01-11 Martin Sebor <msebor@gmail.com>
18064 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18066 PR tree-optimization/83501
18067 PR tree-optimization/81703
18069 * tree-ssa-strlen.c (get_string_cst): Rename...
18070 (get_string_len): ...to this. Handle global constants.
18071 (handle_char_store): Adjust.
18073 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
18074 Jim Wilson <jimw@sifive.com>
18076 * config/riscv/riscv-protos.h (riscv_output_return): New.
18077 * config/riscv/riscv.c (struct machine_function): New naked_p field.
18078 (riscv_attribute_table, riscv_output_return),
18079 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
18080 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
18081 (riscv_compute_frame_info): Only compute frame->mask if not a naked
18083 (riscv_expand_prologue): Add early return for naked function.
18084 (riscv_expand_epilogue): Likewise.
18085 (riscv_function_ok_for_sibcall): Return false for naked function.
18086 (riscv_set_current_function): New.
18087 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
18088 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
18089 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
18090 * doc/extend.texi (RISC-V Function Attributes): New.
18092 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
18094 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
18095 check for 128-bit long double before checking TCmode.
18096 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
18097 128-bit long doubles before checking TFmode or TCmode.
18098 (FLOAT128_IBM_P): Likewise.
18100 2018-01-10 Martin Sebor <msebor@redhat.com>
18102 PR tree-optimization/83671
18103 * builtins.c (c_strlen): Unconditionally return zero for the empty
18105 Use -Warray-bounds for warnings.
18106 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
18107 for non-constant array indices with COMPONENT_REF, arrays of
18108 arrays, and pointers to arrays.
18109 (gimple_fold_builtin_strlen): Determine and set length range for
18110 non-constant character arrays.
18112 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
18114 PR middle-end/81897
18115 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
18118 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
18120 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
18122 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
18125 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
18126 VECTOR_MEM_ALTIVEC_OR_VSX_P.
18127 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
18128 indexed_or_indirect_operand predicate.
18129 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
18130 (*vsx_le_perm_load_v8hi): Likewise.
18131 (*vsx_le_perm_load_v16qi): Likewise.
18132 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
18133 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
18134 (*vsx_le_perm_store_v8hi): Likewise.
18135 (*vsx_le_perm_store_v16qi): Likewise.
18136 (eight unnamed splitters): Likewise.
18138 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
18140 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
18141 * config/rs6000/emmintrin.h: Likewise.
18142 * config/rs6000/mmintrin.h: Likewise.
18143 * config/rs6000/xmmintrin.h: Likewise.
18145 2018-01-10 David Malcolm <dmalcolm@redhat.com>
18148 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
18150 * tree.c (tree_nop_conversion): Return true for location wrapper
18152 (maybe_wrap_with_location): New function.
18153 (selftest::check_strip_nops): New function.
18154 (selftest::test_location_wrappers): New function.
18155 (selftest::tree_c_tests): Call it.
18156 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
18157 (maybe_wrap_with_location): New decl.
18158 (EXPR_LOCATION_WRAPPER_P): New macro.
18159 (location_wrapper_p): New inline function.
18160 (tree_strip_any_location_wrapper): New inline function.
18162 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
18165 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
18166 stack_realign_offset for the largest alignment of stack slot
18168 (ix86_find_max_used_stack_alignment): New function.
18169 (ix86_finalize_stack_frame_flags): Use it. Set
18170 max_used_stack_alignment if we don't realign stack.
18171 * config/i386/i386.h (machine_function): Add
18172 max_used_stack_alignment.
18174 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
18176 * config/arm/arm.opt (-mbranch-cost): New option.
18177 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
18180 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
18183 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
18184 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
18186 2018-01-10 Richard Biener <rguenther@suse.de>
18189 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
18190 early out so it also covers the case where we have a non-NULL
18193 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
18195 PR tree-optimization/83753
18196 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
18197 for non-strided grouped accesses if the number of elements is 1.
18199 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
18202 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
18203 * i386.h (TARGET_USE_GATHER): Define.
18204 * x86-tune.def (X86_TUNE_USE_GATHER): New.
18206 2018-01-10 Martin Liska <mliska@suse.cz>
18209 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
18210 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
18212 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
18213 CLEANUP_NO_PARTITIONING is not set.
18215 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
18217 * doc/rtl.texi: Remove documentation of (const ...) wrappers
18218 for vectors, as a partial revert of r254296.
18219 * rtl.h (const_vec_p): Delete.
18220 (const_vec_duplicate_p): Don't test for vector CONSTs.
18221 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
18222 * expmed.c (make_tree): Likewise.
18225 * common.md (E, F): Use CONSTANT_P instead of checking for
18227 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
18228 checking for CONST_VECTOR.
18230 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18232 PR middle-end/83575
18233 * predict.c (force_edge_cold): Handle in more sane way edges
18234 with no prediction.
18236 2018-01-09 Carl Love <cel@us.ibm.com>
18238 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
18240 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
18241 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
18242 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
18243 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
18244 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
18245 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
18246 * config/rs6000/rs6000-protos.h: Add extern defition for
18247 rs6000_generate_float2_double_code.
18248 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
18250 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
18251 (float2_v2df): Add define_expand.
18253 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
18256 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
18257 op_mode in the force_to_mode call.
18259 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
18261 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
18262 instead of checking each element individually.
18263 (aarch64_evpc_uzp): Likewise.
18264 (aarch64_evpc_zip): Likewise.
18265 (aarch64_evpc_ext): Likewise.
18266 (aarch64_evpc_rev): Likewise.
18267 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
18268 instead of checking each element individually. Return true without
18270 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
18271 whether all selected elements come from the same input, instead of
18272 checking each element individually. Remove calls to gen_rtx_REG,
18273 start_sequence and end_sequence and instead assert that no rtl is
18276 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
18278 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
18279 order of HIGH and CONST checks.
18281 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
18283 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
18284 if the destination isn't an SSA_NAME.
18286 2018-01-09 Richard Biener <rguenther@suse.de>
18288 PR tree-optimization/83668
18289 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
18291 (canonicalize_loop_form): ... here, renamed from ...
18292 (canonicalize_loop_closed_ssa_form): ... this and amended to
18293 swap successor edges for loop exit blocks to make us use
18294 the RPO order we need for initial schedule generation.
18296 2018-01-09 Joseph Myers <joseph@codesourcery.com>
18298 PR tree-optimization/64811
18299 * match.pd: When optimizing comparisons with Inf, avoid
18300 introducing or losing exceptions from comparisons with NaN.
18302 2018-01-09 Martin Liska <mliska@suse.cz>
18305 * asan.c (shadow_mem_size): Add gcc_assert.
18307 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
18309 Don't save registers in main().
18312 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
18313 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
18314 * config/avr/avr.c (avr_set_current_function): Don't error if
18315 naked, OS_task or OS_main are specified at the same time.
18316 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
18318 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
18320 * common/config/avr/avr-common.c (avr_option_optimization_table):
18321 Switch on -mmain-is-OS_task for optimizing compilations.
18323 2018-01-09 Richard Biener <rguenther@suse.de>
18325 PR tree-optimization/83572
18326 * graphite.c: Include cfganal.h.
18327 (graphite_transform_loops): Connect infinite loops to exit
18328 and remove fake edges at the end.
18330 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18332 * ipa-inline.c (edge_badness): Revert accidental checkin.
18334 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18337 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
18338 symbols; not inline clones.
18340 2018-01-09 Jakub Jelinek <jakub@redhat.com>
18343 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
18344 hard registers. Formatting fixes.
18346 PR preprocessor/83722
18347 * gcc.c (try_generate_repro): Pass
18348 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
18349 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
18352 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
18353 Kito Cheng <kito.cheng@gmail.com>
18355 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
18356 (riscv_leaf_function_p): Delete.
18357 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
18359 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18361 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
18363 (do_ifelse): New function.
18364 (do_isel): New function.
18365 (do_sub3): New function.
18366 (do_add3): New function.
18367 (do_load_mask_compare): New function.
18368 (do_overlap_load_compare): New function.
18369 (expand_compare_loop): New function.
18370 (expand_block_compare): Call expand_compare_loop() when appropriate.
18371 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
18372 option description.
18373 (-mblock-compare-inline-loop-limit): New option.
18375 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18378 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
18379 Reverse order of second and third operands in first alternative.
18380 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
18381 of first and second elements in UNSPEC_VPERMR vector.
18382 (altivec_expand_vec_perm_le): Likewise.
18384 2018-01-08 Jeff Law <law@redhat.com>
18386 PR rtl-optimizatin/81308
18387 * tree-switch-conversion.c (cfg_altered): New file scoped static.
18388 (process_switch): If group_case_labels makes a change, then set
18390 (pass_convert_switch::execute): If a switch is converted, then
18391 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
18393 PR rtl-optimization/81308
18394 * recog.c (split_all_insns): Conditionally cleanup the CFG after
18397 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
18399 PR target/83663 - Revert r255946
18400 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
18401 generation for cases where splatting a value is not useful.
18402 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
18403 across a vec_duplicate and a paradoxical subreg forming a vector
18404 mode to a vec_concat.
