| blob | a136e541e13f576531bcbd5d87ccfd2db7a5ec54 |
1 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4 New prototype.
5 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6 Add note to error message to explain internal mapping of overloaded
7 built-in function name to non-overloaded built-in function name.
8 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
9 function.
11 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
13 PR target/85424
14 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
15 where the inputs overlap with the output.
17 2018-04-17 Jakub Jelinek <jakub@redhat.com>
19 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
20 (=v, v) alternative and explicit "memory" attribute.
21 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
22 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
23 attributes.
24 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
25 "sselog1" type instead of "sselog".
26 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
27 "sselog". Remove explicit "memory" attribute.
28 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
29 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
30 attributes.
31 (vec_extract_hi_v32hi): Merge all alternatives into one, use
32 "sselog1" type instead of "sselog". Remove explicit "memory"
33 attribute.
34 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
35 use "sselog1" type instead of "sselog". Remove explicit "memory"
36 attribute.
37 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
38 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
39 attributes.
40 (vec_extract_hi_v64qi): Merge all alternatives into one, use
41 "sselog1" type instead of "sselog". Remove explicit "memory"
42 attribute.
43 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
44 use "sselog1" type instead of "sselog". Remove explicit "memory"
45 attribute.
47 PR target/85430
48 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
50 PR middle-end/85414
51 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
52 on a SUBREG.
54 2018-04-17 Martin Jambor <mjambor@suse.cz>
56 PR ipa/85421
57 * ipa-cp.c (create_specialized_node): Call
58 expand_all_artificial_thunks if necessary.
60 2018-04-17 Martin Liska <mliska@suse.cz>
62 PR lto/85405
63 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
64 in message, remote space in between '_G' and '('.
66 2018-04-17 Jakub Jelinek <jakub@redhat.com>
68 PR target/85281
69 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
70 avx512f_vmcmp<mode>3<round_saeonly_name>,
71 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
72 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
73 avx512f_rndscale<mode><round_saeonly_name>,
74 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
75 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
76 Use %<iptr>2 instead of %2 for -masm=intel.
77 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
78 avx512f_vcvttss2usi<round_saeonly_name>,
79 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
80 -masm=intel.
81 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
82 avx512f_vcvttsd2usi<round_saeonly_name>,
83 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
84 Use %q1 instead of %1 for -masm=intel.
85 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
86 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
87 of %3 for -masm=intel.
88 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
89 -masm=intel.
90 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
91 -masm=intel.
92 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
93 -masm=intel.
94 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
95 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
96 %g1.
97 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
98 -masm=intel.
99 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
100 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
101 %g1 and one with %0 and %1.
102 (avx512er_vmrcp28<mode><round_saeonly_name>,
103 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
104 %1 for -masm=intel.
105 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
106 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
107 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
108 of %0 and %{%4%} for -masm=intel.
109 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
110 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
111 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
112 order of %0 and %{%5%}%{z%} for -masm=intel.
114 2018-04-17 Jan Hubicka <jh@suse.cz>
116 PR lto/85405
117 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
119 2018-04-17 Martin Liska <mliska@suse.cz>
121 PR ipa/85329
122 * multiple_target.c (create_dispatcher_calls): Set apostrophes
123 for target_clone error message. Make default implementation
124 clone to be a local declaration.
125 (separate_attrs): Add new argument and check for an empty
126 string.
127 (expand_target_clones): Handle it.
128 (ipa_target_clone): Make redirection just for target_clones
129 functions.
131 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
132 Tom de Vries <tom@codesourcery.com>
134 PR middle-end/84955
135 * omp-expand.c (expand_oacc_for): Add dummy false branch for
136 tiled basic blocks without omp continue statements.
138 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
140 PR target/83660
141 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
142 vec_extract expression as having side effects to make sure it gets
143 a cleanup point.
145 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
147 PR target/85403
148 * config/i386/i386.c (get_builtin_code_for_version): Check
149 error_mark_node.
151 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
153 PR target/84331
154 * gcc/config.gcc: Support "skylake".
155 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
156 PROCESSOR_SKYLAKE.
157 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
158 (processor_target_table): Add "skylake".
159 (ix86_option_override_internal): Add "skylake".
160 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
161 PROCESSOR_CANNONLAKE.
162 (get_builtin_code_for_version): Fix priority for
163 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
164 PROCESSOR_SKYLAKE-AVX512.
165 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
166 (processor_type): Add PROCESSOR_SKYLAKE.
168 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
169 Jason Merrill <jason@redhat.com>
171 PR c++/85112
172 * convert.c (convert_to_integer_1): Use direct recursion for
173 enumeral types and types with a precision less than the number
174 of bits in their mode.
176 2018-04-16 Julia Koval <julia.koval@intel.com>
178 PR target/84413
179 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
180 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
182 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
184 PR target/85293
185 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
186 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
187 and -mno-direct-move.
189 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
191 PR target/83402
192 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
193 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
194 Ensure negative shifts result in {0}.
196 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
198 PR rtl-optimization/79916
199 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
200 regs (if any) to define how to gnerate SD moves when LRA is in
201 progress.
203 2018-04-13 Jakub Jelinek <jakub@redhat.com>
205 PR rtl-optimization/85393
206 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
207 * except.c (expand_dw2_landing_pad_for_region): Make static.
208 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
209 a label and unconditional jump to old_bb, rather than
210 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
211 basic block.
213 PR rtl-optimization/85376
214 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
215 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
216 instead of a specific value.
218 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
219 Bin Cheng <bin.cheng@arm.com>
221 PR tree-optimization/82965
222 PR tree-optimization/83991
223 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
224 by_profile_only parameter.
225 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
226 information if the loop was predicted to iterate too many times.
227 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
229 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
231 PR lto/71991
232 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
233 always inline.
235 2018-04-13 Martin Liska <mliska@suse.cz>
236 Jakub Jelinek <jakub@redhat.com>
238 PR middle-end/81657
239 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
240 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
241 * builtins.c (expand_builtin_memory_copy_args): Use
242 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
243 handle dest_addr == pc_rtx.
245 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
247 PR target/85291
248 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
249 asked to not generate direct moves.
250 (fix_trunc<mode>si2_stfiwx): Similar.
251 (fix_trunc<mode>si2_internal): Similar.
253 2018-04-12 Jakub Jelinek <jakub@redhat.com>
255 PR debug/83157
256 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
257 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
258 lookup if dest in some wider mode is known to be const0_rtx and
259 if so, record permanent equivalence for it to be ZERO_EXTEND of
260 the narrower mode destination.
262 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
264 * lto-streamer-out.c (output_function): Revert 259346.
265 * omp-expand.c (expand_oacc_for): Likewise.
267 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
269 PR rtl-optimization/85354
270 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
271 * sel-sched.c (sel_global_init): ... here.
273 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
275 PR target/85238
276 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
277 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
278 mode for PE-COFF targets.
279 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
280 (i386_pe_asm_lto_end): Likewise.
281 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
282 (TARGET_ASM_LTO_END): Likewise.
283 * config/i386/winnt.c (saved_debug_info_level): New static variable.
284 (i386_pe_asm_lto_start): New function.
285 (i386_pe_asm_lto_end): Likewise.
287 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
288 Richard Biener <rguenther@suse.de>
290 PR middle-end/84955
291 * lto-streamer-out.c (output_function): Fix CFG loop state before
292 streaming out.
293 * omp-expand.c (expand_oacc_for): Handle calls to internal
294 functions like regular functions.
296 2018-04-12 Richard Biener <rguenther@suse.de>
298 PR lto/85371
299 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
300 for the early LTO debug to properly generate references to it
301 during DIE emission. Do not re-use that for the skeleton for
302 split-dwarf.
303 (dwarf2out_early_finish): Likewise.
305 2018-04-12 Jakub Jelinek <jakub@redhat.com>
307 PR target/85328
308 * config/i386/sse.md
309 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
310 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
311 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
312 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
313 and output is a reg, avoid creating invalid lowpart subreg, but
314 instead split into a 512-bit move. Don't split if not AVX512VL,
315 input is xmm16+ reg and output is a mem.
316 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
317 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
318 xmm16+ reg and output is a mem.
320 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
322 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
323 also for flag_dwarf2_cfi_asm.
325 2018-04-12 Jakub Jelinek <jakub@redhat.com>
327 PR rtl-optimization/85342
328 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
329 a bool scalar var inside of the loop instead. Don't try to update
330 recog_data.operand after failed apply_change_group.
332 2018-04-12 Tom de Vries <tom@codesourcery.com>
334 PR target/85296
335 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
336 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
337 array with flexible array member as array without given dimension.
338 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
339 argument for undefined param to true.
341 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
343 PR target/85321
344 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
345 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
346 from PowerPC section.
347 * config/rs6000/sysv4.opt (mcall-): Improve help text.
348 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
349 help text that is too long.
350 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
351 help text that is too long.
352 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
353 help text that is too long.
355 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
357 * config/alpha/alpha.md (stack_probe_internal): Rename
358 from "probe_stack". Update all callers.
360 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
362 PR rtl-optimization/84566
363 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
364 sched_macro_fuse_insns.
366 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
368 PR target/84301
369 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
370 (compute_block_dependences): ... from here.
372 2018-04-11 Jakub Jelinek <jakub@redhat.com>
374 PR tree-optimization/85331
375 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
376 from int to HOST_WIDE_INT.
378 2018-04-11 Martin Jambor <mjambor@suse.cz>
380 PR ipa/84149
381 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
382 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
383 not the same as the source val.
384 (cgraph_edge_brings_value_p): New parameter.
385 (gather_edges_for_value): Pass destination value to
386 cgraph_edge_brings_value_p.
387 (perhaps_add_new_callers): Likewise.
388 (get_info_about_necessary_edges): Likewise and exclude values brought
389 only by self-recursive edges.
390 (create_specialized_node): Redirect only clones of self-calling edges.
391 (+self_recursive_pass_through_p): New function.
392 (find_more_scalar_values_for_callers_subset): Use it.
393 (find_aggregate_values_for_callers_subset): Likewise.
394 (known_aggs_to_agg_replacement_list): Removed.
395 (decide_whether_version_node): Re-calculate known constants for all
396 remaining context clones.
398 2018-04-11 Richard Biener <rguenther@suse.de>
400 PR lto/85339
401 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
402 from early DWARF output.
403 (dwarf2out_early_finish): Output line info unconditionally into
404 early DWARF and add reference to it.
406 2018-04-11 Jakub Jelinek <jakub@redhat.com>
408 PR target/85281
409 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
410 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
411 other than V2DFmode using iptr mode attribute.
412 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
414 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
416 PR rtl-optimization/84659
417 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
419 2018-04-11 Jakub Jelinek <jakub@redhat.com>
421 PR debug/85302
422 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
423 SIZEP is NULL.
424 (output_loc_list): Pass address of a dummy size variable even in the
425 locview handling loop.
426 (index_location_lists): Add comment on why skip_loc_list_entry can't
427 call size_of_locs.
429 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
431 PR target/85261
432 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
433 into register.
435 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
437 PR target/85321
438 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
439 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
440 and -mstring-compare-inline-limit.
442 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
444 PR target/85287
445 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
446 for stack clash protection in a register whenever we need it to be in
447 a register.
449 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
451 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
452 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
454 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
456 PR target/85321
457 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
458 the help text.
459 (mlong-double-): Ditto.
460 * config/rs6000/sysv4.opt (msdata=): Ditto.
461 (mtls-size=): Ditto.
463 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
465 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
466 erroneous entries for
467 "vector int vec_ldl (int, long int *)", and
468 "vector unsigned int vec_ldl (int, unsigned long int *)".
469 Add comments and entries for
470 "vector bool char vec_ldl (int, bool char *)",
471 "vector bool short vec_ldl (int, bool short *)",
472 "vector bool int vec_ldl (int, bool int *)",
473 "vector bool long long vec_ldl (int, bool long long *)",
474 "vector pixel vec_ldl (int, pixel *)",
475 "vector long long vec_ldl (int, long long *)",
476 "vector unsigned long long vec_ldl (int, unsigned long long *)".
477 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
478 type tree bool_long_long_type_node and correct definition of
479 bool_V2DI_type_node to make reference to this new type tree.
480 (rs6000_mangle_type): Replace erroneous reference to
481 bool_long_type_node with bool_long_long_type_node.
482 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
483 comments to emphasize sign distinctions for char and int types and
484 replace RS6000_BTI_bool_long constant with
485 RS6000_BTI_bool_long_long constant. Also add comment to restrict
486 use of RS6000_BTI_pixel.
487 (bool_long_type_node): Remove this macro definition.
488 (bool_long_long_type_node): New macro definition
490 2018-04-10 Jakub Jelinek <jakub@redhat.com>
492 PR rtl-optimization/85300
493 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
494 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
495 simplify_unary_operation fails.
497 2018-04-10 Martin Liska <mliska@suse.cz>
499 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
500 cgraph_edge and ipa_ref.
502 2018-04-10 Jakub Jelinek <jakub@redhat.com>
504 PR target/85177
505 PR target/85255
506 * config/i386/sse.md
507 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
508 computation of the VEC_MERGE selector from mask.
509 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
510 Fix decoding of the VEC_MERGE selector into mask.
512 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
514 PR tree-optimization/85286
515 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
517 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
519 * final.c (final_1): Set insn_last_address as well as
520 insn_current_address.
522 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
524 PR target/85173
525 * explow.c (emit_stack_probe): Call validize_mem on memory location
526 before passing it to gen_probe_stack. Create address operand and
527 legitimize it for the probe_stack_address case.
529 2018-04-09 Jan Hubicka <jh@suse.cz>
531 PR lto/85078
532 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
533 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
534 * tree.c (free_lang_data_in_type): Fix handling of binfos;
535 walk basetypes.
536 (free_lang_data): Rebuild type inheritance graph.
538 2018-04-09 Martin Sebor <msebor@redhat.com>
540 * invoke.texi (-finline-small-functions): Mention other optimization
541 options.
542 (-findirect-inlining, -fpartial-inlining): Same.
543 (-finline-functions-called-once): Same.
544 (-freorder-blocks-and-partition): Same.
546 2018-04-09 Jan Hubicka <jh@suse.cz>
548 PR rtl/84058
549 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
550 jumps; choose last target that matches the criteria (i.e.
551 no partition changes for non-crossing jumps).
552 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
553 support for redirecting crossing jumps to non-crossing.
555 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
557 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
558 also for naked functions.
560 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
562 * config/arc/arc.md (add_shift): New pattern.
563 (add_shift2): Likewise.
564 (sub_shift): Likewise.
565 (sub_shift_cmp0_noout): Likewise.
566 (compare_si_ashiftsi): Likewise.
567 (xbfu_cmp0_noout): New combine pattern.
568 (xbfu_cmp0"): Likewise.
569 (movsi_set_cc_insn): Place the predicable variant first.
570 (commutative_binary_cmp0_noout): Remove clobber.
571 (commutative_binary_cmp0): New pattern.
572 (noncommutative_binary_cmp0): Likewise.
573 (noncommutative_binary_cmp0_noout): Likewise.
574 (noncommutative_binary_comparison_result_used): Removed.
575 (rsub_cmp0): New pattern.
576 (rsub_cmp0_noout): Likewise.
577 (extzvsi): Changed, keep only meaningful variants.
578 (SQH, SEZ): New iterators.
579 (SQH_postfix): New mode attribute.
580 (SEZ_prefix): New code attribute.
581 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
582 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
583 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
584 of numerical value.
585 (noncommutative_operator): Check the availability of barrel
586 shifter option.
588 2018-04-09 Richard Biener <rguenther@suse.de>
590 PR tree-optimization/85284
591 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
592 Only use the niter constraining form of simple_iv when the exit
593 is always executed.
595 2018-04-09 Tom de Vries <tom@codesourcery.com>
597 PR target/84041
598 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
599 (define_expand "*memory_barrier"): New define_expand.
600 (define_insn "memory_barrier"): New insn.
602 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
604 PR rtl-optimization/80463
605 PR rtl-optimization/83972
606 PR rtl-optimization/83480
608 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
609 correct producer for the insn.
610 (tidy_control_flow): Fixup seqnos in case of debug insns.
612 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
614 PR rtl-optimization/83913
616 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
617 different sched-times when merging exprs.
619 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
621 PR rtl-optimization/83962
623 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
624 tidy_fallthru_edge and tidy_control_flow.
626 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
628 PR rtl-optimization/83530
630 * sel-sched.c (force_next_insn): New global variable.
631 (remove_insn_for_debug): When force_next_insn is true, also leave only
632 next insn in the ready list.
633 (sel_sched_region): When the region wasn't scheduled, make another pass
634 over it with force_next_insn set to 1.
636 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
638 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
639 into tm_file.
640 * config/nds32/constants.md (unspec_volatile_element): Add enum values
641 for interrupt control.
642 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
643 functions for interrupt control.
644 * config/nds32/nds32-intrinsic.md: Likewise.
645 * config/nds32/nds32_intrinsic.h: Likewise.
646 * config/nds32/nds32.h (nds32_builtins): Likewise.
648 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
650 * config/nds32/nds32.c (nds32_init_machine_status,
651 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
652 strict_aligned_p field.
653 (nds32_expand_to_rtl_hook): New function.
654 (TARGET_EXPAND_TO_RTL_HOOK): Define.
655 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
657 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
658 Chung-Ju Wu <jasonwucj@gmail.com>
660 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
661 * config/nds32/nds32-n7.md: New file.
662 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
663 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
664 pipeline.
665 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
666 * config/nds32/nds32.md (pipeline_model): Add n7.
667 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
668 * config/nds32/pipelines.md: Include n7 settings.
670 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
671 Chung-Ju Wu <jasonwucj@gmail.com>
673 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
674 * config/nds32/nds32-e8.md: New file.
675 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
676 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
677 pipeline.
678 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
679 * config/nds32/nds32.md (pipeline_model): Add e8.
680 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
681 * config/nds32/pipelines.md: Include e8 settings.
683 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
684 Chung-Ju Wu <jasonwucj@gmail.com>
686 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
687 * config/nds32/nds32-n8.md: New file.
688 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
689 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
690 pipeline.
691 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
692 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
693 * config/nds32/nds32.md (pipeline_model): Add n8.
694 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
695 * config/nds32/pipelines.md: Include n8 settings.
697 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
698 Chung-Ju Wu <jasonwucj@gmail.com>
700 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
701 * config/nds32/nds32-n9-2r1w.md: New file.
702 * config/nds32/nds32-n9-3r2w.md: New file.
703 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
704 nds32_register_ports): New or modify for cpu n9.
705 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
706 pipeline.
707 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
708 * config/nds32/nds32-utils.c: New file.
709 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
710 TARGET_MUL_SLOW): Define.
711 * config/nds32/nds32.md (pipeline_model): New attribute.
712 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
713 New options that support cpu n9.
714 * config/nds32/pipelines.md: Include n9 settings.
715 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
717 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
719 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
720 information if necessary.
721 (output_cond_branch_compare_zero): Likewise.
722 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
723 (nds32_target_alignment): Refine for alignment.
724 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
725 (FUNCTION_BOUNDARY): Modify.
726 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
727 align case.
728 * config/nds32/nds32.opt (malways-align, malign-functions): New.
730 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
732 * config/nds32/constants.md (unspec_volatile_element): Add values for
733 TLB operation and data prefetch.
734 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
735 functions for TLB operation and data prefetch.
736 * config/nds32/nds32-intrinsic.md: Likewise.
737 * config/nds32/nds32_intrinsic.h: Likewise.
738 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
739 (nds32_print_operand): Likewise.
740 * config/nds32/nds32.h (nds32_builtins): Likewise.
742 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
743 Andrew Pinski <pinsika@gcc.gnu.org>
745 PR middle-end/82976
746 * match.pd: Use constant_boolean_node of correct type instead of
747 boolean_true_node or boolean_false_node for simplifying
748 pointer comparisons to zero.
750 2018-04-07 Jakub Jelinek <jakub@redhat.com>
752 PR tree-optimization/80021
753 * tree.c (verify_type_variant): Make error call in verify_variant_match
754 translatable and remove final full stop.
756 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
758 * config/nds32/constants.md (unspec_volatile_element): Add
759 UNSPEC_VOLATILE_EH_RETURN.
760 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
761 nds32_output_stack_pop): Support dwarf exception handling process.
762 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
763 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
764 exception handling process.
765 (nds32_compute_stack_frame): Likewise.
766 (nds32_return_addr_rtx): Likewise.
767 (nds32_initial_elimination_offset): Likewise.
768 (nds32_expand_prologue): Likewise.
769 (nds32_expand_epilogue): Likewise.
770 (nds32_dynamic_chain_address): New function.
771 * config/nds32/nds32.h (machine_function): Add fields for dwarf
772 exception handling.
773 (DYNAMIC_CHAIN_ADDRESS): Define.
774 (EH_RETURN_DATA_REGNO): Define.
775 (EH_RETURN_STACKADJ_RTX): Define.
776 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
777 patterns for dwarf exception handling.
779 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
781 * config/nds32/nds32.h: Clean up obsolete macros.
783 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
785 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
786 Add enum values for particular instructions.
787 * config/nds32/nds32-intrinsic.c: Implementation of expanding
788 particular intrinsic functions.
789 * config/nds32/nds32-intrinsic.md: Likewise.
790 * config/nds32/nds32_intrinsic.h: Likewise.
791 * config/nds32/nds32.h (nds32_builtins): Likewise.
792 * config/nds32/nds32.md (type): Add pbsad and pbsada.
793 (btst, ave): New patterns for particular instructions.
795 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
797 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
798 Add enum values for atomic load/store and memory sync.
799 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
800 and memory sync.
801 * config/nds32/nds32-intrinsic.md: Likewise.
802 * config/nds32/nds32_intrinsic.h: Likewise.
803 * config/nds32/nds32.h (nds32_builtins): Likewise.
805 2018-04-07 Jakub Jelinek <jakub@redhat.com>
807 PR tree-optimization/85257
808 * fold-const.c (native_encode_vector): If not all elts could fit
809 and off is -1, return 0 rather than offset.
810 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
811 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
812 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
813 adjust buffer in native_interpret_expr call.
815 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
817 * config/nds32/constants.md (unspec_volatile_element): Add cache
818 control enum values.
819 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
820 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
821 * config/nds32/nds32.c (nds32_cctl_names): New.
822 (nds32_print_operand): Handle cache control register names.
823 * config/nds32/nds32.h (nds32_builtins): New enum values.
824 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
825 macros.
826 * config/nds32/nds32.md (type): Add mmu.
827 * config/nds32/pipelines.md (simple_insn): Add mmu.
829 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
831 * config/nds32/nds32.md (type): Remove call.
832 * config/nds32/pipelines.md (simple_insn): Likewise.
834 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
836 * config/nds32/constants.md (unspec_volatile_element): Add
837 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
838 UNSPEC_VOLATILE_FMFCFG.
839 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
840 description for fmfcfg and fmfcsr.
841 (bdesc_1arg): Add fmtcsr.
842 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
843 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
844 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
845 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
846 unspec_fmfcfg): New patterns.
847 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
848 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
849 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
850 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
851 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
852 __nds32__fmfcfg): Define.
854 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
856 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
857 intrinsic register names.
858 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
859 intrinsic register enum values and macros.
861 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
863 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
864 for load/store addressing form.
865 (nds32_print_operand_address): Likewise.
867 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
869 PR target/85196
870 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
871 based on LABEL_REF. Remove useless assertion.
872 (pic_address_needs_scratch): Fix formatting.
873 (sparc_legitimize_pic_address): Minor tweaks.
874 (sparc_delegitimize_address): Adjust assertion accordingly.
875 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
876 into symbolic_operand.
877 (movsi_high_pic_label_ref): Likewise.
878 (movsi_lo_sum_pic_label_ref): Likewise.
879 (movdi_pic_label_ref): Likewise.
880 (movdi_high_pic_label_ref): Likewise.
881 (movdi_lo_sum_pic_label_ref): Likewise.
883 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
885 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
886 custom LIB_SPEC setup.
888 2018-04-06 Ruslan Bukin <br@bsdpad.com>
889 Kito Cheng <kito.cheng@gmail.com>
891 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
892 * config/riscv/freebsd.h: New.
894 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
896 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
897 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
898 file.
900 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
901 Kito Cheng <kito.cheng@gmail.com>
903 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
904 nds32_output_call, nds32_symbol_binds_local_p): New functions.
905 * config/nds32/nds32-protos.h (nds32_output_call,
906 nds32_output_return): Declare.
907 * config/nds32/nds32.md: Refine all the call and return patterns.
909 2018-04-06 Jakub Jelinek <jakub@redhat.com>
911 PR debug/85252
912 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
913 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
915 PR rtl-optimization/84872
916 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
917 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
918 EDGE_CROSSING edge.
920 2018-04-06 Tamar Christina <tamar.christina@arm.com>
922 * expr.c (copy_blkmode_to_reg): Revert 254862.
923 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
925 2018-04-06 Richard Biener <rguenther@suse.de>
927 PR middle-end/85244
928 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
929 after seeing a component reference with an adjacent field. Treat
930 refs to arrays at struct end of external decls similar to
931 refs to unconstrained commons.
933 2018-04-06 Jakub Jelinek <jakub@redhat.com>
935 PR sanitizer/85213
936 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
937 look through SAVE_EXPRs with non-side-effects argument. Adjust
938 recursive calls.
939 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
940 save_p here.
942 2018-04-06 Richard Biener <rguenther@suse.de>
944 PR middle-end/85180
945 * alias.c (find_base_term): New wrapper around find_base_term
946 unwinding CSELIB_VAL_PTR changes.
947 (find_base_term): Do not restore CSELIB_VAL_PTR during the
948 recursion.
950 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
952 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
953 instructions.
954 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
955 constant definitions.
956 ("nop"): lr 0,0 -> nopr r0
957 ("nop_lr0", "nop_lr1"): New insn definitions.
959 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
961 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
962 NDS32_V3PUSH_AVAILABLE_P macro.
964 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
965 Chung-Ju Wu <jasonwucj@gmail.com>
967 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
968 (nds32*-*-*): Add float and fpu_config into supported_defaults.
969 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
970 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
971 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
972 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
973 * config/nds32/constraints.md: New constraints and checking for hard
974 float configuration.
975 * config/nds32/iterators.md: New mode iterator and attribute for hard
976 float configuration.
977 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
978 patterns.
979 * config/nds32/nds32-fpu.md: New file.
980 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
981 deal with hard float code generation.
982 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
983 ARCH_V3S.
984 (abi_type, float_reg_number): New enum type.
985 * config/nds32/nds32-predicates.c: New predicates for hard float.
986 * config/nds32/nds32-protos.h: Declare functions for hard float.
987 * config/nds32/nds32.c: Implementation for hard float configuration.
988 * config/nds32/nds32.h: Definitions for hard float configuration.
989 * config/nds32/nds32.md: Include hard float machine description and
990 modify patterns for hard float configuration.
991 * config/nds32/nds32.opt: New options for hard float configuration.
992 * config/nds32/predicates.md: New predicates for hard float
993 configuration.
995 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
997 * common/config/nds32/nds32-common.c
998 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1000 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1002 PR middle-end/85195
1003 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1004 CONSTRUCTOR_ELT (ctor, ...)->value.
1006 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1008 PR target/85193
1009 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1011 2018-04-05 Tom de Vries <tom@codesourcery.com>
1013 PR target/85204
1014 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1015 cond jump.
1017 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1018 Kito Cheng <kito.cheng@gmail.com>
1020 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1021 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1022 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1023 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1025 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1026 Kito Cheng <kito.cheng@gmail.com>
1028 * config/nds32/constraints.md (Ufe): New memory constraint.
1029 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1030 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1031 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1032 operands.
1033 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1034 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1036 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1038 * config/nds32/nds32.md: Use optimize_size in the condition for
1039 alu-shift instructions.
1041 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1043 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1045 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1047 * config/nds32/nds32.md (negsi2): Refine pattern.
1049 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1050 Chung-Ju Wu <jasonwucj@gmail.com>
1052 * config/nds32/iterators.md (shift_rotate): New code iterator.
1053 (shift): New code attribute.
1054 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1055 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1056 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1057 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1058 bit-wise operations.
1059 (andsi3, *andsi3): Ditto.
1060 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1061 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1062 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1063 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1064 nds32_ior_operand, nds32_xor_operand): New predicates.
1066 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1068 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1069 (addsi3, subsi3): ... this.
1071 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1073 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1075 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1077 * config/nds32/nds32.md: Adjust indention.
1079 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1081 * config/nds32/nds32.md (feature): New attribute.
1083 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1085 * config/nds32/nds32.md (subtype): New attribute.
1087 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1089 PR target/85203
1090 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1091 expansion to perform a bitwise AND of the argument followed by a
1092 boolean negation of the result.
1094 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1096 PR rtl-optimization/84878
1097 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1098 the basic block. Assert the use reference is not artificial and that
1099 it has an associated insn.
1101 2018-04-04 Michael Matz <matz@suse.de>
1103 * builtins.c (compute_objsize): Pass correct operand
1104 to array_at_struct_end_p.
1106 2018-04-04 Richard Biener <rguenther@suse.de>
1108 PR lto/85176
1109 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1110 from contexts for DINFO_LEVEL_TERSE and below.
1112 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1114 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1115 resiter_operand condition.
1116 * config/nds32/nds32.md (*move<mode>): Ditto.
1118 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1119 Monk Chiang <sh.chiang04@gmail.com>
1121 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1123 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1125 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1127 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1128 Kito Cheng <kito.cheng@gmail.com>
1130 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1131 nds32_cond_code_str, output_cond_branch,
1132 output_cond_branch_compare_zero, nds32_expand_cbranch,
1133 nds32_expand_cstore, nds32_expand_movcc,
1134 nds32_output_cbranchsi4_equality_zero,
1135 nds32_output_cbranchsi4_equality_reg,
1136 nds32_output_cbranchsi4_equality_reg_or_const_int,
1137 nds32_output_cbranchsi4_greater_less_zero: New functions.
1138 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1139 nds32_expand_cstore, nds32_expand_movcc,
1140 nds32_output_cbranchsi4_equality_zero,
1141 nds32_output_cbranchsi4_equality_reg,
1142 nds32_output_cbranchsi4_equality_reg_or_const_int,
1143 nds32_output_cbranchsi4_greater_less_zero): Declare.
1144 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1145 nds32_rimm11s_operand): New predicates.
1146 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1147 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1148 patterns.
1150 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1152 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1153 * config/nds32/nds32.md: Ditto.
1154 * config/nds32/pipelines.md: Ditto.
1156 2018-04-04 Richard Biener <rguenther@suse.de>
1158 PR tree-optimization/85168
1159 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1160 propagating abnormals.
1162 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1164 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1166 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1167 Kito Cheng <kito.cheng@gmail.com>
1169 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1170 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1171 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1172 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1173 * config/nds32/nds32.md (sibcall_internal): New.
1174 (sibcall_register): Remove.
1175 (sibcall_immediate): Remove.
1176 (sibcall_value_internal): New.
1177 (sibcall_value_register): Remove.
1178 (sibcall_value_immediate): Remove.
1179 * config/nds32/predicates.md (nds32_general_register_operand): New.
1180 (nds32_call_address_operand): New.
1182 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1184 PR rtl-optimization/85167
1185 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1186 bb_defs if *split_p, instead preinitialize it to NULL.
1188 PR tree-optimization/85156
1189 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1190 evaluating the argument multiple times.
1192 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1194 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1195 than vector.
1196 (_mm_cvtpd_ps): Likewise.
1197 (_mm_cvttpd_epi32): Likewise.
1198 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1199 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1200 vector, pixel, and bool following altivec.h include.
1202 2018-04-03 Martin Sebor <msebor@redhat.com>
1204 * doc/extend.texi (Common Function Attributes): Clarify.
1205 (const attribute): Likewise.
1206 (pure attribute): Likewise.
1208 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1210 PR target/85169
1211 * config/i386/i386.c (ix86_expand_vector_set): Use
1212 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1214 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1216 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1217 instructions when changing rounding bits to preserve precision bits
1218 in the x87 control word.
1220 2018-04-03 Martin Liska <mliska@suse.cz>
1222 PR tree-optimization/82491
1223 * rtl.h (strip_offset_and_add): Replace += suboffset with
1224 poly_uint64 () + suboffset.
1226 2018-03-29 Martin Liska <mliska@suse.cz>
1227 Martin Jambor <mjambor@suse.cz>
1229 PR ipa/84947
1230 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1231 param_type is not an integral or pointer type.
1233 2018-04-03 Richard Biener <rguenther@suse.de>
1235 * sese.h (recompute_all_dominators): Remove.
1237 2018-04-02 Martin Sebor <msebor@redhat.com>
1239 * doc/invoke.texi (-Wrestrict): Fix typos.
1241 2018-04-02 Jim Wilson <jimw@sifive.com>
1243 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1244 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1245 (<optab>di3, <optab>si3_extend): Likewise.
1246 (<optab>si3_mask, <optab>si3_mask_1): New.
1247 (<optab>di3_mask, <optab>di3_mask_1): New.
1248 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1249 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1250 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1252 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1254 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1255 example.
1257 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1259 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1260 (nds32_canonicalize_comparison): New function.
1262 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1263 Kito Cheng <kito.cheng@gmail.com>
1264 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1266 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1267 * config/nds32/constants.md (unspec_volatile_element): Add
1268 UNSPEC_VOLATILE_RELAX_GROUP.
1269 * config/nds32/nds32-relax-opt.c: New file.
1270 * config/nds32/nds32-predicates.c
1271 (nds32_symbol_load_store_p): New function.
1272 * config/nds32/nds32-protos.h
1273 (nds32_symbol_load_store_p): Declare function.
1274 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1275 * config/nds32/nds32.c
1276 (nds32_register_pass): New function to register pass.
1277 (nds32_register_passes): New function to register passes.
1278 * config/nds32/nds32.md (relax_group): New pattern.
1279 * config/nds32/nds32.opt (mrelax-hint): New option.
1280 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1282 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1284 * config/nds32/t-nds32: Modify files dependency.
1286 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1288 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1289 (PROFILE_HOOK): Define its implementation.
1291 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1293 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1294 type and 32-bit size.
1296 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1298 PR middle-end/85090
1299 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1300 (V_128_256): New mode iterator.
1301 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1302 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1303 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1304 of V.
1305 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1306 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1308 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1310 PR target/83315
1311 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1312 NaN inputs correctly.
1314 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1316 PR target/80546
1317 * config/rs6000/vsx.md (??r): New mode attribute.
1318 (*vsx_mov<mode>_64bit): Use it.
1319 (*vsx_mov<mode>_32bit): Likewise.
1321 2018-03-30 Martin Sebor <msebor@redhat.com>
1323 PR tree-optimization/84818
1324 * builtins.c (check_access): Use warning_n.
1326 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1328 PR target/83822
1329 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1330 condition.
1331 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1332 condition.
1334 2018-03-30 Julia Koval <julia.koval@intel.com>
1336 PR target/84413
1337 * x86-tune.def (movx, partial_reg_dependency): Enable for
1338 m_SKYLAKE_AVX512.
1340 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1342 PR inline-asm/84985
1343 * lra-constraints.c (process_alt_operands): Move setting
1344 this_alternative_matches below.
1346 2018-03-29 Martin Liska <mliska@suse.cz>
1348 PR lto/84995.
1349 * doc/invoke.texi: Document how LTO works with debug info.
1350 Describe auto-load support of binutils. Mention 'x86-64'
1351 as valid option value of -march option.
1353 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1355 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1357 PR c/85094
1358 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1359 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1360 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1361 checking.
1363 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1365 PR target/84912
1366 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1367 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1368 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1369 for RS6000_BTM_POWERPC64.
1370 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1371 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1372 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1373 definition.
1374 (DIVDE): Use it.
1375 (DIVDEU): Likewise.
1377 2018-03-28 Carl Love <cel@us.ibm.com>
1379 Revert
1380 2017-09-27 Carl Love <cel@us.ibm.com>
1382 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1383 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1384 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1385 fctiw instruction.
1387 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1389 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1390 instead of __vector bool.
1391 (_mm_max_pu8): Likewise.
1392 (_mm_min_pi16): Likewise.
1394 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1396 PR target/84912
1397 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1398 (DIVWEUO): Likewise.
1399 (DIVDEO): Likewise.
1400 (DIVDEUO): Likewise.
1401 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1402 DIVWEUO and DIVDEUO.
1403 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1404 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1405 (div_extend): Likewise.
1406 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1407 builtin function.
1408 (__builtin_divweuo): Likewise.
1409 (__builtin_divdeo): Likewise.
1410 (__builtin_divdeuo): Likewise.
1412 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1414 PR target/85095
1415 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1416 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1418 PR tree-optimization/82004
1419 * gimple-match-head.c (optimize_pow_to_exp): New function.
1420 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1421 Don't fold to exp if optimize_pow_to_exp is false.
1423 2018-03-28 Martin Liska <mliska@suse.cz>
1425 PR other/84819
1426 * calls.c (initialize_argument_information): Fix trailing space.
1427 * common.opt: Fix typo and provide better explanation for
1428 -fsanitize-coverage option.
1429 * config/i386/i386.opt: Fix typo.
1431 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1432 Martin Liska <mliska@suse.cz>
1434 PR sanitizer/85081
1435 * gimplify.c (asan_poison_variable): Don't do the check for
1436 gimplify_omp_ctxp here.
1437 (gimplify_decl_expr): Do it here.
1438 (gimplify_target_expr): Likewise.
1440 2018-03-28 Martin Liska <mliska@suse.cz>
1442 PR target/84988
1443 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1444 chkp_type_bounds_count if MPX is not enabled.
1446 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1448 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1450 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1452 PR target/84914
1453 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1454 function to create the function decl for complex long double
1455 multiply and divide for -mabi=ieeelongdouble.
1456 (init_float128_ieee): Call it.
1458 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
1460 PR target/85044
1461 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1462 -fcf-protection=branch -mibt.
1463 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
1465 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1467 PR target/81863
1468 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
1470 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
1472 PR target/85056
1473 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
1474 extern array declarations.
1476 2018-03-27 Richard Biener <rguenther@suse.de>
1478 PR middle-end/84067
1479 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
1480 explicit single_use checks.
1482 2018-03-27 Richard Biener <rguenther@suse.de>
1484 PR tree-optimization/85082
1485 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
1486 Valueize the VUSE.
1488 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1490 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
1491 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
1492 Turn on fasynchronous-unwind-tables and funwind-tables.
1494 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
1496 PR target/85073
1497 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1498 (*bmi_blsr_<mode>_ccz): Ditto.
1500 2018-03-26 Tom de Vries <tom@codesourcery.com>
1502 PR tree-optimization/85063
1503 * omp-general.c (offloading_function_p): New function. Factor out
1504 of ...
1505 * omp-offload.c (pass_omp_target_link::gate): ... here.
1506 * omp-general.h (offloading_function_p): Declare.
1507 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1508 with attribute omp declare target for offloading functions.
1510 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
1512 PR tree-optimization/84005
1513 * tree-data-ref.h (get_base_for_alignment): Declare.
1514 * tree-data-ref.c (get_base_for_alignment_1): New function.
1515 (get_base_for_alignment): Likewise.
1516 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1517 get_base_for_alignment to find a suitable base object, instead
1518 of always using drb->base_address.
1520 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1522 PR inline-asm/85022
1523 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1524 known size by default.
1526 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1528 PR inline-asm/85030
1529 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1530 and non BLKmode operands.
1532 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1534 PR target/85026
1535 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1536 Clean up attributes.
1538 2018-03-23 Richard Biener <rguenther@suse.de>
1540 PR debug/85020
1541 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1542 we are going to emit early debug for LTO.
1544 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1546 PR inline-asm/85034
1547 * function.c (match_asm_constraints_1): Don't optimize if input
1548 doesn't satisfy general_operand predicate for output's mode.
1550 PR inline-asm/85022
1551 * alias.c (write_dependence_p): Don't require for x_canonicalized
1552 non-VOIDmode if x has VOIDmode.
1554 PR sanitizer/85029
1555 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1556 just don't try to optimize it rather than assert it never happens.
1558 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1560 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1561 macro expansions for definition of ST_INTERNAL_<mode> and
1562 LD_INTERNAL_<mode> builtins.
1563 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1564 Remove prototype.
1565 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1566 function.
1567 (altivec_expand_st_builtin): Likewise.
