| blob | 14326080acdd97edd89dedd4aa66929412e2776e |
1 2018-02-21 Martin Liska <mliska@suse.cz>
3 PR driver/83193
4 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5 Add "native" as a possible value.
6 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7 the macro when native cpu detection is available.
9 2018-02-21 Martin Liska <mliska@suse.cz>
11 PR driver/83193
12 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
13 Add "native" as a possible value.
14 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
15 when native cpu detection is available.
17 2018-02-21 Jakub Jelinek <jakub@redhat.com>
18 Martin Sebor <msebor@redhat.com>
20 PR tree-optimization/84478
21 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
22 false.
23 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
24 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
25 support which is conservatively correct, for 2 only stay conservative
26 for maxlen. Formatting and comment capitalization fixes. Add STRICT
27 argument to the 2 argument get_range_strlen, adjust 6 arg
28 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
29 false.
30 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
31 (gimple_fold_builtin_strlen): Pass true as last argument to
32 get_range_strlen.
34 2018-02-20 Martin Sebor <msebor@redhat.com>
36 PR middle-end/84095
37 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
38 (builtin_memref::set_base_and_offset): Same. Handle inner references.
39 (builtin_memref::builtin_memref): Factor out parts into
40 set_base_and_offset and call it.
42 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
44 PR middle-end/84406
45 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
46 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
47 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
48 search at the associated MODE_INT.
50 2018-02-20 Jeff Law <law@redhat.com>
52 PR middle-end/82123
53 PR tree-optimization/81592
54 PR middle-end/79257
55 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
56 for range data rather than using global data.
57 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
58 range data rather than using global data.
59 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
60 pass it to children as needed.
61 (struct directive::fmtresult): Similarly.
62 (struct directive::set_width): Similarly.
63 (struct directive::set_precision): Similarly.
64 (format_integer, format_directive, parse_directive): Similarly.
65 (format_none): Accept unnamed vr_values parameter.
66 (format_percent, format_floating, format_character): Similarly.
67 (format_string, format_plain): Similarly.
68 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
69 the EVRP range analyzer for range data rather than using global data.
70 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
71 gimple-ssa-evrp-analyze.h
72 (class sprintf_dom_walker): Add after_dom_children member function.
73 Add evrp_range_analyzer member.
74 (sprintf_dom_walker::before_dom_children): Call into the EVRP
75 range analyzer as needed.
76 (sprintf_dom_walker::after_dom_children): New member function.
77 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
78 if not optimizing.
79 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
80 (evrp_range_analyzer::pop_to_marker): Likewise.
82 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
84 PR tree-optimization/84419
85 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
86 with the required type if its current type is compatible but
87 different.
89 2018-02-20 Jakub Jelinek <jakub@redhat.com>
91 PR middle-end/82004
92 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
93 after vectorization.
95 2018-02-20 Martin Liska <mliska@suse.cz>
97 PR driver/83193
98 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
99 possible values if we don't have a hint.
101 2018-02-20 Martin Liska <mliska@suse.cz>
103 PR c/84310
104 PR target/79747
105 * final.c (shorten_branches): Build align_tab array with one
106 more element.
107 * opts.c (finish_options): Add alignment option limit check.
108 (MAX_CODE_ALIGN): Likewise.
109 (MAX_CODE_ALIGN_VALUE): Likewise.
110 * doc/invoke.texi: Document maximum allowed option value for
111 all -falign-* options.
113 2018-02-19 Jakub Jelinek <jakub@redhat.com>
115 PR target/84146
116 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
117 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
118 * var-tracking.c (emit_note_insn_var_location): Remove all references
119 to NOTE_INSN_CALL_ARG_LOCATION.
120 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
121 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
122 Use copy_rtx_if_shared.
123 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
124 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
125 (dwarf2out_var_location): Remove handling of
126 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
127 on call_insn.
128 * final.c (final_scan_insn): Remove all references to
129 NOTE_INSN_CALL_ARG_LOCATION.
130 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
131 before dumping final insns.
132 * except.c (emit_note_eh_region_end): Remove all references to
133 NOTE_INSN_CALL_ARG_LOCATION.
134 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
135 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
136 * config/arc/arc.c (hwloop_optimize): Likewise.
137 * config/arm/arm.c (create_fix_barrier): Likewise.
138 * config/s390/s390.c (s390_chunkify_start): Likewise.
139 * config/sh/sh.c (find_barrier): Likewise.
140 * config/i386/i386.c (rest_of_insert_endbranch,
141 ix86_seh_fixup_eh_fallthru): Likewise.
142 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
143 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
144 * config/frv/frv.c (frv_function_prologue): Likewise.
145 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
146 reg note.
147 (note_outside_basic_block_p): Remove all references to
148 NOTE_INSN_CALL_ARG_LOCATION.
149 * gengtype.c (adjust_field_rtx_def): Likewise.
150 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
151 Likewise.
152 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
153 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
155 PR c++/84444
156 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
157 is ADDR_EXPR.
159 PR tree-optimization/84452
160 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
161 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
162 is NULL.
164 2018-02-19 Martin Liska <mliska@suse.cz>
166 PR sanitizer/82183
167 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
169 2018-02-19 Martin Liska <mliska@suse.cz>
170 Richard Sandiford <richard.sandiford@linaro.org>
172 PR tree-optimization/82491
173 * gimple-fold.c (get_base_constructor): Make earlier bail out
174 to prevent ubsan.
176 2018-02-19 Carl Love <cel@us.ibm.com>
178 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
179 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
180 BU_P8V_OVERLOAD_1.
181 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
182 P8V_BUILTIN_VEC_NEG.
184 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
186 * config/rl78/rl78.md (movdf): New define expand.
188 2018-02-19 Martin Liska <mliska@suse.cz>
190 PR other/80589
191 * doc/invoke.texi: Fix typo.
192 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
194 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
196 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
197 handle rs6000_single_float and rs6000_double_float specially for
198 e500 family CPUs.
200 2018-02-16 Jeff Law <law@redhat.com>
202 * config/rx/rx.c (add_pop_cfi_notes): New function.;
203 (pop_regs): Use it.
205 2018-02-16 Jakub Jelinek <jakub@redhat.com>
207 PR ipa/84425
208 * ipa-inline.c (inline_small_functions): Fix a typo.
210 2018-02-16 Nathan Sidwell <nathan@acm.org>
212 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
214 2018-02-16 Carl Love <cel@us.ibm.com>
216 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
217 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
218 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
219 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
220 expansion to P8V_BUILTIN_VEC_FLOAT2.
222 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
224 PR rtl-optimization/70023
225 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
226 src_regno into account.
228 2018-02-16 Carl Love <cel@us.ibm.com>
230 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
231 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
232 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
233 * config/rs6000/rs6000.c: Remove case statements for
234 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
235 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
236 and P9V_BUILTIN_VEC_VINSERT4B.
237 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
238 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
239 * config/rs6000/vsx.md:
240 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
241 vec_insert4b.
243 2018-02-16 Carl Love <cel@us.ibm.com>
245 * config/rs6000/altivec.h: Add builtin names vec_extract4b
246 vec_insert4b.
247 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
248 definitions.
249 * config/rs6000/rs6000-c.c: Add the definitions for
250 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
251 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
252 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
253 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
254 definition for insert4b and define insn *insert3b_internal.
255 * doc/extend.texi: Add documentation for vec_extract4b.
257 2018-02-16 Nathan Sidwell <nathan@acm.org>
259 * doc/extend.texi (Backwards Compatibility): Mention friend
260 injection. Note for-scope is deprecated.
261 * doc/invoke.texi (-ffriend-injection): Deprecate.
263 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
265 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
266 that moved to I2, also allow destinations that are a paradoxical
267 subreg (instead of a normal reg).
269 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
271 PR target/83831
272 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
273 to QImode.
275 2018-02-16 Richard Biener <rguenther@suse.de>
277 PR tree-optimization/84037
278 PR tree-optimization/84016
279 PR target/82862
280 * config/i386/i386.c (ix86_builtin_vectorization_cost):
281 Adjust vec_construct for the fact we need additional higher latency
282 128bit inserts for AVX256 and AVX512 vector builds.
283 (ix86_add_stmt_cost): Scale vector construction cost for
284 elementwise loads.
286 2018-02-16 Richard Biener <rguenther@suse.de>
288 PR tree-optimization/84417
289 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
290 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
291 (non_rewritable_lvalue_p): Likewise, use poly-ints.
293 2018-02-16 Martin Liska <mliska@suse.cz>
295 PR sanitizer/84307
296 * internal-fn.def (ASAN_CHECK): Set proper flags.
297 (ASAN_MARK): Likewise.
299 2018-02-16 Julia Koval <julia.koval@intel.com>
301 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
302 from PTA_CANNONLAKE.
304 2018-02-16 Jakub Jelinek <jakub@redhat.com>
306 PR target/84272
307 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
308 Use ++iter rather than iter++ for std::list iterators.
309 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
310 defer deleting them until all nodes in the forest are processed. Do
311 free even leaf nodes. Change to_process into auto_vec.
313 PR bootstrap/84405
314 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
315 * vec.h (vec_default_construct): Use memset instead of placement new
316 if BROKEN_VALUE_INITIALIZATION is defined.
317 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
318 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
319 is defined.
321 PR rtl-optimization/83723
322 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
323 * lra.c (lra_substitute_pseudo): Likewise. If true, use
324 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
325 recursive calls.
326 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
327 callers.
328 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
330 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
332 PR rtl-optimization/81443
333 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
334 from inner REGs to paradoxical SUBREGs.
336 2018-02-16 Richard Biener <rguenther@suse.de>
338 PR tree-optimization/84399
339 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
340 For operands we can analyze at their definition make sure we can
341 analyze them at each use as well.
343 2018-02-16 Richard Biener <rguenther@suse.de>
345 PR tree-optimization/84190
346 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
347 volatile accesses if the decl isn't volatile.
349 2018-02-15 Jason Merrill <jason@redhat.com>
351 PR c++/84314 - ICE with templates and fastcall attribute.
352 * attribs.c (build_type_attribute_qual_variant): Don't clobber
353 TYPE_CANONICAL on an existing type.
355 2018-02-15 Jakub Jelinek <jakub@redhat.com>
357 PR tree-optimization/84383
358 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
359 dstoff nor call operand_equal_p if dstbase is NULL.
361 PR tree-optimization/84334
362 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
363 also a CONSTANT_CLASS_P, punt.
365 2018-02-14 Jim Wilson <jimw@sifive.com>
367 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
368 first SMALL_OPERAND check. New local min_second_step. Move assert
369 to where locals are set. Add TARGET_RVC support.
370 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
372 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
374 * doc/invoke.texi: Correct -Wformat-overflow code sample.
376 2018-02-14 Martin Sebor <msebor@redhat.com>
378 PR tree-optimization/83698
379 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
380 arrays constrain the offset range to their bounds.
381 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
382 (builtin_access::overlap): Avoid setting the size of overlap if it's
383 already been set.
384 (maybe_diag_overlap): Also consider arrays when deciding what values
385 of offsets to include in diagnostics.
387 2018-02-14 Martin Sebor <msebor@redhat.com>
389 PR c/84108
390 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
391 that correspond to the kind of a declaration.
393 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
395 PR target/83984
396 * config/pa/pa.md: Load address of PIC label using the linkage table
397 if the label is nonlocal.
399 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
401 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
402 warning message if user requests -maltivec=be.
403 * doc/invoke.texi: Document deprecation of -maltivec=be.
405 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
407 PR target/84220
408 * config/rs6000/rs6000-c.c: Update definitions for
409 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
410 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
412 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
414 PR target/84239
415 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
416 add _get_ssp intrinsics. Remove argument from
417 __builtin_ia32_rdssp[d|q].
418 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
419 * config/i386/i386-builtin.def: Remove argument from
420 __builtin_ia32_rdssp[d|q].
421 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
422 ix86_expand_special_args_builtin for _rdssp[d|q].
423 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
424 Clear register before usage.
425 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
426 Add documentation for new _get_ssp and _inc_ssp intrinsics.
428 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
430 PR tree-optimization/84357
431 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
432 operand 1 of an ARRAY_REF too.
434 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
436 PR target/83831
437 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
438 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
439 declarations.
440 (set_of_reg): New struct.
441 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
442 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
443 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
444 functions.
445 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
446 Split into bitclr, bitset, bitinvert patterns if appropriate.
447 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
448 use rx_fuse_in_memory_bitop.
449 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
450 to named insn, correct maximum insn length.
452 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
454 PR target/79242
455 * machmode.def: Define a complex mode for PARTIAL_INT.
456 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
457 MODE_PARTIAL_INT.
458 * doc/rtl.texi: Document CSPImode.
459 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
460 handling.
461 (msp430_hard_regno_nregs_with_padding): Likewise.
463 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
465 PR target/84279
466 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
468 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
470 PR rtl-optimization/84169
471 * combine.c (try_combine): New variable split_i2i3. Set it to true if
472 we generated a parallel as new i3 and we split that to new i2 and i3
473 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
474 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
475 those to i2, not i1. Partially rewrite this scan code.
477 2018-02-13 Jakub Jelinek <jakub@redhat.com>
479 PR c/82210
480 * stor-layout.c (place_field): For variable length fields, adjust
481 offset_align afterwards not just based on the field's alignment,
482 but also on the size.
484 PR middle-end/84309
485 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
486 of exps and logs in the use_exp2 case.
488 2018-02-13 Jeff Law <law@redhat.com>
490 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
491 entry for "vector".
493 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
494 ARGS as unused.
496 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
498 PR debug/84342
499 PR debug/84319
500 * common.opt (gas-loc-support, gas-locview-support): New.
501 (ginline-points, ginternal-reset-location-views): New.
502 * doc/invoke.texi: Document them. Use @itemx where intended.
503 (gvariable-location-views): Adjust.
504 * target.def (reset_location_view): New.
505 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
506 (TARGET_RESET_LOCATION_VIEW): New.
507 * doc/tm.texi: Rebuilt.
508 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
509 (dwarf2out_default_as_locview_support): New.
510 (output_asm_line_debug_info): Use option variables.
511 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
512 (output_loc_list): Likewise.
513 (add_high_low_attributes): Check option variables.
514 Don't output entry view attribute in strict mode.
515 (gen_inlined_subroutine_die): Check option variables.
516 (dwarf2out_inline_entry): Likewise.
517 (init_sections_and_labels): Likewise.
518 (dwarf2out_early_finish): Likewise.
519 (maybe_reset_location_view): New, from...
520 (dwarf2out_var_location): ... here. Call it.
521 * debug.h (dwarf2out_default_as_loc_support): Declare.
522 (dwarf2out_default_as_locview_support): Declare.
523 * hooks.c (hook_int_rtx_insn_0): New.
524 * hooks.h (hook_int_rtx_insn_0): Declare.
525 * toplev.c (process_options): Take -gas-loc-support and
526 -gas-locview-support from dwarf2out. Enable
527 -gvariable-location-views by default only with locview
528 assembler support. Enable -ginternal-reset-location-views by
529 default only if the target defines the corresponding hook.
530 Enable -ginline-points by default if location views are
531 enabled; force it disabled if statement frontiers are
532 disabled.
533 * tree-inline.c (expand_call_inline): Check option variables.
534 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
536 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
538 PR tree-optimization/84321
539 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
540 handling. Also check whether the anti-range contains any values
541 that satisfy the mask; switch to a VR_RANGE if not.
543 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
545 PR sanitizer/84340
546 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
548 2018-02-13 Martin Jambor <mjambor@suse.cz>
550 PR c++/83990
551 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
552 of call statements, also set location of a load to a temporary.
554 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
556 * config/rl78/rl78.c (add_vector_labels): New function.
557 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
558 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
559 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
560 which checks that no arguments are passed.
561 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
562 * doc/extend.texi: Documentation for the new attribute.
564 2018-02-13 Andreas Schwab <schwab@suse.de>
566 * config/riscv/linux.h (CPP_SPEC): Define.
568 2018-02-13 Jakub Jelinek <jakub@redhat.com>
570 PR target/84335
571 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
572 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
573 OPTION_MASK_ISA_AES as first argument to def_builtin_const
574 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
575 instead of OPTION_MASK_ISA_PCLMUL as first argument to
576 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
577 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
578 temporarily for AES and PCLMUL builtins.
580 PR tree-optimization/84339
581 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
582 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
583 Formatting fixes.
585 PR middle-end/84309
586 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
587 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
588 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
589 inline function.
590 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
591 inline function.
592 * omp-simd-clone.h: New file.
593 * omp-simd-clone.c: Include omp-simd-clone.h.
594 (expand_simd_clones): No longer static.
595 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
596 cgraph.h and omp-simd-clone.h.
597 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
598 (vect_recog_widen_shift_pattern): Formatting fix.
599 (vect_pattern_recog_1): Don't check optab for calls.
601 PR target/84336
602 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
603 operands[2] into a REG before using gen_lowpart on it.
605 2018-02-12 Jeff Law <law@redhat.com>
607 PR target/83760
608 * config/sh/sh.c (find_barrier): Consider a sibling call
609 a barrier as well.
611 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
612 successfully back substituting a reg.
614 2018-02-12 Richard Biener <rguenther@suse.de>
616 PR tree-optimization/84037
617 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
618 parameter, move visited init to caller.
619 (vect_slp_analyze_operations): Separate cost from validity
620 check, initialize visited once for all instances.
621 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
622 for all instances.
623 * tree-vect-stmts.c (vect_model_simple_cost): Make early
624 out an assert.
625 (vect_model_promotion_demotion_cost): Likewise.
626 (vectorizable_bswap): Guard cost modeling with !slp_node
627 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
628 SLP stmts.
629 (vectorizable_call): Likewise.
630 (vectorizable_conversion): Likewise.
631 (vectorizable_assignment): Likewise.
632 (vectorizable_shift): Likewise.
633 (vectorizable_operation): Likewise.
634 (vectorizable_store): Likewise.
635 (vectorizable_load): Likewise.
636 (vectorizable_condition): Likewise.
637 (vectorizable_comparison): Likewise.
639 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
641 PR sanitizer/84307
642 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
643 (ASAN_MARK): Fix fnspec to account for return value, change pointer
644 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
646 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
648 PR middle-end/83665
649 * params.def (inline-min-speedup): Increase from 8 to 15.
650 (max-inline-insns-auto): Decrease from 40 to 30.
651 * ipa-split.c (consider_split): Add some buffer for function to
652 be considered inlining candidate.
653 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
654 default values.
656 2018-02-12 Richard Biener <rguenther@suse.de>
658 PR tree-optimization/84037
659 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
660 matched stmts if we cannot swap the non-matched ones.
662 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
664 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
665 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
666 _mm_maskz_scalef_round_ss): New intrinsics.
667 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
668 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
669 __builtin_ia32_scalefss_round): Remove.
670 (__builtin_ia32_scalefsd_mask_round,
671 __builtin_ia32_scalefss_mask_round): New intrinsics.
672 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
673 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
674 ((match_operand:VF_128 2 "<round_nimm_predicate>"
675 "<round_constraint>")): Changed to ...
676 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
677 "<round_scalar_constraint>")): ... this.
678 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
679 %0, %1, %2<round_op3>}"): Changed to ...
680 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
681 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
682 %2<round_scalar_mask_op3>}"): ... this.
683 * config/i386/subst.md (round_scalar_nimm_predicate): New.
685 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
687 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
688 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
689 (_mm_maskz_sqrt_round_ss): New intrinsics.
690 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
691 (__builtin_ia32_sqrtsd_mask_round)
692 (__builtin_ia32_sqrtss_mask_round): New builtins.
693 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
694 (__builtin_ia32_sqrtss_round): Remove.
695 (__builtin_ia32_sqrtsd_mask_round)
696 (__builtin_ia32_sqrtss_mask_round): New builtins.
697 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
698 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
699 ((match_operand:VF_128 1 "vector_operand"
700 "xBm,<round_constraint>")): Changed to ...
701 ((match_operand:VF_128 1 "vector_operand"
702 "xBm,<round_scalar_constraint>")): ... this.
703 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
704 %0, %2, %<iptr>1<round_op3>}): Changed to ...
705 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
706 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
707 %<iptr>1<round_scalar_mask_op3>}): ... this.
708 ((set_attr "prefix" "<round_prefix>")): Changed to ...
709 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
711 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
713 PR target/84266
714 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
715 Cast vec_cmpeq result to correct type.
716 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
717 Cast vec_cmpgt result to correct type.
719 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
721 * final.c (final_scan_insn_1): Renamed from...
722 (final_scan_insn): ... this. New wrapper, to recover
723 seen from the outermost call in recursive ones.
724 * config/sparc/sparc.c (output_return): Drop seen from call.
725 (output_sibcall): Likewise.
726 * config/visium/visium.c (output_branch): Likewise.
728 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
730 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
731 function label.
733 2018-02-10 Alan Modra <amodra@gmail.com>
735 PR target/84300
736 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
737 Specify LR as an input.
739 2018-02-10 Jakub Jelinek <jakub@redhat.com>
741 PR sanitizer/83987
742 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
743 remove_member_access_dummy_vars): New functions.
744 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
745 lower_omp_1, execute_lower_omp): Use them.
747 PR rtl-optimization/84308
748 * shrink-wrap.c (spread_components): Release todo vector.
750 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
752 PR rtl-optimization/57193
753 * ira-color.c (struct allocno_color_data): Add member
754 conflict_allocno_hard_prefs.
755 (update_conflict_allocno_hard_prefs): New.
756 (bucket_allocno_compare_func): Add a preference based on
757 conflict_allocno_hard_prefs.
758 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
759 (color_allocnos): Remove a dead code. Initiate
760 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
762 2018-02-09 Jakub Jelinek <jakub@redhat.com>
764 PR target/84226
765 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
766 constraint from =wa to wa. Avoid a subreg on the output operand,
767 instead use a pseudo and subreg it in a move.
768 (p9_xxbrd_<mode>): Changed to ...
769 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
770 (p9_xxbrd_v2df): New expander.
771 (p9_xxbrw_<mode>): Changed to ...
772 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
773 (p9_xxbrw_v4sf): New expander.
775 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
777 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
779 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
781 PR target/83926
782 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
783 multiply in 32-bit mode.
784 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
785 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
786 mode.
788 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
790 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
791 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
792 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
793 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
795 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
797 PR lto/84213
798 * dwarf2out.c (is_trivial_indirect_ref): New function.
799 (dwarf2out_late_global_decl): Do not generate a location
800 attribute for variables that have a non-trivial DECL_VALUE_EXPR
801 and that are not defined in the current unit.
803 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
805 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
806 instead of a libcall for UNORDERED.
808 2018-02-09 Tamar Christina <tamar.christina@arm.com>
810 PR target/82641
811 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
812 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
814 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
816 PR target/PR84295
817 * config/s390/s390.c (s390_set_current_function): Invoke
818 s390_indirect_branch_settings also if fndecl didn't change.
820 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
822 * config/rs6000/rs6000.md (blockage): Set length to zero.
824 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
826 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
828 2018-02-09 Jakub Jelinek <jakub@redhat.com>
830 PR sanitizer/84285
831 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
832 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
833 -static-lib*san.
835 PR debug/84252
836 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
837 PARALLEL incoming that failed vt_get_decl_and_offset check.
839 PR middle-end/84237
840 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
841 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
842 TREE_READONLY bit.
843 (get_variable_section): For decls in named .bss* sections pass true as
844 second argument to bss_initializer_p.
846 2018-02-09 Marek Polacek <polacek@redhat.com>
847 Jakub Jelinek <jakub@redhat.com>
849 PR c++/83659
850 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
851 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
852 Sync some changes from cxx_fold_indirect_ref.
854 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
856 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
857 markers.
858 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
859 (BLOCK_INLINE_ENTRY_LABEL): New.
860 (dwarf2out_var_location): Disregard inline entry markers.
861 (inline_entry_data): New struct.
862 (inline_entry_data_hasher): New hashtable type.
863 (inline_entry_data_hasher::hash): New.
864 (inline_entry_data_hasher::equal): New.
865 (inline_entry_data_table): New variable.
866 (add_high_low_attributes): Add DW_AT_entry_pc and
867 DW_AT_GNU_entry_view attributes if a pending entry is found
868 in inline_entry_data_table. Add old entry_pc attribute only
869 if debug nonbinding markers are disabled.
870 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
871 markers are enabled.
872 (block_within_block_p, dwarf2out_inline_entry): New.
873 (dwarf2out_finish): Check that no entries remained in
874 inline_entry_data_table.
875 * final.c (reemit_insn_block_notes): Handle inline entry notes.
876 (final_scan_insn, notice_source_line): Likewise.
877 (rest_of_clean_state): Skip inline entry markers.
878 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
879 markers.
880 * gimple.c (gimple_build_debug_inline_entry): New.
881 * gimple.h (enum gimple_debug_subcode): Add
882 GIMPLE_DEBUG_INLINE_ENTRY.
883 (gimple_build_debug_inline_entry): Declare.
884 (gimple_debug_inline_entry_p): New.
885 (gimple_debug_nonbind_marker_p): Adjust.
886 * insn-notes.def (INLINE_ENTRY): New.
887 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
888 inline entry marker notes.
889 (print_insn): Likewise.
890 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
891 (INSN_DEBUG_MARKER_KIND): Likewise.
892 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
893 * tree-inline.c (expand_call_inline): Build and insert
894 debug_inline_entry stmt.
895 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
896 inline entry blocks early, if nonbind markers are enabled.
897 (dump_scope_block): Dump fragment info.
898 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
899 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
900 (gimple_build_debug_inline_entry): New.
901 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
902 Enable/disable inline entry points too.
903 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
904 (DEBUG_INSN): Describe inline entry markers.
906 * common.opt (gvariable-location-views): New.
907 (gvariable-location-views=incompat5): New.
908 * config.in: Rebuilt.
909 * configure: Rebuilt.
910 * configure.ac: Test assembler for view support.
911 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
912 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
913 * dwarf2out.c (var_loc_view): New typedef.
914 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
915 (dwarf2out_locviews_in_attribute): New.
916 (dwarf2out_locviews_in_loclist): New.
917 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
918 (enum dw_line_info_opcode): Add LI_adv_address.
919 (struct dw_line_info_table): Add view.
920 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
921 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
922 (zero_view_p): New variable.
923 (ZERO_VIEW_P): New macro.
924 (output_asm_line_debug_info): New.
925 (struct var_loc_node): Add view.
926 (add_AT_view_list, AT_loc_list): New.
927 (add_var_loc_to_decl): Add view param. Test it against last.
928 (new_loc_list): Add view params. Record them.
929 (AT_loc_list_ptr): Handle loc and view lists.
930 (view_list_to_loc_list_val_node): New.
931 (print_dw_val): Handle dw_val_class_view_list.
932 (size_of_die): Likewise.
933 (value_format): Likewise.
934 (loc_list_has_views): New.
935 (gen_llsym): Set vl_symbol too.
936 (maybe_gen_llsym, skip_loc_list_entry): New.
937 (dwarf2out_maybe_output_loclist_view_pair): New.
938 (output_loc_list): Output view list or entries too.
939 (output_view_list_offset): New.
940 (output_die): Handle dw_val_class_view_list.
941 (output_dwarf_version): New.
942 (output_compilation_unit_header): Use it.
943 (output_skeleton_debug_sections): Likewise.
944 (output_rnglists, output_line_info): Likewise.
945 (output_pubnames, output_aranges): Update version comments.
946 (output_one_line_info_table): Output view numbers in asm comments.
947 (dw_loc_list): Determine current endview, pass it to new_loc_list.
948 Call maybe_gen_llsym.
949 (loc_list_from_tree_1): Adjust.
950 (add_AT_location_description): Create view list attribute if
951 needed, check it's absent otherwise.
952 (convert_cfa_to_fb_loc_list): Adjust.
953 (maybe_emit_file): Call output_asm_line_debug_info for test.
954 (dwarf2out_var_location): Reset views as needed. Precompute
955 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
956 attribute. Set view.
957 (new_line_info_table): Reset next view.
958 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
959 (dwarf2out_source_line): Likewise. Output view resets and labels to
960 the assembler, or select appropriate line info opcodes.
961 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
962 (optimize_string_length): Catch it. Adjust.
963 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
964 dw_val_class_view_list, and remove it if no longer needed.
965 (hash_loc_list): Hash view numbers.
966 (loc_list_hasher::equal): Compare them.
967 (optimize_location_lists): Check whether a view list symbol is
968 needed, and whether the locview attribute is present, and
969 whether they match. Remove the locview attribute if no longer
970 needed.
971 (index_location_lists): Call skip_loc_list_entry for test.
972 (dwarf2out_finish): Call output_asm_line_debug_info for test.
973 Use output_dwarf_version.
974 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
975 (struct dw_val_node): Add val_view_list.
976 * final.c (SEEN_NEXT_VIEW): New.
977 (set_next_view_needed): New.
978 (clear_next_view_needed): New.
979 (maybe_output_next_view): New.
980 (final_start_function): Rename to...
981 (final_start_function_1): ... this. Take pointer to FIRST,
982 add SEEN parameter. Emit param bindings in the initial view.
983 (final_start_function): Reintroduce SEEN-less interface.
984 (final): Rename to...
985 (final_1): ... this. Take SEEN parameter. Output final pending
986 next view at the end.
987 (final): Reintroduce seen-less interface.
988 (final_scan_insn): Output pending next view before switching
989 sections or ending a block. Mark the next view as needed when
990 outputting variable locations. Notify debug backend of section
991 changes, and of location view changes.
992 (rest_of_handle_final): Adjust.
993 * toplev.c (process_options): Autodetect value for debug variable
994 location views option. Warn on incompat5 without -gdwarf-5.
995 * doc/invoke.texi (gvariable-location-views): New.
996 (gvariable-location-views=incompat5): New.
997 (gno-variable-location-views): New.