18406 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18408 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
18409 -march=armv8.3-a variants.
18410 * config/arm/t-multilib: Likewise.
18411 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
18413 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18415 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
18417 (cceq_ior_compare_complement): Give it a name so I can use it, and
18418 change boolean_or_operator predicate to boolean_operator so it can
18419 be used to generate a crand.
18420 (eqne): New code iterator.
18421 (bd/bd_neg): New code_attrs.
18422 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
18423 a single define_insn.
18424 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
18425 decrement (bdnzt/bdnzf/bdzt/bdzf).
18426 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
18427 with the new names of the branch decrement patterns, and added the
18428 names of the branch decrement conditional patterns.
18430 2018-01-08 Richard Biener <rguenther@suse.de>
18432 PR tree-optimization/83563
18433 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
18436 2018-01-08 Richard Biener <rguenther@suse.de>
18438 PR middle-end/83713
18439 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
18441 2018-01-08 Richard Biener <rguenther@suse.de>
18443 PR tree-optimization/83685
18444 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
18445 references to abnormals.
18447 2018-01-08 Richard Biener <rguenther@suse.de>
18450 * dwarf2out.c (output_indirect_strings): Handle empty
18451 skeleton_debug_str_hash.
18452 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
18454 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
18456 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
18457 (emit_store_direct): Likewise.
18458 (arc_trampoline_adjust_address): Likewise.
18459 (arc_asm_trampoline_template): New function.
18460 (arc_initialize_trampoline): Use asm_trampoline_template.
18461 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
18462 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
18463 * config/arc/arc.md (flush_icache): Delete pattern.
18465 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
18467 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
18468 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
18471 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
18474 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
18475 by not USED_FOR_TARGET.
18476 (make_pass_resolve_sw_modes): Likewise.
18478 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
18480 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
18483 2018-01-08 Richard Biener <rguenther@suse.de>
18485 PR middle-end/83580
18486 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
18488 2018-01-08 Richard Biener <rguenther@suse.de>
18490 PR middle-end/83517
18491 * match.pd ((t * 2) / 2) -> t): Add missing :c.
18493 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
18495 PR middle-end/81897
18496 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
18497 basic blocks with a small number of successors.
18498 (convert_control_dep_chain_into_preds): Improve handling of
18500 (dump_predicates): Split apart into...
18501 (dump_pred_chain): ...here...
18502 (dump_pred_info): ...and here.
18503 (can_one_predicate_be_invalidated_p): Add debugging printfs.
18504 (can_chain_union_be_invalidated_p): Improve check for invalidation
18506 (uninit_uses_cannot_happen): Avoid unnecessary if
18507 convert_control_dep_chain_into_preds yielded nothing.
18509 2018-01-06 Martin Sebor <msebor@redhat.com>
18511 PR tree-optimization/83640
18512 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
18513 subtracting negative offset from size.
18514 (builtin_access::overlap): Adjust offset bounds of the access to fall
18515 within the size of the object if possible.
18517 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
18519 PR rtl-optimization/83699
18520 * expmed.c (extract_bit_field_1): Restrict the vector usage of
18521 extract_bit_field_as_subreg to cases in which the extracted
18522 value is also a vector.
18524 * lra-constraints.c (process_alt_operands): Test for the equivalence
18525 substitutions when detecting a possible reload cycle.
18527 2018-01-06 Jakub Jelinek <jakub@redhat.com>
18530 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
18531 by default if flag_selective_schedling{,2}. Formatting fixes.
18533 PR rtl-optimization/83682
18534 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
18535 if it has non-VECTOR_MODE element mode.
18536 (vec_duplicate_p): Likewise.
18538 PR middle-end/83694
18539 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
18540 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
18542 2018-01-05 Jakub Jelinek <jakub@redhat.com>
18545 * config/i386/i386-builtin.def
18546 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
18547 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
18548 Require also OPTION_MASK_ISA_AVX512F in addition to
18549 OPTION_MASK_ISA_GFNI.
18550 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
18551 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
18552 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
18553 to OPTION_MASK_ISA_GFNI.
18554 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
18555 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
18556 OPTION_MASK_ISA_AVX512BW.
18557 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
18558 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
18559 addition to OPTION_MASK_ISA_GFNI.
18560 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
18561 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
18562 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
18563 to OPTION_MASK_ISA_GFNI.
18564 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
18565 a requirement for all ISAs rather than any of them with a few
18567 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
18569 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
18570 bitmasks to be enabled with 3 exceptions, instead of requiring any
18571 enabled ISA with lots of exceptions.
18572 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
18573 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
18574 Change avx512bw in isa attribute to avx512f.
18575 * config/i386/sgxintrin.h: Add license boilerplate.
18576 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
18577 to __AVX512F__ and __AVX512VL to __AVX512VL__.
18578 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
18579 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
18581 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
18582 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
18583 temporarily sse2 rather than sse if not enabled already.
18586 * config/i386/sse.md (VI248_VLBW): Rename to ...
18587 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
18588 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
18589 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
18590 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
18591 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
18592 mode iterator instead of VI248_VLBW.
18594 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
18596 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
18597 (record_modified): Skip clobbers; add debug output.
18598 (param_change_prob): Use sreal frequencies.
18600 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
18602 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
18603 punt for user-aligned variables.
18605 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
18607 * tree-chrec.c (chrec_contains_symbols): Return true for
18610 2018-01-05 Sudakshina Das <sudi.das@arm.com>
18613 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
18614 of (x|y) == x for BICS pattern.
18616 2018-01-05 Jakub Jelinek <jakub@redhat.com>
18618 PR tree-optimization/83605
18619 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
18620 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
18623 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
18625 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
18626 * config/epiphany/rtems.h: New file.
18628 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18629 Uros Bizjak <ubizjak@gmail.com>
18632 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
18633 QIreg_operand instead of register_operand predicate.
18634 * config/i386/i386.c (ix86_rop_should_change_byte_p,
18635 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
18636 comments instead of -fmitigate[-_]rop.
18638 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18641 * cgraphunit.c (symbol_table::compile): Switch to text_section
18642 before calling assembly_start debug hook.
18643 * run-rtl-passes.c (run_rtl_passes): Likewise.
18646 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18648 * tree-vrp.c (extract_range_from_binary_expr_1): Check
18649 range_int_cst_p rather than !symbolic_range_p before calling
18650 extract_range_from_multiplicative_op_1.
18652 2018-01-04 Jeff Law <law@redhat.com>
18654 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
18655 redundant test in assertion.
18657 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18659 * doc/rtl.texi: Document machine_mode wrapper classes.
18661 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18663 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
18664 using tree_to_uhwi.
18666 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18668 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
18669 the VEC_PERM_EXPR fold to fail.
18671 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18674 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
18675 to switched_sections.
18677 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18680 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
18681 test for d.testing.
18683 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
18686 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
18687 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
18689 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18692 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
18693 is BLKmode and bitpos not zero or mode change is needed.
18695 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18698 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
18701 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
18704 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
18705 instead of MULT rtx. Update all corresponding splitters.
18706 (*saddl_se): Ditto.
18707 (*ssub<modesuffix>): Ditto.
18708 (*ssubl_se): Ditto.
18709 (*cmp_sadd_di): Update split patterns.
18710 (*cmp_sadd_si): Ditto.
18711 (*cmp_sadd_sidi): Ditto.
18712 (*cmp_ssub_di): Ditto.
18713 (*cmp_ssub_si): Ditto.
18714 (*cmp_ssub_sidi): Ditto.
18715 * config/alpha/predicates.md (const23_operand): New predicate.
18716 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
18717 Look for ASHIFT, not MULT inner operand.
18718 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
18720 2018-01-04 Martin Liska <mliska@suse.cz>
18722 PR gcov-profile/83669
18723 * gcov.c (output_intermediate_file): Add version to intermediate
18725 * doc/gcov.texi: Document new field 'version' in intermediate
18726 file format. Fix location of '-k' option of gcov command.
18728 2018-01-04 Martin Liska <mliska@suse.cz>
18731 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
18733 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18735 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
18737 2018-01-03 Martin Sebor <msebor@redhat.com>
18739 PR tree-optimization/83655
18740 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
18741 checking calls with invalid arguments.
18743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18745 * tree-vect-stmts.c (vect_get_store_rhs): New function.
18746 (vectorizable_mask_load_store): Delete.
18747 (vectorizable_call): Return false for masked loads and stores.
18748 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
18749 instead of gimple_assign_rhs1.
18750 (vectorizable_load): Handle IFN_MASK_LOAD.
18751 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
18753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18755 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
18757 (vectorizable_mask_load_store): ...here.
18758 (vectorizable_load): ...and here.
18760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18762 * tree-vect-stmts.c (vect_build_all_ones_mask)
18763 (vect_build_zero_merge_argument): New functions, split out from...
18764 (vectorizable_load): ...here.
18766 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18768 * tree-vect-stmts.c (vect_check_store_rhs): New function,
18770 (vectorizable_mask_load_store): ...here.
18771 (vectorizable_store): ...and here.
18773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18775 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
18777 (vectorizable_mask_load_store): ...here.