1568 (altivec_expand_builtin): Remove calls to deleted functions.
1569 (rs6000_address_for_altivec): Delete this function.
1570 * config/rs6000/vector.md: Remove expands for
1571 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1573 2018-03-22 Sudakshina Das <sudi.das@arm.com>
1575 PR target/84826
1576 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1577 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1578 re-computing once computed.
1579 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1580 (arm_init_machine_status): Initialize
1581 machine->static_chain_stack_bytes.
1583 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1585 PR target/84760
1586 * doc/extend.texi: Add four new prototypes for vec_ld.
1587 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1588 definitions for more logical presentation.
1589 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1590 entries for V1TI variants of __builtin_altivec_ld builtin.
1591 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1592 handling of V1TI variant of LVX icode pattern.
1593 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1594 (rs6000_gimple_fold_builtin): Likewise.
1595 (altivec_init_builtins): Add code to define
1596 __builtin_altivec_lvx_v1ti function.
1598 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1600 PR inline-asm/84941
1601 * function.c (match_asm_constraints_1): Don't do the optimization
1602 if input isn't a REG, SUBREG, MEM or constant.
1604 2018-03-22 Tom de Vries <tom@codesourcery.com>
1606 PR tree-optimization/84956
1607 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1608 bb_has_abnormal_pred.
1610 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1612 PR sanitizer/85018
1613 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1614 DECL_INITIAL (decl) to decl at the end.
1615 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1616 adjust the comment.
1618 2018-03-21 Joseph Myers <joseph@codesourcery.com>
1620 * doc/extend.texi (__builtin_tgmath): Document when complex
1621 integer types are treated as _Complex _Float64.
1623 2018-03-21 Tom de Vries <tom@codesourcery.com>
1625 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1627 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1629 PR tree-optimization/84960
1630 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1631 if it is ENTRY block, move them into single succ of ENTRY in that case.
1633 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
1635 PR tree-optimization/84811
1636 * poly-int.h (poly_span_traits): Remove the T3 parameter and
1637 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1638 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1639 (known_subrange_p): Update accordingly. Cast each value involved
1640 in the size comparison, rather than casting the result of the
1641 subtraction.
1643 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1645 PR tree-optimization/84982
1646 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1647 by flipping the least significant bit rather than all bits from
1648 bitpos to bitpos + bitsize - 1.
1650 2018-03-21 Nathan Sidwell <nathan@acm.org>
1652 * doc/extend.texi (Deprecated Features): Remove mention of
1653 long-deleted deprecations.
1655 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1657 PR jit/84288
1658 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1659 * configure: Regenerate.
1661 2018-03-21 Tom de Vries <tom@codesourcery.com>
1663 PR tree-optimization/83126
1664 * tree-parloops.c (num_phis): New function.
1665 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1667 2018-03-21 Nathan Sidwell <nathan@acm.org>
1669 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1670 mention anon-struct/union members and trailing attributes.
1672 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1674 PR tree-optimization/84969
1675 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1676 builtin memset partitions if they set different rhs values.
1678 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1680 PR rtl-optimization/84989
1681 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1682 VEC_DUPLICATE with scalar result mode.
1684 2018-03-21 Martin Liska <mliska@suse.cz>
1686 PR ipa/84963
1687 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1688 not intended return statement.
1690 2018-03-21 Martin Liska <mliska@suse.cz>
1692 PR target/84988
1693 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1694 (chkp_find_bound_slots_1): Limit number of iterations.
1696 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
1698 PR target/84838
1699 * Minor grammar fixes for x86 options.
1701 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1703 PR debug/84875
1704 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1705 holding REG_CFA_RESTORE notes, instead turn them into a USE.
1707 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
1709 PR target/83789
1710 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1711 (altivec_lvx_<mode>_1op): Likewise.
1712 (altivec_stvx_<mode>_2op): Likewise.
1713 (altivec_stvx_<mode>_1op): Likewise.
1714 (altivec_lvx_<VM2:mode>): New define_expand.
1715 (altivec_stvx_<VM2:mode>): Likewise.
1716 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1717 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1718 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1719 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1720 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1721 (rs6000_gen_lvx): Likewise.
1722 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1723 (altivec_expand_stv_builtin): Likewise.
1724 (altivec_expand_builtin): Likewise.
1725 * config/rs6000/vector.md: Likewise.
1727 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1729 PR target/82518
1730 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1731 BYTES_BIG_ENDIAN.
1733 2018-03-20 Richard Biener <rguenther@suse.de>
1735 PR target/84986
1736 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1737 sign-conversions as zero, fall back to standard scalar_stmt
1738 cost for the rest.
1740 2018-03-20 Martin Liska <mliska@suse.cz>
1742 PR ipa/84825
1743 * predict.c (rebuild_frequencies): Handle case when we have
1744 PROFILE_ABSENT, but flag_guess_branch_prob is false.
1746 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1748 PR target/84990
1749 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1750 flag_section_anchors.
1751 * varasm.c (use_blocks_for_decl_p): Remove hack for
1752 dw2_force_const_mem.
1754 PR target/84845
1755 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
1756 to ...
1757 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
1758 be created, use lowpart_subreg of operands[0] rather than operands[0]
1759 itself.
1760 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
1761 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
1762 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
1763 and n constraint instead of aarch64_shift_imm_di and Usd.
1764 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
1765 (*aarch64_<optab>_reg_minus<mode>3): ... this.
1767 2018-03-20 Sudakshina Das <sudi.das@arm.com>
1769 PR target/82989
1770 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
1771 to favor GPR over NEON registers.
1772 (<shift>di3_neon): Likewise.
1774 2018-03-20 Tom de Vries <tom@codesourcery.com>
1776 PR target/84952
1777 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
1778 (nvptx_process_pars): Emit bar.sync asap and alap.
1780 2018-03-20 Tom de Vries <tom@codesourcery.com>
1782 PR target/84954
1783 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
1784 seen_label if seen_label is already set.
1786 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1788 PR target/84945
1789 * config/i386/i386.c (fold_builtin_cpu): For features above 31
1790 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
1791 Use 1U instead of 1. Formatting fixes.
1793 PR c/84953
1794 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
1795 instead of TREE_TYPE (s1) for the return value.
1797 2018-03-19 Jakub Jelinek <jakub@redhat.com>
1799 PR tree-optimization/84946
1800 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
1801 bitsize + bitsize in poly_uint64 rather than poly_int64.
1803 PR sanitizer/78651
1804 * dwarf2asm.c: Include fold-const.c.
1805 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1806 of decl rather than decl itself.
1808 PR rtl-optimization/84643
1809 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1811 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
1813 PR sanitizer/78651
1814 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1815 calling assemble_variable.
1817 2018-03-19 Sudakshina Das <sudi.das@arm.com>
1819 PR target/81647
1820 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1821 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1823 2018-03-19 Jim Wilson <jimw@sifive.com>
1825 PR bootstrap/84856
1826 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1827 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1828 (riscv_first_stack_step): Likewise.
1829 (riscv_option_override): Use STACK_BOUNDARY instead of
1830 MIN_STACK_BOUNDARY.
1831 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1832 MIN_STACK_BOUNDARY.
1833 (BIGGEST_ALIGNMENT): Set to 128.
1834 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1835 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1836 STACK_BOUNDARY.
1838 2018-03-19 Richard Biener <rguenther@suse.de>
1840 PR tree-optimization/84933
1841 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1842 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1844 2018-03-19 Richard Biener <rguenther@suse.de>
1846 PR tree-optimization/84859
1847 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1848 (cond_if_else_store_replacement): Perform sinking operation on
1849 single-store BBs regardless of MAX_STORES_TO_SINK setting.
1850 Generalize what a BB with a single eligible store is.
1852 2018-03-19 Richard Biener <rguenther@suse.de>
1854 PR tree-optimization/84929
1855 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1856 chrec_is_positive against non-chrec arg.
1858 2018-03-19 Tamar Christina <tamar.christina@arm.com>
1860 PR target/84711
1861 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1863 2018-03-18 Martin Liska <mliska@suse.cz>
1865 PR rtl-optimization/84635
1866 * regrename.c (build_def_use): Use matches_mode only when
1867 matches >= 0.
1869 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1871 PR tree-optimization/84913
1872 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1873 vectorize chains of COND_EXPRs.
1875 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1877 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1879 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1881 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1883 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1885 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1887 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1888 Kito Cheng <kito.cheng@gmail.com>
1890 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1891 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1892 (nds32_adjust_reg_alloc_order): New function.
1893 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1895 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1897 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1898 nds32_print_operand, nds32_print_operand_address): Use
1899 HOST_WIDE_INT_PRINT_DEC instead.
1901 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1903 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1905 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1907 PR target/84902
1908 * config/i386/i386.c (initial_ix86_tune_features,
1909 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1910 unsigned long long.
1911 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1912 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1913 rather than 1u << ix86_tune. Formatting fix.
1914 (ix86_option_override_internal): Change ix86_arch_mask from
1915 unsigned int to unsigned HOST_WIDE_INT, initialize to
1916 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1917 (ix86_function_specific_restore): Likewise.
1919 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1921 PR target/84899
1922 * postreload.c (reload_combine_recognize_pattern): Perform
1923 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1924 truncate_int_for_mode the result for the destination's mode.
1926 PR c/84909
1927 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1928 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1929 Likewise.
1930 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1931 Likewise.
1933 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
1935 PR target/84876
1936 * lra-assigns.c (lra_split_hard_reg_for): Don't use
1937 regno_allocno_class_array and sorted_pseudos.
1938 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
1939 insns where regno is used.
1941 2018-03-16 Martin Liska <mliska@suse.cz>
1943 PR ipa/84833
1944 * multiple_target.c (create_dispatcher_calls): Redirect
1945 reference in the symbol table.
1947 2018-03-16 Martin Liska <mliska@suse.cz>
1949 PR ipa/84722
1950 * multiple_target.c (create_dispatcher_calls): Redirect also
1951 an alias.
1953 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1955 PR c++/79937
1956 PR c++/82410
1957 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
1958 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
1959 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
1961 2018-03-16 Julia Koval <julia.koval@intel.com>
1963 * doc/invoke.texi (Skylake Server): Add CLWB.
1964 Cannonlake): Remove CLWB.
1966 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1968 PR tree-optimization/84841
1969 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
1970 1 << 3.
1971 (FLOAT_ONE_CONST_TYPE): Define.
1972 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
1973 (sort_by_operand_rank): Put entries with higher constant_type last
1974 rather than first to match comments.
1976 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
1978 * config/nios2/nios2.md (movsi_internal): Fix thinko in
1979 split predicate.
1981 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1983 PR c++/79085
1984 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
1985 check and use address of target always.
1987 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
1989 PR target/84574
1990 * config/i386/i386.c (indirect_thunk_needed): Update comments.
1991 (indirect_thunk_bnd_needed): Likewise.
1992 (indirect_thunks_used): Likewise.
1993 (indirect_thunks_bnd_used): Likewise.
1994 (indirect_return_needed): New.
1995 (indirect_return_bnd_needed): Likewise.
1996 (output_indirect_thunk_function): Add a bool argument for
1997 function return.
1998 (output_indirect_thunk_function): Don't generate alias for
1999 function return thunk.
2000 (ix86_code_end): Call output_indirect_thunk_function to generate
2001 function return thunks.
2002 (ix86_output_function_return): Set indirect_return_bnd_needed
2003 and indirect_return_needed instead of indirect_thunk_bnd_needed
2004 and indirect_thunk_needed.
2006 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2008 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2009 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2010 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2012 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2013 Paul Hua <paul.hua.gm@gmail.com>
2015 PR c/84852
2016 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2018 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2020 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2021 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2022 resp. SFmode cases.
2024 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2026 PR target/84711
2027 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2028 instead of GET_MODE_SIZE when comparing Units.
2030 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2032 PR target/68256
2033 * varasm.c (hash_section): Return an unchangeble hash value
2034 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2035 Return !aarch64_can_use_per_function_literal_pools_p ().
2037 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2039 PR target/84860
2040 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2041 rather than address of cmode as last argument to prepare_cmp_insn.
2043 2018-03-15 Julia Koval <julia.koval@intel.com>
2045 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2046 F_AVX512VNNI, F_AVX512BITALG): New.
2048 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2050 PR target/83451
2051 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2052 insn for floating-point loads and stores.
2054 2018-03-14 Carl Love <cel@us.ibm.com>
2056 * config/rs6000/rs6000-c.c: Add macro definitions for
2057 ALTIVEC_BUILTIN_VEC_PERMXOR.
2058 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2059 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2060 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2061 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2062 UNSPEC_VPERMXOR.
2063 * config/doc/extend.texi: Add prototypes for vec_permxor.
2065 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2067 PR c/84852
2068 * diagnostic-show-locus.c (class layout_point): Convert m_line
2069 from int to linenum_type.
2070 (line_span::comparator): Use linenum "compare" function when
2071 comparing line numbers.
2072 (test_line_span): New function.
2073 (layout_range::contains_point): Convert param "row" from int to
2074 linenum_type.
2075 (layout_range::intersects_line_p): Likewise.
2076 (layout::will_show_line_p): Likewise.
2077 (layout::print_source_line): Likewise.
2078 (layout::should_print_annotation_line_p): Likewise.
2079 (layout::print_annotation_line): Likewise.
2080 (layout::print_leading_fixits): Likewise.
2081 (layout::annotation_line_showed_range_p): Likewise.
2082 (struct line_corrections): Likewise for field m_row.
2083 (line_corrections::line_corrections): Likewise for param "row".
2084 (layout::print_trailing_fixits): Likewise.
2085 (layout::get_state_at_point): Likewise.
2086 (layout::get_x_bound_for_row): Likewise.
2087 (layout::print_line): Likewise.
2088 (diagnostic_show_locus): Likewise for locals "last_line" and
2089 "row".
2090 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2091 * input.c (selftest::test_linenum_comparisons): New function.
2092 (selftest::input_c_tests): Call it.
2093 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2094 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2095 * selftest.h (ASSERT_GT): New macro.
2096 (ASSERT_GT_AT): New macro.
2097 (ASSERT_LT): New macro.
2098 (ASSERT_LT_AT): New macro.
2100 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2102 PR rtl-optimization/84780
2103 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2105 2018-03-14 Martin Liska <mliska@suse.cz>
2107 * tree.c (record_node_allocation_statistics): Use
2108 get_stats_node_kind.
2109 (get_stats_node_kind): New function extracted from
2110 record_node_allocation_statistics.
2111 (free_node): Use get_stats_node_kind.
2113 2018-03-14 Richard Biener <rguenther@suse.de>
2115 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2116 that the value-set of ANTIC_IN doesn't grow.
2118 Revert
2119 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2120 member.
2121 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2122 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2124 2018-03-14 Julia Koval <julia.koval@intel.com>
2126 * config.gcc (icelake-client, icelake-server): New.
2127 (icelake): Remove.
2128 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2129 (initial_ix86_arch_features): Ditto.
2130 (PTA_SKYLAKE): Add SGX.
2131 (PTA_ICELAKE): Remove.
2132 (PTA_ICELAKE_CLIENT): New.
2133 (PTA_ICELAKE_SERVER): New.
2134 (ix86_option_override_internal): Split up icelake on icelake client and
2135 icelake server.
2136 (get_builtin_code_for_version): Ditto.
2137 (fold_builtin_cpu): Ditto.
2138 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2139 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2140 * config/i386/i386.h (processor_type): Ditto.
2141 * doc/invoke.texi: Ditto.
2143 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2145 PR sanitizer/83392
2146 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2147 INTEGER_CST offset, add it together with bitpos / 8 and
2148 sign extend based on POINTER_SIZE.
2150 PR target/84844
2151 Revert
2152 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2154 PR target/78090
2155 * config/i386/constraints.md (Yc): New register constraint.
2156 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2157 Use Yc constraint for alternative 2 of operand 0. Remove
2158 preferred_for_speed attribute.
2160 2018-03-14 Richard Biener <rguenther@suse.de>
2162 PR tree-optimization/84830
2163 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2164 with the old one to avoid oscillations.
2166 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2168 PR target/83712
2169 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2170 pseudos.
2171 (assign_by_spills): Return a flag of reload assignment failure.
2172 Do not process the reload assignment failures. Do not spill other
2173 reload pseudos if they has the same reg class. Update n if
2174 necessary.
2175 (lra_assign): Add a return arg. Set up from the result of
2176 assign_by_spills call.
2177 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2178 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2179 usage_insns if it is not NULL.
2180 (spill_hard_reg_in_range): New function.
2181 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2182 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2183 function prototypes.
2184 (lra_assign): Change prototype.
2185 * lra.c (lra): Add code to deal with fails by splitting hard reg
2186 live ranges.
2188 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2190 * config/riscv/riscv.opt (mrelax): New option.
2191 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2192 "norelax" when riscv_mrelax is disabled.
2193 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2195 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2197 PR target/84743
2198 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2199 reassociation for int modes.
2201 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2203 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2204 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2205 for big-endian.
2206 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2207 * config/aarch64/aarch64-sve.md
2208 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2209 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2210 (*extend<mode><Vwide>2): Rename to...
2211 (aarch64_sve_extend<mode><Vwide>2): ...this.
2212 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2213 renaming the old pattern to...
2214 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2215 unsigned packs.
2216 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2217 define_expand, renaming the old pattern to...
2218 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2219 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2220 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2221 account when deciding which SVE instruction the optab should use.
2222 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2224 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2226 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2227 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2228 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2229 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2230 (tlsdesc_small_<mode>): Turn a define_expand and use
2231 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2232 (tlsdesc_small_advsimd_<mode>): ...this.
2233 (tlsdesc_small_sve_<mode>): New pattern.
2235 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2237 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2238 (UNSPEC_UMUL_HIGHPART): New constants.
2239 (MUL_HIGHPART): New int iteraor.
2240 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2241 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2242 define_expand.
2243 (*<su>mul<mode>3_highpart): New define_insn.
2245 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2247 PR lto/84805
2248 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2249 incomplete types.
2251 2018-03-13 Martin Liska <mliska@suse.cz>
2253 PR ipa/84658.
2254 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2255 vector.
2256 (sem_item_optimizer::~sem_item_optimizer): Release it.
2257 (sem_item_optimizer::merge_classes): Register variable aliases.
2258 (sem_item_optimizer::fixup_pt_set): New function.
2259 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2260 * ipa-icf.h: Declare new variables and functions.
2262 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2264 PR middle-end/84834
2265 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2266 integer_pow2p@2 and test integer_pow2p in condition.
2267 (A < 0 ? C : 0): Similarly for @1.
2269 PR middle-end/84831
2270 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2271 characters starting at p contain '\0' character, don't look beyond
2272 that.
2274 PR target/84827
2275 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2276 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2278 PR target/84828
2279 * reg-stack.c (change_stack): Change update_end var from int to
2280 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2281 also call set_block_for_insn on the newly added insns and rescan.
2283 PR target/84786
2284 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2285 on the last operand.
2287 PR c++/84704
2288 * tree.c (stabilize_reference_1): Return save_expr (e) for
2289 STATEMENT_LIST even if it doesn't have side-effects.
2291 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2293 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2295 2018-03-12 Renlin Li <renlin.li@arm.com>
2297 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2298 aarch64_output_scalar_simd_mov_immediate.
2300 2018-03-12 Martin Sebor <msebor@redhat.com>
2302 PR tree-optimization/83456
2303 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2304 for perfectly overlapping calls to memcpy.
2305 (gimple_fold_builtin_memory_chk): Same.
2306 (gimple_fold_builtin_strcpy): Handle no-warning.
2307 (gimple_fold_builtin_stxcpy_chk): Same.
2308 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2310 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2312 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2313 parameter. Use it for SFmode.
2314 (rs6000_function_arg_advance_1): Adjust.
2315 (rs6000_function_arg): Adjust.
2316 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2318 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2320 PR rtl-optimization/84169
2321 PR rtl-optimization/84780
2322 * combine.c (can_combine_p): Check for a 2-insn combination whether
2323 the destination register is used between the two insns, too.
2325 2018-03-12 Richard Biener <rguenther@suse.de>
2327 PR tree-optimization/84803
2328 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2329 for refs DR analysis didn't process.
2331 2018-03-12 Richard Biener <rguenther@suse.de>
2333 PR tree-optimization/84777
2334 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2335 force-vectorize loops ignore whether we are optimizing for size.
2337 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2339 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2340 (TARGET_MD_ASM_ADJUST): Define.
2342 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2343 Kito Cheng <kito.cheng@gmail.com>
2344 Chung-Ju Wu <jasonwucj@gmail.com>
2346 * config/nds32/nds32.c (nds32_compute_stack_frame,
2347 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2348 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2349 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2350 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2351 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2352 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2353 * config/nds32/nds32.md (prologue, epilogue): Use macro
2354 NDS32_V3PUSH_AVAILABLE_P to do checking.
2356 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2358 PR debug/58150
2359 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2360 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2361 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2362 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2363 addition of most attributes on !orig_type_die or the attribute not
2364 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2366 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2367 Chung-Ju Wu <jasonwucj@gmail.com>
2369 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2370 __NDS32_VH__ macro.
2371 * config/nds32/nds32.opt (mvh): New option.
2373 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2374 Chung-Ju Wu <jasonwucj@gmail.com>
2376 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2377 function.
2378 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2379 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2380 definition.
2382 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2383 Chung-Ju Wu <jasonwucj@gmail.com>
2385 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2386 function.
2387 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2388 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2390 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2391 Kito Cheng <kito.cheng@gmail.com>
2392 Chung-Ju Wu <jasonwucj@gmail.com>
2394 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2395 UNSPEC_FFMISM and UNSPEC_FLMISM.
2396 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2397 for ffb, ffmism and flmism.
2398 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2399 (unspec_ffmism): Ditto.
2400 (unspec_flmism): Ditto.
2401 (nds32_expand_builtin_impl): Check if string extension is available.
2402 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2403 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2405 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2407 Reverting patch:
2408 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2410 PR target/83712
2411 * lra-assigns.c (assign_by_spills): Return a flag of reload
2412 assignment failure. Do not process the reload assignment
2413 failures. Do not spill other reload pseudos if they has the same
2414 reg class.
2415 (lra_assign): Add a return arg. Set up from the result of
2416 assign_by_spills call.
2417 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2418 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2419 usage_insns if it is not NULL.
2420 (spill_hard_reg_in_range): New function.
2421 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2422 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2423 function prototypes.
2424 (lra_assign): Change prototype.
2425 * lra.c (lra): Add code to deal with fails by splitting hard reg
2426 live ranges.
2428 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
2430 PR target/84807
2431 * config/i386/i386.opt: Replace Enforcment with Enforcement.
2433 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2435 PR debug/84620
2436 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2437 (dw_val_node): Add val_symbolic_view.
2438 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2439 (symview_upper_bound): New.
2440 (new_line_info_table): Initialize symviews_since_reset.
2441 (dwarf2out_source_line): Count symviews_since_reset and set
2442 symview_upper_bound.
2443 (dw_val_equal_p): Handle symview.
2444 (add_AT_symview): New.
2445 (print_dw_val): Handle symview.
2446 (attr_checksum, attr_checksum_ordered): Likewise.
2447 (same_dw_val_p, size_of_die): Likewise.
2448 (value_format, output_die): Likewise.
2449 (add_high_low_attributes): Use add_AT_symview for entry_view.
2450 (dwarf2out_finish): Reset symview_upper_bound, clear
2451 zero_view_p.
2453 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2455 PR target/83969
2456 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2457 Add strict argument and use it.
2458 (rs6000_split_multireg_move): Update for new strict argument.
2459 (mem_operand_gpr): Disallow all non-offsettable addresses.
2460 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2462 2018-03-09 Jakub Jelinek <jakub@redhat.com>
2464 PR target/84772
2465 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
2466 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
2467 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
2469 PR c++/84767
2470 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
2471 decl, use remap_type if we want to use the type.
2473 2018-03-09 Martin Sebor <msebor@redhat.com>
2475 PR tree-optimization/84526
2476 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2477 Remove dead code.
2478 (builtin_access::generic_overlap): Be prepared to handle non-array
2479 base objects.
2481 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
2483 PR rtl-optimization/84682
2484 * lra-constraints.c (process_address_1): Check is_address flag
2485 for address constraints.
2486 (process_alt_operands): Likewise.
2487 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
2488 preprocess_constraints.
2489 * recog.h (preprocess_constraints): Add oploc parameter.
2490 Adjust callers.
2491 * recog.c (preprocess_constraints): Test address_operand for
2492 CT_ADDRESS constraints.
2494 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2496 PR target/83712
2497 * lra-assigns.c (assign_by_spills): Return a flag of reload
2498 assignment failure. Do not process the reload assignment
2499 failures. Do not spill other reload pseudos if they has the same
2500 reg class.
2501 (lra_assign): Add a return arg. Set up from the result of
2502 assign_by_spills call.
2503 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2504 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2505 usage_insns if it is not NULL.
2506 (spill_hard_reg_in_range): New function.
2507 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2508 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2509 function prototypes.
2510 (lra_assign): Change prototype.
2511 * lra.c (lra): Add code to deal with fails by splitting hard reg
2512 live ranges.
2514 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2516 PR target/83193
2517 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2518 Accept complain bool parameter. Only emit errors if it is true.
2519 (arm_parse_cpu_option_name): Likewise.
2520 (arm_target_thumb_only): Adjust callers of the above.
2521 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2522 prototype to take a default true bool parameter.
2523 (arm_parse_arch_option_name): Likewise.
2525 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2526 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2528 PR jit/64089
2529 PR jit/84288
2530 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2531 * configure: Regenerate.
2532 * configure.ac ("linker --version-script option"): New.
2533 ("linker soname option"): New.
2535 2018-03-09 Richard Biener <rguenther@suse.de>
2537 PR tree-optimization/84775
2538 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2539 immediate uses of predicate stmts and mark them modified.
2541 Revert
2542 PR tree-optimization/84178
2543 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2544 to caller.
2545 (version_loop_for_if_conversion): Delay update_ssa call.
2546 (tree_if_conversion): Delay update_ssa until after predicate
2547 insertion.
2549 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2551 PR target/84763
2552 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2553 when the function accesses prior frames.
2555 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2557 PR debug/84456
2558 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2559 gen_llsym, otherwise call maybe_gen_llsym.
2561 PR inline-asm/84742
2562 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2563 has ',' character inside of it.
2565 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2567 PR target/84748
2568 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2569 as clobbering CC_REGNUM.
2571 2018-03-08 Richard Biener <rguenther@suse.de>
2573 PR middle-end/84552
2574 * tree-scalar-evolution.c: Include tree-into-ssa.h.
2575 (follow_copies_to_constant): Do not follow SSA names registered
2576 for update.
2578 2018-03-08 Richard Biener <rguenther@suse.de>
2580 PR tree-optimization/84178
2581 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2582 to caller.
2583 (version_loop_for_if_conversion): Delay update_ssa call.
2584 (tree_if_conversion): Delay update_ssa until after predicate
2585 insertion.
2587 2018-03-08 David Malcolm <dmalcolm@redhat.com>
2589 PR tree-optimization/84178
2590 * tree-if-conv.c (release_bb_predicate): Remove the
2591 the assertion that the stmts have NULL use_ops.
2592 Discard the statements, asserting that they haven't
2593 yet been added to a BB.
2595 2018-03-08 Richard Biener <rguenther@suse.de>
2597 PR tree-optimization/84746
2598 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2599 (phi_translate): Pass in destination ANTIC_OUT set.
2600 (phi_translate_1): Likewise. For a simplified result lookup
2601 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2602 (phi_translate_set): Adjust.
2603 (do_pre_regular_insertion): Likewise.
2604 (do_pre_partial_partial_insertion): Likewise.
2606 2018-03-08 Martin Liska <mliska@suse.cz>
2608 PR gcov-profile/84735
2609 * doc/gcov.texi: Document usage of profile files.
2610 * gcov-io.h: Document changes in the format.
2612 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
2614 PR debug/84404
2615 PR debug/84408
2616 * dwarf2out.c (struct dw_line_info_table): Update comments for
2617 view == -1.
2618 (FORCE_RESET_NEXT_VIEW): New.
2619 (FORCE_RESETTING_VIEW_P): New.
2620 (RESETTING_VIEW_P): Check for -1 too.
2621 (ZERO_VIEW_P): Likewise.
2622 (new_line_info_table): Force-reset next view.
2623 (dwarf2out_begin_function): Likewise.
2624 (dwarf2out_source_line): Simplify zero_view_p initialization.
2625 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2626 view directly. Omit view when omitting .loc at line 0.
2628 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2630 PR tree-optimization/84740
2631 * tree-switch-conversion.c (process_switch): Call build_constructors
2632 only if info.phi_count is non-zero.
2634 PR tree-optimization/84739
2635 * tree-tailcall.c (find_tail_calls): Check call arguments against
2636 DECL_ARGUMENTS (current_function_decl) rather than
2637 DECL_ARGUMENTS (func) when checking for tail recursion.
2639 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2641 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2642 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2643 Volker Reichelt's entry and add entries for people that perform
2644 GCC fuzzy testing and report numerous bugs.
2646 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2648 PR target/82411
2649 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2650 readonly data in sdata, if that is disabled.
2651 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2652 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2653 -mreadonly-in-sdata option.
2655 2018-03-07 Martin Sebor <msebor@redhat.com>
2657 PR tree-optimization/84468
2658 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2659 basic block when looking for nul assignment.
2661 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2663 PR target/84277
2664 * except.h (output_function_exception_table): Adjust prototype.
2665 * except.c (output_function_exception_table): Remove FNNAME parameter
2666 and add SECTION parameter. Ouput one part of the table at a time.
2667 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2668 the first part of the exception table and emit unwind directives.
2669 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2670 (i386_pe_seh_cold_init): Likewise.
2671 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2672 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2673 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2674 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2675 * config/i386/winnt.c: Include except.h.
2676 (struct seh_frame_state): Add reg_offset, after_prologue and
2677 in_cold_section fields.
2678 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2679 (i386_pe_seh_cold_init): New function.
2680 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2681 to seh->in_cold_section.
2682 (seh_emit_push): Record the offset of the push.
2683 (seh_emit_save): Record the offet of the save.
2684 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2685 Test seh->after_prologue to disregard the epilogue.
2686 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2687 (i386_pe_end_cold_function): New function.
2689 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2691 PR fortran/84565
2692 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2693 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2695 PR c++/84704
2696 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2697 on tmp_var.
2698 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2699 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2701 PR middle-end/84723
2702 * multiple_target.c: Include tree-inline.h and intl.h.
2703 (expand_target_clones): Diagnose and fail if node->definition and
2704 !tree_versionable_function_p (node->decl).
2706 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
2708 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2709 sprint_ul.
2710 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2711 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2712 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2714 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2716 PR target/84710
2717 * combine.c (try_combine): Use reg_or_subregno instead of handling
2718 just paradoxical SUBREGs and REGs.
2720 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
2722 * config/arc/arc.c (arc_finalize_pic): Remove function.
2723 (arc_must_save_register): We use single base PIC register, remove
2724 checks to save/restore the PIC register.
2725 (arc_expand_prologue): Likewise.
2726 * config/arc/arc-protos.h (arc_set_default_type_attributes):
2727 Remove.
2728 (arc_verify_short): Likewise.
2729 (arc_attr_type): Likewise.
2730 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2731 (walk_stores): Likewise.
2732 (arc_address_cost): Make it static.
2733 (arc_verify_short): Likewise.
2734 (branch_dest): Likewise.
2735 (arc_attr_type): Likewise.
2736 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2737 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2738 (arc_final_prescan_insn): Remove inserting the nops due to
2739 hardware hazards. It is done in reorg step.
2740 (insn_length_variant_t): Remove.
2741 (insn_length_parameters_t): Likewise.
2742 (arc_insn_length_parameters): Likewise.
2743 (arc_get_insn_variants): Likewise.
2744 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2746 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2748 PR inline-asm/84683
2749 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2750 assertion failure.
2752 PR tree-optimization/84687
2753 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
2754 on new_node->decl.
2755 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
2757 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2759 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
2760 Rename to ppc_speculation_barrier.
2761 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
2762 __builtin_ppc_speculation_barrier.
2764 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2766 PR target/84700
2767 * combine.c (combine_simplify_rtx): Don't try to simplify if
2768 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
2769 are equal to x.
2771 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
2773 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
2774 to 32 bytes when compiling for POWER9.
2776 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2778 PR target/84564
2779 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
2780 regparm >= 3 with no arg reg available also for calls with
2781 flag_force_indirect_call. Pass decl to ix86_function_regparm.
2783 PR target/84524
2784 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
2785 orig,vex.
2786 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
2788 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
2790 PR target/84264
2791 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
2793 2018-03-05 Richard Biener <rguenther@suse.de>
2795 PR tree-optimization/84486
2796 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
2797 When inserting a __builtin_assume_aligned call set the LHS
2798 SSA name alignment info accordingly.
2800 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
2802 PR tree-optimization/84114
2803 * config/aarch64/aarch64.c (aarch64_reassociation_width)
2804 Avoid reassociation of FLOAT_MODE addition.
2806 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
2808 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2809 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2810 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2811 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2812 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2813 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2814 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2815 and -mwbnoinvd.
2816 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2817 __builtin_ia32_wbinvd): New builtins.
2818 (SPECIAL_ARGS2): New.
2819 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2820 (SPECIAL_ARGS2): New.
2821 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2822 (ix86_valid_target_attribute_inner_p): Ditto.
2823 (ix86_init_mmx_sse_builtins): Add special_args2.
2824 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2825 TARGET_WBNOINVD_P): New.
2826 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2827 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2828 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2829 * config/i386/immintrin.h (_wbinvd): New intrinsic.
2830 * config/i386/pconfigintrin.h: New file.
2831 * config/i386/wbnoinvdintrin.h: Ditto.
2832 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2833 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2835 2018-03-05 Richard Biener <rguenther@suse.de>
2837 PR tree-optimization/84670
2838 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2839 member.
2840 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2841 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2842 (compute_antic_aux): Only assert the number of values in ANTIC_IN
2843 doesn't grow if all successors (recursively) were visited at least
2844 once.
2846 2018-03-05 Richard Biener <rguenther@suse.de>
2848 PR tree-optimization/84650
2849 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2850 if executed in the loop pipeline.
2852 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
2854 * doc/configfiles.texi (Configuration Files): Move info about
2855 conditionalizing $target-protos.h to...
2856 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
2857 differs from $target-protos.h.
2859 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
2860 Chung-Ju Wu <jasonwucj@gmail.com>
2862 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2863 * config/nds32/nds32-multiple.md (setmemsi): Define.
2864 * config/nds32/nds32-memory-manipulation.c
2865 (nds32_gen_dup_4_byte_to_word_value): New.
2866 (emit_setmem_word_loop): New.
2867 (emit_setmem_byte_loop): New.
2868 (nds32_expand_setmem_loop): New.
2869 (nds32_expand_setmem_loop_v3m): New.
2870 (nds32_expand_setmem_unroll): New.
2871 (nds32_expand_setmem): New.
2873 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2874 Chung-Ju Wu <jasonwucj@gmail.com>
2876 * config/nds32/nds32-memory-manipulation.c
2877 (nds32_emit_load_store): New.
2878 (nds32_emit_post_inc_load_store): New.
2879 (nds32_emit_mem_move): New.
2880 (nds32_emit_mem_move_block): New.
2881 (nds32_expand_movmemsi_loop_unknown_size): New.
2882 (nds32_expand_movmemsi_loop_known_size): New.
2883 (nds32_expand_movmemsi_loop): New.
2884 (nds32_expand_movmemsi_unroll): New.
2885 (nds32_expand_movmemqi): Rename ...
2886 (nds32_expand_movmemsi): ... to this.
2887 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2888 (movmemsi): ... to this.
2889 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2890 (nds32_expand_movmemsi): ... to this.
2892 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2893 Monk Chiang <sh.chiang04@gmail.com>
2894 Chung-Ju Wu <jasonwucj@gmail.com>
2896 * config/nds32/nds32-protos.h
2897 (nds32_expand_load_multiple): New arguments.
2898 (nds32_expand_store_multiple): Ditto.
2899 (nds32_valid_multiple_load_store): Rename ...
2900 (nds32_valid_multiple_load_store_p): ... to this.
2901 * config/nds32/nds32-memory-manipulation.c
2902 (nds32_expand_load_multiple): Refine implementation.
2903 (nds32_expand_store_multiple): Ditto.
2904 * config/nds32/nds32-multiple.md
2905 (load_multiple): Update nds32_expand_load_multiple interface.
2906 (store_multiple): Update nds32_expand_store_multiple interface.
2907 * config/nds32/nds32-predicates.c
2908 (nds32_valid_multiple_load_store): Rename ...
2909 (nds32_valid_multiple_load_store_p): ... to this and refine
2910 implementation.
2911 * config/nds32/predicates.md
2912 (nds32_load_multiple_and_update_address_operation): New predicate.
2913 (nds32_store_multiple_and_update_address_operation): New predicate.
2915 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2916 Chung-Ju Wu <jasonwucj@gmail.com>
2918 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2919 (combo): New attribute.
2920 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2922 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2924 * config/nds32/nds32.opt: Change -mcmodel= default value.
2926 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2927 Monk Chiang <sh.chiang04@gmail.com>
2928 Chung-Ju Wu <jasonwucj@gmail.com>
2930 * config/nds32/constants.md (unspec_element): New enum.
2931 * config/nds32/constraints.md (Umw): New constraint.
2932 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2933 * config/nds32/nds32-intrinsic.md: Likewise.
2934 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
2935 (nds32_valid_smw_lwm_base_p): New.
2936 (nds32_output_smw_single_word): New.
2937 (nds32_output_lmw_single_word): New.
2938 (nds32_expand_unaligned_load): New.
2939 (nds32_expand_unaligned_store): New.
2940 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
2941 (nds32_output_smw_single_word): Declare.
2942 (nds32_output_lmw_single_word): Declare.
2943 (nds32_expand_unaligned_load): Declare.
2944 (nds32_expand_unaligned_store): Declare.
2945 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
2946 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
2947 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
2948 NDS32_BUILTIN_UASTORE_DW.
2949 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
2950 predicate.
2952 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
2953 Kito Cheng <kito.cheng@gmail.com>
2954 Chung-Ju Wu <jasonwucj@gmail.com>
2956 * config/nds32/nds32-intrinsic.c
2957 (nds32_expand_builtin_null_ftype_reg): Delete.
2958 (nds32_expand_builtin_reg_ftype_imm): Ditto.
2959 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
2960 (nds32_read_argument): New.
2961 (nds32_legitimize_target): Ditto.
2962 (nds32_legitimize_argument): Ditto.
2963 (nds32_check_constant_argument): Ditto.
2964 (nds32_expand_unop_builtin): Ditto.
2965 (nds32_expand_unopimm_builtin): Ditto.
2966 (nds32_expand_binop_builtin): Ditto.
2967 (nds32_builtin_decl_impl): Ditto.
2968 (builtin_description): Ditto.
2969 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
2970 (nds32_init_builtins_impl): Ditto.
2971 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
2972 (nds32_builtin_decl): New.
2973 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
2974 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
2976 2018-03-02 Jeff Law <law@redhat.com>
2978 * reorg.c (stop_search_p): Handle DEBUG_INSN.
2979 (redundant_insn, fill_simple_delay_slots): Likewise.
2980 (fill_slots_from_thread): Likewise.
2981 * resource.c (mark_referenced_resources): Likewise.
2982 (mark_set_resources, find_dead_or_set_registers): Likewise.
2984 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2986 * substring-locations.h (format_warning_va): Formatting fix for
2987 ATTRIBUTE_GCC_DIAG.
2988 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
2989 argument.
2990 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
2991 * substring-locations.c: Include intl.h.
2992 (format_warning_va): Turned into small wrapper around
2993 format_warning_n_va, renamed to ...
2994 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
2995 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
2996 use ngettext.
2997 (format_warning_at_substring_n): New function.
2998 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
2999 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3000 format_warning_at_substring with just a shorter name instead of
3001 const function pointer.
3002 (fmtwarn_n): New function.
3003 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3004 appropriate, get rid of all the fmtstr temporaries, move conditionals
3005 with G_() wrapped string literals directly into fmtwarn arguments,
3006 cast dir.len to (int), formatting fixes.
3008 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3010 * doc/invoke.texi: Remove "Cilk Plus" references.
