| blob | f98824ec3028ca7d2f9a5758306af9604c00dfa7 |
1 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4 (=v, v) alternative and explicit "memory" attribute.
5 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7 attributes.
8 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
9 "sselog1" type instead of "sselog".
10 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
11 "sselog". Remove explicit "memory" attribute.
12 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
13 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14 attributes.
15 (vec_extract_hi_v32hi): Merge all alternatives into one, use
16 "sselog1" type instead of "sselog". Remove explicit "memory"
17 attribute.
18 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
19 use "sselog1" type instead of "sselog". Remove explicit "memory"
20 attribute.
21 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
22 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
23 attributes.
24 (vec_extract_hi_v64qi): Merge all alternatives into one, use
25 "sselog1" type instead of "sselog". Remove explicit "memory"
26 attribute.
27 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
28 use "sselog1" type instead of "sselog". Remove explicit "memory"
29 attribute.
31 PR target/85430
32 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
34 PR middle-end/85414
35 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
36 on a SUBREG.
38 2018-04-17 Martin Jambor <mjambor@suse.cz>
40 PR ipa/85421
41 * ipa-cp.c (create_specialized_node): Call
42 expand_all_artificial_thunks if necessary.
44 2018-04-17 Martin Liska <mliska@suse.cz>
46 PR lto/85405
47 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
48 in message, remote space in between '_G' and '('.
50 2018-04-17 Jakub Jelinek <jakub@redhat.com>
52 PR target/85281
53 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
54 avx512f_vmcmp<mode>3<round_saeonly_name>,
55 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
56 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
57 avx512f_rndscale<mode><round_saeonly_name>,
58 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
59 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
60 Use %<iptr>2 instead of %2 for -masm=intel.
61 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
62 avx512f_vcvttss2usi<round_saeonly_name>,
63 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
64 -masm=intel.
65 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
66 avx512f_vcvttsd2usi<round_saeonly_name>,
67 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
68 Use %q1 instead of %1 for -masm=intel.
69 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
70 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
71 of %3 for -masm=intel.
72 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
73 -masm=intel.
74 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
75 -masm=intel.
76 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
77 -masm=intel.
78 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
79 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
80 %g1.
81 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
82 -masm=intel.
83 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
84 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
85 %g1 and one with %0 and %1.
86 (avx512er_vmrcp28<mode><round_saeonly_name>,
87 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
88 %1 for -masm=intel.
89 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
90 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
91 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
92 of %0 and %{%4%} for -masm=intel.
93 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
94 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
95 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
96 order of %0 and %{%5%}%{z%} for -masm=intel.
98 2018-04-17 Jan Hubicka <jh@suse.cz>
100 PR lto/85405
101 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
103 2018-04-17 Martin Liska <mliska@suse.cz>
105 PR ipa/85329
106 * multiple_target.c (create_dispatcher_calls): Set apostrophes
107 for target_clone error message. Make default implementation
108 clone to be a local declaration.
109 (separate_attrs): Add new argument and check for an empty
110 string.
111 (expand_target_clones): Handle it.
112 (ipa_target_clone): Make redirection just for target_clones
113 functions.
115 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
116 Tom de Vries <tom@codesourcery.com>
118 PR middle-end/84955
119 * omp-expand.c (expand_oacc_for): Add dummy false branch for
120 tiled basic blocks without omp continue statements.
122 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
124 PR target/83660
125 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
126 vec_extract expression as having side effects to make sure it gets
127 a cleanup point.
129 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
131 PR target/85403
132 * config/i386/i386.c (get_builtin_code_for_version): Check
133 error_mark_node.
135 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
137 PR target/84331
138 * gcc/config.gcc: Support "skylake".
139 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
140 PROCESSOR_SKYLAKE.
141 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
142 (processor_target_table): Add "skylake".
143 (ix86_option_override_internal): Add "skylake".
144 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
145 PROCESSOR_CANNONLAKE.
146 (get_builtin_code_for_version): Fix priority for
147 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
148 PROCESSOR_SKYLAKE-AVX512.
149 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
150 (processor_type): Add PROCESSOR_SKYLAKE.
152 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
153 Jason Merrill <jason@redhat.com>
155 PR c++/85112
156 * convert.c (convert_to_integer_1): Use direct recursion for
157 enumeral types and types with a precision less than the number
158 of bits in their mode.
160 2018-04-16 Julia Koval <julia.koval@intel.com>
162 PR target/84413
163 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
164 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
166 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
168 PR target/85293
169 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
170 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
171 and -mno-direct-move.
173 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
175 PR target/83402
176 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
177 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
178 Ensure negative shifts result in {0}.
180 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
182 PR rtl-optimization/79916
183 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
184 regs (if any) to define how to gnerate SD moves when LRA is in
185 progress.
187 2018-04-13 Jakub Jelinek <jakub@redhat.com>
189 PR rtl-optimization/85393
190 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
191 * except.c (expand_dw2_landing_pad_for_region): Make static.
192 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
193 a label and unconditional jump to old_bb, rather than
194 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
195 basic block.
197 PR rtl-optimization/85376
198 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
199 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
200 instead of a specific value.
202 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
203 Bin Cheng <bin.cheng@arm.com>
205 PR tree-optimization/82965
206 PR tree-optimization/83991
207 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
208 by_profile_only parameter.
209 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
210 information if the loop was predicted to iterate too many times.
211 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
213 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
215 PR lto/71991
216 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
217 always inline.
219 2018-04-13 Martin Liska <mliska@suse.cz>
220 Jakub Jelinek <jakub@redhat.com>
222 PR middle-end/81657
223 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
224 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
225 * builtins.c (expand_builtin_memory_copy_args): Use
226 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
227 handle dest_addr == pc_rtx.
229 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
231 PR target/85291
232 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
233 asked to not generate direct moves.
234 (fix_trunc<mode>si2_stfiwx): Similar.
235 (fix_trunc<mode>si2_internal): Similar.
237 2018-04-12 Jakub Jelinek <jakub@redhat.com>
239 PR debug/83157
240 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
241 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
242 lookup if dest in some wider mode is known to be const0_rtx and
243 if so, record permanent equivalence for it to be ZERO_EXTEND of
244 the narrower mode destination.
246 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
248 * lto-streamer-out.c (output_function): Revert 259346.
249 * omp-expand.c (expand_oacc_for): Likewise.
251 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
253 PR rtl-optimization/85354
254 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
255 * sel-sched.c (sel_global_init): ... here.
257 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
259 PR target/85238
260 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
261 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
262 mode for PE-COFF targets.
263 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
264 (i386_pe_asm_lto_end): Likewise.
265 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
266 (TARGET_ASM_LTO_END): Likewise.
267 * config/i386/winnt.c (saved_debug_info_level): New static variable.
268 (i386_pe_asm_lto_start): New function.
269 (i386_pe_asm_lto_end): Likewise.
271 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
272 Richard Biener <rguenther@suse.de>
274 PR middle-end/84955
275 * lto-streamer-out.c (output_function): Fix CFG loop state before
276 streaming out.
277 * omp-expand.c (expand_oacc_for): Handle calls to internal
278 functions like regular functions.
280 2018-04-12 Richard Biener <rguenther@suse.de>
282 PR lto/85371
283 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
284 for the early LTO debug to properly generate references to it
285 during DIE emission. Do not re-use that for the skeleton for
286 split-dwarf.
287 (dwarf2out_early_finish): Likewise.
289 2018-04-12 Jakub Jelinek <jakub@redhat.com>
291 PR target/85328
292 * config/i386/sse.md
293 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
294 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
295 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
296 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
297 and output is a reg, avoid creating invalid lowpart subreg, but
298 instead split into a 512-bit move. Don't split if not AVX512VL,
299 input is xmm16+ reg and output is a mem.
300 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
301 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
302 xmm16+ reg and output is a mem.
304 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
306 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
307 also for flag_dwarf2_cfi_asm.
309 2018-04-12 Jakub Jelinek <jakub@redhat.com>
311 PR rtl-optimization/85342
312 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
313 a bool scalar var inside of the loop instead. Don't try to update
314 recog_data.operand after failed apply_change_group.
316 2018-04-12 Tom de Vries <tom@codesourcery.com>
318 PR target/85296
319 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
320 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
321 array with flexible array member as array without given dimension.
322 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
323 argument for undefined param to true.
325 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
327 PR target/85321
328 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
329 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
330 from PowerPC section.
331 * config/rs6000/sysv4.opt (mcall-): Improve help text.
332 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
333 help text that is too long.
334 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
335 help text that is too long.
336 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
337 help text that is too long.
339 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
341 * config/alpha/alpha.md (stack_probe_internal): Rename
342 from "probe_stack". Update all callers.
344 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
346 PR rtl-optimization/84566
347 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
348 sched_macro_fuse_insns.
350 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
352 PR target/84301
353 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
354 (compute_block_dependences): ... from here.
356 2018-04-11 Jakub Jelinek <jakub@redhat.com>
358 PR tree-optimization/85331
359 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
360 from int to HOST_WIDE_INT.
362 2018-04-11 Martin Jambor <mjambor@suse.cz>
364 PR ipa/84149
365 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
366 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
367 not the same as the source val.
368 (cgraph_edge_brings_value_p): New parameter.
369 (gather_edges_for_value): Pass destination value to
370 cgraph_edge_brings_value_p.
371 (perhaps_add_new_callers): Likewise.
372 (get_info_about_necessary_edges): Likewise and exclude values brought
373 only by self-recursive edges.
374 (create_specialized_node): Redirect only clones of self-calling edges.
375 (+self_recursive_pass_through_p): New function.
376 (find_more_scalar_values_for_callers_subset): Use it.
377 (find_aggregate_values_for_callers_subset): Likewise.
378 (known_aggs_to_agg_replacement_list): Removed.
379 (decide_whether_version_node): Re-calculate known constants for all
380 remaining context clones.
382 2018-04-11 Richard Biener <rguenther@suse.de>
384 PR lto/85339
385 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
386 from early DWARF output.
387 (dwarf2out_early_finish): Output line info unconditionally into
388 early DWARF and add reference to it.
390 2018-04-11 Jakub Jelinek <jakub@redhat.com>
392 PR target/85281
393 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
394 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
395 other than V2DFmode using iptr mode attribute.
396 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
398 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
400 PR rtl-optimization/84659
401 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
403 2018-04-11 Jakub Jelinek <jakub@redhat.com>
405 PR debug/85302
406 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
407 SIZEP is NULL.
408 (output_loc_list): Pass address of a dummy size variable even in the
409 locview handling loop.
410 (index_location_lists): Add comment on why skip_loc_list_entry can't
411 call size_of_locs.
413 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
415 PR target/85261
416 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
417 into register.
419 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
421 PR target/85321
422 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
423 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
424 and -mstring-compare-inline-limit.
426 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
428 PR target/85287
429 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
430 for stack clash protection in a register whenever we need it to be in
431 a register.
433 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
435 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
436 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
438 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
440 PR target/85321
441 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
442 the help text.
443 (mlong-double-): Ditto.
444 * config/rs6000/sysv4.opt (msdata=): Ditto.
445 (mtls-size=): Ditto.
447 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
449 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
450 erroneous entries for
451 "vector int vec_ldl (int, long int *)", and
452 "vector unsigned int vec_ldl (int, unsigned long int *)".
453 Add comments and entries for
454 "vector bool char vec_ldl (int, bool char *)",
455 "vector bool short vec_ldl (int, bool short *)",
456 "vector bool int vec_ldl (int, bool int *)",
457 "vector bool long long vec_ldl (int, bool long long *)",
458 "vector pixel vec_ldl (int, pixel *)",
459 "vector long long vec_ldl (int, long long *)",
460 "vector unsigned long long vec_ldl (int, unsigned long long *)".
461 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
462 type tree bool_long_long_type_node and correct definition of
463 bool_V2DI_type_node to make reference to this new type tree.
464 (rs6000_mangle_type): Replace erroneous reference to
465 bool_long_type_node with bool_long_long_type_node.
466 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
467 comments to emphasize sign distinctions for char and int types and
468 replace RS6000_BTI_bool_long constant with
469 RS6000_BTI_bool_long_long constant. Also add comment to restrict
470 use of RS6000_BTI_pixel.
471 (bool_long_type_node): Remove this macro definition.
472 (bool_long_long_type_node): New macro definition
474 2018-04-10 Jakub Jelinek <jakub@redhat.com>
476 PR rtl-optimization/85300
477 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
478 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
479 simplify_unary_operation fails.
481 2018-04-10 Martin Liska <mliska@suse.cz>
483 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
484 cgraph_edge and ipa_ref.
486 2018-04-10 Jakub Jelinek <jakub@redhat.com>
488 PR target/85177
489 PR target/85255
490 * config/i386/sse.md
491 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
492 computation of the VEC_MERGE selector from mask.
493 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
494 Fix decoding of the VEC_MERGE selector into mask.
496 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
498 PR tree-optimization/85286
499 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
501 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
503 * final.c (final_1): Set insn_last_address as well as
504 insn_current_address.
506 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
508 PR target/85173
509 * explow.c (emit_stack_probe): Call validize_mem on memory location
510 before passing it to gen_probe_stack. Create address operand and
511 legitimize it for the probe_stack_address case.
513 2018-04-09 Jan Hubicka <jh@suse.cz>
515 PR lto/85078
516 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
517 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
518 * tree.c (free_lang_data_in_type): Fix handling of binfos;
519 walk basetypes.
520 (free_lang_data): Rebuild type inheritance graph.
522 2018-04-09 Martin Sebor <msebor@redhat.com>
524 * invoke.texi (-finline-small-functions): Mention other optimization
525 options.
526 (-findirect-inlining, -fpartial-inlining): Same.
527 (-finline-functions-called-once): Same.
528 (-freorder-blocks-and-partition): Same.
530 2018-04-09 Jan Hubicka <jh@suse.cz>
532 PR rtl/84058
533 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
534 jumps; choose last target that matches the criteria (i.e.
535 no partition changes for non-crossing jumps).
536 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
537 support for redirecting crossing jumps to non-crossing.
539 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
541 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
542 also for naked functions.
544 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
546 * config/arc/arc.md (add_shift): New pattern.
547 (add_shift2): Likewise.
548 (sub_shift): Likewise.
549 (sub_shift_cmp0_noout): Likewise.
550 (compare_si_ashiftsi): Likewise.
551 (xbfu_cmp0_noout): New combine pattern.
552 (xbfu_cmp0"): Likewise.
553 (movsi_set_cc_insn): Place the predicable variant first.
554 (commutative_binary_cmp0_noout): Remove clobber.
555 (commutative_binary_cmp0): New pattern.
556 (noncommutative_binary_cmp0): Likewise.
557 (noncommutative_binary_cmp0_noout): Likewise.
558 (noncommutative_binary_comparison_result_used): Removed.
559 (rsub_cmp0): New pattern.
560 (rsub_cmp0_noout): Likewise.
561 (extzvsi): Changed, keep only meaningful variants.
562 (SQH, SEZ): New iterators.
563 (SQH_postfix): New mode attribute.
564 (SEZ_prefix): New code attribute.
565 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
566 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
567 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
568 of numerical value.
569 (noncommutative_operator): Check the availability of barrel
570 shifter option.
572 2018-04-09 Richard Biener <rguenther@suse.de>
574 PR tree-optimization/85284
575 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
576 Only use the niter constraining form of simple_iv when the exit
577 is always executed.
579 2018-04-09 Tom de Vries <tom@codesourcery.com>
581 PR target/84041
582 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
583 (define_expand "*memory_barrier"): New define_expand.
584 (define_insn "memory_barrier"): New insn.
586 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
588 PR rtl-optimization/80463
589 PR rtl-optimization/83972
590 PR rtl-optimization/83480
592 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
593 correct producer for the insn.
594 (tidy_control_flow): Fixup seqnos in case of debug insns.
596 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
598 PR rtl-optimization/83913
600 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
601 different sched-times when merging exprs.
603 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
605 PR rtl-optimization/83962
607 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
608 tidy_fallthru_edge and tidy_control_flow.
610 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
612 PR rtl-optimization/83530
614 * sel-sched.c (force_next_insn): New global variable.
615 (remove_insn_for_debug): When force_next_insn is true, also leave only
616 next insn in the ready list.
617 (sel_sched_region): When the region wasn't scheduled, make another pass
618 over it with force_next_insn set to 1.
620 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
622 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
623 into tm_file.
624 * config/nds32/constants.md (unspec_volatile_element): Add enum values
625 for interrupt control.
626 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
627 functions for interrupt control.
628 * config/nds32/nds32-intrinsic.md: Likewise.
629 * config/nds32/nds32_intrinsic.h: Likewise.
630 * config/nds32/nds32.h (nds32_builtins): Likewise.
632 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
634 * config/nds32/nds32.c (nds32_init_machine_status,
635 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
636 strict_aligned_p field.
637 (nds32_expand_to_rtl_hook): New function.
638 (TARGET_EXPAND_TO_RTL_HOOK): Define.
639 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
641 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
642 Chung-Ju Wu <jasonwucj@gmail.com>
644 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
645 * config/nds32/nds32-n7.md: New file.
646 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
647 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
648 pipeline.
649 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
650 * config/nds32/nds32.md (pipeline_model): Add n7.
651 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
652 * config/nds32/pipelines.md: Include n7 settings.
654 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
655 Chung-Ju Wu <jasonwucj@gmail.com>
657 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
658 * config/nds32/nds32-e8.md: New file.
659 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
660 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
661 pipeline.
662 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
663 * config/nds32/nds32.md (pipeline_model): Add e8.
664 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
665 * config/nds32/pipelines.md: Include e8 settings.
667 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
668 Chung-Ju Wu <jasonwucj@gmail.com>
670 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
671 * config/nds32/nds32-n8.md: New file.
672 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
673 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
674 pipeline.
675 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
676 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
677 * config/nds32/nds32.md (pipeline_model): Add n8.
678 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
679 * config/nds32/pipelines.md: Include n8 settings.
681 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
682 Chung-Ju Wu <jasonwucj@gmail.com>
684 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
685 * config/nds32/nds32-n9-2r1w.md: New file.
686 * config/nds32/nds32-n9-3r2w.md: New file.
687 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
688 nds32_register_ports): New or modify for cpu n9.
689 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
690 pipeline.
691 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
692 * config/nds32/nds32-utils.c: New file.
693 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
694 TARGET_MUL_SLOW): Define.
695 * config/nds32/nds32.md (pipeline_model): New attribute.
696 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
697 New options that support cpu n9.
698 * config/nds32/pipelines.md: Include n9 settings.
699 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
701 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
703 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
704 information if necessary.
705 (output_cond_branch_compare_zero): Likewise.
706 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
707 (nds32_target_alignment): Refine for alignment.
708 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
709 (FUNCTION_BOUNDARY): Modify.
710 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
711 align case.
712 * config/nds32/nds32.opt (malways-align, malign-functions): New.
714 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
716 * config/nds32/constants.md (unspec_volatile_element): Add values for
717 TLB operation and data prefetch.
718 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
719 functions for TLB operation and data prefetch.
720 * config/nds32/nds32-intrinsic.md: Likewise.
721 * config/nds32/nds32_intrinsic.h: Likewise.
722 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
723 (nds32_print_operand): Likewise.
724 * config/nds32/nds32.h (nds32_builtins): Likewise.
726 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
727 Andrew Pinski <pinsika@gcc.gnu.org>
729 PR middle-end/82976
730 * match.pd: Use constant_boolean_node of correct type instead of
731 boolean_true_node or boolean_false_node for simplifying
732 pointer comparisons to zero.
734 2018-04-07 Jakub Jelinek <jakub@redhat.com>
736 PR tree-optimization/80021
737 * tree.c (verify_type_variant): Make error call in verify_variant_match
738 translatable and remove final full stop.
740 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
742 * config/nds32/constants.md (unspec_volatile_element): Add
743 UNSPEC_VOLATILE_EH_RETURN.
744 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
745 nds32_output_stack_pop): Support dwarf exception handling process.
746 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
747 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
748 exception handling process.
749 (nds32_compute_stack_frame): Likewise.
750 (nds32_return_addr_rtx): Likewise.
751 (nds32_initial_elimination_offset): Likewise.
752 (nds32_expand_prologue): Likewise.
753 (nds32_expand_epilogue): Likewise.
754 (nds32_dynamic_chain_address): New function.
755 * config/nds32/nds32.h (machine_function): Add fields for dwarf
756 exception handling.
757 (DYNAMIC_CHAIN_ADDRESS): Define.
758 (EH_RETURN_DATA_REGNO): Define.
759 (EH_RETURN_STACKADJ_RTX): Define.