999 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1001 PR tree-optimization/84136
1002 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1003 that the result of find_edge is non-NULL.
1005 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1007 PR target/83008
1008 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1009 storing integer register in SImode. Fix cost of 256 and 512
1010 byte aligned SSE register store.
1012 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1014 * config/i386/i386.c (ix86_multiplication_cost): Fix
1015 multiplication cost for TARGET_AVX512DQ.
1017 2018-02-08 Marek Polacek <polacek@redhat.com>
1019 PR tree-optimization/84238
1020 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1021 get_range_strlen.
1023 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1025 PR tree-optimization/84265
1026 * tree-vect-stmts.c (vectorizable_store): Don't treat
1027 VMAT_CONTIGUOUS accesses as grouped.
1028 (vectorizable_load): Likewise.
1030 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1032 PR tree-optimization/81635
1033 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1034 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1035 (test_round_for_mask): New functions.
1036 (wide_int_cc_tests): Call test_round_for_mask.
1037 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1038 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1039 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1040 range returned by get_range_info.
1042 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1044 PR ipa/81360
1045 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1046 * symtab.c: Include builtins.h
1047 (symtab_node::output_to_lto_symbol_table_p): Move here
1048 from lto-streamer-out.c:output_symbol_p.
1049 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1050 (output_symbol_p): Move all logic to symtab.c
1051 (produce_symtab): Update.
1053 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1055 * config/s390/s390-opts.h (enum indirect_branch): Define.
1056 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1057 (s390_indirect_branch_via_thunk)
1058 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1059 (enum s390_indirect_branch_type): Define.
1060 * config/s390/s390.c (struct s390_frame_layout, struct
1061 machine_function): Remove.
1062 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1063 (indirect_branch_table_label_no, indirect_branch_table_name):
1064 Define variables.
1065 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1066 (enum s390_indirect_branch_option): Define.
1067 (s390_return_addr_from_memory): New function.
1068 (s390_handle_string_attribute): New function.
1069 (s390_attribute_table): Add new attribute handler.
1070 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1071 (s390_indirect_branch_via_thunk): New function.
1072 (s390_indirect_branch_via_inline_thunk): New function.
1073 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1074 sibling call optimization for non z10 compiles.
1075 (s390_emit_call): Force indirect branch target to be a single
1076 register. Add r1 clobber for non-z10 compiles.
1077 (s390_emit_epilogue): Emit return jump via return_use expander.
1078 (s390_reorg): Handle JUMP_INSNs as execute targets.
1079 (s390_option_override_internal): Perform validity checks for the
1080 new command line options.
1081 (s390_indirect_branch_attrvalue): New function.
1082 (s390_indirect_branch_settings): New function.
1083 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1084 (s390_output_indirect_thunk_function): New function.
1085 (s390_code_end): Implement target hook.
1086 (s390_case_values_threshold): Implement target hook.
1087 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1088 macros.
1089 * config/s390/s390.h (struct s390_frame_layout)
1090 (struct machine_function): Move here from s390.c.
1091 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1092 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1093 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1094 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1095 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1096 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1097 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1098 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1099 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1100 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1101 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1102 (mnemonic attribute): Add values which aren't recognized
1103 automatically.
1104 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1105 pattern for branch conversion. Fix mnemonic attribute.
1106 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1107 indirect branch via thunk if requested.
1108 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1109 ("*indirect_jump"): Disable for branch conversion using out of
1110 line thunks.
1111 ("indirect_jump_via_thunk<mode>_z10")
1112 ("indirect_jump_via_thunk<mode>")
1113 ("indirect_jump_via_inlinethunk<mode>_z10")
1114 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
1115 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
1116 ("casesi_jump_via_inlinethunk<mode>_z10")
1117 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
1118 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
1119 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
1120 ("*indirect2_jump"): Disable for branch conversion.
1121 ("casesi_jump"): Turn into expander and expand patterns for branch
1122 conversion.
1123 ("return_use"): New expander.
1124 ("*return"): Emit return via thunk and rename it to ...
1125 ("*return<mode>"): ... this one.
1126 * config/s390/s390.opt: Add new options and and enum for the
1127 option values.
1129 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1131 * lra-constraints.c (match_reload): Unconditionally use
1132 gen_lowpart_SUBREG, rather than selecting between that
1133 and equivalent gen_rtx_SUBREG code.
1135 2018-02-08 Richard Biener <rguenther@suse.de>
1137 PR tree-optimization/84233
1138 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
1139 changed flag instead of boguously re-using phi_inserted.
1141 2018-02-08 Martin Jambor <mjambor@suse.cz>
1143 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
1144 static local variables.
1146 2018-02-08 Richard Biener <rguenther@suse.de>
1148 PR tree-optimization/84278
1149 * tree-vect-stmts.c (vectorizable_store): When looking for
1150 smaller vector types to perform grouped strided loads/stores
1151 make sure the mode is supported by the target.
1152 (vectorizable_load): Likewise.
1154 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1156 * config/aarch64/aarch64.c (aarch64_components_for_bb):
1157 Increase LDP/STP opportunities by adding adjacent callee-saves.
1159 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1161 PR rtl-optimization/84068
1162 PR rtl-optimization/83459
1163 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
1165 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
1167 PR tree-optimization/84224
1168 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
1169 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
1170 non-zero arguments.
1172 2018-02-07 Iain Sandoe <iain@codesourcery.com>
1174 PR target/84113
1175 * config/rs6000/altivec.md (*restore_world): Remove LR use.
1176 * config/rs6000/predicates.md (restore_world_operation): Adjust op
1177 count, remove one USE.
1179 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
1181 * doc/install.texi (Configuration): Document the
1182 --with-long-double-format={ibm,ieee} PowerPC configuration
1183 options.
1185 PR target/84154
1186 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
1187 Convert from define_expand to be define_insn_and_split. Rework
1188 float/double/_Float128 conversions to QI/HI/SImode to work with
1189 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
1190 conversions to QI/HImode types did a store and then a load to
1191 truncate the value. For conversions to VSX registers, don't split
1192 the insn, instead emit the code directly. Use the code iterator
1193 any_fix to combine signed and unsigned conversions.
1194 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
1195 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
1196 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1197 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
1198 (fix_<mode>di2_hw): Likewise.
1199 (fixuns_<mode>di2_hw): Likewise.
1200 (fix_<mode>si2_hw): Likewise.
1201 (fixuns_<mode>si2_hw): Likewise.
1202 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
1203 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1204 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
1205 fix<uns>_trunc<SFDF:mode>si2_p8.
1206 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
1207 used.
1208 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
1209 (fix<uns>_<mode>_mem): Likewise.
1210 (fctiw<u>z_<mode>_mem): Likewise.
1211 (fix<uns>_<mode>_mem): Likewise.
1212 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
1213 the register allocator from doing a direct move to the GPRs to do
1214 a store, and instead use the ISA 3.0 store byte/half-word from
1215 vector register instruction. For IEEE 128-bit floating point,
1216 also optimize stores of 32-bit ints.
1217 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
1219 2018-02-07 Alan Hayward <alan.hayward@arm.com>
1221 * genextract.c (push_pathstr_operand): New function to support
1222 [a-zA-Z].
1223 (walk_rtx): Call push_pathstr_operand.
1224 (print_path): Support [a-zA-Z].
1226 2018-02-07 Richard Biener <rguenther@suse.de>
1228 PR tree-optimization/84037
1229 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
1230 (cse_and_gimplify_to_preheader): Declare.
1231 (vect_get_place_in_interleaving_chain): Likewise.
1232 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1233 ivexpr_map.
1234 (_loop_vec_info::~_loop_vec_info): Delete it.
1235 (cse_and_gimplify_to_preheader): New function.
1236 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
1237 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
1238 (vectorizable_load): Likewise. For grouped stores always base
1239 the IV on the first element.
1240 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
1241 condition before gimplifying.
1243 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1245 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
1246 *DIV_EXPR and *MOD_EXPR.
1248 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
1250 PR target/84248
1251 * config/i386/i386.c (ix86_option_override_internal): Mask out
1252 the CF_SET bit when checking -fcf-protection.
1254 2018-02-07 Tom de Vries <tom@codesourcery.com>
1256 PR libgomp/84217
1257 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
1258 enough.
1260 2018-02-07 Richard Biener <rguenther@suse.de>
1262 PR tree-optimization/84204
1263 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
1264 this place.
1266 PR tree-optimization/84205
1267 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
1268 special-case isl_ast_op_zdiv_r.
1270 PR tree-optimization/84223
1271 * graphite-scop-detection.c (gather_bbs::before_dom_children):
1272 Only add conditions from within the region.
1273 (gather_bbs::after_dom_children): Adjust.
1275 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
1277 PR target/84209
1278 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
1279 * config/avr/avr.md: Only post-reload split REG-REG moves if
1280 either register is REGERAL_REG_P.
1282 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1284 PR tree-optimization/84235
1285 * tree-ssa-scopedtables.c
1286 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
1287 if the subtraction is performed in floating point type where NaNs are
1288 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
1289 build 1. Formatting fix.
1291 2018-02-06 Jakub Jelinek <jakub@redhat.com>
1293 PR target/84146
1294 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
1295 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
1296 and skip it regardless of bb boundaries. Use CALL_P macro,
1297 don't test INSN_P (insn) together with CALL_P or JUMP_P check
1298 unnecessarily, formatting fix.
1300 2018-02-06 Michael Collison <michael.collison@arm.com>
1302 * config/arm/thumb2.md:
1303 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
1304 (*thumb_mov_notscc): Ditto.
1306 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
1308 PR target/84154
1309 * config/rs6000/rs6000.md (su code attribute): Use "u" for
1310 unsigned_fix, not "s".
1312 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1314 * configure.ac (gcc_fn_eh_frame_ro): New function.
1315 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
1316 correct .eh_frame permissions.
1317 * configure: Regenerate.
1319 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
1321 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
1322 irrelevant options.
1324 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1326 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1327 Display warning message for -mno-speculate-indirect-jumps.
1329 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
1331 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
1332 Undocumented.
1333 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
1335 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
1337 PR tree-optimization/84225
1338 * tree-eh.c (find_trapping_overflow): Only call
1339 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
1341 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1343 PR target/84145
1344 * config/i386/i386.c: Reimplement the check of possible options
1345 -mibt/-mshstk conbination. Change error messages.
1346 * doc/invoke.texi: Fix a typo: remove extra '='.
1348 2018-02-06 Marek Polacek <polacek@redhat.com>
1350 PR tree-optimization/84228
1351 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
1353 2018-02-06 Tamar Christina <tamar.christina@arm.com>
1355 PR target/82641
1356 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
1357 emitted arch directives.
1358 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
1359 __ARM_FEATURE_COPROC before changing architectures.
1361 2018-02-06 Richard Biener <rguenther@suse.de>
1363 * config/i386/i386.c (print_reg): Fix typo.
1364 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
1366 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
1368 * configure: Regenerate.
1370 2018-02-05 Martin Sebor <msebor@redhat.com>
1372 PR tree-optimization/83369
1373 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
1374 inlining context.
1376 2018-02-05 Martin Liska <mliska@suse.cz>
1378 * doc/invoke.texi: Cherry-pick upstream r323995.
1380 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
1382 * ira.c (ira_init_register_move_cost): Adjust comment.
1384 2018-02-05 Martin Liska <mliska@suse.cz>
1386 PR gcov-profile/84137
1387 * doc/gcov.texi: Fix typo in documentation.
1389 2018-02-05 Martin Liska <mliska@suse.cz>
1391 PR gcov-profile/83879
1392 * doc/gcov.texi: Document necessity of --dynamic-list-data when
1393 using dlopen functionality.
1395 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
1397 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
1398 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
1399 _mm_maskz_range_ss, _mm_mask_range_round_ss,
1400 _mm_maskz_range_round_ss): New intrinsics.
1401 (__builtin_ia32_rangesd128_round)
1402 (__builtin_ia32_rangess128_round): Remove.
1403 (__builtin_ia32_rangesd128_mask_round,
1404 __builtin_ia32_rangess128_mask_round): New builtins.
1405 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
1406 __builtin_ia32_rangess128_round): Remove.
1407 (__builtin_ia32_rangesd128_mask_round,
1408 __builtin_ia32_rangess128_mask_round): New builtins.
1409 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
1410 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
1411 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
1412 "<round_saeonly_constraint>")): Changed to ...
1413 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
1414 "<round_saeonly_scalar_constraint>")): ... this.
1415 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
1416 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
1417 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
1418 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
1419 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
1421 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
1423 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
1424 options.
1425 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
1426 Remove all values except native, 8540 and 8548.
1428 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
1430 * config/i386/i386.c (ix86_output_function_return): Pass
1431 INVALID_REGNUM, instead of -1, as invalid register number to
1432 indirect_thunk_name and output_indirect_thunk.
1434 2018-02-02 Julia Koval <julia.koval@intel.com>
1436 * config.gcc: Add -march=icelake.
1437 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
1438 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
1439 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
1440 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
1441 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
1442 (processor_target_table): Add icelake.
1443 (ix86_option_override_internal): Handle new PTAs.
1444 (get_builtin_code_for_version): Handle icelake.
1445 (M_INTEL_COREI7_ICELAKE): New.
1446 (fold_builtin_cpu): Handle icelake.
1447 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
1448 * doc/invoke.texi: Add -march=icelake.
1450 2018-02-02 Julia Koval <julia.koval@intel.com>
1452 * config/i386/i386.c (ix86_option_override_internal): Change flags type
1453 to wide_int_bitmask.
1454 * wide-int-bitmask.h: New.
1456 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1458 PR target/84066
1459 * config/i386/i386.md: Replace Pmode with word_mode in
1460 builtin_setjmp_setup and builtin_longjmp to support x32.
1462 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
1464 PR target/56010
1465 PR target/83743
1466 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
1467 #include "opts.h".
1468 (rs6000_supported_cpu_names): New static variable.
1469 (linux_cpu_translation_table): Likewise.
1470 (elf_platform) <cpu>: Define new static variable and use it.
1471 Translate kernel AT_PLATFORM name to canonical name if needed.
1472 Error if platform name is unknown.
1474 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
1476 PR target/84089
1477 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
1479 2018-02-01 Jeff Law <law@redhat.com>
1481 PR target/84128
1482 * config/i386/i386.c (release_scratch_register_on_entry): Add new
1483 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
1484 the scratch if RELEASE_VIA_POP is false.
1485 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
1486 If we have to save a temporary register, decrement SIZE appropriately.
1487 Pass new arguments to release_scratch_register_on_entry.
1488 (ix86_adjust_stack_and_probe): Likewise.
1489 (ix86_emit_probe_stack_range): Pass new arguments to
1490 release_scratch_register_on_entry.
1492 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
1494 PR rtl-optimization/84157
1495 * combine.c (change_zero_ext): Use REG_P predicate in
1496 front of HARD_REGISTER_P predicate.
1498 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
1500 * config/avr/avr.c (avr_option_override): Move disabling of
1501 -fdelete-null-pointer-checks to...
1502 * common/config/avr/avr-common.c (avr_option_optimization_table):
1503 ...here.
1505 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1507 PR tree-optimization/81635
1508 * tree-data-ref.c (split_constant_offset_1): For types that
1509 wrap on overflow, try to use range info to prove that wrapping
1510 cannot occur.
1512 2018-02-01 Renlin Li <renlin.li@arm.com>
1514 PR target/83370
1515 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
1516 TAILCALL_ADDR_REGS.
1517 (aarch64_register_move_cost): Likewise.
1518 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
1519 TAILCALL_ADDR_REGS.
1520 (REG_CLASS_NAMES): Likewise.
1521 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
1522 TAILCALL_ADDR_REGS. Remove IP registers.
1523 * config/aarch64/aarch64.md (Ucs): Update register constraint.
1525 2018-02-01 Richard Biener <rguenther@suse.de>
1527 * domwalk.h (dom_walker::dom_walker): Add additional constructor
1528 for specifying RPO order and allow NULL for that.
1529 * domwalk.c (dom_walker::dom_walker): Likewise.
1530 (dom_walker::walk): Handle NULL RPO order.
1531 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
1532 in RPO order.
1533 (rewrite_update_dom_walker): Likewise.
1534 (mark_def_dom_walker): Likewise.
1536 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1538 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
1539 (aarch64_maybe_expand_sve_subreg_move): Declare.
1540 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
1541 * config/aarch64/predicates.md (aarch64_any_register_operand): New
1542 predicate.
1543 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
1544 that are semantically a reverse operation.
1545 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
1546 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
1547 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
1548 functions.
1549 (aarch64_can_change_mode_class): For big-endian, forbid changes
1550 between two SVE modes if they have different element sizes.
1552 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1554 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
1555 the TImode handling for big-endian targets.
1557 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1559 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
1560 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
1561 not just bytes.
1562 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
1563 Remove BSWAP handing for big-endian targets and use the form of
1564 LD1RQ appropariate for the mode.
1566 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1568 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
1569 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
1570 duplicated element.
1572 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1574 PR tearget/83845
1575 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
1576 check for operands that need to go through aarch64_sve_reload_be.
1578 2018-02-01 Jakub Jelinek <jakub@redhat.com>
1580 PR tree-optimization/81661
1581 PR tree-optimization/84117
1582 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
1583 * tree-eh.c: Include gimplify.h.
1584 (find_trapping_overflow, replace_trapping_overflow,
1585 rewrite_to_non_trapping_overflow): New functions.
1586 * tree-vect-loop.c: Include tree-eh.h.
1587 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
1588 * tree-data-ref.c: Include tree-eh.h.
1589 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
1591 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
1593 PR rtl-optimization/84123
1594 * combine.c (change_zero_ext): Check if hard register satisfies
1595 can_change_dest_mode before calling gen_lowpart_SUBREG.
1597 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
1599 PR target/82444
1600 * ira.c (ira_init_register_move_cost): Remove assert.
1602 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
1604 PR rtl-optimization/84071
1605 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
1606 * doc/tm.texi: Regenerate.
1608 2018-01-31 Richard Biener <rguenther@suse.de>
1610 PR tree-optimization/84132
1611 * tree-data-ref.c (analyze_miv_subscript): Properly
1612 check whether evolution_function_is_affine_multivariate_p
1613 before calling gcd_of_steps_may_divide_p.
1615 2018-01-31 Julia Koval <julia.koval@intel.com>
1617 PR target/83618
1618 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
1619 * config/i386/i386.md (rdpid_rex64) New.
1620 (rdpid): Make 32bit only.
1622 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
1624 PR lto/84105
1625 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
1626 an IDENTIFIER_NODE for FUNCTION_TYPE's.
1628 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
1630 Revert
1631 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
1633 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
1635 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
1637 PR rtl-optimization/84071
1638 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
1639 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
1641 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
1643 * config/arc/arc.c (arc_handle_aux_attribute): New function.
1644 (arc_attribute_table): Add 'aux' attribute.
1645 (arc_in_small_data_p): Consider aux like variables.
1646 (arc_is_aux_reg_p): New function.
1647 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
1648 (arc_get_aux_arg): New function.
1649 (prepare_move_operands): Handle aux-register access.
1650 (arc_handle_aux_attribute): New function.
1651 * doc/extend.texi (ARC Variable attributes): Add subsection.
1653 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
1655 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
1656 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
1657 (arc_attribute_table): Add 'uncached' attribute.
1658 (arc_print_operand): Print '.di' flag for uncached memory
1659 accesses.
1660 (arc_in_small_data_p): Do not consider for small data the uncached
1661 types.
1662 (arc_is_uncached_mem_p): New function.
1663 * config/arc/predicates.md (compact_store_memory_operand): Check
1664 for uncached memory accesses.
1665 (nonvol_nonimm_operand): Likewise.
1666 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
1668 2018-01-31 Jakub Jelinek <jakub@redhat.com>
1670 PR c/84100
1671 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
1672 falign-loops=): Add Optimization flag.
1674 2018-01-30 Jeff Law <law@redhat.com>
1676 PR target/84064
1677 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
1678 INT_REGISTERS_SAVED. Check it prior to calling
1679 get_scratch_register_on_entry.
1680 (ix86_adjust_stack_and_probe): Similarly.
1681 (ix86_emit_probe_stack_range): Similarly.
1682 (ix86_expand_prologue): Corresponding changes.
1684 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1686 PR target/40411
1687 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
1688 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
1690 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
1692 PR target/84112
1693 * lra-constraints.c (curr_insn_transform): Process AND in the
1694 address.
1696 2018-01-30 Jakub Jelinek <jakub@redhat.com>
1698 PR rtl-optimization/83986
1699 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
1700 dependence against last_pending_memory_flush in addition to
1701 pending_jump_insns.
1703 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
1705 PR tree-optimization/81611
1706 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
1707 copies.
1709 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1711 PR target/83758
1712 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
1713 a reg rtx.
1715 2018-01-30 Richard Biener <rguenther@suse.de>
1716 Jakub Jelinek <jakub@redhat.com>
1718 PR tree-optimization/84111
1719 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
1720 inner loops added during recursion, as they don't have up-to-date
1721 SSA form.
1723 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
1725 PR ipa/81360
1726 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
1727 (can_inline_edge_by_limits_p): ... here.
1728 (can_early_inline_edge_p, check_callers,
1729 update_caller_keys, update_callee_keys, recursive_inlining,
1730 add_new_edges_to_heap, speculation_useful_p,
1731 inline_small_functions,
1732 inline_small_functions, flatten_function,
1733 inline_to_all_callers_1): Update.
1735 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
1737 * profile-count.c (profile_count::combine_with_ipa_count): Handle
1738 zeros correctly.
1740 2018-01-30 Richard Biener <rguenther@suse.de>
1742 PR tree-optimization/83008
1743 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
1744 invariant and constant vector uses in stmts when they need
1745 more than one stmt.
1747 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1749 PR bootstrap/84017
1750 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
1751 * configure: Regenerate.
1753 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
1755 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
1756 pattern.
1757 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
1758 Use gen_rtx_REG rather than gen_lowpart.
1760 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
1762 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
1763 rather than 0 when creating partial subregs.
1765 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
1767 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
1768 of usage.
1770 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
1772 PR target/81550
1773 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
1774 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
1775 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
1776 flags. This restores the settings used before the 2017-07-24.
1777 Turning off pre increment/decrement/modify allows IVOPTS to
1778 optimize DF/SF loops where the index is an int.
1780 2018-01-29 Richard Biener <rguenther@suse.de>
1781 Kelvin Nilsen <kelvin@gcc.gnu.org>
1783 PR bootstrap/80867
1784 * tree-vect-stmts.c (vectorizable_call): Don't call
1785 targetm.vectorize_builtin_md_vectorized_function if callee is
1786 NULL.
1788 2018-01-22 Carl Love <cel@us.ibm.com>
1790 * doc/extend.tex: Fix typo in second arg in
1791 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
1793 2018-01-29 Richard Biener <rguenther@suse.de>
1795 PR tree-optimization/84086
1796 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
1797 (flush_ssaname_freelist): When SSA names were released reset
1798 the SCEV hash table.
1800 2018-01-29 Richard Biener <rguenther@suse.de>
1802 PR tree-optimization/84057
1803 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
1804 removed paths when removing edges.
1806 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
1808 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
1809 -mfunction-return=@var{choice}.
1811 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
1813 PR diagnostic/84034
1814 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
1815 Handle CR like TAB.
1816 (layout::print_source_line): Likewise.
1817 (test_get_line_width_without_trailing_whitespace): Add test cases.
1819 2018-01-27 Jakub Jelinek <jakub@redhat.com>
1821 PR middle-end/84040
1822 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
1823 debug insns.
1825 2018-01-26 Jim Wilson <jimw@sifive.com>
1827 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
1829 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
1830 specified.
1832 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1834 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
1835 and CMP + SUB-immediate -> SUBS.
1837 2018-01-26 Martin Sebor <msebor@redhat.com>
1839 PR tree-optimization/83896
1840 * tree-ssa-strlen.c (get_string_len): Rename...
1841 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
1842 Avoid assuming length is constant.
1843 (handle_char_store): Use HOST_WIDE_INT for string length.
1845 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
1847 PR target/81763
1848 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
1849 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
1851 2018-01-26 Richard Biener <rguenther@suse.de>
1853 PR rtl-optimization/84003
1854 * dse.c (record_store): Only record redundant stores when
1855 the earlier store aliases at least all accesses the later one does.
1857 2018-01-26 Jakub Jelinek <jakub@redhat.com>
1859 PR rtl-optimization/83985
1860 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
1861 REG_CFA_RESTORE insns.
1862 (delete_unmarked_insns): Don't ignore separate shrink wrapping
1863 REG_CFA_RESTORE insns here.
1865 PR c/83989
1866 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
1867 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
1869 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
1871 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
1872 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
1873 (arc_init): Likewise.
1874 (arc_override_options): Likewise.
1875 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
1876 value.
1877 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
1878 support.
1879 * config/arc/arc.h (TARGET_DBNZ): Define.
1880 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
1881 properly set the tune attribute.
1882 (dbnz): Use TARGET_DBNZ guard.
1883 * config/arc/arc.opt (mtune): Add core3 option.
1885 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
1887 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
1888 recognize new pic like addresses.
1889 (arc_delegitimize_address): Clean up.
1891 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
1893 * config/arc/arc-arches.def: Option mrf16 valid for all
1894 architectures.
1895 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
1896 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
1897 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
1898 * config/arc/arc-tables.opt: Regenerate.
1899 * config/arc/arc.c (arc_conditional_register_usage): Handle
1900 reduced register file case.
1901 (arc_file_start): Set must have build attributes.
1902 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
1903 mrf16 option value.
1904 * config/arc/arc.opt (mrf16): Add new option.
1905 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
1906 * config/arc/genmultilib.awk: Handle new mrf16 option.
1907 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
1908 * config/arc/t-multilib: Regenerate.
1909 * doc/invoke.texi (ARC Options): Document mrf16 option.
1911 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
1913 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
1914 * config/arc/arc.c (arc_handle_secure_attribute): New function.
1915 (arc_attribute_table): Add 'secure_call' attribute.
1916 (arc_print_operand): Print secure call operand.
1917 (arc_function_ok_for_sibcall): Don't optimize tail calls when
1918 secure.
1919 (arc_is_secure_call_p): New function. * config/arc/arc.md
1920 (call_i): Add support for sjli instruction.
1921 (call_value_i): Likewise.
1922 * config/arc/constraints.md (Csc): New constraint.
1924 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
1925 John Eric Martin <John.Martin@emmicro-us.com>
1927 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
1928 * config/arc/arc.c (_arc_jli_section): New struct.
1929 (arc_jli_section): New type.
1930 (rc_jli_sections): New static variable.
1931 (arc_handle_jli_attribute): New function.
1932 (arc_attribute_table): Add jli_always and jli_fixed attribute.
1933 (arc_file_end): New function.
1934 (TARGET_ASM_FILE_END): Define.
1935 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
1936 (arc_add_jli_section): New function.
1937 (jli_call_scan): Likewise.
1938 (arc_reorg): Call jli_call_scan.
1939 (arc_output_addsi): Remove 'S' from printing asm operand.
1940 (arc_is_jli_call_p): New function.
1941 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
1942 operand.
1943 (movhi_insn): Likewise.
1944 (movsi_insn): Likewise.
1945 (movsi_set_cc_insn): Likewise.
1946 (loadqi_update): Likewise.
1947 (load_zeroextendqisi_update): Likewise.
1948 (load_signextendqisi_update): Likewise.
1949 (loadhi_update): Likewise.
1950 (load_zeroextendhisi_update): Likewise.
1951 (load_signextendhisi_update): Likewise.
1952 (loadsi_update): Likewise.
1953 (loadsf_update): Likewise.
1954 (movsicc_insn): Likewise.
1955 (bset_insn): Likewise.
1956 (bxor_insn): Likewise.
1957 (bclr_insn): Likewise.
1958 (bmsk_insn): Likewise.
1959 (bicsi3_insn): Likewise.
1960 (cmpsi_cc_c_insn): Likewise.
1961 (movsi_ne): Likewise.
1962 (movsi_cond_exec): Likewise.
1963 (clrsbsi2): Likewise.
1964 (norm_f): Likewise.
1965 (normw): Likewise.
1966 (swap): Likewise.
1967 (divaw): Likewise.
1968 (flag): Likewise.
1969 (sr): Likewise.
1970 (kflag): Likewise.
1971 (ffs): Likewise.
1972 (ffs_f): Likewise.
1973 (fls): Likewise.
1974 (call_i): Remove 'S' asm letter, add jli instruction.
1975 (call_value_i): Likewise.
1976 * config/arc/arc.op (mjli-always): New option.
1977 * config/arc/constraints.md (Cji): New constraint.
1978 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
1979 operand.
1980 (subsf3_fpx): Likewise.
1981 (mulsf3_fpx): Likewise.
1982 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
1983 asm operand.
1984 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
1985 function attrbutes.
1986 * doc/invoke.texi (ARC): Document mjli-always option.
1988 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
1990 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
1991 avoid addition with 0 and use incw and decw where possible.
1993 2018-01-26 Richard Biener <rguenther@suse.de>
1995 PR tree-optimization/81082
1996 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
1997 association if it requires casting to unsigned.
1998 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
1999 from fold_plusminus_mult_expr to catch important cases late when
2000 range info is available.
2002 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2004 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2005 * configure.ac (hidden_linkonce): New test.
2006 * configure: Regenerate.
2007 * config.in: Regenerate.
2009 2018-01-26 Julia Koval <julia.koval@intel.com>
2011 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2012 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2013 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2014 _mm_mask_bitshuffle_epi64_mask): Fix type.
2015 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2016 USI_FTYPE_V4DI_V4DI_USI): Remove.
2017 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2018 __builtin_ia32_vpshufbitqmb256_mask,
2019 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2020 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2021 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2023 2018-01-26 Alan Modra <amodra@gmail.com>
2025 PR target/84033
2026 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2027 UNSPEC_VBPERMQ. Sort other unspecs.