3012 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3013 Richard Biener <rguenther@suse.de>
3015 PR ipa/84628
3016 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3017 for error or warning attributes if CALL_FROM_THUNK_P is set.
3018 Formatting fixes.
3020 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3022 PR target/56540
3023 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3024 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3026 PR target/56540
3027 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3028 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3030 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3031 instead of -1U in last predictors element's probability member.
3033 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3035 PR ipa/83983
3036 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3037 arguments if they are comparable.
3039 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3041 PR tree-optimization/84634
3042 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3043 masks and masked_loop_p with a single loop_masks, making sure it's
3044 null for bb vectorization.
3046 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3048 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3049 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3050 loop->force_vectorize to check whether there is no alias.
3052 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3054 PR target/84614
3055 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3056 prototypes.
3057 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3058 comments.
3059 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3060 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3061 instead of a loop around prev_real_insn.
3062 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3063 prev_real_insn.
3065 PR inline-asm/84625
3066 * config/i386/i386.c (ix86_print_operand): Use conditional
3067 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3068 zero vector.
3070 2018-03-02 Richard Biener <rguenther@suse.de>
3072 PR tree-optimization/84427
3073 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3074 (bitmap_set_subtract_values): Rewrite to handle multiple
3075 exprs per value.
3076 (clean): Likewise.
3077 (prune_clobbered_mems): Likewise.
3078 (phi_translate): Take edge instead of pred/phiblock.
3079 (phi_translate_1): Likewise.
3080 (phi_translate_set): Likewise. Insert all translated
3081 exprs for a value into the set, keeping possibly multiple
3082 expressions per value.
3083 (compute_antic_aux): Adjust for phi_translate changes.
3084 When intersecting union the expressions and prune those
3085 not in the final value set, keeping possibly multiple
3086 expressions per value. Do not use value-insertion
3087 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3088 all expressions. Add verification that the value-sets
3089 only shrink during iteration.
3090 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3091 (do_pre_regular_insertion): Likewise.
3092 (do_pre_partial_partial_insertion): Likewise.
3094 2018-03-02 Richard Biener <rguenther@suse.de>
3096 PR target/82005
3097 * config/darwin.c (saved_debug_info_level): New static global.
3098 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3099 (darwin_asm_lto_end): Restore debug info generation settings.
3101 2018-03-01 Martin Liska <mliska@suse.cz>
3103 PR sanitizer/82484
3104 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3105 volatile arguments.
3107 2018-03-01 Richard Biener <rguenther@suse.de>
3109 PR debug/84645
3110 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3111 type annotation with LTO when debug was disabled at compile-time.
3113 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3115 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3116 XINT with INTVAL.
3117 (mips_final_postscan_insn): Likewise.
3119 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3121 PR rtl-optimization/84528
3122 * alias.c (init_alias_target): Add commentary.
3123 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3124 a unique base value if the frame pointer is not eliminated
3125 to the stack pointer.
3127 2018-03-01 Tom de Vries <tom@codesourcery.com>
3129 PR rtl-optimization/83327
3130 * lra-int.h (hard_regs_spilled_into): Declare.
3131 * lra.c (hard_regs_spilled_into): Define.
3132 (init_reg_info): Init hard_regs_spilled_into.
3133 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3134 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3135 (process_bb_lives): Handle hard_regs_spilled_into.
3136 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3137 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3139 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3141 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3142 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3143 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3144 * config/rs6000/aix72.h: New file.
3146 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3148 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3149 instead of warning_at with conditional singular and plural messages
3150 where possible.
3152 PR target/52991
3153 * stor-layout.c (update_alignment_for_field): For
3154 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3155 && !DECL_PACKED (field), do the alignment update, just use
3156 only desired_align instead of MAX (type_align, desired_align)
3157 as the alignment.
3158 (place_field): Don't do known_align < desired_align handling
3159 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3160 is non-NULL, instead do it after rli->prev_field handling and
3161 only if not within a bitfield word. For DECL_PACKED (field)
3162 use type_align of BITS_PER_UNIT.
3164 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3166 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3167 superfluous parentheses and trailing spaces.
3169 2018-02-28 Richard Biener <rguenther@suse.de>
3171 PR tree-optimization/84584
3172 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3173 SCoPs with fake exit edge.
3175 2018-02-28 Martin Liska <mliska@suse.cz>
3177 PR testsuite/84597
3178 * timevar.c (timer::print): Fix format to properly print 100%
3179 values.
3181 2018-02-28 Richard Biener <rguenther@suse.de>
3183 PR middle-end/84607
3184 * genmatch.c (capture_info::walk_match): Do not mark
3185 captured expressions without operands as expr_p given
3186 they act more like predicates and should be subject to
3187 "lost tail" side-effect preserving.
3189 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3191 PR rtl-optimization/81611
3192 * auto-inc-dec.c (attempt_change): Move dead note from
3193 mem_insn if it's the next use of regno
3194 (find_address): Take address use of reg holding
3195 non-incremented value. Add parm to limit search to the named
3196 reg only.
3197 (merge_in_block): Attempt to use a mem insn that is the next
3198 use of the original regno.
3200 2018-02-27 Martin Sebor <msebor@redhat.com>
3202 PR c++/83871
3203 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3204 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3206 2018-02-27 Martin Sebor <msebor@redhat.com>
3208 PR translation/84207
3209 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3210 n argument to unsigned HOST_WIDE_INT.
3211 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3212 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3213 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3214 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3216 2018-02-27 Richard Biener <rguenther@suse.de>
3218 PR tree-optimization/84512
3219 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3220 Do not use the estimate returned from record_stmt_cost for
3221 the scalar iteration cost but sum properly using add_stmt_cost.
3223 2018-02-27 Richard Biener <rguenther@suse.de>
3225 PR tree-optimization/84466
3226 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3227 Adjust last change to less strictly validate use operands.
3229 2018-02-27 Martin Liska <mliska@suse.cz>
3231 PR gcov-profile/84548
3232 * gcov.c (process_file): Allow partial overlap and consider it
3233 also as group functions.
3234 (output_lines): Properly calculate range of lines for a group.
3236 2018-02-27 Martin Liska <mliska@suse.cz>
3238 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3239 'ggc' suffixes. Change first column width.
3240 (timer::print): Fix formatting of the column.
3242 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3244 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3245 preserve inline entry blocks for the sake of debug inline
3246 entry point markers alone.
3247 (remove_unused_locals): Suggest in comments a better place to
3248 force the preservation of inline entry blocks that are
3249 otherwise unused, but do not preserve them.
3251 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3253 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3255 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3257 PR target/84039
3258 * config/i386/constraints.md (Bs): Replace
3259 ix86_indirect_branch_register with
3260 TARGET_INDIRECT_BRANCH_REGISTER.
3261 (Bw): Likewise.
3262 * config/i386/i386.md (indirect_jump): Likewise.
3263 (tablejump): Likewise.
3264 (*sibcall_memory): Likewise.
3265 (*sibcall_value_memory): Likewise.
3266 Peepholes of indirect call and jump via memory: Likewise.
3267 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3268 (*sibcall_value_GOT_32): Likewise.
3269 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3270 (GOT_memory_operand): Likewise.
3271 (call_insn_operand): Likewise.
3272 (sibcall_insn_operand): Likewise.
3273 (GOT32_symbol_operand): Likewise.
3274 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3276 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3278 PR rtl-optimization/83496
3279 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3280 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3281 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3282 redundant insn, if any.
3283 (relax_delay_slots): Likewise.
3284 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3286 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3288 PR tree-optimization/83965
3289 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3290 that grouped statements are part of a reduction chain. Return
3291 true if the statement is not marked as a reduction itself but
3292 is part of a group.
3293 (vect_recog_dot_prod_pattern): Don't check whether the statement
3294 is part of a group here.
3295 (vect_recog_sad_pattern): Likewise.
3296 (vect_recog_widen_sum_pattern): Likewise.
3298 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3300 PR debug/84545
3301 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3303 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3305 PR target/84530
3306 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3307 the bool argument.
3308 (ix86_output_indirect_function_return): New prototype.
3309 (ix86_split_simple_return_pop_internal): Likewise.
3310 * config/i386/i386.c (indirect_return_via_cx): New.
3311 (indirect_return_via_cx_bnd): Likewise.
3312 (indirect_thunk_name): Handle return va CX_REG.
3313 (output_indirect_thunk_function): Create alias for
3314 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3315 (ix86_output_indirect_jmp): Remove the bool argument.
3316 (ix86_output_indirect_function_return): New function.
3317 (ix86_split_simple_return_pop_internal): Likewise.
3318 * config/i386/i386.md (*indirect_jump): Don't pass false
3319 to ix86_output_indirect_jmp.
3320 (*tablejump_1): Likewise.
3321 (simple_return_pop_internal): Change it to define_insn_and_split.
3322 Call ix86_split_simple_return_pop_internal to split it for
3323 -mfunction-return=.
3324 (simple_return_indirect_internal): Call
3325 ix86_output_indirect_function_return instead of
3326 ix86_output_indirect_jmp.
3328 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3330 PR bootstrap/84405
3331 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3332 memset and value initialization afterwards.
3334 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3336 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3338 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3340 PR target/84521
3341 * common/config/aarch64/aarch64-common.c
3342 (aarch_option_optimization_table[]): Switch
3343 off fomit-frame-pointer
3345 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3346 Chung-Ju Wu <jasonwucj@gmail.com>
3348 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3349 volatile memory.
3350 (store_multiple): Ditto.
3352 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3354 * config.gcc: Add --with-cpu support for nds32 target.
3355 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3356 * config/nds32/nds32.opt: Add -mcpu= option.
3358 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3360 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3361 isel=yes): Warn for these deprecated options.
3363 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3365 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3366 ISA_2_5_MASKS_EMBEDDED.
3368 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3370 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3371 p->max as pointers rather than using iterative_hash_expr.
3373 2018-02-23 Carl Love <cel@us.ibm.com>
3375 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3376 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3377 BU_P8V_OVERLOAD_2.
3378 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3379 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3380 P8V_BUILTIN_VEC_VUNSIGNED2.
3382 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3384 PR target/81572
3385 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3386 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3387 LRA_UNKNOWN_ALT.
3388 * lra-constraints.c (curr_insn_transform): Set up
3389 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3390 LRA_UNKNOWN_ALT.
3391 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3392 * lra-eliminations.c (spill_pseudos): Ditto.
3393 (process_insn_for_elimination): Ditto.
3394 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3395 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3396 LRA_NON_CLOBBERED_ALT.
3398 2018-02-22 Martin Sebor <msebor@redhat.com>
3400 PR tree-optimization/84480
3401 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3402 to maybe_diag_stxncpy_trunc. Call it.
3403 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3404 from gimple_fold_builtin_strcpy. Print inlining stack.
3405 (handle_builtin_stxncpy): Print inlining stack.
3406 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3408 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3410 PR target/84176
3411 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3412 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3413 and -fcheck-pointer-bounds are used together.
3414 (indirect_thunk_prefix): New enum.
3415 (indirect_thunk_need_prefix): New function.
3416 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3417 "_nt" instead of "_bnd" for NOTRACK prefix.
3418 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3419 (output_indirect_thunk_function): Likewise.
3420 (): Likewise.
3421 (ix86_code_end): Update output_indirect_thunk_function calls.
3422 (ix86_output_indirect_branch_via_reg): Replace
3423 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3424 (ix86_output_indirect_branch_via_push): Likewise.
3425 (ix86_output_function_return): Likewise.
3426 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3427 incompatible with -fcf-protection=branch and
3428 -fcheck-pointer-bounds.
3430 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3432 PR target/83335
3433 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3434 Change gcc_assert call to output_operand_lossage.
3436 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3438 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3440 2018-02-22 DJ Delorie <dj@redhat.com>
3441 Sebastian Perta <sebastian.perta@renesas.com>
3442 Oleg Endo <olegendo@gcc.gnu.org>
3444 * config/rx/rx.c (rx_rtx_costs): New function.
3445 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3447 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3449 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3451 2018-02-22 Martin Liska <mliska@suse.cz>
3453 PR driver/83193
3454 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3455 Add "native" as a possible value.
3457 2018-02-22 Martin Liska <mliska@suse.cz>
3459 PR driver/83193
3460 * config/i386/i386.c (ix86_option_override_internal):
3461 Add "native" as a possible value for -march and -mtune.
3463 2018-02-22 Jakub Jelinek <jakub@redhat.com>
3465 PR target/84502
3466 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
3467 to all type variants.
3469 PR tree-optimization/84503
3470 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
3471 width as info->bitpos + info->bitsize - start.
3472 (merged_store_group::merge_overlapping): Simplify width computation.
3473 (check_no_overlap): New function.
3474 (imm_store_chain_info::try_coalesce_bswap): Compute expected
3475 start + width and last_order of the group, fail if check_no_overlap
3476 fails.
3477 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
3478 to group if check_no_overlap fails.
3480 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3482 * config/rs6000/altivec.md: Delete contraint arguments to
3483 define_expand, define_split, and define_peephole2, and in
3484 define_insn_and_split if always unused.
3485 * config/rs6000/darwin.md: Ditto.
3486 * config/rs6000/dfp.md: Ditto.
3487 * config/rs6000/rs6000.md: Ditto.
3488 * config/rs6000/sync.md: Ditto.
3489 * config/rs6000/vector.md: Ditto.
3490 * config/rs6000/vsx.md: Ditto.
3492 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3494 * config/rs6000/altivec.md: Write output control strings as braced
3495 blocks instead of double-quoted strings.
3496 * config/rs6000/darwin.md: Ditto.
3497 * config/rs6000/rs6000.md: Ditto.
3498 * config/rs6000/vector.md: Ditto.
3499 * config/rs6000/vsx.md: Ditto.
3501 2018-02-21 Jason Merrill <jason@redhat.com>
3503 PR c++/84314 - ICE with templates and fastcall attribute.
3504 * attribs.c (build_type_attribute_qual_variant): Remove assert.
3506 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3508 * ipa-cp.c (determine_versionability): Fix comment typos.
3510 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3512 PR c/84229
3513 * ipa-cp.c (determine_versionability): Do not version functions caling
3514 va_arg_pack.
3516 2018-02-21 Martin Liska <mliska@suse.cz>
3518 PR driver/83193
3519 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3520 Add "native" as a possible value.
3521 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3522 the macro when native cpu detection is available.
3524 2018-02-21 Martin Liska <mliska@suse.cz>
3526 PR driver/83193
3527 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3528 Add "native" as a possible value.
3529 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3530 when native cpu detection is available.
3532 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3533 Martin Sebor <msebor@redhat.com>
3535 PR tree-optimization/84478
3536 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3537 false.
3538 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3539 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3540 support which is conservatively correct, for 2 only stay conservative
3541 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3542 argument to the 2 argument get_range_strlen, adjust 6 arg
3543 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3544 false.
3545 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3546 (gimple_fold_builtin_strlen): Pass true as last argument to
3547 get_range_strlen.
3549 2018-02-20 Martin Sebor <msebor@redhat.com>
3551 PR middle-end/84095
3552 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3553 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3554 (builtin_memref::builtin_memref): Factor out parts into
3555 set_base_and_offset and call it.
3557 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3559 PR middle-end/84406
3560 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3561 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3562 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3563 search at the associated MODE_INT.
3565 2018-02-20 Jeff Law <law@redhat.com>
3567 PR middle-end/82123
3568 PR tree-optimization/81592
3569 PR middle-end/79257
3570 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3571 for range data rather than using global data.
3572 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3573 range data rather than using global data.
3574 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3575 pass it to children as needed.
3576 (struct directive::fmtresult): Similarly.
3577 (struct directive::set_width): Similarly.
3578 (struct directive::set_precision): Similarly.
3579 (format_integer, format_directive, parse_directive): Similarly.
3580 (format_none): Accept unnamed vr_values parameter.
3581 (format_percent, format_floating, format_character): Similarly.
3582 (format_string, format_plain): Similarly.
3583 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3584 the EVRP range analyzer for range data rather than using global data.
3585 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3586 gimple-ssa-evrp-analyze.h
3587 (class sprintf_dom_walker): Add after_dom_children member function.
3588 Add evrp_range_analyzer member.
3589 (sprintf_dom_walker::before_dom_children): Call into the EVRP
3590 range analyzer as needed.
3591 (sprintf_dom_walker::after_dom_children): New member function.
3592 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3593 if not optimizing.
3594 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3595 (evrp_range_analyzer::pop_to_marker): Likewise.
3597 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3599 PR tree-optimization/84419
3600 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3601 with the required type if its current type is compatible but
3602 different.
3604 2018-02-20 Jakub Jelinek <jakub@redhat.com>
3606 PR middle-end/82004
3607 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3608 after vectorization.
3610 2018-02-20 Martin Liska <mliska@suse.cz>
3612 PR driver/83193
3613 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3614 possible values if we don't have a hint.
3616 2018-02-20 Martin Liska <mliska@suse.cz>
3618 PR c/84310
3619 PR target/79747
3620 * final.c (shorten_branches): Build align_tab array with one
3621 more element.
3622 * opts.c (finish_options): Add alignment option limit check.
3623 (MAX_CODE_ALIGN): Likewise.
3624 (MAX_CODE_ALIGN_VALUE): Likewise.
3625 * doc/invoke.texi: Document maximum allowed option value for
3626 all -falign-* options.
3628 2018-02-19 Jakub Jelinek <jakub@redhat.com>
3630 PR target/84146
3631 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3632 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3633 * var-tracking.c (emit_note_insn_var_location): Remove all references
3634 to NOTE_INSN_CALL_ARG_LOCATION.
3635 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3636 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3637 Use copy_rtx_if_shared.
3638 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3639 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3640 (dwarf2out_var_location): Remove handling of
3641 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3642 on call_insn.
3643 * final.c (final_scan_insn): Remove all references to
3644 NOTE_INSN_CALL_ARG_LOCATION.
3645 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
3646 before dumping final insns.
3647 * except.c (emit_note_eh_region_end): Remove all references to
3648 NOTE_INSN_CALL_ARG_LOCATION.
3649 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3650 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3651 * config/arc/arc.c (hwloop_optimize): Likewise.
3652 * config/arm/arm.c (create_fix_barrier): Likewise.
3653 * config/s390/s390.c (s390_chunkify_start): Likewise.
3654 * config/sh/sh.c (find_barrier): Likewise.
3655 * config/i386/i386.c (rest_of_insert_endbranch,
3656 ix86_seh_fixup_eh_fallthru): Likewise.
3657 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3658 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3659 * config/frv/frv.c (frv_function_prologue): Likewise.
3660 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3661 reg note.
3662 (note_outside_basic_block_p): Remove all references to
3663 NOTE_INSN_CALL_ARG_LOCATION.
3664 * gengtype.c (adjust_field_rtx_def): Likewise.
3665 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3666 Likewise.
3667 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3668 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3670 PR c++/84444
3671 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3672 is ADDR_EXPR.
3674 PR tree-optimization/84452
3675 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3676 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3677 is NULL.
3679 2018-02-19 Martin Liska <mliska@suse.cz>
3681 PR sanitizer/82183
3682 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3684 2018-02-19 Martin Liska <mliska@suse.cz>
3685 Richard Sandiford <richard.sandiford@linaro.org>
3687 PR tree-optimization/82491
3688 * gimple-fold.c (get_base_constructor): Make earlier bail out
3689 to prevent ubsan.
3691 2018-02-19 Carl Love <cel@us.ibm.com>
3693 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3694 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3695 BU_P8V_OVERLOAD_1.
3696 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3697 P8V_BUILTIN_VEC_NEG.
3699 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
3701 * config/rl78/rl78.md (movdf): New define expand.
3703 2018-02-19 Martin Liska <mliska@suse.cz>
3705 PR other/80589
3706 * doc/invoke.texi: Fix typo.
3707 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3709 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
3711 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3712 handle rs6000_single_float and rs6000_double_float specially for
3713 e500 family CPUs.
3715 2018-02-16 Jeff Law <law@redhat.com>
3717 * config/rx/rx.c (add_pop_cfi_notes): New function.;
3718 (pop_regs): Use it.
3720 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3722 PR ipa/84425
3723 * ipa-inline.c (inline_small_functions): Fix a typo.
3725 2018-02-16 Nathan Sidwell <nathan@acm.org>
3727 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3729 2018-02-16 Carl Love <cel@us.ibm.com>
3731 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3732 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3733 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3734 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3735 expansion to P8V_BUILTIN_VEC_FLOAT2.
3737 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
3739 PR rtl-optimization/70023
3740 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3741 src_regno into account.
3743 2018-02-16 Carl Love <cel@us.ibm.com>
3745 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3746 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3747 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3748 * config/rs6000/rs6000.c: Remove case statements for
3749 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3750 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3751 and P9V_BUILTIN_VEC_VINSERT4B.
3752 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
3753 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
3754 * config/rs6000/vsx.md:
3755 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
3756 vec_insert4b.
3758 2018-02-16 Carl Love <cel@us.ibm.com>
3760 * config/rs6000/altivec.h: Add builtin names vec_extract4b
3761 vec_insert4b.
3762 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
3763 definitions.
3764 * config/rs6000/rs6000-c.c: Add the definitions for
3765 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
3766 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
3767 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
3768 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
3769 definition for insert4b and define insn *insert3b_internal.
3770 * doc/extend.texi: Add documentation for vec_extract4b.
3772 2018-02-16 Nathan Sidwell <nathan@acm.org>
3774 * doc/extend.texi (Backwards Compatibility): Mention friend
3775 injection. Note for-scope is deprecated.
3776 * doc/invoke.texi (-ffriend-injection): Deprecate.
3778 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
3780 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
3781 that moved to I2, also allow destinations that are a paradoxical
3782 subreg (instead of a normal reg).
3784 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
3786 PR target/83831
3787 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
3788 to QImode.
3790 2018-02-16 Richard Biener <rguenther@suse.de>
3792 PR tree-optimization/84037
3793 PR tree-optimization/84016
3794 PR target/82862
3795 * config/i386/i386.c (ix86_builtin_vectorization_cost):
3796 Adjust vec_construct for the fact we need additional higher latency
3797 128bit inserts for AVX256 and AVX512 vector builds.
3798 (ix86_add_stmt_cost): Scale vector construction cost for
3799 elementwise loads.
3801 2018-02-16 Richard Biener <rguenther@suse.de>
3803 PR tree-optimization/84417
3804 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3805 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3806 (non_rewritable_lvalue_p): Likewise, use poly-ints.
3808 2018-02-16 Martin Liska <mliska@suse.cz>
3810 PR sanitizer/84307
3811 * internal-fn.def (ASAN_CHECK): Set proper flags.
3812 (ASAN_MARK): Likewise.
3814 2018-02-16 Julia Koval <julia.koval@intel.com>
3816 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3817 from PTA_CANNONLAKE.
3819 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3821 PR target/84272
3822 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3823 Use ++iter rather than iter++ for std::list iterators.
3824 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
3825 defer deleting them until all nodes in the forest are processed. Do
3826 free even leaf nodes. Change to_process into auto_vec.
3828 PR bootstrap/84405
3829 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3830 * vec.h (vec_default_construct): Use memset instead of placement new
3831 if BROKEN_VALUE_INITIALIZATION is defined.
3832 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3833 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3834 is defined.
3836 PR rtl-optimization/83723
3837 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3838 * lra.c (lra_substitute_pseudo): Likewise. If true, use
3839 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
3840 recursive calls.
3841 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3842 callers.
3843 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3845 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
3847 PR rtl-optimization/81443
3848 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3849 from inner REGs to paradoxical SUBREGs.
3851 2018-02-16 Richard Biener <rguenther@suse.de>
3853 PR tree-optimization/84399
3854 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3855 For operands we can analyze at their definition make sure we can
3856 analyze them at each use as well.
3858 2018-02-16 Richard Biener <rguenther@suse.de>
3860 PR tree-optimization/84190
3861 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3862 volatile accesses if the decl isn't volatile.
3864 2018-02-15 Jason Merrill <jason@redhat.com>
3866 PR c++/84314 - ICE with templates and fastcall attribute.
3867 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3868 TYPE_CANONICAL on an existing type.
3870 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3872 PR tree-optimization/84383
3873 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3874 dstoff nor call operand_equal_p if dstbase is NULL.
3876 PR tree-optimization/84334
3877 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3878 also a CONSTANT_CLASS_P, punt.
3880 2018-02-14 Jim Wilson <jimw@sifive.com>
3882 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3883 first SMALL_OPERAND check. New local min_second_step. Move assert
3884 to where locals are set. Add TARGET_RVC support.
3885 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3887 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3889 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3891 2018-02-14 Martin Sebor <msebor@redhat.com>
3893 PR tree-optimization/83698
3894 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3895 arrays constrain the offset range to their bounds.
3896 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3897 (builtin_access::overlap): Avoid setting the size of overlap if it's
3898 already been set.
3899 (maybe_diag_overlap): Also consider arrays when deciding what values
3900 of offsets to include in diagnostics.
3902 2018-02-14 Martin Sebor <msebor@redhat.com>
3904 PR c/84108
3905 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3906 that correspond to the kind of a declaration.
3908 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3910 PR target/83984
3911 * config/pa/pa.md: Load address of PIC label using the linkage table
3912 if the label is nonlocal.
3914 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3916 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3917 warning message if user requests -maltivec=be.
3918 * doc/invoke.texi: Document deprecation of -maltivec=be.
3920 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3922 PR target/84220
3923 * config/rs6000/rs6000-c.c: Update definitions for
3924 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3925 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3927 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3929 PR target/84239
3930 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3931 add _get_ssp intrinsics. Remove argument from
3932 __builtin_ia32_rdssp[d|q].
3933 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3934 * config/i386/i386-builtin.def: Remove argument from
3935 __builtin_ia32_rdssp[d|q].
3936 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
3937 ix86_expand_special_args_builtin for _rdssp[d|q].
3938 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
3939 Clear register before usage.
3940 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
3941 Add documentation for new _get_ssp and _inc_ssp intrinsics.
3943 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
3945 PR tree-optimization/84357
3946 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
3947 operand 1 of an ARRAY_REF too.
3949 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
3951 PR target/83831
3952 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
3953 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3954 declarations.
3955 (set_of_reg): New struct.
3956 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
3957 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
3958 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3959 functions.
3960 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
3961 Split into bitclr, bitset, bitinvert patterns if appropriate.
3962 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
3963 use rx_fuse_in_memory_bitop.
3964 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
3965 to named insn, correct maximum insn length.
3967 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
3969 PR target/79242
3970 * machmode.def: Define a complex mode for PARTIAL_INT.
3971 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
3972 MODE_PARTIAL_INT.
3973 * doc/rtl.texi: Document CSPImode.
3974 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
3975 handling.
3976 (msp430_hard_regno_nregs_with_padding): Likewise.
3978 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
3980 PR target/84279
3981 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
3983 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
3985 PR rtl-optimization/84169
3986 * combine.c (try_combine): New variable split_i2i3. Set it to true if
3987 we generated a parallel as new i3 and we split that to new i2 and i3
3988 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
3989 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
3990 those to i2, not i1. Partially rewrite this scan code.
3992 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3994 PR c/82210
3995 * stor-layout.c (place_field): For variable length fields, adjust
3996 offset_align afterwards not just based on the field's alignment,
3997 but also on the size.
3999 PR middle-end/84309
4000 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4001 of exps and logs in the use_exp2 case.
4003 2018-02-13 Jeff Law <law@redhat.com>
4005 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4006 entry for "vector".
4008 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4009 ARGS as unused.
4011 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4013 PR debug/84342
4014 PR debug/84319
4015 * common.opt (gas-loc-support, gas-locview-support): New.
4016 (ginline-points, ginternal-reset-location-views): New.
4017 * doc/invoke.texi: Document them. Use @itemx where intended.
4018 (gvariable-location-views): Adjust.
4019 * target.def (reset_location_view): New.
4020 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4021 (TARGET_RESET_LOCATION_VIEW): New.
4022 * doc/tm.texi: Rebuilt.
4023 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4024 (dwarf2out_default_as_locview_support): New.
4025 (output_asm_line_debug_info): Use option variables.
4026 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4027 (output_loc_list): Likewise.
4028 (add_high_low_attributes): Check option variables.
4029 Don't output entry view attribute in strict mode.
4030 (gen_inlined_subroutine_die): Check option variables.
4031 (dwarf2out_inline_entry): Likewise.
4032 (init_sections_and_labels): Likewise.
4033 (dwarf2out_early_finish): Likewise.
4034 (maybe_reset_location_view): New, from...
4035 (dwarf2out_var_location): ... here. Call it.
4036 * debug.h (dwarf2out_default_as_loc_support): Declare.
4037 (dwarf2out_default_as_locview_support): Declare.
4038 * hooks.c (hook_int_rtx_insn_0): New.
4039 * hooks.h (hook_int_rtx_insn_0): Declare.
4040 * toplev.c (process_options): Take -gas-loc-support and
4041 -gas-locview-support from dwarf2out. Enable
4042 -gvariable-location-views by default only with locview
4043 assembler support. Enable -ginternal-reset-location-views by
4044 default only if the target defines the corresponding hook.
4045 Enable -ginline-points by default if location views are
4046 enabled; force it disabled if statement frontiers are
4047 disabled.
4048 * tree-inline.c (expand_call_inline): Check option variables.
4049 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4051 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4053 PR tree-optimization/84321
4054 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4055 handling. Also check whether the anti-range contains any values
4056 that satisfy the mask; switch to a VR_RANGE if not.
4058 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4060 PR sanitizer/84340
4061 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4063 2018-02-13 Martin Jambor <mjambor@suse.cz>
4065 PR c++/83990
4066 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4067 of call statements, also set location of a load to a temporary.
4069 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4071 * config/rl78/rl78.c (add_vector_labels): New function.
4072 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4073 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4074 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4075 which checks that no arguments are passed.
4076 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4077 * doc/extend.texi: Documentation for the new attribute.
4079 2018-02-13 Andreas Schwab <schwab@suse.de>
4081 * config/riscv/linux.h (CPP_SPEC): Define.
4083 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4085 PR target/84335
4086 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4087 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4088 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4089 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4090 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4091 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4092 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4093 temporarily for AES and PCLMUL builtins.
4095 PR tree-optimization/84339
4096 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4097 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4098 Formatting fixes.
4100 PR middle-end/84309
4101 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4102 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4103 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4104 inline function.
4105 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4106 inline function.
4107 * omp-simd-clone.h: New file.
4108 * omp-simd-clone.c: Include omp-simd-clone.h.
4109 (expand_simd_clones): No longer static.
4110 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4111 cgraph.h and omp-simd-clone.h.
4112 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4113 (vect_recog_widen_shift_pattern): Formatting fix.
4114 (vect_pattern_recog_1): Don't check optab for calls.
4116 PR target/84336
4117 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4118 operands[2] into a REG before using gen_lowpart on it.
4120 2018-02-12 Jeff Law <law@redhat.com>
4122 PR target/83760
4123 * config/sh/sh.c (find_barrier): Consider a sibling call
4124 a barrier as well.
4126 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4127 successfully back substituting a reg.
4129 2018-02-12 Richard Biener <rguenther@suse.de>
4131 PR tree-optimization/84037
4132 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4133 parameter, move visited init to caller.
4134 (vect_slp_analyze_operations): Separate cost from validity
4135 check, initialize visited once for all instances.
4136 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4137 for all instances.
4138 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4139 out an assert.
4140 (vect_model_promotion_demotion_cost): Likewise.
4141 (vectorizable_bswap): Guard cost modeling with !slp_node
4142 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4143 SLP stmts.
4144 (vectorizable_call): Likewise.
4145 (vectorizable_conversion): Likewise.
4146 (vectorizable_assignment): Likewise.
4147 (vectorizable_shift): Likewise.
4148 (vectorizable_operation): Likewise.
4149 (vectorizable_store): Likewise.
4150 (vectorizable_load): Likewise.
4151 (vectorizable_condition): Likewise.
4152 (vectorizable_comparison): Likewise.
4154 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4156 PR sanitizer/84307
4157 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4158 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4159 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4161 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4163 PR middle-end/83665
4164 * params.def (inline-min-speedup): Increase from 8 to 15.
4165 (max-inline-insns-auto): Decrease from 40 to 30.
4166 * ipa-split.c (consider_split): Add some buffer for function to
4167 be considered inlining candidate.
4168 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4169 default values.
4171 2018-02-12 Richard Biener <rguenther@suse.de>
4173 PR tree-optimization/84037
4174 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4175 matched stmts if we cannot swap the non-matched ones.
4177 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4179 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4180 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4181 _mm_maskz_scalef_round_ss): New intrinsics.
4182 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4183 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4184 __builtin_ia32_scalefss_round): Remove.
4185 (__builtin_ia32_scalefsd_mask_round,
4186 __builtin_ia32_scalefss_mask_round): New intrinsics.
4187 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4188 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4189 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4190 "<round_constraint>")): Changed to ...
4191 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4192 "<round_scalar_constraint>")): ... this.
4193 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4194 %0, %1, %2<round_op3>}"): Changed to ...
4195 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4196 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4197 %2<round_scalar_mask_op3>}"): ... this.
4198 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4200 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4202 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4203 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4204 (_mm_maskz_sqrt_round_ss): New intrinsics.
4205 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4206 (__builtin_ia32_sqrtsd_mask_round)
4207 (__builtin_ia32_sqrtss_mask_round): New builtins.
4208 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4209 (__builtin_ia32_sqrtss_round): Remove.
4210 (__builtin_ia32_sqrtsd_mask_round)
4211 (__builtin_ia32_sqrtss_mask_round): New builtins.
4212 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4213 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4214 ((match_operand:VF_128 1 "vector_operand"
4215 "xBm,<round_constraint>")): Changed to ...
4216 ((match_operand:VF_128 1 "vector_operand"
4217 "xBm,<round_scalar_constraint>")): ... this.
4218 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4219 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4220 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4221 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4222 %<iptr>1<round_scalar_mask_op3>}): ... this.
4223 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4224 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4226 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4228 PR target/84266
4229 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4230 Cast vec_cmpeq result to correct type.
4231 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4232 Cast vec_cmpgt result to correct type.
4234 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4236 * final.c (final_scan_insn_1): Renamed from...
4237 (final_scan_insn): ... this. New wrapper, to recover
4238 seen from the outermost call in recursive ones.
4239 * config/sparc/sparc.c (output_return): Drop seen from call.
4240 (output_sibcall): Likewise.
4241 * config/visium/visium.c (output_branch): Likewise.
4243 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4245 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4246 function label.
4248 2018-02-10 Alan Modra <amodra@gmail.com>
4250 PR target/84300
4251 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4252 Specify LR as an input.
4254 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4256 PR sanitizer/83987
4257 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4258 remove_member_access_dummy_vars): New functions.
4259 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4260 lower_omp_1, execute_lower_omp): Use them.
4262 PR rtl-optimization/84308
4263 * shrink-wrap.c (spread_components): Release todo vector.
4265 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4267 PR rtl-optimization/57193
4268 * ira-color.c (struct allocno_color_data): Add member
4269 conflict_allocno_hard_prefs.
4270 (update_conflict_allocno_hard_prefs): New.
4271 (bucket_allocno_compare_func): Add a preference based on
4272 conflict_allocno_hard_prefs.
4273 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4274 (color_allocnos): Remove a dead code. Initiate
4275 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4277 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4279 PR target/84226
4280 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4281 constraint from =wa to wa. Avoid a subreg on the output operand,
4282 instead use a pseudo and subreg it in a move.
4283 (p9_xxbrd_<mode>): Changed to ...
4284 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4285 (p9_xxbrd_v2df): New expander.
4286 (p9_xxbrw_<mode>): Changed to ...
4287 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4288 (p9_xxbrw_v4sf): New expander.
4290 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4292 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4294 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4296 PR target/83926
4297 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4298 multiply in 32-bit mode.
4299 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4300 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4301 mode.
4303 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4305 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4306 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4307 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4308 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4310 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4312 PR lto/84213
4313 * dwarf2out.c (is_trivial_indirect_ref): New function.
4314 (dwarf2out_late_global_decl): Do not generate a location
4315 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4316 and that are not defined in the current unit.
4318 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4320 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4321 instead of a libcall for UNORDERED.
4323 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4325 PR target/82641
4326 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4327 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4329 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4331 PR target/PR84295
4332 * config/s390/s390.c (s390_set_current_function): Invoke
4333 s390_indirect_branch_settings also if fndecl didn't change.
4335 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4337 * config/rs6000/rs6000.md (blockage): Set length to zero.
4339 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4341 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4343 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4345 PR sanitizer/84285
4346 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4347 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4348 -static-lib*san.
4350 PR debug/84252
4351 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4352 PARALLEL incoming that failed vt_get_decl_and_offset check.
4354 PR middle-end/84237
4355 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4356 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4357 TREE_READONLY bit.
4358 (get_variable_section): For decls in named .bss* sections pass true as
4359 second argument to bss_initializer_p.
4361 2018-02-09 Marek Polacek <polacek@redhat.com>
4362 Jakub Jelinek <jakub@redhat.com>
4364 PR c++/83659
4365 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4366 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4367 Sync some changes from cxx_fold_indirect_ref.
4369 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4371 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4372 markers.
4373 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4374 (BLOCK_INLINE_ENTRY_LABEL): New.
4375 (dwarf2out_var_location): Disregard inline entry markers.
4376 (inline_entry_data): New struct.
4377 (inline_entry_data_hasher): New hashtable type.
4378 (inline_entry_data_hasher::hash): New.
4379 (inline_entry_data_hasher::equal): New.
4380 (inline_entry_data_table): New variable.
4381 (add_high_low_attributes): Add DW_AT_entry_pc and
4382 DW_AT_GNU_entry_view attributes if a pending entry is found
4383 in inline_entry_data_table. Add old entry_pc attribute only
4384 if debug nonbinding markers are disabled.
4385 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4386 markers are enabled.
4387 (block_within_block_p, dwarf2out_inline_entry): New.
4388 (dwarf2out_finish): Check that no entries remained in
4389 inline_entry_data_table.
4390 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4391 (final_scan_insn, notice_source_line): Likewise.
4392 (rest_of_clean_state): Skip inline entry markers.
4393 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4394 markers.
4395 * gimple.c (gimple_build_debug_inline_entry): New.
4396 * gimple.h (enum gimple_debug_subcode): Add
4397 GIMPLE_DEBUG_INLINE_ENTRY.
4398 (gimple_build_debug_inline_entry): Declare.
4399 (gimple_debug_inline_entry_p): New.
4400 (gimple_debug_nonbind_marker_p): Adjust.
4401 * insn-notes.def (INLINE_ENTRY): New.
4402 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4403 inline entry marker notes.
4404 (print_insn): Likewise.
4405 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4406 (INSN_DEBUG_MARKER_KIND): Likewise.
4407 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4408 * tree-inline.c (expand_call_inline): Build and insert
4409 debug_inline_entry stmt.
4410 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4411 inline entry blocks early, if nonbind markers are enabled.
4412 (dump_scope_block): Dump fragment info.
4413 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4414 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4415 (gimple_build_debug_inline_entry): New.
4416 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4417 Enable/disable inline entry points too.
4418 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4419 (DEBUG_INSN): Describe inline entry markers.
4421 * common.opt (gvariable-location-views): New.
4422 (gvariable-location-views=incompat5): New.
4423 * config.in: Rebuilt.
4424 * configure: Rebuilt.
4425 * configure.ac: Test assembler for view support.
4426 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4427 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4428 * dwarf2out.c (var_loc_view): New typedef.
4429 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4430 (dwarf2out_locviews_in_attribute): New.
4431 (dwarf2out_locviews_in_loclist): New.
4432 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4433 (enum dw_line_info_opcode): Add LI_adv_address.
4434 (struct dw_line_info_table): Add view.
4435 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4436 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4437 (zero_view_p): New variable.
4438 (ZERO_VIEW_P): New macro.
4439 (output_asm_line_debug_info): New.
4440 (struct var_loc_node): Add view.
4441 (add_AT_view_list, AT_loc_list): New.
4442 (add_var_loc_to_decl): Add view param. Test it against last.
4443 (new_loc_list): Add view params. Record them.
4444 (AT_loc_list_ptr): Handle loc and view lists.
4445 (view_list_to_loc_list_val_node): New.
4446 (print_dw_val): Handle dw_val_class_view_list.
4447 (size_of_die): Likewise.
4448 (value_format): Likewise.
4449 (loc_list_has_views): New.
4450 (gen_llsym): Set vl_symbol too.