760 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
761 patterns for dwarf exception handling.
763 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
765 * config/nds32/nds32.h: Clean up obsolete macros.
767 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
769 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
770 Add enum values for particular instructions.
771 * config/nds32/nds32-intrinsic.c: Implementation of expanding
772 particular intrinsic functions.
773 * config/nds32/nds32-intrinsic.md: Likewise.
774 * config/nds32/nds32_intrinsic.h: Likewise.
775 * config/nds32/nds32.h (nds32_builtins): Likewise.
776 * config/nds32/nds32.md (type): Add pbsad and pbsada.
777 (btst, ave): New patterns for particular instructions.
779 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
781 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
782 Add enum values for atomic load/store and memory sync.
783 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
784 and memory sync.
785 * config/nds32/nds32-intrinsic.md: Likewise.
786 * config/nds32/nds32_intrinsic.h: Likewise.
787 * config/nds32/nds32.h (nds32_builtins): Likewise.
789 2018-04-07 Jakub Jelinek <jakub@redhat.com>
791 PR tree-optimization/85257
792 * fold-const.c (native_encode_vector): If not all elts could fit
793 and off is -1, return 0 rather than offset.
794 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
795 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
796 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
797 adjust buffer in native_interpret_expr call.
799 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
801 * config/nds32/constants.md (unspec_volatile_element): Add cache
802 control enum values.
803 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
804 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
805 * config/nds32/nds32.c (nds32_cctl_names): New.
806 (nds32_print_operand): Handle cache control register names.
807 * config/nds32/nds32.h (nds32_builtins): New enum values.
808 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
809 macros.
810 * config/nds32/nds32.md (type): Add mmu.
811 * config/nds32/pipelines.md (simple_insn): Add mmu.
813 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
815 * config/nds32/nds32.md (type): Remove call.
816 * config/nds32/pipelines.md (simple_insn): Likewise.
818 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
820 * config/nds32/constants.md (unspec_volatile_element): Add
821 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
822 UNSPEC_VOLATILE_FMFCFG.
823 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
824 description for fmfcfg and fmfcsr.
825 (bdesc_1arg): Add fmtcsr.
826 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
827 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
828 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
829 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
830 unspec_fmfcfg): New patterns.
831 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
832 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
833 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
834 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
835 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
836 __nds32__fmfcfg): Define.
838 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
840 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
841 intrinsic register names.
842 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
843 intrinsic register enum values and macros.
845 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
847 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
848 for load/store addressing form.
849 (nds32_print_operand_address): Likewise.
851 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
853 PR target/85196
854 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
855 based on LABEL_REF. Remove useless assertion.
856 (pic_address_needs_scratch): Fix formatting.
857 (sparc_legitimize_pic_address): Minor tweaks.
858 (sparc_delegitimize_address): Adjust assertion accordingly.
859 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
860 into symbolic_operand.
861 (movsi_high_pic_label_ref): Likewise.
862 (movsi_lo_sum_pic_label_ref): Likewise.
863 (movdi_pic_label_ref): Likewise.
864 (movdi_high_pic_label_ref): Likewise.
865 (movdi_lo_sum_pic_label_ref): Likewise.
867 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
869 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
870 custom LIB_SPEC setup.
872 2018-04-06 Ruslan Bukin <br@bsdpad.com>
873 Kito Cheng <kito.cheng@gmail.com>
875 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
876 * config/riscv/freebsd.h: New.
878 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
880 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
881 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
882 file.
884 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
885 Kito Cheng <kito.cheng@gmail.com>
887 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
888 nds32_output_call, nds32_symbol_binds_local_p): New functions.
889 * config/nds32/nds32-protos.h (nds32_output_call,
890 nds32_output_return): Declare.
891 * config/nds32/nds32.md: Refine all the call and return patterns.
893 2018-04-06 Jakub Jelinek <jakub@redhat.com>
895 PR debug/85252
896 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
897 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
899 PR rtl-optimization/84872
900 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
901 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
902 EDGE_CROSSING edge.
904 2018-04-06 Tamar Christina <tamar.christina@arm.com>
906 * expr.c (copy_blkmode_to_reg): Revert 254862.
907 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
909 2018-04-06 Richard Biener <rguenther@suse.de>
911 PR middle-end/85244
912 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
913 after seeing a component reference with an adjacent field. Treat
914 refs to arrays at struct end of external decls similar to
915 refs to unconstrained commons.
917 2018-04-06 Jakub Jelinek <jakub@redhat.com>
919 PR sanitizer/85213
920 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
921 look through SAVE_EXPRs with non-side-effects argument. Adjust
922 recursive calls.
923 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
924 save_p here.
926 2018-04-06 Richard Biener <rguenther@suse.de>
928 PR middle-end/85180
929 * alias.c (find_base_term): New wrapper around find_base_term
930 unwinding CSELIB_VAL_PTR changes.
931 (find_base_term): Do not restore CSELIB_VAL_PTR during the
932 recursion.
934 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
936 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
937 instructions.
938 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
939 constant definitions.
940 ("nop"): lr 0,0 -> nopr r0
941 ("nop_lr0", "nop_lr1"): New insn definitions.
943 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
945 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
946 NDS32_V3PUSH_AVAILABLE_P macro.
948 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
949 Chung-Ju Wu <jasonwucj@gmail.com>
951 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
952 (nds32*-*-*): Add float and fpu_config into supported_defaults.
953 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
954 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
955 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
956 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
957 * config/nds32/constraints.md: New constraints and checking for hard
958 float configuration.
959 * config/nds32/iterators.md: New mode iterator and attribute for hard
960 float configuration.
961 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
962 patterns.
963 * config/nds32/nds32-fpu.md: New file.
964 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
965 deal with hard float code generation.
966 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
967 ARCH_V3S.
968 (abi_type, float_reg_number): New enum type.
969 * config/nds32/nds32-predicates.c: New predicates for hard float.
970 * config/nds32/nds32-protos.h: Declare functions for hard float.
971 * config/nds32/nds32.c: Implementation for hard float configuration.
972 * config/nds32/nds32.h: Definitions for hard float configuration.
973 * config/nds32/nds32.md: Include hard float machine description and
974 modify patterns for hard float configuration.
975 * config/nds32/nds32.opt: New options for hard float configuration.
976 * config/nds32/predicates.md: New predicates for hard float
977 configuration.
979 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
981 * common/config/nds32/nds32-common.c
982 (nds32_option_optimization_table): Enable -mreleax-hint by default.
984 2018-04-05 Jakub Jelinek <jakub@redhat.com>
986 PR middle-end/85195
987 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
988 CONSTRUCTOR_ELT (ctor, ...)->value.
990 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
992 PR target/85193
993 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
995 2018-04-05 Tom de Vries <tom@codesourcery.com>
997 PR target/85204
998 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
999 cond jump.
1001 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1002 Kito Cheng <kito.cheng@gmail.com>
1004 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1005 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1006 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1007 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1009 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1010 Kito Cheng <kito.cheng@gmail.com>
1012 * config/nds32/constraints.md (Ufe): New memory constraint.
1013 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1014 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1015 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1016 operands.
1017 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1018 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1020 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1022 * config/nds32/nds32.md: Use optimize_size in the condition for
1023 alu-shift instructions.
1025 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1027 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1029 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1031 * config/nds32/nds32.md (negsi2): Refine pattern.
1033 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1034 Chung-Ju Wu <jasonwucj@gmail.com>
1036 * config/nds32/iterators.md (shift_rotate): New code iterator.
1037 (shift): New code attribute.
1038 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1039 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1040 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1041 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1042 bit-wise operations.
1043 (andsi3, *andsi3): Ditto.
1044 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1045 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1046 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1047 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1048 nds32_ior_operand, nds32_xor_operand): New predicates.
1050 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1052 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1053 (addsi3, subsi3): ... this.
1055 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1057 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1059 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1061 * config/nds32/nds32.md: Adjust indention.
1063 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1065 * config/nds32/nds32.md (feature): New attribute.
1067 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1069 * config/nds32/nds32.md (subtype): New attribute.
1071 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1073 PR target/85203
1074 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1075 expansion to perform a bitwise AND of the argument followed by a
1076 boolean negation of the result.
1078 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1080 PR rtl-optimization/84878
1081 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1082 the basic block. Assert the use reference is not artificial and that
1083 it has an associated insn.
1085 2018-04-04 Michael Matz <matz@suse.de>
1087 * builtins.c (compute_objsize): Pass correct operand
1088 to array_at_struct_end_p.
1090 2018-04-04 Richard Biener <rguenther@suse.de>
1092 PR lto/85176
1093 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1094 from contexts for DINFO_LEVEL_TERSE and below.
1096 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1098 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1099 resiter_operand condition.
1100 * config/nds32/nds32.md (*move<mode>): Ditto.
1102 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1103 Monk Chiang <sh.chiang04@gmail.com>
1105 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1107 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1109 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1111 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1112 Kito Cheng <kito.cheng@gmail.com>
1114 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1115 nds32_cond_code_str, output_cond_branch,
1116 output_cond_branch_compare_zero, nds32_expand_cbranch,
1117 nds32_expand_cstore, nds32_expand_movcc,
1118 nds32_output_cbranchsi4_equality_zero,
1119 nds32_output_cbranchsi4_equality_reg,
1120 nds32_output_cbranchsi4_equality_reg_or_const_int,
1121 nds32_output_cbranchsi4_greater_less_zero: New functions.
1122 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1123 nds32_expand_cstore, nds32_expand_movcc,
1124 nds32_output_cbranchsi4_equality_zero,
1125 nds32_output_cbranchsi4_equality_reg,
1126 nds32_output_cbranchsi4_equality_reg_or_const_int,
1127 nds32_output_cbranchsi4_greater_less_zero): Declare.
1128 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1129 nds32_rimm11s_operand): New predicates.
1130 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1131 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1132 patterns.
1134 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1136 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1137 * config/nds32/nds32.md: Ditto.
1138 * config/nds32/pipelines.md: Ditto.
1140 2018-04-04 Richard Biener <rguenther@suse.de>
1142 PR tree-optimization/85168
1143 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1144 propagating abnormals.
1146 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1148 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1150 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1151 Kito Cheng <kito.cheng@gmail.com>
1153 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1154 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1155 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1156 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1157 * config/nds32/nds32.md (sibcall_internal): New.
1158 (sibcall_register): Remove.
1159 (sibcall_immediate): Remove.
1160 (sibcall_value_internal): New.
1161 (sibcall_value_register): Remove.
1162 (sibcall_value_immediate): Remove.
1163 * config/nds32/predicates.md (nds32_general_register_operand): New.
1164 (nds32_call_address_operand): New.
1166 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1168 PR rtl-optimization/85167
1169 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1170 bb_defs if *split_p, instead preinitialize it to NULL.
1172 PR tree-optimization/85156
1173 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1174 evaluating the argument multiple times.
1176 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1178 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1179 than vector.
1180 (_mm_cvtpd_ps): Likewise.
1181 (_mm_cvttpd_epi32): Likewise.
1182 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1183 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1184 vector, pixel, and bool following altivec.h include.
1186 2018-04-03 Martin Sebor <msebor@redhat.com>
1188 * doc/extend.texi (Common Function Attributes): Clarify.
1189 (const attribute): Likewise.
1190 (pure attribute): Likewise.
1192 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1194 PR target/85169
1195 * config/i386/i386.c (ix86_expand_vector_set): Use
1196 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1198 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1200 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1201 instructions when changing rounding bits to preserve precision bits
1202 in the x87 control word.
1204 2018-04-03 Martin Liska <mliska@suse.cz>
1206 PR tree-optimization/82491
1207 * rtl.h (strip_offset_and_add): Replace += suboffset with
1208 poly_uint64 () + suboffset.
1210 2018-03-29 Martin Liska <mliska@suse.cz>
1211 Martin Jambor <mjambor@suse.cz>
1213 PR ipa/84947
1214 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1215 param_type is not an integral or pointer type.
1217 2018-04-03 Richard Biener <rguenther@suse.de>
1219 * sese.h (recompute_all_dominators): Remove.
1221 2018-04-02 Martin Sebor <msebor@redhat.com>
1223 * doc/invoke.texi (-Wrestrict): Fix typos.
1225 2018-04-02 Jim Wilson <jimw@sifive.com>
1227 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1228 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1229 (<optab>di3, <optab>si3_extend): Likewise.
1230 (<optab>si3_mask, <optab>si3_mask_1): New.
1231 (<optab>di3_mask, <optab>di3_mask_1): New.
1232 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1233 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1234 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1236 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1238 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1239 example.
1241 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1243 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1244 (nds32_canonicalize_comparison): New function.
1246 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1247 Kito Cheng <kito.cheng@gmail.com>
1248 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1250 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1251 * config/nds32/constants.md (unspec_volatile_element): Add
1252 UNSPEC_VOLATILE_RELAX_GROUP.
1253 * config/nds32/nds32-relax-opt.c: New file.
1254 * config/nds32/nds32-predicates.c
1255 (nds32_symbol_load_store_p): New function.
1256 * config/nds32/nds32-protos.h
1257 (nds32_symbol_load_store_p): Declare function.
1258 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1259 * config/nds32/nds32.c
1260 (nds32_register_pass): New function to register pass.
1261 (nds32_register_passes): New function to register passes.
1262 * config/nds32/nds32.md (relax_group): New pattern.
1263 * config/nds32/nds32.opt (mrelax-hint): New option.
1264 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1266 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1268 * config/nds32/t-nds32: Modify files dependency.
1270 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1272 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1273 (PROFILE_HOOK): Define its implementation.
1275 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1277 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1278 type and 32-bit size.
1280 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1282 PR middle-end/85090
1283 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1284 (V_128_256): New mode iterator.
1285 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1286 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1287 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1288 of V.
1289 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1290 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1292 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1294 PR target/83315
1295 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1296 NaN inputs correctly.
1298 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1300 PR target/80546
1301 * config/rs6000/vsx.md (??r): New mode attribute.
1302 (*vsx_mov<mode>_64bit): Use it.
1303 (*vsx_mov<mode>_32bit): Likewise.
1305 2018-03-30 Martin Sebor <msebor@redhat.com>
1307 PR tree-optimization/84818
1308 * builtins.c (check_access): Use warning_n.
1310 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1312 PR target/83822
1313 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1314 condition.
1315 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1316 condition.
1318 2018-03-30 Julia Koval <julia.koval@intel.com>
1320 PR target/84413
1321 * x86-tune.def (movx, partial_reg_dependency): Enable for
1322 m_SKYLAKE_AVX512.
1324 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1326 PR inline-asm/84985
1327 * lra-constraints.c (process_alt_operands): Move setting
1328 this_alternative_matches below.
1330 2018-03-29 Martin Liska <mliska@suse.cz>
1332 PR lto/84995.
1333 * doc/invoke.texi: Document how LTO works with debug info.
1334 Describe auto-load support of binutils. Mention 'x86-64'
1335 as valid option value of -march option.
1337 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1339 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1341 PR c/85094
1342 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1343 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1344 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1345 checking.
1347 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1349 PR target/84912
1350 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1351 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1352 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1353 for RS6000_BTM_POWERPC64.
1354 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1355 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1356 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1357 definition.
1358 (DIVDE): Use it.
1359 (DIVDEU): Likewise.
1361 2018-03-28 Carl Love <cel@us.ibm.com>
1363 Revert
1364 2017-09-27 Carl Love <cel@us.ibm.com>
1366 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1367 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1368 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1369 fctiw instruction.
1371 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1373 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1374 instead of __vector bool.
1375 (_mm_max_pu8): Likewise.
1376 (_mm_min_pi16): Likewise.
1378 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1380 PR target/84912
1381 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1382 (DIVWEUO): Likewise.
1383 (DIVDEO): Likewise.
1384 (DIVDEUO): Likewise.
1385 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1386 DIVWEUO and DIVDEUO.
1387 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1388 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1389 (div_extend): Likewise.
1390 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1391 builtin function.
1392 (__builtin_divweuo): Likewise.
1393 (__builtin_divdeo): Likewise.
1394 (__builtin_divdeuo): Likewise.
1396 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1398 PR target/85095
1399 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1400 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1402 PR tree-optimization/82004
1403 * gimple-match-head.c (optimize_pow_to_exp): New function.
1404 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1405 Don't fold to exp if optimize_pow_to_exp is false.
1407 2018-03-28 Martin Liska <mliska@suse.cz>
1409 PR other/84819
1410 * calls.c (initialize_argument_information): Fix trailing space.
1411 * common.opt: Fix typo and provide better explanation for
1412 -fsanitize-coverage option.
1413 * config/i386/i386.opt: Fix typo.
1415 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1416 Martin Liska <mliska@suse.cz>
1418 PR sanitizer/85081
1419 * gimplify.c (asan_poison_variable): Don't do the check for
1420 gimplify_omp_ctxp here.
1421 (gimplify_decl_expr): Do it here.
1422 (gimplify_target_expr): Likewise.
1424 2018-03-28 Martin Liska <mliska@suse.cz>
1426 PR target/84988
1427 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1428 chkp_type_bounds_count if MPX is not enabled.
1430 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1432 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1434 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1436 PR target/84914
1437 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1438 function to create the function decl for complex long double
1439 multiply and divide for -mabi=ieeelongdouble.
1440 (init_float128_ieee): Call it.
1442 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
1444 PR target/85044
1445 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1446 -fcf-protection=branch -mibt.
1447 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
1449 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1451 PR target/81863
1452 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
1454 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
1456 PR target/85056
1457 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
1458 extern array declarations.
1460 2018-03-27 Richard Biener <rguenther@suse.de>
1462 PR middle-end/84067
1463 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
1464 explicit single_use checks.
1466 2018-03-27 Richard Biener <rguenther@suse.de>
1468 PR tree-optimization/85082
1469 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
1470 Valueize the VUSE.
1472 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1474 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
1475 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
1476 Turn on fasynchronous-unwind-tables and funwind-tables.
1478 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
1480 PR target/85073
1481 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1482 (*bmi_blsr_<mode>_ccz): Ditto.
1484 2018-03-26 Tom de Vries <tom@codesourcery.com>
1486 PR tree-optimization/85063
1487 * omp-general.c (offloading_function_p): New function. Factor out
1488 of ...
1489 * omp-offload.c (pass_omp_target_link::gate): ... here.
1490 * omp-general.h (offloading_function_p): Declare.
1491 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1492 with attribute omp declare target for offloading functions.
1494 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
1496 PR tree-optimization/84005
1497 * tree-data-ref.h (get_base_for_alignment): Declare.
1498 * tree-data-ref.c (get_base_for_alignment_1): New function.
1499 (get_base_for_alignment): Likewise.
1500 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1501 get_base_for_alignment to find a suitable base object, instead
1502 of always using drb->base_address.
1504 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1506 PR inline-asm/85022
1507 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1508 known size by default.
1510 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1512 PR inline-asm/85030
1513 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1514 and non BLKmode operands.
1516 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1518 PR target/85026
1519 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1520 Clean up attributes.
1522 2018-03-23 Richard Biener <rguenther@suse.de>
1524 PR debug/85020
1525 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1526 we are going to emit early debug for LTO.
1528 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1530 PR inline-asm/85034
1531 * function.c (match_asm_constraints_1): Don't optimize if input
1532 doesn't satisfy general_operand predicate for output's mode.
1534 PR inline-asm/85022
1535 * alias.c (write_dependence_p): Don't require for x_canonicalized
1536 non-VOIDmode if x has VOIDmode.
1538 PR sanitizer/85029
1539 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1540 just don't try to optimize it rather than assert it never happens.
1542 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1544 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1545 macro expansions for definition of ST_INTERNAL_<mode> and
1546 LD_INTERNAL_<mode> builtins.
1547 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1548 Remove prototype.
1549 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1550 function.
1551 (altivec_expand_st_builtin): Likewise.
1552 (altivec_expand_builtin): Remove calls to deleted functions.
1553 (rs6000_address_for_altivec): Delete this function.
1554 * config/rs6000/vector.md: Remove expands for
1555 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1557 2018-03-22 Sudakshina Das <sudi.das@arm.com>
1559 PR target/84826
1560 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1561 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1562 re-computing once computed.
1563 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1564 (arm_init_machine_status): Initialize
1565 machine->static_chain_stack_bytes.
1567 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1569 PR target/84760
1570 * doc/extend.texi: Add four new prototypes for vec_ld.
1571 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1572 definitions for more logical presentation.
1573 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1574 entries for V1TI variants of __builtin_altivec_ld builtin.
1575 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1576 handling of V1TI variant of LVX icode pattern.
1577 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1578 (rs6000_gimple_fold_builtin): Likewise.