18779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18781 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
18782 (vect_model_store_cost): Take a vec_load_store_type instead of a
18784 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
18785 (vect_model_store_cost): Take a vec_load_store_type instead of a
18787 (vectorizable_mask_load_store): Update accordingly.
18788 (vectorizable_store): Likewise.
18789 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
18791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18793 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
18794 IFN_MASK_LOAD calls here rather than...
18795 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
18797 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18798 Alan Hayward <alan.hayward@arm.com>
18799 David Sherwood <david.sherwood@arm.com>
18801 * expmed.c (extract_bit_field_1): For vector extracts,
18802 fall back to extract_bit_field_as_subreg if vec_extract
18805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18806 Alan Hayward <alan.hayward@arm.com>
18807 David Sherwood <david.sherwood@arm.com>
18809 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
18810 they are variable or constant sized.
18811 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
18812 slots for constant-sized data.
18814 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18815 Alan Hayward <alan.hayward@arm.com>
18816 David Sherwood <david.sherwood@arm.com>
18818 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
18819 handling COND_EXPRs with boolean comparisons, try to find a better
18820 basis for the mask type than the boolean itself.
18822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18824 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
18825 is calculated and how it can be overridden.
18826 * genmodes.c (max_bitsize_mode_any_mode): New variable.
18827 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
18829 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
18832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18833 Alan Hayward <alan.hayward@arm.com>
18834 David Sherwood <david.sherwood@arm.com>
18836 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
18837 Remove the mode argument.
18838 (aarch64_simd_valid_immediate): Remove the mode and inverse
18840 * config/aarch64/iterators.md (bitsize): New iterator.
18841 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
18842 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
18843 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
18844 aarch64_simd_valid_immediate.
18845 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
18846 (aarch64_reg_or_bic_imm): Likewise.
18847 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
18848 with an insn_type enum and msl with a modifier_type enum.
18849 Replace element_width with a scalar_mode. Change the shift
18850 to unsigned int. Add constructors for scalar_float_mode and
18851 scalar_int_mode elements.
18852 (aarch64_vect_float_const_representable_p): Delete.
18853 (aarch64_can_const_movi_rtx_p)
18854 (aarch64_simd_scalar_immediate_valid_for_move)
18855 (aarch64_simd_make_constant): Update call to
18856 aarch64_simd_valid_immediate.
18857 (aarch64_advsimd_valid_immediate_hs): New function.
18858 (aarch64_advsimd_valid_immediate): Likewise.
18859 (aarch64_simd_valid_immediate): Remove mode and inverse
18860 arguments. Rewrite to use the above. Use const_vec_duplicate_p
18861 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
18862 and aarch64_float_const_representable_p on the result.
18863 (aarch64_output_simd_mov_immediate): Remove mode argument.
18864 Update call to aarch64_simd_valid_immediate and use of
18865 simd_immediate_info.
18866 (aarch64_output_scalar_simd_mov_immediate): Update call
18869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18870 Alan Hayward <alan.hayward@arm.com>
18871 David Sherwood <david.sherwood@arm.com>
18873 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
18874 (mode_nunits): Likewise CONST_MODE_NUNITS.
18875 * machmode.def (ADJUST_NUNITS): Document.
18876 * genmodes.c (mode_data::need_nunits_adj): New field.
18877 (blank_mode): Update accordingly.
18878 (adj_nunits): New variable.
18879 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
18881 (emit_mode_size_inline): Set need_bytesize_adj for all modes
18882 listed in adj_nunits.
18883 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
18884 listed in adj_nunits. Don't emit case statements for such modes.
18885 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
18886 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
18887 nothing if adj_nunits is nonnull.
18888 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
18889 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
18890 (emit_mode_fbit): Update use of print_maybe_const_decl.
18891 (emit_move_size): Likewise. Treat the array as non-const
18893 (emit_mode_adjustments): Handle adj_nunits.
18895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18897 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
18898 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
18899 (VECTOR_MODES): Use it.
18900 (make_vector_modes): Take the prefix as an argument.
18902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18903 Alan Hayward <alan.hayward@arm.com>
18904 David Sherwood <david.sherwood@arm.com>
18906 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
18907 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
18908 for MODE_VECTOR_BOOL.
18909 * machmode.def (VECTOR_BOOL_MODE): Document.
18910 * genmodes.c (VECTOR_BOOL_MODE): New macro.
18911 (make_vector_bool_mode): New function.
18912 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
18914 * lto-streamer-in.c (lto_input_mode_table): Likewise.
18915 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
18917 * stor-layout.c (int_mode_for_mode): Likewise.
18918 * tree.c (build_vector_type_for_mode): Likewise.
18919 * varasm.c (output_constant_pool_2): Likewise.
18920 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
18921 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
18922 for MODE_VECTOR_BOOL.
18923 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
18924 of mode class checks.
18925 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
18926 instead of a list of mode class checks.
18927 (expand_vector_scalar_condition): Likewise.
18928 (type_for_widest_vector_mode): Handle BImode as an inner mode.
18930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18931 Alan Hayward <alan.hayward@arm.com>
18932 David Sherwood <david.sherwood@arm.com>
18934 * machmode.h (mode_size): Change from unsigned short to
18936 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
18937 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18938 or if measurement_type is not polynomial.
18939 (fixed_size_mode::includes_p): Check for constant-sized modes.
18940 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
18941 return a poly_uint16 rather than an unsigned short.
18942 (emit_mode_size): Change the type of mode_size from unsigned short
18943 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
18944 (emit_mode_adjustments): Cope with polynomial vector sizes.
18945 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18947 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18949 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
18950 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
18951 * caller-save.c (setup_save_areas): Likewise.
18952 (replace_reg_with_saved_mem): Likewise.
18953 * calls.c (emit_library_call_value_1): Likewise.
18954 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
18955 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
18956 (gen_lowpart_for_combine): Likewise.
18957 * convert.c (convert_to_integer_1): Likewise.
18958 * cse.c (equiv_constant, cse_insn): Likewise.
18959 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
18960 (cselib_subst_to_values): Likewise.
18961 * dce.c (word_dce_process_block): Likewise.
18962 * df-problems.c (df_word_lr_mark_ref): Likewise.
18963 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
18964 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
18965 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
18966 (rtl_for_decl_location): Likewise.
18967 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
18968 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
18969 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
18970 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
18971 (expand_expr_real_1): Likewise.
18972 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
18973 (pad_below): Likewise.
18974 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18975 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
18976 * ira.c (get_subreg_tracking_sizes): Likewise.
18977 * ira-build.c (ira_create_allocno_objects): Likewise.
18978 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
18979 (ira_sort_regnos_for_alter_reg): Likewise.
18980 * ira-costs.c (record_operand_costs): Likewise.
18981 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
18982 (resolve_simple_move): Likewise.
18983 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
18984 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
18985 (lra_constraints): Likewise.
18986 (CONST_POOL_OK_P): Reject variable-sized modes.
18987 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
18988 (add_pseudo_to_slot, lra_spill): Likewise.
18989 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18990 * optabs-query.c (get_best_extraction_insn): Likewise.
18991 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18992 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
18993 (expand_mult_highpart, valid_multiword_target_p): Likewise.
18994 * recog.c (offsettable_address_addr_space_p): Likewise.
18995 * regcprop.c (maybe_mode_change): Likewise.
18996 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
18997 * regrename.c (build_def_use): Likewise.
18998 * regstat.c (dump_reg_info): Likewise.
18999 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
19000 (find_reloads, find_reloads_subreg_address): Likewise.
19001 * reload1.c (eliminate_regs_1): Likewise.
19002 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
19003 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
19004 (simplify_binary_operation_1, simplify_subreg): Likewise.
19005 * targhooks.c (default_function_arg_padding): Likewise.
19006 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
19007 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
19008 (verify_gimple_assign_ternary): Likewise.
19009 * tree-inline.c (estimate_move_cost): Likewise.
19010 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19011 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
19012 (get_address_cost_ainc): Likewise.
19013 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
19014 (vect_supportable_dr_alignment): Likewise.
19015 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
19016 (vectorizable_reduction): Likewise.
19017 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
19018 (vectorizable_operation, vectorizable_load): Likewise.
19019 * tree.c (build_same_sized_truth_vector_type): Likewise.
19020 * valtrack.c (cleanup_auto_inc_dec): Likewise.
19021 * var-tracking.c (emit_note_insn_var_location): Likewise.
19022 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
19023 (ADDR_VEC_ALIGN): Likewise.
19025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19026 Alan Hayward <alan.hayward@arm.com>
19027 David Sherwood <david.sherwood@arm.com>
19029 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
19031 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
19032 or if measurement_type is polynomial.
19033 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
19034 * combine.c (make_extraction): Likewise.
19035 * dse.c (find_shift_sequence): Likewise.
19036 * dwarf2out.c (mem_loc_descriptor): Likewise.
19037 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
19038 (extract_bit_field, extract_low_bits): Likewise.
19039 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
19040 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
19041 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
19042 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
19043 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
19044 * reload.c (find_reloads): Likewise.
19045 * reload1.c (alter_reg): Likewise.