4451 (maybe_gen_llsym, skip_loc_list_entry): New.
4452 (dwarf2out_maybe_output_loclist_view_pair): New.
4453 (output_loc_list): Output view list or entries too.
4454 (output_view_list_offset): New.
4455 (output_die): Handle dw_val_class_view_list.
4456 (output_dwarf_version): New.
4457 (output_compilation_unit_header): Use it.
4458 (output_skeleton_debug_sections): Likewise.
4459 (output_rnglists, output_line_info): Likewise.
4460 (output_pubnames, output_aranges): Update version comments.
4461 (output_one_line_info_table): Output view numbers in asm comments.
4462 (dw_loc_list): Determine current endview, pass it to new_loc_list.
4463 Call maybe_gen_llsym.
4464 (loc_list_from_tree_1): Adjust.
4465 (add_AT_location_description): Create view list attribute if
4466 needed, check it's absent otherwise.
4467 (convert_cfa_to_fb_loc_list): Adjust.
4468 (maybe_emit_file): Call output_asm_line_debug_info for test.
4469 (dwarf2out_var_location): Reset views as needed. Precompute
4470 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
4471 attribute. Set view.
4472 (new_line_info_table): Reset next view.
4473 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
4474 (dwarf2out_source_line): Likewise. Output view resets and labels to
4475 the assembler, or select appropriate line info opcodes.
4476 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
4477 (optimize_string_length): Catch it. Adjust.
4478 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
4479 dw_val_class_view_list, and remove it if no longer needed.
4480 (hash_loc_list): Hash view numbers.
4481 (loc_list_hasher::equal): Compare them.
4482 (optimize_location_lists): Check whether a view list symbol is
4483 needed, and whether the locview attribute is present, and
4484 whether they match. Remove the locview attribute if no longer
4485 needed.
4486 (index_location_lists): Call skip_loc_list_entry for test.
4487 (dwarf2out_finish): Call output_asm_line_debug_info for test.
4488 Use output_dwarf_version.
4489 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
4490 (struct dw_val_node): Add val_view_list.
4491 * final.c (SEEN_NEXT_VIEW): New.
4492 (set_next_view_needed): New.
4493 (clear_next_view_needed): New.
4494 (maybe_output_next_view): New.
4495 (final_start_function): Rename to...
4496 (final_start_function_1): ... this. Take pointer to FIRST,
4497 add SEEN parameter. Emit param bindings in the initial view.
4498 (final_start_function): Reintroduce SEEN-less interface.
4499 (final): Rename to...
4500 (final_1): ... this. Take SEEN parameter. Output final pending
4501 next view at the end.
4502 (final): Reintroduce seen-less interface.
4503 (final_scan_insn): Output pending next view before switching
4504 sections or ending a block. Mark the next view as needed when
4505 outputting variable locations. Notify debug backend of section
4506 changes, and of location view changes.
4507 (rest_of_handle_final): Adjust.
4508 * toplev.c (process_options): Autodetect value for debug variable
4509 location views option. Warn on incompat5 without -gdwarf-5.
4510 * doc/invoke.texi (gvariable-location-views): New.
4511 (gvariable-location-views=incompat5): New.
4512 (gno-variable-location-views): New.
4514 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4516 PR tree-optimization/84136
4517 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4518 that the result of find_edge is non-NULL.
4520 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4522 PR target/83008
4523 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4524 storing integer register in SImode. Fix cost of 256 and 512
4525 byte aligned SSE register store.
4527 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4529 * config/i386/i386.c (ix86_multiplication_cost): Fix
4530 multiplication cost for TARGET_AVX512DQ.
4532 2018-02-08 Marek Polacek <polacek@redhat.com>
4534 PR tree-optimization/84238
4535 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4536 get_range_strlen.
4538 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4540 PR tree-optimization/84265
4541 * tree-vect-stmts.c (vectorizable_store): Don't treat
4542 VMAT_CONTIGUOUS accesses as grouped.
4543 (vectorizable_load): Likewise.
4545 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4547 PR tree-optimization/81635
4548 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4549 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4550 (test_round_for_mask): New functions.
4551 (wide_int_cc_tests): Call test_round_for_mask.
4552 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4553 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4554 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4555 range returned by get_range_info.
4557 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4559 PR ipa/81360
4560 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4561 * symtab.c: Include builtins.h
4562 (symtab_node::output_to_lto_symbol_table_p): Move here
4563 from lto-streamer-out.c:output_symbol_p.
4564 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4565 (output_symbol_p): Move all logic to symtab.c
4566 (produce_symtab): Update.
4568 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4570 * config/s390/s390-opts.h (enum indirect_branch): Define.
4571 * config/s390/s390-protos.h (s390_return_addr_from_memory)
4572 (s390_indirect_branch_via_thunk)
4573 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4574 (enum s390_indirect_branch_type): Define.
4575 * config/s390/s390.c (struct s390_frame_layout, struct
4576 machine_function): Remove.
4577 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4578 (indirect_branch_table_label_no, indirect_branch_table_name):
4579 Define variables.
4580 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4581 (enum s390_indirect_branch_option): Define.
4582 (s390_return_addr_from_memory): New function.
4583 (s390_handle_string_attribute): New function.
4584 (s390_attribute_table): Add new attribute handler.
4585 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4586 (s390_indirect_branch_via_thunk): New function.
4587 (s390_indirect_branch_via_inline_thunk): New function.
4588 (s390_function_ok_for_sibcall): When jumping via thunk disallow
4589 sibling call optimization for non z10 compiles.
4590 (s390_emit_call): Force indirect branch target to be a single
4591 register. Add r1 clobber for non-z10 compiles.
4592 (s390_emit_epilogue): Emit return jump via return_use expander.
4593 (s390_reorg): Handle JUMP_INSNs as execute targets.
4594 (s390_option_override_internal): Perform validity checks for the
4595 new command line options.
4596 (s390_indirect_branch_attrvalue): New function.
4597 (s390_indirect_branch_settings): New function.
4598 (s390_set_current_function): Invoke s390_indirect_branch_settings.
4599 (s390_output_indirect_thunk_function): New function.
4600 (s390_code_end): Implement target hook.
4601 (s390_case_values_threshold): Implement target hook.
4602 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4603 macros.
4604 * config/s390/s390.h (struct s390_frame_layout)
4605 (struct machine_function): Move here from s390.c.
4606 (TARGET_INDIRECT_BRANCH_NOBP_RET)
4607 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4608 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4609 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4610 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4611 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4612 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4613 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4614 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4615 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4616 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4617 (mnemonic attribute): Add values which aren't recognized
4618 automatically.
4619 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4620 pattern for branch conversion. Fix mnemonic attribute.
4621 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4622 indirect branch via thunk if requested.
4623 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4624 ("*indirect_jump"): Disable for branch conversion using out of
4625 line thunks.
4626 ("indirect_jump_via_thunk<mode>_z10")
4627 ("indirect_jump_via_thunk<mode>")
4628 ("indirect_jump_via_inlinethunk<mode>_z10")
4629 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4630 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4631 ("casesi_jump_via_inlinethunk<mode>_z10")
4632 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4633 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4634 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4635 ("*indirect2_jump"): Disable for branch conversion.
4636 ("casesi_jump"): Turn into expander and expand patterns for branch
4637 conversion.
4638 ("return_use"): New expander.
4639 ("*return"): Emit return via thunk and rename it to ...
4640 ("*return<mode>"): ... this one.
4641 * config/s390/s390.opt: Add new options and and enum for the
4642 option values.
4644 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4646 * lra-constraints.c (match_reload): Unconditionally use
4647 gen_lowpart_SUBREG, rather than selecting between that
4648 and equivalent gen_rtx_SUBREG code.
4650 2018-02-08 Richard Biener <rguenther@suse.de>
4652 PR tree-optimization/84233
4653 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4654 changed flag instead of boguously re-using phi_inserted.
4656 2018-02-08 Martin Jambor <mjambor@suse.cz>
4658 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4659 static local variables.
4661 2018-02-08 Richard Biener <rguenther@suse.de>
4663 PR tree-optimization/84278
4664 * tree-vect-stmts.c (vectorizable_store): When looking for
4665 smaller vector types to perform grouped strided loads/stores
4666 make sure the mode is supported by the target.
4667 (vectorizable_load): Likewise.
4669 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4671 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4672 Increase LDP/STP opportunities by adding adjacent callee-saves.
4674 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4676 PR rtl-optimization/84068
4677 PR rtl-optimization/83459
4678 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4680 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4682 PR tree-optimization/84224
4683 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4684 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4685 non-zero arguments.
4687 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4689 PR target/84113
4690 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4691 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4692 count, remove one USE.
4694 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
4696 * doc/install.texi (Configuration): Document the
4697 --with-long-double-format={ibm,ieee} PowerPC configuration
4698 options.
4700 PR target/84154
4701 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4702 Convert from define_expand to be define_insn_and_split. Rework
4703 float/double/_Float128 conversions to QI/HI/SImode to work with
4704 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
4705 conversions to QI/HImode types did a store and then a load to
4706 truncate the value. For conversions to VSX registers, don't split
4707 the insn, instead emit the code directly. Use the code iterator
4708 any_fix to combine signed and unsigned conversions.
4709 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4710 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4711 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4712 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4713 (fix_<mode>di2_hw): Likewise.
4714 (fixuns_<mode>di2_hw): Likewise.
4715 (fix_<mode>si2_hw): Likewise.
4716 (fixuns_<mode>si2_hw): Likewise.
4717 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4718 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4719 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4720 fix<uns>_trunc<SFDF:mode>si2_p8.
4721 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4722 used.
4723 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4724 (fix<uns>_<mode>_mem): Likewise.
4725 (fctiw<u>z_<mode>_mem): Likewise.
4726 (fix<uns>_<mode>_mem): Likewise.
4727 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4728 the register allocator from doing a direct move to the GPRs to do
4729 a store, and instead use the ISA 3.0 store byte/half-word from
4730 vector register instruction. For IEEE 128-bit floating point,
4731 also optimize stores of 32-bit ints.
4732 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4734 2018-02-07 Alan Hayward <alan.hayward@arm.com>
4736 * genextract.c (push_pathstr_operand): New function to support
4737 [a-zA-Z].
4738 (walk_rtx): Call push_pathstr_operand.
4739 (print_path): Support [a-zA-Z].
4741 2018-02-07 Richard Biener <rguenther@suse.de>
4743 PR tree-optimization/84037
4744 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4745 (cse_and_gimplify_to_preheader): Declare.
4746 (vect_get_place_in_interleaving_chain): Likewise.
4747 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4748 ivexpr_map.
4749 (_loop_vec_info::~_loop_vec_info): Delete it.
4750 (cse_and_gimplify_to_preheader): New function.
4751 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
4752 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
4753 (vectorizable_load): Likewise. For grouped stores always base
4754 the IV on the first element.
4755 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
4756 condition before gimplifying.
4758 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4760 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
4761 *DIV_EXPR and *MOD_EXPR.
4763 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
4765 PR target/84248
4766 * config/i386/i386.c (ix86_option_override_internal): Mask out
4767 the CF_SET bit when checking -fcf-protection.
4769 2018-02-07 Tom de Vries <tom@codesourcery.com>
4771 PR libgomp/84217
4772 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
4773 enough.
4775 2018-02-07 Richard Biener <rguenther@suse.de>
4777 PR tree-optimization/84204
4778 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4779 this place.
4781 PR tree-optimization/84205
4782 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
4783 special-case isl_ast_op_zdiv_r.
4785 PR tree-optimization/84223
4786 * graphite-scop-detection.c (gather_bbs::before_dom_children):
4787 Only add conditions from within the region.
4788 (gather_bbs::after_dom_children): Adjust.
4790 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
4792 PR target/84209
4793 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
4794 * config/avr/avr.md: Only post-reload split REG-REG moves if
4795 either register is GENERAL_REG_P.
4797 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4799 PR tree-optimization/84235
4800 * tree-ssa-scopedtables.c
4801 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
4802 if the subtraction is performed in floating point type where NaNs are
4803 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4804 build 1. Formatting fix.
4806 2018-02-06 Jakub Jelinek <jakub@redhat.com>
4808 PR target/84146
4809 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4810 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4811 and skip it regardless of bb boundaries. Use CALL_P macro,
4812 don't test INSN_P (insn) together with CALL_P or JUMP_P check
4813 unnecessarily, formatting fix.
4815 2018-02-06 Michael Collison <michael.collison@arm.com>
4817 * config/arm/thumb2.md:
4818 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4819 (*thumb_mov_notscc): Ditto.
4821 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
4823 PR target/84154
4824 * config/rs6000/rs6000.md (su code attribute): Use "u" for
4825 unsigned_fix, not "s".
4827 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4829 * configure.ac (gcc_fn_eh_frame_ro): New function.
4830 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4831 correct .eh_frame permissions.
4832 * configure: Regenerate.
4834 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
4836 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4837 irrelevant options.
4839 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4841 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4842 Display warning message for -mno-speculate-indirect-jumps.
4844 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
4846 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4847 Undocumented.
4848 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4850 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
4852 PR tree-optimization/84225
4853 * tree-eh.c (find_trapping_overflow): Only call
4854 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4856 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4858 PR target/84145
4859 * config/i386/i386.c: Reimplement the check of possible options
4860 -mibt/-mshstk conbination. Change error messages.
4861 * doc/invoke.texi: Fix a typo: remove extra '='.
4863 2018-02-06 Marek Polacek <polacek@redhat.com>
4865 PR tree-optimization/84228
4866 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4868 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4870 PR target/82641
4871 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4872 emitted arch directives.
4873 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4874 __ARM_FEATURE_COPROC before changing architectures.
4876 2018-02-06 Richard Biener <rguenther@suse.de>
4878 * config/i386/i386.c (print_reg): Fix typo.
4879 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4881 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4883 * configure: Regenerate.
4885 2018-02-05 Martin Sebor <msebor@redhat.com>
4887 PR tree-optimization/83369
4888 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4889 inlining context.
4891 2018-02-05 Martin Liska <mliska@suse.cz>
4893 * doc/invoke.texi: Cherry-pick upstream r323995.
4895 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4897 * ira.c (ira_init_register_move_cost): Adjust comment.
4899 2018-02-05 Martin Liska <mliska@suse.cz>
4901 PR gcov-profile/84137
4902 * doc/gcov.texi: Fix typo in documentation.
4904 2018-02-05 Martin Liska <mliska@suse.cz>
4906 PR gcov-profile/83879
4907 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4908 using dlopen functionality.
4910 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4912 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4913 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4914 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4915 _mm_maskz_range_round_ss): New intrinsics.
4916 (__builtin_ia32_rangesd128_round)
4917 (__builtin_ia32_rangess128_round): Remove.
4918 (__builtin_ia32_rangesd128_mask_round,
4919 __builtin_ia32_rangess128_mask_round): New builtins.
4920 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4921 __builtin_ia32_rangess128_round): Remove.
4922 (__builtin_ia32_rangesd128_mask_round,
4923 __builtin_ia32_rangess128_mask_round): New builtins.
4924 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4925 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4926 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4927 "<round_saeonly_constraint>")): Changed to ...
4928 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4929 "<round_saeonly_scalar_constraint>")): ... this.
4930 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4931 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4932 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4933 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4934 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
4936 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
4938 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
4939 options.
4940 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
4941 Remove all values except native, 8540 and 8548.
4943 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
4945 * config/i386/i386.c (ix86_output_function_return): Pass
4946 INVALID_REGNUM, instead of -1, as invalid register number to
4947 indirect_thunk_name and output_indirect_thunk.
4949 2018-02-02 Julia Koval <julia.koval@intel.com>
4951 * config.gcc: Add -march=icelake.
4952 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
4953 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
4954 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
4955 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
4956 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
4957 (processor_target_table): Add icelake.
4958 (ix86_option_override_internal): Handle new PTAs.
4959 (get_builtin_code_for_version): Handle icelake.
4960 (M_INTEL_COREI7_ICELAKE): New.
4961 (fold_builtin_cpu): Handle icelake.
4962 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
4963 * doc/invoke.texi: Add -march=icelake.
4965 2018-02-02 Julia Koval <julia.koval@intel.com>
4967 * config/i386/i386.c (ix86_option_override_internal): Change flags type
4968 to wide_int_bitmask.
4969 * wide-int-bitmask.h: New.
4971 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4973 PR target/84066
4974 * config/i386/i386.md: Replace Pmode with word_mode in
4975 builtin_setjmp_setup and builtin_longjmp to support x32.
4977 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
4979 PR target/56010
4980 PR target/83743
4981 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
4982 #include "opts.h".
4983 (rs6000_supported_cpu_names): New static variable.
4984 (linux_cpu_translation_table): Likewise.
4985 (elf_platform) <cpu>: Define new static variable and use it.
4986 Translate kernel AT_PLATFORM name to canonical name if needed.
4987 Error if platform name is unknown.
4989 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
4991 PR target/84089
4992 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
4994 2018-02-01 Jeff Law <law@redhat.com>
4996 PR target/84128
4997 * config/i386/i386.c (release_scratch_register_on_entry): Add new
4998 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
4999 the scratch if RELEASE_VIA_POP is false.
5000 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5001 If we have to save a temporary register, decrement SIZE appropriately.
5002 Pass new arguments to release_scratch_register_on_entry.
5003 (ix86_adjust_stack_and_probe): Likewise.
5004 (ix86_emit_probe_stack_range): Pass new arguments to
5005 release_scratch_register_on_entry.
5007 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5009 PR rtl-optimization/84157
5010 * combine.c (change_zero_ext): Use REG_P predicate in
5011 front of HARD_REGISTER_P predicate.
5013 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5015 * config/avr/avr.c (avr_option_override): Move disabling of
5016 -fdelete-null-pointer-checks to...
5017 * common/config/avr/avr-common.c (avr_option_optimization_table):
5018 ...here.
5020 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5022 PR tree-optimization/81635
5023 * tree-data-ref.c (split_constant_offset_1): For types that
5024 wrap on overflow, try to use range info to prove that wrapping
5025 cannot occur.
5027 2018-02-01 Renlin Li <renlin.li@arm.com>
5029 PR target/83370
5030 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5031 TAILCALL_ADDR_REGS.
5032 (aarch64_register_move_cost): Likewise.
5033 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5034 TAILCALL_ADDR_REGS.
5035 (REG_CLASS_NAMES): Likewise.
5036 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5037 TAILCALL_ADDR_REGS. Remove IP registers.
5038 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5040 2018-02-01 Richard Biener <rguenther@suse.de>
5042 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5043 for specifying RPO order and allow NULL for that.
5044 * domwalk.c (dom_walker::dom_walker): Likewise.
5045 (dom_walker::walk): Handle NULL RPO order.
5046 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5047 in RPO order.
5048 (rewrite_update_dom_walker): Likewise.
5049 (mark_def_dom_walker): Likewise.
5051 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5053 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5054 (aarch64_maybe_expand_sve_subreg_move): Declare.
5055 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5056 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5057 predicate.
5058 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5059 that are semantically a reverse operation.
5060 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5061 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5062 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5063 functions.
5064 (aarch64_can_change_mode_class): For big-endian, forbid changes
5065 between two SVE modes if they have different element sizes.
5067 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5069 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5070 the TImode handling for big-endian targets.
5072 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5074 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5075 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5076 not just bytes.
5077 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5078 Remove BSWAP handing for big-endian targets and use the form of
5079 LD1RQ appropariate for the mode.
5081 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5083 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5084 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5085 duplicated element.
5087 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5089 PR tearget/83845
5090 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5091 check for operands that need to go through aarch64_sve_reload_be.
5093 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5095 PR tree-optimization/81661
5096 PR tree-optimization/84117
5097 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5098 * tree-eh.c: Include gimplify.h.
5099 (find_trapping_overflow, replace_trapping_overflow,
5100 rewrite_to_non_trapping_overflow): New functions.
5101 * tree-vect-loop.c: Include tree-eh.h.
5102 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5103 * tree-data-ref.c: Include tree-eh.h.
5104 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5106 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5108 PR rtl-optimization/84123
5109 * combine.c (change_zero_ext): Check if hard register satisfies
5110 can_change_dest_mode before calling gen_lowpart_SUBREG.
5112 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5114 PR target/82444
5115 * ira.c (ira_init_register_move_cost): Remove assert.
5117 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5119 PR rtl-optimization/84071
5120 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5121 * doc/tm.texi: Regenerate.
5123 2018-01-31 Richard Biener <rguenther@suse.de>
5125 PR tree-optimization/84132
5126 * tree-data-ref.c (analyze_miv_subscript): Properly
5127 check whether evolution_function_is_affine_multivariate_p
5128 before calling gcd_of_steps_may_divide_p.
5130 2018-01-31 Julia Koval <julia.koval@intel.com>
5132 PR target/83618
5133 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5134 * config/i386/i386.md (rdpid_rex64) New.
5135 (rdpid): Make 32bit only.
5137 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5139 PR lto/84105
5140 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5141 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5143 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5145 Revert
5146 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5148 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5150 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5152 PR rtl-optimization/84071
5153 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5154 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5156 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5158 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5159 (arc_attribute_table): Add 'aux' attribute.
5160 (arc_in_small_data_p): Consider aux like variables.
5161 (arc_is_aux_reg_p): New function.
5162 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5163 (arc_get_aux_arg): New function.
5164 (prepare_move_operands): Handle aux-register access.
5165 (arc_handle_aux_attribute): New function.
5166 * doc/extend.texi (ARC Variable attributes): Add subsection.
5168 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5170 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5171 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5172 (arc_attribute_table): Add 'uncached' attribute.
5173 (arc_print_operand): Print '.di' flag for uncached memory
5174 accesses.
5175 (arc_in_small_data_p): Do not consider for small data the uncached
5176 types.
5177 (arc_is_uncached_mem_p): New function.
5178 * config/arc/predicates.md (compact_store_memory_operand): Check
5179 for uncached memory accesses.
5180 (nonvol_nonimm_operand): Likewise.
5181 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5183 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5185 PR c/84100
5186 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5187 falign-loops=): Add Optimization flag.
5189 2018-01-30 Jeff Law <law@redhat.com>
5191 PR target/84064
5192 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5193 INT_REGISTERS_SAVED. Check it prior to calling
5194 get_scratch_register_on_entry.
5195 (ix86_adjust_stack_and_probe): Similarly.
5196 (ix86_emit_probe_stack_range): Similarly.
5197 (ix86_expand_prologue): Corresponding changes.
5199 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5201 PR target/40411
5202 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5203 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5205 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5207 PR target/84112
5208 * lra-constraints.c (curr_insn_transform): Process AND in the
5209 address.
5211 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5213 PR rtl-optimization/83986
5214 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5215 dependence against last_pending_memory_flush in addition to
5216 pending_jump_insns.
5218 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5220 PR tree-optimization/81611
5221 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5222 copies.
5224 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5226 PR target/83758
5227 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5228 a reg rtx.
5230 2018-01-30 Richard Biener <rguenther@suse.de>
5231 Jakub Jelinek <jakub@redhat.com>
5233 PR tree-optimization/84111
5234 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5235 inner loops added during recursion, as they don't have up-to-date
5236 SSA form.
5238 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5240 PR ipa/81360
5241 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5242 (can_inline_edge_by_limits_p): ... here.
5243 (can_early_inline_edge_p, check_callers,
5244 update_caller_keys, update_callee_keys, recursive_inlining,
5245 add_new_edges_to_heap, speculation_useful_p,
5246 inline_small_functions,
5247 inline_small_functions, flatten_function,
5248 inline_to_all_callers_1): Update.
5250 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5252 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5253 zeros correctly.
5255 2018-01-30 Richard Biener <rguenther@suse.de>
5257 PR tree-optimization/83008
5258 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5259 invariant and constant vector uses in stmts when they need
5260 more than one stmt.
5262 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5264 PR bootstrap/84017
5265 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5266 * configure: Regenerate.
5268 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5270 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5271 pattern.
5272 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5273 Use gen_rtx_REG rather than gen_lowpart.
5275 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5277 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5278 rather than 0 when creating partial subregs.
5280 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5282 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5283 of usage.
5285 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5287 PR target/81550
5288 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5289 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5290 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5291 flags. This restores the settings used before the 2017-07-24.
5292 Turning off pre increment/decrement/modify allows IVOPTS to
5293 optimize DF/SF loops where the index is an int.
5295 2018-01-29 Richard Biener <rguenther@suse.de>
5296 Kelvin Nilsen <kelvin@gcc.gnu.org>
5298 PR bootstrap/80867
5299 * tree-vect-stmts.c (vectorizable_call): Don't call
5300 targetm.vectorize_builtin_md_vectorized_function if callee is
5301 NULL.
5303 2018-01-22 Carl Love <cel@us.ibm.com>
5305 * doc/extend.tex: Fix typo in second arg in
5306 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5308 2018-01-29 Richard Biener <rguenther@suse.de>
5310 PR tree-optimization/84086
5311 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5312 (flush_ssaname_freelist): When SSA names were released reset
5313 the SCEV hash table.
5315 2018-01-29 Richard Biener <rguenther@suse.de>
5317 PR tree-optimization/84057
5318 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5319 removed paths when removing edges.
5321 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5323 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5324 -mfunction-return=@var{choice}.
5326 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5328 PR diagnostic/84034
5329 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5330 Handle CR like TAB.
5331 (layout::print_source_line): Likewise.
5332 (test_get_line_width_without_trailing_whitespace): Add test cases.
5334 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5336 PR middle-end/84040
5337 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5338 debug insns.
5340 2018-01-26 Jim Wilson <jimw@sifive.com>
5342 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5344 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5345 specified.
5347 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5349 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5350 and CMP + SUB-immediate -> SUBS.
5352 2018-01-26 Martin Sebor <msebor@redhat.com>
5354 PR tree-optimization/83896
5355 * tree-ssa-strlen.c (get_string_len): Rename...
5356 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5357 Avoid assuming length is constant.
5358 (handle_char_store): Use HOST_WIDE_INT for string length.
5360 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5362 PR target/81763
5363 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5364 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5366 2018-01-26 Richard Biener <rguenther@suse.de>
5368 PR rtl-optimization/84003
5369 * dse.c (record_store): Only record redundant stores when
5370 the earlier store aliases at least all accesses the later one does.
5372 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5374 PR rtl-optimization/83985
5375 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5376 REG_CFA_RESTORE insns.
5377 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5378 REG_CFA_RESTORE insns here.
5380 PR c/83989
5381 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5382 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5384 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5386 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5387 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5388 (arc_init): Likewise.
5389 (arc_override_options): Likewise.
5390 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5391 value.
5392 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5393 support.
5394 * config/arc/arc.h (TARGET_DBNZ): Define.
5395 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5396 properly set the tune attribute.
5397 (dbnz): Use TARGET_DBNZ guard.
5398 * config/arc/arc.opt (mtune): Add core3 option.
5400 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5402 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5403 recognize new pic like addresses.
5404 (arc_delegitimize_address): Clean up.
5406 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5408 * config/arc/arc-arches.def: Option mrf16 valid for all
5409 architectures.
5410 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5411 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5412 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5413 * config/arc/arc-tables.opt: Regenerate.
5414 * config/arc/arc.c (arc_conditional_register_usage): Handle
5415 reduced register file case.
5416 (arc_file_start): Set must have build attributes.
5417 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5418 mrf16 option value.
5419 * config/arc/arc.opt (mrf16): Add new option.
5420 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5421 * config/arc/genmultilib.awk: Handle new mrf16 option.
5422 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5423 * config/arc/t-multilib: Regenerate.
5424 * doc/invoke.texi (ARC Options): Document mrf16 option.
5426 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5428 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5429 * config/arc/arc.c (arc_handle_secure_attribute): New function.
5430 (arc_attribute_table): Add 'secure_call' attribute.
5431 (arc_print_operand): Print secure call operand.
5432 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5433 secure.
5434 (arc_is_secure_call_p): New function. * config/arc/arc.md
5435 (call_i): Add support for sjli instruction.
5436 (call_value_i): Likewise.
5437 * config/arc/constraints.md (Csc): New constraint.
5439 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5440 John Eric Martin <John.Martin@emmicro-us.com>
5442 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5443 * config/arc/arc.c (_arc_jli_section): New struct.
5444 (arc_jli_section): New type.
5445 (rc_jli_sections): New static variable.
5446 (arc_handle_jli_attribute): New function.
5447 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5448 (arc_file_end): New function.
5449 (TARGET_ASM_FILE_END): Define.
5450 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5451 (arc_add_jli_section): New function.
5452 (jli_call_scan): Likewise.
5453 (arc_reorg): Call jli_call_scan.
5454 (arc_output_addsi): Remove 'S' from printing asm operand.
5455 (arc_is_jli_call_p): New function.
5456 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5457 operand.
5458 (movhi_insn): Likewise.
5459 (movsi_insn): Likewise.
5460 (movsi_set_cc_insn): Likewise.
5461 (loadqi_update): Likewise.
5462 (load_zeroextendqisi_update): Likewise.
5463 (load_signextendqisi_update): Likewise.
5464 (loadhi_update): Likewise.
5465 (load_zeroextendhisi_update): Likewise.
5466 (load_signextendhisi_update): Likewise.
5467 (loadsi_update): Likewise.
5468 (loadsf_update): Likewise.
5469 (movsicc_insn): Likewise.
5470 (bset_insn): Likewise.
5471 (bxor_insn): Likewise.
5472 (bclr_insn): Likewise.
5473 (bmsk_insn): Likewise.
5474 (bicsi3_insn): Likewise.
5475 (cmpsi_cc_c_insn): Likewise.
5476 (movsi_ne): Likewise.
5477 (movsi_cond_exec): Likewise.
5478 (clrsbsi2): Likewise.
5479 (norm_f): Likewise.
5480 (normw): Likewise.
5481 (swap): Likewise.
5482 (divaw): Likewise.
5483 (flag): Likewise.
5484 (sr): Likewise.
5485 (kflag): Likewise.
5486 (ffs): Likewise.
5487 (ffs_f): Likewise.
5488 (fls): Likewise.
5489 (call_i): Remove 'S' asm letter, add jli instruction.
5490 (call_value_i): Likewise.
5491 * config/arc/arc.op (mjli-always): New option.
5492 * config/arc/constraints.md (Cji): New constraint.
5493 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
5494 operand.
5495 (subsf3_fpx): Likewise.
5496 (mulsf3_fpx): Likewise.
5497 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5498 asm operand.
5499 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5500 function attrbutes.
5501 * doc/invoke.texi (ARC): Document mjli-always option.
5503 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
5505 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5506 avoid addition with 0 and use incw and decw where possible.
5508 2018-01-26 Richard Biener <rguenther@suse.de>
5510 PR tree-optimization/81082
5511 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5512 association if it requires casting to unsigned.
5513 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5514 from fold_plusminus_mult_expr to catch important cases late when
5515 range info is available.
5517 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5519 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5520 * configure.ac (hidden_linkonce): New test.
5521 * configure: Regenerate.
5522 * config.in: Regenerate.
5524 2018-01-26 Julia Koval <julia.koval@intel.com>
5526 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5527 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5528 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5529 _mm_mask_bitshuffle_epi64_mask): Fix type.
5530 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5531 USI_FTYPE_V4DI_V4DI_USI): Remove.
5532 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5533 __builtin_ia32_vpshufbitqmb256_mask,
5534 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5535 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5536 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5538 2018-01-26 Alan Modra <amodra@gmail.com>
5540 PR target/84033
5541 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5542 UNSPEC_VBPERMQ. Sort other unspecs.
5544 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5546 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5548 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5550 PR middle-end/83055
5551 * predict.c (drop_profile): Do not push/pop cfun; update also
5552 node->count.
5553 (handle_missing_profiles): Fix logic looking for zero profiles.
5555 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5557 PR middle-end/83977
5558 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5559 on functions with #pragma omp declare simd or functions with simd
5560 attribute.
5561 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5562 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5563 Remove trailing \n from warning_at calls.
5565 2018-01-25 Tom de Vries <tom@codesourcery.com>
5567 PR target/84028
5568 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5569 for neutered workers.
5571 2018-01-24 Joseph Myers <joseph@codesourcery.com>
5573 PR target/68467
5574 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5575 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5577 2018-01-24 Jeff Law <law@redhat.com>
5579 PR target/83994
5580 * i386.c (get_probe_interval): Move to earlier point.
5581 (ix86_compute_frame_layout): If -fstack-clash-protection and
5582 the frame is larger than the probe interval, then use pushes
5583 to save registers rather than reg->mem moves.
5584 (ix86_expand_prologue): Remove conditional for int_registers_saved
5585 assertion.
5587 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
5589 PR target/84014
5590 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5591 min/max for never referenced object.
5593 2018-01-24 Jakub Jelinek <jakub@redhat.com>
5595 PR middle-end/83977
5596 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5597 here.
5598 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5599 attributes from DECL_ATTRIBUTES (decl) without affecting
5600 DECL_ATTRIBUTES (current_function_decl).
5601 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5602 functions with non-NULL DECL_ABSTRACT_ORIGIN.
5604 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
5606 PR tree-optimization/83979
5607 * fold-const.c (fold_comparison): Use constant_boolean_node
5608 instead of boolean_{true,false}_node.
5610 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
5612 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5613 with zero counts.
5615 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5617 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5618 Simplify the clause that sets the length attribute.
5619 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5620 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5621 clause that sets the length attribute.
5622 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5624 2018-01-24 Tom de Vries <tom@codesourcery.com>
5626 PR target/83589
5627 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5628 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5629 Add strict parameter.
5630 (prevent_branch_around_nothing): Insert dummy insn between branch to
5631 label and label with no ptx insn inbetween.
5632 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5634 2018-01-24 Tom de Vries <tom@codesourcery.com>
5636 PR target/81352
5637 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5638 for neutered threads in warp.
5639 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5641 2018-01-24 Richard Biener <rguenther@suse.de>
5643 PR tree-optimization/83176
5644 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5645 operands.
5647 2018-01-24 Richard Biener <rguenther@suse.de>
5649 PR tree-optimization/82819
5650 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5651 code generating pluses that are no-ops in the target precision.
5653 2018-01-24 Richard Biener <rguenther@suse.de>
5655 PR middle-end/84000
5656 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5658 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5660 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5661 to merge probabilities.
5662 * predict.c (probably_never_executed): Also mark as cold functions
5663 with global 0 profile and guessed local profile.
5664 * profile-count.c (profile_probability::combine_with_count): New
5665 member function.
5666 * profile-count.h (profile_probability::operator*,
5667 profile_probability::operator*=, profile_probability::operator/,
5668 profile_probability::operator/=): Reduce precision to adjusted
5669 and set value to guessed on contradictory divisions.
5670 (profile_probability::combine_with_freq): Remove.
5671 (profile_probability::combine_wiht_count): Declare.
5672 (profile_count::force_nonzero):: Set to adjusted.
5673 (profile_count::probability_in):: Set quality to adjusted.
5674 * tree-ssa-tail-merge.c (replace_block_by): Use
5675 combine_with_count.
5677 2018-01-23 Andrew Waterman <andrew@sifive.com>
5678 Jim Wilson <jimw@sifive.com>
5680 * config/riscv/riscv.c (riscv_stack_boundary): New.
5681 (riscv_option_override): Set riscv_stack_boundary. Handle
5682 riscv_preferred_stack_boundary_arg.
5683 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5684 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5685 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5686 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5687 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5688 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5690 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5692 PR target/83905
5693 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5694 of struct ix86_frame.
5695 (ix86_expand_epilogue): Likewise. Add a local variable for
5696 the reg_save_offset field in struct ix86_frame.
5698 2018-01-23 Bin Cheng <bin.cheng@arm.com>
5700 PR tree-optimization/82604
5701 * tree-loop-distribution.c (enum partition_kind): New enum item
5702 PKIND_PARTIAL_MEMSET.
5703 (partition_builtin_p): Support above new enum item.
5704 (generate_code_for_partition): Ditto.
5705 (compute_access_range): Differentiate cases that equality can be
5706 proven at all loops, the innermost loops or no loops.
5707 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5708 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
5709 (finalize_partitions, distribute_loop): Don't fuse partition of
5710 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5711 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5712 parloop is enabled.
5714 2018-01-23 Martin Liska <mliska@suse.cz>
5716 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5717 order to ignore the predictor.
5718 (PRED_POLYMORPHIC_CALL): Likewise.
5719 (PRED_RECURSIVE_CALL): Likewise.
5721 2018-01-23 Martin Liska <mliska@suse.cz>
5723 * tree-profile.c (tree_profiling): Print function header to
5724 aware reader which function we are working on.
5725 * value-prof.c (gimple_find_values_to_profile): Do not print
5726 not interesting value histograms.
5728 2018-01-23 Martin Liska <mliska@suse.cz>
5730 * profile-count.h (enum profile_quality): Add
5731 profile_uninitialized as the first value. Do not number values
5732 as they are zero based.
5733 (profile_count::verify): Update sanity check.
5734 (profile_probability::verify): Likewise.
5736 2018-01-23 Nathan Sidwell <nathan@acm.org>
5738 * doc/invoke.texi (ffor-scope): Deprecate.
5740 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5742 PR tree-optimization/83510
5743 * domwalk.c (set_all_edges_as_executable): New function.
5744 (dom_walker::dom_walker): Convert bool param
5745 "skip_unreachable_blocks" to enum reachability. Move setup of
5746 edge flags to set_all_edges_as_executable and only do it when
5747 reachability is REACHABLE_BLOCKS.
5748 * domwalk.h (enum dom_walker::reachability): New enum.
5749 (dom_walker::dom_walker): Convert bool param
5750 "skip_unreachable_blocks" to enum reachability.
5751 (set_all_edges_as_executable): New decl.
5752 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
5753 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
5754 "reachability".
5755 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
5756 but converting true to REACHABLE_BLOCKS.
5757 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
5758 * tree-vrp.c
5759 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
5760 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
5761 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
5762 REACHABLE_BLOCKS.
5763 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
5764 if check_all_array_refs will be called.
5766 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5768 * tree.c (selftest::test_location_wrappers): Add more test
5769 coverage.
5771 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5773 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
5774 (selftest::test_bit_in_range): Likewise.
5776 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5778 PR testsuite/83888
5779 * doc/sourcebuild.texi (vect_float): Say that the selector
5780 only describes the situation when -funsafe-math-optimizations is on.
5781 (vect_float_strict): Document.
5783 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5785 PR tree-optimization/83965
5786 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5787 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
5788 instead of checking only for a reduction.
5789 (vect_recog_widen_sum_pattern): Likewise.
5791 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5793 * predict.c (probably_never_executed): Only use precise profile info.
5794 (compute_function_frequency): Skip after inlining hack since we now
5795 have quality checking.
5797 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5799 * profile-count.h (profile_probability::very_unlikely,
5800 profile_probability::unlikely, profile_probability::even): Set
5801 precision to guessed.
5803 2018-01-23 Richard Biener <rguenther@suse.de>
5805 PR tree-optimization/83963
5806 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5807 Properly terminate dominator walk when crossing the exit edge not
5808 when visiting its source block.
5810 2018-01-23 Jakub Jelinek <jakub@redhat.com>
5812 PR c++/83918
5813 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5814 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5816 2018-01-22 Jakub Jelinek <jakub@redhat.com>
5818 PR tree-optimization/83957
5819 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
5820 semicolon after for body surrounded by braces.
5822 PR tree-optimization/83081
5823 * profile-count.h (profile_probability::split): New method.
5824 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5825 Use profile_probability::split.
5826 (do_compare_rtx_and_jump): Fix adjustment of probabilities
5827 when splitting a single conditional jump into 2.
5829 2018-01-22 David Malcolm <dmalcolm@redhat.com>
5831 PR tree-optimization/69452
5832 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5833 decl.
5835 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5837 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5838 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5839 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5841 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5843 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5844 * config/rl78/rl78.md (movdi): New define_expand.
5845 * config/rl78/rl78.c (rl78_split_movdi): New function.
5847 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
5849 PR target/83862
5850 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5851 no longer used.
5852 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5853 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5854 128-bit to produce an UNSPEC move to get the double word with the
5855 signbit and then a shift directly to do signbit.
5856 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5857 implementation with a new version that just does either a direct
5858 move or a regular move. Move memory interface to separate insns.
5859 Move insns so they are next to the expander.
5860 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5861 with signbit move. Split big and little endian case.
5862 (signbit<mode>2_dm_mem_le): Likewise.
5863 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5864 (signbit<mode>2_dm2): Likewise.