1579 (altivec_init_builtins): Add code to define
1580 __builtin_altivec_lvx_v1ti function.
1582 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1584 PR inline-asm/84941
1585 * function.c (match_asm_constraints_1): Don't do the optimization
1586 if input isn't a REG, SUBREG, MEM or constant.
1588 2018-03-22 Tom de Vries <tom@codesourcery.com>
1590 PR tree-optimization/84956
1591 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1592 bb_has_abnormal_pred.
1594 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1596 PR sanitizer/85018
1597 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1598 DECL_INITIAL (decl) to decl at the end.
1599 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1600 adjust the comment.
1602 2018-03-21 Joseph Myers <joseph@codesourcery.com>
1604 * doc/extend.texi (__builtin_tgmath): Document when complex
1605 integer types are treated as _Complex _Float64.
1607 2018-03-21 Tom de Vries <tom@codesourcery.com>
1609 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1611 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1613 PR tree-optimization/84960
1614 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1615 if it is ENTRY block, move them into single succ of ENTRY in that case.
1617 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
1619 PR tree-optimization/84811
1620 * poly-int.h (poly_span_traits): Remove the T3 parameter and
1621 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1622 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1623 (known_subrange_p): Update accordingly. Cast each value involved
1624 in the size comparison, rather than casting the result of the
1625 subtraction.
1627 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1629 PR tree-optimization/84982
1630 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1631 by flipping the least significant bit rather than all bits from
1632 bitpos to bitpos + bitsize - 1.
1634 2018-03-21 Nathan Sidwell <nathan@acm.org>
1636 * doc/extend.texi (Deprecated Features): Remove mention of
1637 long-deleted deprecations.
1639 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1641 PR jit/84288
1642 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1643 * configure: Regenerate.
1645 2018-03-21 Tom de Vries <tom@codesourcery.com>
1647 PR tree-optimization/83126
1648 * tree-parloops.c (num_phis): New function.
1649 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1651 2018-03-21 Nathan Sidwell <nathan@acm.org>
1653 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1654 mention anon-struct/union members and trailing attributes.
1656 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1658 PR tree-optimization/84969
1659 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1660 builtin memset partitions if they set different rhs values.
1662 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1664 PR rtl-optimization/84989
1665 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1666 VEC_DUPLICATE with scalar result mode.
1668 2018-03-21 Martin Liska <mliska@suse.cz>
1670 PR ipa/84963
1671 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1672 not intended return statement.
1674 2018-03-21 Martin Liska <mliska@suse.cz>
1676 PR target/84988
1677 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1678 (chkp_find_bound_slots_1): Limit number of iterations.
1680 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
1682 PR target/84838
1683 * Minor grammar fixes for x86 options.
1685 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1687 PR debug/84875
1688 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1689 holding REG_CFA_RESTORE notes, instead turn them into a USE.
1691 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
1693 PR target/83789
1694 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1695 (altivec_lvx_<mode>_1op): Likewise.
1696 (altivec_stvx_<mode>_2op): Likewise.
1697 (altivec_stvx_<mode>_1op): Likewise.
1698 (altivec_lvx_<VM2:mode>): New define_expand.
1699 (altivec_stvx_<VM2:mode>): Likewise.
1700 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1701 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1702 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1703 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1704 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1705 (rs6000_gen_lvx): Likewise.
1706 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1707 (altivec_expand_stv_builtin): Likewise.
1708 (altivec_expand_builtin): Likewise.
1709 * config/rs6000/vector.md: Likewise.
1711 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1713 PR target/82518
1714 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1715 BYTES_BIG_ENDIAN.
1717 2018-03-20 Richard Biener <rguenther@suse.de>
1719 PR target/84986
1720 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1721 sign-conversions as zero, fall back to standard scalar_stmt
1722 cost for the rest.
1724 2018-03-20 Martin Liska <mliska@suse.cz>
1726 PR ipa/84825
1727 * predict.c (rebuild_frequencies): Handle case when we have
1728 PROFILE_ABSENT, but flag_guess_branch_prob is false.
1730 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1732 PR target/84990
1733 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1734 flag_section_anchors.
1735 * varasm.c (use_blocks_for_decl_p): Remove hack for
1736 dw2_force_const_mem.
1738 PR target/84845
1739 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
1740 to ...
1741 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
1742 be created, use lowpart_subreg of operands[0] rather than operands[0]
1743 itself.
1744 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
1745 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
1746 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
1747 and n constraint instead of aarch64_shift_imm_di and Usd.
1748 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
1749 (*aarch64_<optab>_reg_minus<mode>3): ... this.
1751 2018-03-20 Sudakshina Das <sudi.das@arm.com>
1753 PR target/82989
1754 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
1755 to favor GPR over NEON registers.
1756 (<shift>di3_neon): Likewise.
1758 2018-03-20 Tom de Vries <tom@codesourcery.com>
1760 PR target/84952
1761 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
1762 (nvptx_process_pars): Emit bar.sync asap and alap.
1764 2018-03-20 Tom de Vries <tom@codesourcery.com>
1766 PR target/84954
1767 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
1768 seen_label if seen_label is already set.
1770 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1772 PR target/84945
1773 * config/i386/i386.c (fold_builtin_cpu): For features above 31
1774 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
1775 Use 1U instead of 1. Formatting fixes.
1777 PR c/84953
1778 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
1779 instead of TREE_TYPE (s1) for the return value.
1781 2018-03-19 Jakub Jelinek <jakub@redhat.com>
1783 PR tree-optimization/84946
1784 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
1785 bitsize + bitsize in poly_uint64 rather than poly_int64.
1787 PR sanitizer/78651
1788 * dwarf2asm.c: Include fold-const.c.
1789 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1790 of decl rather than decl itself.
1792 PR rtl-optimization/84643
1793 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1795 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
1797 PR sanitizer/78651
1798 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1799 calling assemble_variable.
1801 2018-03-19 Sudakshina Das <sudi.das@arm.com>
1803 PR target/81647
1804 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1805 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1807 2018-03-19 Jim Wilson <jimw@sifive.com>
1809 PR bootstrap/84856
1810 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1811 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1812 (riscv_first_stack_step): Likewise.
1813 (riscv_option_override): Use STACK_BOUNDARY instead of
1814 MIN_STACK_BOUNDARY.
1815 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1816 MIN_STACK_BOUNDARY.
1817 (BIGGEST_ALIGNMENT): Set to 128.
1818 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1819 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1820 STACK_BOUNDARY.
1822 2018-03-19 Richard Biener <rguenther@suse.de>
1824 PR tree-optimization/84933
1825 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1826 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1828 2018-03-19 Richard Biener <rguenther@suse.de>
1830 PR tree-optimization/84859
1831 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1832 (cond_if_else_store_replacement): Perform sinking operation on
1833 single-store BBs regardless of MAX_STORES_TO_SINK setting.
1834 Generalize what a BB with a single eligible store is.
1836 2018-03-19 Richard Biener <rguenther@suse.de>
1838 PR tree-optimization/84929
1839 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1840 chrec_is_positive against non-chrec arg.
1842 2018-03-19 Tamar Christina <tamar.christina@arm.com>
1844 PR target/84711
1845 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1847 2018-03-18 Martin Liska <mliska@suse.cz>
1849 PR rtl-optimization/84635
1850 * regrename.c (build_def_use): Use matches_mode only when
1851 matches >= 0.
1853 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1855 PR tree-optimization/84913
1856 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1857 vectorize chains of COND_EXPRs.
1859 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1861 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1863 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1865 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1867 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1869 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1871 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1872 Kito Cheng <kito.cheng@gmail.com>
1874 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1875 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1876 (nds32_adjust_reg_alloc_order): New function.
1877 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1879 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1881 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1882 nds32_print_operand, nds32_print_operand_address): Use
1883 HOST_WIDE_INT_PRINT_DEC instead.
1885 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1887 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1889 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1891 PR target/84902
1892 * config/i386/i386.c (initial_ix86_tune_features,
1893 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1894 unsigned long long.
1895 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1896 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1897 rather than 1u << ix86_tune. Formatting fix.
1898 (ix86_option_override_internal): Change ix86_arch_mask from
1899 unsigned int to unsigned HOST_WIDE_INT, initialize to
1900 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1901 (ix86_function_specific_restore): Likewise.
1903 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1905 PR target/84899
1906 * postreload.c (reload_combine_recognize_pattern): Perform
1907 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1908 truncate_int_for_mode the result for the destination's mode.
1910 PR c/84909
1911 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1912 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1913 Likewise.
1914 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1915 Likewise.
1917 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
1919 PR target/84876
1920 * lra-assigns.c (lra_split_hard_reg_for): Don't use
1921 regno_allocno_class_array and sorted_pseudos.
1922 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
1923 insns where regno is used.
1925 2018-03-16 Martin Liska <mliska@suse.cz>
1927 PR ipa/84833
1928 * multiple_target.c (create_dispatcher_calls): Redirect
1929 reference in the symbol table.
1931 2018-03-16 Martin Liska <mliska@suse.cz>
1933 PR ipa/84722
1934 * multiple_target.c (create_dispatcher_calls): Redirect also
1935 an alias.
1937 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1939 PR c++/79937
1940 PR c++/82410
1941 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
1942 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
1943 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
1945 2018-03-16 Julia Koval <julia.koval@intel.com>
1947 * doc/invoke.texi (Skylake Server): Add CLWB.
1948 Cannonlake): Remove CLWB.
1950 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1952 PR tree-optimization/84841
1953 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
1954 1 << 3.
1955 (FLOAT_ONE_CONST_TYPE): Define.
1956 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
1957 (sort_by_operand_rank): Put entries with higher constant_type last
1958 rather than first to match comments.
1960 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
1962 * config/nios2/nios2.md (movsi_internal): Fix thinko in
1963 split predicate.
1965 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1967 PR c++/79085
1968 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
1969 check and use address of target always.
1971 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
1973 PR target/84574
1974 * config/i386/i386.c (indirect_thunk_needed): Update comments.
1975 (indirect_thunk_bnd_needed): Likewise.
1976 (indirect_thunks_used): Likewise.
1977 (indirect_thunks_bnd_used): Likewise.
1978 (indirect_return_needed): New.
1979 (indirect_return_bnd_needed): Likewise.
1980 (output_indirect_thunk_function): Add a bool argument for
1981 function return.
1982 (output_indirect_thunk_function): Don't generate alias for
1983 function return thunk.
1984 (ix86_code_end): Call output_indirect_thunk_function to generate
1985 function return thunks.
1986 (ix86_output_function_return): Set indirect_return_bnd_needed
1987 and indirect_return_needed instead of indirect_thunk_bnd_needed
1988 and indirect_thunk_needed.
1990 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
1992 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
1993 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
1994 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
1996 2018-03-15 David Malcolm <dmalcolm@redhat.com>
1997 Paul Hua <paul.hua.gm@gmail.com>
1999 PR c/84852
2000 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2002 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2004 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2005 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2006 resp. SFmode cases.
2008 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2010 PR target/84711
2011 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2012 instead of GET_MODE_SIZE when comparing Units.
2014 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2016 PR target/68256
2017 * varasm.c (hash_section): Return an unchangeble hash value
2018 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2019 Return !aarch64_can_use_per_function_literal_pools_p ().
2021 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2023 PR target/84860
2024 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2025 rather than address of cmode as last argument to prepare_cmp_insn.
2027 2018-03-15 Julia Koval <julia.koval@intel.com>
2029 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2030 F_AVX512VNNI, F_AVX512BITALG): New.
2032 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2034 PR target/83451
2035 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2036 insn for floating-point loads and stores.
2038 2018-03-14 Carl Love <cel@us.ibm.com>
2040 * config/rs6000/rs6000-c.c: Add macro definitions for
2041 ALTIVEC_BUILTIN_VEC_PERMXOR.
2042 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2043 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2044 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2045 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2046 UNSPEC_VPERMXOR.
2047 * config/doc/extend.texi: Add prototypes for vec_permxor.
2049 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2051 PR c/84852
2052 * diagnostic-show-locus.c (class layout_point): Convert m_line
2053 from int to linenum_type.
2054 (line_span::comparator): Use linenum "compare" function when
2055 comparing line numbers.
2056 (test_line_span): New function.
2057 (layout_range::contains_point): Convert param "row" from int to
2058 linenum_type.
2059 (layout_range::intersects_line_p): Likewise.
2060 (layout::will_show_line_p): Likewise.
2061 (layout::print_source_line): Likewise.
2062 (layout::should_print_annotation_line_p): Likewise.
2063 (layout::print_annotation_line): Likewise.
2064 (layout::print_leading_fixits): Likewise.
2065 (layout::annotation_line_showed_range_p): Likewise.
2066 (struct line_corrections): Likewise for field m_row.
2067 (line_corrections::line_corrections): Likewise for param "row".
2068 (layout::print_trailing_fixits): Likewise.
2069 (layout::get_state_at_point): Likewise.
2070 (layout::get_x_bound_for_row): Likewise.
2071 (layout::print_line): Likewise.
2072 (diagnostic_show_locus): Likewise for locals "last_line" and
2073 "row".
2074 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2075 * input.c (selftest::test_linenum_comparisons): New function.
2076 (selftest::input_c_tests): Call it.
2077 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2078 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2079 * selftest.h (ASSERT_GT): New macro.
2080 (ASSERT_GT_AT): New macro.
2081 (ASSERT_LT): New macro.
2082 (ASSERT_LT_AT): New macro.
2084 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2086 PR rtl-optimization/84780
2087 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2089 2018-03-14 Martin Liska <mliska@suse.cz>
2091 * tree.c (record_node_allocation_statistics): Use
2092 get_stats_node_kind.
2093 (get_stats_node_kind): New function extracted from
2094 record_node_allocation_statistics.
2095 (free_node): Use get_stats_node_kind.
2097 2018-03-14 Richard Biener <rguenther@suse.de>
2099 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2100 that the value-set of ANTIC_IN doesn't grow.
2102 Revert
2103 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2104 member.
2105 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2106 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2108 2018-03-14 Julia Koval <julia.koval@intel.com>
2110 * config.gcc (icelake-client, icelake-server): New.
2111 (icelake): Remove.
2112 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2113 (initial_ix86_arch_features): Ditto.
2114 (PTA_SKYLAKE): Add SGX.
2115 (PTA_ICELAKE): Remove.
2116 (PTA_ICELAKE_CLIENT): New.
2117 (PTA_ICELAKE_SERVER): New.
2118 (ix86_option_override_internal): Split up icelake on icelake client and
2119 icelake server.
2120 (get_builtin_code_for_version): Ditto.
2121 (fold_builtin_cpu): Ditto.
2122 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2123 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2124 * config/i386/i386.h (processor_type): Ditto.
2125 * doc/invoke.texi: Ditto.
2127 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2129 PR sanitizer/83392
2130 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2131 INTEGER_CST offset, add it together with bitpos / 8 and
2132 sign extend based on POINTER_SIZE.
2134 PR target/84844
2135 Revert
2136 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2138 PR target/78090
2139 * config/i386/constraints.md (Yc): New register constraint.
2140 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2141 Use Yc constraint for alternative 2 of operand 0. Remove
2142 preferred_for_speed attribute.
2144 2018-03-14 Richard Biener <rguenther@suse.de>
2146 PR tree-optimization/84830
2147 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2148 with the old one to avoid oscillations.
2150 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2152 PR target/83712
2153 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2154 pseudos.
2155 (assign_by_spills): Return a flag of reload assignment failure.
2156 Do not process the reload assignment failures. Do not spill other
2157 reload pseudos if they has the same reg class. Update n if
2158 necessary.
2159 (lra_assign): Add a return arg. Set up from the result of
2160 assign_by_spills call.
2161 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2162 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2163 usage_insns if it is not NULL.
2164 (spill_hard_reg_in_range): New function.
2165 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2166 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2167 function prototypes.
2168 (lra_assign): Change prototype.
2169 * lra.c (lra): Add code to deal with fails by splitting hard reg
2170 live ranges.
2172 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2174 * config/riscv/riscv.opt (mrelax): New option.
2175 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2176 "norelax" when riscv_mrelax is disabled.
2177 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2179 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2181 PR target/84743
2182 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2183 reassociation for int modes.
2185 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2187 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2188 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2189 for big-endian.
2190 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2191 * config/aarch64/aarch64-sve.md
2192 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2193 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2194 (*extend<mode><Vwide>2): Rename to...
2195 (aarch64_sve_extend<mode><Vwide>2): ...this.
2196 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2197 renaming the old pattern to...
2198 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2199 unsigned packs.
2200 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2201 define_expand, renaming the old pattern to...
2202 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2203 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2204 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2205 account when deciding which SVE instruction the optab should use.
2206 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2208 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2210 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2211 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2212 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2213 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2214 (tlsdesc_small_<mode>): Turn a define_expand and use
2215 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2216 (tlsdesc_small_advsimd_<mode>): ...this.
2217 (tlsdesc_small_sve_<mode>): New pattern.
2219 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2221 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2222 (UNSPEC_UMUL_HIGHPART): New constants.
2223 (MUL_HIGHPART): New int iteraor.
2224 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2225 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2226 define_expand.
2227 (*<su>mul<mode>3_highpart): New define_insn.
2229 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2231 PR lto/84805
2232 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2233 incomplete types.
2235 2018-03-13 Martin Liska <mliska@suse.cz>
2237 PR ipa/84658.
2238 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2239 vector.
2240 (sem_item_optimizer::~sem_item_optimizer): Release it.
2241 (sem_item_optimizer::merge_classes): Register variable aliases.
2242 (sem_item_optimizer::fixup_pt_set): New function.
2243 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2244 * ipa-icf.h: Declare new variables and functions.
2246 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2248 PR middle-end/84834
2249 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2250 integer_pow2p@2 and test integer_pow2p in condition.
2251 (A < 0 ? C : 0): Similarly for @1.
2253 PR middle-end/84831
2254 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2255 characters starting at p contain '\0' character, don't look beyond
2256 that.
2258 PR target/84827
2259 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2260 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2262 PR target/84828
2263 * reg-stack.c (change_stack): Change update_end var from int to
2264 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2265 also call set_block_for_insn on the newly added insns and rescan.
2267 PR target/84786
2268 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2269 on the last operand.
2271 PR c++/84704
2272 * tree.c (stabilize_reference_1): Return save_expr (e) for
2273 STATEMENT_LIST even if it doesn't have side-effects.
2275 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2277 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2279 2018-03-12 Renlin Li <renlin.li@arm.com>
2281 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2282 aarch64_output_scalar_simd_mov_immediate.
2284 2018-03-12 Martin Sebor <msebor@redhat.com>
2286 PR tree-optimization/83456
2287 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2288 for perfectly overlapping calls to memcpy.
2289 (gimple_fold_builtin_memory_chk): Same.
2290 (gimple_fold_builtin_strcpy): Handle no-warning.
2291 (gimple_fold_builtin_stxcpy_chk): Same.
2292 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2294 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2296 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2297 parameter. Use it for SFmode.
2298 (rs6000_function_arg_advance_1): Adjust.
2299 (rs6000_function_arg): Adjust.
2300 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2302 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2304 PR rtl-optimization/84169
2305 PR rtl-optimization/84780
2306 * combine.c (can_combine_p): Check for a 2-insn combination whether
2307 the destination register is used between the two insns, too.
2309 2018-03-12 Richard Biener <rguenther@suse.de>
2311 PR tree-optimization/84803
2312 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2313 for refs DR analysis didn't process.
2315 2018-03-12 Richard Biener <rguenther@suse.de>
2317 PR tree-optimization/84777
2318 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2319 force-vectorize loops ignore whether we are optimizing for size.
2321 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2323 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2324 (TARGET_MD_ASM_ADJUST): Define.
2326 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2327 Kito Cheng <kito.cheng@gmail.com>
2328 Chung-Ju Wu <jasonwucj@gmail.com>
2330 * config/nds32/nds32.c (nds32_compute_stack_frame,
2331 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2332 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2333 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2334 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2335 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2336 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2337 * config/nds32/nds32.md (prologue, epilogue): Use macro
2338 NDS32_V3PUSH_AVAILABLE_P to do checking.
2340 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2342 PR debug/58150
2343 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2344 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2345 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2346 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2347 addition of most attributes on !orig_type_die or the attribute not
2348 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2350 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2351 Chung-Ju Wu <jasonwucj@gmail.com>
2353 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2354 __NDS32_VH__ macro.
2355 * config/nds32/nds32.opt (mvh): New option.
2357 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2358 Chung-Ju Wu <jasonwucj@gmail.com>
2360 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2361 function.