19046 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
19047 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
19048 * tree-if-conv.c (predicate_mem_writes): Likewise.
19049 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19050 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
19051 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
19052 * valtrack.c (dead_debug_insert_temp): Likewise.
19053 * varasm.c (mergeable_constant_section): Likewise.
19054 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
19056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19057 Alan Hayward <alan.hayward@arm.com>
19058 David Sherwood <david.sherwood@arm.com>
19060 * expr.c (expand_assignment): Cope with polynomial mode sizes
19061 when assigning to a CONCAT.
19063 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19064 Alan Hayward <alan.hayward@arm.com>
19065 David Sherwood <david.sherwood@arm.com>
19067 * machmode.h (mode_precision): Change from unsigned short to
19069 (mode_to_precision): Return a poly_uint16 rather than an unsigned
19071 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
19072 or if measurement_type is not polynomial.
19073 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
19074 in which the mode is already known to be a scalar_int_mode.
19075 * genmodes.c (emit_mode_precision): Change the type of mode_precision
19076 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
19078 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
19079 for GET_MODE_PRECISION.
19080 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
19081 for GET_MODE_PRECISION.
19082 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
19084 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
19085 (expand_field_assignment, make_extraction): Likewise.
19086 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
19087 (get_last_value): Likewise.
19088 * convert.c (convert_to_integer_1): Likewise.
19089 * cse.c (cse_insn): Likewise.
19090 * expr.c (expand_expr_real_1): Likewise.
19091 * lra-constraints.c (simplify_operand_subreg): Likewise.
19092 * optabs-query.c (can_atomic_load_p): Likewise.
19093 * optabs.c (expand_atomic_load): Likewise.
19094 (expand_atomic_store): Likewise.
19095 * ree.c (combine_reaching_defs): Likewise.
19096 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
19097 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
19098 * tree.h (type_has_mode_precision_p): Likewise.
19099 * ubsan.c (instrument_si_overflow): Likewise.
19101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19102 Alan Hayward <alan.hayward@arm.com>
19103 David Sherwood <david.sherwood@arm.com>
19105 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
19106 polynomial numbers of units.
19107 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
19108 (valid_vector_subparts_p): New function.
19109 (build_vector_type): Remove temporary shim and take the number
19110 of units as a poly_uint64 rather than an int.
19111 (build_opaque_vector_type): Take the number of units as a
19112 poly_uint64 rather than an int.
19113 * tree.c (build_vector_from_ctor): Handle polynomial
19114 TYPE_VECTOR_SUBPARTS.
19115 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
19116 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
19117 (build_vector_from_val): If the number of units is variable,
19118 use build_vec_duplicate_cst for constant operands and
19119 VEC_DUPLICATE_EXPR otherwise.
19120 (make_vector_type): Remove temporary is_constant ().
19121 (build_vector_type, build_opaque_vector_type): Take the number of
19122 units as a poly_uint64 rather than an int.
19123 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
19125 * cfgexpand.c (expand_debug_expr): Likewise.
19126 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
19127 (store_constructor, expand_expr_real_1): Likewise.
19128 (const_scalar_mask_from_tree): Likewise.
19129 * fold-const-call.c (fold_const_reduction): Likewise.
19130 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
19131 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
19132 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
19133 (fold_relational_const): Likewise.
19134 (native_interpret_vector): Likewise. Change the size from an
19135 int to an unsigned int.
19136 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
19137 TYPE_VECTOR_SUBPARTS.
19138 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
19139 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
19140 duplicating a non-constant operand into a variable-length vector.
19141 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
19142 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
19143 * ipa-icf.c (sem_variable::equals): Likewise.
19144 * match.pd: Likewise.
19145 * omp-simd-clone.c (simd_clone_subparts): Likewise.
19146 * print-tree.c (print_node): Likewise.
19147 * stor-layout.c (layout_type): Likewise.
19148 * targhooks.c (default_builtin_vectorization_cost): Likewise.
19149 * tree-cfg.c (verify_gimple_comparison): Likewise.
19150 (verify_gimple_assign_binary): Likewise.
19151 (verify_gimple_assign_ternary): Likewise.
19152 (verify_gimple_assign_single): Likewise.
19153 * tree-pretty-print.c (dump_generic_node): Likewise.
19154 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19155 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
19156 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
19157 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
19158 (vect_shift_permute_load_chain): Likewise.
19159 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
19160 (expand_vector_condition, optimize_vector_constructor): Likewise.
19161 (lower_vec_perm, get_compute_type): Likewise.
19162 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
19163 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
19164 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
19165 (vect_recog_mask_conversion_pattern): Likewise.
19166 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
19167 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
19168 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19169 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
19170 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
19171 (vectorizable_shift, vectorizable_operation, vectorizable_store)
19172 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
19173 (supportable_widening_operation): Likewise.
19174 (supportable_narrowing_operation): Likewise.
19175 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
19177 * varasm.c (output_constant): Likewise.
19179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19180 Alan Hayward <alan.hayward@arm.com>
19181 David Sherwood <david.sherwood@arm.com>
19183 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
19184 so that both the length == 3 and length != 3 cases set up their
19185 own permute vectors. Add comments explaining why we know the
19186 number of elements is constant.
19187 (vect_permute_load_chain): Likewise.
19189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19190 Alan Hayward <alan.hayward@arm.com>
19191 David Sherwood <david.sherwood@arm.com>
19193 * machmode.h (mode_nunits): Change from unsigned char to
19195 (ONLY_FIXED_SIZE_MODES): New macro.
19196 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
19197 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
19198 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
19200 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
19201 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
19202 or if measurement_type is not polynomial.
19203 * genmodes.c (ZERO_COEFFS): New macro.
19204 (emit_mode_nunits_inline): Make mode_nunits_inline return a
19206 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
19207 Use ZERO_COEFFS when emitting initializers.
19208 * data-streamer.h (bp_pack_poly_value): New function.
19209 (bp_unpack_poly_value): Likewise.
19210 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
19211 for GET_MODE_NUNITS.
19212 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
19213 for GET_MODE_NUNITS.
19214 * tree.c (make_vector_type): Remove temporary shim and make
19215 the real function take the number of units as a poly_uint64
19216 rather than an int.
19217 (build_vector_type_for_mode): Handle polynomial nunits.
19218 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
19219 * emit-rtl.c (const_vec_series_p_1): Likewise.
19220 (gen_rtx_CONST_VECTOR): Likewise.
19221 * fold-const.c (test_vec_duplicate_folding): Likewise.
19222 * genrecog.c (validate_pattern): Likewise.
19223 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
19224 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
19225 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
19226 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
19227 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
19228 * rtlanal.c (subreg_get_info): Likewise.
19229 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19230 (vect_grouped_load_supported): Likewise.
19231 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
19232 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
19233 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19234 (simplify_const_unary_operation, simplify_binary_operation_1)
19235 (simplify_const_binary_operation, simplify_ternary_operation)
19236 (test_vector_ops_duplicate, test_vector_ops): Likewise.
19237 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
19238 instead of CONST_VECTOR_NUNITS.
19239 * varasm.c (output_constant_pool_2): Likewise.
19240 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
19241 explicit-encoded elements in the XVEC for variable-length vectors.
19243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19245 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
19247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19248 Alan Hayward <alan.hayward@arm.com>
19249 David Sherwood <david.sherwood@arm.com>
19251 * coretypes.h (fixed_size_mode): Declare.
19252 (fixed_size_mode_pod): New typedef.
19253 * builtins.h (target_builtins::x_apply_args_mode)
19254 (target_builtins::x_apply_result_mode): Change type to
19255 fixed_size_mode_pod.
19256 * builtins.c (apply_args_size, apply_result_size, result_vector)
19257 (expand_builtin_apply_args_1, expand_builtin_apply)
19258 (expand_builtin_return): Update accordingly.
19260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19262 * cse.c (hash_rtx_cb): Hash only the encoded elements.
19263 * cselib.c (cselib_hash_rtx): Likewise.
19264 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
19265 CONST_VECTOR encoding.
19267 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19268 Jeff Law <law@redhat.com>
19271 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
19272 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
19273 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
19274 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
19277 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
19278 explicitly probe *sp in a noreturn function if there were any callee
19279 register saves or frame pointer is needed.
19281 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19284 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
19285 BLKmode for ternary, binary or unary expressions.
19288 * var-tracking.c (delete_vta_debug_insn): New inline function.
19289 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
19290 insns from get_insns () to NULL instead of each bb separately.
19291 Use delete_vta_debug_insn. No longer static.
19292 (vt_debug_insns_local, variable_tracking_main_1): Adjust
19293 delete_vta_debug_insns callers.
19294 * rtl.h (delete_vta_debug_insns): Declare.
19295 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
19296 instead of variable_tracking_main.
19298 2018-01-03 Martin Sebor <msebor@redhat.com>
19300 PR tree-optimization/83603
19301 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
19302 arguments past the endof the argument list in functions declared
19303 without a prototype.
19304 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
19305 Avoid checking when arguments are null.
19307 2018-01-03 Martin Sebor <msebor@redhat.com>
19310 * doc/extend.texi (attribute const): Fix a typo.