5866 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5868 * config/rl78/rl78.md (anddi3): New define_expand.
5870 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5872 * config/rl78/rl78.md (umindi3): New define_expand.
5874 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5876 * config/rl78/rl78.md (smindi3): New define_expand.
5878 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5880 * config/rl78/rl78.md (smaxdi3): New define_expand.
5882 2018-01-22 Carl Love <cel@us.ibm.com>
5884 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5885 LVX_V1TI): Add macro expansion.
5886 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5887 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5888 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5889 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5890 Change check to determine if the instruction is a byte reversing
5891 entry. Fix typo in comment.
5892 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5893 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5894 Add def_builtin calls for new builtins.
5895 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5896 Add define_insn expansion.
5898 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5900 * config/rl78/rl78.md (umaxdi3): New define_expand.
5902 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5904 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5905 for non-QImode registers.
5907 2018-01-22 Richard Biener <rguenther@suse.de>
5909 PR tree-optimization/83963
5910 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5911 including the loop exit block.
5912 (scop_detection::merge_sese): Likewise.
5913 (scop_detection::add_scop): Do it here instead.
5915 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5917 * doc/sourcebuild.texi (arm_softfloat): Document.
5919 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5921 PR gcc/77734
5922 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5923 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5924 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5926 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5927 David Edelsohn <dje.gcc@gmail.com>
5929 PR target/83946
5930 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5931 Change "crset eq" to "crset 2".
5932 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5933 (*call_indirect_aix<mode>_nospec): Likewise.
5934 (*call_value_indirect_aix<mode>_nospec): Likewise.
5935 (*call_indirect_elfv2<mode>_nospec): Likewise.
5936 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
5937 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
5938 change assembly output from . to $.
5939 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5940 (indirect_jump<mode>_nospec): Change assembly output from . to $.
5941 (*tablejump<mode>_internal1_nospec): Likewise.
5943 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
5945 PR target/80870
5946 * config/sh/sh_optimize_sett_clrt.cc:
5947 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
5949 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
5951 PR tree-optimization/83940
5952 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
5953 offset_dt to vect_constant_def rather than vect_unknown_def_type.
5954 (vect_check_load_store_mask): Add a mask_dt_out parameter and
5955 use it to pass back the definition type.
5956 (vect_check_store_rhs): Likewise rhs_dt_out.
5957 (vect_build_gather_load_calls): Add a mask_dt argument and use
5958 it instead of a call to vect_is_simple_use.
5959 (vectorizable_store): Update calls to vect_check_load_store_mask
5960 and vect_check_store_rhs. Use the dt returned by the latter instead
5961 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
5962 instead of calls to vect_is_simple_use. Pass the scalar rather
5963 than the vector operand to vect_is_simple_use when handling
5964 second and subsequent copies of an rhs value.
5965 (vectorizable_load): Update calls to vect_check_load_store_mask
5966 and vect_build_gather_load_calls. Use the cached mask_dt and
5967 gs_info.offset_dt instead of calls to vect_is_simple_use.
5969 2018-01-20 Jakub Jelinek <jakub@redhat.com>
5971 PR middle-end/83945
5972 * tree-emutls.c: Include gimplify.h.
5973 (lower_emutls_2): New function.
5974 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
5975 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
5976 it before further processing.
5978 PR target/83930
5979 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
5980 UINTVAL (trueop1) instead of INTVAL (op1).
5982 2018-01-19 Jakub Jelinek <jakub@redhat.com>
5984 PR debug/81570
5985 PR debug/83728
5986 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
5987 INCOMING_FRAME_SP_OFFSET if not defined.
5988 (scan_trace): Add ENTRY argument. If true and
5989 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
5990 emit a note to adjust the CFA offset.
5991 (create_cfi_notes): Adjust scan_trace callers.
5992 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
5993 INCOMING_FRAME_SP_OFFSET in the CIE.
5994 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
5995 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
5996 Likewise.
5997 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
5998 * doc/tm.texi: Regenerated.
6000 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6002 PR rtl-optimization/83147
6003 * lra-constraints.c (remove_inheritance_pseudos): Use
6004 lra_substitute_pseudo_within_insn.
6006 2018-01-19 Tom de Vries <tom@codesourcery.com>
6007 Cesar Philippidis <cesar@codesourcery.com>
6009 PR target/83920
6010 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6012 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6014 PR target/83790
6015 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6016 spaces for function labels.
6018 2018-01-19 Martin Liska <mliska@suse.cz>
6020 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6021 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6022 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6023 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6024 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6025 (PRED_CONST_RETURN): Change from 69 to 65.
6026 (PRED_NULL_RETURN): Change from 91 to 71.
6027 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6028 (PRED_LOOP_GUARD): Change from 66 to 73.
6030 2018-01-19 Martin Liska <mliska@suse.cz>
6032 * predict.c (predict_insn_def): Add new assert.
6033 (struct branch_predictor): Change type to signed integer.
6034 (test_prediction_value_range): Amend test to cover
6035 PROB_UNINITIALIZED.
6036 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6037 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6038 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6039 (PRED_LOOP_IV_COMPARE): Likewise.
6040 * predict.h (PROB_UNINITIALIZED): Define new constant.
6042 2018-01-19 Martin Liska <mliska@suse.cz>
6044 * predict.c (dump_prediction): Add new format for
6045 analyze_brprob.py script which is enabled with -details
6046 suboption.
6047 * profile-count.h (precise_p): New function.
6049 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6051 PR tree-optimization/83922
6052 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6053 there are no statements that need masking.
6054 (vect_active_double_reduction_p): New function.
6055 (vect_analyze_loop_operations): Use it when handling phis that
6056 are not in the loop header.
6058 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6060 PR tree-optimization/83914
6061 * tree-vect-loop.c (vectorizable_induction): Don't convert
6062 init_expr or apply the peeling adjustment for inductions
6063 that are nested within the vectorized loop.
6065 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6067 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6068 instead of NEG.
6070 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6072 PR sanitizer/81715
6073 PR testsuite/83882
6074 * function.h (gimplify_parameters): Add gimple_seq * argument.
6075 * function.c: Include gimple.h and options.h.
6076 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6077 for the added local temporaries if needed.
6078 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6079 if there are any parameter cleanups, wrap whole body into a
6080 try/finally with the cleanups.
6082 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6084 PR target/82964
6085 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6086 Use GET_MODE_CLASS for scalar floating point.
6088 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6090 PR ipa/82256
6091 patch by PaX Team
6092 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6093 Fix call of call_cgraph_insertion_hooks.
6095 2018-01-18 Martin Sebor <msebor@redhat.com>
6097 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6099 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6101 PR ipa/83619
6102 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6103 frequencies.
6105 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6107 PR other/70268
6108 * common.opt: (-ffile-prefix-map): New option.
6109 * opts.c (common_handle_option): Defer it.
6110 * opts-global.c (handle_common_deferred_options): Handle it.
6111 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6112 * file-prefix-map.h: New file.
6113 (remap_debug_filename, add_debug_prefix_map): ...here.
6114 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6115 * final.c (debug_prefix_map, add_debug_prefix_map
6116 remap_debug_filename): Move to...
6117 * file-prefix-map.c: New file.
6118 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6119 generalize, get rid of alloca(), use strrchr() instead of strchr().
6120 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6121 Implement in terms of add_prefix_map().
6122 (remap_macro_filename, remap_debug_filename): Implement in term of
6123 remap_filename().
6124 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6125 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6126 * dbxout.c: Include file-prefix-map.h.
6127 * varasm.c: Likewise.
6128 * vmsdbgout.c: Likewise.
6129 * xcoffout.c: Likewise.
6130 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6131 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6132 * doc/invoke.texi (-ffile-prefix-map): Document.
6133 (-fdebug-prefix-map): Update description.
6135 2018-01-18 Martin Liska <mliska@suse.cz>
6137 * config/i386/i386.c (indirect_thunk_name): Document that also
6138 lfence is emitted.
6139 (output_indirect_thunk): Document why both instructions
6140 (pause and lfence) are generated.
6142 2018-01-18 Richard Biener <rguenther@suse.de>
6144 PR tree-optimization/83887
6145 * graphite-scop-detection.c
6146 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6147 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6148 (scop_detection::merge_sese): Re-implement with a flood-fill
6149 algorithm that properly finds a SESE region if it exists.
6151 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6153 PR c/61240
6154 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6155 pointer_diff optimizations use view_convert instead of convert.
6157 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6159 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6160 Generate different code for -mno-speculate-indirect-jumps.
6161 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6162 (*call_indirect_aix<mode>): Disable for
6163 -mno-speculate-indirect-jumps.
6164 (*call_indirect_aix<mode>_nospec): New define_insn.
6165 (*call_value_indirect_aix<mode>): Disable for
6166 -mno-speculate-indirect-jumps.
6167 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6168 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6169 -mno-speculate-indirect-jumps.
6170 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6172 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6174 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6175 long double type, set the flags for noting the default long double
6176 type, even if we don't pass or return a long double type.
6178 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6180 PR ipa/83051
6181 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6182 failure.
6184 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6186 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6187 support for merge[hl].
6188 (fold_mergehl_helper): New helper function.
6189 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6190 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6191 (altivec_vmrglw_direct): Add xxmrglw insn.
6193 2018-01-17 Andrew Waterman <andrew@sifive.com>
6195 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6196 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6198 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6200 PR lto/83121
6201 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6202 call the lto_location_cache before reading the
6203 DECL_SOURCE_LOCATION of the types.
6205 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6206 Richard Sandiford <richard.sandiford@linaro.org>
6208 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6209 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6210 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6211 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6212 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6213 Add declaration.
6214 * config/aarch64/constraints.md (aarch64_movti_operand):
6215 Limit immediates.
6216 * config/aarch64/predicates.md (Uti): Add new constraint.
6218 2018-01-17 Carl Love <cel@us.ibm.com>
6220 * config/rs6000/vsx.md (define_expand xl_len_r,
6221 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6222 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6223 lxvll.
6224 (define_expand, define_insn): Move the shift left from the
6225 define_insn to the define_expand for lxvl and stxvl instructions.
6226 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6227 and XL_LEN_R definitions to PURE.
6229 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6231 * config/i386/i386.c (indirect_thunk_name): Declare regno
6232 as unsigned int. Compare regno with INVALID_REGNUM.
6233 (output_indirect_thunk): Ditto.
6234 (output_indirect_thunk_function): Ditto.
6235 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6236 in the call to output_indirect_thunk_function.
6238 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6240 PR middle-end/83884
6241 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6242 rather than the size of inner_type to determine the stack slot size
6243 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6245 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6247 PR target/83546
6248 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6249 to PTA_SILVERMONT.
6251 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6253 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6254 endian Linux systems to optionally enable multilibs for selecting
6255 the long double type if the user configured an explicit type.
6256 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6257 have no long double multilibs if not defined.
6258 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6259 warn if the user used -mabi={ieee,ibm}longdouble and we built
6260 multilibs for long double.
6261 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6262 appropriate multilib option.
6263 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6264 multilib options.
6265 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6266 for building long double multilibs.
6267 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6269 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6271 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6272 copies.
6274 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6275 64 bits.
6276 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6277 128 bits.
6279 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6280 variables.
6282 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6283 return value.
6285 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6287 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6288 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6290 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6292 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6293 different rtl trees depending on TARGET_64BIT.
6294 (rs6000_gen_lvx): Likewise.
6296 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6298 * config/visium/visium.md (nop): Tweak comment.
6299 (hazard_nop): Likewise.
6301 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6303 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6304 -mspeculate-indirect-jumps.
6305 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6306 for -mno-speculate-indirect-jumps.
6307 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6308 (*call_value_indirect_elfv2<mode>): Disable for
6309 -mno-speculate-indirect-jumps.
6310 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6311 (indirect_jump): Emit different RTL for
6312 -mno-speculate-indirect-jumps.
6313 (*indirect_jump<mode>): Disable for
6314 -mno-speculate-indirect-jumps.
6315 (*indirect_jump<mode>_nospec): New define_insn.
6316 (tablejump): Emit different RTL for
6317 -mno-speculate-indirect-jumps.
6318 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6319 (tablejumpsi_nospec): New define_expand.
6320 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6321 (tablejumpdi_nospec): New define_expand.
6322 (*tablejump<mode>_internal1): Disable for
6323 -mno-speculate-indirect-jumps.
6324 (*tablejump<mode>_internal1_nospec): New define_insn.
6325 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6326 option.
6328 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6330 * caller-save.c (insert_save): Drop unnecessary parameter. All
6331 callers updated.
6333 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6334 Richard Biener <rguenth@suse.de>
6336 PR libgomp/83590
6337 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6338 return early, inline manually is_gimple_sizepos. Make sure if we
6339 call gimplify_expr we don't end up with a gimple constant.
6340 * tree.c (variably_modified_type_p): Don't return true for
6341 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6342 * gimplify.h (is_gimple_sizepos): Remove.
6344 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6346 PR tree-optimization/83857
6347 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6348 vectorizable_live_operation for pure SLP statements.
6349 (vectorizable_live_operation): Handle PHIs.
6351 2018-01-16 Richard Biener <rguenther@suse.de>
6353 PR tree-optimization/83867
6354 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6355 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6357 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6359 PR c/83844
6360 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6361 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6362 If off is not INTEGER_CST, issue a may not be aligned warning
6363 rather than isn't aligned. Use isn%'t rather than isn't.
6364 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6365 into MULT_EXPR.
6366 <case MULT_EXPR>: Improve the case when bottom and one of the
6367 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6368 operand, in that case check if the other operand is multiple of
6369 bottom divided by the INTEGER_CST operand.
6371 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6373 PR target/83858
6374 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6375 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6376 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6377 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6378 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6379 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6380 (pa_function_arg_size): New function.
6382 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6384 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6385 in a separate statement.
6387 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6389 PR tree-optimization/83847
6390 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6391 group gathers and scatters.
6393 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6395 PR rtl-optimization/86620
6396 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6398 PR rtl-optimization/83213
6399 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6400 to last if both are JUMP_INSNs.
6402 PR tree-optimization/83843
6403 * gimple-ssa-store-merging.c
6404 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6405 store_immediate_info for bswap/nop orig_stores.
6407 2018-01-15 Andrew Waterman <andrew@sifive.com>
6409 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6410 !TARGET_MUL.
6411 <UDIV>: Increase cost if !TARGET_DIV.
6413 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6415 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6416 (define_attr "cr_logical_3op"): New.
6417 (cceq_ior_compare): Adjust.
6418 (cceq_ior_compare_complement): Adjust.
6419 (*cceq_rev_compare): Adjust.
6420 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6421 (is_cracked_insn): Adjust.
6422 (insn_must_be_first_in_group): Adjust.
6423 * config/rs6000/40x.md: Adjust.
6424 * config/rs6000/440.md: Adjust.
6425 * config/rs6000/476.md: Adjust.
6426 * config/rs6000/601.md: Adjust.
6427 * config/rs6000/603.md: Adjust.
6428 * config/rs6000/6xx.md: Adjust.
6429 * config/rs6000/7450.md: Adjust.
6430 * config/rs6000/7xx.md: Adjust.
6431 * config/rs6000/8540.md: Adjust.
6432 * config/rs6000/cell.md: Adjust.
6433 * config/rs6000/e300c2c3.md: Adjust.
6434 * config/rs6000/e500mc.md: Adjust.
6435 * config/rs6000/e500mc64.md: Adjust.
6436 * config/rs6000/e5500.md: Adjust.
6437 * config/rs6000/e6500.md: Adjust.
6438 * config/rs6000/mpc.md: Adjust.
6439 * config/rs6000/power4.md: Adjust.
6440 * config/rs6000/power5.md: Adjust.
6441 * config/rs6000/power6.md: Adjust.
6442 * config/rs6000/power7.md: Adjust.
6443 * config/rs6000/power8.md: Adjust.
6444 * config/rs6000/power9.md: Adjust.
6445 * config/rs6000/rs64.md: Adjust.
6446 * config/rs6000/titan.md: Adjust.
6448 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6450 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6451 ix86_indirect_branch_register logic.
6453 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6455 * config/i386/constraints.md (Bs): Update
6456 ix86_indirect_branch_register check. Don't check
6457 ix86_indirect_branch_register with GOT_memory_operand.
6458 (Bw): Likewise.
6459 * config/i386/predicates.md (GOT_memory_operand): Don't check
6460 ix86_indirect_branch_register here.
6461 (GOT32_symbol_operand): Likewise.
6463 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6465 * config/i386/predicates.md (constant_call_address_operand):
6466 Rewrite ix86_indirect_branch_register logic.
6467 (sibcall_insn_operand): Likewise.
6469 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6471 * config/i386/constraints.md (Bs): Replace
6472 ix86_indirect_branch_thunk_register with
6473 ix86_indirect_branch_register.
6474 (Bw): Likewise.
6475 * config/i386/i386.md (indirect_jump): Likewise.
6476 (tablejump): Likewise.
6477 (*sibcall_memory): Likewise.
6478 (*sibcall_value_memory): Likewise.
6479 Peepholes of indirect call and jump via memory: Likewise.
6480 * config/i386/i386.opt: Likewise.
6481 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6482 (GOT_memory_operand): Likewise.
6483 (call_insn_operand): Likewise.
6484 (sibcall_insn_operand): Likewise.
6485 (GOT32_symbol_operand): Likewise.
6487 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6489 PR middle-end/83837
6490 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
6491 type rather than type addr's type points to.
6492 (expand_omp_atomic_mutex): Likewise.
6493 (expand_omp_atomic): Likewise.
6495 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6497 PR target/83839
6498 * config/i386/i386.c (output_indirect_thunk_function): Use
6499 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6500 for __x86_return_thunk.
6502 2018-01-15 Richard Biener <rguenther@suse.de>
6504 PR middle-end/83850
6505 * expmed.c (extract_bit_field_1): Fix typo.
6507 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6509 PR target/83687
6510 * config/arm/iterators.md (VF): New mode iterator.
6511 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6512 Remove integer-related logic from pattern.
6513 (neon_vabd<mode>_3): Likewise.
6515 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6517 PR middle-end/82694
6518 * common.opt (fstrict-overflow): No longer an alias.
6519 (fwrapv-pointer): New option.
6520 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6521 also for pointer types based on flag_wrapv_pointer.
6522 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6523 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6524 opts->x_flag_wrapv got set.
6525 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6526 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6527 POINTER_TYPE_OVERFLOW_UNDEFINED.
6528 * match.pd: Likewise in address comparison pattern.
6529 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6531 2018-01-15 Richard Biener <rguenther@suse.de>
6533 PR lto/83804
6534 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6535 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6536 Reset type names to their identifier if their TYPE_DECL doesn't
6537 have linkage (and thus is used for ODR and devirt).
6538 (save_debug_info_for_decl): Remove.
6539 (save_debug_info_for_type): Likewise.
6540 (add_tree_to_fld_list): Adjust.
6541 * tree-pretty-print.c (dump_generic_node): Make dumping of
6542 type names more robust.
6544 2018-01-15 Richard Biener <rguenther@suse.de>
6546 * BASE-VER: Bump to 8.0.1.
6548 2018-01-14 Martin Sebor <msebor@redhat.com>
6550 PR other/83508
6551 * builtins.c (check_access): Avoid warning when the no-warning bit
6552 is set.
6554 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6556 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6557 * ira-color (allocno_hard_regs_compare): Likewise.
6559 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6561 PR target/83013
6562 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6563 Use .pushsection/.popsection.
6565 2018-01-14 Martin Sebor <msebor@redhat.com>
6567 PR c++/81327
6568 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6570 2018-01-14 Jakub Jelinek <jakub@redhat.com>
6572 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6573 entry from extra_headers.
6574 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6575 extra_headers, make the list bitwise identical to the i?86-*-* one.
6577 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6579 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6580 -mcmodel=large with -mindirect-branch=thunk,
6581 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6582 -mfunction-return=thunk-extern.
6583 * doc/invoke.texi: Document -mcmodel=large is incompatible with
6584 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6585 -mfunction-return=thunk and -mfunction-return=thunk-extern.
6587 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6589 * config/i386/i386.c (print_reg): Print the name of the full
6590 integer register without '%'.
6591 (ix86_print_operand): Handle 'V'.
6592 * doc/extend.texi: Document 'V' modifier.
6594 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6596 * config/i386/constraints.md (Bs): Disallow memory operand for
6597 -mindirect-branch-register.
6598 (Bw): Likewise.
6599 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6600 (GOT_memory_operand): Likewise.
6601 (call_insn_operand): Likewise.
6602 (sibcall_insn_operand): Likewise.
6603 (GOT32_symbol_operand): Likewise.
6604 * config/i386/i386.md (indirect_jump): Call convert_memory_address
6605 for -mindirect-branch-register.
6606 (tablejump): Likewise.
6607 (*sibcall_memory): Likewise.
6608 (*sibcall_value_memory): Likewise.
6609 Disallow peepholes of indirect call and jump via memory for
6610 -mindirect-branch-register.
6611 (*call_pop): Replace m with Bw.
6612 (*call_value_pop): Likewise.
6613 (*sibcall_pop_memory): Replace m with Bs.
6614 * config/i386/i386.opt (mindirect-branch-register): New option.
6615 * doc/invoke.texi: Document -mindirect-branch-register option.
6617 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6619 * config/i386/i386-protos.h (ix86_output_function_return): New.
6620 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6621 set function_return_type.
6622 (indirect_thunk_name): Add ret_p to indicate thunk for function
6623 return.
6624 (output_indirect_thunk_function): Pass false to
6625 indirect_thunk_name.
6626 (ix86_output_indirect_branch_via_reg): Likewise.
6627 (ix86_output_indirect_branch_via_push): Likewise.
6628 (output_indirect_thunk_function): Create alias for function
6629 return thunk if regno < 0.
6630 (ix86_output_function_return): New function.
6631 (ix86_handle_fndecl_attribute): Handle function_return.
6632 (ix86_attribute_table): Add function_return.
6633 * config/i386/i386.h (machine_function): Add
6634 function_return_type.
6635 * config/i386/i386.md (simple_return_internal): Use
6636 ix86_output_function_return.
6637 (simple_return_internal_long): Likewise.
6638 * config/i386/i386.opt (mfunction-return=): New option.
6639 (indirect_branch): Mention -mfunction-return=.
6640 * doc/extend.texi: Document function_return function attribute.
6641 * doc/invoke.texi: Document -mfunction-return= option.
6643 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6645 * config/i386/i386-opts.h (indirect_branch): New.
6646 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6647 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6648 with local indirect jump when converting indirect call and jump.
6649 (ix86_set_indirect_branch_type): New.
6650 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6651 (indirectlabelno): New.
6652 (indirect_thunk_needed): Likewise.
6653 (indirect_thunk_bnd_needed): Likewise.
6654 (indirect_thunks_used): Likewise.
6655 (indirect_thunks_bnd_used): Likewise.
6656 (INDIRECT_LABEL): Likewise.
6657 (indirect_thunk_name): Likewise.
6658 (output_indirect_thunk): Likewise.
6659 (output_indirect_thunk_function): Likewise.
6660 (ix86_output_indirect_branch_via_reg): Likewise.
6661 (ix86_output_indirect_branch_via_push): Likewise.
6662 (ix86_output_indirect_branch): Likewise.
6663 (ix86_output_indirect_jmp): Likewise.
6664 (ix86_code_end): Call output_indirect_thunk_function if needed.
6665 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6666 needed.
6667 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6668 (ix86_attribute_table): Add indirect_branch.
6669 * config/i386/i386.h (machine_function): Add indirect_branch_type
6670 and has_local_indirect_jump.
6671 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6672 to true.
6673 (tablejump): Likewise.
6674 (*indirect_jump): Use ix86_output_indirect_jmp.
6675 (*tablejump_1): Likewise.
6676 (simple_return_indirect_internal): Likewise.
6677 * config/i386/i386.opt (mindirect-branch=): New option.
6678 (indirect_branch): New.
6679 (keep): Likewise.
6680 (thunk): Likewise.
6681 (thunk-inline): Likewise.
6682 (thunk-extern): Likewise.
6683 * doc/extend.texi: Document indirect_branch function attribute.
6684 * doc/invoke.texi: Document -mindirect-branch= option.
6686 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6688 PR ipa/83051
6689 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6691 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6693 * ipa-inline.c (want_inline_small_function_p): Return false if
6694 inlining has already failed with CIF_FINAL_ERROR.
6695 (update_caller_keys): Call want_inline_small_function_p before
6696 can_inline_edge_p.
6697 (update_callee_keys): Likewise.
6699 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6701 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6702 New function.
6703 (rs6000_quadword_masked_address_p): Likewise.
6704 (quad_aligned_load_p): Likewise.
6705 (quad_aligned_store_p): Likewise.
6706 (const_load_sequence_p): Add comment to describe the outer-most loop.
6707 (mimic_memory_attributes_and_flags): New function.
6708 (rs6000_gen_stvx): Likewise.
6709 (replace_swapped_aligned_store): Likewise.
6710 (rs6000_gen_lvx): Likewise.
6711 (replace_swapped_aligned_load): Likewise.
6712 (replace_swapped_load_constant): Capitalize argument name in
6713 comment describing this function.
6714 (rs6000_analyze_swaps): Add a third pass to search for vector loads
6715 and stores that access quad-word aligned addresses and replace
6716 with stvx or lvx instructions when appropriate.
6717 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6718 New function prototype.
6719 (rs6000_quadword_masked_address_p): Likewise.
6720 (rs6000_gen_lvx): Likewise.
6721 (rs6000_gen_stvx): Likewise.
6722 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6723 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6724 when memory address is aligned.
6725 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6726 this split to select lvx instruction when memory address is aligned.
6727 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6728 instruction when memory address is aligned.
6729 (*vsx_le_perm_load_v16qi): Likewise.
6730 (four unnamed splitters): Modify to select the stvx instruction
6731 when memory is aligned.
6733 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
6735 * predict.c (determine_unlikely_bbs): Handle correctly BBs
6736 which appears in the queue multiple times.
6738 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6739 Alan Hayward <alan.hayward@arm.com>
6740 David Sherwood <david.sherwood@arm.com>
6742 * tree-vectorizer.h (vec_lower_bound): New structure.
6743 (_loop_vec_info): Add check_nonzero and lower_bounds.
6744 (LOOP_VINFO_CHECK_NONZERO): New macro.
6745 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6746 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6747 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6748 fields. Make seg_len the distance travelled, not including the
6749 access size.
6750 (dr_direction_indicator): Declare.
6751 (dr_zero_step_indicator): Likewise.
6752 (dr_known_forward_stride_p): Likewise.
6753 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
6754 tree-ssanames.h.
6755 (runtime_alias_check_p): Allow runtime alias checks with
6756 variable strides.
6757 (operator ==): Compare access_size and align.
6758 (prune_runtime_alias_test_list): Rework for new distinction between
6759 the access_size and seg_len.
6760 (create_intersect_range_checks_index): Likewise. Cope with polynomial
6761 segment lengths.
6762 (get_segment_min_max): New function.
6763 (create_intersect_range_checks): Use it.
6764 (dr_step_indicator): New function.
6765 (dr_direction_indicator): Likewise.
6766 (dr_zero_step_indicator): Likewise.
6767 (dr_known_forward_stride_p): Likewise.
6768 * tree-loop-distribution.c (data_ref_segment_size): Return
6769 DR_STEP * (niters - 1).
6770 (compute_alias_check_pairs): Update call to the dr_with_seg_len
6771 constructor.
6772 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
6773 (vect_preserves_scalar_order_p): New function, split out from...
6774 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
6775 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
6776 (vect_vfa_access_size): New function.
6777 (vect_vfa_align): Likewise.
6778 (vect_compile_time_alias): Take access_size_a and access_b arguments.
6779 (dump_lower_bound): New function.
6780 (vect_check_lower_bound): Likewise.
6781 (vect_small_gap_p): Likewise.
6782 (vectorizable_with_step_bound_p): Likewise.
6783 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
6784 depencies if the vectorization factor is 1. Convert the checks
6785 for nonzero steps into checks on the bounds of DR_STEP. Try using
6786 a bunds check for variable steps if the minimum required step is
6787 relatively small. Update calls to the dr_with_seg_len
6788 constructor and to vect_compile_time_alias.
6789 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
6790 function.
6791 (vect_loop_versioning): Call it.
6792 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
6793 when retrying.
6794 (vect_estimate_min_profitable_iters): Account for any bounds checks.
6796 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6797 Alan Hayward <alan.hayward@arm.com>
6798 David Sherwood <david.sherwood@arm.com>
6800 * doc/sourcebuild.texi (vect_scatter_store): Document.
6801 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
6802 optabs.
6803 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6804 Document.
6805 * genopinit.c (main): Add supports_vec_scatter_store and
6806 supports_vec_scatter_store_cached to target_optabs.
6807 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6808 IFN_MASK_SCATTER_STORE.
6809 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6810 functions.
6811 * internal-fn.h (internal_store_fn_p): Declare.
6812 (internal_fn_stored_value_index): Likewise.
6813 * internal-fn.c (scatter_store_direct): New macro.
6814 (expand_scatter_store_optab_fn): New function.
6815 (direct_scatter_store_optab_supported_p): New macro.
6816 (internal_store_fn_p): New function.
6817 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6818 IFN_MASK_SCATTER_STORE.
6819 (internal_fn_mask_index): Likewise.
6820 (internal_fn_stored_value_index): New function.
6821 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6822 for scatter stores.
6823 * optabs-query.h (supports_vec_scatter_store_p): Declare.
6824 * optabs-query.c (supports_vec_scatter_store_p): New function.
6825 * tree-vectorizer.h (vect_get_store_rhs): Declare.
6826 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6827 true for scatter stores.
6828 (vect_gather_scatter_fn_p): Handle scatter stores too.
6829 (vect_check_gather_scatter): Consider using scatter stores if
6830 supports_vec_scatter_store_p.
6831 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6832 scatter stores too.
6833 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6834 internal_fn_stored_value_index.
6835 (check_load_store_masking): Handle scatter stores too.
6836 (vect_get_store_rhs): Make public.
6837 (vectorizable_call): Use internal_store_fn_p.
6838 (vectorizable_store): Handle scatter store internal functions.
6839 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6840 when deciding whether the end of the group has been reached.
6841 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6842 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6843 (mask_scatter_store<mode>): New insns.
6845 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6846 Alan Hayward <alan.hayward@arm.com>
6847 David Sherwood <david.sherwood@arm.com>
6849 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6850 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6851 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6852 function.
6853 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6854 Use vect_truncate_gather_scatter_offset if we can't treat the
6855 operation as a normal gather load or scatter store.
6856 (get_group_load_store_type): Take the gather_scatter_info
6857 as argument. Try using a gather load or scatter store for
6858 single-element groups.
6859 (get_load_store_type): Update calls to get_group_load_store_type
6860 and vect_use_strided_gather_scatters_p.
6862 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6863 Alan Hayward <alan.hayward@arm.com>
6864 David Sherwood <david.sherwood@arm.com>
6866 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6867 optional tree argument.
6868 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6869 null target hooks.
6870 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6871 but continue to use the current value as a fallback.
6872 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6873 to compare the updates.
6874 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6875 (get_load_store_type): Use it when handling a strided access.
6876 (vect_get_strided_load_store_ops): New function.
6877 (vect_get_data_ptr_increment): Likewise.
6878 (vectorizable_load): Handle strided gather loads. Always pass
6879 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6881 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6882 Alan Hayward <alan.hayward@arm.com>
6883 David Sherwood <david.sherwood@arm.com>
6885 * doc/md.texi (gather_load@var{m}): Document.
6886 (mask_gather_load@var{m}): Likewise.
6887 * genopinit.c (main): Add supports_vec_gather_load and
6888 supports_vec_gather_load_cached to target_optabs.
6889 * optabs-tree.c (init_tree_optimization_optabs): Use
6890 ggc_cleared_alloc to allocate target_optabs.
6891 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6892 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6893 functions.
6894 * internal-fn.h (internal_load_fn_p): Declare.
6895 (internal_gather_scatter_fn_p): Likewise.
6896 (internal_fn_mask_index): Likewise.
6897 (internal_gather_scatter_fn_supported_p): Likewise.
6898 * internal-fn.c (gather_load_direct): New macro.
6899 (expand_gather_load_optab_fn): New function.
6900 (direct_gather_load_optab_supported_p): New macro.
6901 (direct_internal_fn_optab): New function.
6902 (internal_load_fn_p): Likewise.
6903 (internal_gather_scatter_fn_p): Likewise.
6904 (internal_fn_mask_index): Likewise.
6905 (internal_gather_scatter_fn_supported_p): Likewise.
6906 * optabs-query.c (supports_at_least_one_mode_p): New function.
6907 (supports_vec_gather_load_p): Likewise.
6908 * optabs-query.h (supports_vec_gather_load_p): Declare.
6909 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6910 and memory_type field.
6911 (NUM_PATTERNS): Bump to 15.
6912 * tree-vect-data-refs.c: Include internal-fn.h.
6913 (vect_gather_scatter_fn_p): New function.
6914 (vect_describe_gather_scatter_call): Likewise.
6915 (vect_check_gather_scatter): Try using internal functions for
6916 gather loads. Recognize existing calls to a gather load function.
6917 (vect_analyze_data_refs): Consider using gather loads if
6918 supports_vec_gather_load_p.
6919 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6920 (vect_get_gather_scatter_offset_type): Likewise.
6921 (vect_convert_mask_for_vectype): Likewise.
6922 (vect_add_conversion_to_patterm): Likewise.
6923 (vect_try_gather_scatter_pattern): Likewise.
6924 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6925 (vect_vect_recog_func_ptrs): Add it.
6926 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6927 internal_fn_mask_index and internal_gather_scatter_fn_p.
6928 (check_load_store_masking): Take the gather_scatter_info as an
6929 argument and handle gather loads.
6930 (vect_get_gather_scatter_ops): New function.
6931 (vectorizable_call): Check internal_load_fn_p.
6932 (vectorizable_load): Likewise. Handle gather load internal
6933 functions.
6934 (vectorizable_store): Update call to check_load_store_masking.
6935 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
6936 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
6937 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
6938 (aarch64_gather_scale_operand_d): New predicates.
6939 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
6940 (mask_gather_load<mode>): New insns.
6942 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6943 Alan Hayward <alan.hayward@arm.com>
6944 David Sherwood <david.sherwood@arm.com>
6946 * optabs.def (fold_left_plus_optab): New optab.
6947 * doc/md.texi (fold_left_plus_@var{m}): Document.
6948 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
6949 * internal-fn.c (fold_left_direct): Define.
6950 (expand_fold_left_optab_fn): Likewise.
6951 (direct_fold_left_optab_supported_p): Likewise.
6952 * fold-const-call.c (fold_const_fold_left): New function.
6953 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
6954 * tree-parloops.c (valid_reduction_p): New function.
6955 (gather_scalar_reductions): Use it.
6956 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
6957 (vect_finish_replace_stmt): Declare.
6958 * tree-vect-loop.c (fold_left_reduction_fn): New function.
6959 (needs_fold_left_reduction_p): New function, split out from...
6960 (vect_is_simple_reduction): ...here. Accept reductions that
6961 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
6962 (vect_force_simple_reduction): Also store the reduction type in
6963 the assignment's STMT_VINFO_REDUC_TYPE.
6964 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
6965 (merge_with_identity): New function.
6966 (vect_expand_fold_left): Likewise.
6967 (vectorize_fold_left_reduction): Likewise.
6968 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
6969 scalar phi in place for it. Check for target support and reject
6970 cases that would reassociate the operation. Defer the transform
6971 phase to vectorize_fold_left_reduction.
6972 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
6973 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
6974 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
6976 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6978 * tree-if-conv.c (predicate_mem_writes): Remove redundant
6979 call to ifc_temp_var.
6981 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6982 Alan Hayward <alan.hayward@arm.com>
6983 David Sherwood <david.sherwood@arm.com>
6985 * target.def (legitimize_address_displacement): Take the original
6986 offset as a poly_int.
6987 * targhooks.h (default_legitimize_address_displacement): Update
6988 accordingly.
6989 * targhooks.c (default_legitimize_address_displacement): Likewise.
6990 * doc/tm.texi: Regenerate.
6991 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
6992 as an argument, moving assert of ad->disp == ad->disp_term to...
6993 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
6994 Try calling targetm.legitimize_address_displacement before expanding
6995 the address rather than afterwards, and adjust for the new interface.
6996 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
6997 Match the new hook interface. Handle SVE addresses.
6998 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
6999 new hook interface.
7001 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7003 * Makefile.in (OBJS): Add early-remat.o.
7004 * target.def (select_early_remat_modes): New hook.
7005 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7006 * doc/tm.texi: Regenerate.
7007 * targhooks.h (default_select_early_remat_modes): Declare.
7008 * targhooks.c (default_select_early_remat_modes): New function.
7009 * timevar.def (TV_EARLY_REMAT): New timevar.
7010 * passes.def (pass_early_remat): New pass.
7011 * tree-pass.h (make_pass_early_remat): Declare.
7012 * early-remat.c: New file.
7013 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7014 function.
7015 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7017 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7018 Alan Hayward <alan.hayward@arm.com>
7019 David Sherwood <david.sherwood@arm.com>
7021 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7022 vfm1 with a bound_epilog parameter.
7023 (vect_do_peeling): Update calls accordingly, and move the prologue
7024 call earlier in the function. Treat the base bound_epilog as 0 for
7025 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7026 this base when peeling for gaps.
7027 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7028 with fully-masked loops.
7029 (vect_estimate_min_profitable_iters): Handle the single peeled
7030 iteration in that case.
7032 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7033 Alan Hayward <alan.hayward@arm.com>
7034 David Sherwood <david.sherwood@arm.com>
7036 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7037 single-element interleaving even if the size is not a power of 2.
7038 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7039 accesses for single-element interleaving if the group size is
7040 not a power of 2.
7042 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7043 Alan Hayward <alan.hayward@arm.com>
7044 David Sherwood <david.sherwood@arm.com>
7046 * doc/md.texi (fold_extract_last_@var{m}): Document.
7047 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7048 * optabs.def (fold_extract_last_optab): New optab.
7049 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7050 * internal-fn.c (fold_extract_direct): New macro.
7051 (expand_fold_extract_optab_fn): Likewise.
7052 (direct_fold_extract_optab_supported_p): Likewise.
7053 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7054 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7055 EXTRACT_LAST_REDUCTION.
7056 (get_initial_def_for_reduction): Do not create an initial vector
7057 for EXTRACT_LAST_REDUCTION reductions.
7058 (vectorizable_reduction): Leave the scalar phi in place for
7059 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7060 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7061 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7062 transform phase to vectorizable_condition.
7063 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7064 split out from...
7065 (vect_finish_stmt_generation): ...here.
7066 (vect_finish_replace_stmt): New function.
7067 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7068 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7069 pattern.
7070 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7072 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7073 Alan Hayward <alan.hayward@arm.com>
7074 David Sherwood <david.sherwood@arm.com>
7076 * doc/md.texi (extract_last_@var{m}): Document.
7077 * optabs.def (extract_last_optab): New optab.
7078 * internal-fn.def (EXTRACT_LAST): New internal function.
7079 * internal-fn.c (cond_unary_direct): New macro.
7080 (expand_cond_unary_optab_fn): Likewise.
7081 (direct_cond_unary_optab_supported_p): Likewise.
7082 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7083 loops using EXTRACT_LAST.
7084 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7085 (extract_last_<mode>): ...this optab.
7086 (vec_extract<mode><Vel>): Update accordingly.
7088 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7089 Alan Hayward <alan.hayward@arm.com>
7090 David Sherwood <david.sherwood@arm.com>
7092 * target.def (empty_mask_is_expensive): New hook.
7093 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7094 * doc/tm.texi: Regenerate.
7095 * targhooks.h (default_empty_mask_is_expensive): Declare.
7096 * targhooks.c (default_empty_mask_is_expensive): New function.
7097 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7098 if the target says that empty masks are expensive.
7099 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7100 New function.
7101 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7103 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7104 Alan Hayward <alan.hayward@arm.com>
7105 David Sherwood <david.sherwood@arm.com>
7107 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7108 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7109 (vect_use_loop_mask_for_alignment_p): New function.
7110 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7111 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7112 niters_skip argument. Make sure that the first niters_skip elements
7113 of the first iteration are inactive.
7114 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7115 Update call to vect_set_loop_masks_directly.