2029 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2031 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2033 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2035 PR middle-end/83055
2036 * predict.c (drop_profile): Do not push/pop cfun; update also
2037 node->count.
2038 (handle_missing_profiles): Fix logic looking for zero profiles.
2040 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2042 PR middle-end/83977
2043 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2044 on functions with #pragma omp declare simd or functions with simd
2045 attribute.
2046 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2047 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2048 Remove trailing \n from warning_at calls.
2050 2018-01-25 Tom de Vries <tom@codesourcery.com>
2052 PR target/84028
2053 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2054 for neutered workers.
2056 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2058 PR target/68467
2059 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2060 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2062 2018-01-24 Jeff Law <law@redhat.com>
2064 PR target/83994
2065 * i386.c (get_probe_interval): Move to earlier point.
2066 (ix86_compute_frame_layout): If -fstack-clash-protection and
2067 the frame is larger than the probe interval, then use pushes
2068 to save registers rather than reg->mem moves.
2069 (ix86_expand_prologue): Remove conditional for int_registers_saved
2070 assertion.
2072 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2074 PR target/84014
2075 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2076 min/max for never referenced object.
2078 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2080 PR middle-end/83977
2081 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2082 here.
2083 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2084 attributes from DECL_ATTRIBUTES (decl) without affecting
2085 DECL_ATTRIBUTES (current_function_decl).
2086 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2087 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2089 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2091 PR tree-optimization/83979
2092 * fold-const.c (fold_comparison): Use constant_boolean_node
2093 instead of boolean_{true,false}_node.
2095 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2097 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2098 with zero counts.
2100 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2102 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2103 Simplify the clause that sets the length attribute.
2104 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2105 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2106 clause that sets the length attribute.
2107 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2109 2018-01-24 Tom de Vries <tom@codesourcery.com>
2111 PR target/83589
2112 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
2113 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
2114 Add strict parameter.
2115 (prevent_branch_around_nothing): Insert dummy insn between branch to
2116 label and label with no ptx insn inbetween.
2117 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
2119 2018-01-24 Tom de Vries <tom@codesourcery.com>
2121 PR target/81352
2122 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2123 for neutered threads in warp.
2124 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
2126 2018-01-24 Richard Biener <rguenther@suse.de>
2128 PR tree-optimization/83176
2129 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
2130 operands.
2132 2018-01-24 Richard Biener <rguenther@suse.de>
2134 PR tree-optimization/82819
2135 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
2136 code generating pluses that are no-ops in the target precision.
2138 2018-01-24 Richard Biener <rguenther@suse.de>
2140 PR middle-end/84000
2141 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
2143 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2145 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
2146 to merge probabilities.
2147 * predict.c (probably_never_executed): Also mark as cold functions
2148 with global 0 profile and guessed local profile.
2149 * profile-count.c (profile_probability::combine_with_count): New
2150 member function.
2151 * profile-count.h (profile_probability::operator*,
2152 profile_probability::operator*=, profile_probability::operator/,
2153 profile_probability::operator/=): Reduce precision to adjusted
2154 and set value to guessed on contradictory divisions.
2155 (profile_probability::combine_with_freq): Remove.
2156 (profile_probability::combine_wiht_count): Declare.
2157 (profile_count::force_nonzero):: Set to adjusted.
2158 (profile_count::probability_in):: Set quality to adjusted.
2159 * tree-ssa-tail-merge.c (replace_block_by): Use
2160 combine_with_count.
2162 2018-01-23 Andrew Waterman <andrew@sifive.com>
2163 Jim Wilson <jimw@sifive.com>
2165 * config/riscv/riscv.c (riscv_stack_boundary): New.
2166 (riscv_option_override): Set riscv_stack_boundary. Handle
2167 riscv_preferred_stack_boundary_arg.
2168 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
2169 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
2170 (STACK_BOUNDARY): Set to riscv_stack_boundary.
2171 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
2172 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
2173 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
2175 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
2177 PR target/83905
2178 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
2179 of struct ix86_frame.
2180 (ix86_expand_epilogue): Likewise. Add a local variable for
2181 the reg_save_offset field in struct ix86_frame.
2183 2018-01-23 Bin Cheng <bin.cheng@arm.com>
2185 PR tree-optimization/82604
2186 * tree-loop-distribution.c (enum partition_kind): New enum item
2187 PKIND_PARTIAL_MEMSET.
2188 (partition_builtin_p): Support above new enum item.
2189 (generate_code_for_partition): Ditto.
2190 (compute_access_range): Differentiate cases that equality can be
2191 proven at all loops, the innermost loops or no loops.
2192 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
2193 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
2194 (finalize_partitions, distribute_loop): Don't fuse partition of
2195 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
2196 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
2197 parloop is enabled.
2199 2018-01-23 Martin Liska <mliska@suse.cz>
2201 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
2202 order to ignore the predictor.
2203 (PRED_POLYMORPHIC_CALL): Likewise.
2204 (PRED_RECURSIVE_CALL): Likewise.
2206 2018-01-23 Martin Liska <mliska@suse.cz>
2208 * tree-profile.c (tree_profiling): Print function header to
2209 aware reader which function we are working on.
2210 * value-prof.c (gimple_find_values_to_profile): Do not print
2211 not interesting value histograms.
2213 2018-01-23 Martin Liska <mliska@suse.cz>
2215 * profile-count.h (enum profile_quality): Add
2216 profile_uninitialized as the first value. Do not number values
2217 as they are zero based.
2218 (profile_count::verify): Update sanity check.
2219 (profile_probability::verify): Likewise.
2221 2018-01-23 Nathan Sidwell <nathan@acm.org>
2223 * doc/invoke.texi (ffor-scope): Deprecate.
2225 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2227 PR tree-optimization/83510
2228 * domwalk.c (set_all_edges_as_executable): New function.
2229 (dom_walker::dom_walker): Convert bool param
2230 "skip_unreachable_blocks" to enum reachability. Move setup of
2231 edge flags to set_all_edges_as_executable and only do it when
2232 reachability is REACHABLE_BLOCKS.
2233 * domwalk.h (enum dom_walker::reachability): New enum.
2234 (dom_walker::dom_walker): Convert bool param
2235 "skip_unreachable_blocks" to enum reachability.
2236 (set_all_edges_as_executable): New decl.
2237 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
2238 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
2239 "reachability".
2240 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
2241 but converting true to REACHABLE_BLOCKS.
2242 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
2243 * tree-vrp.c
2244 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
2245 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
2246 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
2247 REACHABLE_BLOCKS.
2248 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
2249 if check_all_array_refs will be called.
2251 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2253 * tree.c (selftest::test_location_wrappers): Add more test
2254 coverage.
2256 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2258 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
2259 (selftest::test_bit_in_range): Likewise.
2261 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2263 PR testsuite/83888
2264 * doc/sourcebuild.texi (vect_float): Say that the selector
2265 only describes the situation when -funsafe-math-optimizations is on.
2266 (vect_float_strict): Document.
2268 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2270 PR tree-optimization/83965
2271 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
2272 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
2273 instead of checking only for a reduction.
2274 (vect_recog_widen_sum_pattern): Likewise.
2276 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2278 * predict.c (probably_never_executed): Only use precise profile info.
2279 (compute_function_frequency): Skip after inlining hack since we now
2280 have quality checking.
2282 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2284 * profile-count.h (profile_probability::very_unlikely,
2285 profile_probability::unlikely, profile_probability::even): Set
2286 precision to guessed.
2288 2018-01-23 Richard Biener <rguenther@suse.de>
2290 PR tree-optimization/83963
2291 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
2292 Properly terminate dominator walk when crossing the exit edge not
2293 when visiting its source block.
2295 2018-01-23 Jakub Jelinek <jakub@redhat.com>
2297 PR c++/83918
2298 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
2299 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
2301 2018-01-22 Jakub Jelinek <jakub@redhat.com>
2303 PR tree-optimization/83957
2304 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
2305 semicolon after for body surrounded by braces.
2307 PR tree-optimization/83081
2308 * profile-count.h (profile_probability::split): New method.
2309 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
2310 Use profile_probability::split.
2311 (do_compare_rtx_and_jump): Fix adjustment of probabilities
2312 when splitting a single conditional jump into 2.
2314 2018-01-22 David Malcolm <dmalcolm@redhat.com>
2316 PR tree-optimization/69452
2317 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
2318 decl.
2320 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2322 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
2323 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
2324 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
2326 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2328 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
2329 * config/rl78/rl78.md (movdi): New define_expand.
2330 * config/rl78/rl78.c (rl78_split_movdi): New function.
2332 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
2334 PR target/83862
2335 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
2336 no longer used.
2337 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
2338 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
2339 128-bit to produce an UNSPEC move to get the double word with the
2340 signbit and then a shift directly to do signbit.
2341 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
2342 implementation with a new version that just does either a direct
2343 move or a regular move. Move memory interface to separate insns.
2344 Move insns so they are next to the expander.
2345 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
2346 with signbit move. Split big and little endian case.
2347 (signbit<mode>2_dm_mem_le): Likewise.
2348 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
2349 (signbit<mode>2_dm2): Likewise.
2351 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2353 * config/rl78/rl78.md (anddi3): New define_expand.
2355 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2357 * config/rl78/rl78.md (umindi3): New define_expand.
2359 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2361 * config/rl78/rl78.md (smindi3): New define_expand.
2363 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2365 * config/rl78/rl78.md (smaxdi3): New define_expand.
2367 2018-01-22 Carl Love <cel@us.ibm.com>
2369 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
2370 LVX_V1TI): Add macro expansion.
2371 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
2372 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
2373 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
2374 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
2375 Change check to determine if the instruction is a byte reversing
2376 entry. Fix typo in comment.
2377 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
2378 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
2379 Add def_builtin calls for new builtins.
2380 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
2381 Add define_insn expansion.
2383 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2385 * config/rl78/rl78.md (umaxdi3): New define_expand.
2387 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2389 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
2390 for non-QImode registers.
2392 2018-01-22 Richard Biener <rguenther@suse.de>
2394 PR tree-optimization/83963
2395 * graphite-scop-detection.c (scop_detection::get_sese): Delay
2396 including the loop exit block.
2397 (scop_detection::merge_sese): Likewise.
2398 (scop_detection::add_scop): Do it here instead.
2400 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2402 * doc/sourcebuild.texi (arm_softfloat): Document.
2404 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
2406 PR gcc/77734
2407 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
2408 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
2409 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
2411 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2412 David Edelsohn <dje.gcc@gmail.com>
2414 PR target/83946
2415 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2416 Change "crset eq" to "crset 2".
2417 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2418 (*call_indirect_aix<mode>_nospec): Likewise.
2419 (*call_value_indirect_aix<mode>_nospec): Likewise.
2420 (*call_indirect_elfv2<mode>_nospec): Likewise.
2421 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
2422 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
2423 change assembly output from . to $.
2424 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2425 (indirect_jump<mode>_nospec): Change assembly output from . to $.
2426 (*tablejump<mode>_internal1_nospec): Likewise.
2428 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
2430 PR target/80870
2431 * config/sh/sh_optimize_sett_clrt.cc:
2432 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
2434 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
2436 PR tree-optimization/83940
2437 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
2438 offset_dt to vect_constant_def rather than vect_unknown_def_type.
2439 (vect_check_load_store_mask): Add a mask_dt_out parameter and
2440 use it to pass back the definition type.
2441 (vect_check_store_rhs): Likewise rhs_dt_out.
2442 (vect_build_gather_load_calls): Add a mask_dt argument and use
2443 it instead of a call to vect_is_simple_use.
2444 (vectorizable_store): Update calls to vect_check_load_store_mask
2445 and vect_check_store_rhs. Use the dt returned by the latter instead
2446 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
2447 instead of calls to vect_is_simple_use. Pass the scalar rather
2448 than the vector operand to vect_is_simple_use when handling
2449 second and subsequent copies of an rhs value.
2450 (vectorizable_load): Update calls to vect_check_load_store_mask
2451 and vect_build_gather_load_calls. Use the cached mask_dt and
2452 gs_info.offset_dt instead of calls to vect_is_simple_use.
2454 2018-01-20 Jakub Jelinek <jakub@redhat.com>
2456 PR middle-end/83945
2457 * tree-emutls.c: Include gimplify.h.
2458 (lower_emutls_2): New function.
2459 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
2460 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
2461 it before further processing.
2463 PR target/83930
2464 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
2465 UINTVAL (trueop1) instead of INTVAL (op1).
2467 2018-01-19 Jakub Jelinek <jakub@redhat.com>
2469 PR debug/81570
2470 PR debug/83728
2471 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
2472 INCOMING_FRAME_SP_OFFSET if not defined.
2473 (scan_trace): Add ENTRY argument. If true and
2474 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
2475 emit a note to adjust the CFA offset.
2476 (create_cfi_notes): Adjust scan_trace callers.
2477 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
2478 INCOMING_FRAME_SP_OFFSET in the CIE.
2479 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
2480 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
2481 Likewise.
2482 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
2483 * doc/tm.texi: Regenerated.
2485 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2487 PR rtl-optimization/83147
2488 * lra-constraints.c (remove_inheritance_pseudos): Use
2489 lra_substitute_pseudo_within_insn.
2491 2018-01-19 Tom de Vries <tom@codesourcery.com>
2492 Cesar Philippidis <cesar@codesourcery.com>
2494 PR target/83920
2495 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
2497 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
2499 PR target/83790
2500 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
2501 spaces for function labels.
2503 2018-01-19 Martin Liska <mliska@suse.cz>
2505 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
2506 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
2507 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
2508 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
2509 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
2510 (PRED_CONST_RETURN): Change from 69 to 65.
2511 (PRED_NULL_RETURN): Change from 91 to 71.
2512 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
2513 (PRED_LOOP_GUARD): Change from 66 to 73.
2515 2018-01-19 Martin Liska <mliska@suse.cz>
2517 * predict.c (predict_insn_def): Add new assert.
2518 (struct branch_predictor): Change type to signed integer.
2519 (test_prediction_value_range): Amend test to cover
2520 PROB_UNINITIALIZED.
2521 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
2522 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
2523 (PRED_LOOP_ITERATIONS_MAX): Likewise.
2524 (PRED_LOOP_IV_COMPARE): Likewise.
2525 * predict.h (PROB_UNINITIALIZED): Define new constant.
2527 2018-01-19 Martin Liska <mliska@suse.cz>
2529 * predict.c (dump_prediction): Add new format for
2530 analyze_brprob.py script which is enabled with -details
2531 suboption.
2532 * profile-count.h (precise_p): New function.
2534 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
2536 PR tree-optimization/83922
2537 * tree-vect-loop.c (vect_verify_full_masking): Return false if
2538 there are no statements that need masking.
2539 (vect_active_double_reduction_p): New function.
2540 (vect_analyze_loop_operations): Use it when handling phis that
2541 are not in the loop header.
2543 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
2545 PR tree-optimization/83914
2546 * tree-vect-loop.c (vectorizable_induction): Don't convert
2547 init_expr or apply the peeling adjustment for inductions
2548 that are nested within the vectorized loop.
2550 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2552 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
2553 instead of NEG.
2555 2018-01-18 Jakub Jelinek <jakub@redhat.com>
2557 PR sanitizer/81715
2558 PR testsuite/83882
2559 * function.h (gimplify_parameters): Add gimple_seq * argument.
2560 * function.c: Include gimple.h and options.h.
2561 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
2562 for the added local temporaries if needed.
2563 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
2564 if there are any parameter cleanups, wrap whole body into a
2565 try/finally with the cleanups.
2567 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
2569 PR target/82964
2570 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
2571 Use GET_MODE_CLASS for scalar floating point.
2573 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
2575 PR ipa/82256
2576 patch by PaX Team
2577 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
2578 Fix call of call_cgraph_insertion_hooks.
2580 2018-01-18 Martin Sebor <msebor@redhat.com>
2582 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
2584 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
2586 PR ipa/83619
2587 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
2588 frequencies.
2590 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
2592 PR other/70268
2593 * common.opt: (-ffile-prefix-map): New option.
2594 * opts.c (common_handle_option): Defer it.
2595 * opts-global.c (handle_common_deferred_options): Handle it.
2596 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
2597 * file-prefix-map.h: New file.
2598 (remap_debug_filename, add_debug_prefix_map): ...here.
2599 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
2600 * final.c (debug_prefix_map, add_debug_prefix_map
2601 remap_debug_filename): Move to...
2602 * file-prefix-map.c: New file.
2603 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
2604 generalize, get rid of alloca(), use strrchr() instead of strchr().
2605 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
2606 Implement in terms of add_prefix_map().
2607 (remap_macro_filename, remap_debug_filename): Implement in term of
2608 remap_filename().
2609 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
2610 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
2611 * dbxout.c: Include file-prefix-map.h.
2612 * varasm.c: Likewise.
2613 * vmsdbgout.c: Likewise.
2614 * xcoffout.c: Likewise.
2615 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
2616 * doc/cppopts.texi (-fmacro-prefix-map): Document.
2617 * doc/invoke.texi (-ffile-prefix-map): Document.
2618 (-fdebug-prefix-map): Update description.
2620 2018-01-18 Martin Liska <mliska@suse.cz>
2622 * config/i386/i386.c (indirect_thunk_name): Document that also
2623 lfence is emitted.
2624 (output_indirect_thunk): Document why both instructions
2625 (pause and lfence) are generated.
2627 2018-01-18 Richard Biener <rguenther@suse.de>
2629 PR tree-optimization/83887
2630 * graphite-scop-detection.c
2631 (scop_detection::get_nearest_dom_with_single_entry): Remove.
2632 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
2633 (scop_detection::merge_sese): Re-implement with a flood-fill
2634 algorithm that properly finds a SESE region if it exists.
2636 2018-01-18 Jakub Jelinek <jakub@redhat.com>
2638 PR c/61240
2639 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
2640 pointer_diff optimizations use view_convert instead of convert.
2642 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2644 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2645 Generate different code for -mno-speculate-indirect-jumps.
2646 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2647 (*call_indirect_aix<mode>): Disable for
2648 -mno-speculate-indirect-jumps.
2649 (*call_indirect_aix<mode>_nospec): New define_insn.
2650 (*call_value_indirect_aix<mode>): Disable for
2651 -mno-speculate-indirect-jumps.
2652 (*call_value_indirect_aix<mode>_nospec): New define_insn.
2653 (*sibcall_nonlocal_sysv<mode>): Generate different code for
2654 -mno-speculate-indirect-jumps.
2655 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2657 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2659 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
2660 long double type, set the flags for noting the default long double
2661 type, even if we don't pass or return a long double type.
2663 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
2665 PR ipa/83051
2666 * ipa-inline.c (flatten_function): Do not overwrite final inlining
2667 failure.
2669 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
2671 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
2672 support for merge[hl].
2673 (fold_mergehl_helper): New helper function.
2674 (tree-vector-builder.h): New #include for tree_vector_builder usage.
2675 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
2676 (altivec_vmrglw_direct): Add xxmrglw insn.
2678 2018-01-17 Andrew Waterman <andrew@sifive.com>
2680 * config/riscv/riscv.c (riscv_conditional_register_usage): If
2681 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
2683 2018-01-17 David Malcolm <dmalcolm@redhat.com>
2685 PR lto/83121
2686 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
2687 call the lto_location_cache before reading the
2688 DECL_SOURCE_LOCATION of the types.
2690 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2691 Richard Sandiford <richard.sandiford@linaro.org>
2693 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
2694 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
2695 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
2696 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
2697 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
2698 Add declaration.
2699 * config/aarch64/constraints.md (aarch64_movti_operand):
2700 Limit immediates.
2701 * config/aarch64/predicates.md (Uti): Add new constraint.
2703 2018-01-17 Carl Love <cel@us.ibm.com>
2705 * config/rs6000/vsx.md (define_expand xl_len_r,
2706 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
2707 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
2708 lxvll.
2709 (define_expand, define_insn): Move the shift left from the
2710 define_insn to the define_expand for lxvl and stxvl instructions.
2711 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
2712 and XL_LEN_R definitions to PURE.
2714 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
2716 * config/i386/i386.c (indirect_thunk_name): Declare regno
2717 as unsigned int. Compare regno with INVALID_REGNUM.
2718 (output_indirect_thunk): Ditto.
2719 (output_indirect_thunk_function): Ditto.
2720 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
2721 in the call to output_indirect_thunk_function.
2723 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
2725 PR middle-end/83884
2726 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
2727 rather than the size of inner_type to determine the stack slot size
2728 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
2730 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
2732 PR target/83546
2733 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
2734 to PTA_SILVERMONT.
2736 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
2738 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
2739 endian Linux systems to optionally enable multilibs for selecting
2740 the long double type if the user configured an explicit type.
2741 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
2742 have no long double multilibs if not defined.
2743 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
2744 warn if the user used -mabi={ieee,ibm}longdouble and we built
2745 multilibs for long double.
2746 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
2747 appropriate multilib option.
2748 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
2749 multilib options.
2750 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
2751 for building long double multilibs.
2752 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
2754 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
2756 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
2757 copies.
2759 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
2760 64 bits.
2761 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
2762 128 bits.
2764 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
2765 variables.
2767 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
2768 return value.
2770 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
2772 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
2773 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
2775 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
2777 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
2778 different rtl trees depending on TARGET_64BIT.
2779 (rs6000_gen_lvx): Likewise.
2781 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
2783 * config/visium/visium.md (nop): Tweak comment.
2784 (hazard_nop): Likewise.
2786 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2788 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
2789 -mspeculate-indirect-jumps.
2790 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
2791 for -mno-speculate-indirect-jumps.
2792 (*call_indirect_elfv2<mode>_nospec): New define_insn.
2793 (*call_value_indirect_elfv2<mode>): Disable for
2794 -mno-speculate-indirect-jumps.
2795 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
2796 (indirect_jump): Emit different RTL for
2797 -mno-speculate-indirect-jumps.
2798 (*indirect_jump<mode>): Disable for
2799 -mno-speculate-indirect-jumps.
2800 (*indirect_jump<mode>_nospec): New define_insn.
2801 (tablejump): Emit different RTL for
2802 -mno-speculate-indirect-jumps.
2803 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
2804 (tablejumpsi_nospec): New define_expand.
2805 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
2806 (tablejumpdi_nospec): New define_expand.
2807 (*tablejump<mode>_internal1): Disable for
2808 -mno-speculate-indirect-jumps.
2809 (*tablejump<mode>_internal1_nospec): New define_insn.
2810 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
2811 option.
2813 2018-01-16 Artyom Skrobov tyomitch@gmail.com
2815 * caller-save.c (insert_save): Drop unnecessary parameter. All
2816 callers updated.
2818 2018-01-16 Jakub Jelinek <jakub@redhat.com>
2819 Richard Biener <rguenth@suse.de>
2821 PR libgomp/83590
2822 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
2823 return early, inline manually is_gimple_sizepos. Make sure if we
2824 call gimplify_expr we don't end up with a gimple constant.
2825 * tree.c (variably_modified_type_p): Don't return true for
2826 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
2827 * gimplify.h (is_gimple_sizepos): Remove.
2829 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
2831 PR tree-optimization/83857
2832 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
2833 vectorizable_live_operation for pure SLP statements.
2834 (vectorizable_live_operation): Handle PHIs.
2836 2018-01-16 Richard Biener <rguenther@suse.de>
2838 PR tree-optimization/83867
2839 * tree-vect-stmts.c (vect_transform_stmt): Precompute
2840 nested_in_vect_loop_p since the scalar stmt may get invalidated.
2842 2018-01-16 Jakub Jelinek <jakub@redhat.com>
2844 PR c/83844
2845 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
2846 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
2847 If off is not INTEGER_CST, issue a may not be aligned warning
2848 rather than isn't aligned. Use isn%'t rather than isn't.
2849 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
2850 into MULT_EXPR.
2851 <case MULT_EXPR>: Improve the case when bottom and one of the
2852 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
2853 operand, in that case check if the other operand is multiple of
2854 bottom divided by the INTEGER_CST operand.
2856 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
2858 PR target/83858
2859 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
2860 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
2861 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
2862 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
2863 * config/pa/pa.c (pa_function_arg_advance): Likewise.
2864 (pa_function_arg, pa_arg_partial_bytes): Likewise.
2865 (pa_function_arg_size): New function.
2867 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
2869 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
2870 in a separate statement.
2872 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
2874 PR tree-optimization/83847
2875 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
2876 group gathers and scatters.
2878 2018-01-16 Jakub Jelinek <jakub@redhat.com>
2880 PR rtl-optimization/86620
2881 * params.def (max-sched-ready-insns): Bump minimum value to 1.
2883 PR rtl-optimization/83213
2884 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
2885 to last if both are JUMP_INSNs.
2887 PR tree-optimization/83843
2888 * gimple-ssa-store-merging.c
2889 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
2890 store_immediate_info for bswap/nop orig_stores.
2892 2018-01-15 Andrew Waterman <andrew@sifive.com>
2894 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
2895 !TARGET_MUL.
2896 <UDIV>: Increase cost if !TARGET_DIV.
2898 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
2900 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
2901 (define_attr "cr_logical_3op"): New.
2902 (cceq_ior_compare): Adjust.
2903 (cceq_ior_compare_complement): Adjust.
2904 (*cceq_rev_compare): Adjust.
2905 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
2906 (is_cracked_insn): Adjust.
2907 (insn_must_be_first_in_group): Adjust.
2908 * config/rs6000/40x.md: Adjust.
2909 * config/rs6000/440.md: Adjust.
2910 * config/rs6000/476.md: Adjust.
2911 * config/rs6000/601.md: Adjust.
2912 * config/rs6000/603.md: Adjust.
2913 * config/rs6000/6xx.md: Adjust.
2914 * config/rs6000/7450.md: Adjust.
2915 * config/rs6000/7xx.md: Adjust.
2916 * config/rs6000/8540.md: Adjust.
2917 * config/rs6000/cell.md: Adjust.
2918 * config/rs6000/e300c2c3.md: Adjust.
2919 * config/rs6000/e500mc.md: Adjust.
2920 * config/rs6000/e500mc64.md: Adjust.
2921 * config/rs6000/e5500.md: Adjust.
2922 * config/rs6000/e6500.md: Adjust.
2923 * config/rs6000/mpc.md: Adjust.
2924 * config/rs6000/power4.md: Adjust.
2925 * config/rs6000/power5.md: Adjust.
2926 * config/rs6000/power6.md: Adjust.
2927 * config/rs6000/power7.md: Adjust.
2928 * config/rs6000/power8.md: Adjust.
2929 * config/rs6000/power9.md: Adjust.
2930 * config/rs6000/rs64.md: Adjust.
2931 * config/rs6000/titan.md: Adjust.
2933 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
2935 * config/i386/predicates.md (indirect_branch_operand): Rewrite
2936 ix86_indirect_branch_register logic.
2938 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
2940 * config/i386/constraints.md (Bs): Update
2941 ix86_indirect_branch_register check. Don't check
2942 ix86_indirect_branch_register with GOT_memory_operand.
2943 (Bw): Likewise.
2944 * config/i386/predicates.md (GOT_memory_operand): Don't check
2945 ix86_indirect_branch_register here.
2946 (GOT32_symbol_operand): Likewise.
2948 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
2950 * config/i386/predicates.md (constant_call_address_operand):
2951 Rewrite ix86_indirect_branch_register logic.
2952 (sibcall_insn_operand): Likewise.
2954 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
2956 * config/i386/constraints.md (Bs): Replace
2957 ix86_indirect_branch_thunk_register with
2958 ix86_indirect_branch_register.
2959 (Bw): Likewise.
2960 * config/i386/i386.md (indirect_jump): Likewise.
2961 (tablejump): Likewise.
2962 (*sibcall_memory): Likewise.
2963 (*sibcall_value_memory): Likewise.
2964 Peepholes of indirect call and jump via memory: Likewise.
2965 * config/i386/i386.opt: Likewise.
2966 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2967 (GOT_memory_operand): Likewise.
2968 (call_insn_operand): Likewise.
2969 (sibcall_insn_operand): Likewise.
2970 (GOT32_symbol_operand): Likewise.
2972 2018-01-15 Jakub Jelinek <jakub@redhat.com>
2974 PR middle-end/83837
2975 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
2976 type rather than type addr's type points to.
2977 (expand_omp_atomic_mutex): Likewise.
2978 (expand_omp_atomic): Likewise.
2980 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
2982 PR target/83839
2983 * config/i386/i386.c (output_indirect_thunk_function): Use
2984 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
2985 for __x86_return_thunk.
2987 2018-01-15 Richard Biener <rguenther@suse.de>
2989 PR middle-end/83850
2990 * expmed.c (extract_bit_field_1): Fix typo.
2992 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2994 PR target/83687
2995 * config/arm/iterators.md (VF): New mode iterator.
2996 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
2997 Remove integer-related logic from pattern.
2998 (neon_vabd<mode>_3): Likewise.
3000 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3002 PR middle-end/82694
3003 * common.opt (fstrict-overflow): No longer an alias.
3004 (fwrapv-pointer): New option.
3005 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3006 also for pointer types based on flag_wrapv_pointer.
3007 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3008 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3009 opts->x_flag_wrapv got set.
3010 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3011 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3012 POINTER_TYPE_OVERFLOW_UNDEFINED.
3013 * match.pd: Likewise in address comparison pattern.
3014 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3016 2018-01-15 Richard Biener <rguenther@suse.de>
3018 PR lto/83804
3019 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3020 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3021 Reset type names to their identifier if their TYPE_DECL doesn't
3022 have linkage (and thus is used for ODR and devirt).
3023 (save_debug_info_for_decl): Remove.
3024 (save_debug_info_for_type): Likewise.
3025 (add_tree_to_fld_list): Adjust.
3026 * tree-pretty-print.c (dump_generic_node): Make dumping of
3027 type names more robust.