19311 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
19312 issuing -Wsuggest-attribute for void functions.
19314 2018-01-03 Martin Sebor <msebor@redhat.com>
19316 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
19317 offset_int::from instead of wide_int::to_shwi.
19318 (maybe_diag_overlap): Remove assertion.
19319 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
19320 * gimple-ssa-sprintf.c (format_directive): Same.
19321 (parse_directive): Same.
19322 (sprintf_dom_walker::compute_format_length): Same.
19323 (try_substitute_return_value): Same.
19325 2018-01-03 Jeff Law <law@redhat.com>
19327 PR middle-end/83654
19328 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
19329 non-constant residual for zero at runtime and avoid probing in
19330 that case. Reorganize code for trailing problem to mirror handling
19333 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19335 PR tree-optimization/83501
19336 * tree-ssa-strlen.c (get_string_cst): New.
19337 (handle_char_store): Call get_string_cst.
19339 2018-01-03 Martin Liska <mliska@suse.cz>
19341 PR tree-optimization/83593
19342 * tree-ssa-strlen.c: Include tree-cfg.h.
19343 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
19344 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
19345 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
19347 (strlen_dom_walker::before_dom_children): Call
19348 gimple_purge_dead_eh_edges. Dump tranformation with details
19350 (strlen_dom_walker::before_dom_children): Update call by adding
19351 new argument cleanup_eh.
19352 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
19354 2018-01-03 Martin Liska <mliska@suse.cz>
19357 * cif-code.def (VARIADIC_THUNK): New enum value.
19358 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
19361 2018-01-03 Jan Beulich <jbeulich@suse.com>
19363 * sse.md (mov<mode>_internal): Tighten condition for when to use
19364 vmovdqu<ssescalarsize> for TI and OI modes.
19366 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19368 Update copyright years.
19370 2018-01-03 Martin Liska <mliska@suse.cz>
19373 * ipa-visibility.c (function_and_variable_visibility): Skip
19374 functions with noipa attribure.
19376 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19378 * gcc.c (process_command): Update copyright notice dates.
19379 * gcov-dump.c (print_version): Ditto.
19380 * gcov.c (print_version): Ditto.
19381 * gcov-tool.c (print_version): Ditto.
19382 * gengtype.c (create_file): Ditto.
19383 * doc/cpp.texi: Bump @copying's copyright year.
19384 * doc/cppinternals.texi: Ditto.
19385 * doc/gcc.texi: Ditto.
19386 * doc/gccint.texi: Ditto.
19387 * doc/gcov.texi: Ditto.
19388 * doc/install.texi: Ditto.
19389 * doc/invoke.texi: Ditto.
19391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19393 * vector-builder.h (vector_builder::m_full_nelts): Change from
19394 unsigned int to poly_uint64.
19395 (vector_builder::full_nelts): Update prototype accordingly.
19396 (vector_builder::new_vector): Likewise.
19397 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
19398 (vector_builder::operator ==): Likewise.
19399 (vector_builder::finalize): Likewise.
19400 * int-vector-builder.h (int_vector_builder::int_vector_builder):
19401 Take the number of elements as a poly_uint64 rather than an
19403 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
19404 from unsigned int to poly_uint64.
19405 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
19406 (vec_perm_indices::new_vector): Likewise.
19407 (vec_perm_indices::length): Likewise.
19408 (vec_perm_indices::nelts_per_input): Likewise.
19409 (vec_perm_indices::input_nelts): Likewise.
19410 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
19411 number of elements per input as a poly_uint64 rather than an
19412 unsigned int. Use the original encoding for variable-length
19413 vectors, rather than clamping each individual element.
19414 For the second and subsequent elements in each pattern,
19415 clamp the step and base before clamping their sum.
19416 (vec_perm_indices::series_p): Handle polynomial element counts.
19417 (vec_perm_indices::all_in_range_p): Likewise.
19418 (vec_perm_indices_to_tree): Likewise.
19419 (vec_perm_indices_to_rtx): Likewise.
19420 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19421 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
19422 (tree_vector_builder::new_binary_operation): Handle polynomial
19423 element counts. Return false if we need to know the number
19424 of elements at compile time.
19425 * fold-const.c (fold_vec_perm): Punt if the number of elements
19426 isn't known at compile time.
19428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19430 * vec-perm-indices.h (vec_perm_builder): Change element type
19431 from HOST_WIDE_INT to poly_int64.
19432 (vec_perm_indices::element_type): Update accordingly.
19433 (vec_perm_indices::clamp): Handle polynomial element_types.
19434 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19435 (vec_perm_indices::all_in_range_p): Likewise.
19436 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
19438 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
19439 polynomial vec_perm_indices element types.
19440 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
19441 * fold-const.c (fold_vec_perm): Likewise.
19442 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
19443 * tree-vect-generic.c (lower_vec_perm): Likewise.
19444 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19445 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
19446 element type to HOST_WIDE_INT.
19448 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19449 Alan Hayward <alan.hayward@arm.com>
19450 David Sherwood <david.sherwood@arm.com>
19452 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
19453 rather than an int. Use plus_constant.
19454 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
19455 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
19457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19458 Alan Hayward <alan.hayward@arm.com>
19459 David Sherwood <david.sherwood@arm.com>
19461 * calls.c (emit_call_1, expand_call): Change struct_value_size from
19462 a HOST_WIDE_INT to a poly_int64.
19464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19465 Alan Hayward <alan.hayward@arm.com>
19466 David Sherwood <david.sherwood@arm.com>
19468 * calls.c (load_register_parameters): Cope with polynomial
19469 mode sizes. Require a constant size for BLKmode parameters
19470 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
19471 forces a parameter to be padded at the lsb end in order to
19472 fill a complete number of words, require the parameter size
19473 to be ordered wrt UNITS_PER_WORD.
19475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19476 Alan Hayward <alan.hayward@arm.com>
19477 David Sherwood <david.sherwood@arm.com>
19479 * reload1.c (spill_stack_slot_width): Change element type
19480 from unsigned int to poly_uint64_pod.
19481 (alter_reg): Treat mode sizes as polynomial.
19483 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19484 Alan Hayward <alan.hayward@arm.com>
19485 David Sherwood <david.sherwood@arm.com>
19487 * reload.c (complex_word_subreg_p): New function.
19488 (reload_inner_reg_of_subreg, push_reload): Use it.
19490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19491 Alan Hayward <alan.hayward@arm.com>
19492 David Sherwood <david.sherwood@arm.com>
19494 * lra-constraints.c (process_alt_operands): Reject matched
19495 operands whose sizes aren't ordered.
19496 (match_reload): Refer to this check here.
19498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19499 Alan Hayward <alan.hayward@arm.com>
19500 David Sherwood <david.sherwood@arm.com>
19502 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
19503 that the mode size is in the set {1, 2, 4, 8, 16}.
19505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19506 Alan Hayward <alan.hayward@arm.com>
19507 David Sherwood <david.sherwood@arm.com>
19509 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
19510 Use plus_constant instead of gen_rtx_PLUS.
19512 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19513 Alan Hayward <alan.hayward@arm.com>
19514 David Sherwood <david.sherwood@arm.com>
19516 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
19517 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
19518 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
19519 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
19520 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
19521 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
19522 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
19523 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
19524 * config/i386/i386.c (ix86_push_rounding): ...this new function.
19525 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
19527 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
19528 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
19529 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
19530 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
19531 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
19532 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
19533 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
19534 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
19535 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
19536 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
19538 * expr.c (emit_move_resolve_push): Treat the input and result
19539 of PUSH_ROUNDING as a poly_int64.
19540 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
19541 (emit_push_insn): Likewise.
19542 * lra-eliminations.c (mark_not_eliminable): Likewise.
19543 * recog.c (push_operand): Likewise.
19544 * reload1.c (elimination_effects): Likewise.
19545 * rtlanal.c (nonzero_bits1): Likewise.
19546 * calls.c (store_one_arg): Likewise. Require the padding to be
19547 known at compile time.
19549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19550 Alan Hayward <alan.hayward@arm.com>
19551 David Sherwood <david.sherwood@arm.com>
19553 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
19554 Use plus_constant instead of gen_rtx_PLUS.
19556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19557 Alan Hayward <alan.hayward@arm.com>
19558 David Sherwood <david.sherwood@arm.com>
19560 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
19561 rather than an int.
19563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19564 Alan Hayward <alan.hayward@arm.com>
19565 David Sherwood <david.sherwood@arm.com>
19567 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
19568 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
19569 via stack temporaries. Treat the mode size as polynomial too.
19571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19572 Alan Hayward <alan.hayward@arm.com>
19573 David Sherwood <david.sherwood@arm.com>
19575 * expr.c (expand_expr_real_2): When handling conversions involving
19576 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
19577 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
19578 as a poly_uint64 too.
19580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19581 Alan Hayward <alan.hayward@arm.com>
19582 David Sherwood <david.sherwood@arm.com>
19584 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
19586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19587 Alan Hayward <alan.hayward@arm.com>
19588 David Sherwood <david.sherwood@arm.com>
19590 * combine.c (can_change_dest_mode): Handle polynomial
19591 REGMODE_NATURAL_SIZE.
19592 * expmed.c (store_bit_field_1): Likewise.