2362 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2363 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2364 definition.
2366 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2367 Chung-Ju Wu <jasonwucj@gmail.com>
2369 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2370 function.
2371 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2372 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2374 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2375 Kito Cheng <kito.cheng@gmail.com>
2376 Chung-Ju Wu <jasonwucj@gmail.com>
2378 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2379 UNSPEC_FFMISM and UNSPEC_FLMISM.
2380 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2381 for ffb, ffmism and flmism.
2382 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2383 (unspec_ffmism): Ditto.
2384 (unspec_flmism): Ditto.
2385 (nds32_expand_builtin_impl): Check if string extension is available.
2386 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2387 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2389 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2391 Reverting patch:
2392 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2394 PR target/83712
2395 * lra-assigns.c (assign_by_spills): Return a flag of reload
2396 assignment failure. Do not process the reload assignment
2397 failures. Do not spill other reload pseudos if they has the same
2398 reg class.
2399 (lra_assign): Add a return arg. Set up from the result of
2400 assign_by_spills call.
2401 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2402 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2403 usage_insns if it is not NULL.
2404 (spill_hard_reg_in_range): New function.
2405 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2406 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2407 function prototypes.
2408 (lra_assign): Change prototype.
2409 * lra.c (lra): Add code to deal with fails by splitting hard reg
2410 live ranges.
2412 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
2414 PR target/84807
2415 * config/i386/i386.opt: Replace Enforcment with Enforcement.
2417 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2419 PR debug/84620
2420 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2421 (dw_val_node): Add val_symbolic_view.
2422 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2423 (symview_upper_bound): New.
2424 (new_line_info_table): Initialize symviews_since_reset.
2425 (dwarf2out_source_line): Count symviews_since_reset and set
2426 symview_upper_bound.
2427 (dw_val_equal_p): Handle symview.
2428 (add_AT_symview): New.
2429 (print_dw_val): Handle symview.
2430 (attr_checksum, attr_checksum_ordered): Likewise.
2431 (same_dw_val_p, size_of_die): Likewise.
2432 (value_format, output_die): Likewise.
2433 (add_high_low_attributes): Use add_AT_symview for entry_view.
2434 (dwarf2out_finish): Reset symview_upper_bound, clear
2435 zero_view_p.
2437 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2439 PR target/83969
2440 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2441 Add strict argument and use it.
2442 (rs6000_split_multireg_move): Update for new strict argument.
2443 (mem_operand_gpr): Disallow all non-offsettable addresses.
2444 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2446 2018-03-09 Jakub Jelinek <jakub@redhat.com>
2448 PR target/84772
2449 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
2450 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
2451 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
2453 PR c++/84767
2454 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
2455 decl, use remap_type if we want to use the type.
2457 2018-03-09 Martin Sebor <msebor@redhat.com>
2459 PR tree-optimization/84526
2460 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2461 Remove dead code.
2462 (builtin_access::generic_overlap): Be prepared to handle non-array
2463 base objects.
2465 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
2467 PR rtl-optimization/84682
2468 * lra-constraints.c (process_address_1): Check is_address flag
2469 for address constraints.
2470 (process_alt_operands): Likewise.
2471 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
2472 preprocess_constraints.
2473 * recog.h (preprocess_constraints): Add oploc parameter.
2474 Adjust callers.
2475 * recog.c (preprocess_constraints): Test address_operand for
2476 CT_ADDRESS constraints.
2478 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2480 PR target/83712
2481 * lra-assigns.c (assign_by_spills): Return a flag of reload
2482 assignment failure. Do not process the reload assignment
2483 failures. Do not spill other reload pseudos if they has the same
2484 reg class.
2485 (lra_assign): Add a return arg. Set up from the result of
2486 assign_by_spills call.
2487 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2488 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2489 usage_insns if it is not NULL.
2490 (spill_hard_reg_in_range): New function.
2491 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2492 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2493 function prototypes.
2494 (lra_assign): Change prototype.
2495 * lra.c (lra): Add code to deal with fails by splitting hard reg
2496 live ranges.
2498 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2500 PR target/83193
2501 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2502 Accept complain bool parameter. Only emit errors if it is true.
2503 (arm_parse_cpu_option_name): Likewise.
2504 (arm_target_thumb_only): Adjust callers of the above.
2505 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2506 prototype to take a default true bool parameter.
2507 (arm_parse_arch_option_name): Likewise.
2509 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2510 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2512 PR jit/64089
2513 PR jit/84288
2514 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2515 * configure: Regenerate.
2516 * configure.ac ("linker --version-script option"): New.
2517 ("linker soname option"): New.
2519 2018-03-09 Richard Biener <rguenther@suse.de>
2521 PR tree-optimization/84775
2522 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2523 immediate uses of predicate stmts and mark them modified.
2525 Revert
2526 PR tree-optimization/84178
2527 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2528 to caller.
2529 (version_loop_for_if_conversion): Delay update_ssa call.
2530 (tree_if_conversion): Delay update_ssa until after predicate
2531 insertion.
2533 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2535 PR target/84763
2536 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2537 when the function accesses prior frames.
2539 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2541 PR debug/84456
2542 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2543 gen_llsym, otherwise call maybe_gen_llsym.
2545 PR inline-asm/84742
2546 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2547 has ',' character inside of it.
2549 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2551 PR target/84748
2552 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2553 as clobbering CC_REGNUM.
2555 2018-03-08 Richard Biener <rguenther@suse.de>
2557 PR middle-end/84552
2558 * tree-scalar-evolution.c: Include tree-into-ssa.h.
2559 (follow_copies_to_constant): Do not follow SSA names registered
2560 for update.
2562 2018-03-08 Richard Biener <rguenther@suse.de>
2564 PR tree-optimization/84178
2565 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2566 to caller.
2567 (version_loop_for_if_conversion): Delay update_ssa call.
2568 (tree_if_conversion): Delay update_ssa until after predicate
2569 insertion.
2571 2018-03-08 David Malcolm <dmalcolm@redhat.com>
2573 PR tree-optimization/84178
2574 * tree-if-conv.c (release_bb_predicate): Remove the
2575 the assertion that the stmts have NULL use_ops.
2576 Discard the statements, asserting that they haven't
2577 yet been added to a BB.
2579 2018-03-08 Richard Biener <rguenther@suse.de>
2581 PR tree-optimization/84746
2582 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2583 (phi_translate): Pass in destination ANTIC_OUT set.
2584 (phi_translate_1): Likewise. For a simplified result lookup
2585 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2586 (phi_translate_set): Adjust.
2587 (do_pre_regular_insertion): Likewise.
2588 (do_pre_partial_partial_insertion): Likewise.
2590 2018-03-08 Martin Liska <mliska@suse.cz>
2592 PR gcov-profile/84735
2593 * doc/gcov.texi: Document usage of profile files.
2594 * gcov-io.h: Document changes in the format.
2596 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
2598 PR debug/84404
2599 PR debug/84408
2600 * dwarf2out.c (struct dw_line_info_table): Update comments for
2601 view == -1.
2602 (FORCE_RESET_NEXT_VIEW): New.
2603 (FORCE_RESETTING_VIEW_P): New.
2604 (RESETTING_VIEW_P): Check for -1 too.
2605 (ZERO_VIEW_P): Likewise.
2606 (new_line_info_table): Force-reset next view.
2607 (dwarf2out_begin_function): Likewise.
2608 (dwarf2out_source_line): Simplify zero_view_p initialization.
2609 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2610 view directly. Omit view when omitting .loc at line 0.
2612 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2614 PR tree-optimization/84740
2615 * tree-switch-conversion.c (process_switch): Call build_constructors
2616 only if info.phi_count is non-zero.
2618 PR tree-optimization/84739
2619 * tree-tailcall.c (find_tail_calls): Check call arguments against
2620 DECL_ARGUMENTS (current_function_decl) rather than
2621 DECL_ARGUMENTS (func) when checking for tail recursion.
2623 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2625 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2626 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2627 Volker Reichelt's entry and add entries for people that perform
2628 GCC fuzzy testing and report numerous bugs.
2630 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2632 PR target/82411
2633 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2634 readonly data in sdata, if that is disabled.
2635 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2636 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2637 -mreadonly-in-sdata option.
2639 2018-03-07 Martin Sebor <msebor@redhat.com>
2641 PR tree-optimization/84468
2642 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2643 basic block when looking for nul assignment.
2645 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2647 PR target/84277
2648 * except.h (output_function_exception_table): Adjust prototype.
2649 * except.c (output_function_exception_table): Remove FNNAME parameter
2650 and add SECTION parameter. Ouput one part of the table at a time.
2651 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2652 the first part of the exception table and emit unwind directives.
2653 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2654 (i386_pe_seh_cold_init): Likewise.
2655 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2656 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2657 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2658 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2659 * config/i386/winnt.c: Include except.h.
2660 (struct seh_frame_state): Add reg_offset, after_prologue and
2661 in_cold_section fields.
2662 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2663 (i386_pe_seh_cold_init): New function.
2664 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2665 to seh->in_cold_section.
2666 (seh_emit_push): Record the offset of the push.
2667 (seh_emit_save): Record the offet of the save.
2668 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2669 Test seh->after_prologue to disregard the epilogue.
2670 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2671 (i386_pe_end_cold_function): New function.
2673 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2675 PR fortran/84565
2676 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2677 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2679 PR c++/84704
2680 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2681 on tmp_var.
2682 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2683 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2685 PR middle-end/84723
2686 * multiple_target.c: Include tree-inline.h and intl.h.
2687 (expand_target_clones): Diagnose and fail if node->definition and
2688 !tree_versionable_function_p (node->decl).
2690 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
2692 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2693 sprint_ul.
2694 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2695 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2696 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2698 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2700 PR target/84710
2701 * combine.c (try_combine): Use reg_or_subregno instead of handling
2702 just paradoxical SUBREGs and REGs.
2704 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
2706 * config/arc/arc.c (arc_finalize_pic): Remove function.
2707 (arc_must_save_register): We use single base PIC register, remove
2708 checks to save/restore the PIC register.
2709 (arc_expand_prologue): Likewise.
2710 * config/arc/arc-protos.h (arc_set_default_type_attributes):
2711 Remove.
2712 (arc_verify_short): Likewise.
2713 (arc_attr_type): Likewise.
2714 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2715 (walk_stores): Likewise.
2716 (arc_address_cost): Make it static.
2717 (arc_verify_short): Likewise.
2718 (branch_dest): Likewise.
2719 (arc_attr_type): Likewise.
2720 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2721 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2722 (arc_final_prescan_insn): Remove inserting the nops due to
2723 hardware hazards. It is done in reorg step.
2724 (insn_length_variant_t): Remove.
2725 (insn_length_parameters_t): Likewise.
2726 (arc_insn_length_parameters): Likewise.
2727 (arc_get_insn_variants): Likewise.
2728 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2730 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2732 PR inline-asm/84683
2733 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2734 assertion failure.
2736 PR tree-optimization/84687
2737 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
2738 on new_node->decl.
2739 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
2741 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2743 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
2744 Rename to ppc_speculation_barrier.
2745 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
2746 __builtin_ppc_speculation_barrier.
2748 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2750 PR target/84700
2751 * combine.c (combine_simplify_rtx): Don't try to simplify if
2752 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
2753 are equal to x.
2755 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
2757 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
2758 to 32 bytes when compiling for POWER9.
2760 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2762 PR target/84564
2763 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
2764 regparm >= 3 with no arg reg available also for calls with
2765 flag_force_indirect_call. Pass decl to ix86_function_regparm.
2767 PR target/84524
2768 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
2769 orig,vex.
2770 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
2772 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
2774 PR target/84264
2775 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
2777 2018-03-05 Richard Biener <rguenther@suse.de>
2779 PR tree-optimization/84486
2780 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
2781 When inserting a __builtin_assume_aligned call set the LHS
2782 SSA name alignment info accordingly.
2784 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
2786 PR tree-optimization/84114
2787 * config/aarch64/aarch64.c (aarch64_reassociation_width)
2788 Avoid reassociation of FLOAT_MODE addition.
2790 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
2792 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2793 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2794 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2795 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2796 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2797 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2798 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2799 and -mwbnoinvd.
2800 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2801 __builtin_ia32_wbinvd): New builtins.
2802 (SPECIAL_ARGS2): New.
2803 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2804 (SPECIAL_ARGS2): New.
2805 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2806 (ix86_valid_target_attribute_inner_p): Ditto.
2807 (ix86_init_mmx_sse_builtins): Add special_args2.
2808 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2809 TARGET_WBNOINVD_P): New.
2810 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2811 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2812 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2813 * config/i386/immintrin.h (_wbinvd): New intrinsic.
2814 * config/i386/pconfigintrin.h: New file.
2815 * config/i386/wbnoinvdintrin.h: Ditto.
2816 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2817 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2819 2018-03-05 Richard Biener <rguenther@suse.de>
2821 PR tree-optimization/84670
2822 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2823 member.
2824 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2825 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2826 (compute_antic_aux): Only assert the number of values in ANTIC_IN
2827 doesn't grow if all successors (recursively) were visited at least
2828 once.
2830 2018-03-05 Richard Biener <rguenther@suse.de>
2832 PR tree-optimization/84650
2833 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2834 if executed in the loop pipeline.
2836 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
2838 * doc/configfiles.texi (Configuration Files): Move info about
2839 conditionalizing $target-protos.h to...
2840 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
2841 differs from $target-protos.h.
2843 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
2844 Chung-Ju Wu <jasonwucj@gmail.com>
2846 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2847 * config/nds32/nds32-multiple.md (setmemsi): Define.
2848 * config/nds32/nds32-memory-manipulation.c
2849 (nds32_gen_dup_4_byte_to_word_value): New.
2850 (emit_setmem_word_loop): New.
2851 (emit_setmem_byte_loop): New.
2852 (nds32_expand_setmem_loop): New.
2853 (nds32_expand_setmem_loop_v3m): New.
2854 (nds32_expand_setmem_unroll): New.
2855 (nds32_expand_setmem): New.
2857 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2858 Chung-Ju Wu <jasonwucj@gmail.com>
2860 * config/nds32/nds32-memory-manipulation.c
2861 (nds32_emit_load_store): New.
2862 (nds32_emit_post_inc_load_store): New.
2863 (nds32_emit_mem_move): New.
2864 (nds32_emit_mem_move_block): New.
2865 (nds32_expand_movmemsi_loop_unknown_size): New.
2866 (nds32_expand_movmemsi_loop_known_size): New.
2867 (nds32_expand_movmemsi_loop): New.
2868 (nds32_expand_movmemsi_unroll): New.
2869 (nds32_expand_movmemqi): Rename ...
2870 (nds32_expand_movmemsi): ... to this.
2871 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2872 (movmemsi): ... to this.
2873 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2874 (nds32_expand_movmemsi): ... to this.
2876 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2877 Monk Chiang <sh.chiang04@gmail.com>
2878 Chung-Ju Wu <jasonwucj@gmail.com>
2880 * config/nds32/nds32-protos.h
2881 (nds32_expand_load_multiple): New arguments.
2882 (nds32_expand_store_multiple): Ditto.
2883 (nds32_valid_multiple_load_store): Rename ...
2884 (nds32_valid_multiple_load_store_p): ... to this.
2885 * config/nds32/nds32-memory-manipulation.c
2886 (nds32_expand_load_multiple): Refine implementation.
2887 (nds32_expand_store_multiple): Ditto.
2888 * config/nds32/nds32-multiple.md
2889 (load_multiple): Update nds32_expand_load_multiple interface.
2890 (store_multiple): Update nds32_expand_store_multiple interface.
2891 * config/nds32/nds32-predicates.c
2892 (nds32_valid_multiple_load_store): Rename ...
2893 (nds32_valid_multiple_load_store_p): ... to this and refine
2894 implementation.
2895 * config/nds32/predicates.md
2896 (nds32_load_multiple_and_update_address_operation): New predicate.
2897 (nds32_store_multiple_and_update_address_operation): New predicate.
2899 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2900 Chung-Ju Wu <jasonwucj@gmail.com>
2902 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2903 (combo): New attribute.
2904 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2906 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2908 * config/nds32/nds32.opt: Change -mcmodel= default value.
2910 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2911 Monk Chiang <sh.chiang04@gmail.com>
2912 Chung-Ju Wu <jasonwucj@gmail.com>
2914 * config/nds32/constants.md (unspec_element): New enum.
2915 * config/nds32/constraints.md (Umw): New constraint.
2916 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2917 * config/nds32/nds32-intrinsic.md: Likewise.
2918 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
2919 (nds32_valid_smw_lwm_base_p): New.
2920 (nds32_output_smw_single_word): New.
2921 (nds32_output_lmw_single_word): New.
2922 (nds32_expand_unaligned_load): New.
2923 (nds32_expand_unaligned_store): New.
2924 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
2925 (nds32_output_smw_single_word): Declare.
2926 (nds32_output_lmw_single_word): Declare.
2927 (nds32_expand_unaligned_load): Declare.
2928 (nds32_expand_unaligned_store): Declare.
2929 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
2930 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
2931 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
2932 NDS32_BUILTIN_UASTORE_DW.
2933 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
2934 predicate.
2936 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
2937 Kito Cheng <kito.cheng@gmail.com>
2938 Chung-Ju Wu <jasonwucj@gmail.com>
2940 * config/nds32/nds32-intrinsic.c
2941 (nds32_expand_builtin_null_ftype_reg): Delete.
2942 (nds32_expand_builtin_reg_ftype_imm): Ditto.
2943 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
2944 (nds32_read_argument): New.
2945 (nds32_legitimize_target): Ditto.
2946 (nds32_legitimize_argument): Ditto.
2947 (nds32_check_constant_argument): Ditto.
2948 (nds32_expand_unop_builtin): Ditto.
2949 (nds32_expand_unopimm_builtin): Ditto.
2950 (nds32_expand_binop_builtin): Ditto.
2951 (nds32_builtin_decl_impl): Ditto.
2952 (builtin_description): Ditto.
2953 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
2954 (nds32_init_builtins_impl): Ditto.
2955 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
2956 (nds32_builtin_decl): New.
2957 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
2958 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
2960 2018-03-02 Jeff Law <law@redhat.com>
2962 * reorg.c (stop_search_p): Handle DEBUG_INSN.
2963 (redundant_insn, fill_simple_delay_slots): Likewise.
2964 (fill_slots_from_thread): Likewise.
2965 * resource.c (mark_referenced_resources): Likewise.
2966 (mark_set_resources, find_dead_or_set_registers): Likewise.
2968 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2970 * substring-locations.h (format_warning_va): Formatting fix for
2971 ATTRIBUTE_GCC_DIAG.
2972 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
2973 argument.
2974 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
2975 * substring-locations.c: Include intl.h.
2976 (format_warning_va): Turned into small wrapper around
2977 format_warning_n_va, renamed to ...
2978 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
2979 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
2980 use ngettext.
2981 (format_warning_at_substring_n): New function.
2982 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
2983 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
2984 format_warning_at_substring with just a shorter name instead of
2985 const function pointer.
2986 (fmtwarn_n): New function.
2987 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
2988 appropriate, get rid of all the fmtstr temporaries, move conditionals
2989 with G_() wrapped string literals directly into fmtwarn arguments,
2990 cast dir.len to (int), formatting fixes.
2992 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
2994 * doc/invoke.texi: Remove "Cilk Plus" references.
2996 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2997 Richard Biener <rguenther@suse.de>
2999 PR ipa/84628
3000 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3001 for error or warning attributes if CALL_FROM_THUNK_P is set.
3002 Formatting fixes.
3004 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3006 PR target/56540
3007 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3008 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3010 PR target/56540
3011 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3012 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3014 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3015 instead of -1U in last predictors element's probability member.
3017 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3019 PR ipa/83983
3020 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3021 arguments if they are comparable.
3023 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3025 PR tree-optimization/84634
3026 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3027 masks and masked_loop_p with a single loop_masks, making sure it's
3028 null for bb vectorization.
3030 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3032 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3033 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3034 loop->force_vectorize to check whether there is no alias.
3036 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3038 PR target/84614
3039 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3040 prototypes.
3041 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3042 comments.
3043 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3044 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3045 instead of a loop around prev_real_insn.
3046 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3047 prev_real_insn.
3049 PR inline-asm/84625
3050 * config/i386/i386.c (ix86_print_operand): Use conditional
3051 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3052 zero vector.
3054 2018-03-02 Richard Biener <rguenther@suse.de>
3056 PR tree-optimization/84427
3057 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3058 (bitmap_set_subtract_values): Rewrite to handle multiple
3059 exprs per value.
3060 (clean): Likewise.