3029 2018-01-15 Richard Biener <rguenther@suse.de>
3031 * BASE-VER: Bump to 8.0.1.
3033 2018-01-14 Martin Sebor <msebor@redhat.com>
3035 PR other/83508
3036 * builtins.c (check_access): Avoid warning when the no-warning bit
3037 is set.
3039 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3041 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3042 * ira-color (allocno_hard_regs_compare): Likewise.
3044 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3046 PR target/83013
3047 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3048 Use .pushsection/.popsection.
3050 2018-01-14 Martin Sebor <msebor@redhat.com>
3052 PR c++/81327
3053 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3055 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3057 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3058 entry from extra_headers.
3059 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3060 extra_headers, make the list bitwise identical to the i?86-*-* one.
3062 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3064 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3065 -mcmodel=large with -mindirect-branch=thunk,
3066 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3067 -mfunction-return=thunk-extern.
3068 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3069 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3070 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3072 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3074 * config/i386/i386.c (print_reg): Print the name of the full
3075 integer register without '%'.
3076 (ix86_print_operand): Handle 'V'.
3077 * doc/extend.texi: Document 'V' modifier.
3079 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3081 * config/i386/constraints.md (Bs): Disallow memory operand for
3082 -mindirect-branch-register.
3083 (Bw): Likewise.
3084 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3085 (GOT_memory_operand): Likewise.
3086 (call_insn_operand): Likewise.
3087 (sibcall_insn_operand): Likewise.
3088 (GOT32_symbol_operand): Likewise.
3089 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3090 for -mindirect-branch-register.
3091 (tablejump): Likewise.
3092 (*sibcall_memory): Likewise.
3093 (*sibcall_value_memory): Likewise.
3094 Disallow peepholes of indirect call and jump via memory for
3095 -mindirect-branch-register.
3096 (*call_pop): Replace m with Bw.
3097 (*call_value_pop): Likewise.
3098 (*sibcall_pop_memory): Replace m with Bs.
3099 * config/i386/i386.opt (mindirect-branch-register): New option.
3100 * doc/invoke.texi: Document -mindirect-branch-register option.
3102 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3104 * config/i386/i386-protos.h (ix86_output_function_return): New.
3105 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3106 set function_return_type.
3107 (indirect_thunk_name): Add ret_p to indicate thunk for function
3108 return.
3109 (output_indirect_thunk_function): Pass false to
3110 indirect_thunk_name.
3111 (ix86_output_indirect_branch_via_reg): Likewise.
3112 (ix86_output_indirect_branch_via_push): Likewise.
3113 (output_indirect_thunk_function): Create alias for function
3114 return thunk if regno < 0.
3115 (ix86_output_function_return): New function.
3116 (ix86_handle_fndecl_attribute): Handle function_return.
3117 (ix86_attribute_table): Add function_return.
3118 * config/i386/i386.h (machine_function): Add
3119 function_return_type.
3120 * config/i386/i386.md (simple_return_internal): Use
3121 ix86_output_function_return.
3122 (simple_return_internal_long): Likewise.
3123 * config/i386/i386.opt (mfunction-return=): New option.
3124 (indirect_branch): Mention -mfunction-return=.
3125 * doc/extend.texi: Document function_return function attribute.
3126 * doc/invoke.texi: Document -mfunction-return= option.
3128 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3130 * config/i386/i386-opts.h (indirect_branch): New.
3131 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
3132 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
3133 with local indirect jump when converting indirect call and jump.
3134 (ix86_set_indirect_branch_type): New.
3135 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
3136 (indirectlabelno): New.
3137 (indirect_thunk_needed): Likewise.
3138 (indirect_thunk_bnd_needed): Likewise.
3139 (indirect_thunks_used): Likewise.
3140 (indirect_thunks_bnd_used): Likewise.
3141 (INDIRECT_LABEL): Likewise.
3142 (indirect_thunk_name): Likewise.
3143 (output_indirect_thunk): Likewise.
3144 (output_indirect_thunk_function): Likewise.
3145 (ix86_output_indirect_branch_via_reg): Likewise.
3146 (ix86_output_indirect_branch_via_push): Likewise.
3147 (ix86_output_indirect_branch): Likewise.
3148 (ix86_output_indirect_jmp): Likewise.
3149 (ix86_code_end): Call output_indirect_thunk_function if needed.
3150 (ix86_output_call_insn): Call ix86_output_indirect_branch if
3151 needed.
3152 (ix86_handle_fndecl_attribute): Handle indirect_branch.
3153 (ix86_attribute_table): Add indirect_branch.
3154 * config/i386/i386.h (machine_function): Add indirect_branch_type
3155 and has_local_indirect_jump.
3156 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
3157 to true.
3158 (tablejump): Likewise.
3159 (*indirect_jump): Use ix86_output_indirect_jmp.
3160 (*tablejump_1): Likewise.
3161 (simple_return_indirect_internal): Likewise.
3162 * config/i386/i386.opt (mindirect-branch=): New option.
3163 (indirect_branch): New.
3164 (keep): Likewise.
3165 (thunk): Likewise.
3166 (thunk-inline): Likewise.
3167 (thunk-extern): Likewise.
3168 * doc/extend.texi: Document indirect_branch function attribute.
3169 * doc/invoke.texi: Document -mindirect-branch= option.
3171 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
3173 PR ipa/83051
3174 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
3176 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
3178 * ipa-inline.c (want_inline_small_function_p): Return false if
3179 inlining has already failed with CIF_FINAL_ERROR.
3180 (update_caller_keys): Call want_inline_small_function_p before
3181 can_inline_edge_p.
3182 (update_callee_keys): Likewise.
3184 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3186 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
3187 New function.
3188 (rs6000_quadword_masked_address_p): Likewise.
3189 (quad_aligned_load_p): Likewise.
3190 (quad_aligned_store_p): Likewise.
3191 (const_load_sequence_p): Add comment to describe the outer-most loop.
3192 (mimic_memory_attributes_and_flags): New function.
3193 (rs6000_gen_stvx): Likewise.
3194 (replace_swapped_aligned_store): Likewise.
3195 (rs6000_gen_lvx): Likewise.
3196 (replace_swapped_aligned_load): Likewise.
3197 (replace_swapped_load_constant): Capitalize argument name in
3198 comment describing this function.
3199 (rs6000_analyze_swaps): Add a third pass to search for vector loads
3200 and stores that access quad-word aligned addresses and replace
3201 with stvx or lvx instructions when appropriate.
3202 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
3203 New function prototype.
3204 (rs6000_quadword_masked_address_p): Likewise.
3205 (rs6000_gen_lvx): Likewise.
3206 (rs6000_gen_stvx): Likewise.
3207 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
3208 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
3209 when memory address is aligned.
3210 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
3211 this split to select lvx instruction when memory address is aligned.
3212 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
3213 instruction when memory address is aligned.
3214 (*vsx_le_perm_load_v16qi): Likewise.
3215 (four unnamed splitters): Modify to select the stvx instruction
3216 when memory is aligned.
3218 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
3220 * predict.c (determine_unlikely_bbs): Handle correctly BBs
3221 which appears in the queue multiple times.
3223 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3224 Alan Hayward <alan.hayward@arm.com>
3225 David Sherwood <david.sherwood@arm.com>
3227 * tree-vectorizer.h (vec_lower_bound): New structure.
3228 (_loop_vec_info): Add check_nonzero and lower_bounds.
3229 (LOOP_VINFO_CHECK_NONZERO): New macro.
3230 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
3231 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
3232 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
3233 fields. Make seg_len the distance travelled, not including the
3234 access size.
3235 (dr_direction_indicator): Declare.
3236 (dr_zero_step_indicator): Likewise.
3237 (dr_known_forward_stride_p): Likewise.
3238 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
3239 tree-ssanames.h.
3240 (runtime_alias_check_p): Allow runtime alias checks with
3241 variable strides.
3242 (operator ==): Compare access_size and align.
3243 (prune_runtime_alias_test_list): Rework for new distinction between
3244 the access_size and seg_len.
3245 (create_intersect_range_checks_index): Likewise. Cope with polynomial
3246 segment lengths.
3247 (get_segment_min_max): New function.
3248 (create_intersect_range_checks): Use it.
3249 (dr_step_indicator): New function.
3250 (dr_direction_indicator): Likewise.
3251 (dr_zero_step_indicator): Likewise.
3252 (dr_known_forward_stride_p): Likewise.
3253 * tree-loop-distribution.c (data_ref_segment_size): Return
3254 DR_STEP * (niters - 1).
3255 (compute_alias_check_pairs): Update call to the dr_with_seg_len
3256 constructor.
3257 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
3258 (vect_preserves_scalar_order_p): New function, split out from...
3259 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
3260 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
3261 (vect_vfa_access_size): New function.
3262 (vect_vfa_align): Likewise.
3263 (vect_compile_time_alias): Take access_size_a and access_b arguments.
3264 (dump_lower_bound): New function.
3265 (vect_check_lower_bound): Likewise.
3266 (vect_small_gap_p): Likewise.
3267 (vectorizable_with_step_bound_p): Likewise.
3268 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
3269 depencies if the vectorization factor is 1. Convert the checks
3270 for nonzero steps into checks on the bounds of DR_STEP. Try using
3271 a bunds check for variable steps if the minimum required step is
3272 relatively small. Update calls to the dr_with_seg_len
3273 constructor and to vect_compile_time_alias.
3274 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
3275 function.
3276 (vect_loop_versioning): Call it.
3277 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
3278 when retrying.
3279 (vect_estimate_min_profitable_iters): Account for any bounds checks.
3281 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3282 Alan Hayward <alan.hayward@arm.com>
3283 David Sherwood <david.sherwood@arm.com>
3285 * doc/sourcebuild.texi (vect_scatter_store): Document.
3286 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
3287 optabs.
3288 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
3289 Document.
3290 * genopinit.c (main): Add supports_vec_scatter_store and
3291 supports_vec_scatter_store_cached to target_optabs.
3292 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
3293 IFN_MASK_SCATTER_STORE.
3294 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
3295 functions.
3296 * internal-fn.h (internal_store_fn_p): Declare.
3297 (internal_fn_stored_value_index): Likewise.
3298 * internal-fn.c (scatter_store_direct): New macro.
3299 (expand_scatter_store_optab_fn): New function.
3300 (direct_scatter_store_optab_supported_p): New macro.
3301 (internal_store_fn_p): New function.
3302 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
3303 IFN_MASK_SCATTER_STORE.
3304 (internal_fn_mask_index): Likewise.
3305 (internal_fn_stored_value_index): New function.
3306 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
3307 for scatter stores.
3308 * optabs-query.h (supports_vec_scatter_store_p): Declare.
3309 * optabs-query.c (supports_vec_scatter_store_p): New function.
3310 * tree-vectorizer.h (vect_get_store_rhs): Declare.
3311 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
3312 true for scatter stores.
3313 (vect_gather_scatter_fn_p): Handle scatter stores too.
3314 (vect_check_gather_scatter): Consider using scatter stores if
3315 supports_vec_scatter_store_p.
3316 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
3317 scatter stores too.
3318 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3319 internal_fn_stored_value_index.
3320 (check_load_store_masking): Handle scatter stores too.
3321 (vect_get_store_rhs): Make public.
3322 (vectorizable_call): Use internal_store_fn_p.
3323 (vectorizable_store): Handle scatter store internal functions.
3324 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
3325 when deciding whether the end of the group has been reached.
3326 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
3327 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
3328 (mask_scatter_store<mode>): New insns.
3330 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3331 Alan Hayward <alan.hayward@arm.com>
3332 David Sherwood <david.sherwood@arm.com>
3334 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
3335 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
3336 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
3337 function.
3338 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
3339 Use vect_truncate_gather_scatter_offset if we can't treat the
3340 operation as a normal gather load or scatter store.
3341 (get_group_load_store_type): Take the gather_scatter_info
3342 as argument. Try using a gather load or scatter store for
3343 single-element groups.
3344 (get_load_store_type): Update calls to get_group_load_store_type
3345 and vect_use_strided_gather_scatters_p.
3347 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3348 Alan Hayward <alan.hayward@arm.com>
3349 David Sherwood <david.sherwood@arm.com>
3351 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
3352 optional tree argument.
3353 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
3354 null target hooks.
3355 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
3356 but continue to use the current value as a fallback.
3357 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
3358 to compare the updates.
3359 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
3360 (get_load_store_type): Use it when handling a strided access.
3361 (vect_get_strided_load_store_ops): New function.
3362 (vect_get_data_ptr_increment): Likewise.
3363 (vectorizable_load): Handle strided gather loads. Always pass
3364 a step to vect_create_data_ref_ptr and bump_vector_ptr.
3366 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3367 Alan Hayward <alan.hayward@arm.com>
3368 David Sherwood <david.sherwood@arm.com>
3370 * doc/md.texi (gather_load@var{m}): Document.
3371 (mask_gather_load@var{m}): Likewise.
3372 * genopinit.c (main): Add supports_vec_gather_load and
3373 supports_vec_gather_load_cached to target_optabs.
3374 * optabs-tree.c (init_tree_optimization_optabs): Use
3375 ggc_cleared_alloc to allocate target_optabs.
3376 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
3377 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
3378 functions.
3379 * internal-fn.h (internal_load_fn_p): Declare.
3380 (internal_gather_scatter_fn_p): Likewise.
3381 (internal_fn_mask_index): Likewise.
3382 (internal_gather_scatter_fn_supported_p): Likewise.
3383 * internal-fn.c (gather_load_direct): New macro.
3384 (expand_gather_load_optab_fn): New function.
3385 (direct_gather_load_optab_supported_p): New macro.
3386 (direct_internal_fn_optab): New function.
3387 (internal_load_fn_p): Likewise.
3388 (internal_gather_scatter_fn_p): Likewise.
3389 (internal_fn_mask_index): Likewise.
3390 (internal_gather_scatter_fn_supported_p): Likewise.
3391 * optabs-query.c (supports_at_least_one_mode_p): New function.
3392 (supports_vec_gather_load_p): Likewise.
3393 * optabs-query.h (supports_vec_gather_load_p): Declare.
3394 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
3395 and memory_type field.
3396 (NUM_PATTERNS): Bump to 15.
3397 * tree-vect-data-refs.c: Include internal-fn.h.
3398 (vect_gather_scatter_fn_p): New function.
3399 (vect_describe_gather_scatter_call): Likewise.
3400 (vect_check_gather_scatter): Try using internal functions for
3401 gather loads. Recognize existing calls to a gather load function.
3402 (vect_analyze_data_refs): Consider using gather loads if
3403 supports_vec_gather_load_p.
3404 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
3405 (vect_get_gather_scatter_offset_type): Likewise.
3406 (vect_convert_mask_for_vectype): Likewise.
3407 (vect_add_conversion_to_patterm): Likewise.
3408 (vect_try_gather_scatter_pattern): Likewise.
3409 (vect_recog_gather_scatter_pattern): New pattern recognizer.
3410 (vect_vect_recog_func_ptrs): Add it.
3411 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3412 internal_fn_mask_index and internal_gather_scatter_fn_p.
3413 (check_load_store_masking): Take the gather_scatter_info as an
3414 argument and handle gather loads.
3415 (vect_get_gather_scatter_ops): New function.
3416 (vectorizable_call): Check internal_load_fn_p.
3417 (vectorizable_load): Likewise. Handle gather load internal
3418 functions.
3419 (vectorizable_store): Update call to check_load_store_masking.
3420 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
3421 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
3422 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
3423 (aarch64_gather_scale_operand_d): New predicates.
3424 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
3425 (mask_gather_load<mode>): New insns.
3427 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3428 Alan Hayward <alan.hayward@arm.com>
3429 David Sherwood <david.sherwood@arm.com>
3431 * optabs.def (fold_left_plus_optab): New optab.
3432 * doc/md.texi (fold_left_plus_@var{m}): Document.
3433 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
3434 * internal-fn.c (fold_left_direct): Define.
3435 (expand_fold_left_optab_fn): Likewise.
3436 (direct_fold_left_optab_supported_p): Likewise.
3437 * fold-const-call.c (fold_const_fold_left): New function.
3438 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
3439 * tree-parloops.c (valid_reduction_p): New function.
3440 (gather_scalar_reductions): Use it.
3441 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
3442 (vect_finish_replace_stmt): Declare.
3443 * tree-vect-loop.c (fold_left_reduction_fn): New function.
3444 (needs_fold_left_reduction_p): New function, split out from...
3445 (vect_is_simple_reduction): ...here. Accept reductions that
3446 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
3447 (vect_force_simple_reduction): Also store the reduction type in
3448 the assignment's STMT_VINFO_REDUC_TYPE.
3449 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
3450 (merge_with_identity): New function.
3451 (vect_expand_fold_left): Likewise.
3452 (vectorize_fold_left_reduction): Likewise.
3453 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
3454 scalar phi in place for it. Check for target support and reject
3455 cases that would reassociate the operation. Defer the transform
3456 phase to vectorize_fold_left_reduction.
3457 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
3458 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
3459 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
3461 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3463 * tree-if-conv.c (predicate_mem_writes): Remove redundant
3464 call to ifc_temp_var.
3466 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3467 Alan Hayward <alan.hayward@arm.com>
3468 David Sherwood <david.sherwood@arm.com>
3470 * target.def (legitimize_address_displacement): Take the original
3471 offset as a poly_int.
3472 * targhooks.h (default_legitimize_address_displacement): Update
3473 accordingly.
3474 * targhooks.c (default_legitimize_address_displacement): Likewise.
3475 * doc/tm.texi: Regenerate.
3476 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
3477 as an argument, moving assert of ad->disp == ad->disp_term to...
3478 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
3479 Try calling targetm.legitimize_address_displacement before expanding
3480 the address rather than afterwards, and adjust for the new interface.
3481 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
3482 Match the new hook interface. Handle SVE addresses.
3483 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
3484 new hook interface.
3486 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3488 * Makefile.in (OBJS): Add early-remat.o.
3489 * target.def (select_early_remat_modes): New hook.
3490 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
3491 * doc/tm.texi: Regenerate.
3492 * targhooks.h (default_select_early_remat_modes): Declare.
3493 * targhooks.c (default_select_early_remat_modes): New function.
3494 * timevar.def (TV_EARLY_REMAT): New timevar.
3495 * passes.def (pass_early_remat): New pass.
3496 * tree-pass.h (make_pass_early_remat): Declare.
3497 * early-remat.c: New file.
3498 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
3499 function.
3500 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
3502 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3503 Alan Hayward <alan.hayward@arm.com>
3504 David Sherwood <david.sherwood@arm.com>
3506 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
3507 vfm1 with a bound_epilog parameter.
3508 (vect_do_peeling): Update calls accordingly, and move the prologue
3509 call earlier in the function. Treat the base bound_epilog as 0 for
3510 fully-masked loops and retain vf - 1 for other loops. Add 1 to
3511 this base when peeling for gaps.
3512 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
3513 with fully-masked loops.
3514 (vect_estimate_min_profitable_iters): Handle the single peeled
3515 iteration in that case.
3517 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3518 Alan Hayward <alan.hayward@arm.com>
3519 David Sherwood <david.sherwood@arm.com>
3521 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
3522 single-element interleaving even if the size is not a power of 2.
3523 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
3524 accesses for single-element interleaving if the group size is
3525 not a power of 2.
3527 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3528 Alan Hayward <alan.hayward@arm.com>
3529 David Sherwood <david.sherwood@arm.com>
3531 * doc/md.texi (fold_extract_last_@var{m}): Document.
3532 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
3533 * optabs.def (fold_extract_last_optab): New optab.
3534 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
3535 * internal-fn.c (fold_extract_direct): New macro.
3536 (expand_fold_extract_optab_fn): Likewise.
3537 (direct_fold_extract_optab_supported_p): Likewise.
3538 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
3539 * tree-vect-loop.c (vect_model_reduction_cost): Handle
3540 EXTRACT_LAST_REDUCTION.
3541 (get_initial_def_for_reduction): Do not create an initial vector
3542 for EXTRACT_LAST_REDUCTION reductions.
3543 (vectorizable_reduction): Leave the scalar phi in place for
3544 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
3545 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
3546 epilogue code for EXTRACT_LAST_REDUCTION and defer the
3547 transform phase to vectorizable_condition.
3548 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
3549 split out from...
3550 (vect_finish_stmt_generation): ...here.
3551 (vect_finish_replace_stmt): New function.
3552 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
3553 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
3554 pattern.
3555 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
3557 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3558 Alan Hayward <alan.hayward@arm.com>
3559 David Sherwood <david.sherwood@arm.com>
3561 * doc/md.texi (extract_last_@var{m}): Document.
3562 * optabs.def (extract_last_optab): New optab.
3563 * internal-fn.def (EXTRACT_LAST): New internal function.
3564 * internal-fn.c (cond_unary_direct): New macro.
3565 (expand_cond_unary_optab_fn): Likewise.
3566 (direct_cond_unary_optab_supported_p): Likewise.
3567 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
3568 loops using EXTRACT_LAST.
3569 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
3570 (extract_last_<mode>): ...this optab.
3571 (vec_extract<mode><Vel>): Update accordingly.
3573 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3574 Alan Hayward <alan.hayward@arm.com>
3575 David Sherwood <david.sherwood@arm.com>
3577 * target.def (empty_mask_is_expensive): New hook.
3578 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
3579 * doc/tm.texi: Regenerate.
3580 * targhooks.h (default_empty_mask_is_expensive): Declare.
3581 * targhooks.c (default_empty_mask_is_expensive): New function.
3582 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
3583 if the target says that empty masks are expensive.
3584 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
3585 New function.
3586 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
3588 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3589 Alan Hayward <alan.hayward@arm.com>
3590 David Sherwood <david.sherwood@arm.com>
3592 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
3593 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
3594 (vect_use_loop_mask_for_alignment_p): New function.
3595 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
3596 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
3597 niters_skip argument. Make sure that the first niters_skip elements
3598 of the first iteration are inactive.
3599 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
3600 Update call to vect_set_loop_masks_directly.
3601 (get_misalign_in_elems): New function, split out from...
3602 (vect_gen_prolog_loop_niters): ...here.
3603 (vect_update_init_of_dr): Take a code argument that specifies whether
3604 the adjustment should be added or subtracted.
3605 (vect_update_init_of_drs): Likewise.
3606 (vect_prepare_for_masked_peels): New function.
3607 (vect_do_peeling): Skip prologue peeling if we're using a mask
3608 instead. Update call to vect_update_inits_of_drs.
3609 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3610 mask_skip_niters.
3611 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
3612 alignment. Do not include the number of peeled iterations in
3613 the minimum threshold in that case.
3614 (vectorizable_induction): Adjust the start value down by
3615 LOOP_VINFO_MASK_SKIP_NITERS iterations.
3616 (vect_transform_loop): Call vect_prepare_for_masked_peels.
3617 Take the number of skipped iterations into account when calculating
3618 the loop bounds.
3619 * tree-vect-stmts.c (vect_gen_while_not): New function.
3621 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3622 Alan Hayward <alan.hayward@arm.com>
3623 David Sherwood <david.sherwood@arm.com>
3625 * doc/sourcebuild.texi (vect_fully_masked): Document.
3626 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
3627 default value to 0.
3628 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
3629 split out from...
3630 (vect_analyze_loop_2): ...here. Don't check the vectorization
3631 factor against the number of loop iterations if the loop is
3632 fully-masked.
3634 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3635 Alan Hayward <alan.hayward@arm.com>
3636 David Sherwood <david.sherwood@arm.com>
3638 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
3639 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
3640 (dump_groups): Update accordingly.
3641 (iv_use::mem_type): New member variable.
3642 (address_p): New function.
3643 (record_use): Add a mem_type argument and initialize the new
3644 mem_type field.
3645 (record_group_use): Add a mem_type argument. Use address_p.
3646 Remove obsolete null checks of base_object. Update call to record_use.
3647 (find_interesting_uses_op): Update call to record_group_use.
3648 (find_interesting_uses_cond): Likewise.
3649 (find_interesting_uses_address): Likewise.
3650 (get_mem_type_for_internal_fn): New function.
3651 (find_address_like_use): Likewise.
3652 (find_interesting_uses_stmt): Try find_address_like_use before
3653 calling find_interesting_uses_op.
3654 (addr_offset_valid_p): Use the iv mem_type field as the type
3655 of the addressed memory.
3656 (add_autoinc_candidates): Likewise.
3657 (get_address_cost): Likewise.
3658 (split_small_address_groups_p): Use address_p.
3659 (split_address_groups): Likewise.
3660 (add_iv_candidate_for_use): Likewise.
3661 (autoinc_possible_for_pair): Likewise.
3662 (rewrite_groups): Likewise.
3663 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
3664 (determine_group_iv_cost): Update after split of USE_ADDRESS.
3665 (get_alias_ptr_type_for_ptr_address): New function.
3666 (rewrite_use_address): Rewrite address uses in calls that were
3667 identified by find_address_like_use.
3669 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3670 Alan Hayward <alan.hayward@arm.com>
3671 David Sherwood <david.sherwood@arm.com>
3673 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
3674 TARGET_MEM_REFs.
3675 * gimple-expr.h (is_gimple_addressable: Likewise.
3676 * gimple-expr.c (is_gimple_address): Likewise.
3677 * internal-fn.c (expand_call_mem_ref): New function.
3678 (expand_mask_load_optab_fn): Use it.
3679 (expand_mask_store_optab_fn): Likewise.
3681 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3682 Alan Hayward <alan.hayward@arm.com>
3683 David Sherwood <david.sherwood@arm.com>
3685 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
3686 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
3687 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
3688 (cond_umax@var{mode}): Document.
3689 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
3690 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
3691 (cond_umin_optab, cond_umax_optab): New optabs.
3692 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
3693 (COND_IOR, COND_XOR): New internal functions.
3694 * internal-fn.h (get_conditional_internal_fn): Declare.
3695 * internal-fn.c (cond_binary_direct): New macro.
3696 (expand_cond_binary_optab_fn): Likewise.
3697 (direct_cond_binary_optab_supported_p): Likewise.
3698 (get_conditional_internal_fn): New function.
3699 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
3700 Cope with reduction statements that are vectorized as calls rather
3701 than assignments.
3702 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
3703 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
3704 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
3705 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3706 (UNSPEC_COND_EOR): New unspecs.
3707 (optab): Add mappings for them.
3708 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
3709 (sve_int_op, sve_fp_op): New int attributes.
3711 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3712 Alan Hayward <alan.hayward@arm.com>
3713 David Sherwood <david.sherwood@arm.com>
3715 * optabs.def (while_ult_optab): New optab.
3716 * doc/md.texi (while_ult@var{m}@var{n}): Document.
3717 * internal-fn.def (WHILE_ULT): New internal function.
3718 * internal-fn.h (direct_internal_fn_supported_p): New override
3719 that takes two types as argument.
3720 * internal-fn.c (while_direct): New macro.
3721 (expand_while_optab_fn): New function.
3722 (convert_optab_supported_p): Likewise.
3723 (direct_while_optab_supported_p): New macro.
3724 * wide-int.h (wi::udiv_ceil): New function.
3725 * tree-vectorizer.h (rgroup_masks): New structure.
3726 (vec_loop_masks): New typedef.
3727 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
3728 and fully_masked_p.
3729 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
3730 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
3731 (vect_max_vf): New function.
3732 (slpeel_make_loop_iterate_ntimes): Delete.
3733 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
3734 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
3735 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
3736 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
3737 internal-fn.h, stor-layout.h and optabs-query.h.
3738 (vect_set_loop_mask): New function.
3739 (add_preheader_seq): Likewise.
3740 (add_header_seq): Likewise.
3741 (interleave_supported_p): Likewise.
3742 (vect_maybe_permute_loop_masks): Likewise.
3743 (vect_set_loop_masks_directly): Likewise.
3744 (vect_set_loop_condition_masked): Likewise.
3745 (vect_set_loop_condition_unmasked): New function, split out from
3746 slpeel_make_loop_iterate_ntimes.
3747 (slpeel_make_loop_iterate_ntimes): Rename to..
3748 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
3749 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
3750 (vect_do_peeling): Update call accordingly.
3751 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
3752 loops.
3753 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3754 mask_compare_type, can_fully_mask_p and fully_masked_p.
3755 (release_vec_loop_masks): New function.
3756 (_loop_vec_info): Use it to free the loop masks.
3757 (can_produce_all_loop_masks_p): New function.
3758 (vect_get_max_nscalars_per_iter): Likewise.
3759 (vect_verify_full_masking): Likewise.
3760 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
3761 retries, and free the mask rgroups before retrying. Check loop-wide
3762 reasons for disallowing fully-masked loops. Make the final decision
3763 about whether use a fully-masked loop or not.
3764 (vect_estimate_min_profitable_iters): Do not assume that peeling
3765 for the number of iterations will be needed for fully-masked loops.
3766 (vectorizable_reduction): Disable fully-masked loops.
3767 (vectorizable_live_operation): Likewise.
3768 (vect_halve_mask_nunits): New function.
3769 (vect_double_mask_nunits): Likewise.
3770 (vect_record_loop_mask): Likewise.
3771 (vect_get_loop_mask): Likewise.
3772 (vect_transform_loop): Handle the case in which the final loop
3773 iteration might handle a partial vector. Call vect_set_loop_condition
3774 instead of slpeel_make_loop_iterate_ntimes.
3775 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
3776 (check_load_store_masking): New function.
3777 (prepare_load_store_mask): Likewise.
3778 (vectorizable_store): Handle fully-masked loops.
3779 (vectorizable_load): Likewise.
3780 (supportable_widening_operation): Use vect_halve_mask_nunits for
3781 booleans.
3782 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
3783 (vect_gen_while): New function.
3784 * config/aarch64/aarch64.md (umax<mode>3): New expander.
3785 (aarch64_uqdec<mode>): New insn.