19593 * expr.c (store_constructor): Likewise.
19594 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
19595 and polynomial REGMODE_NATURAL_SIZE.
19596 (gen_lowpart_common): Likewise.
19597 * reginfo.c (record_subregs_of_mode): Likewise.
19598 * rtlanal.c (read_modify_subreg_p): Likewise.
19600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19601 Alan Hayward <alan.hayward@arm.com>
19602 David Sherwood <david.sherwood@arm.com>
19604 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
19605 numbers of elements.
19607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19608 Alan Hayward <alan.hayward@arm.com>
19609 David Sherwood <david.sherwood@arm.com>
19611 * match.pd: Cope with polynomial numbers of vector elements.
19613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19614 Alan Hayward <alan.hayward@arm.com>
19615 David Sherwood <david.sherwood@arm.com>
19617 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
19618 in a POINTER_PLUS_EXPR.
19620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19621 Alan Hayward <alan.hayward@arm.com>
19622 David Sherwood <david.sherwood@arm.com>
19624 * omp-simd-clone.c (simd_clone_subparts): New function.
19625 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
19626 (ipa_simd_modify_function_body): Likewise.
19628 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19629 Alan Hayward <alan.hayward@arm.com>
19630 David Sherwood <david.sherwood@arm.com>
19632 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
19633 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
19634 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
19635 (expand_vector_condition, vector_element): Likewise.
19636 (subparts_gt): New function.
19637 (get_compute_type): Use subparts_gt.
19638 (count_type_subparts): Delete.
19639 (expand_vector_operations_1): Use subparts_gt instead of
19640 count_type_subparts.
19642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19643 Alan Hayward <alan.hayward@arm.com>
19644 David Sherwood <david.sherwood@arm.com>
19646 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
19647 (vect_compile_time_alias): ...this new function. Do the calculation
19648 on poly_ints rather than trees.
19649 (vect_prune_runtime_alias_test_list): Update call accordingly.
19651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19652 Alan Hayward <alan.hayward@arm.com>
19653 David Sherwood <david.sherwood@arm.com>
19655 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
19657 (vect_schedule_slp_instance): Likewise.
19659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19660 Alan Hayward <alan.hayward@arm.com>
19661 David Sherwood <david.sherwood@arm.com>
19663 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
19664 constant and extern definitions for variable-length vectors.
19665 (vect_get_constant_vectors): Note that the number of units
19666 is known to be constant.
19668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19669 Alan Hayward <alan.hayward@arm.com>
19670 David Sherwood <david.sherwood@arm.com>
19672 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
19673 of units as polynomial. Choose between WIDE and NARROW based
19676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19677 Alan Hayward <alan.hayward@arm.com>
19678 David Sherwood <david.sherwood@arm.com>
19680 * tree-vect-stmts.c (simd_clone_subparts): New function.
19681 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
19683 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19684 Alan Hayward <alan.hayward@arm.com>
19685 David Sherwood <david.sherwood@arm.com>
19687 * tree-vect-stmts.c (vectorizable_call): Treat the number of
19688 vectors as polynomial. Use build_index_vector for
19689 IFN_GOMP_SIMD_LANE.
19691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19692 Alan Hayward <alan.hayward@arm.com>
19693 David Sherwood <david.sherwood@arm.com>
19695 * tree-vect-stmts.c (get_load_store_type): Treat the number of
19696 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
19697 for variable-length vectors.
19698 (vectorizable_mask_load_store): Treat the number of units as
19699 polynomial, asserting that it is constant if the condition has
19700 already been enforced.
19701 (vectorizable_store, vectorizable_load): Likewise.
19703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19704 Alan Hayward <alan.hayward@arm.com>
19705 David Sherwood <david.sherwood@arm.com>
19707 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
19708 of units as polynomial. Punt if we can't tell at compile time
19709 which vector contains the final result.
19711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19712 Alan Hayward <alan.hayward@arm.com>
19713 David Sherwood <david.sherwood@arm.com>
19715 * tree-vect-loop.c (vectorizable_induction): Treat the number
19716 of units as polynomial. Punt on SLP inductions. Use an integer
19717 VEC_SERIES_EXPR for variable-length integer reductions. Use a
19718 cast of such a series for variable-length floating-point
19721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19722 Alan Hayward <alan.hayward@arm.com>
19723 David Sherwood <david.sherwood@arm.com>
19725 * tree.h (build_index_vector): Declare.
19726 * tree.c (build_index_vector): New function.
19727 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
19728 of units as polynomial, forcibly converting it to a constant if
19729 vectorizable_reduction has already enforced the condition.
19730 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
19731 to create a {1,2,3,...} vector.
19732 (vectorizable_reduction): Treat the number of units as polynomial.
19733 Choose vectype_in based on the largest scalar element size rather
19734 than the smallest number of units. Enforce the restrictions
19737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19738 Alan Hayward <alan.hayward@arm.com>
19739 David Sherwood <david.sherwood@arm.com>
19741 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
19742 number of units as polynomial.
19744 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19745 Alan Hayward <alan.hayward@arm.com>
19746 David Sherwood <david.sherwood@arm.com>
19748 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
19749 * target.def (autovectorize_vector_sizes): Return the vector sizes
19750 by pointer, using vector_sizes rather than a bitmask.
19751 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
19752 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
19753 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
19755 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
19756 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
19757 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
19758 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
19759 * omp-general.c (omp_max_vf): Likewise.
19760 * omp-low.c (omp_clause_aligned_alignment): Likewise.
19761 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
19762 * tree-vect-loop.c (vect_analyze_loop): Likewise.
19763 * tree-vect-slp.c (vect_slp_bb): Likewise.
19764 * doc/tm.texi: Regenerate.
19765 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
19767 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
19768 the vector size as a poly_uint64 rather than an unsigned int.
19769 (current_vector_size): Change from an unsigned int to a poly_uint64.
19770 (get_vectype_for_scalar_type): Update accordingly.
19771 * tree.h (build_truth_vector_type): Take the size and number of
19772 units as a poly_uint64 rather than an unsigned int.
19773 (build_vector_type): Add a temporary overload that takes
19774 the number of units as a poly_uint64 rather than an unsigned int.
19775 * tree.c (make_vector_type): Likewise.
19776 (build_truth_vector_type): Take the number of units as a poly_uint64
19777 rather than an unsigned int.
19779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19780 Alan Hayward <alan.hayward@arm.com>
19781 David Sherwood <david.sherwood@arm.com>
19783 * target.def (get_mask_mode): Take the number of units and length
19784 as poly_uint64s rather than unsigned ints.
19785 * targhooks.h (default_get_mask_mode): Update accordingly.
19786 * targhooks.c (default_get_mask_mode): Likewise.
19787 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19788 * doc/tm.texi: Regenerate.
19790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19791 Alan Hayward <alan.hayward@arm.com>
19792 David Sherwood <david.sherwood@arm.com>
19794 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
19795 * omp-general.c (omp_max_vf): Likewise.
19796 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
19797 (expand_omp_simd): Handle polynomial safelen.
19798 * omp-low.c (omplow_simd_context): Add a default constructor.
19799 (omplow_simd_context::max_vf): Change from int to poly_uint64.
19800 (lower_rec_simd_input_clauses): Update accordingly.
19801 (lower_rec_input_clauses): Likewise.
19803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19804 Alan Hayward <alan.hayward@arm.com>
19805 David Sherwood <david.sherwood@arm.com>
19807 * tree-vectorizer.h (vect_nunits_for_cost): New function.
19808 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
19809 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
19810 (vect_analyze_slp_cost): Likewise.
19811 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
19812 (vect_model_load_cost): Likewise.
19814 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19815 Alan Hayward <alan.hayward@arm.com>
19816 David Sherwood <david.sherwood@arm.com>
19818 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
19819 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
19820 from an unsigned int * to a poly_uint64_pod *.
19821 (calculate_unrolling_factor): New function.
19822 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
19824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19825 Alan Hayward <alan.hayward@arm.com>
19826 David Sherwood <david.sherwood@arm.com>
19828 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
19829 from an unsigned int to a poly_uint64.
19830 (_loop_vec_info::slp_unrolling_factor): Likewise.
19831 (_loop_vec_info::vectorization_factor): Change from an int
19833 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
19834 (vect_get_num_vectors): New function.
19835 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
19836 (vect_get_num_copies): Use vect_get_num_vectors.
19837 (vect_analyze_data_ref_dependences): Change max_vf from an int *
19838 to an unsigned int *.
19839 (vect_analyze_data_refs): Change min_vf from an int * to a
19841 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19842 than an unsigned HOST_WIDE_INT.
19843 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
19844 (vect_analyze_data_ref_dependence): Change max_vf from an int *
19845 to an unsigned int *.
19846 (vect_analyze_data_ref_dependences): Likewise.
19847 (vect_compute_data_ref_alignment): Handle polynomial vf.
19848 (vect_enhance_data_refs_alignment): Likewise.