7116 (get_misalign_in_elems): New function, split out from...
7117 (vect_gen_prolog_loop_niters): ...here.
7118 (vect_update_init_of_dr): Take a code argument that specifies whether
7119 the adjustment should be added or subtracted.
7120 (vect_update_init_of_drs): Likewise.
7121 (vect_prepare_for_masked_peels): New function.
7122 (vect_do_peeling): Skip prologue peeling if we're using a mask
7123 instead. Update call to vect_update_inits_of_drs.
7124 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7125 mask_skip_niters.
7126 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7127 alignment. Do not include the number of peeled iterations in
7128 the minimum threshold in that case.
7129 (vectorizable_induction): Adjust the start value down by
7130 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7131 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7132 Take the number of skipped iterations into account when calculating
7133 the loop bounds.
7134 * tree-vect-stmts.c (vect_gen_while_not): New function.
7136 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7137 Alan Hayward <alan.hayward@arm.com>
7138 David Sherwood <david.sherwood@arm.com>
7140 * doc/sourcebuild.texi (vect_fully_masked): Document.
7141 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7142 default value to 0.
7143 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7144 split out from...
7145 (vect_analyze_loop_2): ...here. Don't check the vectorization
7146 factor against the number of loop iterations if the loop is
7147 fully-masked.
7149 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7150 Alan Hayward <alan.hayward@arm.com>
7151 David Sherwood <david.sherwood@arm.com>
7153 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7154 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7155 (dump_groups): Update accordingly.
7156 (iv_use::mem_type): New member variable.
7157 (address_p): New function.
7158 (record_use): Add a mem_type argument and initialize the new
7159 mem_type field.
7160 (record_group_use): Add a mem_type argument. Use address_p.
7161 Remove obsolete null checks of base_object. Update call to record_use.
7162 (find_interesting_uses_op): Update call to record_group_use.
7163 (find_interesting_uses_cond): Likewise.
7164 (find_interesting_uses_address): Likewise.
7165 (get_mem_type_for_internal_fn): New function.
7166 (find_address_like_use): Likewise.
7167 (find_interesting_uses_stmt): Try find_address_like_use before
7168 calling find_interesting_uses_op.
7169 (addr_offset_valid_p): Use the iv mem_type field as the type
7170 of the addressed memory.
7171 (add_autoinc_candidates): Likewise.
7172 (get_address_cost): Likewise.
7173 (split_small_address_groups_p): Use address_p.
7174 (split_address_groups): Likewise.
7175 (add_iv_candidate_for_use): Likewise.
7176 (autoinc_possible_for_pair): Likewise.
7177 (rewrite_groups): Likewise.
7178 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7179 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7180 (get_alias_ptr_type_for_ptr_address): New function.
7181 (rewrite_use_address): Rewrite address uses in calls that were
7182 identified by find_address_like_use.
7184 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7185 Alan Hayward <alan.hayward@arm.com>
7186 David Sherwood <david.sherwood@arm.com>
7188 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7189 TARGET_MEM_REFs.
7190 * gimple-expr.h (is_gimple_addressable: Likewise.
7191 * gimple-expr.c (is_gimple_address): Likewise.
7192 * internal-fn.c (expand_call_mem_ref): New function.
7193 (expand_mask_load_optab_fn): Use it.
7194 (expand_mask_store_optab_fn): Likewise.
7196 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7197 Alan Hayward <alan.hayward@arm.com>
7198 David Sherwood <david.sherwood@arm.com>
7200 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7201 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7202 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7203 (cond_umax@var{mode}): Document.
7204 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7205 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7206 (cond_umin_optab, cond_umax_optab): New optabs.
7207 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7208 (COND_IOR, COND_XOR): New internal functions.
7209 * internal-fn.h (get_conditional_internal_fn): Declare.
7210 * internal-fn.c (cond_binary_direct): New macro.
7211 (expand_cond_binary_optab_fn): Likewise.
7212 (direct_cond_binary_optab_supported_p): Likewise.
7213 (get_conditional_internal_fn): New function.
7214 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7215 Cope with reduction statements that are vectorized as calls rather
7216 than assignments.
7217 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7218 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7219 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7220 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7221 (UNSPEC_COND_EOR): New unspecs.
7222 (optab): Add mappings for them.
7223 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7224 (sve_int_op, sve_fp_op): New int attributes.
7226 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7227 Alan Hayward <alan.hayward@arm.com>
7228 David Sherwood <david.sherwood@arm.com>
7230 * optabs.def (while_ult_optab): New optab.
7231 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7232 * internal-fn.def (WHILE_ULT): New internal function.
7233 * internal-fn.h (direct_internal_fn_supported_p): New override
7234 that takes two types as argument.
7235 * internal-fn.c (while_direct): New macro.
7236 (expand_while_optab_fn): New function.
7237 (convert_optab_supported_p): Likewise.
7238 (direct_while_optab_supported_p): New macro.
7239 * wide-int.h (wi::udiv_ceil): New function.
7240 * tree-vectorizer.h (rgroup_masks): New structure.
7241 (vec_loop_masks): New typedef.
7242 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7243 and fully_masked_p.
7244 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7245 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7246 (vect_max_vf): New function.
7247 (slpeel_make_loop_iterate_ntimes): Delete.
7248 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7249 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7250 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7251 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7252 internal-fn.h, stor-layout.h and optabs-query.h.
7253 (vect_set_loop_mask): New function.
7254 (add_preheader_seq): Likewise.
7255 (add_header_seq): Likewise.
7256 (interleave_supported_p): Likewise.
7257 (vect_maybe_permute_loop_masks): Likewise.
7258 (vect_set_loop_masks_directly): Likewise.
7259 (vect_set_loop_condition_masked): Likewise.
7260 (vect_set_loop_condition_unmasked): New function, split out from
7261 slpeel_make_loop_iterate_ntimes.
7262 (slpeel_make_loop_iterate_ntimes): Rename to..
7263 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7264 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7265 (vect_do_peeling): Update call accordingly.
7266 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7267 loops.
7268 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7269 mask_compare_type, can_fully_mask_p and fully_masked_p.
7270 (release_vec_loop_masks): New function.
7271 (_loop_vec_info): Use it to free the loop masks.
7272 (can_produce_all_loop_masks_p): New function.
7273 (vect_get_max_nscalars_per_iter): Likewise.
7274 (vect_verify_full_masking): Likewise.
7275 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7276 retries, and free the mask rgroups before retrying. Check loop-wide
7277 reasons for disallowing fully-masked loops. Make the final decision
7278 about whether use a fully-masked loop or not.
7279 (vect_estimate_min_profitable_iters): Do not assume that peeling
7280 for the number of iterations will be needed for fully-masked loops.
7281 (vectorizable_reduction): Disable fully-masked loops.
7282 (vectorizable_live_operation): Likewise.
7283 (vect_halve_mask_nunits): New function.
7284 (vect_double_mask_nunits): Likewise.
7285 (vect_record_loop_mask): Likewise.
7286 (vect_get_loop_mask): Likewise.
7287 (vect_transform_loop): Handle the case in which the final loop
7288 iteration might handle a partial vector. Call vect_set_loop_condition
7289 instead of slpeel_make_loop_iterate_ntimes.
7290 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7291 (check_load_store_masking): New function.
7292 (prepare_load_store_mask): Likewise.
7293 (vectorizable_store): Handle fully-masked loops.
7294 (vectorizable_load): Likewise.
7295 (supportable_widening_operation): Use vect_halve_mask_nunits for
7296 booleans.
7297 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7298 (vect_gen_while): New function.
7299 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7300 (aarch64_uqdec<mode>): New insn.
7302 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7303 Alan Hayward <alan.hayward@arm.com>
7304 David Sherwood <david.sherwood@arm.com>
7306 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7307 (reduc_xor_scal_optab): New optabs.
7308 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7309 (reduc_xor_scal_@var{m}): Document.
7310 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7311 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7312 internal functions.
7313 * fold-const-call.c (fold_const_call): Handle them.
7314 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7315 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7316 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7317 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7318 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7319 (UNSPEC_XORV): New unspecs.
7320 (optab): Add entries for them.
7321 (BITWISEV): New int iterator.
7322 (bit_reduc_op): New int attributes.
7324 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7325 Alan Hayward <alan.hayward@arm.com>
7326 David Sherwood <david.sherwood@arm.com>
7328 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7329 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7330 * optabs.def (vec_shl_insert_optab): New optab.
7331 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7332 (duplicate_and_interleave): Likewise.
7333 * tree-vect-loop.c: Include internal-fn.h.
7334 (neutral_op_for_slp_reduction): New function, split out from
7335 get_initial_defs_for_reduction.
7336 (get_initial_def_for_reduction): Handle option 2 for variable-length
7337 vectors by loading the neutral value into a vector and then shifting
7338 the initial value into element 0.
7339 (get_initial_defs_for_reduction): Replace the code argument with
7340 the neutral value calculated by neutral_op_for_slp_reduction.
7341 Use gimple_build_vector for constant-length vectors.
7342 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7343 but the first group_size elements have a neutral value.
7344 Use duplicate_and_interleave otherwise.
7345 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7346 Update call to get_initial_defs_for_reduction. Handle SLP
7347 reductions for variable-length vectors by creating one vector
7348 result for each scalar result, with the elements associated
7349 with other scalar results stubbed out with the neutral value.
7350 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7351 Require IFN_VEC_SHL_INSERT for double reductions on
7352 variable-length vectors, or SLP reductions that have
7353 a neutral value. Require can_duplicate_and_interleave_p
7354 support for variable-length unchained SLP reductions if there
7355 is no neutral value, such as for MIN/MAX reductions. Also require
7356 the number of vector elements to be a multiple of the number of
7357 SLP statements when doing variable-length unchained SLP reductions.
7358 Update call to vect_create_epilog_for_reduction.
7359 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7360 and remove initial values.
7361 (duplicate_and_interleave): Make public.
7362 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7363 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7365 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7366 Alan Hayward <alan.hayward@arm.com>
7367 David Sherwood <david.sherwood@arm.com>
7369 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7370 (can_duplicate_and_interleave_p): New function.
7371 (vect_get_and_check_slp_defs): Take the vector of statements
7372 rather than just the current one. Remove excess parentheses.
7373 Restriction rejectinon of vect_constant_def and vect_external_def
7374 for variable-length vectors to boolean types, or types for which
7375 can_duplicate_and_interleave_p is false.
7376 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7377 (duplicate_and_interleave): New function.
7378 (vect_get_constant_vectors): Use gimple_build_vector for
7379 constant-length vectors and suitable variable-length constant
7380 vectors. Use duplicate_and_interleave for other variable-length
7381 vectors. Don't defer the update when inserting new statements.
7383 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7384 Alan Hayward <alan.hayward@arm.com>
7385 David Sherwood <david.sherwood@arm.com>
7387 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7388 min_profitable_iters doesn't go negative.
7390 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7391 Alan Hayward <alan.hayward@arm.com>
7392 David Sherwood <david.sherwood@arm.com>
7394 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7395 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7396 * optabs.def (vec_mask_load_lanes_optab): New optab.
7397 (vec_mask_store_lanes_optab): Likewise.
7398 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7399 (MASK_STORE_LANES): Likewise.
7400 * internal-fn.c (mask_load_lanes_direct): New macro.
7401 (mask_store_lanes_direct): Likewise.
7402 (expand_mask_load_optab_fn): Handle masked operations.
7403 (expand_mask_load_lanes_optab_fn): New macro.
7404 (expand_mask_store_optab_fn): Handle masked operations.
7405 (expand_mask_store_lanes_optab_fn): New macro.
7406 (direct_mask_load_lanes_optab_supported_p): Likewise.
7407 (direct_mask_store_lanes_optab_supported_p): Likewise.
7408 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7409 parameter.
7410 (vect_load_lanes_supported): Likewise.
7411 * tree-vect-data-refs.c (strip_conversion): New function.
7412 (can_group_stmts_p): Likewise.
7413 (vect_analyze_data_ref_accesses): Use it instead of checking
7414 for a pair of assignments.
7415 (vect_store_lanes_supported): Take a masked_p parameter.
7416 (vect_load_lanes_supported): Likewise.
7417 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7418 vect_store_lanes_supported and vect_load_lanes_supported.
7419 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7420 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7421 parameter. Don't allow gaps for masked accesses.
7422 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7423 and vect_load_lanes_supported.
7424 (get_load_store_type): Take a masked_p parameter and update
7425 call to get_group_load_store_type.
7426 (vectorizable_store): Update call to get_load_store_type.
7427 Handle IFN_MASK_STORE_LANES.
7428 (vectorizable_load): Update call to get_load_store_type.
7429 Handle IFN_MASK_LOAD_LANES.
7431 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7432 Alan Hayward <alan.hayward@arm.com>
7433 David Sherwood <david.sherwood@arm.com>
7435 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7436 modes for SVE.
7437 * config/aarch64/aarch64-protos.h
7438 (aarch64_sve_struct_memory_operand_p): Declare.
7439 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7440 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7441 (VPRED, vpred): Handle SVE structure modes.
7442 * config/aarch64/constraints.md (Utx): New constraint.
7443 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7444 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7445 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7446 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7447 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7448 structure modes. Split into pieces after RA.
7449 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7450 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7451 New patterns.
7452 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7453 SVE structure modes.
7454 (aarch64_classify_address): Likewise.
7455 (sizetochar): Move earlier in file.
7456 (aarch64_print_operand): Handle SVE register lists.
7457 (aarch64_array_mode): New function.
7458 (aarch64_sve_struct_memory_operand_p): Likewise.
7459 (TARGET_ARRAY_MODE): Redefine.
7461 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7462 Alan Hayward <alan.hayward@arm.com>
7463 David Sherwood <david.sherwood@arm.com>
7465 * target.def (array_mode): New target hook.
7466 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
7467 * doc/tm.texi: Regenerate.
7468 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
7469 * hooks.c (hook_optmode_mode_uhwi_none): New function.
7470 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
7471 targetm.array_mode.
7472 * stor-layout.c (mode_for_array): Likewise. Support polynomial
7473 type sizes.
7475 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7476 Alan Hayward <alan.hayward@arm.com>
7477 David Sherwood <david.sherwood@arm.com>
7479 * fold-const.c (fold_binary_loc): Check the argument types
7480 rather than the result type when testing for a vector operation.
7482 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7484 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
7485 * doc/tm.texi: Regenerate.
7487 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7488 Alan Hayward <alan.hayward@arm.com>
7489 David Sherwood <david.sherwood@arm.com>
7491 * doc/invoke.texi (-msve-vector-bits=): Document new option.
7492 (sve): Document new AArch64 extension.
7493 * doc/md.texi (w): Extend the description of the AArch64
7494 constraint to include SVE vectors.
7495 (Upl, Upa): Document new AArch64 predicate constraints.
7496 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7497 enum.
7498 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7499 (msve-vector-bits=): New option.
7500 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7501 SVE when these are disabled.
7502 (sve): New extension.
7503 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7504 modes. Adjust their number of units based on aarch64_sve_vg.
7505 (MAX_BITSIZE_MODE_ANY_MODE): Define.
7506 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7507 aarch64_addr_query_type.
7508 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7509 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7510 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7511 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7512 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7513 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7514 (aarch64_simd_imm_zero_p): Delete.
7515 (aarch64_check_zero_based_sve_index_immediate): Declare.
7516 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7517 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7518 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7519 (aarch64_sve_float_mul_immediate_p): Likewise.
7520 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7521 rather than an rtx.
7522 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7523 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7524 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7525 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7526 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7527 (aarch64_regmode_natural_size): Likewise.
7528 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7529 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7530 left one place.
7531 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7532 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7533 for VG and the SVE predicate registers.
7534 (V_ALIASES): Add a "z"-prefixed alias.
7535 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7536 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7537 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7538 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7539 (REG_CLASS_NAMES): Add entries for them.
7540 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7541 and the predicate registers.
7542 (aarch64_sve_vg): Declare.
7543 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7544 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7545 (REGMODE_NATURAL_SIZE): Define.
7546 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7547 SVE macros.
7548 * config/aarch64/aarch64.c: Include cfgrtl.h.
7549 (simd_immediate_info): Add a constructor for series vectors,
7550 and an associated step field.
7551 (aarch64_sve_vg): New variable.
7552 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7553 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7554 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7555 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7556 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7557 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7558 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7559 (aarch64_get_mask_mode): New functions.
7560 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7561 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7562 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7563 predicate modes and predicate registers. Explicitly restrict
7564 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7565 to store a vector mode if it is recognized by
7566 aarch64_classify_vector_mode.
7567 (aarch64_regmode_natural_size): New function.
7568 (aarch64_hard_regno_caller_save_mode): Return the original mode
7569 for predicates.
7570 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7571 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7572 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7573 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7574 functions.
7575 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
7576 does not overlap dest if the function is frame-related. Handle
7577 SVE constants.
7578 (aarch64_split_add_offset): New function.
7579 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7580 them aarch64_add_offset.
7581 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7582 and update call to aarch64_sub_sp.
7583 (aarch64_add_cfa_expression): New function.
7584 (aarch64_expand_prologue): Pass extra temporary registers to the
7585 functions above. Handle the case in which we need to emit new
7586 DW_CFA_expressions for registers that were originally saved
7587 relative to the stack pointer, but now have to be expressed
7588 relative to the frame pointer.
7589 (aarch64_output_mi_thunk): Pass extra temporary registers to the
7590 functions above.
7591 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
7592 IP0 and IP1 values for SVE frames.
7593 (aarch64_expand_vec_series): New function.
7594 (aarch64_expand_sve_widened_duplicate): Likewise.
7595 (aarch64_expand_sve_const_vector): Likewise.
7596 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7597 Handle SVE constants. Use emit_move_insn to move a force_const_mem
7598 into the register, rather than emitting a SET directly.
7599 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7600 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7601 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7602 (offset_9bit_signed_scaled_p): New functions.
7603 (aarch64_replicate_bitmask_imm): New function.
7604 (aarch64_bitmask_imm): Use it.
7605 (aarch64_cannot_force_const_mem): Reject expressions involving
7606 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
7607 (aarch64_classify_index): Handle SVE indices, by requiring
7608 a plain register index with a scale that matches the element size.
7609 (aarch64_classify_address): Handle SVE addresses. Assert that
7610 the mode of the address is VOIDmode or an integer mode.
7611 Update call to aarch64_classify_symbol.
7612 (aarch64_classify_symbolic_expression): Update call to
7613 aarch64_classify_symbol.
7614 (aarch64_const_vec_all_in_range_p): New function.
7615 (aarch64_print_vector_float_operand): Likewise.
7616 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
7617 "vN" for FP registers with SVE modes. Handle (const ...) vectors
7618 and the FP immediates 1.0 and 0.5.
7619 (aarch64_print_address_internal): Handle SVE addresses.
7620 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7621 (aarch64_regno_regclass): Handle predicate registers.
7622 (aarch64_secondary_reload): Handle big-endian reloads of SVE
7623 data modes.
7624 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7625 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7626 (aarch64_convert_sve_vector_bits): New function.
7627 (aarch64_override_options): Use it to handle -msve-vector-bits=.
7628 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7629 rather than an rtx.
7630 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7631 Handle SVE vector and predicate modes. Accept VL-based constants
7632 that need only one temporary register, and VL offsets that require
7633 no temporary registers.
7634 (aarch64_conditional_register_usage): Mark the predicate registers
7635 as fixed if SVE isn't available.
7636 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7637 Return true for SVE vector and predicate modes.
7638 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7639 rather than an unsigned int. Handle SVE modes.
7640 (aarch64_preferred_simd_mode): Update call accordingly. Handle
7641 SVE modes.
7642 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7643 if SVE is enabled.
7644 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7645 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7646 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7647 (aarch64_sve_float_mul_immediate_p): New functions.
7648 (aarch64_sve_valid_immediate): New function.
7649 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7650 Explicitly reject structure modes. Check for INDEX constants.
7651 Handle PTRUE and PFALSE constants.
7652 (aarch64_check_zero_based_sve_index_immediate): New function.
7653 (aarch64_simd_imm_zero_p): Delete.
7654 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7655 vector modes. Accept constants in the range of CNT[BHWD].
7656 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7657 ask for an Advanced SIMD mode.
7658 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7659 (aarch64_simd_vector_alignment): Handle SVE predicates.
7660 (aarch64_vectorize_preferred_vector_alignment): New function.
7661 (aarch64_simd_vector_alignment_reachable): Use it instead of
7662 the vector size.
7663 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7664 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7665 functions.
7666 (MAX_VECT_LEN): Delete.
7667 (expand_vec_perm_d): Add a vec_flags field.
7668 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7669 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7670 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7671 for SVE modes.
7672 (aarch64_evpc_rev): Rename to...
7673 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7674 (aarch64_evpc_rev_global): New function.
7675 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7676 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7677 MAX_VECT_LEN.
7678 (aarch64_evpc_sve_tbl): New function.
7679 (aarch64_expand_vec_perm_const_1): Update after rename of
7680 aarch64_evpc_rev. Handle SVE permutes too, trying
7681 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7682 than aarch64_evpc_tbl.
7683 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7684 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7685 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7686 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7687 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7688 (aarch64_expand_sve_vcond): New functions.
7689 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7690 of aarch64_vector_mode_p.
7691 (aarch64_dwarf_poly_indeterminate_value): New function.
7692 (aarch64_compute_pressure_classes): Likewise.
7693 (aarch64_can_change_mode_class): Likewise.
7694 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7695 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7696 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7697 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7698 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7699 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7700 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7701 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7702 constraints.
7703 (Dn, Dl, Dr): Accept const as well as const_vector.
7704 (Dz): Likewise. Compare against CONST0_RTX.
7705 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7706 of "vector" where appropriate.
7707 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7708 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7709 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7710 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7711 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7712 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7713 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7714 (v_int_equiv): Extend to SVE modes.
7715 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7716 mode attributes.
7717 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7718 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7719 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7720 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7721 (SVE_COND_FP_CMP): New int iterators.
7722 (perm_hilo): Handle the new unpack unspecs.
7723 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7724 attributes.
7725 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7726 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7727 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7728 (aarch64_equality_operator, aarch64_constant_vector_operand)
7729 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7730 (aarch64_sve_nonimmediate_operand): Likewise.
7731 (aarch64_sve_general_operand): Likewise.
7732 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7733 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7734 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7735 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7736 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7737 (aarch64_sve_float_arith_immediate): Likewise.
7738 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7739 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7740 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7741 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7742 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7743 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7744 (aarch64_sve_float_arith_operand): Likewise.
7745 (aarch64_sve_float_arith_with_sub_operand): Likewise.
7746 (aarch64_sve_float_mul_operand): Likewise.
7747 (aarch64_sve_vec_perm_operand): Likewise.
7748 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7749 (aarch64_mov_operand): Accept const_poly_int and const_vector.
7750 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7751 as well as const_vector.
7752 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
7753 in file. Use CONST0_RTX and CONSTM1_RTX.
7754 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
7755 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
7756 Use aarch64_simd_imm_zero.
7757 * config/aarch64/aarch64-sve.md: New file.
7758 * config/aarch64/aarch64.md: Include it.
7759 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
7760 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
7761 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
7762 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
7763 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
7764 (sve): New attribute.
7765 (enabled): Disable instructions with the sve attribute unless
7766 TARGET_SVE.
7767 (movqi, movhi): Pass CONST_POLY_INT operaneds through
7768 aarch64_expand_mov_immediate.
7769 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
7770 CNT[BHSD] immediates.
7771 (movti): Split CONST_POLY_INT moves into two halves.
7772 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
7773 Split additions that need a temporary here if the destination
7774 is the stack pointer.
7775 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
7776 (*add<mode>3_poly_1): New instruction.
7777 (set_clobber_cc): New expander.
7779 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7781 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
7782 parameter and use it instead of GET_MODE_SIZE (innermode). Use
7783 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
7784 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
7785 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
7786 Change innermode from fixed_mode_size to machine_mode.
7787 (simplify_subreg): Update call accordingly. Handle a constant-sized
7788 subreg of a variable-length CONST_VECTOR.
7790 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7791 Alan Hayward <alan.hayward@arm.com>
7792 David Sherwood <david.sherwood@arm.com>
7794 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
7795 (add_offset_to_base): New function, split out from...
7796 (create_mem_ref): ...here. When handling a scale other than 1,
7797 check first whether the address is valid without the offset.
7798 Add it into the base if so, leaving the index and scale as-is.
7800 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7802 PR c++/83778
7803 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7804 fold_for_warn before checking if arg2 is INTEGER_CST.
7806 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
7808 * config/rs6000/predicates.md (load_multiple_operation): Delete.
7809 (store_multiple_operation): Delete.
7810 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7811 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7812 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7813 guarded by TARGET_STRING.
7814 (rs6000_output_load_multiple): Delete.
7815 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7816 OPTION_MASK_STRING / TARGET_STRING handling.
7817 (print_operand) <'N', 'O'>: Add comment that these are unused now.
7818 (const rs6000_opt_masks) <"string">: Change mask to 0.
7819 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7820 (MASK_STRING): Delete.
7821 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7822 parts. Simplify.
7823 (load_multiple): Delete.
7824 (*ldmsi8): Delete.
7825 (*ldmsi7): Delete.
7826 (*ldmsi6): Delete.
7827 (*ldmsi5): Delete.
7828 (*ldmsi4): Delete.
7829 (*ldmsi3): Delete.
7830 (store_multiple): Delete.
7831 (*stmsi8): Delete.
7832 (*stmsi7): Delete.
7833 (*stmsi6): Delete.
7834 (*stmsi5): Delete.
7835 (*stmsi4): Delete.
7836 (*stmsi3): Delete.
7837 (movmemsi_8reg): Delete.
7838 (corresponding unnamed define_insn): Delete.
7839 (movmemsi_6reg): Delete.
7840 (corresponding unnamed define_insn): Delete.
7841 (movmemsi_4reg): Delete.
7842 (corresponding unnamed define_insn): Delete.
7843 (movmemsi_2reg): Delete.
7844 (corresponding unnamed define_insn): Delete.
7845 (movmemsi_1reg): Delete.
7846 (corresponding unnamed define_insn): Delete.
7847 * config/rs6000/rs6000.opt (mno-string): New.
7848 (mstring): Replace by deprecation warning stub.
7849 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7851 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7853 * regrename.c (regrename_do_replace): If replacing the same
7854 reg multiple times, try to reuse last created gen_raw_REG.
7856 PR debug/81155
7857 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7858 main to workaround a bug in GDB.
7860 2018-01-12 Tom de Vries <tom@codesourcery.com>
7862 PR target/83737
7863 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7865 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7867 PR rtl-optimization/80481
7868 * ira-color.c (get_cap_member): New function.
7869 (allocnos_conflict_by_live_ranges_p): Use it.
7870 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7871 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7873 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7875 PR target/83628
7876 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7877 (*saddl_se_1): Ditto.
7878 (*ssubsi_1): Ditto.
7879 (*ssubl_se_1): Ditto.
7881 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7883 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7884 rather than wi::to_widest for DR_INITs.
7885 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7886 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7887 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7888 INTEGER_CSTs.
7889 (vect_analyze_group_access_1): Note that here.
7891 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7893 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7894 polynomial type sizes.
7896 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7898 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7899 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7900 (gimple_add_tmp_var): Likewise.
7902 2018-01-12 Martin Liska <mliska@suse.cz>
7904 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7905 (gimple_alloc_sizes): Likewise.
7906 (dump_gimple_statistics): Use PRIu64 in printf format.
7907 * gimple.h: Change uint64_t to int.
7909 2018-01-12 Martin Liska <mliska@suse.cz>
7911 * tree-core.h: Use uint64_t instead of int.
7912 * tree.c (tree_node_counts): Likewise.
7913 (tree_node_sizes): Likewise.
7914 (dump_tree_statistics): Use PRIu64 in printf format.
7916 2018-01-12 Martin Liska <mliska@suse.cz>
7918 * Makefile.in: As qsort_chk is implemented in vec.c, add
7919 vec.o to linkage of gencfn-macros.
7920 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7921 passing the info to record_node_allocation_statistics.
7922 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7923 and pass the info.
7924 * ggc-common.c (struct ggc_usage): Add operator== and use
7925 it in operator< and compare function.
7926 * mem-stats.h (struct mem_usage): Likewise.
7927 * vec.c (struct vec_usage): Remove operator< and compare
7928 function. Can be simply inherited.
7930 2018-01-12 Martin Jambor <mjambor@suse.cz>
7932 PR target/81616
7933 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7934 * tree-ssa-math-opts.c: Include domwalk.h.
7935 (convert_mult_to_fma_1): New function.
7936 (fma_transformation_info): New type.
7937 (fma_deferring_state): Likewise.
7938 (cancel_fma_deferring): New function.
7939 (result_of_phi): Likewise.
7940 (last_fma_candidate_feeds_initial_phi): Likewise.
7941 (convert_mult_to_fma): Added deferring logic, split actual
7942 transformation to convert_mult_to_fma_1.
7943 (math_opts_dom_walker): New type.
7944 (math_opts_dom_walker::after_dom_children): New method, body moved
7945 here from pass_optimize_widening_mul::execute, added deferring logic
7946 bits.
7947 (pass_optimize_widening_mul::execute): Moved most of code to
7948 math_opts_dom_walker::after_dom_children.
7949 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
7950 * config/i386/i386.c (ix86_option_override_internal): Added
7951 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
7953 2018-01-12 Richard Biener <rguenther@suse.de>
7955 PR debug/83157
7956 * dwarf2out.c (gen_variable_die): Do not reset old_die for
7957 inline instance vars.
7959 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
7961 PR target/81819
7962 * config/rx/rx.c (rx_is_restricted_memory_address):
7963 Handle SUBREG case.
7965 2018-01-12 Richard Biener <rguenther@suse.de>
7967 PR tree-optimization/80846
7968 * target.def (split_reduction): New target hook.
7969 * targhooks.c (default_split_reduction): New function.
7970 * targhooks.h (default_split_reduction): Declare.
7971 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
7972 target requests first reduce vectors by combining low and high
7973 parts.
7974 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
7975 (get_vectype_for_scalar_type_and_size): Export.
7976 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
7977 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
7978 * doc/tm.texi: Regenerate.
7979 * config/i386/i386.c (ix86_split_reduction): Implement
7980 TARGET_VECTORIZE_SPLIT_REDUCTION.
7982 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7984 PR target/83368
7985 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
7986 in PIC mode except for TARGET_VXWORKS_RTP.
7987 * config/sparc/sparc.c: Include cfgrtl.h.
7988 (TARGET_INIT_PIC_REG): Define.
7989 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
7990 (sparc_pic_register_p): New predicate.
7991 (sparc_legitimate_address_p): Use it.
7992 (sparc_legitimize_pic_address): Likewise.
7993 (sparc_delegitimize_address): Likewise.
7994 (sparc_mode_dependent_address_p): Likewise.
7995 (gen_load_pcrel_sym): Remove 4th parameter.
7996 (load_got_register): Adjust call to above. Remove obsolete stuff.
7997 (sparc_expand_prologue): Do not call load_got_register here.
7998 (sparc_flat_expand_prologue): Likewise.
7999 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8000 (sparc_use_pseudo_pic_reg): New function.
8001 (sparc_init_pic_reg): Likewise.
8002 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8003 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8005 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8007 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8008 Add item for branch_cost.
8010 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8012 PR rtl-optimization/83565
8013 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8014 not extend the result to a larger mode for rotate operations.
8015 (num_sign_bit_copies1): Likewise.
8017 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8019 PR target/40411
8020 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8021 -symbolic.
8022 Use values-Xc.o for -pedantic.
8023 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8025 2018-01-12 Martin Liska <mliska@suse.cz>
8027 PR ipa/83054
8028 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8029 New function.
8030 (possible_polymorphic_call_targets): Use it.
8031 (ipa_devirt): Likewise.
8033 2018-01-12 Martin Liska <mliska@suse.cz>
8035 * profile-count.h (enum profile_quality): Use 0 as invalid
8036 enum value of profile_quality.
8038 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8040 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8041 -mext-string options.
8043 2018-01-12 Richard Biener <rguenther@suse.de>
8045 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8046 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8047 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8048 Likewise.
8049 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8051 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8053 * configure.ac (--with-long-double-format): Add support for the
8054 configuration option to change the default long double format on
8055 PowerPC systems.
8056 * config.gcc (powerpc*-linux*-*): Likewise.
8057 * configure: Regenerate.
8058 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8059 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8060 used without modification.
8062 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8064 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8065 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8066 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8067 MISC_BUILTIN_SPEC_BARRIER.
8068 (rs6000_init_builtins): Likewise.
8069 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8070 enum value.
8071 (speculation_barrier): New define_insn.
8072 * doc/extend.texi: Document __builtin_speculation_barrier.
8074 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8076 PR target/83203
8077 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8078 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8079 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8080 iterators.
8081 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8082 integral modes instead of "ss" and "sd".
8083 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8084 vectors with 32-bit and 64-bit elements.
8085 (vecdupssescalarmodesuffix): New mode attribute.
8086 (vec_dup<mode>): Use it.
8088 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8090 PR target/83330
8091 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8092 frame if argument is passed on stack.
8094 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8096 PR target/82682
8097 * ree.c (combine_reaching_defs): Optimize also
8098 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8099 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8101 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8103 PR middle-end/83189
8104 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8106 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8108 PR middle-end/83718
8109 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8110 after they are computed.
8112 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8114 PR tree-optimization/83695
8115 * gimple-loop-linterchange.cc
8116 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8117 reset cached scev information after interchange.
8118 (pass_linterchange::execute): Remove call to scev_reset_htab.
8120 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8122 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8123 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8124 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8125 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8126 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8127 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8128 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8129 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8130 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8131 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8132 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8133 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8134 (V_lane_reg): Likewise.
8135 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8136 New define_expand.
8137 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8138 (vfmal_lane_low<mode>_intrinsic,
8139 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8140 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8141 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8142 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8143 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8144 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8146 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8148 * config/arm/arm-cpus.in (fp16fml): New feature.
8149 (ALL_SIMD): Add fp16fml.
8150 (armv8.2-a): Add fp16fml as an option.
8151 (armv8.3-a): Likewise.
8152 (armv8.4-a): Add fp16fml as part of fp16.
8153 * config/arm/arm.h (TARGET_FP16FML): Define.
8154 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8155 when appropriate.
8156 * config/arm/arm-modes.def (V2HF): Define.
8157 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8158 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8159 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8160 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8161 vfmsl_low, vfmsl_high): New set of builtins.
8162 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8163 (vfml_op): New code attribute.
8164 (VFMLHALVES): New int iterator.
8165 (VFML, VFMLSEL): New mode attributes.
8166 (V_reg): Define mapping for V2HF.
8167 (V_hi, V_lo): New mode attributes.
8168 (VF_constraint): Likewise.
8169 (vfml_half, vfml_half_selector): New int attributes.
8170 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8171 define_expand.
8172 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8173 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8174 New define_insn.
8175 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8176 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8177 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8178 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8179 documentation.
8180 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8181 Document new effective target and option set.
8183 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8185 * config/arm/arm-cpus.in (armv8_4): New feature.
8186 (ARMv8_4a): New fgroup.
8187 (armv8.4-a): New arch.
8188 * config/arm/arm-tables.opt: Regenerate.
8189 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8190 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8191 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8192 Add matching rules for -march=armv8.4-a and extensions.
8193 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8195 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8197 PR target/81821
8198 * config/rx/rx.md (BW): New mode attribute.
8199 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8201 2018-01-11 Richard Biener <rguenther@suse.de>
8203 PR tree-optimization/83435
8204 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8205 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8206 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8208 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8209 Alan Hayward <alan.hayward@arm.com>
8210 David Sherwood <david.sherwood@arm.com>
8212 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8213 field.
8214 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8215 (aarch64_print_address_internal): Use it to check for a zero offset.
8217 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8218 Alan Hayward <alan.hayward@arm.com>
8219 David Sherwood <david.sherwood@arm.com>
8221 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8222 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8223 Return a poly_int64 rather than a HOST_WIDE_INT.
8224 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8225 rather than a HOST_WIDE_INT.
8226 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8227 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8228 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8229 final_offset from HOST_WIDE_INT to poly_int64.
8230 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8231 to_constant when getting the number of units in an Advanced SIMD
8232 mode.
8233 (aarch64_builtin_vectorized_function): Check for a constant number
8234 of units.
8235 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8236 GET_MODE_SIZE.
8237 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8238 attribute instead of GET_MODE_NUNITS.
8239 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8240 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8241 GET_MODE_SIZE for fixed-size registers.
8242 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8243 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8244 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8245 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8246 (aarch64_print_operand, aarch64_print_address_internal)
8247 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8248 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8249 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8250 Handle polynomial GET_MODE_SIZE.
8251 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8252 wider than SImode without modification.
8253 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8254 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8255 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8256 passing and returning SVE modes.
8257 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8258 rather than GEN_INT.
8259 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8260 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8261 (aarch64_allocate_and_probe_stack_space): Likewise.
8262 (aarch64_layout_frame): Cope with polynomial offsets.
8263 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8264 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8265 polynomial offsets.
8266 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8267 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8268 poly_int64 rather than a HOST_WIDE_INT.
8269 (aarch64_get_separate_components, aarch64_process_components)
8270 (aarch64_expand_prologue, aarch64_expand_epilogue)
8271 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8272 (aarch64_anchor_offset): New function, split out from...
8273 (aarch64_legitimize_address): ...here.
8274 (aarch64_builtin_vectorization_cost): Handle polynomial
8275 TYPE_VECTOR_SUBPARTS.
8276 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8277 GET_MODE_NUNITS.
8278 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8279 number of elements from the PARALLEL rather than the mode.
8280 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8281 rather than GET_MODE_BITSIZE.
8282 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8283 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8284 (aarch64_expand_vec_perm_const_1): Handle polynomial
8285 d->perm.length () and d->perm elements.
8286 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8287 Apply to_constant to d->perm elements.
8288 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8289 polynomial CONST_VECTOR_NUNITS.
8290 (aarch64_move_pointer): Take amount as a poly_int64 rather
8291 than an int.
8292 (aarch64_progress_pointer): Avoid temporary variable.
8293 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8294 the mode attribute instead of GET_MODE.
8296 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8297 Alan Hayward <alan.hayward@arm.com>
8298 David Sherwood <david.sherwood@arm.com>
8300 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8301 x exists before using it.
8302 (aarch64_add_constant_internal): Rename to...
8303 (aarch64_add_offset_1): ...this. Replace regnum with separate
8304 src and dest rtxes. Handle the case in which they're different,
8305 including when the offset is zero. Replace scratchreg with an rtx.
8306 Use 2 additions if there is no spare register into which we can
8307 move a 16-bit constant.
8308 (aarch64_add_constant): Delete.
8309 (aarch64_add_offset): Replace reg with separate src and dest
8310 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8311 Use aarch64_add_offset_1.
8312 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8313 an rtx rather than an int. Take the delta as a poly_int64
8314 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8315 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8316 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8317 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8318 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8319 and aarch64_add_sp.
8320 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8321 aarch64_add_constant.
8323 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8325 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8326 Use scalar_float_mode.
8328 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8330 * config/aarch64/aarch64-simd.md
8331 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8332 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8333 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8334 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8335 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8336 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8337 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8338 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8339 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8340 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8342 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8344 PR target/83514
8345 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8346 targ_options->x_arm_arch_string is non NULL.
8348 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8350 * config/aarch64/aarch64.h
8351 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8353 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8355 PR target/82096
8356 * expmed.c (emit_store_flag_force): Swap if const op0
8357 and change VOIDmode to mode of op0.
8359 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8361 PR rtl-optimization/83761
8362 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8363 than bytes to mode_for_size.
8365 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8367 PR middle-end/83189
8368 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8369 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8370 profile.
8372 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8374 PR middle-end/83575
8375 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8376 when in layout mode.
8377 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8378 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8379 partition fixup.
8381 2018-01-10 Michael Collison <michael.collison@arm.com>
8383 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8384 * config/aarch64/aarch64-option-extension.def: Add
8385 AARCH64_OPT_EXTENSION of 'fp16fml'.
8386 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8387 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8388 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8389 * config/aarch64/constraints.md (Ui7): New constraint.
8390 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8391 (VFMLA_SEL_W): Ditto.
8392 (f16quad): Ditto.
8393 (f16mac1): Ditto.
8394 (VFMLA16_LOW): New int iterator.
8395 (VFMLA16_HIGH): Ditto.