3061 (prune_clobbered_mems): Likewise.
3062 (phi_translate): Take edge instead of pred/phiblock.
3063 (phi_translate_1): Likewise.
3064 (phi_translate_set): Likewise. Insert all translated
3065 exprs for a value into the set, keeping possibly multiple
3066 expressions per value.
3067 (compute_antic_aux): Adjust for phi_translate changes.
3068 When intersecting union the expressions and prune those
3069 not in the final value set, keeping possibly multiple
3070 expressions per value. Do not use value-insertion
3071 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3072 all expressions. Add verification that the value-sets
3073 only shrink during iteration.
3074 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3075 (do_pre_regular_insertion): Likewise.
3076 (do_pre_partial_partial_insertion): Likewise.
3078 2018-03-02 Richard Biener <rguenther@suse.de>
3080 PR target/82005
3081 * config/darwin.c (saved_debug_info_level): New static global.
3082 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3083 (darwin_asm_lto_end): Restore debug info generation settings.
3085 2018-03-01 Martin Liska <mliska@suse.cz>
3087 PR sanitizer/82484
3088 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3089 volatile arguments.
3091 2018-03-01 Richard Biener <rguenther@suse.de>
3093 PR debug/84645
3094 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3095 type annotation with LTO when debug was disabled at compile-time.
3097 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3099 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3100 XINT with INTVAL.
3101 (mips_final_postscan_insn): Likewise.
3103 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3105 PR rtl-optimization/84528
3106 * alias.c (init_alias_target): Add commentary.
3107 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3108 a unique base value if the frame pointer is not eliminated
3109 to the stack pointer.
3111 2018-03-01 Tom de Vries <tom@codesourcery.com>
3113 PR rtl-optimization/83327
3114 * lra-int.h (hard_regs_spilled_into): Declare.
3115 * lra.c (hard_regs_spilled_into): Define.
3116 (init_reg_info): Init hard_regs_spilled_into.
3117 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3118 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3119 (process_bb_lives): Handle hard_regs_spilled_into.
3120 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3121 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3123 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3125 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3126 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3127 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3128 * config/rs6000/aix72.h: New file.
3130 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3132 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3133 instead of warning_at with conditional singular and plural messages
3134 where possible.
3136 PR target/52991
3137 * stor-layout.c (update_alignment_for_field): For
3138 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3139 && !DECL_PACKED (field), do the alignment update, just use
3140 only desired_align instead of MAX (type_align, desired_align)
3141 as the alignment.
3142 (place_field): Don't do known_align < desired_align handling
3143 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3144 is non-NULL, instead do it after rli->prev_field handling and
3145 only if not within a bitfield word. For DECL_PACKED (field)
3146 use type_align of BITS_PER_UNIT.
3148 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3150 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3151 superfluous parentheses and trailing spaces.
3153 2018-02-28 Richard Biener <rguenther@suse.de>
3155 PR tree-optimization/84584
3156 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3157 SCoPs with fake exit edge.
3159 2018-02-28 Martin Liska <mliska@suse.cz>
3161 PR testsuite/84597
3162 * timevar.c (timer::print): Fix format to properly print 100%
3163 values.
3165 2018-02-28 Richard Biener <rguenther@suse.de>
3167 PR middle-end/84607
3168 * genmatch.c (capture_info::walk_match): Do not mark
3169 captured expressions without operands as expr_p given
3170 they act more like predicates and should be subject to
3171 "lost tail" side-effect preserving.
3173 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3175 PR rtl-optimization/81611
3176 * auto-inc-dec.c (attempt_change): Move dead note from
3177 mem_insn if it's the next use of regno
3178 (find_address): Take address use of reg holding
3179 non-incremented value. Add parm to limit search to the named
3180 reg only.
3181 (merge_in_block): Attempt to use a mem insn that is the next
3182 use of the original regno.
3184 2018-02-27 Martin Sebor <msebor@redhat.com>
3186 PR c++/83871
3187 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3188 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3190 2018-02-27 Martin Sebor <msebor@redhat.com>
3192 PR translation/84207
3193 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3194 n argument to unsigned HOST_WIDE_INT.
3195 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3196 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3197 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3198 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3200 2018-02-27 Richard Biener <rguenther@suse.de>
3202 PR tree-optimization/84512
3203 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3204 Do not use the estimate returned from record_stmt_cost for
3205 the scalar iteration cost but sum properly using add_stmt_cost.
3207 2018-02-27 Richard Biener <rguenther@suse.de>
3209 PR tree-optimization/84466
3210 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3211 Adjust last change to less strictly validate use operands.
3213 2018-02-27 Martin Liska <mliska@suse.cz>
3215 PR gcov-profile/84548
3216 * gcov.c (process_file): Allow partial overlap and consider it
3217 also as group functions.
3218 (output_lines): Properly calculate range of lines for a group.
3220 2018-02-27 Martin Liska <mliska@suse.cz>
3222 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3223 'ggc' suffixes. Change first column width.
3224 (timer::print): Fix formatting of the column.
3226 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3228 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3229 preserve inline entry blocks for the sake of debug inline
3230 entry point markers alone.
3231 (remove_unused_locals): Suggest in comments a better place to
3232 force the preservation of inline entry blocks that are
3233 otherwise unused, but do not preserve them.
3235 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3237 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3239 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3241 PR target/84039
3242 * config/i386/constraints.md (Bs): Replace
3243 ix86_indirect_branch_register with
3244 TARGET_INDIRECT_BRANCH_REGISTER.
3245 (Bw): Likewise.
3246 * config/i386/i386.md (indirect_jump): Likewise.
3247 (tablejump): Likewise.
3248 (*sibcall_memory): Likewise.
3249 (*sibcall_value_memory): Likewise.
3250 Peepholes of indirect call and jump via memory: Likewise.
3251 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3252 (*sibcall_value_GOT_32): Likewise.
3253 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3254 (GOT_memory_operand): Likewise.
3255 (call_insn_operand): Likewise.
3256 (sibcall_insn_operand): Likewise.
3257 (GOT32_symbol_operand): Likewise.
3258 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3260 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3262 PR rtl-optimization/83496
3263 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3264 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3265 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3266 redundant insn, if any.
3267 (relax_delay_slots): Likewise.
3268 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3270 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3272 PR tree-optimization/83965
3273 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3274 that grouped statements are part of a reduction chain. Return
3275 true if the statement is not marked as a reduction itself but
3276 is part of a group.
3277 (vect_recog_dot_prod_pattern): Don't check whether the statement
3278 is part of a group here.
3279 (vect_recog_sad_pattern): Likewise.
3280 (vect_recog_widen_sum_pattern): Likewise.
3282 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3284 PR debug/84545
3285 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3287 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3289 PR target/84530
3290 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3291 the bool argument.
3292 (ix86_output_indirect_function_return): New prototype.
3293 (ix86_split_simple_return_pop_internal): Likewise.
3294 * config/i386/i386.c (indirect_return_via_cx): New.
3295 (indirect_return_via_cx_bnd): Likewise.
3296 (indirect_thunk_name): Handle return va CX_REG.
3297 (output_indirect_thunk_function): Create alias for
3298 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3299 (ix86_output_indirect_jmp): Remove the bool argument.
3300 (ix86_output_indirect_function_return): New function.
3301 (ix86_split_simple_return_pop_internal): Likewise.
3302 * config/i386/i386.md (*indirect_jump): Don't pass false
3303 to ix86_output_indirect_jmp.
3304 (*tablejump_1): Likewise.
3305 (simple_return_pop_internal): Change it to define_insn_and_split.
3306 Call ix86_split_simple_return_pop_internal to split it for
3307 -mfunction-return=.
3308 (simple_return_indirect_internal): Call
3309 ix86_output_indirect_function_return instead of
3310 ix86_output_indirect_jmp.
3312 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3314 PR bootstrap/84405
3315 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3316 memset and value initialization afterwards.
3318 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3320 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3322 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3324 PR target/84521
3325 * common/config/aarch64/aarch64-common.c
3326 (aarch_option_optimization_table[]): Switch
3327 off fomit-frame-pointer
3329 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3330 Chung-Ju Wu <jasonwucj@gmail.com>
3332 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3333 volatile memory.
3334 (store_multiple): Ditto.
3336 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3338 * config.gcc: Add --with-cpu support for nds32 target.
3339 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3340 * config/nds32/nds32.opt: Add -mcpu= option.
3342 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3344 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3345 isel=yes): Warn for these deprecated options.
3347 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3349 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3350 ISA_2_5_MASKS_EMBEDDED.
3352 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3354 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3355 p->max as pointers rather than using iterative_hash_expr.
3357 2018-02-23 Carl Love <cel@us.ibm.com>
3359 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3360 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3361 BU_P8V_OVERLOAD_2.
3362 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3363 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3364 P8V_BUILTIN_VEC_VUNSIGNED2.
3366 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3368 PR target/81572
3369 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3370 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3371 LRA_UNKNOWN_ALT.
3372 * lra-constraints.c (curr_insn_transform): Set up
3373 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3374 LRA_UNKNOWN_ALT.
3375 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3376 * lra-eliminations.c (spill_pseudos): Ditto.
3377 (process_insn_for_elimination): Ditto.
3378 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3379 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3380 LRA_NON_CLOBBERED_ALT.
3382 2018-02-22 Martin Sebor <msebor@redhat.com>
3384 PR tree-optimization/84480
3385 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3386 to maybe_diag_stxncpy_trunc. Call it.
3387 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3388 from gimple_fold_builtin_strcpy. Print inlining stack.
3389 (handle_builtin_stxncpy): Print inlining stack.
3390 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3392 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3394 PR target/84176
3395 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3396 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3397 and -fcheck-pointer-bounds are used together.
3398 (indirect_thunk_prefix): New enum.
3399 (indirect_thunk_need_prefix): New function.
3400 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3401 "_nt" instead of "_bnd" for NOTRACK prefix.
3402 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3403 (output_indirect_thunk_function): Likewise.
3404 (): Likewise.
3405 (ix86_code_end): Update output_indirect_thunk_function calls.
3406 (ix86_output_indirect_branch_via_reg): Replace
3407 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3408 (ix86_output_indirect_branch_via_push): Likewise.
3409 (ix86_output_function_return): Likewise.
3410 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3411 incompatible with -fcf-protection=branch and
3412 -fcheck-pointer-bounds.
3414 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3416 PR target/83335
3417 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3418 Change gcc_assert call to output_operand_lossage.
3420 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3422 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3424 2018-02-22 DJ Delorie <dj@redhat.com>
3425 Sebastian Perta <sebastian.perta@renesas.com>
3426 Oleg Endo <olegendo@gcc.gnu.org>
3428 * config/rx/rx.c (rx_rtx_costs): New function.
3429 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3431 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3433 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3435 2018-02-22 Martin Liska <mliska@suse.cz>
3437 PR driver/83193
3438 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3439 Add "native" as a possible value.
3441 2018-02-22 Martin Liska <mliska@suse.cz>
3443 PR driver/83193
3444 * config/i386/i386.c (ix86_option_override_internal):
3445 Add "native" as a possible value for -march and -mtune.
3447 2018-02-22 Jakub Jelinek <jakub@redhat.com>
3449 PR target/84502
3450 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
3451 to all type variants.
3453 PR tree-optimization/84503
3454 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
3455 width as info->bitpos + info->bitsize - start.
3456 (merged_store_group::merge_overlapping): Simplify width computation.
3457 (check_no_overlap): New function.
3458 (imm_store_chain_info::try_coalesce_bswap): Compute expected
3459 start + width and last_order of the group, fail if check_no_overlap
3460 fails.
3461 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
3462 to group if check_no_overlap fails.
3464 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3466 * config/rs6000/altivec.md: Delete contraint arguments to
3467 define_expand, define_split, and define_peephole2, and in
3468 define_insn_and_split if always unused.
3469 * config/rs6000/darwin.md: Ditto.
3470 * config/rs6000/dfp.md: Ditto.
3471 * config/rs6000/rs6000.md: Ditto.
3472 * config/rs6000/sync.md: Ditto.
3473 * config/rs6000/vector.md: Ditto.
3474 * config/rs6000/vsx.md: Ditto.
3476 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3478 * config/rs6000/altivec.md: Write output control strings as braced
3479 blocks instead of double-quoted strings.
3480 * config/rs6000/darwin.md: Ditto.
3481 * config/rs6000/rs6000.md: Ditto.
3482 * config/rs6000/vector.md: Ditto.
3483 * config/rs6000/vsx.md: Ditto.
3485 2018-02-21 Jason Merrill <jason@redhat.com>
3487 PR c++/84314 - ICE with templates and fastcall attribute.
3488 * attribs.c (build_type_attribute_qual_variant): Remove assert.
3490 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3492 * ipa-cp.c (determine_versionability): Fix comment typos.
3494 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3496 PR c/84229
3497 * ipa-cp.c (determine_versionability): Do not version functions caling
3498 va_arg_pack.
3500 2018-02-21 Martin Liska <mliska@suse.cz>
3502 PR driver/83193
3503 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3504 Add "native" as a possible value.
3505 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3506 the macro when native cpu detection is available.
3508 2018-02-21 Martin Liska <mliska@suse.cz>
3510 PR driver/83193
3511 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3512 Add "native" as a possible value.
3513 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3514 when native cpu detection is available.
3516 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3517 Martin Sebor <msebor@redhat.com>
3519 PR tree-optimization/84478
3520 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3521 false.
3522 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3523 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3524 support which is conservatively correct, for 2 only stay conservative
3525 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3526 argument to the 2 argument get_range_strlen, adjust 6 arg
3527 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3528 false.
3529 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3530 (gimple_fold_builtin_strlen): Pass true as last argument to
3531 get_range_strlen.
3533 2018-02-20 Martin Sebor <msebor@redhat.com>
3535 PR middle-end/84095
3536 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3537 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3538 (builtin_memref::builtin_memref): Factor out parts into
3539 set_base_and_offset and call it.
3541 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3543 PR middle-end/84406
3544 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3545 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3546 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3547 search at the associated MODE_INT.
3549 2018-02-20 Jeff Law <law@redhat.com>
3551 PR middle-end/82123
3552 PR tree-optimization/81592
3553 PR middle-end/79257
3554 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3555 for range data rather than using global data.
3556 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3557 range data rather than using global data.
3558 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3559 pass it to children as needed.
3560 (struct directive::fmtresult): Similarly.
3561 (struct directive::set_width): Similarly.
3562 (struct directive::set_precision): Similarly.
3563 (format_integer, format_directive, parse_directive): Similarly.
3564 (format_none): Accept unnamed vr_values parameter.
3565 (format_percent, format_floating, format_character): Similarly.
3566 (format_string, format_plain): Similarly.
3567 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3568 the EVRP range analyzer for range data rather than using global data.
3569 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3570 gimple-ssa-evrp-analyze.h
3571 (class sprintf_dom_walker): Add after_dom_children member function.
3572 Add evrp_range_analyzer member.
3573 (sprintf_dom_walker::before_dom_children): Call into the EVRP
3574 range analyzer as needed.
3575 (sprintf_dom_walker::after_dom_children): New member function.
3576 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3577 if not optimizing.
3578 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3579 (evrp_range_analyzer::pop_to_marker): Likewise.
3581 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3583 PR tree-optimization/84419
3584 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3585 with the required type if its current type is compatible but
3586 different.
3588 2018-02-20 Jakub Jelinek <jakub@redhat.com>
3590 PR middle-end/82004
3591 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3592 after vectorization.
3594 2018-02-20 Martin Liska <mliska@suse.cz>
3596 PR driver/83193
3597 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3598 possible values if we don't have a hint.
3600 2018-02-20 Martin Liska <mliska@suse.cz>
3602 PR c/84310
3603 PR target/79747
3604 * final.c (shorten_branches): Build align_tab array with one
3605 more element.
3606 * opts.c (finish_options): Add alignment option limit check.
3607 (MAX_CODE_ALIGN): Likewise.
3608 (MAX_CODE_ALIGN_VALUE): Likewise.
3609 * doc/invoke.texi: Document maximum allowed option value for
3610 all -falign-* options.
3612 2018-02-19 Jakub Jelinek <jakub@redhat.com>
3614 PR target/84146
3615 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3616 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3617 * var-tracking.c (emit_note_insn_var_location): Remove all references
3618 to NOTE_INSN_CALL_ARG_LOCATION.
3619 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3620 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3621 Use copy_rtx_if_shared.
3622 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3623 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3624 (dwarf2out_var_location): Remove handling of
3625 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3626 on call_insn.
3627 * final.c (final_scan_insn): Remove all references to
3628 NOTE_INSN_CALL_ARG_LOCATION.
3629 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
3630 before dumping final insns.
3631 * except.c (emit_note_eh_region_end): Remove all references to
3632 NOTE_INSN_CALL_ARG_LOCATION.
3633 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3634 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3635 * config/arc/arc.c (hwloop_optimize): Likewise.
3636 * config/arm/arm.c (create_fix_barrier): Likewise.
3637 * config/s390/s390.c (s390_chunkify_start): Likewise.
3638 * config/sh/sh.c (find_barrier): Likewise.
3639 * config/i386/i386.c (rest_of_insert_endbranch,
3640 ix86_seh_fixup_eh_fallthru): Likewise.
3641 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3642 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3643 * config/frv/frv.c (frv_function_prologue): Likewise.
3644 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3645 reg note.
3646 (note_outside_basic_block_p): Remove all references to
3647 NOTE_INSN_CALL_ARG_LOCATION.
3648 * gengtype.c (adjust_field_rtx_def): Likewise.
3649 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3650 Likewise.
3651 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3652 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3654 PR c++/84444
3655 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3656 is ADDR_EXPR.
3658 PR tree-optimization/84452
3659 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3660 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3661 is NULL.
3663 2018-02-19 Martin Liska <mliska@suse.cz>
3665 PR sanitizer/82183
3666 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3668 2018-02-19 Martin Liska <mliska@suse.cz>
3669 Richard Sandiford <richard.sandiford@linaro.org>
3671 PR tree-optimization/82491
3672 * gimple-fold.c (get_base_constructor): Make earlier bail out
3673 to prevent ubsan.
3675 2018-02-19 Carl Love <cel@us.ibm.com>
3677 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3678 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3679 BU_P8V_OVERLOAD_1.
3680 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3681 P8V_BUILTIN_VEC_NEG.
3683 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
3685 * config/rl78/rl78.md (movdf): New define expand.
3687 2018-02-19 Martin Liska <mliska@suse.cz>
3689 PR other/80589
3690 * doc/invoke.texi: Fix typo.
3691 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3693 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
3695 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3696 handle rs6000_single_float and rs6000_double_float specially for
3697 e500 family CPUs.
3699 2018-02-16 Jeff Law <law@redhat.com>
3701 * config/rx/rx.c (add_pop_cfi_notes): New function.;
3702 (pop_regs): Use it.
3704 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3706 PR ipa/84425
3707 * ipa-inline.c (inline_small_functions): Fix a typo.
3709 2018-02-16 Nathan Sidwell <nathan@acm.org>
3711 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3713 2018-02-16 Carl Love <cel@us.ibm.com>
3715 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3716 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3717 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3718 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3719 expansion to P8V_BUILTIN_VEC_FLOAT2.
3721 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
3723 PR rtl-optimization/70023
3724 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3725 src_regno into account.
3727 2018-02-16 Carl Love <cel@us.ibm.com>
3729 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3730 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3731 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3732 * config/rs6000/rs6000.c: Remove case statements for
3733 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3734 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3735 and P9V_BUILTIN_VEC_VINSERT4B.
3736 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
3737 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
3738 * config/rs6000/vsx.md:
3739 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
3740 vec_insert4b.
3742 2018-02-16 Carl Love <cel@us.ibm.com>
3744 * config/rs6000/altivec.h: Add builtin names vec_extract4b
3745 vec_insert4b.
3746 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
3747 definitions.
3748 * config/rs6000/rs6000-c.c: Add the definitions for
3749 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
3750 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
3751 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
3752 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
3753 definition for insert4b and define insn *insert3b_internal.
3754 * doc/extend.texi: Add documentation for vec_extract4b.
3756 2018-02-16 Nathan Sidwell <nathan@acm.org>
3758 * doc/extend.texi (Backwards Compatibility): Mention friend
3759 injection. Note for-scope is deprecated.
3760 * doc/invoke.texi (-ffriend-injection): Deprecate.
3762 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
3764 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
3765 that moved to I2, also allow destinations that are a paradoxical
3766 subreg (instead of a normal reg).
3768 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
3770 PR target/83831
3771 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
3772 to QImode.