19849 (vect_prune_runtime_alias_test_list): Likewise.
19850 (vect_shift_permute_load_chain): Likewise.
19851 (vect_supportable_dr_alignment): Likewise.
19852 (dependence_distance_ge_vf): Take the vectorization factor as a
19853 poly_uint64 rather than an unsigned HOST_WIDE_INT.
19854 (vect_analyze_data_refs): Change min_vf from an int * to a
19856 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
19857 vfm1 as a poly_uint64 rather than an int. Make the same change
19858 for the returned bound_scalar.
19859 (vect_gen_vector_loop_niters): Handle polynomial vf.
19860 (vect_do_peeling): Likewise. Update call to
19861 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
19862 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
19864 * tree-vect-loop.c (vect_determine_vectorization_factor)
19865 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
19866 (vect_get_known_peeling_cost): Likewise.
19867 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
19868 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
19869 (vect_transform_loop): Likewise. Use the lowest possible VF when
19870 updating the upper bounds of the loop.
19871 (vect_min_worthwhile_factor): Make static. Return an unsigned int
19872 rather than an int.
19873 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
19874 polynomial unroll factors.
19875 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
19876 (vect_make_slp_decision): Likewise.
19877 (vect_supported_load_permutation_p): Likewise, and polynomial
19879 (vect_analyze_slp_cost): Handle polynomial vf.
19880 (vect_slp_analyze_node_operations): Likewise.
19881 (vect_slp_analyze_bb_1): Likewise.
19882 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19883 than an unsigned HOST_WIDE_INT.
19884 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
19885 (vectorizable_load): Handle polynomial vf.
19886 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
19888 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
19890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19891 Alan Hayward <alan.hayward@arm.com>
19892 David Sherwood <david.sherwood@arm.com>
19894 * match.pd: Handle bit operations involving three constants
19895 and try to fold one pair.
19897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19899 * tree-vect-loop-manip.c: Include gimple-fold.h.
19900 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
19901 niters_maybe_zero parameters. Handle other cases besides a step of 1.
19902 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
19903 Add a path that uses a step of VF instead of 1, but disable it
19905 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
19906 and niters_no_overflow parameters. Update calls to
19907 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
19908 Create a new SSA name if the latter choses to use a ste other
19909 than zero, and return it via niters_vector_mult_vf_var.
19910 * tree-vect-loop.c (vect_transform_loop): Update calls to
19911 vect_do_peeling, vect_gen_vector_loop_niters and
19912 slpeel_make_loop_iterate_ntimes.
19913 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
19914 (vect_gen_vector_loop_niters): Update declarations after above changes.
19916 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
19918 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
19919 128-bit round to integer instructions.
19920 (ceil<mode>2): Likewise.
19921 (btrunc<mode>2): Likewise.
19922 (round<mode>2): Likewise.
19924 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19926 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
19927 unaligned VSX load/store on P8/P9.
19928 (expand_block_clear): Allow the use of unaligned VSX
19929 load/store on P8/P9.
19931 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19933 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
19935 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
19936 swap associated with both a load and a store.
19938 2018-01-02 Andrew Waterman <andrew@sifive.com>
19940 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
19941 * config/riscv/riscv.md (clear_cache): Use it.
19943 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
19945 * web.c: Remove out-of-date comment.
19947 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19949 * expr.c (fixup_args_size_notes): Check that any existing
19950 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
19951 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
19952 (emit_single_push_insn): ...here.
19954 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19956 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
19957 (const_vector_encoded_nelts): New function.
19958 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
19959 (const_vector_int_elt, const_vector_elt): Declare.
19960 * emit-rtl.c (const_vector_int_elt_1): New function.
19961 (const_vector_elt): Likewise.
19962 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
19963 of CONST_VECTOR_ELT.
19965 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19967 * expr.c: Include rtx-vector-builder.h.
19968 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
19969 directly on the tree encoding.
19970 (const_vector_from_tree): Likewise.
19971 * optabs.c: Include rtx-vector-builder.h.
19972 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
19973 sequence of "u" values.
19974 * vec-perm-indices.c: Include rtx-vector-builder.h.
19975 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
19976 directly on the vec_perm_indices encoding.
19978 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19980 * doc/rtl.texi (const_vector): Describe new encoding scheme.
19981 * Makefile.in (OBJS): Add rtx-vector-builder.o.
19982 * rtx-vector-builder.h: New file.
19983 * rtx-vector-builder.c: Likewise.
19984 * rtl.h (rtx_def::u2): Add a const_vector field.
19985 (CONST_VECTOR_NPATTERNS): New macro.
19986 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
19987 (CONST_VECTOR_DUPLICATE_P): Likewise.
19988 (CONST_VECTOR_STEPPED_P): Likewise.
19989 (CONST_VECTOR_ENCODED_ELT): Likewise.
19990 (const_vec_duplicate_p): Check for a duplicated vector encoding.
19991 (unwrap_const_vec_duplicate): Likewise.
19992 (const_vec_series_p): Check for a non-duplicated vector encoding.
19993 Say that the function only returns true for integer vectors.
19994 * emit-rtl.c: Include rtx-vector-builder.h.
19995 (gen_const_vec_duplicate_1): Delete.
19996 (gen_const_vector): Call gen_const_vec_duplicate instead of
19997 gen_const_vec_duplicate_1.
19998 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
19999 (gen_const_vec_duplicate): Use rtx_vector_builder.
20000 (gen_const_vec_series): Likewise.
20001 (gen_rtx_CONST_VECTOR): Likewise.
20002 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
20003 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
20004 Build a new vector rather than modifying a CONST_VECTOR in-place.
20005 (handle_special_swappables): Update call accordingly.
20006 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
20007 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
20008 Build a new vector rather than modifying a CONST_VECTOR in-place.
20009 (handle_special_swappables): Update call accordingly.
20011 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20013 * simplify-rtx.c (simplify_const_binary_operation): Use
20014 CONST_VECTOR_ELT instead of XVECEXP.
20016 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20018 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
20019 the selector elements to be different from the data elements
20020 if the selector is a VECTOR_CST.
20021 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
20022 ssizetype for the selector.
20024 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20026 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
20027 before testing each element individually.
20028 * tree-vect-generic.c (lower_vec_perm): Likewise.
20030 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20032 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
20033 * selftest-run-tests.c (selftest::run_tests): Call it.
20034 * vector-builder.h (vector_builder::operator ==): New function.
20035 (vector_builder::operator !=): Likewise.
20036 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
20037 (vec_perm_indices::all_from_input_p): New function.
20038 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
20039 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
20040 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
20041 instead of reading the VECTOR_CST directly. Detect whether both
20042 vector inputs are the same before constructing the vec_perm_indices,
20043 and update the number of inputs argument accordingly. Use the
20044 utility functions added above. Only construct sel2 if we need to.
20046 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20048 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
20049 the broadcast of the low byte.
20050 (expand_mult_highpart): Use an explicit encoding for the permutes.
20051 * optabs-query.c (can_mult_highpart_p): Likewise.
20052 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
20053 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20054 (vectorizable_bswap): Likewise.
20055 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
20056 explicit encoding for the power-of-2 permutes.
20057 (vect_permute_store_chain): Likewise.
20058 (vect_grouped_load_supported): Likewise.
20059 (vect_permute_load_chain): Likewise.
20061 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20063 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
20064 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
20065 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
20066 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
20067 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
20068 (vect_gen_perm_mask_any): Likewise.
20070 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20072 * int-vector-builder.h: New file.
20073 * vec-perm-indices.h: Include int-vector-builder.h.
20074 (vec_perm_indices): Redefine as an int_vector_builder.
20075 (auto_vec_perm_indices): Delete.
20076 (vec_perm_builder): Redefine as a stand-alone class.
20077 (vec_perm_indices::vec_perm_indices): New function.
20078 (vec_perm_indices::clamp): Likewise.
20079 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
20080 (vec_perm_indices::new_vector): New function.
20081 (vec_perm_indices::new_expanded_vector): Update for new
20082 vec_perm_indices class.
20083 (vec_perm_indices::rotate_inputs): New function.
20084 (vec_perm_indices::all_in_range_p): Operate directly on the
20085 encoded form, without computing elided elements.
20086 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
20087 encoding. Update for new vec_perm_indices class.
20088 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
20089 the given vec_perm_builder.
20090 (expand_vec_perm_var): Update vec_perm_builder constructor.
20091 (expand_mult_highpart): Use vec_perm_builder instead of
20092 auto_vec_perm_indices.
20093 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
20094 vec_perm_indices instead of auto_vec_perm_indices. Use a single
20095 or double series encoding as appropriate.
20096 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
20097 vec_perm_indices instead of auto_vec_perm_indices.
20098 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20099 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20100 (vect_permute_store_chain): Likewise.
20101 (vect_grouped_load_supported): Likewise.
20102 (vect_permute_load_chain): Likewise.
20103 (vect_shift_permute_load_chain): Likewise.
20104 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
20105 (vect_transform_slp_perm_load): Likewise.
20106 (vect_schedule_slp_instance): Likewise.
20107 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20108 (vectorizable_mask_load_store): Likewise.