8396 (UNSPEC_FMLAL): New unspec.
8397 (UNSPEC_FMLSL): Ditto.
8398 (UNSPEC_FMLAL2): Ditto.
8399 (UNSPEC_FMLSL2): Ditto.
8400 (f16mac): New code attribute.
8401 * config/aarch64/aarch64-simd-builtins.def
8402 (aarch64_fmlal_lowv2sf): Ditto.
8403 (aarch64_fmlsl_lowv2sf): Ditto.
8404 (aarch64_fmlalq_lowv4sf): Ditto.
8405 (aarch64_fmlslq_lowv4sf): Ditto.
8406 (aarch64_fmlal_highv2sf): Ditto.
8407 (aarch64_fmlsl_highv2sf): Ditto.
8408 (aarch64_fmlalq_highv4sf): Ditto.
8409 (aarch64_fmlslq_highv4sf): Ditto.
8410 (aarch64_fmlal_lane_lowv2sf): Ditto.
8411 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8412 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8413 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8414 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8415 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8416 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8417 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8418 (aarch64_fmlal_lane_highv2sf): Ditto.
8419 (aarch64_fmlsl_lane_highv2sf): Ditto.
8420 (aarch64_fmlal_laneq_highv2sf): Ditto.
8421 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8422 (aarch64_fmlalq_lane_highv4sf): Ditto.
8423 (aarch64_fmlsl_lane_highv4sf): Ditto.
8424 (aarch64_fmlalq_laneq_highv4sf): Ditto.
8425 (aarch64_fmlsl_laneq_highv4sf): Ditto.
8426 * config/aarch64/aarch64-simd.md:
8427 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8428 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8429 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8430 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8431 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8432 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8433 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8434 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8435 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8436 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8437 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8438 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8439 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8440 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8441 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8442 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8443 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8444 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8445 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8446 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8447 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8448 (vfmlsl_low_u32): Ditto.
8449 (vfmlalq_low_u32): Ditto.
8450 (vfmlslq_low_u32): Ditto.
8451 (vfmlal_high_u32): Ditto.
8452 (vfmlsl_high_u32): Ditto.
8453 (vfmlalq_high_u32): Ditto.
8454 (vfmlslq_high_u32): Ditto.
8455 (vfmlal_lane_low_u32): Ditto.
8456 (vfmlsl_lane_low_u32): Ditto.
8457 (vfmlal_laneq_low_u32): Ditto.
8458 (vfmlsl_laneq_low_u32): Ditto.
8459 (vfmlalq_lane_low_u32): Ditto.
8460 (vfmlslq_lane_low_u32): Ditto.
8461 (vfmlalq_laneq_low_u32): Ditto.
8462 (vfmlslq_laneq_low_u32): Ditto.
8463 (vfmlal_lane_high_u32): Ditto.
8464 (vfmlsl_lane_high_u32): Ditto.
8465 (vfmlal_laneq_high_u32): Ditto.
8466 (vfmlsl_laneq_high_u32): Ditto.
8467 (vfmlalq_lane_high_u32): Ditto.
8468 (vfmlslq_lane_high_u32): Ditto.
8469 (vfmlalq_laneq_high_u32): Ditto.
8470 (vfmlslq_laneq_high_u32): Ditto.
8471 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
8472 (AARCH64_FL_FOR_ARCH8_4): New.
8473 (AARCH64_ISA_F16FML): New ISA flag.
8474 (TARGET_F16FML): New feature flag for fp16fml.
8475 (doc/invoke.texi): Document new fp16fml option.
8477 2018-01-10 Michael Collison <michael.collison@arm.com>
8479 * config/aarch64/aarch64-builtins.c:
8480 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
8481 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8482 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
8483 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
8484 (AARCH64_ISA_SHA3): New ISA flag.
8485 (TARGET_SHA3): New feature flag for sha3.
8486 * config/aarch64/iterators.md (sha512_op): New int attribute.
8487 (CRYPTO_SHA512): New int iterator.
8488 (UNSPEC_SHA512H): New unspec.
8489 (UNSPEC_SHA512H2): Ditto.
8490 (UNSPEC_SHA512SU0): Ditto.
8491 (UNSPEC_SHA512SU1): Ditto.
8492 * config/aarch64/aarch64-simd-builtins.def
8493 (aarch64_crypto_sha512hqv2di): New builtin.
8494 (aarch64_crypto_sha512h2qv2di): Ditto.
8495 (aarch64_crypto_sha512su0qv2di): Ditto.
8496 (aarch64_crypto_sha512su1qv2di): Ditto.
8497 (aarch64_eor3qv8hi): Ditto.
8498 (aarch64_rax1qv2di): Ditto.
8499 (aarch64_xarqv2di): Ditto.
8500 (aarch64_bcaxqv8hi): Ditto.
8501 * config/aarch64/aarch64-simd.md:
8502 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8503 (aarch64_crypto_sha512su0qv2di): Ditto.
8504 (aarch64_crypto_sha512su1qv2di): Ditto.
8505 (aarch64_eor3qv8hi): Ditto.
8506 (aarch64_rax1qv2di): Ditto.
8507 (aarch64_xarqv2di): Ditto.
8508 (aarch64_bcaxqv8hi): Ditto.
8509 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8510 (vsha512h2q_u64): Ditto.
8511 (vsha512su0q_u64): Ditto.
8512 (vsha512su1q_u64): Ditto.
8513 (veor3q_u16): Ditto.
8514 (vrax1q_u64): Ditto.
8515 (vxarq_u64): Ditto.
8516 (vbcaxq_u16): Ditto.
8517 * config/arm/types.md (crypto_sha512): New type attribute.
8518 (crypto_sha3): Ditto.
8519 (doc/invoke.texi): Document new sha3 option.
8521 2018-01-10 Michael Collison <michael.collison@arm.com>
8523 * config/aarch64/aarch64-builtins.c:
8524 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8525 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8526 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8527 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8528 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8529 (AARCH64_ISA_SM4): New ISA flag.
8530 (TARGET_SM4): New feature flag for sm4.
8531 * config/aarch64/aarch64-simd-builtins.def
8532 (aarch64_sm3ss1qv4si): Ditto.
8533 (aarch64_sm3tt1aq4si): Ditto.
8534 (aarch64_sm3tt1bq4si): Ditto.
8535 (aarch64_sm3tt2aq4si): Ditto.
8536 (aarch64_sm3tt2bq4si): Ditto.
8537 (aarch64_sm3partw1qv4si): Ditto.
8538 (aarch64_sm3partw2qv4si): Ditto.
8539 (aarch64_sm4eqv4si): Ditto.
8540 (aarch64_sm4ekeyqv4si): Ditto.
8541 * config/aarch64/aarch64-simd.md:
8542 (aarch64_sm3ss1qv4si): Ditto.
8543 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8544 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8545 (aarch64_sm4eqv4si): Ditto.
8546 (aarch64_sm4ekeyqv4si): Ditto.
8547 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8548 (sm3part_op): Ditto.
8549 (CRYPTO_SM3TT): Ditto.
8550 (CRYPTO_SM3PART): Ditto.
8551 (UNSPEC_SM3SS1): New unspec.
8552 (UNSPEC_SM3TT1A): Ditto.
8553 (UNSPEC_SM3TT1B): Ditto.
8554 (UNSPEC_SM3TT2A): Ditto.
8555 (UNSPEC_SM3TT2B): Ditto.
8556 (UNSPEC_SM3PARTW1): Ditto.
8557 (UNSPEC_SM3PARTW2): Ditto.
8558 (UNSPEC_SM4E): Ditto.
8559 (UNSPEC_SM4EKEY): Ditto.
8560 * config/aarch64/constraints.md (Ui2): New constraint.
8561 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8562 * config/arm/types.md (crypto_sm3): New type attribute.
8563 (crypto_sm4): Ditto.
8564 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8565 (vsm3tt1aq_u32): Ditto.
8566 (vsm3tt1bq_u32): Ditto.
8567 (vsm3tt2aq_u32): Ditto.
8568 (vsm3tt2bq_u32): Ditto.
8569 (vsm3partw1q_u32): Ditto.
8570 (vsm3partw2q_u32): Ditto.
8571 (vsm4eq_u32): Ditto.
8572 (vsm4ekeyq_u32): Ditto.
8573 (doc/invoke.texi): Document new sm4 option.
8575 2018-01-10 Michael Collison <michael.collison@arm.com>
8577 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8578 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8579 (AARCH64_FL_FOR_ARCH8_4): New.
8580 (AARCH64_FL_V8_4): New flag.
8581 (doc/invoke.texi): Document new armv8.4-a option.
8583 2018-01-10 Michael Collison <michael.collison@arm.com>
8585 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8586 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8587 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8588 * config/aarch64/aarch64-option-extension.def: Add
8589 AARCH64_OPT_EXTENSION of 'sha2'.
8590 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8591 (crypto): Disable sha2 and aes if crypto disabled.
8592 (crypto): Enable aes and sha2 if enabled.
8593 (simd): Disable sha2 and aes if simd disabled.
8594 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8595 New flags.
8596 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8597 (TARGET_SHA2): New feature flag for sha2.
8598 (TARGET_AES): New feature flag for aes.
8599 * config/aarch64/aarch64-simd.md:
8600 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8601 conditional on TARGET_AES.
8602 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8603 (aarch64_crypto_sha1hsi): Make pattern conditional
8604 on TARGET_SHA2.
8605 (aarch64_crypto_sha1hv4si): Ditto.
8606 (aarch64_be_crypto_sha1hv4si): Ditto.
8607 (aarch64_crypto_sha1su1v4si): Ditto.
8608 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8609 (aarch64_crypto_sha1su0v4si): Ditto.
8610 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8611 (aarch64_crypto_sha256su0v4si): Ditto.
8612 (aarch64_crypto_sha256su1v4si): Ditto.
8613 (doc/invoke.texi): Document new aes and sha2 options.
8615 2018-01-10 Martin Sebor <msebor@redhat.com>
8617 PR tree-optimization/83781
8618 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8619 as string arrays.
8621 2018-01-11 Martin Sebor <msebor@gmail.com>
8622 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8624 PR tree-optimization/83501
8625 PR tree-optimization/81703
8627 * tree-ssa-strlen.c (get_string_cst): Rename...
8628 (get_string_len): ...to this. Handle global constants.
8629 (handle_char_store): Adjust.
8631 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
8632 Jim Wilson <jimw@sifive.com>
8634 * config/riscv/riscv-protos.h (riscv_output_return): New.
8635 * config/riscv/riscv.c (struct machine_function): New naked_p field.
8636 (riscv_attribute_table, riscv_output_return),
8637 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8638 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8639 (riscv_compute_frame_info): Only compute frame->mask if not a naked
8640 function.
8641 (riscv_expand_prologue): Add early return for naked function.
8642 (riscv_expand_epilogue): Likewise.
8643 (riscv_function_ok_for_sibcall): Return false for naked function.
8644 (riscv_set_current_function): New.
8645 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8646 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8647 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8648 * doc/extend.texi (RISC-V Function Attributes): New.
8650 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8652 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8653 check for 128-bit long double before checking TCmode.
8654 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8655 128-bit long doubles before checking TFmode or TCmode.
8656 (FLOAT128_IBM_P): Likewise.
8658 2018-01-10 Martin Sebor <msebor@redhat.com>
8660 PR tree-optimization/83671
8661 * builtins.c (c_strlen): Unconditionally return zero for the empty
8662 string.
8663 Use -Warray-bounds for warnings.
8664 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8665 for non-constant array indices with COMPONENT_REF, arrays of
8666 arrays, and pointers to arrays.
8667 (gimple_fold_builtin_strlen): Determine and set length range for
8668 non-constant character arrays.
8670 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8672 PR middle-end/81897
8673 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8674 empty blocks.
8676 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8678 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8680 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8682 PR target/83399
8683 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8684 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8685 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8686 indexed_or_indirect_operand predicate.
8687 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8688 (*vsx_le_perm_load_v8hi): Likewise.
8689 (*vsx_le_perm_load_v16qi): Likewise.
8690 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8691 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8692 (*vsx_le_perm_store_v8hi): Likewise.
8693 (*vsx_le_perm_store_v16qi): Likewise.
8694 (eight unnamed splitters): Likewise.
8696 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8698 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8699 * config/rs6000/emmintrin.h: Likewise.
8700 * config/rs6000/mmintrin.h: Likewise.
8701 * config/rs6000/xmmintrin.h: Likewise.
8703 2018-01-10 David Malcolm <dmalcolm@redhat.com>
8705 PR c++/43486
8706 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8707 "public_flag".
8708 * tree.c (tree_nop_conversion): Return true for location wrapper
8709 nodes.
8710 (maybe_wrap_with_location): New function.
8711 (selftest::check_strip_nops): New function.
8712 (selftest::test_location_wrappers): New function.
8713 (selftest::tree_c_tests): Call it.
8714 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8715 (maybe_wrap_with_location): New decl.
8716 (EXPR_LOCATION_WRAPPER_P): New macro.
8717 (location_wrapper_p): New inline function.
8718 (tree_strip_any_location_wrapper): New inline function.
8720 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
8722 PR target/83735
8723 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8724 stack_realign_offset for the largest alignment of stack slot
8725 actually used.
8726 (ix86_find_max_used_stack_alignment): New function.
8727 (ix86_finalize_stack_frame_flags): Use it. Set
8728 max_used_stack_alignment if we don't realign stack.
8729 * config/i386/i386.h (machine_function): Add
8730 max_used_stack_alignment.
8732 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
8734 * config/arm/arm.opt (-mbranch-cost): New option.
8735 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8736 account.
8738 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
8740 PR target/83629
8741 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8742 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8744 2018-01-10 Richard Biener <rguenther@suse.de>
8746 PR debug/83765
8747 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8748 early out so it also covers the case where we have a non-NULL
8749 origin.
8751 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8753 PR tree-optimization/83753
8754 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
8755 for non-strided grouped accesses if the number of elements is 1.
8757 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8759 PR target/81616
8760 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
8761 * i386.h (TARGET_USE_GATHER): Define.
8762 * x86-tune.def (X86_TUNE_USE_GATHER): New.
8764 2018-01-10 Martin Liska <mliska@suse.cz>
8766 PR bootstrap/82831
8767 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
8768 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
8769 partitioning.
8770 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
8771 CLEANUP_NO_PARTITIONING is not set.
8773 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8775 * doc/rtl.texi: Remove documentation of (const ...) wrappers
8776 for vectors, as a partial revert of r254296.
8777 * rtl.h (const_vec_p): Delete.
8778 (const_vec_duplicate_p): Don't test for vector CONSTs.
8779 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
8780 * expmed.c (make_tree): Likewise.
8782 Revert:
8783 * common.md (E, F): Use CONSTANT_P instead of checking for
8784 CONST_VECTOR.
8785 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
8786 checking for CONST_VECTOR.
8788 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8790 PR middle-end/83575
8791 * predict.c (force_edge_cold): Handle in more sane way edges
8792 with no prediction.
8794 2018-01-09 Carl Love <cel@us.ibm.com>
8796 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
8797 V4SI, V4SF types.
8798 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
8799 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
8800 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
8801 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
8802 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
8803 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
8804 * config/rs6000/rs6000-protos.h: Add extern defition for
8805 rs6000_generate_float2_double_code.
8806 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8807 function.
8808 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8809 (float2_v2df): Add define_expand.
8811 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
8813 PR target/83628
8814 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8815 op_mode in the force_to_mode call.
8817 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8819 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8820 instead of checking each element individually.
8821 (aarch64_evpc_uzp): Likewise.
8822 (aarch64_evpc_zip): Likewise.
8823 (aarch64_evpc_ext): Likewise.
8824 (aarch64_evpc_rev): Likewise.
8825 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8826 instead of checking each element individually. Return true without
8827 generating rtl if
8828 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8829 whether all selected elements come from the same input, instead of
8830 checking each element individually. Remove calls to gen_rtx_REG,
8831 start_sequence and end_sequence and instead assert that no rtl is
8832 generated.
8834 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8836 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8837 order of HIGH and CONST checks.
8839 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8841 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8842 if the destination isn't an SSA_NAME.
8844 2018-01-09 Richard Biener <rguenther@suse.de>
8846 PR tree-optimization/83668
8847 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8848 move prologue...
8849 (canonicalize_loop_form): ... here, renamed from ...
8850 (canonicalize_loop_closed_ssa_form): ... this and amended to
8851 swap successor edges for loop exit blocks to make us use
8852 the RPO order we need for initial schedule generation.
8854 2018-01-09 Joseph Myers <joseph@codesourcery.com>
8856 PR tree-optimization/64811
8857 * match.pd: When optimizing comparisons with Inf, avoid
8858 introducing or losing exceptions from comparisons with NaN.
8860 2018-01-09 Martin Liska <mliska@suse.cz>
8862 PR sanitizer/82517
8863 * asan.c (shadow_mem_size): Add gcc_assert.
8865 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8867 Don't save registers in main().
8869 PR target/83738
8870 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8871 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8872 * config/avr/avr.c (avr_set_current_function): Don't error if
8873 naked, OS_task or OS_main are specified at the same time.
8874 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8875 OS_main.
8876 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8877 attribute.
8878 * common/config/avr/avr-common.c (avr_option_optimization_table):
8879 Switch on -mmain-is-OS_task for optimizing compilations.
8881 2018-01-09 Richard Biener <rguenther@suse.de>
8883 PR tree-optimization/83572
8884 * graphite.c: Include cfganal.h.
8885 (graphite_transform_loops): Connect infinite loops to exit
8886 and remove fake edges at the end.
8888 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8890 * ipa-inline.c (edge_badness): Revert accidental checkin.
8892 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8894 PR ipa/80763
8895 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8896 symbols; not inline clones.
8898 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8900 PR target/83507
8901 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8902 hard registers. Formatting fixes.
8904 PR preprocessor/83722
8905 * gcc.c (try_generate_repro): Pass
8906 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8907 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8908 do_report_bug.
8910 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8911 Kito Cheng <kito.cheng@gmail.com>
8913 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8914 (riscv_leaf_function_p): Delete.
8915 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8917 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8919 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8920 function.
8921 (do_ifelse): New function.
8922 (do_isel): New function.
8923 (do_sub3): New function.
8924 (do_add3): New function.
8925 (do_load_mask_compare): New function.
8926 (do_overlap_load_compare): New function.
8927 (expand_compare_loop): New function.
8928 (expand_block_compare): Call expand_compare_loop() when appropriate.
8929 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8930 option description.
8931 (-mblock-compare-inline-loop-limit): New option.
8933 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8935 PR target/83677
8936 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
8937 Reverse order of second and third operands in first alternative.
8938 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
8939 of first and second elements in UNSPEC_VPERMR vector.
8940 (altivec_expand_vec_perm_le): Likewise.
8942 2018-01-08 Jeff Law <law@redhat.com>
8944 PR rtl-optimizatin/81308
8945 * tree-switch-conversion.c (cfg_altered): New file scoped static.
8946 (process_switch): If group_case_labels makes a change, then set
8947 cfg_altered.
8948 (pass_convert_switch::execute): If a switch is converted, then
8949 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
8951 PR rtl-optimization/81308
8952 * recog.c (split_all_insns): Conditionally cleanup the CFG after
8953 splitting insns.
8955 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
8957 PR target/83663 - Revert r255946
8958 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
8959 generation for cases where splatting a value is not useful.
8960 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8961 across a vec_duplicate and a paradoxical subreg forming a vector
8962 mode to a vec_concat.
8964 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8966 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
8967 -march=armv8.3-a variants.
8968 * config/arm/t-multilib: Likewise.
8969 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
8971 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8973 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
8974 to generate rtl.
8975 (cceq_ior_compare_complement): Give it a name so I can use it, and
8976 change boolean_or_operator predicate to boolean_operator so it can
8977 be used to generate a crand.
8978 (eqne): New code iterator.
8979 (bd/bd_neg): New code_attrs.
8980 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
8981 a single define_insn.
8982 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
8983 decrement (bdnzt/bdnzf/bdzt/bdzf).
8984 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
8985 with the new names of the branch decrement patterns, and added the
8986 names of the branch decrement conditional patterns.
8988 2018-01-08 Richard Biener <rguenther@suse.de>
8990 PR tree-optimization/83563
8991 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
8992 cache.
8994 2018-01-08 Richard Biener <rguenther@suse.de>
8996 PR middle-end/83713
8997 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
8999 2018-01-08 Richard Biener <rguenther@suse.de>
9001 PR tree-optimization/83685
9002 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9003 references to abnormals.
9005 2018-01-08 Richard Biener <rguenther@suse.de>
9007 PR lto/83719
9008 * dwarf2out.c (output_indirect_strings): Handle empty
9009 skeleton_debug_str_hash.
9010 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9012 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9014 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9015 (emit_store_direct): Likewise.
9016 (arc_trampoline_adjust_address): Likewise.
9017 (arc_asm_trampoline_template): New function.
9018 (arc_initialize_trampoline): Use asm_trampoline_template.
9019 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9020 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9021 * config/arc/arc.md (flush_icache): Delete pattern.
9023 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9025 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9026 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9027 munaligned-access.
9029 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9031 PR target/83681
9032 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9033 by not USED_FOR_TARGET.
9034 (make_pass_resolve_sw_modes): Likewise.
9036 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9038 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9039 USED_FOR_TARGET.
9041 2018-01-08 Richard Biener <rguenther@suse.de>
9043 PR middle-end/83580
9044 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9046 2018-01-08 Richard Biener <rguenther@suse.de>
9048 PR middle-end/83517
9049 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9051 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9053 PR middle-end/81897
9054 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9055 basic blocks with a small number of successors.
9056 (convert_control_dep_chain_into_preds): Improve handling of
9057 forwarder blocks.
9058 (dump_predicates): Split apart into...
9059 (dump_pred_chain): ...here...
9060 (dump_pred_info): ...and here.
9061 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9062 (can_chain_union_be_invalidated_p): Improve check for invalidation
9063 of paths.
9064 (uninit_uses_cannot_happen): Avoid unnecessary if
9065 convert_control_dep_chain_into_preds yielded nothing.
9067 2018-01-06 Martin Sebor <msebor@redhat.com>
9069 PR tree-optimization/83640
9070 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9071 subtracting negative offset from size.
9072 (builtin_access::overlap): Adjust offset bounds of the access to fall
9073 within the size of the object if possible.
9075 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9077 PR rtl-optimization/83699
9078 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9079 extract_bit_field_as_subreg to cases in which the extracted
9080 value is also a vector.
9082 * lra-constraints.c (process_alt_operands): Test for the equivalence
9083 substitutions when detecting a possible reload cycle.
9085 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9087 PR debug/83480
9088 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9089 by default if flag_selective_schedling{,2}. Formatting fixes.
9091 PR rtl-optimization/83682
9092 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9093 if it has non-VECTOR_MODE element mode.
9094 (vec_duplicate_p): Likewise.
9096 PR middle-end/83694
9097 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9098 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9100 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9102 PR target/83604
9103 * config/i386/i386-builtin.def
9104 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9105 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9106 Require also OPTION_MASK_ISA_AVX512F in addition to
9107 OPTION_MASK_ISA_GFNI.
9108 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9109 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9110 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9111 to OPTION_MASK_ISA_GFNI.
9112 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9113 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9114 OPTION_MASK_ISA_AVX512BW.
9115 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9116 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9117 addition to OPTION_MASK_ISA_GFNI.
9118 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9119 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9120 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9121 to OPTION_MASK_ISA_GFNI.
9122 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9123 a requirement for all ISAs rather than any of them with a few
9124 exceptions.
9125 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9126 processing.
9127 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9128 bitmasks to be enabled with 3 exceptions, instead of requiring any
9129 enabled ISA with lots of exceptions.
9130 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9131 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9132 Change avx512bw in isa attribute to avx512f.
9133 * config/i386/sgxintrin.h: Add license boilerplate.
9134 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9135 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9136 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9137 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9138 defined.
9139 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9140 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9141 temporarily sse2 rather than sse if not enabled already.
9143 PR target/83604
9144 * config/i386/sse.md (VI248_VLBW): Rename to ...
9145 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9146 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9147 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9148 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9149 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9150 mode iterator instead of VI248_VLBW.
9152 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9154 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9155 (record_modified): Skip clobbers; add debug output.
9156 (param_change_prob): Use sreal frequencies.
9158 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9160 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9161 punt for user-aligned variables.
9163 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9165 * tree-chrec.c (chrec_contains_symbols): Return true for
9166 POLY_INT_CST.
9168 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9170 PR target/82439
9171 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9172 of (x|y) == x for BICS pattern.
9174 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9176 PR tree-optimization/83605
9177 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9178 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9179 can throw.
9181 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9183 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9184 * config/epiphany/rtems.h: New file.
9186 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9187 Uros Bizjak <ubizjak@gmail.com>
9189 PR target/83554
9190 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9191 QIreg_operand instead of register_operand predicate.
9192 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9193 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9194 comments instead of -fmitigate[-_]rop.
9196 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9198 PR bootstrap/81926
9199 * cgraphunit.c (symbol_table::compile): Switch to text_section
9200 before calling assembly_start debug hook.
9201 * run-rtl-passes.c (run_rtl_passes): Likewise.
9202 Include output.h.
9204 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9206 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9207 range_int_cst_p rather than !symbolic_range_p before calling
9208 extract_range_from_multiplicative_op_1.
9210 2018-01-04 Jeff Law <law@redhat.com>
9212 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9213 redundant test in assertion.
9215 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9217 * doc/rtl.texi: Document machine_mode wrapper classes.
9219 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9221 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9222 using tree_to_uhwi.
9224 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9226 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9227 the VEC_PERM_EXPR fold to fail.
9229 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9231 PR debug/83585
9232 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9233 to switched_sections.
9235 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9237 PR target/83680
9238 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9239 test for d.testing.
9241 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9243 PR target/83387
9244 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9245 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9247 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9249 PR debug/83666
9250 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9251 is BLKmode and bitpos not zero or mode change is needed.
9253 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9255 PR target/83675
9256 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9257 TARGET_VIS2.
9259 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9261 PR target/83628
9262 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9263 instead of MULT rtx. Update all corresponding splitters.
9264 (*saddl_se): Ditto.
9265 (*ssub<modesuffix>): Ditto.
9266 (*ssubl_se): Ditto.
9267 (*cmp_sadd_di): Update split patterns.
9268 (*cmp_sadd_si): Ditto.
9269 (*cmp_sadd_sidi): Ditto.
9270 (*cmp_ssub_di): Ditto.
9271 (*cmp_ssub_si): Ditto.
9272 (*cmp_ssub_sidi): Ditto.
9273 * config/alpha/predicates.md (const23_operand): New predicate.
9274 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9275 Look for ASHIFT, not MULT inner operand.
9276 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9278 2018-01-04 Martin Liska <mliska@suse.cz>
9280 PR gcov-profile/83669
9281 * gcov.c (output_intermediate_file): Add version to intermediate
9282 gcov file.
9283 * doc/gcov.texi: Document new field 'version' in intermediate
9284 file format. Fix location of '-k' option of gcov command.
9286 2018-01-04 Martin Liska <mliska@suse.cz>
9288 PR ipa/82352
9289 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9291 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9293 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9295 2018-01-03 Martin Sebor <msebor@redhat.com>
9297 PR tree-optimization/83655
9298 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9299 checking calls with invalid arguments.
9301 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9303 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9304 (vectorizable_mask_load_store): Delete.
9305 (vectorizable_call): Return false for masked loads and stores.
9306 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9307 instead of gimple_assign_rhs1.
9308 (vectorizable_load): Handle IFN_MASK_LOAD.
9309 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9313 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9314 split out from..,
9315 (vectorizable_mask_load_store): ...here.
9316 (vectorizable_load): ...and here.
9318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9320 * tree-vect-stmts.c (vect_build_all_ones_mask)
9321 (vect_build_zero_merge_argument): New functions, split out from...
9322 (vectorizable_load): ...here.
9324 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9326 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9327 split out from...
9328 (vectorizable_mask_load_store): ...here.
9329 (vectorizable_store): ...and here.
9331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9333 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9334 split out from...
9335 (vectorizable_mask_load_store): ...here.
9337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9339 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9340 (vect_model_store_cost): Take a vec_load_store_type instead of a
9341 vect_def_type.
9342 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9343 (vect_model_store_cost): Take a vec_load_store_type instead of a
9344 vect_def_type.
9345 (vectorizable_mask_load_store): Update accordingly.
9346 (vectorizable_store): Likewise.
9347 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9351 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9352 IFN_MASK_LOAD calls here rather than...
9353 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9356 Alan Hayward <alan.hayward@arm.com>
9357 David Sherwood <david.sherwood@arm.com>
9359 * expmed.c (extract_bit_field_1): For vector extracts,
9360 fall back to extract_bit_field_as_subreg if vec_extract
9361 isn't available.
9363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9364 Alan Hayward <alan.hayward@arm.com>
9365 David Sherwood <david.sherwood@arm.com>
9367 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9368 they are variable or constant sized.
9369 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9370 slots for constant-sized data.
9372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9373 Alan Hayward <alan.hayward@arm.com>
9374 David Sherwood <david.sherwood@arm.com>
9376 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9377 handling COND_EXPRs with boolean comparisons, try to find a better
9378 basis for the mask type than the boolean itself.
9380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9382 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9383 is calculated and how it can be overridden.
9384 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9385 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9386 if defined.
9387 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9388 if nonzero.
9390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9391 Alan Hayward <alan.hayward@arm.com>
9392 David Sherwood <david.sherwood@arm.com>
9394 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9395 Remove the mode argument.
9396 (aarch64_simd_valid_immediate): Remove the mode and inverse
9397 arguments.
9398 * config/aarch64/iterators.md (bitsize): New iterator.
9399 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9400 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9401 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9402 aarch64_simd_valid_immediate.
9403 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9404 (aarch64_reg_or_bic_imm): Likewise.
9405 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9406 with an insn_type enum and msl with a modifier_type enum.
9407 Replace element_width with a scalar_mode. Change the shift
9408 to unsigned int. Add constructors for scalar_float_mode and
9409 scalar_int_mode elements.
9410 (aarch64_vect_float_const_representable_p): Delete.
9411 (aarch64_can_const_movi_rtx_p)
9412 (aarch64_simd_scalar_immediate_valid_for_move)
9413 (aarch64_simd_make_constant): Update call to
9414 aarch64_simd_valid_immediate.
9415 (aarch64_advsimd_valid_immediate_hs): New function.
9416 (aarch64_advsimd_valid_immediate): Likewise.
9417 (aarch64_simd_valid_immediate): Remove mode and inverse
9418 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9419 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9420 and aarch64_float_const_representable_p on the result.
9421 (aarch64_output_simd_mov_immediate): Remove mode argument.
9422 Update call to aarch64_simd_valid_immediate and use of
9423 simd_immediate_info.
9424 (aarch64_output_scalar_simd_mov_immediate): Update call
9425 accordingly.
9427 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9428 Alan Hayward <alan.hayward@arm.com>
9429 David Sherwood <david.sherwood@arm.com>
9431 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9432 (mode_nunits): Likewise CONST_MODE_NUNITS.
9433 * machmode.def (ADJUST_NUNITS): Document.
9434 * genmodes.c (mode_data::need_nunits_adj): New field.
9435 (blank_mode): Update accordingly.
9436 (adj_nunits): New variable.
9437 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9438 parameter.
9439 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9440 listed in adj_nunits.
9441 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9442 listed in adj_nunits. Don't emit case statements for such modes.
9443 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9444 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9445 nothing if adj_nunits is nonnull.
9446 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9447 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9448 (emit_mode_fbit): Update use of print_maybe_const_decl.
9449 (emit_move_size): Likewise. Treat the array as non-const
9450 if adj_nunits.
9451 (emit_mode_adjustments): Handle adj_nunits.
9453 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9455 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9456 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9457 (VECTOR_MODES): Use it.
9458 (make_vector_modes): Take the prefix as an argument.
9460 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9461 Alan Hayward <alan.hayward@arm.com>
9462 David Sherwood <david.sherwood@arm.com>
9464 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
9465 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
9466 for MODE_VECTOR_BOOL.
9467 * machmode.def (VECTOR_BOOL_MODE): Document.
9468 * genmodes.c (VECTOR_BOOL_MODE): New macro.
9469 (make_vector_bool_mode): New function.
9470 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
9471 MODE_VECTOR_BOOL.
9472 * lto-streamer-in.c (lto_input_mode_table): Likewise.
9473 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
9474 Likewise.
9475 * stor-layout.c (int_mode_for_mode): Likewise.
9476 * tree.c (build_vector_type_for_mode): Likewise.
9477 * varasm.c (output_constant_pool_2): Likewise.
9478 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
9479 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
9480 for MODE_VECTOR_BOOL.
9481 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
9482 of mode class checks.
9483 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
9484 instead of a list of mode class checks.
9485 (expand_vector_scalar_condition): Likewise.
9486 (type_for_widest_vector_mode): Handle BImode as an inner mode.
9488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9489 Alan Hayward <alan.hayward@arm.com>
9490 David Sherwood <david.sherwood@arm.com>
9492 * machmode.h (mode_size): Change from unsigned short to
9493 poly_uint16_pod.
9494 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
9495 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9496 or if measurement_type is not polynomial.
9497 (fixed_size_mode::includes_p): Check for constant-sized modes.
9498 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9499 return a poly_uint16 rather than an unsigned short.
9500 (emit_mode_size): Change the type of mode_size from unsigned short
9501 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
9502 (emit_mode_adjustments): Cope with polynomial vector sizes.
9503 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9504 for GET_MODE_SIZE.
9505 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9506 for GET_MODE_SIZE.
9507 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9508 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9509 * caller-save.c (setup_save_areas): Likewise.
9510 (replace_reg_with_saved_mem): Likewise.
9511 * calls.c (emit_library_call_value_1): Likewise.
9512 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9513 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9514 (gen_lowpart_for_combine): Likewise.
9515 * convert.c (convert_to_integer_1): Likewise.
9516 * cse.c (equiv_constant, cse_insn): Likewise.
9517 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9518 (cselib_subst_to_values): Likewise.
9519 * dce.c (word_dce_process_block): Likewise.
9520 * df-problems.c (df_word_lr_mark_ref): Likewise.
9521 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9522 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9523 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9524 (rtl_for_decl_location): Likewise.
9525 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9526 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9527 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9528 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9529 (expand_expr_real_1): Likewise.
9530 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9531 (pad_below): Likewise.
9532 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9533 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9534 * ira.c (get_subreg_tracking_sizes): Likewise.
9535 * ira-build.c (ira_create_allocno_objects): Likewise.
9536 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9537 (ira_sort_regnos_for_alter_reg): Likewise.
9538 * ira-costs.c (record_operand_costs): Likewise.
9539 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9540 (resolve_simple_move): Likewise.
9541 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9542 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9543 (lra_constraints): Likewise.
9544 (CONST_POOL_OK_P): Reject variable-sized modes.
9545 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9546 (add_pseudo_to_slot, lra_spill): Likewise.
9547 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9548 * optabs-query.c (get_best_extraction_insn): Likewise.
9549 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9550 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9551 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9552 * recog.c (offsettable_address_addr_space_p): Likewise.
9553 * regcprop.c (maybe_mode_change): Likewise.
9554 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9555 * regrename.c (build_def_use): Likewise.
9556 * regstat.c (dump_reg_info): Likewise.
9557 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9558 (find_reloads, find_reloads_subreg_address): Likewise.
9559 * reload1.c (eliminate_regs_1): Likewise.
9560 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9561 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9562 (simplify_binary_operation_1, simplify_subreg): Likewise.
9563 * targhooks.c (default_function_arg_padding): Likewise.
9564 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9565 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9566 (verify_gimple_assign_ternary): Likewise.
9567 * tree-inline.c (estimate_move_cost): Likewise.
9568 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9569 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9570 (get_address_cost_ainc): Likewise.
9571 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9572 (vect_supportable_dr_alignment): Likewise.
9573 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9574 (vectorizable_reduction): Likewise.
9575 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9576 (vectorizable_operation, vectorizable_load): Likewise.
9577 * tree.c (build_same_sized_truth_vector_type): Likewise.
9578 * valtrack.c (cleanup_auto_inc_dec): Likewise.
9579 * var-tracking.c (emit_note_insn_var_location): Likewise.
9580 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9581 (ADDR_VEC_ALIGN): Likewise.
9583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9584 Alan Hayward <alan.hayward@arm.com>
9585 David Sherwood <david.sherwood@arm.com>
9587 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9588 unsigned short.
9589 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9590 or if measurement_type is polynomial.
9591 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9592 * combine.c (make_extraction): Likewise.
9593 * dse.c (find_shift_sequence): Likewise.
9594 * dwarf2out.c (mem_loc_descriptor): Likewise.
9595 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9596 (extract_bit_field, extract_low_bits): Likewise.
9597 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9598 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9599 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9600 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9601 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9602 * reload.c (find_reloads): Likewise.
9603 * reload1.c (alter_reg): Likewise.
9604 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9605 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9606 * tree-if-conv.c (predicate_mem_writes): Likewise.
9607 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9608 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9609 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9610 * valtrack.c (dead_debug_insert_temp): Likewise.
9611 * varasm.c (mergeable_constant_section): Likewise.
9612 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
9614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9615 Alan Hayward <alan.hayward@arm.com>
9616 David Sherwood <david.sherwood@arm.com>
9618 * expr.c (expand_assignment): Cope with polynomial mode sizes
9619 when assigning to a CONCAT.
9621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9622 Alan Hayward <alan.hayward@arm.com>
9623 David Sherwood <david.sherwood@arm.com>
9625 * machmode.h (mode_precision): Change from unsigned short to
9626 poly_uint16_pod.
9627 (mode_to_precision): Return a poly_uint16 rather than an unsigned
9628 short.
9629 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9630 or if measurement_type is not polynomial.
9631 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
9632 in which the mode is already known to be a scalar_int_mode.
9633 * genmodes.c (emit_mode_precision): Change the type of mode_precision
9634 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
9635 initializer.
9636 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9637 for GET_MODE_PRECISION.
9638 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9639 for GET_MODE_PRECISION.
9640 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9641 as polynomial.
9642 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9643 (expand_field_assignment, make_extraction): Likewise.
9644 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9645 (get_last_value): Likewise.
9646 * convert.c (convert_to_integer_1): Likewise.
9647 * cse.c (cse_insn): Likewise.
9648 * expr.c (expand_expr_real_1): Likewise.
9649 * lra-constraints.c (simplify_operand_subreg): Likewise.
9650 * optabs-query.c (can_atomic_load_p): Likewise.
9651 * optabs.c (expand_atomic_load): Likewise.
9652 (expand_atomic_store): Likewise.
9653 * ree.c (combine_reaching_defs): Likewise.
9654 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9655 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9656 * tree.h (type_has_mode_precision_p): Likewise.
9657 * ubsan.c (instrument_si_overflow): Likewise.
9659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9660 Alan Hayward <alan.hayward@arm.com>
9661 David Sherwood <david.sherwood@arm.com>
9663 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9664 polynomial numbers of units.
9665 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9666 (valid_vector_subparts_p): New function.
9667 (build_vector_type): Remove temporary shim and take the number
9668 of units as a poly_uint64 rather than an int.
9669 (build_opaque_vector_type): Take the number of units as a
9670 poly_uint64 rather than an int.
9671 * tree.c (build_vector_from_ctor): Handle polynomial
9672 TYPE_VECTOR_SUBPARTS.
9673 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9674 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9675 (build_vector_from_val): If the number of units is variable,
9676 use build_vec_duplicate_cst for constant operands and
9677 VEC_DUPLICATE_EXPR otherwise.
9678 (make_vector_type): Remove temporary is_constant ().
9679 (build_vector_type, build_opaque_vector_type): Take the number of
9680 units as a poly_uint64 rather than an int.
9681 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9682 VECTOR_CST_NELTS.
9683 * cfgexpand.c (expand_debug_expr): Likewise.
9684 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9685 (store_constructor, expand_expr_real_1): Likewise.
9686 (const_scalar_mask_from_tree): Likewise.
9687 * fold-const-call.c (fold_const_reduction): Likewise.
9688 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9689 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9690 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9691 (fold_relational_const): Likewise.
9692 (native_interpret_vector): Likewise. Change the size from an
9693 int to an unsigned int.
9694 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9695 TYPE_VECTOR_SUBPARTS.