3774 2018-02-16 Richard Biener <rguenther@suse.de>
3776 PR tree-optimization/84037
3777 PR tree-optimization/84016
3778 PR target/82862
3779 * config/i386/i386.c (ix86_builtin_vectorization_cost):
3780 Adjust vec_construct for the fact we need additional higher latency
3781 128bit inserts for AVX256 and AVX512 vector builds.
3782 (ix86_add_stmt_cost): Scale vector construction cost for
3783 elementwise loads.
3785 2018-02-16 Richard Biener <rguenther@suse.de>
3787 PR tree-optimization/84417
3788 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3789 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3790 (non_rewritable_lvalue_p): Likewise, use poly-ints.
3792 2018-02-16 Martin Liska <mliska@suse.cz>
3794 PR sanitizer/84307
3795 * internal-fn.def (ASAN_CHECK): Set proper flags.
3796 (ASAN_MARK): Likewise.
3798 2018-02-16 Julia Koval <julia.koval@intel.com>
3800 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3801 from PTA_CANNONLAKE.
3803 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3805 PR target/84272
3806 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3807 Use ++iter rather than iter++ for std::list iterators.
3808 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
3809 defer deleting them until all nodes in the forest are processed. Do
3810 free even leaf nodes. Change to_process into auto_vec.
3812 PR bootstrap/84405
3813 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3814 * vec.h (vec_default_construct): Use memset instead of placement new
3815 if BROKEN_VALUE_INITIALIZATION is defined.
3816 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3817 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3818 is defined.
3820 PR rtl-optimization/83723
3821 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3822 * lra.c (lra_substitute_pseudo): Likewise. If true, use
3823 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
3824 recursive calls.
3825 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3826 callers.
3827 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3829 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
3831 PR rtl-optimization/81443
3832 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3833 from inner REGs to paradoxical SUBREGs.
3835 2018-02-16 Richard Biener <rguenther@suse.de>
3837 PR tree-optimization/84399
3838 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3839 For operands we can analyze at their definition make sure we can
3840 analyze them at each use as well.
3842 2018-02-16 Richard Biener <rguenther@suse.de>
3844 PR tree-optimization/84190
3845 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3846 volatile accesses if the decl isn't volatile.
3848 2018-02-15 Jason Merrill <jason@redhat.com>
3850 PR c++/84314 - ICE with templates and fastcall attribute.
3851 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3852 TYPE_CANONICAL on an existing type.
3854 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3856 PR tree-optimization/84383
3857 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3858 dstoff nor call operand_equal_p if dstbase is NULL.
3860 PR tree-optimization/84334
3861 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3862 also a CONSTANT_CLASS_P, punt.
3864 2018-02-14 Jim Wilson <jimw@sifive.com>
3866 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3867 first SMALL_OPERAND check. New local min_second_step. Move assert
3868 to where locals are set. Add TARGET_RVC support.
3869 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3871 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3873 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3875 2018-02-14 Martin Sebor <msebor@redhat.com>
3877 PR tree-optimization/83698
3878 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3879 arrays constrain the offset range to their bounds.
3880 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3881 (builtin_access::overlap): Avoid setting the size of overlap if it's
3882 already been set.
3883 (maybe_diag_overlap): Also consider arrays when deciding what values
3884 of offsets to include in diagnostics.
3886 2018-02-14 Martin Sebor <msebor@redhat.com>
3888 PR c/84108
3889 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3890 that correspond to the kind of a declaration.
3892 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3894 PR target/83984
3895 * config/pa/pa.md: Load address of PIC label using the linkage table
3896 if the label is nonlocal.
3898 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3900 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3901 warning message if user requests -maltivec=be.
3902 * doc/invoke.texi: Document deprecation of -maltivec=be.
3904 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3906 PR target/84220
3907 * config/rs6000/rs6000-c.c: Update definitions for
3908 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3909 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3911 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3913 PR target/84239
3914 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3915 add _get_ssp intrinsics. Remove argument from
3916 __builtin_ia32_rdssp[d|q].
3917 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3918 * config/i386/i386-builtin.def: Remove argument from
3919 __builtin_ia32_rdssp[d|q].
3920 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
3921 ix86_expand_special_args_builtin for _rdssp[d|q].
3922 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
3923 Clear register before usage.
3924 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
3925 Add documentation for new _get_ssp and _inc_ssp intrinsics.
3927 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
3929 PR tree-optimization/84357
3930 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
3931 operand 1 of an ARRAY_REF too.
3933 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
3935 PR target/83831
3936 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
3937 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3938 declarations.
3939 (set_of_reg): New struct.
3940 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
3941 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
3942 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3943 functions.
3944 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
3945 Split into bitclr, bitset, bitinvert patterns if appropriate.
3946 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
3947 use rx_fuse_in_memory_bitop.
3948 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
3949 to named insn, correct maximum insn length.
3951 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
3953 PR target/79242
3954 * machmode.def: Define a complex mode for PARTIAL_INT.
3955 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
3956 MODE_PARTIAL_INT.
3957 * doc/rtl.texi: Document CSPImode.
3958 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
3959 handling.
3960 (msp430_hard_regno_nregs_with_padding): Likewise.
3962 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
3964 PR target/84279
3965 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
3967 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
3969 PR rtl-optimization/84169
3970 * combine.c (try_combine): New variable split_i2i3. Set it to true if
3971 we generated a parallel as new i3 and we split that to new i2 and i3
3972 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
3973 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
3974 those to i2, not i1. Partially rewrite this scan code.
3976 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3978 PR c/82210
3979 * stor-layout.c (place_field): For variable length fields, adjust
3980 offset_align afterwards not just based on the field's alignment,
3981 but also on the size.
3983 PR middle-end/84309
3984 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
3985 of exps and logs in the use_exp2 case.
3987 2018-02-13 Jeff Law <law@redhat.com>
3989 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
3990 entry for "vector".
3992 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
3993 ARGS as unused.
3995 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
3997 PR debug/84342
3998 PR debug/84319
3999 * common.opt (gas-loc-support, gas-locview-support): New.
4000 (ginline-points, ginternal-reset-location-views): New.
4001 * doc/invoke.texi: Document them. Use @itemx where intended.
4002 (gvariable-location-views): Adjust.
4003 * target.def (reset_location_view): New.
4004 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4005 (TARGET_RESET_LOCATION_VIEW): New.
4006 * doc/tm.texi: Rebuilt.
4007 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4008 (dwarf2out_default_as_locview_support): New.
4009 (output_asm_line_debug_info): Use option variables.
4010 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4011 (output_loc_list): Likewise.
4012 (add_high_low_attributes): Check option variables.
4013 Don't output entry view attribute in strict mode.
4014 (gen_inlined_subroutine_die): Check option variables.
4015 (dwarf2out_inline_entry): Likewise.
4016 (init_sections_and_labels): Likewise.
4017 (dwarf2out_early_finish): Likewise.
4018 (maybe_reset_location_view): New, from...
4019 (dwarf2out_var_location): ... here. Call it.
4020 * debug.h (dwarf2out_default_as_loc_support): Declare.
4021 (dwarf2out_default_as_locview_support): Declare.
4022 * hooks.c (hook_int_rtx_insn_0): New.
4023 * hooks.h (hook_int_rtx_insn_0): Declare.
4024 * toplev.c (process_options): Take -gas-loc-support and
4025 -gas-locview-support from dwarf2out. Enable
4026 -gvariable-location-views by default only with locview
4027 assembler support. Enable -ginternal-reset-location-views by
4028 default only if the target defines the corresponding hook.
4029 Enable -ginline-points by default if location views are
4030 enabled; force it disabled if statement frontiers are
4031 disabled.
4032 * tree-inline.c (expand_call_inline): Check option variables.
4033 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4035 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4037 PR tree-optimization/84321
4038 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4039 handling. Also check whether the anti-range contains any values
4040 that satisfy the mask; switch to a VR_RANGE if not.
4042 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4044 PR sanitizer/84340
4045 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4047 2018-02-13 Martin Jambor <mjambor@suse.cz>
4049 PR c++/83990
4050 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4051 of call statements, also set location of a load to a temporary.
4053 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4055 * config/rl78/rl78.c (add_vector_labels): New function.
4056 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4057 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4058 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4059 which checks that no arguments are passed.
4060 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4061 * doc/extend.texi: Documentation for the new attribute.
4063 2018-02-13 Andreas Schwab <schwab@suse.de>
4065 * config/riscv/linux.h (CPP_SPEC): Define.
4067 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4069 PR target/84335
4070 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4071 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4072 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4073 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4074 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4075 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4076 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4077 temporarily for AES and PCLMUL builtins.
4079 PR tree-optimization/84339
4080 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4081 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4082 Formatting fixes.
4084 PR middle-end/84309
4085 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4086 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4087 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4088 inline function.
4089 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4090 inline function.
4091 * omp-simd-clone.h: New file.
4092 * omp-simd-clone.c: Include omp-simd-clone.h.
4093 (expand_simd_clones): No longer static.
4094 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4095 cgraph.h and omp-simd-clone.h.
4096 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4097 (vect_recog_widen_shift_pattern): Formatting fix.
4098 (vect_pattern_recog_1): Don't check optab for calls.
4100 PR target/84336
4101 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4102 operands[2] into a REG before using gen_lowpart on it.
4104 2018-02-12 Jeff Law <law@redhat.com>
4106 PR target/83760
4107 * config/sh/sh.c (find_barrier): Consider a sibling call
4108 a barrier as well.
4110 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4111 successfully back substituting a reg.
4113 2018-02-12 Richard Biener <rguenther@suse.de>
4115 PR tree-optimization/84037
4116 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4117 parameter, move visited init to caller.
4118 (vect_slp_analyze_operations): Separate cost from validity
4119 check, initialize visited once for all instances.
4120 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4121 for all instances.
4122 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4123 out an assert.
4124 (vect_model_promotion_demotion_cost): Likewise.
4125 (vectorizable_bswap): Guard cost modeling with !slp_node
4126 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4127 SLP stmts.
4128 (vectorizable_call): Likewise.
4129 (vectorizable_conversion): Likewise.
4130 (vectorizable_assignment): Likewise.
4131 (vectorizable_shift): Likewise.
4132 (vectorizable_operation): Likewise.
4133 (vectorizable_store): Likewise.
4134 (vectorizable_load): Likewise.
4135 (vectorizable_condition): Likewise.
4136 (vectorizable_comparison): Likewise.
4138 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4140 PR sanitizer/84307
4141 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4142 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4143 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4145 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4147 PR middle-end/83665
4148 * params.def (inline-min-speedup): Increase from 8 to 15.
4149 (max-inline-insns-auto): Decrease from 40 to 30.
4150 * ipa-split.c (consider_split): Add some buffer for function to
4151 be considered inlining candidate.
4152 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4153 default values.
4155 2018-02-12 Richard Biener <rguenther@suse.de>
4157 PR tree-optimization/84037
4158 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4159 matched stmts if we cannot swap the non-matched ones.
4161 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4163 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4164 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4165 _mm_maskz_scalef_round_ss): New intrinsics.
4166 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4167 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4168 __builtin_ia32_scalefss_round): Remove.
4169 (__builtin_ia32_scalefsd_mask_round,
4170 __builtin_ia32_scalefss_mask_round): New intrinsics.
4171 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4172 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4173 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4174 "<round_constraint>")): Changed to ...
4175 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4176 "<round_scalar_constraint>")): ... this.
4177 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4178 %0, %1, %2<round_op3>}"): Changed to ...
4179 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4180 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4181 %2<round_scalar_mask_op3>}"): ... this.
4182 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4184 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4186 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4187 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4188 (_mm_maskz_sqrt_round_ss): New intrinsics.
4189 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4190 (__builtin_ia32_sqrtsd_mask_round)
4191 (__builtin_ia32_sqrtss_mask_round): New builtins.
4192 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4193 (__builtin_ia32_sqrtss_round): Remove.
4194 (__builtin_ia32_sqrtsd_mask_round)
4195 (__builtin_ia32_sqrtss_mask_round): New builtins.
4196 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4197 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4198 ((match_operand:VF_128 1 "vector_operand"
4199 "xBm,<round_constraint>")): Changed to ...
4200 ((match_operand:VF_128 1 "vector_operand"
4201 "xBm,<round_scalar_constraint>")): ... this.
4202 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4203 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4204 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4205 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4206 %<iptr>1<round_scalar_mask_op3>}): ... this.
4207 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4208 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4210 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4212 PR target/84266
4213 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4214 Cast vec_cmpeq result to correct type.
4215 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4216 Cast vec_cmpgt result to correct type.
4218 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4220 * final.c (final_scan_insn_1): Renamed from...
4221 (final_scan_insn): ... this. New wrapper, to recover
4222 seen from the outermost call in recursive ones.
4223 * config/sparc/sparc.c (output_return): Drop seen from call.
4224 (output_sibcall): Likewise.
4225 * config/visium/visium.c (output_branch): Likewise.
4227 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4229 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4230 function label.
4232 2018-02-10 Alan Modra <amodra@gmail.com>
4234 PR target/84300
4235 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4236 Specify LR as an input.
4238 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4240 PR sanitizer/83987
4241 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4242 remove_member_access_dummy_vars): New functions.
4243 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4244 lower_omp_1, execute_lower_omp): Use them.
4246 PR rtl-optimization/84308
4247 * shrink-wrap.c (spread_components): Release todo vector.
4249 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4251 PR rtl-optimization/57193
4252 * ira-color.c (struct allocno_color_data): Add member
4253 conflict_allocno_hard_prefs.
4254 (update_conflict_allocno_hard_prefs): New.
4255 (bucket_allocno_compare_func): Add a preference based on
4256 conflict_allocno_hard_prefs.
4257 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4258 (color_allocnos): Remove a dead code. Initiate
4259 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4261 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4263 PR target/84226
4264 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4265 constraint from =wa to wa. Avoid a subreg on the output operand,
4266 instead use a pseudo and subreg it in a move.
4267 (p9_xxbrd_<mode>): Changed to ...
4268 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4269 (p9_xxbrd_v2df): New expander.
4270 (p9_xxbrw_<mode>): Changed to ...
4271 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4272 (p9_xxbrw_v4sf): New expander.
4274 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4276 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4278 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4280 PR target/83926
4281 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4282 multiply in 32-bit mode.
4283 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4284 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4285 mode.
4287 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4289 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4290 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4291 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4292 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4294 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4296 PR lto/84213
4297 * dwarf2out.c (is_trivial_indirect_ref): New function.
4298 (dwarf2out_late_global_decl): Do not generate a location
4299 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4300 and that are not defined in the current unit.
4302 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4304 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4305 instead of a libcall for UNORDERED.
4307 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4309 PR target/82641
4310 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4311 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4313 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4315 PR target/PR84295
4316 * config/s390/s390.c (s390_set_current_function): Invoke
4317 s390_indirect_branch_settings also if fndecl didn't change.
4319 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4321 * config/rs6000/rs6000.md (blockage): Set length to zero.
4323 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4325 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4327 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4329 PR sanitizer/84285
4330 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4331 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4332 -static-lib*san.
4334 PR debug/84252
4335 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4336 PARALLEL incoming that failed vt_get_decl_and_offset check.
4338 PR middle-end/84237
4339 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4340 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4341 TREE_READONLY bit.
4342 (get_variable_section): For decls in named .bss* sections pass true as
4343 second argument to bss_initializer_p.
4345 2018-02-09 Marek Polacek <polacek@redhat.com>
4346 Jakub Jelinek <jakub@redhat.com>
4348 PR c++/83659
4349 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4350 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4351 Sync some changes from cxx_fold_indirect_ref.
4353 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4355 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4356 markers.
4357 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4358 (BLOCK_INLINE_ENTRY_LABEL): New.
4359 (dwarf2out_var_location): Disregard inline entry markers.
4360 (inline_entry_data): New struct.
4361 (inline_entry_data_hasher): New hashtable type.
4362 (inline_entry_data_hasher::hash): New.
4363 (inline_entry_data_hasher::equal): New.
4364 (inline_entry_data_table): New variable.
4365 (add_high_low_attributes): Add DW_AT_entry_pc and
4366 DW_AT_GNU_entry_view attributes if a pending entry is found
4367 in inline_entry_data_table. Add old entry_pc attribute only
4368 if debug nonbinding markers are disabled.
4369 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4370 markers are enabled.
4371 (block_within_block_p, dwarf2out_inline_entry): New.
4372 (dwarf2out_finish): Check that no entries remained in
4373 inline_entry_data_table.
4374 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4375 (final_scan_insn, notice_source_line): Likewise.
4376 (rest_of_clean_state): Skip inline entry markers.
4377 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4378 markers.
4379 * gimple.c (gimple_build_debug_inline_entry): New.
4380 * gimple.h (enum gimple_debug_subcode): Add
4381 GIMPLE_DEBUG_INLINE_ENTRY.
4382 (gimple_build_debug_inline_entry): Declare.
4383 (gimple_debug_inline_entry_p): New.
4384 (gimple_debug_nonbind_marker_p): Adjust.
4385 * insn-notes.def (INLINE_ENTRY): New.
4386 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4387 inline entry marker notes.
4388 (print_insn): Likewise.
4389 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4390 (INSN_DEBUG_MARKER_KIND): Likewise.
4391 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4392 * tree-inline.c (expand_call_inline): Build and insert
4393 debug_inline_entry stmt.
4394 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4395 inline entry blocks early, if nonbind markers are enabled.
4396 (dump_scope_block): Dump fragment info.
4397 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4398 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4399 (gimple_build_debug_inline_entry): New.
4400 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4401 Enable/disable inline entry points too.
4402 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4403 (DEBUG_INSN): Describe inline entry markers.
4405 * common.opt (gvariable-location-views): New.
4406 (gvariable-location-views=incompat5): New.
4407 * config.in: Rebuilt.
4408 * configure: Rebuilt.
4409 * configure.ac: Test assembler for view support.
4410 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4411 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4412 * dwarf2out.c (var_loc_view): New typedef.
4413 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4414 (dwarf2out_locviews_in_attribute): New.
4415 (dwarf2out_locviews_in_loclist): New.
4416 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4417 (enum dw_line_info_opcode): Add LI_adv_address.
4418 (struct dw_line_info_table): Add view.
4419 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4420 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4421 (zero_view_p): New variable.
4422 (ZERO_VIEW_P): New macro.
4423 (output_asm_line_debug_info): New.
4424 (struct var_loc_node): Add view.
4425 (add_AT_view_list, AT_loc_list): New.
4426 (add_var_loc_to_decl): Add view param. Test it against last.
4427 (new_loc_list): Add view params. Record them.
4428 (AT_loc_list_ptr): Handle loc and view lists.
4429 (view_list_to_loc_list_val_node): New.
4430 (print_dw_val): Handle dw_val_class_view_list.
4431 (size_of_die): Likewise.
4432 (value_format): Likewise.
4433 (loc_list_has_views): New.
4434 (gen_llsym): Set vl_symbol too.
4435 (maybe_gen_llsym, skip_loc_list_entry): New.
4436 (dwarf2out_maybe_output_loclist_view_pair): New.
4437 (output_loc_list): Output view list or entries too.
4438 (output_view_list_offset): New.
4439 (output_die): Handle dw_val_class_view_list.
4440 (output_dwarf_version): New.
4441 (output_compilation_unit_header): Use it.
4442 (output_skeleton_debug_sections): Likewise.
4443 (output_rnglists, output_line_info): Likewise.
4444 (output_pubnames, output_aranges): Update version comments.
4445 (output_one_line_info_table): Output view numbers in asm comments.
4446 (dw_loc_list): Determine current endview, pass it to new_loc_list.
4447 Call maybe_gen_llsym.
4448 (loc_list_from_tree_1): Adjust.
4449 (add_AT_location_description): Create view list attribute if
4450 needed, check it's absent otherwise.
4451 (convert_cfa_to_fb_loc_list): Adjust.
4452 (maybe_emit_file): Call output_asm_line_debug_info for test.
4453 (dwarf2out_var_location): Reset views as needed. Precompute
4454 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
4455 attribute. Set view.
4456 (new_line_info_table): Reset next view.
4457 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
4458 (dwarf2out_source_line): Likewise. Output view resets and labels to
4459 the assembler, or select appropriate line info opcodes.
4460 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
4461 (optimize_string_length): Catch it. Adjust.
4462 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
4463 dw_val_class_view_list, and remove it if no longer needed.
4464 (hash_loc_list): Hash view numbers.
4465 (loc_list_hasher::equal): Compare them.
4466 (optimize_location_lists): Check whether a view list symbol is
4467 needed, and whether the locview attribute is present, and
4468 whether they match. Remove the locview attribute if no longer
4469 needed.