3787 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3788 Alan Hayward <alan.hayward@arm.com>
3789 David Sherwood <david.sherwood@arm.com>
3791 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
3792 (reduc_xor_scal_optab): New optabs.
3793 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
3794 (reduc_xor_scal_@var{m}): Document.
3795 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
3796 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
3797 internal functions.
3798 * fold-const-call.c (fold_const_call): Handle them.
3799 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
3800 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
3801 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
3802 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
3803 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
3804 (UNSPEC_XORV): New unspecs.
3805 (optab): Add entries for them.
3806 (BITWISEV): New int iterator.
3807 (bit_reduc_op): New int attributes.
3809 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3810 Alan Hayward <alan.hayward@arm.com>
3811 David Sherwood <david.sherwood@arm.com>
3813 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
3814 * internal-fn.def (VEC_SHL_INSERT): New internal function.
3815 * optabs.def (vec_shl_insert_optab): New optab.
3816 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
3817 (duplicate_and_interleave): Likewise.
3818 * tree-vect-loop.c: Include internal-fn.h.
3819 (neutral_op_for_slp_reduction): New function, split out from
3820 get_initial_defs_for_reduction.
3821 (get_initial_def_for_reduction): Handle option 2 for variable-length
3822 vectors by loading the neutral value into a vector and then shifting
3823 the initial value into element 0.
3824 (get_initial_defs_for_reduction): Replace the code argument with
3825 the neutral value calculated by neutral_op_for_slp_reduction.
3826 Use gimple_build_vector for constant-length vectors.
3827 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
3828 but the first group_size elements have a neutral value.
3829 Use duplicate_and_interleave otherwise.
3830 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
3831 Update call to get_initial_defs_for_reduction. Handle SLP
3832 reductions for variable-length vectors by creating one vector
3833 result for each scalar result, with the elements associated
3834 with other scalar results stubbed out with the neutral value.
3835 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
3836 Require IFN_VEC_SHL_INSERT for double reductions on
3837 variable-length vectors, or SLP reductions that have
3838 a neutral value. Require can_duplicate_and_interleave_p
3839 support for variable-length unchained SLP reductions if there
3840 is no neutral value, such as for MIN/MAX reductions. Also require
3841 the number of vector elements to be a multiple of the number of
3842 SLP statements when doing variable-length unchained SLP reductions.
3843 Update call to vect_create_epilog_for_reduction.
3844 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
3845 and remove initial values.
3846 (duplicate_and_interleave): Make public.
3847 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
3848 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
3850 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3851 Alan Hayward <alan.hayward@arm.com>
3852 David Sherwood <david.sherwood@arm.com>
3854 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
3855 (can_duplicate_and_interleave_p): New function.
3856 (vect_get_and_check_slp_defs): Take the vector of statements
3857 rather than just the current one. Remove excess parentheses.
3858 Restriction rejectinon of vect_constant_def and vect_external_def
3859 for variable-length vectors to boolean types, or types for which
3860 can_duplicate_and_interleave_p is false.
3861 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
3862 (duplicate_and_interleave): New function.
3863 (vect_get_constant_vectors): Use gimple_build_vector for
3864 constant-length vectors and suitable variable-length constant
3865 vectors. Use duplicate_and_interleave for other variable-length
3866 vectors. Don't defer the update when inserting new statements.
3868 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3869 Alan Hayward <alan.hayward@arm.com>
3870 David Sherwood <david.sherwood@arm.com>
3872 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
3873 min_profitable_iters doesn't go negative.
3875 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3876 Alan Hayward <alan.hayward@arm.com>
3877 David Sherwood <david.sherwood@arm.com>
3879 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
3880 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
3881 * optabs.def (vec_mask_load_lanes_optab): New optab.
3882 (vec_mask_store_lanes_optab): Likewise.
3883 * internal-fn.def (MASK_LOAD_LANES): New internal function.
3884 (MASK_STORE_LANES): Likewise.
3885 * internal-fn.c (mask_load_lanes_direct): New macro.
3886 (mask_store_lanes_direct): Likewise.
3887 (expand_mask_load_optab_fn): Handle masked operations.
3888 (expand_mask_load_lanes_optab_fn): New macro.
3889 (expand_mask_store_optab_fn): Handle masked operations.
3890 (expand_mask_store_lanes_optab_fn): New macro.
3891 (direct_mask_load_lanes_optab_supported_p): Likewise.
3892 (direct_mask_store_lanes_optab_supported_p): Likewise.
3893 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
3894 parameter.
3895 (vect_load_lanes_supported): Likewise.
3896 * tree-vect-data-refs.c (strip_conversion): New function.
3897 (can_group_stmts_p): Likewise.
3898 (vect_analyze_data_ref_accesses): Use it instead of checking
3899 for a pair of assignments.
3900 (vect_store_lanes_supported): Take a masked_p parameter.
3901 (vect_load_lanes_supported): Likewise.
3902 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
3903 vect_store_lanes_supported and vect_load_lanes_supported.
3904 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3905 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
3906 parameter. Don't allow gaps for masked accesses.
3907 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
3908 and vect_load_lanes_supported.
3909 (get_load_store_type): Take a masked_p parameter and update
3910 call to get_group_load_store_type.
3911 (vectorizable_store): Update call to get_load_store_type.
3912 Handle IFN_MASK_STORE_LANES.
3913 (vectorizable_load): Update call to get_load_store_type.
3914 Handle IFN_MASK_LOAD_LANES.
3916 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3917 Alan Hayward <alan.hayward@arm.com>
3918 David Sherwood <david.sherwood@arm.com>
3920 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
3921 modes for SVE.
3922 * config/aarch64/aarch64-protos.h
3923 (aarch64_sve_struct_memory_operand_p): Declare.
3924 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
3925 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
3926 (VPRED, vpred): Handle SVE structure modes.
3927 * config/aarch64/constraints.md (Utx): New constraint.
3928 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
3929 (aarch64_sve_struct_nonimmediate_operand): New predicates.
3930 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
3931 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
3932 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
3933 structure modes. Split into pieces after RA.
3934 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
3935 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
3936 New patterns.
3937 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
3938 SVE structure modes.
3939 (aarch64_classify_address): Likewise.
3940 (sizetochar): Move earlier in file.
3941 (aarch64_print_operand): Handle SVE register lists.
3942 (aarch64_array_mode): New function.
3943 (aarch64_sve_struct_memory_operand_p): Likewise.
3944 (TARGET_ARRAY_MODE): Redefine.
3946 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3947 Alan Hayward <alan.hayward@arm.com>
3948 David Sherwood <david.sherwood@arm.com>
3950 * target.def (array_mode): New target hook.
3951 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
3952 * doc/tm.texi: Regenerate.
3953 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
3954 * hooks.c (hook_optmode_mode_uhwi_none): New function.
3955 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
3956 targetm.array_mode.
3957 * stor-layout.c (mode_for_array): Likewise. Support polynomial
3958 type sizes.
3960 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3961 Alan Hayward <alan.hayward@arm.com>
3962 David Sherwood <david.sherwood@arm.com>
3964 * fold-const.c (fold_binary_loc): Check the argument types
3965 rather than the result type when testing for a vector operation.
3967 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3969 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
3970 * doc/tm.texi: Regenerate.
3972 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3973 Alan Hayward <alan.hayward@arm.com>
3974 David Sherwood <david.sherwood@arm.com>
3976 * doc/invoke.texi (-msve-vector-bits=): Document new option.
3977 (sve): Document new AArch64 extension.
3978 * doc/md.texi (w): Extend the description of the AArch64
3979 constraint to include SVE vectors.
3980 (Upl, Upa): Document new AArch64 predicate constraints.
3981 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
3982 enum.
3983 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
3984 (msve-vector-bits=): New option.
3985 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
3986 SVE when these are disabled.
3987 (sve): New extension.
3988 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
3989 modes. Adjust their number of units based on aarch64_sve_vg.
3990 (MAX_BITSIZE_MODE_ANY_MODE): Define.
3991 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
3992 aarch64_addr_query_type.
3993 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
3994 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
3995 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
3996 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
3997 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
3998 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
3999 (aarch64_simd_imm_zero_p): Delete.
4000 (aarch64_check_zero_based_sve_index_immediate): Declare.
4001 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4002 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4003 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4004 (aarch64_sve_float_mul_immediate_p): Likewise.
4005 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4006 rather than an rtx.
4007 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4008 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4009 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4010 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4011 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4012 (aarch64_regmode_natural_size): Likewise.
4013 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4014 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4015 left one place.
4016 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4017 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4018 for VG and the SVE predicate registers.
4019 (V_ALIASES): Add a "z"-prefixed alias.
4020 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4021 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4022 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4023 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4024 (REG_CLASS_NAMES): Add entries for them.
4025 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4026 and the predicate registers.
4027 (aarch64_sve_vg): Declare.
4028 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4029 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4030 (REGMODE_NATURAL_SIZE): Define.
4031 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4032 SVE macros.
4033 * config/aarch64/aarch64.c: Include cfgrtl.h.
4034 (simd_immediate_info): Add a constructor for series vectors,
4035 and an associated step field.
4036 (aarch64_sve_vg): New variable.
4037 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4038 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4039 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4040 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4041 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4042 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4043 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4044 (aarch64_get_mask_mode): New functions.
4045 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4046 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4047 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4048 predicate modes and predicate registers. Explicitly restrict
4049 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4050 to store a vector mode if it is recognized by
4051 aarch64_classify_vector_mode.
4052 (aarch64_regmode_natural_size): New function.
4053 (aarch64_hard_regno_caller_save_mode): Return the original mode
4054 for predicates.
4055 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4056 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4057 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4058 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4059 functions.
4060 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4061 does not overlap dest if the function is frame-related. Handle
4062 SVE constants.
4063 (aarch64_split_add_offset): New function.
4064 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4065 them aarch64_add_offset.
4066 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4067 and update call to aarch64_sub_sp.
4068 (aarch64_add_cfa_expression): New function.
4069 (aarch64_expand_prologue): Pass extra temporary registers to the
4070 functions above. Handle the case in which we need to emit new
4071 DW_CFA_expressions for registers that were originally saved
4072 relative to the stack pointer, but now have to be expressed
4073 relative to the frame pointer.
4074 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4075 functions above.
4076 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4077 IP0 and IP1 values for SVE frames.
4078 (aarch64_expand_vec_series): New function.
4079 (aarch64_expand_sve_widened_duplicate): Likewise.
4080 (aarch64_expand_sve_const_vector): Likewise.
4081 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4082 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4083 into the register, rather than emitting a SET directly.
4084 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4085 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4086 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4087 (offset_9bit_signed_scaled_p): New functions.
4088 (aarch64_replicate_bitmask_imm): New function.
4089 (aarch64_bitmask_imm): Use it.
4090 (aarch64_cannot_force_const_mem): Reject expressions involving
4091 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4092 (aarch64_classify_index): Handle SVE indices, by requiring
4093 a plain register index with a scale that matches the element size.
4094 (aarch64_classify_address): Handle SVE addresses. Assert that
4095 the mode of the address is VOIDmode or an integer mode.
4096 Update call to aarch64_classify_symbol.
4097 (aarch64_classify_symbolic_expression): Update call to
4098 aarch64_classify_symbol.
4099 (aarch64_const_vec_all_in_range_p): New function.
4100 (aarch64_print_vector_float_operand): Likewise.
4101 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4102 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4103 and the FP immediates 1.0 and 0.5.
4104 (aarch64_print_address_internal): Handle SVE addresses.
4105 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4106 (aarch64_regno_regclass): Handle predicate registers.
4107 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4108 data modes.
4109 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4110 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4111 (aarch64_convert_sve_vector_bits): New function.
4112 (aarch64_override_options): Use it to handle -msve-vector-bits=.
4113 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4114 rather than an rtx.
4115 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
4116 Handle SVE vector and predicate modes. Accept VL-based constants
4117 that need only one temporary register, and VL offsets that require
4118 no temporary registers.
4119 (aarch64_conditional_register_usage): Mark the predicate registers
4120 as fixed if SVE isn't available.
4121 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
4122 Return true for SVE vector and predicate modes.
4123 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
4124 rather than an unsigned int. Handle SVE modes.
4125 (aarch64_preferred_simd_mode): Update call accordingly. Handle
4126 SVE modes.
4127 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
4128 if SVE is enabled.
4129 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4130 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4131 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4132 (aarch64_sve_float_mul_immediate_p): New functions.
4133 (aarch64_sve_valid_immediate): New function.
4134 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
4135 Explicitly reject structure modes. Check for INDEX constants.
4136 Handle PTRUE and PFALSE constants.
4137 (aarch64_check_zero_based_sve_index_immediate): New function.
4138 (aarch64_simd_imm_zero_p): Delete.
4139 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
4140 vector modes. Accept constants in the range of CNT[BHWD].
4141 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
4142 ask for an Advanced SIMD mode.
4143 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
4144 (aarch64_simd_vector_alignment): Handle SVE predicates.
4145 (aarch64_vectorize_preferred_vector_alignment): New function.
4146 (aarch64_simd_vector_alignment_reachable): Use it instead of
4147 the vector size.
4148 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
4149 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
4150 functions.
4151 (MAX_VECT_LEN): Delete.
4152 (expand_vec_perm_d): Add a vec_flags field.
4153 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
4154 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
4155 (aarch64_evpc_ext): Don't apply a big-endian lane correction
4156 for SVE modes.
4157 (aarch64_evpc_rev): Rename to...
4158 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
4159 (aarch64_evpc_rev_global): New function.
4160 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
4161 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
4162 MAX_VECT_LEN.
4163 (aarch64_evpc_sve_tbl): New function.
4164 (aarch64_expand_vec_perm_const_1): Update after rename of
4165 aarch64_evpc_rev. Handle SVE permutes too, trying
4166 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
4167 than aarch64_evpc_tbl.
4168 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
4169 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
4170 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
4171 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
4172 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
4173 (aarch64_expand_sve_vcond): New functions.
4174 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
4175 of aarch64_vector_mode_p.
4176 (aarch64_dwarf_poly_indeterminate_value): New function.
4177 (aarch64_compute_pressure_classes): Likewise.
4178 (aarch64_can_change_mode_class): Likewise.
4179 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
4180 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
4181 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
4182 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
4183 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
4184 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
4185 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
4186 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
4187 constraints.
4188 (Dn, Dl, Dr): Accept const as well as const_vector.
4189 (Dz): Likewise. Compare against CONST0_RTX.
4190 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
4191 of "vector" where appropriate.
4192 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
4193 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
4194 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
4195 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
4196 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
4197 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
4198 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
4199 (v_int_equiv): Extend to SVE modes.
4200 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
4201 mode attributes.
4202 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
4203 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
4204 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
4205 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
4206 (SVE_COND_FP_CMP): New int iterators.
4207 (perm_hilo): Handle the new unpack unspecs.
4208 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
4209 attributes.
4210 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
4211 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
4212 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
4213 (aarch64_equality_operator, aarch64_constant_vector_operand)
4214 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
4215 (aarch64_sve_nonimmediate_operand): Likewise.
4216 (aarch64_sve_general_operand): Likewise.
4217 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
4218 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
4219 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
4220 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
4221 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
4222 (aarch64_sve_float_arith_immediate): Likewise.
4223 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
4224 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
4225 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
4226 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
4227 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
4228 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
4229 (aarch64_sve_float_arith_operand): Likewise.
4230 (aarch64_sve_float_arith_with_sub_operand): Likewise.
4231 (aarch64_sve_float_mul_operand): Likewise.
4232 (aarch64_sve_vec_perm_operand): Likewise.
4233 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
4234 (aarch64_mov_operand): Accept const_poly_int and const_vector.
4235 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
4236 as well as const_vector.
4237 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
4238 in file. Use CONST0_RTX and CONSTM1_RTX.
4239 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
4240 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
4241 Use aarch64_simd_imm_zero.
4242 * config/aarch64/aarch64-sve.md: New file.
4243 * config/aarch64/aarch64.md: Include it.
4244 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
4245 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
4246 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
4247 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
4248 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
4249 (sve): New attribute.
4250 (enabled): Disable instructions with the sve attribute unless
4251 TARGET_SVE.
4252 (movqi, movhi): Pass CONST_POLY_INT operaneds through
4253 aarch64_expand_mov_immediate.
4254 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
4255 CNT[BHSD] immediates.
4256 (movti): Split CONST_POLY_INT moves into two halves.
4257 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
4258 Split additions that need a temporary here if the destination
4259 is the stack pointer.
4260 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
4261 (*add<mode>3_poly_1): New instruction.
4262 (set_clobber_cc): New expander.
4264 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4266 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
4267 parameter and use it instead of GET_MODE_SIZE (innermode). Use
4268 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
4269 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
4270 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
4271 Change innermode from fixed_mode_size to machine_mode.
4272 (simplify_subreg): Update call accordingly. Handle a constant-sized
4273 subreg of a variable-length CONST_VECTOR.
4275 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4276 Alan Hayward <alan.hayward@arm.com>
4277 David Sherwood <david.sherwood@arm.com>
4279 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
4280 (add_offset_to_base): New function, split out from...
4281 (create_mem_ref): ...here. When handling a scale other than 1,
4282 check first whether the address is valid without the offset.
4283 Add it into the base if so, leaving the index and scale as-is.
4285 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4287 PR c++/83778
4288 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
4289 fold_for_warn before checking if arg2 is INTEGER_CST.
4291 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
4293 * config/rs6000/predicates.md (load_multiple_operation): Delete.
4294 (store_multiple_operation): Delete.
4295 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
4296 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
4297 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
4298 guarded by TARGET_STRING.
4299 (rs6000_output_load_multiple): Delete.
4300 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
4301 OPTION_MASK_STRING / TARGET_STRING handling.
4302 (print_operand) <'N', 'O'>: Add comment that these are unused now.
4303 (const rs6000_opt_masks) <"string">: Change mask to 0.
4304 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
4305 (MASK_STRING): Delete.
4306 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
4307 parts. Simplify.
4308 (load_multiple): Delete.
4309 (*ldmsi8): Delete.
4310 (*ldmsi7): Delete.
4311 (*ldmsi6): Delete.
4312 (*ldmsi5): Delete.
4313 (*ldmsi4): Delete.
4314 (*ldmsi3): Delete.
4315 (store_multiple): Delete.
4316 (*stmsi8): Delete.
4317 (*stmsi7): Delete.
4318 (*stmsi6): Delete.
4319 (*stmsi5): Delete.
4320 (*stmsi4): Delete.
4321 (*stmsi3): Delete.
4322 (movmemsi_8reg): Delete.
4323 (corresponding unnamed define_insn): Delete.
4324 (movmemsi_6reg): Delete.
4325 (corresponding unnamed define_insn): Delete.
4326 (movmemsi_4reg): Delete.
4327 (corresponding unnamed define_insn): Delete.
4328 (movmemsi_2reg): Delete.
4329 (corresponding unnamed define_insn): Delete.
4330 (movmemsi_1reg): Delete.
4331 (corresponding unnamed define_insn): Delete.
4332 * config/rs6000/rs6000.opt (mno-string): New.
4333 (mstring): Replace by deprecation warning stub.
4334 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
4336 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4338 * regrename.c (regrename_do_replace): If replacing the same
4339 reg multiple times, try to reuse last created gen_raw_REG.
4341 PR debug/81155
4342 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
4343 main to workaround a bug in GDB.
4345 2018-01-12 Tom de Vries <tom@codesourcery.com>
4347 PR target/83737
4348 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
4350 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
4352 PR rtl-optimization/80481
4353 * ira-color.c (get_cap_member): New function.
4354 (allocnos_conflict_by_live_ranges_p): Use it.
4355 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
4356 (setup_slot_coalesced_allocno_live_ranges): Ditto.
4358 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
4360 PR target/83628
4361 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
4362 (*saddl_se_1): Ditto.
4363 (*ssubsi_1): Ditto.
4364 (*ssubl_se_1): Ditto.
4366 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4368 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
4369 rather than wi::to_widest for DR_INITs.
4370 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
4371 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
4372 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
4373 INTEGER_CSTs.
4374 (vect_analyze_group_access_1): Note that here.
4376 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4378 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
4379 polynomial type sizes.
4381 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4383 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
4384 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
4385 (gimple_add_tmp_var): Likewise.
4387 2018-01-12 Martin Liska <mliska@suse.cz>
4389 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
4390 (gimple_alloc_sizes): Likewise.
4391 (dump_gimple_statistics): Use PRIu64 in printf format.
4392 * gimple.h: Change uint64_t to int.
4394 2018-01-12 Martin Liska <mliska@suse.cz>
4396 * tree-core.h: Use uint64_t instead of int.
4397 * tree.c (tree_node_counts): Likewise.
4398 (tree_node_sizes): Likewise.
4399 (dump_tree_statistics): Use PRIu64 in printf format.
4401 2018-01-12 Martin Liska <mliska@suse.cz>
4403 * Makefile.in: As qsort_chk is implemented in vec.c, add
4404 vec.o to linkage of gencfn-macros.
4405 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
4406 passing the info to record_node_allocation_statistics.
4407 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
4408 and pass the info.
4409 * ggc-common.c (struct ggc_usage): Add operator== and use
4410 it in operator< and compare function.
4411 * mem-stats.h (struct mem_usage): Likewise.
4412 * vec.c (struct vec_usage): Remove operator< and compare
4413 function. Can be simply inherited.
4415 2018-01-12 Martin Jambor <mjambor@suse.cz>
4417 PR target/81616
4418 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
4419 * tree-ssa-math-opts.c: Include domwalk.h.
4420 (convert_mult_to_fma_1): New function.
4421 (fma_transformation_info): New type.
4422 (fma_deferring_state): Likewise.
4423 (cancel_fma_deferring): New function.
4424 (result_of_phi): Likewise.
4425 (last_fma_candidate_feeds_initial_phi): Likewise.
4426 (convert_mult_to_fma): Added deferring logic, split actual
4427 transformation to convert_mult_to_fma_1.
4428 (math_opts_dom_walker): New type.
4429 (math_opts_dom_walker::after_dom_children): New method, body moved
4430 here from pass_optimize_widening_mul::execute, added deferring logic
4431 bits.
4432 (pass_optimize_widening_mul::execute): Moved most of code to
4433 math_opts_dom_walker::after_dom_children.
4434 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
4435 * config/i386/i386.c (ix86_option_override_internal): Added
4436 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
4438 2018-01-12 Richard Biener <rguenther@suse.de>
4440 PR debug/83157
4441 * dwarf2out.c (gen_variable_die): Do not reset old_die for
4442 inline instance vars.
4444 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
4446 PR target/81819
4447 * config/rx/rx.c (rx_is_restricted_memory_address):
4448 Handle SUBREG case.
4450 2018-01-12 Richard Biener <rguenther@suse.de>
4452 PR tree-optimization/80846
4453 * target.def (split_reduction): New target hook.
4454 * targhooks.c (default_split_reduction): New function.
4455 * targhooks.h (default_split_reduction): Declare.
4456 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
4457 target requests first reduce vectors by combining low and high
4458 parts.
4459 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
4460 (get_vectype_for_scalar_type_and_size): Export.
4461 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
4462 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
4463 * doc/tm.texi: Regenerate.
4464 * config/i386/i386.c (ix86_split_reduction): Implement
4465 TARGET_VECTORIZE_SPLIT_REDUCTION.
4467 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4469 PR target/83368
4470 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
4471 in PIC mode except for TARGET_VXWORKS_RTP.
4472 * config/sparc/sparc.c: Include cfgrtl.h.
4473 (TARGET_INIT_PIC_REG): Define.
4474 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
4475 (sparc_pic_register_p): New predicate.
4476 (sparc_legitimate_address_p): Use it.
4477 (sparc_legitimize_pic_address): Likewise.
4478 (sparc_delegitimize_address): Likewise.
4479 (sparc_mode_dependent_address_p): Likewise.
4480 (gen_load_pcrel_sym): Remove 4th parameter.
4481 (load_got_register): Adjust call to above. Remove obsolete stuff.
4482 (sparc_expand_prologue): Do not call load_got_register here.
4483 (sparc_flat_expand_prologue): Likewise.
4484 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
4485 (sparc_use_pseudo_pic_reg): New function.
4486 (sparc_init_pic_reg): Likewise.
4487 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4488 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
4490 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
4492 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
4493 Add item for branch_cost.
4495 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4497 PR rtl-optimization/83565
4498 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
4499 not extend the result to a larger mode for rotate operations.
4500 (num_sign_bit_copies1): Likewise.
4502 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4504 PR target/40411
4505 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
4506 -symbolic.
4507 Use values-Xc.o for -pedantic.
4508 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
4510 2018-01-12 Martin Liska <mliska@suse.cz>
4512 PR ipa/83054
4513 * ipa-devirt.c (final_warning_record::grow_type_warnings):
4514 New function.
4515 (possible_polymorphic_call_targets): Use it.
4516 (ipa_devirt): Likewise.
4518 2018-01-12 Martin Liska <mliska@suse.cz>
4520 * profile-count.h (enum profile_quality): Use 0 as invalid
4521 enum value of profile_quality.
4523 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
4525 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
4526 -mext-string options.
4528 2018-01-12 Richard Biener <rguenther@suse.de>
4530 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
4531 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
4532 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
4533 Likewise.
4534 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
4536 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4538 * configure.ac (--with-long-double-format): Add support for the
4539 configuration option to change the default long double format on
4540 PowerPC systems.
4541 * config.gcc (powerpc*-linux*-*): Likewise.
4542 * configure: Regenerate.
4543 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
4544 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
4545 used without modification.
4547 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4549 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
4550 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
4551 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
4552 MISC_BUILTIN_SPEC_BARRIER.
4553 (rs6000_init_builtins): Likewise.
4554 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
4555 enum value.
4556 (speculation_barrier): New define_insn.
4557 * doc/extend.texi: Document __builtin_speculation_barrier.
4559 2018-01-11 Jakub Jelinek <jakub@redhat.com>
4561 PR target/83203
4562 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
4563 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
4564 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
4565 iterators.
4566 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
4567 integral modes instead of "ss" and "sd".
4568 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
4569 vectors with 32-bit and 64-bit elements.
4570 (vecdupssescalarmodesuffix): New mode attribute.
4571 (vec_dup<mode>): Use it.
4573 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
4575 PR target/83330
4576 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
4577 frame if argument is passed on stack.
4579 2018-01-11 Jakub Jelinek <jakub@redhat.com>
4581 PR target/82682
4582 * ree.c (combine_reaching_defs): Optimize also
4583 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
4584 reg2=any_extend(exp); reg1=reg2;, formatting fix.
4586 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
4588 PR middle-end/83189
4589 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
4591 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
4593 PR middle-end/83718
4594 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
4595 after they are computed.
4597 2018-01-11 Bin Cheng <bin.cheng@arm.com>
4599 PR tree-optimization/83695
4600 * gimple-loop-linterchange.cc
4601 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
4602 reset cached scev information after interchange.
4603 (pass_linterchange::execute): Remove call to scev_reset_htab.
4605 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4607 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
4608 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
4609 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
4610 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
4611 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
4612 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
4613 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
4614 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
4615 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
4616 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
4617 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
4618 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
4619 (V_lane_reg): Likewise.
4620 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
4621 New define_expand.
4622 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
4623 (vfmal_lane_low<mode>_intrinsic,
4624 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
4625 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
4626 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
4627 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
4628 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
4629 vfmsl_lane_high<mode>_intrinsic): New define_insns.
4631 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4633 * config/arm/arm-cpus.in (fp16fml): New feature.
4634 (ALL_SIMD): Add fp16fml.
4635 (armv8.2-a): Add fp16fml as an option.
4636 (armv8.3-a): Likewise.
4637 (armv8.4-a): Add fp16fml as part of fp16.
4638 * config/arm/arm.h (TARGET_FP16FML): Define.
4639 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
4640 when appropriate.
4641 * config/arm/arm-modes.def (V2HF): Define.
4642 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
4643 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
4644 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
4645 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
4646 vfmsl_low, vfmsl_high): New set of builtins.
4647 * config/arm/iterators.md (PLUSMINUS): New code iterator.
4648 (vfml_op): New code attribute.
4649 (VFMLHALVES): New int iterator.
4650 (VFML, VFMLSEL): New mode attributes.
4651 (V_reg): Define mapping for V2HF.
4652 (V_hi, V_lo): New mode attributes.
4653 (VF_constraint): Likewise.
4654 (vfml_half, vfml_half_selector): New int attributes.
4655 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
4656 define_expand.
4657 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
4658 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
4659 New define_insn.
4660 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
4661 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
4662 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
4663 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
4664 documentation.
4665 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
4666 Document new effective target and option set.
4668 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4670 * config/arm/arm-cpus.in (armv8_4): New feature.
4671 (ARMv8_4a): New fgroup.
4672 (armv8.4-a): New arch.
4673 * config/arm/arm-tables.opt: Regenerate.
4674 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
4675 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
4676 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
4677 Add matching rules for -march=armv8.4-a and extensions.
4678 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
4680 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
4682 PR target/81821
4683 * config/rx/rx.md (BW): New mode attribute.
4684 (sync_lock_test_and_setsi): Add mode suffix to insn output.
4686 2018-01-11 Richard Biener <rguenther@suse.de>
4688 PR tree-optimization/83435
4689 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
4690 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
4691 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
4693 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4694 Alan Hayward <alan.hayward@arm.com>
4695 David Sherwood <david.sherwood@arm.com>
4697 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
4698 field.
4699 (aarch64_classify_address): Initialize it. Track polynomial offsets.
4700 (aarch64_print_address_internal): Use it to check for a zero offset.
4702 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4703 Alan Hayward <alan.hayward@arm.com>
4704 David Sherwood <david.sherwood@arm.com>
4706 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
4707 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
4708 Return a poly_int64 rather than a HOST_WIDE_INT.
4709 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
4710 rather than a HOST_WIDE_INT.