9696 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9697 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9698 duplicating a non-constant operand into a variable-length vector.
9699 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9700 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9701 * ipa-icf.c (sem_variable::equals): Likewise.
9702 * match.pd: Likewise.
9703 * omp-simd-clone.c (simd_clone_subparts): Likewise.
9704 * print-tree.c (print_node): Likewise.
9705 * stor-layout.c (layout_type): Likewise.
9706 * targhooks.c (default_builtin_vectorization_cost): Likewise.
9707 * tree-cfg.c (verify_gimple_comparison): Likewise.
9708 (verify_gimple_assign_binary): Likewise.
9709 (verify_gimple_assign_ternary): Likewise.
9710 (verify_gimple_assign_single): Likewise.
9711 * tree-pretty-print.c (dump_generic_node): Likewise.
9712 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9713 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9714 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9715 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9716 (vect_shift_permute_load_chain): Likewise.
9717 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9718 (expand_vector_condition, optimize_vector_constructor): Likewise.
9719 (lower_vec_perm, get_compute_type): Likewise.
9720 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9721 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9722 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9723 (vect_recog_mask_conversion_pattern): Likewise.
9724 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9725 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9726 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9727 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9728 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9729 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9730 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9731 (supportable_widening_operation): Likewise.
9732 (supportable_narrowing_operation): Likewise.
9733 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9734 Likewise.
9735 * varasm.c (output_constant): Likewise.
9737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9738 Alan Hayward <alan.hayward@arm.com>
9739 David Sherwood <david.sherwood@arm.com>
9741 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9742 so that both the length == 3 and length != 3 cases set up their
9743 own permute vectors. Add comments explaining why we know the
9744 number of elements is constant.
9745 (vect_permute_load_chain): Likewise.
9747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9748 Alan Hayward <alan.hayward@arm.com>
9749 David Sherwood <david.sherwood@arm.com>
9751 * machmode.h (mode_nunits): Change from unsigned char to
9752 poly_uint16_pod.
9753 (ONLY_FIXED_SIZE_MODES): New macro.
9754 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
9755 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
9756 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
9757 New typedefs.
9758 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
9759 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
9760 or if measurement_type is not polynomial.
9761 * genmodes.c (ZERO_COEFFS): New macro.
9762 (emit_mode_nunits_inline): Make mode_nunits_inline return a
9763 poly_uint16.
9764 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
9765 Use ZERO_COEFFS when emitting initializers.
9766 * data-streamer.h (bp_pack_poly_value): New function.
9767 (bp_unpack_poly_value): Likewise.
9768 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9769 for GET_MODE_NUNITS.
9770 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9771 for GET_MODE_NUNITS.
9772 * tree.c (make_vector_type): Remove temporary shim and make
9773 the real function take the number of units as a poly_uint64
9774 rather than an int.
9775 (build_vector_type_for_mode): Handle polynomial nunits.
9776 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
9777 * emit-rtl.c (const_vec_series_p_1): Likewise.
9778 (gen_rtx_CONST_VECTOR): Likewise.
9779 * fold-const.c (test_vec_duplicate_folding): Likewise.
9780 * genrecog.c (validate_pattern): Likewise.
9781 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
9782 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9783 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
9784 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
9785 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
9786 * rtlanal.c (subreg_get_info): Likewise.
9787 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9788 (vect_grouped_load_supported): Likewise.
9789 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
9790 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
9791 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9792 (simplify_const_unary_operation, simplify_binary_operation_1)
9793 (simplify_const_binary_operation, simplify_ternary_operation)
9794 (test_vector_ops_duplicate, test_vector_ops): Likewise.
9795 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
9796 instead of CONST_VECTOR_NUNITS.
9797 * varasm.c (output_constant_pool_2): Likewise.
9798 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
9799 explicit-encoded elements in the XVEC for variable-length vectors.
9801 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9803 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9806 Alan Hayward <alan.hayward@arm.com>
9807 David Sherwood <david.sherwood@arm.com>
9809 * coretypes.h (fixed_size_mode): Declare.
9810 (fixed_size_mode_pod): New typedef.
9811 * builtins.h (target_builtins::x_apply_args_mode)
9812 (target_builtins::x_apply_result_mode): Change type to
9813 fixed_size_mode_pod.
9814 * builtins.c (apply_args_size, apply_result_size, result_vector)
9815 (expand_builtin_apply_args_1, expand_builtin_apply)
9816 (expand_builtin_return): Update accordingly.
9818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9820 * cse.c (hash_rtx_cb): Hash only the encoded elements.
9821 * cselib.c (cselib_hash_rtx): Likewise.
9822 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9823 CONST_VECTOR encoding.
9825 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9826 Jeff Law <law@redhat.com>
9828 PR target/83641
9829 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9830 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9831 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9832 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9834 PR target/83641
9835 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9836 explicitly probe *sp in a noreturn function if there were any callee
9837 register saves or frame pointer is needed.
9839 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9841 PR debug/83621
9842 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9843 BLKmode for ternary, binary or unary expressions.
9845 PR debug/83645
9846 * var-tracking.c (delete_vta_debug_insn): New inline function.
9847 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9848 insns from get_insns () to NULL instead of each bb separately.
9849 Use delete_vta_debug_insn. No longer static.
9850 (vt_debug_insns_local, variable_tracking_main_1): Adjust
9851 delete_vta_debug_insns callers.
9852 * rtl.h (delete_vta_debug_insns): Declare.
9853 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9854 instead of variable_tracking_main.
9856 2018-01-03 Martin Sebor <msebor@redhat.com>
9858 PR tree-optimization/83603
9859 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9860 arguments past the endof the argument list in functions declared
9861 without a prototype.
9862 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9863 Avoid checking when arguments are null.
9865 2018-01-03 Martin Sebor <msebor@redhat.com>
9867 PR c/83559
9868 * doc/extend.texi (attribute const): Fix a typo.
9869 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9870 issuing -Wsuggest-attribute for void functions.
9872 2018-01-03 Martin Sebor <msebor@redhat.com>
9874 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9875 offset_int::from instead of wide_int::to_shwi.
9876 (maybe_diag_overlap): Remove assertion.
9877 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9878 * gimple-ssa-sprintf.c (format_directive): Same.
9879 (parse_directive): Same.
9880 (sprintf_dom_walker::compute_format_length): Same.
9881 (try_substitute_return_value): Same.
9883 2018-01-03 Jeff Law <law@redhat.com>
9885 PR middle-end/83654
9886 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9887 non-constant residual for zero at runtime and avoid probing in
9888 that case. Reorganize code for trailing problem to mirror handling
9889 of the residual.
9891 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9893 PR tree-optimization/83501
9894 * tree-ssa-strlen.c (get_string_cst): New.
9895 (handle_char_store): Call get_string_cst.
9897 2018-01-03 Martin Liska <mliska@suse.cz>
9899 PR tree-optimization/83593
9900 * tree-ssa-strlen.c: Include tree-cfg.h.
9901 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9902 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9903 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9904 to false.
9905 (strlen_dom_walker::before_dom_children): Call
9906 gimple_purge_dead_eh_edges. Dump tranformation with details
9907 dump flags.
9908 (strlen_dom_walker::before_dom_children): Update call by adding
9909 new argument cleanup_eh.
9910 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9912 2018-01-03 Martin Liska <mliska@suse.cz>
9914 PR ipa/83549
9915 * cif-code.def (VARIADIC_THUNK): New enum value.
9916 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9917 thunks.
9919 2018-01-03 Jan Beulich <jbeulich@suse.com>
9921 * sse.md (mov<mode>_internal): Tighten condition for when to use
9922 vmovdqu<ssescalarsize> for TI and OI modes.
9924 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9926 Update copyright years.
9928 2018-01-03 Martin Liska <mliska@suse.cz>
9930 PR ipa/83594
9931 * ipa-visibility.c (function_and_variable_visibility): Skip
9932 functions with noipa attribure.
9934 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9936 * gcc.c (process_command): Update copyright notice dates.
9937 * gcov-dump.c (print_version): Ditto.
9938 * gcov.c (print_version): Ditto.
9939 * gcov-tool.c (print_version): Ditto.
9940 * gengtype.c (create_file): Ditto.
9941 * doc/cpp.texi: Bump @copying's copyright year.
9942 * doc/cppinternals.texi: Ditto.
9943 * doc/gcc.texi: Ditto.
9944 * doc/gccint.texi: Ditto.
9945 * doc/gcov.texi: Ditto.
9946 * doc/install.texi: Ditto.
9947 * doc/invoke.texi: Ditto.
9949 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9951 * vector-builder.h (vector_builder::m_full_nelts): Change from
9952 unsigned int to poly_uint64.
9953 (vector_builder::full_nelts): Update prototype accordingly.
9954 (vector_builder::new_vector): Likewise.
9955 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
9956 (vector_builder::operator ==): Likewise.
9957 (vector_builder::finalize): Likewise.
9958 * int-vector-builder.h (int_vector_builder::int_vector_builder):
9959 Take the number of elements as a poly_uint64 rather than an
9960 unsigned int.
9961 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
9962 from unsigned int to poly_uint64.
9963 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
9964 (vec_perm_indices::new_vector): Likewise.
9965 (vec_perm_indices::length): Likewise.
9966 (vec_perm_indices::nelts_per_input): Likewise.
9967 (vec_perm_indices::input_nelts): Likewise.
9968 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
9969 number of elements per input as a poly_uint64 rather than an
9970 unsigned int. Use the original encoding for variable-length
9971 vectors, rather than clamping each individual element.
9972 For the second and subsequent elements in each pattern,
9973 clamp the step and base before clamping their sum.
9974 (vec_perm_indices::series_p): Handle polynomial element counts.
9975 (vec_perm_indices::all_in_range_p): Likewise.
9976 (vec_perm_indices_to_tree): Likewise.
9977 (vec_perm_indices_to_rtx): Likewise.
9978 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
9979 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
9980 (tree_vector_builder::new_binary_operation): Handle polynomial
9981 element counts. Return false if we need to know the number
9982 of elements at compile time.
9983 * fold-const.c (fold_vec_perm): Punt if the number of elements
9984 isn't known at compile time.
9986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9988 * vec-perm-indices.h (vec_perm_builder): Change element type
9989 from HOST_WIDE_INT to poly_int64.
9990 (vec_perm_indices::element_type): Update accordingly.
9991 (vec_perm_indices::clamp): Handle polynomial element_types.
9992 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9993 (vec_perm_indices::all_in_range_p): Likewise.
9994 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
9995 than shwi trees.
9996 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
9997 polynomial vec_perm_indices element types.
9998 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
9999 * fold-const.c (fold_vec_perm): Likewise.
10000 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10001 * tree-vect-generic.c (lower_vec_perm): Likewise.
10002 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10003 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10004 element type to HOST_WIDE_INT.
10006 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10007 Alan Hayward <alan.hayward@arm.com>
10008 David Sherwood <david.sherwood@arm.com>
10010 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10011 rather than an int. Use plus_constant.
10012 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10013 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10016 Alan Hayward <alan.hayward@arm.com>
10017 David Sherwood <david.sherwood@arm.com>
10019 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10020 a HOST_WIDE_INT to a poly_int64.
10022 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10023 Alan Hayward <alan.hayward@arm.com>
10024 David Sherwood <david.sherwood@arm.com>
10026 * calls.c (load_register_parameters): Cope with polynomial
10027 mode sizes. Require a constant size for BLKmode parameters
10028 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10029 forces a parameter to be padded at the lsb end in order to
10030 fill a complete number of words, require the parameter size
10031 to be ordered wrt UNITS_PER_WORD.
10033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10034 Alan Hayward <alan.hayward@arm.com>
10035 David Sherwood <david.sherwood@arm.com>
10037 * reload1.c (spill_stack_slot_width): Change element type
10038 from unsigned int to poly_uint64_pod.
10039 (alter_reg): Treat mode sizes as polynomial.
10041 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10042 Alan Hayward <alan.hayward@arm.com>
10043 David Sherwood <david.sherwood@arm.com>
10045 * reload.c (complex_word_subreg_p): New function.
10046 (reload_inner_reg_of_subreg, push_reload): Use it.
10048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10049 Alan Hayward <alan.hayward@arm.com>
10050 David Sherwood <david.sherwood@arm.com>
10052 * lra-constraints.c (process_alt_operands): Reject matched
10053 operands whose sizes aren't ordered.
10054 (match_reload): Refer to this check here.
10056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10057 Alan Hayward <alan.hayward@arm.com>
10058 David Sherwood <david.sherwood@arm.com>
10060 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10061 that the mode size is in the set {1, 2, 4, 8, 16}.
10063 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10064 Alan Hayward <alan.hayward@arm.com>
10065 David Sherwood <david.sherwood@arm.com>
10067 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10068 Use plus_constant instead of gen_rtx_PLUS.
10070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10071 Alan Hayward <alan.hayward@arm.com>
10072 David Sherwood <david.sherwood@arm.com>
10074 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10075 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10076 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10077 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10078 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10079 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10080 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10081 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10082 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10083 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10084 a poly_int64.
10085 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10086 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10087 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10088 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10089 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10090 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10091 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10092 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10093 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10094 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10095 function.
10096 * expr.c (emit_move_resolve_push): Treat the input and result
10097 of PUSH_ROUNDING as a poly_int64.
10098 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10099 (emit_push_insn): Likewise.
10100 * lra-eliminations.c (mark_not_eliminable): Likewise.
10101 * recog.c (push_operand): Likewise.
10102 * reload1.c (elimination_effects): Likewise.
10103 * rtlanal.c (nonzero_bits1): Likewise.
10104 * calls.c (store_one_arg): Likewise. Require the padding to be
10105 known at compile time.
10107 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10108 Alan Hayward <alan.hayward@arm.com>
10109 David Sherwood <david.sherwood@arm.com>
10111 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10112 Use plus_constant instead of gen_rtx_PLUS.
10114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10115 Alan Hayward <alan.hayward@arm.com>
10116 David Sherwood <david.sherwood@arm.com>
10118 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10119 rather than an int.
10121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10122 Alan Hayward <alan.hayward@arm.com>
10123 David Sherwood <david.sherwood@arm.com>
10125 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10126 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10127 via stack temporaries. Treat the mode size as polynomial too.
10129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10130 Alan Hayward <alan.hayward@arm.com>
10131 David Sherwood <david.sherwood@arm.com>
10133 * expr.c (expand_expr_real_2): When handling conversions involving
10134 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10135 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10136 as a poly_uint64 too.
10138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10139 Alan Hayward <alan.hayward@arm.com>
10140 David Sherwood <david.sherwood@arm.com>
10142 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10145 Alan Hayward <alan.hayward@arm.com>
10146 David Sherwood <david.sherwood@arm.com>
10148 * combine.c (can_change_dest_mode): Handle polynomial
10149 REGMODE_NATURAL_SIZE.
10150 * expmed.c (store_bit_field_1): Likewise.
10151 * expr.c (store_constructor): Likewise.
10152 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10153 and polynomial REGMODE_NATURAL_SIZE.
10154 (gen_lowpart_common): Likewise.
10155 * reginfo.c (record_subregs_of_mode): Likewise.
10156 * rtlanal.c (read_modify_subreg_p): Likewise.
10158 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10159 Alan Hayward <alan.hayward@arm.com>
10160 David Sherwood <david.sherwood@arm.com>
10162 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10163 numbers of elements.
10165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10166 Alan Hayward <alan.hayward@arm.com>
10167 David Sherwood <david.sherwood@arm.com>
10169 * match.pd: Cope with polynomial numbers of vector elements.
10171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10172 Alan Hayward <alan.hayward@arm.com>
10173 David Sherwood <david.sherwood@arm.com>
10175 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10176 in a POINTER_PLUS_EXPR.
10178 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10179 Alan Hayward <alan.hayward@arm.com>
10180 David Sherwood <david.sherwood@arm.com>
10182 * omp-simd-clone.c (simd_clone_subparts): New function.
10183 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10184 (ipa_simd_modify_function_body): Likewise.
10186 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10187 Alan Hayward <alan.hayward@arm.com>
10188 David Sherwood <david.sherwood@arm.com>
10190 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10191 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10192 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10193 (expand_vector_condition, vector_element): Likewise.
10194 (subparts_gt): New function.
10195 (get_compute_type): Use subparts_gt.
10196 (count_type_subparts): Delete.
10197 (expand_vector_operations_1): Use subparts_gt instead of
10198 count_type_subparts.
10200 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10201 Alan Hayward <alan.hayward@arm.com>
10202 David Sherwood <david.sherwood@arm.com>
10204 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10205 (vect_compile_time_alias): ...this new function. Do the calculation
10206 on poly_ints rather than trees.
10207 (vect_prune_runtime_alias_test_list): Update call accordingly.
10209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10210 Alan Hayward <alan.hayward@arm.com>
10211 David Sherwood <david.sherwood@arm.com>
10213 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10214 numbers of units.
10215 (vect_schedule_slp_instance): Likewise.
10217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10218 Alan Hayward <alan.hayward@arm.com>
10219 David Sherwood <david.sherwood@arm.com>
10221 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10222 constant and extern definitions for variable-length vectors.
10223 (vect_get_constant_vectors): Note that the number of units
10224 is known to be constant.
10226 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10227 Alan Hayward <alan.hayward@arm.com>
10228 David Sherwood <david.sherwood@arm.com>
10230 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10231 of units as polynomial. Choose between WIDE and NARROW based
10232 on multiple_p.
10234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10235 Alan Hayward <alan.hayward@arm.com>
10236 David Sherwood <david.sherwood@arm.com>
10238 * tree-vect-stmts.c (simd_clone_subparts): New function.
10239 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10242 Alan Hayward <alan.hayward@arm.com>
10243 David Sherwood <david.sherwood@arm.com>
10245 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10246 vectors as polynomial. Use build_index_vector for
10247 IFN_GOMP_SIMD_LANE.
10249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10250 Alan Hayward <alan.hayward@arm.com>
10251 David Sherwood <david.sherwood@arm.com>
10253 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10254 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10255 for variable-length vectors.
10256 (vectorizable_mask_load_store): Treat the number of units as
10257 polynomial, asserting that it is constant if the condition has
10258 already been enforced.
10259 (vectorizable_store, vectorizable_load): Likewise.
10261 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10262 Alan Hayward <alan.hayward@arm.com>
10263 David Sherwood <david.sherwood@arm.com>
10265 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10266 of units as polynomial. Punt if we can't tell at compile time
10267 which vector contains the final result.
10269 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10270 Alan Hayward <alan.hayward@arm.com>
10271 David Sherwood <david.sherwood@arm.com>
10273 * tree-vect-loop.c (vectorizable_induction): Treat the number
10274 of units as polynomial. Punt on SLP inductions. Use an integer
10275 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10276 cast of such a series for variable-length floating-point
10277 reductions.
10279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10280 Alan Hayward <alan.hayward@arm.com>
10281 David Sherwood <david.sherwood@arm.com>
10283 * tree.h (build_index_vector): Declare.
10284 * tree.c (build_index_vector): New function.
10285 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10286 of units as polynomial, forcibly converting it to a constant if
10287 vectorizable_reduction has already enforced the condition.
10288 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10289 to create a {1,2,3,...} vector.
10290 (vectorizable_reduction): Treat the number of units as polynomial.
10291 Choose vectype_in based on the largest scalar element size rather
10292 than the smallest number of units. Enforce the restrictions
10293 relied on above.
10295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10296 Alan Hayward <alan.hayward@arm.com>
10297 David Sherwood <david.sherwood@arm.com>
10299 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10300 number of units as polynomial.
10302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10303 Alan Hayward <alan.hayward@arm.com>
10304 David Sherwood <david.sherwood@arm.com>
10306 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10307 * target.def (autovectorize_vector_sizes): Return the vector sizes
10308 by pointer, using vector_sizes rather than a bitmask.
10309 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10310 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10311 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10312 Likewise.
10313 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10314 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10315 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10316 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10317 * omp-general.c (omp_max_vf): Likewise.
10318 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10319 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10320 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10321 * tree-vect-slp.c (vect_slp_bb): Likewise.
10322 * doc/tm.texi: Regenerate.
10323 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10324 to a poly_uint64.
10325 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10326 the vector size as a poly_uint64 rather than an unsigned int.
10327 (current_vector_size): Change from an unsigned int to a poly_uint64.
10328 (get_vectype_for_scalar_type): Update accordingly.
10329 * tree.h (build_truth_vector_type): Take the size and number of
10330 units as a poly_uint64 rather than an unsigned int.
10331 (build_vector_type): Add a temporary overload that takes
10332 the number of units as a poly_uint64 rather than an unsigned int.
10333 * tree.c (make_vector_type): Likewise.
10334 (build_truth_vector_type): Take the number of units as a poly_uint64
10335 rather than an unsigned int.
10337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10338 Alan Hayward <alan.hayward@arm.com>
10339 David Sherwood <david.sherwood@arm.com>
10341 * target.def (get_mask_mode): Take the number of units and length
10342 as poly_uint64s rather than unsigned ints.
10343 * targhooks.h (default_get_mask_mode): Update accordingly.
10344 * targhooks.c (default_get_mask_mode): Likewise.
10345 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10346 * doc/tm.texi: Regenerate.
10348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10349 Alan Hayward <alan.hayward@arm.com>
10350 David Sherwood <david.sherwood@arm.com>
10352 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10353 * omp-general.c (omp_max_vf): Likewise.
10354 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10355 (expand_omp_simd): Handle polynomial safelen.
10356 * omp-low.c (omplow_simd_context): Add a default constructor.
10357 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10358 (lower_rec_simd_input_clauses): Update accordingly.
10359 (lower_rec_input_clauses): Likewise.
10361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10362 Alan Hayward <alan.hayward@arm.com>
10363 David Sherwood <david.sherwood@arm.com>
10365 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10366 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10367 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10368 (vect_analyze_slp_cost): Likewise.
10369 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10370 (vect_model_load_cost): Likewise.
10372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10373 Alan Hayward <alan.hayward@arm.com>
10374 David Sherwood <david.sherwood@arm.com>
10376 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10377 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10378 from an unsigned int * to a poly_uint64_pod *.
10379 (calculate_unrolling_factor): New function.
10380 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10383 Alan Hayward <alan.hayward@arm.com>
10384 David Sherwood <david.sherwood@arm.com>
10386 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10387 from an unsigned int to a poly_uint64.
10388 (_loop_vec_info::slp_unrolling_factor): Likewise.
10389 (_loop_vec_info::vectorization_factor): Change from an int
10390 to a poly_uint64.
10391 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10392 (vect_get_num_vectors): New function.
10393 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10394 (vect_get_num_copies): Use vect_get_num_vectors.
10395 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10396 to an unsigned int *.
10397 (vect_analyze_data_refs): Change min_vf from an int * to a
10398 poly_uint64 *.
10399 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10400 than an unsigned HOST_WIDE_INT.
10401 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10402 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10403 to an unsigned int *.
10404 (vect_analyze_data_ref_dependences): Likewise.
10405 (vect_compute_data_ref_alignment): Handle polynomial vf.
10406 (vect_enhance_data_refs_alignment): Likewise.
10407 (vect_prune_runtime_alias_test_list): Likewise.
10408 (vect_shift_permute_load_chain): Likewise.
10409 (vect_supportable_dr_alignment): Likewise.
10410 (dependence_distance_ge_vf): Take the vectorization factor as a
10411 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10412 (vect_analyze_data_refs): Change min_vf from an int * to a
10413 poly_uint64 *.
10414 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10415 vfm1 as a poly_uint64 rather than an int. Make the same change
10416 for the returned bound_scalar.
10417 (vect_gen_vector_loop_niters): Handle polynomial vf.
10418 (vect_do_peeling): Likewise. Update call to
10419 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10420 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10421 be constant.
10422 * tree-vect-loop.c (vect_determine_vectorization_factor)
10423 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10424 (vect_get_known_peeling_cost): Likewise.
10425 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10426 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10427 (vect_transform_loop): Likewise. Use the lowest possible VF when
10428 updating the upper bounds of the loop.
10429 (vect_min_worthwhile_factor): Make static. Return an unsigned int
10430 rather than an int.
10431 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10432 polynomial unroll factors.
10433 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10434 (vect_make_slp_decision): Likewise.
10435 (vect_supported_load_permutation_p): Likewise, and polynomial
10436 vf too.
10437 (vect_analyze_slp_cost): Handle polynomial vf.
10438 (vect_slp_analyze_node_operations): Likewise.
10439 (vect_slp_analyze_bb_1): Likewise.
10440 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10441 than an unsigned HOST_WIDE_INT.
10442 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10443 (vectorizable_load): Handle polynomial vf.
10444 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10445 a poly_uint64.
10446 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10448 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10449 Alan Hayward <alan.hayward@arm.com>
10450 David Sherwood <david.sherwood@arm.com>
10452 * match.pd: Handle bit operations involving three constants
10453 and try to fold one pair.
10455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10457 * tree-vect-loop-manip.c: Include gimple-fold.h.
10458 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10459 niters_maybe_zero parameters. Handle other cases besides a step of 1.
10460 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10461 Add a path that uses a step of VF instead of 1, but disable it
10462 for now.
10463 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
10464 and niters_no_overflow parameters. Update calls to
10465 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
10466 Create a new SSA name if the latter choses to use a ste other
10467 than zero, and return it via niters_vector_mult_vf_var.
10468 * tree-vect-loop.c (vect_transform_loop): Update calls to
10469 vect_do_peeling, vect_gen_vector_loop_niters and
10470 slpeel_make_loop_iterate_ntimes.
10471 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
10472 (vect_gen_vector_loop_niters): Update declarations after above changes.
10474 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
10476 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
10477 128-bit round to integer instructions.
10478 (ceil<mode>2): Likewise.
10479 (btrunc<mode>2): Likewise.
10480 (round<mode>2): Likewise.
10482 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10484 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
10485 unaligned VSX load/store on P8/P9.
10486 (expand_block_clear): Allow the use of unaligned VSX
10487 load/store on P8/P9.
10489 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10491 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
10492 New function.
10493 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
10494 swap associated with both a load and a store.
10496 2018-01-02 Andrew Waterman <andrew@sifive.com>
10498 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10499 * config/riscv/riscv.md (clear_cache): Use it.
10501 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
10503 * web.c: Remove out-of-date comment.
10505 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10507 * expr.c (fixup_args_size_notes): Check that any existing
10508 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10509 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10510 (emit_single_push_insn): ...here.
10512 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10514 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10515 (const_vector_encoded_nelts): New function.
10516 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10517 (const_vector_int_elt, const_vector_elt): Declare.
10518 * emit-rtl.c (const_vector_int_elt_1): New function.
10519 (const_vector_elt): Likewise.
10520 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10521 of CONST_VECTOR_ELT.
10523 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10525 * expr.c: Include rtx-vector-builder.h.
10526 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10527 directly on the tree encoding.
10528 (const_vector_from_tree): Likewise.
10529 * optabs.c: Include rtx-vector-builder.h.
10530 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10531 sequence of "u" values.
10532 * vec-perm-indices.c: Include rtx-vector-builder.h.
10533 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10534 directly on the vec_perm_indices encoding.
10536 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10538 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10539 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10540 * rtx-vector-builder.h: New file.
10541 * rtx-vector-builder.c: Likewise.
10542 * rtl.h (rtx_def::u2): Add a const_vector field.
10543 (CONST_VECTOR_NPATTERNS): New macro.
10544 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10545 (CONST_VECTOR_DUPLICATE_P): Likewise.
10546 (CONST_VECTOR_STEPPED_P): Likewise.
10547 (CONST_VECTOR_ENCODED_ELT): Likewise.
10548 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10549 (unwrap_const_vec_duplicate): Likewise.
10550 (const_vec_series_p): Check for a non-duplicated vector encoding.
10551 Say that the function only returns true for integer vectors.
10552 * emit-rtl.c: Include rtx-vector-builder.h.
10553 (gen_const_vec_duplicate_1): Delete.
10554 (gen_const_vector): Call gen_const_vec_duplicate instead of
10555 gen_const_vec_duplicate_1.
10556 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10557 (gen_const_vec_duplicate): Use rtx_vector_builder.
10558 (gen_const_vec_series): Likewise.
10559 (gen_rtx_CONST_VECTOR): Likewise.
10560 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10561 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10562 Build a new vector rather than modifying a CONST_VECTOR in-place.
10563 (handle_special_swappables): Update call accordingly.
10564 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10565 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10566 Build a new vector rather than modifying a CONST_VECTOR in-place.
10567 (handle_special_swappables): Update call accordingly.
10569 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10571 * simplify-rtx.c (simplify_const_binary_operation): Use
10572 CONST_VECTOR_ELT instead of XVECEXP.
10574 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10576 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10577 the selector elements to be different from the data elements
10578 if the selector is a VECTOR_CST.
10579 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10580 ssizetype for the selector.
10582 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10584 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10585 before testing each element individually.
10586 * tree-vect-generic.c (lower_vec_perm): Likewise.
10588 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10590 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10591 * selftest-run-tests.c (selftest::run_tests): Call it.
10592 * vector-builder.h (vector_builder::operator ==): New function.
10593 (vector_builder::operator !=): Likewise.
10594 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10595 (vec_perm_indices::all_from_input_p): New function.
10596 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10597 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10598 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10599 instead of reading the VECTOR_CST directly. Detect whether both
10600 vector inputs are the same before constructing the vec_perm_indices,
10601 and update the number of inputs argument accordingly. Use the
10602 utility functions added above. Only construct sel2 if we need to.
10604 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10606 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10607 the broadcast of the low byte.
10608 (expand_mult_highpart): Use an explicit encoding for the permutes.
10609 * optabs-query.c (can_mult_highpart_p): Likewise.
10610 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10611 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10612 (vectorizable_bswap): Likewise.
10613 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10614 explicit encoding for the power-of-2 permutes.
10615 (vect_permute_store_chain): Likewise.
10616 (vect_grouped_load_supported): Likewise.
10617 (vect_permute_load_chain): Likewise.
10619 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10621 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10622 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10623 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10624 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10625 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10626 (vect_gen_perm_mask_any): Likewise.
10628 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10630 * int-vector-builder.h: New file.
10631 * vec-perm-indices.h: Include int-vector-builder.h.
10632 (vec_perm_indices): Redefine as an int_vector_builder.
10633 (auto_vec_perm_indices): Delete.
10634 (vec_perm_builder): Redefine as a stand-alone class.
10635 (vec_perm_indices::vec_perm_indices): New function.
10636 (vec_perm_indices::clamp): Likewise.
10637 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10638 (vec_perm_indices::new_vector): New function.
10639 (vec_perm_indices::new_expanded_vector): Update for new
10640 vec_perm_indices class.
10641 (vec_perm_indices::rotate_inputs): New function.
10642 (vec_perm_indices::all_in_range_p): Operate directly on the
10643 encoded form, without computing elided elements.
10644 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10645 encoding. Update for new vec_perm_indices class.
10646 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10647 the given vec_perm_builder.
10648 (expand_vec_perm_var): Update vec_perm_builder constructor.
10649 (expand_mult_highpart): Use vec_perm_builder instead of
10650 auto_vec_perm_indices.
10651 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10652 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10653 or double series encoding as appropriate.
10654 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10655 vec_perm_indices instead of auto_vec_perm_indices.
10656 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10657 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10658 (vect_permute_store_chain): Likewise.
10659 (vect_grouped_load_supported): Likewise.
10660 (vect_permute_load_chain): Likewise.
10661 (vect_shift_permute_load_chain): Likewise.
10662 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10663 (vect_transform_slp_perm_load): Likewise.
10664 (vect_schedule_slp_instance): Likewise.
10665 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10666 (vectorizable_mask_load_store): Likewise.
10667 (vectorizable_bswap): Likewise.
10668 (vectorizable_store): Likewise.
10669 (vectorizable_load): Likewise.
10670 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10671 vec_perm_indices instead of auto_vec_perm_indices. Use
10672 tree_to_vec_perm_builder to read the vector from a tree.
10673 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10674 vec_perm_builder instead of a vec_perm_indices.
10675 (have_whole_vector_shift): Use vec_perm_builder and
10676 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10677 truncation to calc_vec_perm_mask_for_shift.
10678 (vect_create_epilog_for_reduction): Likewise.
10679 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10680 from auto_vec_perm_indices to vec_perm_indices.
10681 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10682 instead of changing individual elements.
10683 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10684 the vector in d.perm.
10685 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10686 from auto_vec_perm_indices to vec_perm_indices.
10687 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10688 instead of changing individual elements.
10689 (arm_vectorize_vec_perm_const): Use new_vector to install
10690 the vector in d.perm.
10691 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10692 Update vec_perm_builder constructor.
10693 (rs6000_expand_interleave): Likewise.
10694 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10695 (rs6000_expand_interleave): Likewise.
10697 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10699 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10700 to qimode could truncate the indices.
10701 * optabs.c (expand_vec_perm_var): Likewise.
10703 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10705 * Makefile.in (OBJS): Add vec-perm-indices.o.
10706 * vec-perm-indices.h: New file.
10707 * vec-perm-indices.c: Likewise.
10708 * target.h (vec_perm_indices): Replace with a forward class
10709 declaration.
10710 (auto_vec_perm_indices): Move to vec-perm-indices.h.
10711 * optabs.h: Include vec-perm-indices.h.
10712 (expand_vec_perm): Delete.
10713 (selector_fits_mode_p, expand_vec_perm_var): Declare.
10714 (expand_vec_perm_const): Declare.
10715 * target.def (vec_perm_const_ok): Replace with...
10716 (vec_perm_const): ...this new hook.
10717 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10718 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10719 * doc/tm.texi: Regenerate.
10720 * optabs.def (vec_perm_const): Delete.
10721 * doc/md.texi (vec_perm_const): Likewise.
10722 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10723 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10724 expand_vec_perm for constant permutation vectors. Assert that
10725 the mode of variable permutation vectors is the integer equivalent
10726 of the mode that is being permuted.
10727 * optabs-query.h (selector_fits_mode_p): Declare.
10728 * optabs-query.c: Include vec-perm-indices.h.
10729 (selector_fits_mode_p): New function.
10730 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10731 is defined, instead of checking whether the vec_perm_const_optab
10732 exists. Use targetm.vectorize.vec_perm_const instead of
10733 targetm.vectorize.vec_perm_const_ok. Check whether the indices
10734 fit in the vector mode before using a variable permute.
10735 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10736 vec_perm_indices instead of an rtx.
10737 (expand_vec_perm): Replace with...
10738 (expand_vec_perm_const): ...this new function. Take the selector
10739 as a vec_perm_indices rather than an rtx. Also take the mode of
10740 the selector. Update call to shift_amt_for_vec_perm_mask.
10741 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10742 Use vec_perm_indices::new_expanded_vector to expand the original
10743 selector into bytes. Check whether the indices fit in the vector
10744 mode before using a variable permute.
10745 (expand_vec_perm_var): Make global.
10746 (expand_mult_highpart): Use expand_vec_perm_const.
10747 * fold-const.c: Includes vec-perm-indices.h.
10748 * tree-ssa-forwprop.c: Likewise.
10749 * tree-vect-data-refs.c: Likewise.
10750 * tree-vect-generic.c: Likewise.
10751 * tree-vect-loop.c: Likewise.
10752 * tree-vect-slp.c: Likewise.
10753 * tree-vect-stmts.c: Likewise.
10754 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
10755 Delete.
10756 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
10757 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
10758 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
10759 (aarch64_vectorize_vec_perm_const): ...this new function.
10760 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10761 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10762 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
10763 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
10764 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10765 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10766 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
10767 into...
10768 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
10769 check for NEON modes.
10770 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
10771 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
10772 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
10773 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
10774 into...
10775 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
10776 the old VEC_PERM_CONST conditions.
10777 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
10778 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
10779 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
10780 (ia64_vectorize_vec_perm_const_ok): Merge into...
10781 (ia64_vectorize_vec_perm_const): ...this new function.
10782 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
10783 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
10784 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
10785 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
10786 * config/mips/mips.c (mips_expand_vec_perm_const)
10787 (mips_vectorize_vec_perm_const_ok): Merge into...
10788 (mips_vectorize_vec_perm_const): ...this new function.
10789 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
10790 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
10791 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
10792 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
10793 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
10794 (rs6000_expand_vec_perm_const): Delete.
10795 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
10796 Delete.
10797 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10798 (altivec_expand_vec_perm_const_le): Take each operand individually.
10799 Operate on constant selectors rather than rtxes.
10800 (altivec_expand_vec_perm_const): Likewise. Update call to
10801 altivec_expand_vec_perm_const_le.
10802 (rs6000_expand_vec_perm_const): Delete.
10803 (rs6000_vectorize_vec_perm_const_ok): Delete.
10804 (rs6000_vectorize_vec_perm_const): New function.
10805 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10806 an element count and rtx array.
10807 (rs6000_expand_extract_even): Update call accordingly.
10808 (rs6000_expand_interleave): Likewise.
10809 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10810 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10811 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10812 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10813 (rs6000_expand_vec_perm_const): Delete.
10814 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10815 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10816 (altivec_expand_vec_perm_const_le): Take each operand individually.
10817 Operate on constant selectors rather than rtxes.
10818 (altivec_expand_vec_perm_const): Likewise. Update call to
10819 altivec_expand_vec_perm_const_le.
10820 (rs6000_expand_vec_perm_const): Delete.
10821 (rs6000_vectorize_vec_perm_const_ok): Delete.
10822 (rs6000_vectorize_vec_perm_const): New function. Remove stray
10823 reference to the SPE evmerge intructions.
10824 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10825 an element count and rtx array.
10826 (rs6000_expand_extract_even): Update call accordingly.
10827 (rs6000_expand_interleave): Likewise.
10828 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10829 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10830 new function.
10831 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10833 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10835 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10836 vector mode and that that mode matches the mode of the data
10837 being permuted.
10838 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10839 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
10840 directly using expand_vec_perm_1 when forcing selectors into
10841 registers.
10842 (expand_vec_perm_var): New function, split out from expand_vec_perm.
10844 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10846 * optabs-query.h (can_vec_perm_p): Delete.
10847 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10848 * optabs-query.c (can_vec_perm_p): Split into...
10849 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10850 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10851 particular selector is valid.
10852 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10853 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10854 (vect_grouped_load_supported): Likewise.
10855 (vect_shift_permute_load_chain): Likewise.
10856 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10857 (vect_transform_slp_perm_load): Likewise.
10858 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10859 (vectorizable_bswap): Likewise.
10860 (vect_gen_perm_mask_checked): Likewise.
10861 * fold-const.c (fold_ternary_loc): Likewise. Don't take
10862 implementations of variable permutation vectors into account
10863 when deciding which selector to use.
10864 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10865 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10866 with a false third argument.
10867 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10868 to test whether the constant selector is valid and can_vec_perm_var_p
10869 to test whether a variable selector is valid.
10871 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10873 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10874 * optabs-query.c (can_vec_perm_p): Likewise.
10875 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10876 instead of vec_perm_indices.
10877 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10878 (vect_gen_perm_mask_checked): Likewise,
10879 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10880 (vect_gen_perm_mask_checked): Likewise,
10882 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10884 * optabs-query.h (qimode_for_vec_perm): Declare.
10885 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10886 (qimode_for_vec_perm): ...this new function.
10887 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10889 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10891 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10892 does not have a conditional at the top.
10894 2018-01-02 Richard Biener <rguenther@suse.de>
10896 * ipa-inline.c (big_speedup_p): Fix expression.
10898 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10900 PR target/81616
10901 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10902 for generic 4->6.
10904 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10906 PR target/81616
10907 Generic tuning.
10908 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10909 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10910 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10911 cond_taken_branch_cost 3->4.
10913 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10915 PR tree-optimization/83581
10916 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10917 TODO_cleanup_cfg if any changes have been made.
10919 PR middle-end/83608
10920 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10921 convert_modes if target mode has the right side, but different mode
10922 class.
10924 PR middle-end/83609
10925 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10926 last argument when extracting from CONCAT. If either from_real or
10927 from_imag is NULL, use expansion through memory. If result is not
10928 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10929 the parts directly to inner mode, if even that fails, use expansion
10930 through memory.
10932 PR middle-end/83623
10933 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10934 check for bswap in mode rather than HImode and use that in expand_unop
10935 too.
10936 \f
10937 Copyright (C) 2018 Free Software Foundation, Inc.
10939 Copying and distribution of this file, with or without modification,
10940 are permitted in any medium without royalty provided the copyright
10941 notice and this notice are preserved.