20109 (vectorizable_bswap): Likewise.
20110 (vectorizable_store): Likewise.
20111 (vectorizable_load): Likewise.
20112 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
20113 vec_perm_indices instead of auto_vec_perm_indices. Use
20114 tree_to_vec_perm_builder to read the vector from a tree.
20115 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
20116 vec_perm_builder instead of a vec_perm_indices.
20117 (have_whole_vector_shift): Use vec_perm_builder and
20118 vec_perm_indices instead of auto_vec_perm_indices. Leave the
20119 truncation to calc_vec_perm_mask_for_shift.
20120 (vect_create_epilog_for_reduction): Likewise.
20121 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
20122 from auto_vec_perm_indices to vec_perm_indices.
20123 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
20124 instead of changing individual elements.
20125 (aarch64_vectorize_vec_perm_const): Use new_vector to install
20126 the vector in d.perm.
20127 * config/arm/arm.c (expand_vec_perm_d::perm): Change
20128 from auto_vec_perm_indices to vec_perm_indices.
20129 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
20130 instead of changing individual elements.
20131 (arm_vectorize_vec_perm_const): Use new_vector to install
20132 the vector in d.perm.
20133 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
20134 Update vec_perm_builder constructor.
20135 (rs6000_expand_interleave): Likewise.
20136 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
20137 (rs6000_expand_interleave): Likewise.
20139 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20141 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
20142 to qimode could truncate the indices.
20143 * optabs.c (expand_vec_perm_var): Likewise.
20145 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20147 * Makefile.in (OBJS): Add vec-perm-indices.o.
20148 * vec-perm-indices.h: New file.
20149 * vec-perm-indices.c: Likewise.
20150 * target.h (vec_perm_indices): Replace with a forward class
20152 (auto_vec_perm_indices): Move to vec-perm-indices.h.
20153 * optabs.h: Include vec-perm-indices.h.
20154 (expand_vec_perm): Delete.
20155 (selector_fits_mode_p, expand_vec_perm_var): Declare.
20156 (expand_vec_perm_const): Declare.
20157 * target.def (vec_perm_const_ok): Replace with...
20158 (vec_perm_const): ...this new hook.
20159 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
20160 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
20161 * doc/tm.texi: Regenerate.
20162 * optabs.def (vec_perm_const): Delete.
20163 * doc/md.texi (vec_perm_const): Likewise.
20164 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
20165 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
20166 expand_vec_perm for constant permutation vectors. Assert that
20167 the mode of variable permutation vectors is the integer equivalent
20168 of the mode that is being permuted.
20169 * optabs-query.h (selector_fits_mode_p): Declare.
20170 * optabs-query.c: Include vec-perm-indices.h.
20171 (selector_fits_mode_p): New function.
20172 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
20173 is defined, instead of checking whether the vec_perm_const_optab
20174 exists. Use targetm.vectorize.vec_perm_const instead of
20175 targetm.vectorize.vec_perm_const_ok. Check whether the indices
20176 fit in the vector mode before using a variable permute.
20177 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
20178 vec_perm_indices instead of an rtx.
20179 (expand_vec_perm): Replace with...
20180 (expand_vec_perm_const): ...this new function. Take the selector
20181 as a vec_perm_indices rather than an rtx. Also take the mode of
20182 the selector. Update call to shift_amt_for_vec_perm_mask.
20183 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
20184 Use vec_perm_indices::new_expanded_vector to expand the original
20185 selector into bytes. Check whether the indices fit in the vector
20186 mode before using a variable permute.
20187 (expand_vec_perm_var): Make global.
20188 (expand_mult_highpart): Use expand_vec_perm_const.
20189 * fold-const.c: Includes vec-perm-indices.h.
20190 * tree-ssa-forwprop.c: Likewise.
20191 * tree-vect-data-refs.c: Likewise.
20192 * tree-vect-generic.c: Likewise.
20193 * tree-vect-loop.c: Likewise.
20194 * tree-vect-slp.c: Likewise.
20195 * tree-vect-stmts.c: Likewise.
20196 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
20198 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
20199 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
20200 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
20201 (aarch64_vectorize_vec_perm_const): ...this new function.
20202 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
20203 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20204 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
20205 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
20206 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
20207 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20208 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
20210 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
20211 check for NEON modes.
20212 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
20213 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
20214 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
20215 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
20217 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
20218 the old VEC_PERM_CONST conditions.
20219 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
20220 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
20221 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
20222 (ia64_vectorize_vec_perm_const_ok): Merge into...
20223 (ia64_vectorize_vec_perm_const): ...this new function.
20224 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
20225 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
20226 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
20227 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
20228 * config/mips/mips.c (mips_expand_vec_perm_const)
20229 (mips_vectorize_vec_perm_const_ok): Merge into...
20230 (mips_vectorize_vec_perm_const): ...this new function.
20231 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
20232 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
20233 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
20234 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
20235 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
20236 (rs6000_expand_vec_perm_const): Delete.
20237 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
20239 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20240 (altivec_expand_vec_perm_const_le): Take each operand individually.
20241 Operate on constant selectors rather than rtxes.
20242 (altivec_expand_vec_perm_const): Likewise. Update call to
20243 altivec_expand_vec_perm_const_le.
20244 (rs6000_expand_vec_perm_const): Delete.
20245 (rs6000_vectorize_vec_perm_const_ok): Delete.
20246 (rs6000_vectorize_vec_perm_const): New function.
20247 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
20248 an element count and rtx array.
20249 (rs6000_expand_extract_even): Update call accordingly.
20250 (rs6000_expand_interleave): Likewise.
20251 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
20252 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
20253 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
20254 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
20255 (rs6000_expand_vec_perm_const): Delete.
20256 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
20257 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20258 (altivec_expand_vec_perm_const_le): Take each operand individually.
20259 Operate on constant selectors rather than rtxes.
20260 (altivec_expand_vec_perm_const): Likewise. Update call to
20261 altivec_expand_vec_perm_const_le.
20262 (rs6000_expand_vec_perm_const): Delete.
20263 (rs6000_vectorize_vec_perm_const_ok): Delete.
20264 (rs6000_vectorize_vec_perm_const): New function. Remove stray
20265 reference to the SPE evmerge intructions.
20266 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
20267 an element count and rtx array.
20268 (rs6000_expand_extract_even): Update call accordingly.
20269 (rs6000_expand_interleave): Likewise.
20270 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
20271 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
20273 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20275 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20277 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
20278 vector mode and that that mode matches the mode of the data
20280 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
20281 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
20282 directly using expand_vec_perm_1 when forcing selectors into
20284 (expand_vec_perm_var): New function, split out from expand_vec_perm.
20286 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20288 * optabs-query.h (can_vec_perm_p): Delete.
20289 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
20290 * optabs-query.c (can_vec_perm_p): Split into...
20291 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
20292 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
20293 particular selector is valid.
20294 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20295 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20296 (vect_grouped_load_supported): Likewise.
20297 (vect_shift_permute_load_chain): Likewise.
20298 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
20299 (vect_transform_slp_perm_load): Likewise.
20300 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20301 (vectorizable_bswap): Likewise.
20302 (vect_gen_perm_mask_checked): Likewise.
20303 * fold-const.c (fold_ternary_loc): Likewise. Don't take
20304 implementations of variable permutation vectors into account
20305 when deciding which selector to use.
20306 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
20307 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
20308 with a false third argument.
20309 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
20310 to test whether the constant selector is valid and can_vec_perm_var_p
20311 to test whether a variable selector is valid.
20313 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20315 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
20316 * optabs-query.c (can_vec_perm_p): Likewise.
20317 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
20318 instead of vec_perm_indices.
20319 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
20320 (vect_gen_perm_mask_checked): Likewise,
20321 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
20322 (vect_gen_perm_mask_checked): Likewise,
20324 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20326 * optabs-query.h (qimode_for_vec_perm): Declare.
20327 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
20328 (qimode_for_vec_perm): ...this new function.
20329 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
20331 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20333 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
20334 does not have a conditional at the top.
20336 2018-01-02 Richard Biener <rguenther@suse.de>
20338 * ipa-inline.c (big_speedup_p): Fix expression.
20340 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
20343 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
20346 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
20350 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
20351 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
20352 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
20353 cond_taken_branch_cost 3->4.
20355 2018-01-01 Jakub Jelinek <jakub@redhat.com>
20357 PR tree-optimization/83581
20358 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
20359 TODO_cleanup_cfg if any changes have been made.
20361 PR middle-end/83608
20362 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
20363 convert_modes if target mode has the right side, but different mode
20366 PR middle-end/83609
20367 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
20368 last argument when extracting from CONCAT. If either from_real or
20369 from_imag is NULL, use expansion through memory. If result is not
20370 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
20371 the parts directly to inner mode, if even that fails, use expansion
20374 PR middle-end/83623
20375 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
20376 check for bswap in mode rather than HImode and use that in expand_unop
20379 Copyright (C) 2018 Free Software Foundation, Inc.
20381 Copying and distribution of this file, with or without modification,
20382 are permitted in any medium without royalty provided the copyright
20383 notice and this notice are preserved.