4711 * config/aarch64/aarch64.h (aarch64_frame): Protect with
4712 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
4713 hard_fp_offset, frame_size, initial_adjust, callee_offset and
4714 final_offset from HOST_WIDE_INT to poly_int64.
4715 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
4716 to_constant when getting the number of units in an Advanced SIMD
4717 mode.
4718 (aarch64_builtin_vectorized_function): Check for a constant number
4719 of units.
4720 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
4721 GET_MODE_SIZE.
4722 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
4723 attribute instead of GET_MODE_NUNITS.
4724 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
4725 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
4726 GET_MODE_SIZE for fixed-size registers.
4727 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
4728 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
4729 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
4730 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
4731 (aarch64_print_operand, aarch64_print_address_internal)
4732 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
4733 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
4734 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
4735 Handle polynomial GET_MODE_SIZE.
4736 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
4737 wider than SImode without modification.
4738 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
4739 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
4740 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
4741 passing and returning SVE modes.
4742 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
4743 rather than GEN_INT.
4744 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
4745 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
4746 (aarch64_allocate_and_probe_stack_space): Likewise.
4747 (aarch64_layout_frame): Cope with polynomial offsets.
4748 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
4749 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
4750 polynomial offsets.
4751 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
4752 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
4753 poly_int64 rather than a HOST_WIDE_INT.
4754 (aarch64_get_separate_components, aarch64_process_components)
4755 (aarch64_expand_prologue, aarch64_expand_epilogue)
4756 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
4757 (aarch64_anchor_offset): New function, split out from...
4758 (aarch64_legitimize_address): ...here.
4759 (aarch64_builtin_vectorization_cost): Handle polynomial
4760 TYPE_VECTOR_SUBPARTS.
4761 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
4762 GET_MODE_NUNITS.
4763 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
4764 number of elements from the PARALLEL rather than the mode.
4765 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
4766 rather than GET_MODE_BITSIZE.
4767 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
4768 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
4769 (aarch64_expand_vec_perm_const_1): Handle polynomial
4770 d->perm.length () and d->perm elements.
4771 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
4772 Apply to_constant to d->perm elements.
4773 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
4774 polynomial CONST_VECTOR_NUNITS.
4775 (aarch64_move_pointer): Take amount as a poly_int64 rather
4776 than an int.
4777 (aarch64_progress_pointer): Avoid temporary variable.
4778 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
4779 the mode attribute instead of GET_MODE.
4781 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4782 Alan Hayward <alan.hayward@arm.com>
4783 David Sherwood <david.sherwood@arm.com>
4785 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
4786 x exists before using it.
4787 (aarch64_add_constant_internal): Rename to...
4788 (aarch64_add_offset_1): ...this. Replace regnum with separate
4789 src and dest rtxes. Handle the case in which they're different,
4790 including when the offset is zero. Replace scratchreg with an rtx.
4791 Use 2 additions if there is no spare register into which we can
4792 move a 16-bit constant.
4793 (aarch64_add_constant): Delete.
4794 (aarch64_add_offset): Replace reg with separate src and dest
4795 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
4796 Use aarch64_add_offset_1.
4797 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
4798 an rtx rather than an int. Take the delta as a poly_int64
4799 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
4800 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
4801 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
4802 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
4803 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
4804 and aarch64_add_sp.
4805 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
4806 aarch64_add_constant.
4808 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4810 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
4811 Use scalar_float_mode.
4813 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4815 * config/aarch64/aarch64-simd.md
4816 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
4817 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
4818 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
4819 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
4820 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
4821 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
4822 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
4823 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
4824 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
4825 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
4827 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4829 PR target/83514
4830 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
4831 targ_options->x_arm_arch_string is non NULL.
4833 2018-01-11 Tamar Christina <tamar.christina@arm.com>
4835 * config/aarch64/aarch64.h
4836 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
4838 2018-01-11 Sudakshina Das <sudi.das@arm.com>
4840 PR target/82096
4841 * expmed.c (emit_store_flag_force): Swap if const op0
4842 and change VOIDmode to mode of op0.
4844 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4846 PR rtl-optimization/83761
4847 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
4848 than bytes to mode_for_size.
4850 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
4852 PR middle-end/83189
4853 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
4854 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
4855 profile.
4857 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
4859 PR middle-end/83575
4860 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
4861 when in layout mode.
4862 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
4863 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
4864 partition fixup.
4866 2018-01-10 Michael Collison <michael.collison@arm.com>
4868 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
4869 * config/aarch64/aarch64-option-extension.def: Add
4870 AARCH64_OPT_EXTENSION of 'fp16fml'.
4871 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
4872 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
4873 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
4874 * config/aarch64/constraints.md (Ui7): New constraint.
4875 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
4876 (VFMLA_SEL_W): Ditto.
4877 (f16quad): Ditto.
4878 (f16mac1): Ditto.
4879 (VFMLA16_LOW): New int iterator.
4880 (VFMLA16_HIGH): Ditto.
4881 (UNSPEC_FMLAL): New unspec.
4882 (UNSPEC_FMLSL): Ditto.
4883 (UNSPEC_FMLAL2): Ditto.
4884 (UNSPEC_FMLSL2): Ditto.
4885 (f16mac): New code attribute.
4886 * config/aarch64/aarch64-simd-builtins.def
4887 (aarch64_fmlal_lowv2sf): Ditto.
4888 (aarch64_fmlsl_lowv2sf): Ditto.
4889 (aarch64_fmlalq_lowv4sf): Ditto.
4890 (aarch64_fmlslq_lowv4sf): Ditto.
4891 (aarch64_fmlal_highv2sf): Ditto.
4892 (aarch64_fmlsl_highv2sf): Ditto.
4893 (aarch64_fmlalq_highv4sf): Ditto.
4894 (aarch64_fmlslq_highv4sf): Ditto.
4895 (aarch64_fmlal_lane_lowv2sf): Ditto.
4896 (aarch64_fmlsl_lane_lowv2sf): Ditto.
4897 (aarch64_fmlal_laneq_lowv2sf): Ditto.
4898 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
4899 (aarch64_fmlalq_lane_lowv4sf): Ditto.
4900 (aarch64_fmlsl_lane_lowv4sf): Ditto.
4901 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
4902 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
4903 (aarch64_fmlal_lane_highv2sf): Ditto.
4904 (aarch64_fmlsl_lane_highv2sf): Ditto.
4905 (aarch64_fmlal_laneq_highv2sf): Ditto.
4906 (aarch64_fmlsl_laneq_highv2sf): Ditto.
4907 (aarch64_fmlalq_lane_highv4sf): Ditto.
4908 (aarch64_fmlsl_lane_highv4sf): Ditto.
4909 (aarch64_fmlalq_laneq_highv4sf): Ditto.
4910 (aarch64_fmlsl_laneq_highv4sf): Ditto.
4911 * config/aarch64/aarch64-simd.md:
4912 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
4913 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
4914 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
4915 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
4916 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
4917 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
4918 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
4919 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
4920 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
4921 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
4922 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
4923 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
4924 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
4925 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
4926 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
4927 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
4928 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
4929 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
4930 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
4931 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
4932 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
4933 (vfmlsl_low_u32): Ditto.
4934 (vfmlalq_low_u32): Ditto.
4935 (vfmlslq_low_u32): Ditto.
4936 (vfmlal_high_u32): Ditto.
4937 (vfmlsl_high_u32): Ditto.
4938 (vfmlalq_high_u32): Ditto.
4939 (vfmlslq_high_u32): Ditto.
4940 (vfmlal_lane_low_u32): Ditto.
4941 (vfmlsl_lane_low_u32): Ditto.
4942 (vfmlal_laneq_low_u32): Ditto.
4943 (vfmlsl_laneq_low_u32): Ditto.
4944 (vfmlalq_lane_low_u32): Ditto.
4945 (vfmlslq_lane_low_u32): Ditto.
4946 (vfmlalq_laneq_low_u32): Ditto.
4947 (vfmlslq_laneq_low_u32): Ditto.
4948 (vfmlal_lane_high_u32): Ditto.
4949 (vfmlsl_lane_high_u32): Ditto.
4950 (vfmlal_laneq_high_u32): Ditto.
4951 (vfmlsl_laneq_high_u32): Ditto.
4952 (vfmlalq_lane_high_u32): Ditto.
4953 (vfmlslq_lane_high_u32): Ditto.
4954 (vfmlalq_laneq_high_u32): Ditto.
4955 (vfmlslq_laneq_high_u32): Ditto.
4956 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
4957 (AARCH64_FL_FOR_ARCH8_4): New.
4958 (AARCH64_ISA_F16FML): New ISA flag.
4959 (TARGET_F16FML): New feature flag for fp16fml.
4960 (doc/invoke.texi): Document new fp16fml option.
4962 2018-01-10 Michael Collison <michael.collison@arm.com>
4964 * config/aarch64/aarch64-builtins.c:
4965 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
4966 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
4967 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
4968 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
4969 (AARCH64_ISA_SHA3): New ISA flag.
4970 (TARGET_SHA3): New feature flag for sha3.
4971 * config/aarch64/iterators.md (sha512_op): New int attribute.
4972 (CRYPTO_SHA512): New int iterator.
4973 (UNSPEC_SHA512H): New unspec.
4974 (UNSPEC_SHA512H2): Ditto.
4975 (UNSPEC_SHA512SU0): Ditto.
4976 (UNSPEC_SHA512SU1): Ditto.
4977 * config/aarch64/aarch64-simd-builtins.def
4978 (aarch64_crypto_sha512hqv2di): New builtin.
4979 (aarch64_crypto_sha512h2qv2di): Ditto.
4980 (aarch64_crypto_sha512su0qv2di): Ditto.
4981 (aarch64_crypto_sha512su1qv2di): Ditto.
4982 (aarch64_eor3qv8hi): Ditto.
4983 (aarch64_rax1qv2di): Ditto.
4984 (aarch64_xarqv2di): Ditto.
4985 (aarch64_bcaxqv8hi): Ditto.
4986 * config/aarch64/aarch64-simd.md:
4987 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
4988 (aarch64_crypto_sha512su0qv2di): Ditto.
4989 (aarch64_crypto_sha512su1qv2di): Ditto.
4990 (aarch64_eor3qv8hi): Ditto.
4991 (aarch64_rax1qv2di): Ditto.
4992 (aarch64_xarqv2di): Ditto.
4993 (aarch64_bcaxqv8hi): Ditto.
4994 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
4995 (vsha512h2q_u64): Ditto.
4996 (vsha512su0q_u64): Ditto.
4997 (vsha512su1q_u64): Ditto.
4998 (veor3q_u16): Ditto.
4999 (vrax1q_u64): Ditto.
5000 (vxarq_u64): Ditto.
5001 (vbcaxq_u16): Ditto.
5002 * config/arm/types.md (crypto_sha512): New type attribute.
5003 (crypto_sha3): Ditto.
5004 (doc/invoke.texi): Document new sha3 option.
5006 2018-01-10 Michael Collison <michael.collison@arm.com>
5008 * config/aarch64/aarch64-builtins.c:
5009 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5010 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5011 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5012 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5013 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5014 (AARCH64_ISA_SM4): New ISA flag.
5015 (TARGET_SM4): New feature flag for sm4.
5016 * config/aarch64/aarch64-simd-builtins.def
5017 (aarch64_sm3ss1qv4si): Ditto.
5018 (aarch64_sm3tt1aq4si): Ditto.
5019 (aarch64_sm3tt1bq4si): Ditto.
5020 (aarch64_sm3tt2aq4si): Ditto.
5021 (aarch64_sm3tt2bq4si): Ditto.
5022 (aarch64_sm3partw1qv4si): Ditto.
5023 (aarch64_sm3partw2qv4si): Ditto.
5024 (aarch64_sm4eqv4si): Ditto.
5025 (aarch64_sm4ekeyqv4si): Ditto.
5026 * config/aarch64/aarch64-simd.md:
5027 (aarch64_sm3ss1qv4si): Ditto.
5028 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5029 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5030 (aarch64_sm4eqv4si): Ditto.
5031 (aarch64_sm4ekeyqv4si): Ditto.
5032 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5033 (sm3part_op): Ditto.
5034 (CRYPTO_SM3TT): Ditto.
5035 (CRYPTO_SM3PART): Ditto.
5036 (UNSPEC_SM3SS1): New unspec.
5037 (UNSPEC_SM3TT1A): Ditto.
5038 (UNSPEC_SM3TT1B): Ditto.
5039 (UNSPEC_SM3TT2A): Ditto.
5040 (UNSPEC_SM3TT2B): Ditto.
5041 (UNSPEC_SM3PARTW1): Ditto.
5042 (UNSPEC_SM3PARTW2): Ditto.
5043 (UNSPEC_SM4E): Ditto.
5044 (UNSPEC_SM4EKEY): Ditto.
5045 * config/aarch64/constraints.md (Ui2): New constraint.
5046 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5047 * config/arm/types.md (crypto_sm3): New type attribute.
5048 (crypto_sm4): Ditto.
5049 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5050 (vsm3tt1aq_u32): Ditto.
5051 (vsm3tt1bq_u32): Ditto.
5052 (vsm3tt2aq_u32): Ditto.
5053 (vsm3tt2bq_u32): Ditto.
5054 (vsm3partw1q_u32): Ditto.
5055 (vsm3partw2q_u32): Ditto.
5056 (vsm4eq_u32): Ditto.
5057 (vsm4ekeyq_u32): Ditto.
5058 (doc/invoke.texi): Document new sm4 option.
5060 2018-01-10 Michael Collison <michael.collison@arm.com>
5062 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5063 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5064 (AARCH64_FL_FOR_ARCH8_4): New.
5065 (AARCH64_FL_V8_4): New flag.
5066 (doc/invoke.texi): Document new armv8.4-a option.
5068 2018-01-10 Michael Collison <michael.collison@arm.com>
5070 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5071 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5072 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5073 * config/aarch64/aarch64-option-extension.def: Add
5074 AARCH64_OPT_EXTENSION of 'sha2'.
5075 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5076 (crypto): Disable sha2 and aes if crypto disabled.
5077 (crypto): Enable aes and sha2 if enabled.
5078 (simd): Disable sha2 and aes if simd disabled.
5079 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5080 New flags.
5081 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5082 (TARGET_SHA2): New feature flag for sha2.
5083 (TARGET_AES): New feature flag for aes.
5084 * config/aarch64/aarch64-simd.md:
5085 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5086 conditional on TARGET_AES.
5087 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5088 (aarch64_crypto_sha1hsi): Make pattern conditional
5089 on TARGET_SHA2.
5090 (aarch64_crypto_sha1hv4si): Ditto.
5091 (aarch64_be_crypto_sha1hv4si): Ditto.
5092 (aarch64_crypto_sha1su1v4si): Ditto.
5093 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5094 (aarch64_crypto_sha1su0v4si): Ditto.
5095 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5096 (aarch64_crypto_sha256su0v4si): Ditto.
5097 (aarch64_crypto_sha256su1v4si): Ditto.
5098 (doc/invoke.texi): Document new aes and sha2 options.
5100 2018-01-10 Martin Sebor <msebor@redhat.com>
5102 PR tree-optimization/83781
5103 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5104 as string arrays.
5106 2018-01-11 Martin Sebor <msebor@gmail.com>
5107 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5109 PR tree-optimization/83501
5110 PR tree-optimization/81703
5112 * tree-ssa-strlen.c (get_string_cst): Rename...
5113 (get_string_len): ...to this. Handle global constants.
5114 (handle_char_store): Adjust.
5116 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
5117 Jim Wilson <jimw@sifive.com>
5119 * config/riscv/riscv-protos.h (riscv_output_return): New.
5120 * config/riscv/riscv.c (struct machine_function): New naked_p field.
5121 (riscv_attribute_table, riscv_output_return),
5122 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
5123 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
5124 (riscv_compute_frame_info): Only compute frame->mask if not a naked
5125 function.
5126 (riscv_expand_prologue): Add early return for naked function.
5127 (riscv_expand_epilogue): Likewise.
5128 (riscv_function_ok_for_sibcall): Return false for naked function.
5129 (riscv_set_current_function): New.
5130 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
5131 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
5132 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
5133 * doc/extend.texi (RISC-V Function Attributes): New.
5135 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5137 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
5138 check for 128-bit long double before checking TCmode.
5139 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
5140 128-bit long doubles before checking TFmode or TCmode.
5141 (FLOAT128_IBM_P): Likewise.
5143 2018-01-10 Martin Sebor <msebor@redhat.com>
5145 PR tree-optimization/83671
5146 * builtins.c (c_strlen): Unconditionally return zero for the empty
5147 string.
5148 Use -Warray-bounds for warnings.
5149 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
5150 for non-constant array indices with COMPONENT_REF, arrays of
5151 arrays, and pointers to arrays.
5152 (gimple_fold_builtin_strlen): Determine and set length range for
5153 non-constant character arrays.
5155 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
5157 PR middle-end/81897
5158 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
5159 empty blocks.
5161 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
5163 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
5165 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5167 PR target/83399
5168 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
5169 VECTOR_MEM_ALTIVEC_OR_VSX_P.
5170 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
5171 indexed_or_indirect_operand predicate.
5172 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
5173 (*vsx_le_perm_load_v8hi): Likewise.
5174 (*vsx_le_perm_load_v16qi): Likewise.
5175 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
5176 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
5177 (*vsx_le_perm_store_v8hi): Likewise.
5178 (*vsx_le_perm_store_v16qi): Likewise.
5179 (eight unnamed splitters): Likewise.
5181 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5183 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
5184 * config/rs6000/emmintrin.h: Likewise.
5185 * config/rs6000/mmintrin.h: Likewise.
5186 * config/rs6000/xmmintrin.h: Likewise.
5188 2018-01-10 David Malcolm <dmalcolm@redhat.com>
5190 PR c++/43486
5191 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
5192 "public_flag".
5193 * tree.c (tree_nop_conversion): Return true for location wrapper
5194 nodes.
5195 (maybe_wrap_with_location): New function.
5196 (selftest::check_strip_nops): New function.
5197 (selftest::test_location_wrappers): New function.
5198 (selftest::tree_c_tests): Call it.
5199 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
5200 (maybe_wrap_with_location): New decl.
5201 (EXPR_LOCATION_WRAPPER_P): New macro.
5202 (location_wrapper_p): New inline function.
5203 (tree_strip_any_location_wrapper): New inline function.
5205 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
5207 PR target/83735
5208 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
5209 stack_realign_offset for the largest alignment of stack slot
5210 actually used.
5211 (ix86_find_max_used_stack_alignment): New function.
5212 (ix86_finalize_stack_frame_flags): Use it. Set
5213 max_used_stack_alignment if we don't realign stack.
5214 * config/i386/i386.h (machine_function): Add
5215 max_used_stack_alignment.
5217 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
5219 * config/arm/arm.opt (-mbranch-cost): New option.
5220 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
5221 account.
5223 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
5225 PR target/83629
5226 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
5227 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
5229 2018-01-10 Richard Biener <rguenther@suse.de>
5231 PR debug/83765
5232 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
5233 early out so it also covers the case where we have a non-NULL
5234 origin.
5236 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5238 PR tree-optimization/83753
5239 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
5240 for non-strided grouped accesses if the number of elements is 1.
5242 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5244 PR target/81616
5245 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
5246 * i386.h (TARGET_USE_GATHER): Define.
5247 * x86-tune.def (X86_TUNE_USE_GATHER): New.
5249 2018-01-10 Martin Liska <mliska@suse.cz>
5251 PR bootstrap/82831
5252 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
5253 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
5254 partitioning.
5255 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
5256 CLEANUP_NO_PARTITIONING is not set.
5258 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5260 * doc/rtl.texi: Remove documentation of (const ...) wrappers
5261 for vectors, as a partial revert of r254296.
5262 * rtl.h (const_vec_p): Delete.
5263 (const_vec_duplicate_p): Don't test for vector CONSTs.
5264 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
5265 * expmed.c (make_tree): Likewise.
5267 Revert:
5268 * common.md (E, F): Use CONSTANT_P instead of checking for
5269 CONST_VECTOR.
5270 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
5271 checking for CONST_VECTOR.
5273 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5275 PR middle-end/83575
5276 * predict.c (force_edge_cold): Handle in more sane way edges
5277 with no prediction.
5279 2018-01-09 Carl Love <cel@us.ibm.com>
5281 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
5282 V4SI, V4SF types.
5283 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
5284 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
5285 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
5286 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
5287 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
5288 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
5289 * config/rs6000/rs6000-protos.h: Add extern defition for
5290 rs6000_generate_float2_double_code.
5291 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
5292 function.
5293 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
5294 (float2_v2df): Add define_expand.
5296 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
5298 PR target/83628
5299 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
5300 op_mode in the force_to_mode call.
5302 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5304 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
5305 instead of checking each element individually.
5306 (aarch64_evpc_uzp): Likewise.
5307 (aarch64_evpc_zip): Likewise.
5308 (aarch64_evpc_ext): Likewise.
5309 (aarch64_evpc_rev): Likewise.
5310 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
5311 instead of checking each element individually. Return true without
5312 generating rtl if
5313 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
5314 whether all selected elements come from the same input, instead of
5315 checking each element individually. Remove calls to gen_rtx_REG,
5316 start_sequence and end_sequence and instead assert that no rtl is
5317 generated.
5319 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5321 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
5322 order of HIGH and CONST checks.
5324 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5326 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
5327 if the destination isn't an SSA_NAME.
5329 2018-01-09 Richard Biener <rguenther@suse.de>
5331 PR tree-optimization/83668
5332 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
5333 move prologue...
5334 (canonicalize_loop_form): ... here, renamed from ...
5335 (canonicalize_loop_closed_ssa_form): ... this and amended to
5336 swap successor edges for loop exit blocks to make us use
5337 the RPO order we need for initial schedule generation.
5339 2018-01-09 Joseph Myers <joseph@codesourcery.com>
5341 PR tree-optimization/64811
5342 * match.pd: When optimizing comparisons with Inf, avoid
5343 introducing or losing exceptions from comparisons with NaN.
5345 2018-01-09 Martin Liska <mliska@suse.cz>
5347 PR sanitizer/82517
5348 * asan.c (shadow_mem_size): Add gcc_assert.
5350 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
5352 Don't save registers in main().
5354 PR target/83738
5355 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
5356 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
5357 * config/avr/avr.c (avr_set_current_function): Don't error if
5358 naked, OS_task or OS_main are specified at the same time.
5359 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
5360 OS_main.
5361 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
5362 attribute.
5363 * common/config/avr/avr-common.c (avr_option_optimization_table):
5364 Switch on -mmain-is-OS_task for optimizing compilations.
5366 2018-01-09 Richard Biener <rguenther@suse.de>
5368 PR tree-optimization/83572
5369 * graphite.c: Include cfganal.h.
5370 (graphite_transform_loops): Connect infinite loops to exit
5371 and remove fake edges at the end.
5373 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5375 * ipa-inline.c (edge_badness): Revert accidental checkin.
5377 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5379 PR ipa/80763
5380 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
5381 symbols; not inline clones.
5383 2018-01-09 Jakub Jelinek <jakub@redhat.com>
5385 PR target/83507
5386 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
5387 hard registers. Formatting fixes.
5389 PR preprocessor/83722
5390 * gcc.c (try_generate_repro): Pass
5391 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
5392 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
5393 do_report_bug.
5395 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
5396 Kito Cheng <kito.cheng@gmail.com>
5398 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
5399 (riscv_leaf_function_p): Delete.
5400 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
5402 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5404 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
5405 function.
5406 (do_ifelse): New function.
5407 (do_isel): New function.
5408 (do_sub3): New function.
5409 (do_add3): New function.
5410 (do_load_mask_compare): New function.
5411 (do_overlap_load_compare): New function.
5412 (expand_compare_loop): New function.
5413 (expand_block_compare): Call expand_compare_loop() when appropriate.
5414 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
5415 option description.
5416 (-mblock-compare-inline-loop-limit): New option.
5418 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5420 PR target/83677
5421 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
5422 Reverse order of second and third operands in first alternative.
5423 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
5424 of first and second elements in UNSPEC_VPERMR vector.
5425 (altivec_expand_vec_perm_le): Likewise.
5427 2018-01-08 Jeff Law <law@redhat.com>
5429 PR rtl-optimizatin/81308
5430 * tree-switch-conversion.c (cfg_altered): New file scoped static.
5431 (process_switch): If group_case_labels makes a change, then set
5432 cfg_altered.
5433 (pass_convert_switch::execute): If a switch is converted, then
5434 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
5436 PR rtl-optimization/81308
5437 * recog.c (split_all_insns): Conditionally cleanup the CFG after
5438 splitting insns.
5440 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
5442 PR target/83663 - Revert r255946
5443 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
5444 generation for cases where splatting a value is not useful.
5445 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
5446 across a vec_duplicate and a paradoxical subreg forming a vector
5447 mode to a vec_concat.
5449 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5451 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
5452 -march=armv8.3-a variants.
5453 * config/arm/t-multilib: Likewise.
5454 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
5456 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5458 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
5459 to generate rtl.
5460 (cceq_ior_compare_complement): Give it a name so I can use it, and
5461 change boolean_or_operator predicate to boolean_operator so it can
5462 be used to generate a crand.
5463 (eqne): New code iterator.
5464 (bd/bd_neg): New code_attrs.
5465 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
5466 a single define_insn.
5467 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
5468 decrement (bdnzt/bdnzf/bdzt/bdzf).
5469 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
5470 with the new names of the branch decrement patterns, and added the
5471 names of the branch decrement conditional patterns.
5473 2018-01-08 Richard Biener <rguenther@suse.de>
5475 PR tree-optimization/83563
5476 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
5477 cache.
5479 2018-01-08 Richard Biener <rguenther@suse.de>
5481 PR middle-end/83713
5482 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
5484 2018-01-08 Richard Biener <rguenther@suse.de>
5486 PR tree-optimization/83685
5487 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
5488 references to abnormals.
5490 2018-01-08 Richard Biener <rguenther@suse.de>
5492 PR lto/83719
5493 * dwarf2out.c (output_indirect_strings): Handle empty
5494 skeleton_debug_str_hash.
5495 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
5497 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
5499 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
5500 (emit_store_direct): Likewise.
5501 (arc_trampoline_adjust_address): Likewise.
5502 (arc_asm_trampoline_template): New function.
5503 (arc_initialize_trampoline): Use asm_trampoline_template.
5504 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
5505 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
5506 * config/arc/arc.md (flush_icache): Delete pattern.
5508 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
5510 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
5511 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
5512 munaligned-access.
5514 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
5516 PR target/83681
5517 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
5518 by not USED_FOR_TARGET.
5519 (make_pass_resolve_sw_modes): Likewise.
5521 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
5523 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
5524 USED_FOR_TARGET.
5526 2018-01-08 Richard Biener <rguenther@suse.de>
5528 PR middle-end/83580
5529 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
5531 2018-01-08 Richard Biener <rguenther@suse.de>
5533 PR middle-end/83517
5534 * match.pd ((t * 2) / 2) -> t): Add missing :c.
5536 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
5538 PR middle-end/81897
5539 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
5540 basic blocks with a small number of successors.
5541 (convert_control_dep_chain_into_preds): Improve handling of
5542 forwarder blocks.
5543 (dump_predicates): Split apart into...
5544 (dump_pred_chain): ...here...
5545 (dump_pred_info): ...and here.
5546 (can_one_predicate_be_invalidated_p): Add debugging printfs.
5547 (can_chain_union_be_invalidated_p): Improve check for invalidation
5548 of paths.
5549 (uninit_uses_cannot_happen): Avoid unnecessary if
5550 convert_control_dep_chain_into_preds yielded nothing.
5552 2018-01-06 Martin Sebor <msebor@redhat.com>
5554 PR tree-optimization/83640
5555 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
5556 subtracting negative offset from size.
5557 (builtin_access::overlap): Adjust offset bounds of the access to fall
5558 within the size of the object if possible.
5560 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
5562 PR rtl-optimization/83699
5563 * expmed.c (extract_bit_field_1): Restrict the vector usage of
5564 extract_bit_field_as_subreg to cases in which the extracted
5565 value is also a vector.
5567 * lra-constraints.c (process_alt_operands): Test for the equivalence
5568 substitutions when detecting a possible reload cycle.
5570 2018-01-06 Jakub Jelinek <jakub@redhat.com>
5572 PR debug/83480
5573 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
5574 by default if flag_selective_schedling{,2}. Formatting fixes.
5576 PR rtl-optimization/83682
5577 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
5578 if it has non-VECTOR_MODE element mode.
5579 (vec_duplicate_p): Likewise.
5581 PR middle-end/83694
5582 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
5583 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
5585 2018-01-05 Jakub Jelinek <jakub@redhat.com>
5587 PR target/83604
5588 * config/i386/i386-builtin.def
5589 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
5590 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
5591 Require also OPTION_MASK_ISA_AVX512F in addition to
5592 OPTION_MASK_ISA_GFNI.
5593 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
5594 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
5595 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
5596 to OPTION_MASK_ISA_GFNI.
5597 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
5598 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
5599 OPTION_MASK_ISA_AVX512BW.
5600 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
5601 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
5602 addition to OPTION_MASK_ISA_GFNI.
5603 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
5604 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
5605 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
5606 to OPTION_MASK_ISA_GFNI.
5607 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
5608 a requirement for all ISAs rather than any of them with a few
5609 exceptions.
5610 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
5611 processing.
5612 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
5613 bitmasks to be enabled with 3 exceptions, instead of requiring any
5614 enabled ISA with lots of exceptions.
5615 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
5616 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
5617 Change avx512bw in isa attribute to avx512f.
5618 * config/i386/sgxintrin.h: Add license boilerplate.
5619 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
5620 to __AVX512F__ and __AVX512VL to __AVX512VL__.
5621 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
5622 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
5623 defined.
5624 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
5625 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
5626 temporarily sse2 rather than sse if not enabled already.