4470 (index_location_lists): Call skip_loc_list_entry for test.
4471 (dwarf2out_finish): Call output_asm_line_debug_info for test.
4472 Use output_dwarf_version.
4473 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
4474 (struct dw_val_node): Add val_view_list.
4475 * final.c (SEEN_NEXT_VIEW): New.
4476 (set_next_view_needed): New.
4477 (clear_next_view_needed): New.
4478 (maybe_output_next_view): New.
4479 (final_start_function): Rename to...
4480 (final_start_function_1): ... this. Take pointer to FIRST,
4481 add SEEN parameter. Emit param bindings in the initial view.
4482 (final_start_function): Reintroduce SEEN-less interface.
4483 (final): Rename to...
4484 (final_1): ... this. Take SEEN parameter. Output final pending
4485 next view at the end.
4486 (final): Reintroduce seen-less interface.
4487 (final_scan_insn): Output pending next view before switching
4488 sections or ending a block. Mark the next view as needed when
4489 outputting variable locations. Notify debug backend of section
4490 changes, and of location view changes.
4491 (rest_of_handle_final): Adjust.
4492 * toplev.c (process_options): Autodetect value for debug variable
4493 location views option. Warn on incompat5 without -gdwarf-5.
4494 * doc/invoke.texi (gvariable-location-views): New.
4495 (gvariable-location-views=incompat5): New.
4496 (gno-variable-location-views): New.
4498 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4500 PR tree-optimization/84136
4501 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4502 that the result of find_edge is non-NULL.
4504 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4506 PR target/83008
4507 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4508 storing integer register in SImode. Fix cost of 256 and 512
4509 byte aligned SSE register store.
4511 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4513 * config/i386/i386.c (ix86_multiplication_cost): Fix
4514 multiplication cost for TARGET_AVX512DQ.
4516 2018-02-08 Marek Polacek <polacek@redhat.com>
4518 PR tree-optimization/84238
4519 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4520 get_range_strlen.
4522 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4524 PR tree-optimization/84265
4525 * tree-vect-stmts.c (vectorizable_store): Don't treat
4526 VMAT_CONTIGUOUS accesses as grouped.
4527 (vectorizable_load): Likewise.
4529 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4531 PR tree-optimization/81635
4532 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4533 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4534 (test_round_for_mask): New functions.
4535 (wide_int_cc_tests): Call test_round_for_mask.
4536 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4537 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4538 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4539 range returned by get_range_info.
4541 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4543 PR ipa/81360
4544 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4545 * symtab.c: Include builtins.h
4546 (symtab_node::output_to_lto_symbol_table_p): Move here
4547 from lto-streamer-out.c:output_symbol_p.
4548 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4549 (output_symbol_p): Move all logic to symtab.c
4550 (produce_symtab): Update.
4552 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4554 * config/s390/s390-opts.h (enum indirect_branch): Define.
4555 * config/s390/s390-protos.h (s390_return_addr_from_memory)
4556 (s390_indirect_branch_via_thunk)
4557 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4558 (enum s390_indirect_branch_type): Define.
4559 * config/s390/s390.c (struct s390_frame_layout, struct
4560 machine_function): Remove.
4561 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4562 (indirect_branch_table_label_no, indirect_branch_table_name):
4563 Define variables.
4564 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4565 (enum s390_indirect_branch_option): Define.
4566 (s390_return_addr_from_memory): New function.
4567 (s390_handle_string_attribute): New function.
4568 (s390_attribute_table): Add new attribute handler.
4569 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4570 (s390_indirect_branch_via_thunk): New function.
4571 (s390_indirect_branch_via_inline_thunk): New function.
4572 (s390_function_ok_for_sibcall): When jumping via thunk disallow
4573 sibling call optimization for non z10 compiles.
4574 (s390_emit_call): Force indirect branch target to be a single
4575 register. Add r1 clobber for non-z10 compiles.
4576 (s390_emit_epilogue): Emit return jump via return_use expander.
4577 (s390_reorg): Handle JUMP_INSNs as execute targets.
4578 (s390_option_override_internal): Perform validity checks for the
4579 new command line options.
4580 (s390_indirect_branch_attrvalue): New function.
4581 (s390_indirect_branch_settings): New function.
4582 (s390_set_current_function): Invoke s390_indirect_branch_settings.
4583 (s390_output_indirect_thunk_function): New function.
4584 (s390_code_end): Implement target hook.
4585 (s390_case_values_threshold): Implement target hook.
4586 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4587 macros.
4588 * config/s390/s390.h (struct s390_frame_layout)
4589 (struct machine_function): Move here from s390.c.
4590 (TARGET_INDIRECT_BRANCH_NOBP_RET)
4591 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4592 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4593 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4594 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4595 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4596 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4597 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4598 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4599 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4600 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4601 (mnemonic attribute): Add values which aren't recognized
4602 automatically.
4603 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4604 pattern for branch conversion. Fix mnemonic attribute.
4605 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4606 indirect branch via thunk if requested.
4607 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4608 ("*indirect_jump"): Disable for branch conversion using out of
4609 line thunks.
4610 ("indirect_jump_via_thunk<mode>_z10")
4611 ("indirect_jump_via_thunk<mode>")
4612 ("indirect_jump_via_inlinethunk<mode>_z10")
4613 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4614 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4615 ("casesi_jump_via_inlinethunk<mode>_z10")
4616 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4617 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4618 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4619 ("*indirect2_jump"): Disable for branch conversion.
4620 ("casesi_jump"): Turn into expander and expand patterns for branch
4621 conversion.
4622 ("return_use"): New expander.
4623 ("*return"): Emit return via thunk and rename it to ...
4624 ("*return<mode>"): ... this one.
4625 * config/s390/s390.opt: Add new options and and enum for the
4626 option values.
4628 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4630 * lra-constraints.c (match_reload): Unconditionally use
4631 gen_lowpart_SUBREG, rather than selecting between that
4632 and equivalent gen_rtx_SUBREG code.
4634 2018-02-08 Richard Biener <rguenther@suse.de>
4636 PR tree-optimization/84233
4637 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4638 changed flag instead of boguously re-using phi_inserted.
4640 2018-02-08 Martin Jambor <mjambor@suse.cz>
4642 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4643 static local variables.
4645 2018-02-08 Richard Biener <rguenther@suse.de>
4647 PR tree-optimization/84278
4648 * tree-vect-stmts.c (vectorizable_store): When looking for
4649 smaller vector types to perform grouped strided loads/stores
4650 make sure the mode is supported by the target.
4651 (vectorizable_load): Likewise.
4653 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4655 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4656 Increase LDP/STP opportunities by adding adjacent callee-saves.
4658 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4660 PR rtl-optimization/84068
4661 PR rtl-optimization/83459
4662 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4664 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4666 PR tree-optimization/84224
4667 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4668 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4669 non-zero arguments.
4671 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4673 PR target/84113
4674 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4675 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4676 count, remove one USE.
4678 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
4680 * doc/install.texi (Configuration): Document the
4681 --with-long-double-format={ibm,ieee} PowerPC configuration
4682 options.
4684 PR target/84154
4685 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4686 Convert from define_expand to be define_insn_and_split. Rework
4687 float/double/_Float128 conversions to QI/HI/SImode to work with
4688 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
4689 conversions to QI/HImode types did a store and then a load to
4690 truncate the value. For conversions to VSX registers, don't split
4691 the insn, instead emit the code directly. Use the code iterator
4692 any_fix to combine signed and unsigned conversions.
4693 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4694 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4695 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4696 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4697 (fix_<mode>di2_hw): Likewise.
4698 (fixuns_<mode>di2_hw): Likewise.
4699 (fix_<mode>si2_hw): Likewise.
4700 (fixuns_<mode>si2_hw): Likewise.
4701 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4702 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4703 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4704 fix<uns>_trunc<SFDF:mode>si2_p8.
4705 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4706 used.
4707 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4708 (fix<uns>_<mode>_mem): Likewise.
4709 (fctiw<u>z_<mode>_mem): Likewise.
4710 (fix<uns>_<mode>_mem): Likewise.
4711 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4712 the register allocator from doing a direct move to the GPRs to do
4713 a store, and instead use the ISA 3.0 store byte/half-word from
4714 vector register instruction. For IEEE 128-bit floating point,
4715 also optimize stores of 32-bit ints.
4716 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4718 2018-02-07 Alan Hayward <alan.hayward@arm.com>
4720 * genextract.c (push_pathstr_operand): New function to support
4721 [a-zA-Z].
4722 (walk_rtx): Call push_pathstr_operand.
4723 (print_path): Support [a-zA-Z].
4725 2018-02-07 Richard Biener <rguenther@suse.de>
4727 PR tree-optimization/84037
4728 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4729 (cse_and_gimplify_to_preheader): Declare.
4730 (vect_get_place_in_interleaving_chain): Likewise.
4731 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4732 ivexpr_map.
4733 (_loop_vec_info::~_loop_vec_info): Delete it.
4734 (cse_and_gimplify_to_preheader): New function.
4735 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
4736 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
4737 (vectorizable_load): Likewise. For grouped stores always base
4738 the IV on the first element.
4739 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
4740 condition before gimplifying.
4742 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4744 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
4745 *DIV_EXPR and *MOD_EXPR.
4747 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
4749 PR target/84248
4750 * config/i386/i386.c (ix86_option_override_internal): Mask out
4751 the CF_SET bit when checking -fcf-protection.
4753 2018-02-07 Tom de Vries <tom@codesourcery.com>
4755 PR libgomp/84217
4756 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
4757 enough.
4759 2018-02-07 Richard Biener <rguenther@suse.de>
4761 PR tree-optimization/84204
4762 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4763 this place.
4765 PR tree-optimization/84205
4766 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
4767 special-case isl_ast_op_zdiv_r.
4769 PR tree-optimization/84223
4770 * graphite-scop-detection.c (gather_bbs::before_dom_children):
4771 Only add conditions from within the region.
4772 (gather_bbs::after_dom_children): Adjust.
4774 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
4776 PR target/84209
4777 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
4778 * config/avr/avr.md: Only post-reload split REG-REG moves if
4779 either register is GENERAL_REG_P.
4781 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4783 PR tree-optimization/84235
4784 * tree-ssa-scopedtables.c
4785 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
4786 if the subtraction is performed in floating point type where NaNs are
4787 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4788 build 1. Formatting fix.
4790 2018-02-06 Jakub Jelinek <jakub@redhat.com>
4792 PR target/84146
4793 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4794 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4795 and skip it regardless of bb boundaries. Use CALL_P macro,
4796 don't test INSN_P (insn) together with CALL_P or JUMP_P check
4797 unnecessarily, formatting fix.
4799 2018-02-06 Michael Collison <michael.collison@arm.com>
4801 * config/arm/thumb2.md:
4802 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4803 (*thumb_mov_notscc): Ditto.
4805 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
4807 PR target/84154
4808 * config/rs6000/rs6000.md (su code attribute): Use "u" for
4809 unsigned_fix, not "s".
4811 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4813 * configure.ac (gcc_fn_eh_frame_ro): New function.
4814 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4815 correct .eh_frame permissions.
4816 * configure: Regenerate.
4818 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
4820 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4821 irrelevant options.
4823 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4825 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4826 Display warning message for -mno-speculate-indirect-jumps.
4828 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
4830 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4831 Undocumented.
4832 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4834 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
4836 PR tree-optimization/84225
4837 * tree-eh.c (find_trapping_overflow): Only call
4838 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4840 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4842 PR target/84145
4843 * config/i386/i386.c: Reimplement the check of possible options
4844 -mibt/-mshstk conbination. Change error messages.
4845 * doc/invoke.texi: Fix a typo: remove extra '='.
4847 2018-02-06 Marek Polacek <polacek@redhat.com>
4849 PR tree-optimization/84228
4850 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4852 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4854 PR target/82641
4855 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4856 emitted arch directives.
4857 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4858 __ARM_FEATURE_COPROC before changing architectures.
4860 2018-02-06 Richard Biener <rguenther@suse.de>
4862 * config/i386/i386.c (print_reg): Fix typo.
4863 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4865 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4867 * configure: Regenerate.
4869 2018-02-05 Martin Sebor <msebor@redhat.com>
4871 PR tree-optimization/83369
4872 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4873 inlining context.
4875 2018-02-05 Martin Liska <mliska@suse.cz>
4877 * doc/invoke.texi: Cherry-pick upstream r323995.
4879 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4881 * ira.c (ira_init_register_move_cost): Adjust comment.
4883 2018-02-05 Martin Liska <mliska@suse.cz>
4885 PR gcov-profile/84137
4886 * doc/gcov.texi: Fix typo in documentation.
4888 2018-02-05 Martin Liska <mliska@suse.cz>
4890 PR gcov-profile/83879
4891 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4892 using dlopen functionality.
4894 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4896 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4897 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4898 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4899 _mm_maskz_range_round_ss): New intrinsics.
4900 (__builtin_ia32_rangesd128_round)
4901 (__builtin_ia32_rangess128_round): Remove.
4902 (__builtin_ia32_rangesd128_mask_round,
4903 __builtin_ia32_rangess128_mask_round): New builtins.
4904 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4905 __builtin_ia32_rangess128_round): Remove.
4906 (__builtin_ia32_rangesd128_mask_round,
4907 __builtin_ia32_rangess128_mask_round): New builtins.
4908 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4909 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4910 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4911 "<round_saeonly_constraint>")): Changed to ...
4912 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4913 "<round_saeonly_scalar_constraint>")): ... this.
4914 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4915 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4916 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4917 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4918 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
4920 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
4922 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
4923 options.
4924 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
4925 Remove all values except native, 8540 and 8548.
4927 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
4929 * config/i386/i386.c (ix86_output_function_return): Pass
4930 INVALID_REGNUM, instead of -1, as invalid register number to
4931 indirect_thunk_name and output_indirect_thunk.
4933 2018-02-02 Julia Koval <julia.koval@intel.com>
4935 * config.gcc: Add -march=icelake.
4936 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
4937 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
4938 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
4939 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
4940 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
4941 (processor_target_table): Add icelake.
4942 (ix86_option_override_internal): Handle new PTAs.
4943 (get_builtin_code_for_version): Handle icelake.
4944 (M_INTEL_COREI7_ICELAKE): New.
4945 (fold_builtin_cpu): Handle icelake.
4946 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
4947 * doc/invoke.texi: Add -march=icelake.
4949 2018-02-02 Julia Koval <julia.koval@intel.com>
4951 * config/i386/i386.c (ix86_option_override_internal): Change flags type
4952 to wide_int_bitmask.
4953 * wide-int-bitmask.h: New.
4955 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4957 PR target/84066
4958 * config/i386/i386.md: Replace Pmode with word_mode in
4959 builtin_setjmp_setup and builtin_longjmp to support x32.
4961 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
4963 PR target/56010
4964 PR target/83743
4965 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
4966 #include "opts.h".
4967 (rs6000_supported_cpu_names): New static variable.
4968 (linux_cpu_translation_table): Likewise.
4969 (elf_platform) <cpu>: Define new static variable and use it.
4970 Translate kernel AT_PLATFORM name to canonical name if needed.
4971 Error if platform name is unknown.
4973 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
4975 PR target/84089
4976 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
4978 2018-02-01 Jeff Law <law@redhat.com>
4980 PR target/84128
4981 * config/i386/i386.c (release_scratch_register_on_entry): Add new
4982 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
4983 the scratch if RELEASE_VIA_POP is false.
4984 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
4985 If we have to save a temporary register, decrement SIZE appropriately.
4986 Pass new arguments to release_scratch_register_on_entry.
4987 (ix86_adjust_stack_and_probe): Likewise.
4988 (ix86_emit_probe_stack_range): Pass new arguments to
4989 release_scratch_register_on_entry.
4991 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
4993 PR rtl-optimization/84157
4994 * combine.c (change_zero_ext): Use REG_P predicate in
4995 front of HARD_REGISTER_P predicate.
4997 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
4999 * config/avr/avr.c (avr_option_override): Move disabling of
5000 -fdelete-null-pointer-checks to...
5001 * common/config/avr/avr-common.c (avr_option_optimization_table):
5002 ...here.
5004 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5006 PR tree-optimization/81635
5007 * tree-data-ref.c (split_constant_offset_1): For types that
5008 wrap on overflow, try to use range info to prove that wrapping
5009 cannot occur.
5011 2018-02-01 Renlin Li <renlin.li@arm.com>
5013 PR target/83370
5014 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5015 TAILCALL_ADDR_REGS.
5016 (aarch64_register_move_cost): Likewise.
5017 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5018 TAILCALL_ADDR_REGS.
5019 (REG_CLASS_NAMES): Likewise.
5020 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5021 TAILCALL_ADDR_REGS. Remove IP registers.
5022 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5024 2018-02-01 Richard Biener <rguenther@suse.de>
5026 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5027 for specifying RPO order and allow NULL for that.
5028 * domwalk.c (dom_walker::dom_walker): Likewise.
5029 (dom_walker::walk): Handle NULL RPO order.
5030 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5031 in RPO order.
5032 (rewrite_update_dom_walker): Likewise.
5033 (mark_def_dom_walker): Likewise.
5035 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5037 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5038 (aarch64_maybe_expand_sve_subreg_move): Declare.
5039 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5040 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5041 predicate.
5042 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5043 that are semantically a reverse operation.
5044 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5045 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5046 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5047 functions.
5048 (aarch64_can_change_mode_class): For big-endian, forbid changes
5049 between two SVE modes if they have different element sizes.
5051 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5053 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5054 the TImode handling for big-endian targets.
5056 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5058 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5059 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5060 not just bytes.
5061 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5062 Remove BSWAP handing for big-endian targets and use the form of
5063 LD1RQ appropariate for the mode.
5065 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5067 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5068 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5069 duplicated element.
5071 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5073 PR tearget/83845
5074 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5075 check for operands that need to go through aarch64_sve_reload_be.
5077 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5079 PR tree-optimization/81661
5080 PR tree-optimization/84117
5081 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5082 * tree-eh.c: Include gimplify.h.
5083 (find_trapping_overflow, replace_trapping_overflow,
5084 rewrite_to_non_trapping_overflow): New functions.
5085 * tree-vect-loop.c: Include tree-eh.h.
5086 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5087 * tree-data-ref.c: Include tree-eh.h.
5088 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5090 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5092 PR rtl-optimization/84123
5093 * combine.c (change_zero_ext): Check if hard register satisfies
5094 can_change_dest_mode before calling gen_lowpart_SUBREG.
5096 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5098 PR target/82444
5099 * ira.c (ira_init_register_move_cost): Remove assert.
5101 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5103 PR rtl-optimization/84071
5104 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5105 * doc/tm.texi: Regenerate.
5107 2018-01-31 Richard Biener <rguenther@suse.de>
5109 PR tree-optimization/84132
5110 * tree-data-ref.c (analyze_miv_subscript): Properly
5111 check whether evolution_function_is_affine_multivariate_p
5112 before calling gcd_of_steps_may_divide_p.
5114 2018-01-31 Julia Koval <julia.koval@intel.com>
5116 PR target/83618
5117 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5118 * config/i386/i386.md (rdpid_rex64) New.
5119 (rdpid): Make 32bit only.
5121 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5123 PR lto/84105
5124 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5125 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5127 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5129 Revert
5130 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5132 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5134 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5136 PR rtl-optimization/84071
5137 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5138 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5140 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5142 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5143 (arc_attribute_table): Add 'aux' attribute.
5144 (arc_in_small_data_p): Consider aux like variables.
5145 (arc_is_aux_reg_p): New function.
5146 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5147 (arc_get_aux_arg): New function.
5148 (prepare_move_operands): Handle aux-register access.
5149 (arc_handle_aux_attribute): New function.
5150 * doc/extend.texi (ARC Variable attributes): Add subsection.
5152 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5154 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5155 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5156 (arc_attribute_table): Add 'uncached' attribute.
5157 (arc_print_operand): Print '.di' flag for uncached memory
5158 accesses.
5159 (arc_in_small_data_p): Do not consider for small data the uncached
5160 types.
5161 (arc_is_uncached_mem_p): New function.
5162 * config/arc/predicates.md (compact_store_memory_operand): Check
5163 for uncached memory accesses.
5164 (nonvol_nonimm_operand): Likewise.
5165 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5167 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5169 PR c/84100
5170 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5171 falign-loops=): Add Optimization flag.
5173 2018-01-30 Jeff Law <law@redhat.com>
5175 PR target/84064
5176 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5177 INT_REGISTERS_SAVED. Check it prior to calling
5178 get_scratch_register_on_entry.
5179 (ix86_adjust_stack_and_probe): Similarly.