5628 PR target/83604
5629 * config/i386/sse.md (VI248_VLBW): Rename to ...
5630 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
5631 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
5632 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
5633 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
5634 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
5635 mode iterator instead of VI248_VLBW.
5637 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
5639 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
5640 (record_modified): Skip clobbers; add debug output.
5641 (param_change_prob): Use sreal frequencies.
5643 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
5645 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
5646 punt for user-aligned variables.
5648 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
5650 * tree-chrec.c (chrec_contains_symbols): Return true for
5651 POLY_INT_CST.
5653 2018-01-05 Sudakshina Das <sudi.das@arm.com>
5655 PR target/82439
5656 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
5657 of (x|y) == x for BICS pattern.
5659 2018-01-05 Jakub Jelinek <jakub@redhat.com>
5661 PR tree-optimization/83605
5662 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
5663 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
5664 can throw.
5666 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
5668 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
5669 * config/epiphany/rtems.h: New file.
5671 2018-01-04 Jakub Jelinek <jakub@redhat.com>
5672 Uros Bizjak <ubizjak@gmail.com>
5674 PR target/83554
5675 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
5676 QIreg_operand instead of register_operand predicate.
5677 * config/i386/i386.c (ix86_rop_should_change_byte_p,
5678 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
5679 comments instead of -fmitigate[-_]rop.
5681 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5683 PR bootstrap/81926
5684 * cgraphunit.c (symbol_table::compile): Switch to text_section
5685 before calling assembly_start debug hook.
5686 * run-rtl-passes.c (run_rtl_passes): Likewise.
5687 Include output.h.
5689 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5691 * tree-vrp.c (extract_range_from_binary_expr_1): Check
5692 range_int_cst_p rather than !symbolic_range_p before calling
5693 extract_range_from_multiplicative_op_1.
5695 2018-01-04 Jeff Law <law@redhat.com>
5697 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
5698 redundant test in assertion.
5700 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5702 * doc/rtl.texi: Document machine_mode wrapper classes.
5704 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5706 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
5707 using tree_to_uhwi.
5709 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5711 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
5712 the VEC_PERM_EXPR fold to fail.
5714 2018-01-04 Jakub Jelinek <jakub@redhat.com>
5716 PR debug/83585
5717 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
5718 to switched_sections.
5720 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5722 PR target/83680
5723 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
5724 test for d.testing.
5726 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
5728 PR target/83387
5729 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
5730 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
5732 2018-01-04 Jakub Jelinek <jakub@redhat.com>
5734 PR debug/83666
5735 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
5736 is BLKmode and bitpos not zero or mode change is needed.
5738 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5740 PR target/83675
5741 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
5742 TARGET_VIS2.
5744 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
5746 PR target/83628
5747 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
5748 instead of MULT rtx. Update all corresponding splitters.
5749 (*saddl_se): Ditto.
5750 (*ssub<modesuffix>): Ditto.
5751 (*ssubl_se): Ditto.
5752 (*cmp_sadd_di): Update split patterns.
5753 (*cmp_sadd_si): Ditto.
5754 (*cmp_sadd_sidi): Ditto.
5755 (*cmp_ssub_di): Ditto.
5756 (*cmp_ssub_si): Ditto.
5757 (*cmp_ssub_sidi): Ditto.
5758 * config/alpha/predicates.md (const23_operand): New predicate.
5759 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
5760 Look for ASHIFT, not MULT inner operand.
5761 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
5763 2018-01-04 Martin Liska <mliska@suse.cz>
5765 PR gcov-profile/83669
5766 * gcov.c (output_intermediate_file): Add version to intermediate
5767 gcov file.
5768 * doc/gcov.texi: Document new field 'version' in intermediate
5769 file format. Fix location of '-k' option of gcov command.
5771 2018-01-04 Martin Liska <mliska@suse.cz>
5773 PR ipa/82352
5774 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
5776 2018-01-04 Jakub Jelinek <jakub@redhat.com>
5778 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
5780 2018-01-03 Martin Sebor <msebor@redhat.com>
5782 PR tree-optimization/83655
5783 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
5784 checking calls with invalid arguments.
5786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5788 * tree-vect-stmts.c (vect_get_store_rhs): New function.
5789 (vectorizable_mask_load_store): Delete.
5790 (vectorizable_call): Return false for masked loads and stores.
5791 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
5792 instead of gimple_assign_rhs1.
5793 (vectorizable_load): Handle IFN_MASK_LOAD.
5794 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
5796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5798 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
5799 split out from..,
5800 (vectorizable_mask_load_store): ...here.
5801 (vectorizable_load): ...and here.
5803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5805 * tree-vect-stmts.c (vect_build_all_ones_mask)
5806 (vect_build_zero_merge_argument): New functions, split out from...
5807 (vectorizable_load): ...here.
5809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5811 * tree-vect-stmts.c (vect_check_store_rhs): New function,
5812 split out from...
5813 (vectorizable_mask_load_store): ...here.
5814 (vectorizable_store): ...and here.
5816 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5818 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
5819 split out from...
5820 (vectorizable_mask_load_store): ...here.
5822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5824 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
5825 (vect_model_store_cost): Take a vec_load_store_type instead of a
5826 vect_def_type.
5827 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
5828 (vect_model_store_cost): Take a vec_load_store_type instead of a
5829 vect_def_type.
5830 (vectorizable_mask_load_store): Update accordingly.
5831 (vectorizable_store): Likewise.
5832 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
5834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5836 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
5837 IFN_MASK_LOAD calls here rather than...
5838 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
5840 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5841 Alan Hayward <alan.hayward@arm.com>
5842 David Sherwood <david.sherwood@arm.com>
5844 * expmed.c (extract_bit_field_1): For vector extracts,
5845 fall back to extract_bit_field_as_subreg if vec_extract
5846 isn't available.
5848 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5849 Alan Hayward <alan.hayward@arm.com>
5850 David Sherwood <david.sherwood@arm.com>
5852 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
5853 they are variable or constant sized.
5854 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
5855 slots for constant-sized data.
5857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5858 Alan Hayward <alan.hayward@arm.com>
5859 David Sherwood <david.sherwood@arm.com>
5861 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
5862 handling COND_EXPRs with boolean comparisons, try to find a better
5863 basis for the mask type than the boolean itself.
5865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5867 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
5868 is calculated and how it can be overridden.
5869 * genmodes.c (max_bitsize_mode_any_mode): New variable.
5870 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
5871 if defined.
5872 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
5873 if nonzero.
5875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5876 Alan Hayward <alan.hayward@arm.com>
5877 David Sherwood <david.sherwood@arm.com>
5879 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
5880 Remove the mode argument.
5881 (aarch64_simd_valid_immediate): Remove the mode and inverse
5882 arguments.
5883 * config/aarch64/iterators.md (bitsize): New iterator.
5884 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
5885 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
5886 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
5887 aarch64_simd_valid_immediate.
5888 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
5889 (aarch64_reg_or_bic_imm): Likewise.
5890 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
5891 with an insn_type enum and msl with a modifier_type enum.
5892 Replace element_width with a scalar_mode. Change the shift
5893 to unsigned int. Add constructors for scalar_float_mode and
5894 scalar_int_mode elements.
5895 (aarch64_vect_float_const_representable_p): Delete.
5896 (aarch64_can_const_movi_rtx_p)
5897 (aarch64_simd_scalar_immediate_valid_for_move)
5898 (aarch64_simd_make_constant): Update call to
5899 aarch64_simd_valid_immediate.
5900 (aarch64_advsimd_valid_immediate_hs): New function.
5901 (aarch64_advsimd_valid_immediate): Likewise.
5902 (aarch64_simd_valid_immediate): Remove mode and inverse
5903 arguments. Rewrite to use the above. Use const_vec_duplicate_p
5904 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
5905 and aarch64_float_const_representable_p on the result.
5906 (aarch64_output_simd_mov_immediate): Remove mode argument.
5907 Update call to aarch64_simd_valid_immediate and use of
5908 simd_immediate_info.
5909 (aarch64_output_scalar_simd_mov_immediate): Update call
5910 accordingly.
5912 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5913 Alan Hayward <alan.hayward@arm.com>
5914 David Sherwood <david.sherwood@arm.com>
5916 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
5917 (mode_nunits): Likewise CONST_MODE_NUNITS.
5918 * machmode.def (ADJUST_NUNITS): Document.
5919 * genmodes.c (mode_data::need_nunits_adj): New field.
5920 (blank_mode): Update accordingly.
5921 (adj_nunits): New variable.
5922 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
5923 parameter.
5924 (emit_mode_size_inline): Set need_bytesize_adj for all modes
5925 listed in adj_nunits.
5926 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
5927 listed in adj_nunits. Don't emit case statements for such modes.
5928 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
5929 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
5930 nothing if adj_nunits is nonnull.
5931 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
5932 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
5933 (emit_mode_fbit): Update use of print_maybe_const_decl.
5934 (emit_move_size): Likewise. Treat the array as non-const
5935 if adj_nunits.
5936 (emit_mode_adjustments): Handle adj_nunits.
5938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5940 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
5941 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
5942 (VECTOR_MODES): Use it.
5943 (make_vector_modes): Take the prefix as an argument.
5945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5946 Alan Hayward <alan.hayward@arm.com>
5947 David Sherwood <david.sherwood@arm.com>
5949 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
5950 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
5951 for MODE_VECTOR_BOOL.
5952 * machmode.def (VECTOR_BOOL_MODE): Document.
5953 * genmodes.c (VECTOR_BOOL_MODE): New macro.
5954 (make_vector_bool_mode): New function.
5955 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
5956 MODE_VECTOR_BOOL.
5957 * lto-streamer-in.c (lto_input_mode_table): Likewise.
5958 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
5959 Likewise.
5960 * stor-layout.c (int_mode_for_mode): Likewise.
5961 * tree.c (build_vector_type_for_mode): Likewise.
5962 * varasm.c (output_constant_pool_2): Likewise.
5963 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
5964 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
5965 for MODE_VECTOR_BOOL.
5966 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
5967 of mode class checks.
5968 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
5969 instead of a list of mode class checks.
5970 (expand_vector_scalar_condition): Likewise.
5971 (type_for_widest_vector_mode): Handle BImode as an inner mode.
5973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5974 Alan Hayward <alan.hayward@arm.com>
5975 David Sherwood <david.sherwood@arm.com>
5977 * machmode.h (mode_size): Change from unsigned short to
5978 poly_uint16_pod.
5979 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
5980 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
5981 or if measurement_type is not polynomial.
5982 (fixed_size_mode::includes_p): Check for constant-sized modes.
5983 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
5984 return a poly_uint16 rather than an unsigned short.
5985 (emit_mode_size): Change the type of mode_size from unsigned short
5986 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
5987 (emit_mode_adjustments): Cope with polynomial vector sizes.
5988 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
5989 for GET_MODE_SIZE.
5990 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
5991 for GET_MODE_SIZE.
5992 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
5993 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
5994 * caller-save.c (setup_save_areas): Likewise.
5995 (replace_reg_with_saved_mem): Likewise.
5996 * calls.c (emit_library_call_value_1): Likewise.
5997 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
5998 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
5999 (gen_lowpart_for_combine): Likewise.
6000 * convert.c (convert_to_integer_1): Likewise.
6001 * cse.c (equiv_constant, cse_insn): Likewise.
6002 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6003 (cselib_subst_to_values): Likewise.
6004 * dce.c (word_dce_process_block): Likewise.
6005 * df-problems.c (df_word_lr_mark_ref): Likewise.
6006 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6007 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6008 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6009 (rtl_for_decl_location): Likewise.
6010 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6011 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6012 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6013 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6014 (expand_expr_real_1): Likewise.
6015 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6016 (pad_below): Likewise.
6017 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6018 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6019 * ira.c (get_subreg_tracking_sizes): Likewise.
6020 * ira-build.c (ira_create_allocno_objects): Likewise.
6021 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6022 (ira_sort_regnos_for_alter_reg): Likewise.
6023 * ira-costs.c (record_operand_costs): Likewise.
6024 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6025 (resolve_simple_move): Likewise.
6026 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6027 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6028 (lra_constraints): Likewise.
6029 (CONST_POOL_OK_P): Reject variable-sized modes.
6030 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6031 (add_pseudo_to_slot, lra_spill): Likewise.
6032 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6033 * optabs-query.c (get_best_extraction_insn): Likewise.
6034 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6035 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6036 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6037 * recog.c (offsettable_address_addr_space_p): Likewise.
6038 * regcprop.c (maybe_mode_change): Likewise.
6039 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6040 * regrename.c (build_def_use): Likewise.
6041 * regstat.c (dump_reg_info): Likewise.
6042 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6043 (find_reloads, find_reloads_subreg_address): Likewise.
6044 * reload1.c (eliminate_regs_1): Likewise.
6045 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6046 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6047 (simplify_binary_operation_1, simplify_subreg): Likewise.
6048 * targhooks.c (default_function_arg_padding): Likewise.
6049 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6050 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6051 (verify_gimple_assign_ternary): Likewise.
6052 * tree-inline.c (estimate_move_cost): Likewise.
6053 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6054 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6055 (get_address_cost_ainc): Likewise.
6056 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6057 (vect_supportable_dr_alignment): Likewise.
6058 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6059 (vectorizable_reduction): Likewise.
6060 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6061 (vectorizable_operation, vectorizable_load): Likewise.
6062 * tree.c (build_same_sized_truth_vector_type): Likewise.
6063 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6064 * var-tracking.c (emit_note_insn_var_location): Likewise.
6065 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6066 (ADDR_VEC_ALIGN): Likewise.
6068 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6069 Alan Hayward <alan.hayward@arm.com>
6070 David Sherwood <david.sherwood@arm.com>
6072 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6073 unsigned short.
6074 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6075 or if measurement_type is polynomial.
6076 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6077 * combine.c (make_extraction): Likewise.
6078 * dse.c (find_shift_sequence): Likewise.
6079 * dwarf2out.c (mem_loc_descriptor): Likewise.
6080 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6081 (extract_bit_field, extract_low_bits): Likewise.
6082 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6083 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6084 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6085 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6086 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6087 * reload.c (find_reloads): Likewise.
6088 * reload1.c (alter_reg): Likewise.
6089 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6090 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6091 * tree-if-conv.c (predicate_mem_writes): Likewise.
6092 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6093 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6094 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6095 * valtrack.c (dead_debug_insert_temp): Likewise.
6096 * varasm.c (mergeable_constant_section): Likewise.
6097 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
6099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6100 Alan Hayward <alan.hayward@arm.com>
6101 David Sherwood <david.sherwood@arm.com>
6103 * expr.c (expand_assignment): Cope with polynomial mode sizes
6104 when assigning to a CONCAT.
6106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6107 Alan Hayward <alan.hayward@arm.com>
6108 David Sherwood <david.sherwood@arm.com>
6110 * machmode.h (mode_precision): Change from unsigned short to
6111 poly_uint16_pod.
6112 (mode_to_precision): Return a poly_uint16 rather than an unsigned
6113 short.
6114 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
6115 or if measurement_type is not polynomial.
6116 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
6117 in which the mode is already known to be a scalar_int_mode.
6118 * genmodes.c (emit_mode_precision): Change the type of mode_precision
6119 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
6120 initializer.
6121 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6122 for GET_MODE_PRECISION.
6123 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6124 for GET_MODE_PRECISION.
6125 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
6126 as polynomial.
6127 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
6128 (expand_field_assignment, make_extraction): Likewise.
6129 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
6130 (get_last_value): Likewise.
6131 * convert.c (convert_to_integer_1): Likewise.
6132 * cse.c (cse_insn): Likewise.
6133 * expr.c (expand_expr_real_1): Likewise.
6134 * lra-constraints.c (simplify_operand_subreg): Likewise.
6135 * optabs-query.c (can_atomic_load_p): Likewise.
6136 * optabs.c (expand_atomic_load): Likewise.
6137 (expand_atomic_store): Likewise.
6138 * ree.c (combine_reaching_defs): Likewise.
6139 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
6140 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
6141 * tree.h (type_has_mode_precision_p): Likewise.
6142 * ubsan.c (instrument_si_overflow): Likewise.
6144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6145 Alan Hayward <alan.hayward@arm.com>
6146 David Sherwood <david.sherwood@arm.com>
6148 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
6149 polynomial numbers of units.
6150 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
6151 (valid_vector_subparts_p): New function.
6152 (build_vector_type): Remove temporary shim and take the number
6153 of units as a poly_uint64 rather than an int.
6154 (build_opaque_vector_type): Take the number of units as a
6155 poly_uint64 rather than an int.
6156 * tree.c (build_vector_from_ctor): Handle polynomial
6157 TYPE_VECTOR_SUBPARTS.
6158 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
6159 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
6160 (build_vector_from_val): If the number of units is variable,
6161 use build_vec_duplicate_cst for constant operands and
6162 VEC_DUPLICATE_EXPR otherwise.
6163 (make_vector_type): Remove temporary is_constant ().
6164 (build_vector_type, build_opaque_vector_type): Take the number of
6165 units as a poly_uint64 rather than an int.
6166 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
6167 VECTOR_CST_NELTS.
6168 * cfgexpand.c (expand_debug_expr): Likewise.
6169 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
6170 (store_constructor, expand_expr_real_1): Likewise.
6171 (const_scalar_mask_from_tree): Likewise.
6172 * fold-const-call.c (fold_const_reduction): Likewise.
6173 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
6174 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
6175 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
6176 (fold_relational_const): Likewise.
6177 (native_interpret_vector): Likewise. Change the size from an
6178 int to an unsigned int.
6179 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
6180 TYPE_VECTOR_SUBPARTS.
6181 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
6182 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
6183 duplicating a non-constant operand into a variable-length vector.
6184 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
6185 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
6186 * ipa-icf.c (sem_variable::equals): Likewise.
6187 * match.pd: Likewise.
6188 * omp-simd-clone.c (simd_clone_subparts): Likewise.
6189 * print-tree.c (print_node): Likewise.
6190 * stor-layout.c (layout_type): Likewise.
6191 * targhooks.c (default_builtin_vectorization_cost): Likewise.
6192 * tree-cfg.c (verify_gimple_comparison): Likewise.
6193 (verify_gimple_assign_binary): Likewise.
6194 (verify_gimple_assign_ternary): Likewise.
6195 (verify_gimple_assign_single): Likewise.
6196 * tree-pretty-print.c (dump_generic_node): Likewise.
6197 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6198 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
6199 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
6200 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
6201 (vect_shift_permute_load_chain): Likewise.
6202 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
6203 (expand_vector_condition, optimize_vector_constructor): Likewise.
6204 (lower_vec_perm, get_compute_type): Likewise.
6205 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6206 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
6207 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
6208 (vect_recog_mask_conversion_pattern): Likewise.
6209 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
6210 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
6211 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6212 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
6213 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
6214 (vectorizable_shift, vectorizable_operation, vectorizable_store)
6215 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
6216 (supportable_widening_operation): Likewise.
6217 (supportable_narrowing_operation): Likewise.
6218 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
6219 Likewise.
6220 * varasm.c (output_constant): Likewise.
6222 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6223 Alan Hayward <alan.hayward@arm.com>
6224 David Sherwood <david.sherwood@arm.com>
6226 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
6227 so that both the length == 3 and length != 3 cases set up their
6228 own permute vectors. Add comments explaining why we know the
6229 number of elements is constant.
6230 (vect_permute_load_chain): Likewise.
6232 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6233 Alan Hayward <alan.hayward@arm.com>
6234 David Sherwood <david.sherwood@arm.com>
6236 * machmode.h (mode_nunits): Change from unsigned char to
6237 poly_uint16_pod.
6238 (ONLY_FIXED_SIZE_MODES): New macro.
6239 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
6240 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
6241 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
6242 New typedefs.
6243 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
6244 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
6245 or if measurement_type is not polynomial.
6246 * genmodes.c (ZERO_COEFFS): New macro.
6247 (emit_mode_nunits_inline): Make mode_nunits_inline return a
6248 poly_uint16.
6249 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
6250 Use ZERO_COEFFS when emitting initializers.
6251 * data-streamer.h (bp_pack_poly_value): New function.
6252 (bp_unpack_poly_value): Likewise.
6253 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6254 for GET_MODE_NUNITS.
6255 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6256 for GET_MODE_NUNITS.
6257 * tree.c (make_vector_type): Remove temporary shim and make
6258 the real function take the number of units as a poly_uint64
6259 rather than an int.
6260 (build_vector_type_for_mode): Handle polynomial nunits.
6261 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
6262 * emit-rtl.c (const_vec_series_p_1): Likewise.
6263 (gen_rtx_CONST_VECTOR): Likewise.
6264 * fold-const.c (test_vec_duplicate_folding): Likewise.
6265 * genrecog.c (validate_pattern): Likewise.
6266 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
6267 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6268 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
6269 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
6270 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
6271 * rtlanal.c (subreg_get_info): Likewise.
6272 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
6273 (vect_grouped_load_supported): Likewise.
6274 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
6275 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
6276 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
6277 (simplify_const_unary_operation, simplify_binary_operation_1)
6278 (simplify_const_binary_operation, simplify_ternary_operation)
6279 (test_vector_ops_duplicate, test_vector_ops): Likewise.
6280 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
6281 instead of CONST_VECTOR_NUNITS.
6282 * varasm.c (output_constant_pool_2): Likewise.
6283 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
6284 explicit-encoded elements in the XVEC for variable-length vectors.
6286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6288 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
6290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6291 Alan Hayward <alan.hayward@arm.com>
6292 David Sherwood <david.sherwood@arm.com>
6294 * coretypes.h (fixed_size_mode): Declare.
6295 (fixed_size_mode_pod): New typedef.
6296 * builtins.h (target_builtins::x_apply_args_mode)
6297 (target_builtins::x_apply_result_mode): Change type to
6298 fixed_size_mode_pod.
6299 * builtins.c (apply_args_size, apply_result_size, result_vector)
6300 (expand_builtin_apply_args_1, expand_builtin_apply)
6301 (expand_builtin_return): Update accordingly.
6303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6305 * cse.c (hash_rtx_cb): Hash only the encoded elements.
6306 * cselib.c (cselib_hash_rtx): Likewise.
6307 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
6308 CONST_VECTOR encoding.
6310 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6311 Jeff Law <law@redhat.com>
6313 PR target/83641
6314 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
6315 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
6316 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
6317 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
6319 PR target/83641
6320 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
6321 explicitly probe *sp in a noreturn function if there were any callee
6322 register saves or frame pointer is needed.
6324 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6326 PR debug/83621
6327 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
6328 BLKmode for ternary, binary or unary expressions.
6330 PR debug/83645
6331 * var-tracking.c (delete_vta_debug_insn): New inline function.
6332 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
6333 insns from get_insns () to NULL instead of each bb separately.
6334 Use delete_vta_debug_insn. No longer static.
6335 (vt_debug_insns_local, variable_tracking_main_1): Adjust
6336 delete_vta_debug_insns callers.
6337 * rtl.h (delete_vta_debug_insns): Declare.
6338 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
6339 instead of variable_tracking_main.
6341 2018-01-03 Martin Sebor <msebor@redhat.com>
6343 PR tree-optimization/83603
6344 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
6345 arguments past the endof the argument list in functions declared
6346 without a prototype.
6347 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
6348 Avoid checking when arguments are null.
6350 2018-01-03 Martin Sebor <msebor@redhat.com>
6352 PR c/83559
6353 * doc/extend.texi (attribute const): Fix a typo.
6354 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
6355 issuing -Wsuggest-attribute for void functions.
6357 2018-01-03 Martin Sebor <msebor@redhat.com>
6359 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
6360 offset_int::from instead of wide_int::to_shwi.
6361 (maybe_diag_overlap): Remove assertion.
6362 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
6363 * gimple-ssa-sprintf.c (format_directive): Same.
6364 (parse_directive): Same.
6365 (sprintf_dom_walker::compute_format_length): Same.
6366 (try_substitute_return_value): Same.
6368 2018-01-03 Jeff Law <law@redhat.com>
6370 PR middle-end/83654
6371 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
6372 non-constant residual for zero at runtime and avoid probing in
6373 that case. Reorganize code for trailing problem to mirror handling
6374 of the residual.
6376 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6378 PR tree-optimization/83501
6379 * tree-ssa-strlen.c (get_string_cst): New.
6380 (handle_char_store): Call get_string_cst.
6382 2018-01-03 Martin Liska <mliska@suse.cz>
6384 PR tree-optimization/83593
6385 * tree-ssa-strlen.c: Include tree-cfg.h.
6386 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
6387 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
6388 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
6389 to false.
6390 (strlen_dom_walker::before_dom_children): Call
6391 gimple_purge_dead_eh_edges. Dump tranformation with details
6392 dump flags.
6393 (strlen_dom_walker::before_dom_children): Update call by adding
6394 new argument cleanup_eh.
6395 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
6397 2018-01-03 Martin Liska <mliska@suse.cz>
6399 PR ipa/83549
6400 * cif-code.def (VARIADIC_THUNK): New enum value.
6401 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
6402 thunks.
6404 2018-01-03 Jan Beulich <jbeulich@suse.com>
6406 * sse.md (mov<mode>_internal): Tighten condition for when to use
6407 vmovdqu<ssescalarsize> for TI and OI modes.
6409 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6411 Update copyright years.
6413 2018-01-03 Martin Liska <mliska@suse.cz>
6415 PR ipa/83594
6416 * ipa-visibility.c (function_and_variable_visibility): Skip
6417 functions with noipa attribure.
6419 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6421 * gcc.c (process_command): Update copyright notice dates.
6422 * gcov-dump.c (print_version): Ditto.
6423 * gcov.c (print_version): Ditto.
6424 * gcov-tool.c (print_version): Ditto.
6425 * gengtype.c (create_file): Ditto.
6426 * doc/cpp.texi: Bump @copying's copyright year.
6427 * doc/cppinternals.texi: Ditto.
6428 * doc/gcc.texi: Ditto.
6429 * doc/gccint.texi: Ditto.
6430 * doc/gcov.texi: Ditto.
6431 * doc/install.texi: Ditto.
6432 * doc/invoke.texi: Ditto.
6434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6436 * vector-builder.h (vector_builder::m_full_nelts): Change from
6437 unsigned int to poly_uint64.
6438 (vector_builder::full_nelts): Update prototype accordingly.
6439 (vector_builder::new_vector): Likewise.
6440 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
6441 (vector_builder::operator ==): Likewise.
6442 (vector_builder::finalize): Likewise.
6443 * int-vector-builder.h (int_vector_builder::int_vector_builder):
6444 Take the number of elements as a poly_uint64 rather than an
6445 unsigned int.
6446 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
6447 from unsigned int to poly_uint64.
6448 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
6449 (vec_perm_indices::new_vector): Likewise.
6450 (vec_perm_indices::length): Likewise.
6451 (vec_perm_indices::nelts_per_input): Likewise.
6452 (vec_perm_indices::input_nelts): Likewise.
6453 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
6454 number of elements per input as a poly_uint64 rather than an
6455 unsigned int. Use the original encoding for variable-length
6456 vectors, rather than clamping each individual element.
6457 For the second and subsequent elements in each pattern,
6458 clamp the step and base before clamping their sum.
6459 (vec_perm_indices::series_p): Handle polynomial element counts.
6460 (vec_perm_indices::all_in_range_p): Likewise.
6461 (vec_perm_indices_to_tree): Likewise.
6462 (vec_perm_indices_to_rtx): Likewise.
6463 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
6464 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
6465 (tree_vector_builder::new_binary_operation): Handle polynomial
6466 element counts. Return false if we need to know the number
6467 of elements at compile time.
6468 * fold-const.c (fold_vec_perm): Punt if the number of elements
6469 isn't known at compile time.
6471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6473 * vec-perm-indices.h (vec_perm_builder): Change element type
6474 from HOST_WIDE_INT to poly_int64.
6475 (vec_perm_indices::element_type): Update accordingly.
6476 (vec_perm_indices::clamp): Handle polynomial element_types.
6477 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
6478 (vec_perm_indices::all_in_range_p): Likewise.
6479 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
6480 than shwi trees.
6481 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
6482 polynomial vec_perm_indices element types.
6483 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
6484 * fold-const.c (fold_vec_perm): Likewise.
6485 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
6486 * tree-vect-generic.c (lower_vec_perm): Likewise.
6487 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
6488 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
6489 element type to HOST_WIDE_INT.
6491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6492 Alan Hayward <alan.hayward@arm.com>
6493 David Sherwood <david.sherwood@arm.com>
6495 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
6496 rather than an int. Use plus_constant.
6497 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
6498 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
6500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6501 Alan Hayward <alan.hayward@arm.com>
6502 David Sherwood <david.sherwood@arm.com>
6504 * calls.c (emit_call_1, expand_call): Change struct_value_size from
6505 a HOST_WIDE_INT to a poly_int64.
6507 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6508 Alan Hayward <alan.hayward@arm.com>
6509 David Sherwood <david.sherwood@arm.com>
6511 * calls.c (load_register_parameters): Cope with polynomial
6512 mode sizes. Require a constant size for BLKmode parameters
6513 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
6514 forces a parameter to be padded at the lsb end in order to
6515 fill a complete number of words, require the parameter size
6516 to be ordered wrt UNITS_PER_WORD.
6518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6519 Alan Hayward <alan.hayward@arm.com>
6520 David Sherwood <david.sherwood@arm.com>
6522 * reload1.c (spill_stack_slot_width): Change element type
6523 from unsigned int to poly_uint64_pod.
6524 (alter_reg): Treat mode sizes as polynomial.
6526 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6527 Alan Hayward <alan.hayward@arm.com>
6528 David Sherwood <david.sherwood@arm.com>
6530 * reload.c (complex_word_subreg_p): New function.
6531 (reload_inner_reg_of_subreg, push_reload): Use it.
6533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6534 Alan Hayward <alan.hayward@arm.com>
6535 David Sherwood <david.sherwood@arm.com>
6537 * lra-constraints.c (process_alt_operands): Reject matched
6538 operands whose sizes aren't ordered.
6539 (match_reload): Refer to this check here.
6541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6542 Alan Hayward <alan.hayward@arm.com>
6543 David Sherwood <david.sherwood@arm.com>
6545 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
6546 that the mode size is in the set {1, 2, 4, 8, 16}.
6548 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6549 Alan Hayward <alan.hayward@arm.com>
6550 David Sherwood <david.sherwood@arm.com>
6552 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
6553 Use plus_constant instead of gen_rtx_PLUS.
6555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6556 Alan Hayward <alan.hayward@arm.com>
6557 David Sherwood <david.sherwood@arm.com>
6559 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
6560 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
6561 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
6562 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
6563 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
6564 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
6565 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
6566 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
6567 * config/i386/i386.c (ix86_push_rounding): ...this new function.
6568 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
6569 a poly_int64.
6570 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
6571 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
6572 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
6573 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
6574 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
6575 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
6576 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
6577 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
6578 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
6579 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
6580 function.
6581 * expr.c (emit_move_resolve_push): Treat the input and result
6582 of PUSH_ROUNDING as a poly_int64.
6583 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
6584 (emit_push_insn): Likewise.
6585 * lra-eliminations.c (mark_not_eliminable): Likewise.
6586 * recog.c (push_operand): Likewise.
6587 * reload1.c (elimination_effects): Likewise.
6588 * rtlanal.c (nonzero_bits1): Likewise.
6589 * calls.c (store_one_arg): Likewise. Require the padding to be
6590 known at compile time.
6592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6593 Alan Hayward <alan.hayward@arm.com>
6594 David Sherwood <david.sherwood@arm.com>
6596 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
6597 Use plus_constant instead of gen_rtx_PLUS.
6599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6600 Alan Hayward <alan.hayward@arm.com>
6601 David Sherwood <david.sherwood@arm.com>
6603 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
6604 rather than an int.
6606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6607 Alan Hayward <alan.hayward@arm.com>
6608 David Sherwood <david.sherwood@arm.com>
6610 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
6611 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
6612 via stack temporaries. Treat the mode size as polynomial too.
6614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6615 Alan Hayward <alan.hayward@arm.com>
6616 David Sherwood <david.sherwood@arm.com>
6618 * expr.c (expand_expr_real_2): When handling conversions involving
6619 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
6620 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
6621 as a poly_uint64 too.
6623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6624 Alan Hayward <alan.hayward@arm.com>
6625 David Sherwood <david.sherwood@arm.com>
6627 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
6629 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6630 Alan Hayward <alan.hayward@arm.com>
6631 David Sherwood <david.sherwood@arm.com>
6633 * combine.c (can_change_dest_mode): Handle polynomial
6634 REGMODE_NATURAL_SIZE.
6635 * expmed.c (store_bit_field_1): Likewise.
6636 * expr.c (store_constructor): Likewise.
6637 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
6638 and polynomial REGMODE_NATURAL_SIZE.
6639 (gen_lowpart_common): Likewise.
6640 * reginfo.c (record_subregs_of_mode): Likewise.
6641 * rtlanal.c (read_modify_subreg_p): Likewise.
6643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6644 Alan Hayward <alan.hayward@arm.com>
6645 David Sherwood <david.sherwood@arm.com>
6647 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
6648 numbers of elements.
6650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6651 Alan Hayward <alan.hayward@arm.com>
6652 David Sherwood <david.sherwood@arm.com>
6654 * match.pd: Cope with polynomial numbers of vector elements.
6656 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6657 Alan Hayward <alan.hayward@arm.com>
6658 David Sherwood <david.sherwood@arm.com>
6660 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
6661 in a POINTER_PLUS_EXPR.
6663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6664 Alan Hayward <alan.hayward@arm.com>
6665 David Sherwood <david.sherwood@arm.com>
6667 * omp-simd-clone.c (simd_clone_subparts): New function.
6668 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
6669 (ipa_simd_modify_function_body): Likewise.
6671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6672 Alan Hayward <alan.hayward@arm.com>
6673 David Sherwood <david.sherwood@arm.com>
6675 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
6676 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
6677 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
6678 (expand_vector_condition, vector_element): Likewise.
6679 (subparts_gt): New function.
6680 (get_compute_type): Use subparts_gt.
6681 (count_type_subparts): Delete.
6682 (expand_vector_operations_1): Use subparts_gt instead of
6683 count_type_subparts.
6685 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6686 Alan Hayward <alan.hayward@arm.com>
6687 David Sherwood <david.sherwood@arm.com>
6689 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
6690 (vect_compile_time_alias): ...this new function. Do the calculation
6691 on poly_ints rather than trees.
6692 (vect_prune_runtime_alias_test_list): Update call accordingly.
6694 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6695 Alan Hayward <alan.hayward@arm.com>
6696 David Sherwood <david.sherwood@arm.com>
6698 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
6699 numbers of units.
6700 (vect_schedule_slp_instance): Likewise.
6702 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6703 Alan Hayward <alan.hayward@arm.com>
6704 David Sherwood <david.sherwood@arm.com>
6706 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
6707 constant and extern definitions for variable-length vectors.
6708 (vect_get_constant_vectors): Note that the number of units
6709 is known to be constant.
6711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6712 Alan Hayward <alan.hayward@arm.com>
6713 David Sherwood <david.sherwood@arm.com>
6715 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
6716 of units as polynomial. Choose between WIDE and NARROW based
6717 on multiple_p.
6719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6720 Alan Hayward <alan.hayward@arm.com>
6721 David Sherwood <david.sherwood@arm.com>
6723 * tree-vect-stmts.c (simd_clone_subparts): New function.
6724 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
6726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6727 Alan Hayward <alan.hayward@arm.com>
6728 David Sherwood <david.sherwood@arm.com>
6730 * tree-vect-stmts.c (vectorizable_call): Treat the number of
6731 vectors as polynomial. Use build_index_vector for
6732 IFN_GOMP_SIMD_LANE.
6734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6735 Alan Hayward <alan.hayward@arm.com>
6736 David Sherwood <david.sherwood@arm.com>
6738 * tree-vect-stmts.c (get_load_store_type): Treat the number of
6739 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
6740 for variable-length vectors.
6741 (vectorizable_mask_load_store): Treat the number of units as
6742 polynomial, asserting that it is constant if the condition has
6743 already been enforced.
6744 (vectorizable_store, vectorizable_load): Likewise.
6746 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6747 Alan Hayward <alan.hayward@arm.com>
6748 David Sherwood <david.sherwood@arm.com>
6750 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
6751 of units as polynomial. Punt if we can't tell at compile time
6752 which vector contains the final result.
6754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6755 Alan Hayward <alan.hayward@arm.com>
6756 David Sherwood <david.sherwood@arm.com>
6758 * tree-vect-loop.c (vectorizable_induction): Treat the number
6759 of units as polynomial. Punt on SLP inductions. Use an integer
6760 VEC_SERIES_EXPR for variable-length integer reductions. Use a
6761 cast of such a series for variable-length floating-point
6762 reductions.
6764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6765 Alan Hayward <alan.hayward@arm.com>
6766 David Sherwood <david.sherwood@arm.com>
6768 * tree.h (build_index_vector): Declare.
6769 * tree.c (build_index_vector): New function.
6770 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
6771 of units as polynomial, forcibly converting it to a constant if
6772 vectorizable_reduction has already enforced the condition.
6773 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
6774 to create a {1,2,3,...} vector.
6775 (vectorizable_reduction): Treat the number of units as polynomial.
6776 Choose vectype_in based on the largest scalar element size rather
6777 than the smallest number of units. Enforce the restrictions
6778 relied on above.
6780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6781 Alan Hayward <alan.hayward@arm.com>
6782 David Sherwood <david.sherwood@arm.com>
6784 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
6785 number of units as polynomial.
6787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6788 Alan Hayward <alan.hayward@arm.com>
6789 David Sherwood <david.sherwood@arm.com>
6791 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
6792 * target.def (autovectorize_vector_sizes): Return the vector sizes
6793 by pointer, using vector_sizes rather than a bitmask.
6794 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
6795 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
6796 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
6797 Likewise.
6798 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
6799 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
6800 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
6801 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
6802 * omp-general.c (omp_max_vf): Likewise.
6803 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6804 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
6805 * tree-vect-loop.c (vect_analyze_loop): Likewise.
6806 * tree-vect-slp.c (vect_slp_bb): Likewise.
6807 * doc/tm.texi: Regenerate.
6808 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
6809 to a poly_uint64.
6810 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
6811 the vector size as a poly_uint64 rather than an unsigned int.
6812 (current_vector_size): Change from an unsigned int to a poly_uint64.
6813 (get_vectype_for_scalar_type): Update accordingly.
6814 * tree.h (build_truth_vector_type): Take the size and number of
6815 units as a poly_uint64 rather than an unsigned int.
6816 (build_vector_type): Add a temporary overload that takes
6817 the number of units as a poly_uint64 rather than an unsigned int.
6818 * tree.c (make_vector_type): Likewise.
6819 (build_truth_vector_type): Take the number of units as a poly_uint64
6820 rather than an unsigned int.
6822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6823 Alan Hayward <alan.hayward@arm.com>
6824 David Sherwood <david.sherwood@arm.com>
6826 * target.def (get_mask_mode): Take the number of units and length
6827 as poly_uint64s rather than unsigned ints.
6828 * targhooks.h (default_get_mask_mode): Update accordingly.
6829 * targhooks.c (default_get_mask_mode): Likewise.
6830 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
6831 * doc/tm.texi: Regenerate.
6833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6834 Alan Hayward <alan.hayward@arm.com>
6835 David Sherwood <david.sherwood@arm.com>
6837 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
6838 * omp-general.c (omp_max_vf): Likewise.
6839 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
6840 (expand_omp_simd): Handle polynomial safelen.
6841 * omp-low.c (omplow_simd_context): Add a default constructor.
6842 (omplow_simd_context::max_vf): Change from int to poly_uint64.
6843 (lower_rec_simd_input_clauses): Update accordingly.
6844 (lower_rec_input_clauses): Likewise.
6846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6847 Alan Hayward <alan.hayward@arm.com>
6848 David Sherwood <david.sherwood@arm.com>
6850 * tree-vectorizer.h (vect_nunits_for_cost): New function.
6851 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
6852 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
6853 (vect_analyze_slp_cost): Likewise.
6854 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
6855 (vect_model_load_cost): Likewise.
6857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6858 Alan Hayward <alan.hayward@arm.com>
6859 David Sherwood <david.sherwood@arm.com>
6861 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
6862 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
6863 from an unsigned int * to a poly_uint64_pod *.
6864 (calculate_unrolling_factor): New function.
6865 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
6867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6868 Alan Hayward <alan.hayward@arm.com>
6869 David Sherwood <david.sherwood@arm.com>
6871 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
6872 from an unsigned int to a poly_uint64.
6873 (_loop_vec_info::slp_unrolling_factor): Likewise.
6874 (_loop_vec_info::vectorization_factor): Change from an int
6875 to a poly_uint64.
6876 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
6877 (vect_get_num_vectors): New function.
6878 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
6879 (vect_get_num_copies): Use vect_get_num_vectors.
6880 (vect_analyze_data_ref_dependences): Change max_vf from an int *
6881 to an unsigned int *.
6882 (vect_analyze_data_refs): Change min_vf from an int * to a
6883 poly_uint64 *.
6884 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
6885 than an unsigned HOST_WIDE_INT.
6886 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
6887 (vect_analyze_data_ref_dependence): Change max_vf from an int *
6888 to an unsigned int *.
6889 (vect_analyze_data_ref_dependences): Likewise.
6890 (vect_compute_data_ref_alignment): Handle polynomial vf.
6891 (vect_enhance_data_refs_alignment): Likewise.
6892 (vect_prune_runtime_alias_test_list): Likewise.
6893 (vect_shift_permute_load_chain): Likewise.
6894 (vect_supportable_dr_alignment): Likewise.
6895 (dependence_distance_ge_vf): Take the vectorization factor as a
6896 poly_uint64 rather than an unsigned HOST_WIDE_INT.
6897 (vect_analyze_data_refs): Change min_vf from an int * to a
6898 poly_uint64 *.
6899 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
6900 vfm1 as a poly_uint64 rather than an int. Make the same change
6901 for the returned bound_scalar.
6902 (vect_gen_vector_loop_niters): Handle polynomial vf.
6903 (vect_do_peeling): Likewise. Update call to
6904 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
6905 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
6906 be constant.
6907 * tree-vect-loop.c (vect_determine_vectorization_factor)
6908 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
6909 (vect_get_known_peeling_cost): Likewise.
6910 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
6911 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
6912 (vect_transform_loop): Likewise. Use the lowest possible VF when
6913 updating the upper bounds of the loop.
6914 (vect_min_worthwhile_factor): Make static. Return an unsigned int
6915 rather than an int.
6916 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
6917 polynomial unroll factors.
6918 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
6919 (vect_make_slp_decision): Likewise.
6920 (vect_supported_load_permutation_p): Likewise, and polynomial
6921 vf too.
6922 (vect_analyze_slp_cost): Handle polynomial vf.
6923 (vect_slp_analyze_node_operations): Likewise.
6924 (vect_slp_analyze_bb_1): Likewise.
6925 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
6926 than an unsigned HOST_WIDE_INT.
6927 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
6928 (vectorizable_load): Handle polynomial vf.
6929 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
6930 a poly_uint64.
6931 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
6933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6934 Alan Hayward <alan.hayward@arm.com>
6935 David Sherwood <david.sherwood@arm.com>
6937 * match.pd: Handle bit operations involving three constants
6938 and try to fold one pair.
6940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6942 * tree-vect-loop-manip.c: Include gimple-fold.h.
6943 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
6944 niters_maybe_zero parameters. Handle other cases besides a step of 1.
6945 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
6946 Add a path that uses a step of VF instead of 1, but disable it
6947 for now.
6948 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
6949 and niters_no_overflow parameters. Update calls to
6950 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
6951 Create a new SSA name if the latter choses to use a ste other
6952 than zero, and return it via niters_vector_mult_vf_var.
6953 * tree-vect-loop.c (vect_transform_loop): Update calls to
6954 vect_do_peeling, vect_gen_vector_loop_niters and
6955 slpeel_make_loop_iterate_ntimes.
6956 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
6957 (vect_gen_vector_loop_niters): Update declarations after above changes.
6959 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
6961 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
6962 128-bit round to integer instructions.
6963 (ceil<mode>2): Likewise.
6964 (btrunc<mode>2): Likewise.
6965 (round<mode>2): Likewise.
6967 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6969 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
6970 unaligned VSX load/store on P8/P9.
6971 (expand_block_clear): Allow the use of unaligned VSX
6972 load/store on P8/P9.
6974 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6976 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
6977 New function.
6978 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
6979 swap associated with both a load and a store.
6981 2018-01-02 Andrew Waterman <andrew@sifive.com>
6983 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
6984 * config/riscv/riscv.md (clear_cache): Use it.
6986 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
6988 * web.c: Remove out-of-date comment.
6990 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6992 * expr.c (fixup_args_size_notes): Check that any existing
6993 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
6994 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
6995 (emit_single_push_insn): ...here.
6997 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6999 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7000 (const_vector_encoded_nelts): New function.
7001 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7002 (const_vector_int_elt, const_vector_elt): Declare.
7003 * emit-rtl.c (const_vector_int_elt_1): New function.
7004 (const_vector_elt): Likewise.
7005 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7006 of CONST_VECTOR_ELT.
7008 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7010 * expr.c: Include rtx-vector-builder.h.
7011 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7012 directly on the tree encoding.
7013 (const_vector_from_tree): Likewise.
7014 * optabs.c: Include rtx-vector-builder.h.
7015 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7016 sequence of "u" values.
7017 * vec-perm-indices.c: Include rtx-vector-builder.h.
7018 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7019 directly on the vec_perm_indices encoding.
7021 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7023 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7024 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7025 * rtx-vector-builder.h: New file.
7026 * rtx-vector-builder.c: Likewise.
7027 * rtl.h (rtx_def::u2): Add a const_vector field.
7028 (CONST_VECTOR_NPATTERNS): New macro.
7029 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7030 (CONST_VECTOR_DUPLICATE_P): Likewise.
7031 (CONST_VECTOR_STEPPED_P): Likewise.
7032 (CONST_VECTOR_ENCODED_ELT): Likewise.
7033 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7034 (unwrap_const_vec_duplicate): Likewise.
7035 (const_vec_series_p): Check for a non-duplicated vector encoding.
7036 Say that the function only returns true for integer vectors.
7037 * emit-rtl.c: Include rtx-vector-builder.h.
7038 (gen_const_vec_duplicate_1): Delete.
7039 (gen_const_vector): Call gen_const_vec_duplicate instead of
7040 gen_const_vec_duplicate_1.
7041 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7042 (gen_const_vec_duplicate): Use rtx_vector_builder.
7043 (gen_const_vec_series): Likewise.
7044 (gen_rtx_CONST_VECTOR): Likewise.
7045 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7046 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7047 Build a new vector rather than modifying a CONST_VECTOR in-place.
7048 (handle_special_swappables): Update call accordingly.
7049 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7050 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7051 Build a new vector rather than modifying a CONST_VECTOR in-place.
7052 (handle_special_swappables): Update call accordingly.
7054 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7056 * simplify-rtx.c (simplify_const_binary_operation): Use
7057 CONST_VECTOR_ELT instead of XVECEXP.
7059 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7061 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7062 the selector elements to be different from the data elements
7063 if the selector is a VECTOR_CST.
7064 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7065 ssizetype for the selector.
7067 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7069 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7070 before testing each element individually.
7071 * tree-vect-generic.c (lower_vec_perm): Likewise.
7073 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7075 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7076 * selftest-run-tests.c (selftest::run_tests): Call it.
7077 * vector-builder.h (vector_builder::operator ==): New function.
7078 (vector_builder::operator !=): Likewise.
7079 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7080 (vec_perm_indices::all_from_input_p): New function.
7081 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7082 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7083 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7084 instead of reading the VECTOR_CST directly. Detect whether both
7085 vector inputs are the same before constructing the vec_perm_indices,
7086 and update the number of inputs argument accordingly. Use the
7087 utility functions added above. Only construct sel2 if we need to.
7089 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7091 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7092 the broadcast of the low byte.
7093 (expand_mult_highpart): Use an explicit encoding for the permutes.
7094 * optabs-query.c (can_mult_highpart_p): Likewise.
7095 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7096 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7097 (vectorizable_bswap): Likewise.
7098 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7099 explicit encoding for the power-of-2 permutes.
7100 (vect_permute_store_chain): Likewise.
7101 (vect_grouped_load_supported): Likewise.
7102 (vect_permute_load_chain): Likewise.
7104 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7106 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7107 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7108 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7109 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7110 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7111 (vect_gen_perm_mask_any): Likewise.
7113 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7115 * int-vector-builder.h: New file.
7116 * vec-perm-indices.h: Include int-vector-builder.h.
7117 (vec_perm_indices): Redefine as an int_vector_builder.
7118 (auto_vec_perm_indices): Delete.
7119 (vec_perm_builder): Redefine as a stand-alone class.
7120 (vec_perm_indices::vec_perm_indices): New function.
7121 (vec_perm_indices::clamp): Likewise.
7122 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
7123 (vec_perm_indices::new_vector): New function.
7124 (vec_perm_indices::new_expanded_vector): Update for new
7125 vec_perm_indices class.
7126 (vec_perm_indices::rotate_inputs): New function.
7127 (vec_perm_indices::all_in_range_p): Operate directly on the
7128 encoded form, without computing elided elements.
7129 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
7130 encoding. Update for new vec_perm_indices class.
7131 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
7132 the given vec_perm_builder.
7133 (expand_vec_perm_var): Update vec_perm_builder constructor.
7134 (expand_mult_highpart): Use vec_perm_builder instead of
7135 auto_vec_perm_indices.
7136 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
7137 vec_perm_indices instead of auto_vec_perm_indices. Use a single
7138 or double series encoding as appropriate.
7139 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
7140 vec_perm_indices instead of auto_vec_perm_indices.
7141 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7142 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7143 (vect_permute_store_chain): Likewise.
7144 (vect_grouped_load_supported): Likewise.
7145 (vect_permute_load_chain): Likewise.
7146 (vect_shift_permute_load_chain): Likewise.
7147 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7148 (vect_transform_slp_perm_load): Likewise.
7149 (vect_schedule_slp_instance): Likewise.
7150 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7151 (vectorizable_mask_load_store): Likewise.
7152 (vectorizable_bswap): Likewise.
7153 (vectorizable_store): Likewise.
7154 (vectorizable_load): Likewise.
7155 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
7156 vec_perm_indices instead of auto_vec_perm_indices. Use
7157 tree_to_vec_perm_builder to read the vector from a tree.
7158 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
7159 vec_perm_builder instead of a vec_perm_indices.
7160 (have_whole_vector_shift): Use vec_perm_builder and
7161 vec_perm_indices instead of auto_vec_perm_indices. Leave the
7162 truncation to calc_vec_perm_mask_for_shift.
7163 (vect_create_epilog_for_reduction): Likewise.
7164 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
7165 from auto_vec_perm_indices to vec_perm_indices.
7166 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7167 instead of changing individual elements.
7168 (aarch64_vectorize_vec_perm_const): Use new_vector to install
7169 the vector in d.perm.
7170 * config/arm/arm.c (expand_vec_perm_d::perm): Change
7171 from auto_vec_perm_indices to vec_perm_indices.
7172 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7173 instead of changing individual elements.
7174 (arm_vectorize_vec_perm_const): Use new_vector to install
7175 the vector in d.perm.
7176 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
7177 Update vec_perm_builder constructor.
7178 (rs6000_expand_interleave): Likewise.
7179 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
7180 (rs6000_expand_interleave): Likewise.
7182 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7184 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
7185 to qimode could truncate the indices.
7186 * optabs.c (expand_vec_perm_var): Likewise.
7188 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7190 * Makefile.in (OBJS): Add vec-perm-indices.o.
7191 * vec-perm-indices.h: New file.
7192 * vec-perm-indices.c: Likewise.
7193 * target.h (vec_perm_indices): Replace with a forward class
7194 declaration.
7195 (auto_vec_perm_indices): Move to vec-perm-indices.h.
7196 * optabs.h: Include vec-perm-indices.h.
7197 (expand_vec_perm): Delete.
7198 (selector_fits_mode_p, expand_vec_perm_var): Declare.
7199 (expand_vec_perm_const): Declare.
7200 * target.def (vec_perm_const_ok): Replace with...
7201 (vec_perm_const): ...this new hook.
7202 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
7203 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
7204 * doc/tm.texi: Regenerate.
7205 * optabs.def (vec_perm_const): Delete.
7206 * doc/md.texi (vec_perm_const): Likewise.
7207 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
7208 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
7209 expand_vec_perm for constant permutation vectors. Assert that
7210 the mode of variable permutation vectors is the integer equivalent
7211 of the mode that is being permuted.
7212 * optabs-query.h (selector_fits_mode_p): Declare.
7213 * optabs-query.c: Include vec-perm-indices.h.
7214 (selector_fits_mode_p): New function.
7215 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
7216 is defined, instead of checking whether the vec_perm_const_optab
7217 exists. Use targetm.vectorize.vec_perm_const instead of
7218 targetm.vectorize.vec_perm_const_ok. Check whether the indices
7219 fit in the vector mode before using a variable permute.
7220 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
7221 vec_perm_indices instead of an rtx.
7222 (expand_vec_perm): Replace with...
7223 (expand_vec_perm_const): ...this new function. Take the selector
7224 as a vec_perm_indices rather than an rtx. Also take the mode of
7225 the selector. Update call to shift_amt_for_vec_perm_mask.
7226 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
7227 Use vec_perm_indices::new_expanded_vector to expand the original
7228 selector into bytes. Check whether the indices fit in the vector
7229 mode before using a variable permute.
7230 (expand_vec_perm_var): Make global.
7231 (expand_mult_highpart): Use expand_vec_perm_const.
7232 * fold-const.c: Includes vec-perm-indices.h.
7233 * tree-ssa-forwprop.c: Likewise.
7234 * tree-vect-data-refs.c: Likewise.
7235 * tree-vect-generic.c: Likewise.
7236 * tree-vect-loop.c: Likewise.
7237 * tree-vect-slp.c: Likewise.
7238 * tree-vect-stmts.c: Likewise.
7239 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
7240 Delete.
7241 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
7242 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
7243 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
7244 (aarch64_vectorize_vec_perm_const): ...this new function.
7245 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7246 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7247 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
7248 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
7249 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7250 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7251 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
7252 into...
7253 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
7254 check for NEON modes.
7255 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
7256 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
7257 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
7258 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
7259 into...
7260 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
7261 the old VEC_PERM_CONST conditions.
7262 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
7263 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
7264 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
7265 (ia64_vectorize_vec_perm_const_ok): Merge into...
7266 (ia64_vectorize_vec_perm_const): ...this new function.
7267 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
7268 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
7269 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
7270 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
7271 * config/mips/mips.c (mips_expand_vec_perm_const)
7272 (mips_vectorize_vec_perm_const_ok): Merge into...
7273 (mips_vectorize_vec_perm_const): ...this new function.
7274 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
7275 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
7276 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
7277 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
7278 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
7279 (rs6000_expand_vec_perm_const): Delete.
7280 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
7281 Delete.
7282 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7283 (altivec_expand_vec_perm_const_le): Take each operand individually.
7284 Operate on constant selectors rather than rtxes.
7285 (altivec_expand_vec_perm_const): Likewise. Update call to
7286 altivec_expand_vec_perm_const_le.
7287 (rs6000_expand_vec_perm_const): Delete.
7288 (rs6000_vectorize_vec_perm_const_ok): Delete.
7289 (rs6000_vectorize_vec_perm_const): New function.
7290 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7291 an element count and rtx array.
7292 (rs6000_expand_extract_even): Update call accordingly.
7293 (rs6000_expand_interleave): Likewise.
7294 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
7295 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
7296 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
7297 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
7298 (rs6000_expand_vec_perm_const): Delete.
7299 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7300 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7301 (altivec_expand_vec_perm_const_le): Take each operand individually.
7302 Operate on constant selectors rather than rtxes.
7303 (altivec_expand_vec_perm_const): Likewise. Update call to
7304 altivec_expand_vec_perm_const_le.
7305 (rs6000_expand_vec_perm_const): Delete.
7306 (rs6000_vectorize_vec_perm_const_ok): Delete.
7307 (rs6000_vectorize_vec_perm_const): New function. Remove stray
7308 reference to the SPE evmerge intructions.
7309 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7310 an element count and rtx array.
7311 (rs6000_expand_extract_even): Update call accordingly.
7312 (rs6000_expand_interleave): Likewise.
7313 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
7314 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
7315 new function.
7316 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7318 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7320 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
7321 vector mode and that that mode matches the mode of the data
7322 being permuted.
7323 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
7324 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
7325 directly using expand_vec_perm_1 when forcing selectors into
7326 registers.
7327 (expand_vec_perm_var): New function, split out from expand_vec_perm.
7329 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7331 * optabs-query.h (can_vec_perm_p): Delete.
7332 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
7333 * optabs-query.c (can_vec_perm_p): Split into...
7334 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
7335 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
7336 particular selector is valid.
7337 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7338 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7339 (vect_grouped_load_supported): Likewise.
7340 (vect_shift_permute_load_chain): Likewise.
7341 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7342 (vect_transform_slp_perm_load): Likewise.
7343 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7344 (vectorizable_bswap): Likewise.
7345 (vect_gen_perm_mask_checked): Likewise.
7346 * fold-const.c (fold_ternary_loc): Likewise. Don't take
7347 implementations of variable permutation vectors into account
7348 when deciding which selector to use.
7349 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
7350 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
7351 with a false third argument.
7352 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
7353 to test whether the constant selector is valid and can_vec_perm_var_p
7354 to test whether a variable selector is valid.
7356 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7358 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
7359 * optabs-query.c (can_vec_perm_p): Likewise.
7360 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
7361 instead of vec_perm_indices.
7362 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
7363 (vect_gen_perm_mask_checked): Likewise,
7364 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
7365 (vect_gen_perm_mask_checked): Likewise,
7367 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7369 * optabs-query.h (qimode_for_vec_perm): Declare.
7370 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
7371 (qimode_for_vec_perm): ...this new function.
7372 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
7374 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7376 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
7377 does not have a conditional at the top.
7379 2018-01-02 Richard Biener <rguenther@suse.de>
7381 * ipa-inline.c (big_speedup_p): Fix expression.
7383 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7385 PR target/81616
7386 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
7387 for generic 4->6.
7389 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7391 PR target/81616
7392 Generic tuning.
7393 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
7394 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
7395 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
7396 cond_taken_branch_cost 3->4.
7398 2018-01-01 Jakub Jelinek <jakub@redhat.com>
7400 PR tree-optimization/83581
7401 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
7402 TODO_cleanup_cfg if any changes have been made.
7404 PR middle-end/83608
7405 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
7406 convert_modes if target mode has the right side, but different mode
7407 class.
7409 PR middle-end/83609
7410 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
7411 last argument when extracting from CONCAT. If either from_real or
7412 from_imag is NULL, use expansion through memory. If result is not
7413 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
7414 the parts directly to inner mode, if even that fails, use expansion
7415 through memory.
7417 PR middle-end/83623
7418 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
7419 check for bswap in mode rather than HImode and use that in expand_unop
7420 too.
7421 \f
7422 Copyright (C) 2018 Free Software Foundation, Inc.
7424 Copying and distribution of this file, with or without modification,
7425 are permitted in any medium without royalty provided the copyright
7426 notice and this notice are preserved.