5180 (ix86_emit_probe_stack_range): Similarly.
5181 (ix86_expand_prologue): Corresponding changes.
5183 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5185 PR target/40411
5186 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5187 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5189 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5191 PR target/84112
5192 * lra-constraints.c (curr_insn_transform): Process AND in the
5193 address.
5195 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5197 PR rtl-optimization/83986
5198 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5199 dependence against last_pending_memory_flush in addition to
5200 pending_jump_insns.
5202 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5204 PR tree-optimization/81611
5205 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5206 copies.
5208 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5210 PR target/83758
5211 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5212 a reg rtx.
5214 2018-01-30 Richard Biener <rguenther@suse.de>
5215 Jakub Jelinek <jakub@redhat.com>
5217 PR tree-optimization/84111
5218 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5219 inner loops added during recursion, as they don't have up-to-date
5220 SSA form.
5222 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5224 PR ipa/81360
5225 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5226 (can_inline_edge_by_limits_p): ... here.
5227 (can_early_inline_edge_p, check_callers,
5228 update_caller_keys, update_callee_keys, recursive_inlining,
5229 add_new_edges_to_heap, speculation_useful_p,
5230 inline_small_functions,
5231 inline_small_functions, flatten_function,
5232 inline_to_all_callers_1): Update.
5234 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5236 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5237 zeros correctly.
5239 2018-01-30 Richard Biener <rguenther@suse.de>
5241 PR tree-optimization/83008
5242 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5243 invariant and constant vector uses in stmts when they need
5244 more than one stmt.
5246 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5248 PR bootstrap/84017
5249 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5250 * configure: Regenerate.
5252 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5254 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5255 pattern.
5256 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5257 Use gen_rtx_REG rather than gen_lowpart.
5259 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5261 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5262 rather than 0 when creating partial subregs.
5264 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5266 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5267 of usage.
5269 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5271 PR target/81550
5272 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5273 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5274 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5275 flags. This restores the settings used before the 2017-07-24.
5276 Turning off pre increment/decrement/modify allows IVOPTS to
5277 optimize DF/SF loops where the index is an int.
5279 2018-01-29 Richard Biener <rguenther@suse.de>
5280 Kelvin Nilsen <kelvin@gcc.gnu.org>
5282 PR bootstrap/80867
5283 * tree-vect-stmts.c (vectorizable_call): Don't call
5284 targetm.vectorize_builtin_md_vectorized_function if callee is
5285 NULL.
5287 2018-01-22 Carl Love <cel@us.ibm.com>
5289 * doc/extend.tex: Fix typo in second arg in
5290 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5292 2018-01-29 Richard Biener <rguenther@suse.de>
5294 PR tree-optimization/84086
5295 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5296 (flush_ssaname_freelist): When SSA names were released reset
5297 the SCEV hash table.
5299 2018-01-29 Richard Biener <rguenther@suse.de>
5301 PR tree-optimization/84057
5302 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5303 removed paths when removing edges.
5305 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5307 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5308 -mfunction-return=@var{choice}.
5310 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5312 PR diagnostic/84034
5313 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5314 Handle CR like TAB.
5315 (layout::print_source_line): Likewise.
5316 (test_get_line_width_without_trailing_whitespace): Add test cases.
5318 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5320 PR middle-end/84040
5321 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5322 debug insns.
5324 2018-01-26 Jim Wilson <jimw@sifive.com>
5326 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5328 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5329 specified.
5331 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5333 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5334 and CMP + SUB-immediate -> SUBS.
5336 2018-01-26 Martin Sebor <msebor@redhat.com>
5338 PR tree-optimization/83896
5339 * tree-ssa-strlen.c (get_string_len): Rename...
5340 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5341 Avoid assuming length is constant.
5342 (handle_char_store): Use HOST_WIDE_INT for string length.
5344 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5346 PR target/81763
5347 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5348 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5350 2018-01-26 Richard Biener <rguenther@suse.de>
5352 PR rtl-optimization/84003
5353 * dse.c (record_store): Only record redundant stores when
5354 the earlier store aliases at least all accesses the later one does.
5356 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5358 PR rtl-optimization/83985
5359 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5360 REG_CFA_RESTORE insns.
5361 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5362 REG_CFA_RESTORE insns here.
5364 PR c/83989
5365 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5366 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5368 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5370 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5371 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5372 (arc_init): Likewise.
5373 (arc_override_options): Likewise.
5374 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5375 value.
5376 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5377 support.
5378 * config/arc/arc.h (TARGET_DBNZ): Define.
5379 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5380 properly set the tune attribute.
5381 (dbnz): Use TARGET_DBNZ guard.
5382 * config/arc/arc.opt (mtune): Add core3 option.
5384 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5386 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5387 recognize new pic like addresses.
5388 (arc_delegitimize_address): Clean up.
5390 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5392 * config/arc/arc-arches.def: Option mrf16 valid for all
5393 architectures.
5394 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5395 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5396 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5397 * config/arc/arc-tables.opt: Regenerate.
5398 * config/arc/arc.c (arc_conditional_register_usage): Handle
5399 reduced register file case.
5400 (arc_file_start): Set must have build attributes.
5401 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5402 mrf16 option value.
5403 * config/arc/arc.opt (mrf16): Add new option.
5404 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5405 * config/arc/genmultilib.awk: Handle new mrf16 option.
5406 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5407 * config/arc/t-multilib: Regenerate.
5408 * doc/invoke.texi (ARC Options): Document mrf16 option.
5410 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5412 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5413 * config/arc/arc.c (arc_handle_secure_attribute): New function.
5414 (arc_attribute_table): Add 'secure_call' attribute.
5415 (arc_print_operand): Print secure call operand.
5416 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5417 secure.
5418 (arc_is_secure_call_p): New function. * config/arc/arc.md
5419 (call_i): Add support for sjli instruction.
5420 (call_value_i): Likewise.
5421 * config/arc/constraints.md (Csc): New constraint.
5423 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5424 John Eric Martin <John.Martin@emmicro-us.com>
5426 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5427 * config/arc/arc.c (_arc_jli_section): New struct.
5428 (arc_jli_section): New type.
5429 (rc_jli_sections): New static variable.
5430 (arc_handle_jli_attribute): New function.
5431 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5432 (arc_file_end): New function.
5433 (TARGET_ASM_FILE_END): Define.
5434 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5435 (arc_add_jli_section): New function.
5436 (jli_call_scan): Likewise.
5437 (arc_reorg): Call jli_call_scan.
5438 (arc_output_addsi): Remove 'S' from printing asm operand.
5439 (arc_is_jli_call_p): New function.
5440 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5441 operand.
5442 (movhi_insn): Likewise.
5443 (movsi_insn): Likewise.
5444 (movsi_set_cc_insn): Likewise.
5445 (loadqi_update): Likewise.
5446 (load_zeroextendqisi_update): Likewise.
5447 (load_signextendqisi_update): Likewise.
5448 (loadhi_update): Likewise.
5449 (load_zeroextendhisi_update): Likewise.
5450 (load_signextendhisi_update): Likewise.
5451 (loadsi_update): Likewise.
5452 (loadsf_update): Likewise.
5453 (movsicc_insn): Likewise.
5454 (bset_insn): Likewise.
5455 (bxor_insn): Likewise.
5456 (bclr_insn): Likewise.
5457 (bmsk_insn): Likewise.
5458 (bicsi3_insn): Likewise.
5459 (cmpsi_cc_c_insn): Likewise.
5460 (movsi_ne): Likewise.
5461 (movsi_cond_exec): Likewise.
5462 (clrsbsi2): Likewise.
5463 (norm_f): Likewise.
5464 (normw): Likewise.
5465 (swap): Likewise.
5466 (divaw): Likewise.
5467 (flag): Likewise.
5468 (sr): Likewise.
5469 (kflag): Likewise.
5470 (ffs): Likewise.
5471 (ffs_f): Likewise.
5472 (fls): Likewise.
5473 (call_i): Remove 'S' asm letter, add jli instruction.
5474 (call_value_i): Likewise.
5475 * config/arc/arc.op (mjli-always): New option.
5476 * config/arc/constraints.md (Cji): New constraint.
5477 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
5478 operand.
5479 (subsf3_fpx): Likewise.
5480 (mulsf3_fpx): Likewise.
5481 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5482 asm operand.
5483 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5484 function attrbutes.
5485 * doc/invoke.texi (ARC): Document mjli-always option.
5487 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
5489 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5490 avoid addition with 0 and use incw and decw where possible.
5492 2018-01-26 Richard Biener <rguenther@suse.de>
5494 PR tree-optimization/81082
5495 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5496 association if it requires casting to unsigned.
5497 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5498 from fold_plusminus_mult_expr to catch important cases late when
5499 range info is available.
5501 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5503 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5504 * configure.ac (hidden_linkonce): New test.
5505 * configure: Regenerate.
5506 * config.in: Regenerate.
5508 2018-01-26 Julia Koval <julia.koval@intel.com>
5510 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5511 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5512 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5513 _mm_mask_bitshuffle_epi64_mask): Fix type.
5514 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5515 USI_FTYPE_V4DI_V4DI_USI): Remove.
5516 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5517 __builtin_ia32_vpshufbitqmb256_mask,
5518 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5519 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5520 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5522 2018-01-26 Alan Modra <amodra@gmail.com>
5524 PR target/84033
5525 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5526 UNSPEC_VBPERMQ. Sort other unspecs.
5528 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5530 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5532 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5534 PR middle-end/83055
5535 * predict.c (drop_profile): Do not push/pop cfun; update also
5536 node->count.
5537 (handle_missing_profiles): Fix logic looking for zero profiles.
5539 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5541 PR middle-end/83977
5542 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5543 on functions with #pragma omp declare simd or functions with simd
5544 attribute.
5545 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5546 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5547 Remove trailing \n from warning_at calls.
5549 2018-01-25 Tom de Vries <tom@codesourcery.com>
5551 PR target/84028
5552 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5553 for neutered workers.
5555 2018-01-24 Joseph Myers <joseph@codesourcery.com>
5557 PR target/68467
5558 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5559 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5561 2018-01-24 Jeff Law <law@redhat.com>
5563 PR target/83994
5564 * i386.c (get_probe_interval): Move to earlier point.
5565 (ix86_compute_frame_layout): If -fstack-clash-protection and
5566 the frame is larger than the probe interval, then use pushes
5567 to save registers rather than reg->mem moves.
5568 (ix86_expand_prologue): Remove conditional for int_registers_saved
5569 assertion.
5571 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
5573 PR target/84014
5574 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5575 min/max for never referenced object.
5577 2018-01-24 Jakub Jelinek <jakub@redhat.com>
5579 PR middle-end/83977
5580 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5581 here.
5582 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5583 attributes from DECL_ATTRIBUTES (decl) without affecting
5584 DECL_ATTRIBUTES (current_function_decl).
5585 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5586 functions with non-NULL DECL_ABSTRACT_ORIGIN.
5588 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
5590 PR tree-optimization/83979
5591 * fold-const.c (fold_comparison): Use constant_boolean_node
5592 instead of boolean_{true,false}_node.
5594 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
5596 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5597 with zero counts.
5599 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5601 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5602 Simplify the clause that sets the length attribute.
5603 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5604 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5605 clause that sets the length attribute.
5606 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5608 2018-01-24 Tom de Vries <tom@codesourcery.com>
5610 PR target/83589
5611 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5612 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5613 Add strict parameter.
5614 (prevent_branch_around_nothing): Insert dummy insn between branch to
5615 label and label with no ptx insn inbetween.
5616 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5618 2018-01-24 Tom de Vries <tom@codesourcery.com>
5620 PR target/81352
5621 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5622 for neutered threads in warp.
5623 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5625 2018-01-24 Richard Biener <rguenther@suse.de>
5627 PR tree-optimization/83176
5628 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5629 operands.
5631 2018-01-24 Richard Biener <rguenther@suse.de>
5633 PR tree-optimization/82819
5634 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5635 code generating pluses that are no-ops in the target precision.
5637 2018-01-24 Richard Biener <rguenther@suse.de>
5639 PR middle-end/84000
5640 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5642 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5644 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5645 to merge probabilities.
5646 * predict.c (probably_never_executed): Also mark as cold functions
5647 with global 0 profile and guessed local profile.
5648 * profile-count.c (profile_probability::combine_with_count): New
5649 member function.
5650 * profile-count.h (profile_probability::operator*,
5651 profile_probability::operator*=, profile_probability::operator/,
5652 profile_probability::operator/=): Reduce precision to adjusted
5653 and set value to guessed on contradictory divisions.
5654 (profile_probability::combine_with_freq): Remove.
5655 (profile_probability::combine_wiht_count): Declare.
5656 (profile_count::force_nonzero):: Set to adjusted.
5657 (profile_count::probability_in):: Set quality to adjusted.
5658 * tree-ssa-tail-merge.c (replace_block_by): Use
5659 combine_with_count.
5661 2018-01-23 Andrew Waterman <andrew@sifive.com>
5662 Jim Wilson <jimw@sifive.com>
5664 * config/riscv/riscv.c (riscv_stack_boundary): New.
5665 (riscv_option_override): Set riscv_stack_boundary. Handle
5666 riscv_preferred_stack_boundary_arg.
5667 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5668 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5669 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5670 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5671 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5672 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5674 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5676 PR target/83905
5677 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5678 of struct ix86_frame.
5679 (ix86_expand_epilogue): Likewise. Add a local variable for
5680 the reg_save_offset field in struct ix86_frame.
5682 2018-01-23 Bin Cheng <bin.cheng@arm.com>
5684 PR tree-optimization/82604
5685 * tree-loop-distribution.c (enum partition_kind): New enum item
5686 PKIND_PARTIAL_MEMSET.
5687 (partition_builtin_p): Support above new enum item.
5688 (generate_code_for_partition): Ditto.
5689 (compute_access_range): Differentiate cases that equality can be
5690 proven at all loops, the innermost loops or no loops.
5691 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5692 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
5693 (finalize_partitions, distribute_loop): Don't fuse partition of
5694 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5695 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5696 parloop is enabled.
5698 2018-01-23 Martin Liska <mliska@suse.cz>
5700 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5701 order to ignore the predictor.
5702 (PRED_POLYMORPHIC_CALL): Likewise.
5703 (PRED_RECURSIVE_CALL): Likewise.
5705 2018-01-23 Martin Liska <mliska@suse.cz>
5707 * tree-profile.c (tree_profiling): Print function header to
5708 aware reader which function we are working on.
5709 * value-prof.c (gimple_find_values_to_profile): Do not print
5710 not interesting value histograms.
5712 2018-01-23 Martin Liska <mliska@suse.cz>
5714 * profile-count.h (enum profile_quality): Add
5715 profile_uninitialized as the first value. Do not number values
5716 as they are zero based.
5717 (profile_count::verify): Update sanity check.
5718 (profile_probability::verify): Likewise.
5720 2018-01-23 Nathan Sidwell <nathan@acm.org>
5722 * doc/invoke.texi (ffor-scope): Deprecate.
5724 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5726 PR tree-optimization/83510
5727 * domwalk.c (set_all_edges_as_executable): New function.
5728 (dom_walker::dom_walker): Convert bool param
5729 "skip_unreachable_blocks" to enum reachability. Move setup of
5730 edge flags to set_all_edges_as_executable and only do it when
5731 reachability is REACHABLE_BLOCKS.
5732 * domwalk.h (enum dom_walker::reachability): New enum.
5733 (dom_walker::dom_walker): Convert bool param
5734 "skip_unreachable_blocks" to enum reachability.
5735 (set_all_edges_as_executable): New decl.
5736 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
5737 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
5738 "reachability".
5739 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
5740 but converting true to REACHABLE_BLOCKS.
5741 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
5742 * tree-vrp.c
5743 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
5744 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
5745 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
5746 REACHABLE_BLOCKS.
5747 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
5748 if check_all_array_refs will be called.
5750 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5752 * tree.c (selftest::test_location_wrappers): Add more test
5753 coverage.
5755 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5757 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
5758 (selftest::test_bit_in_range): Likewise.
5760 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5762 PR testsuite/83888
5763 * doc/sourcebuild.texi (vect_float): Say that the selector
5764 only describes the situation when -funsafe-math-optimizations is on.
5765 (vect_float_strict): Document.
5767 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5769 PR tree-optimization/83965
5770 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5771 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
5772 instead of checking only for a reduction.
5773 (vect_recog_widen_sum_pattern): Likewise.
5775 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5777 * predict.c (probably_never_executed): Only use precise profile info.
5778 (compute_function_frequency): Skip after inlining hack since we now
5779 have quality checking.
5781 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5783 * profile-count.h (profile_probability::very_unlikely,
5784 profile_probability::unlikely, profile_probability::even): Set
5785 precision to guessed.
5787 2018-01-23 Richard Biener <rguenther@suse.de>
5789 PR tree-optimization/83963
5790 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5791 Properly terminate dominator walk when crossing the exit edge not
5792 when visiting its source block.
5794 2018-01-23 Jakub Jelinek <jakub@redhat.com>
5796 PR c++/83918
5797 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5798 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5800 2018-01-22 Jakub Jelinek <jakub@redhat.com>
5802 PR tree-optimization/83957
5803 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
5804 semicolon after for body surrounded by braces.
5806 PR tree-optimization/83081
5807 * profile-count.h (profile_probability::split): New method.
5808 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5809 Use profile_probability::split.
5810 (do_compare_rtx_and_jump): Fix adjustment of probabilities
5811 when splitting a single conditional jump into 2.
5813 2018-01-22 David Malcolm <dmalcolm@redhat.com>
5815 PR tree-optimization/69452
5816 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5817 decl.
5819 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5821 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5822 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5823 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5825 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5827 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5828 * config/rl78/rl78.md (movdi): New define_expand.
5829 * config/rl78/rl78.c (rl78_split_movdi): New function.
5831 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
5833 PR target/83862
5834 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5835 no longer used.
5836 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5837 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5838 128-bit to produce an UNSPEC move to get the double word with the
5839 signbit and then a shift directly to do signbit.
5840 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5841 implementation with a new version that just does either a direct
5842 move or a regular move. Move memory interface to separate insns.
5843 Move insns so they are next to the expander.
5844 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5845 with signbit move. Split big and little endian case.
5846 (signbit<mode>2_dm_mem_le): Likewise.
5847 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5848 (signbit<mode>2_dm2): Likewise.
5850 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5852 * config/rl78/rl78.md (anddi3): New define_expand.
5854 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5856 * config/rl78/rl78.md (umindi3): New define_expand.
5858 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5860 * config/rl78/rl78.md (smindi3): New define_expand.
5862 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5864 * config/rl78/rl78.md (smaxdi3): New define_expand.
5866 2018-01-22 Carl Love <cel@us.ibm.com>
5868 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5869 LVX_V1TI): Add macro expansion.
5870 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5871 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5872 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5873 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5874 Change check to determine if the instruction is a byte reversing
5875 entry. Fix typo in comment.
5876 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5877 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5878 Add def_builtin calls for new builtins.
5879 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5880 Add define_insn expansion.
5882 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5884 * config/rl78/rl78.md (umaxdi3): New define_expand.
5886 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5888 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5889 for non-QImode registers.
5891 2018-01-22 Richard Biener <rguenther@suse.de>
5893 PR tree-optimization/83963
5894 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5895 including the loop exit block.
5896 (scop_detection::merge_sese): Likewise.
5897 (scop_detection::add_scop): Do it here instead.
5899 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5901 * doc/sourcebuild.texi (arm_softfloat): Document.
5903 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5905 PR gcc/77734
5906 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5907 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5908 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5910 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5911 David Edelsohn <dje.gcc@gmail.com>
5913 PR target/83946
5914 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5915 Change "crset eq" to "crset 2".
5916 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5917 (*call_indirect_aix<mode>_nospec): Likewise.
5918 (*call_value_indirect_aix<mode>_nospec): Likewise.
5919 (*call_indirect_elfv2<mode>_nospec): Likewise.
5920 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
5921 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
5922 change assembly output from . to $.
5923 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5924 (indirect_jump<mode>_nospec): Change assembly output from . to $.
5925 (*tablejump<mode>_internal1_nospec): Likewise.
5927 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
5929 PR target/80870
5930 * config/sh/sh_optimize_sett_clrt.cc:
5931 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
5933 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
5935 PR tree-optimization/83940
5936 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
5937 offset_dt to vect_constant_def rather than vect_unknown_def_type.
5938 (vect_check_load_store_mask): Add a mask_dt_out parameter and
5939 use it to pass back the definition type.
5940 (vect_check_store_rhs): Likewise rhs_dt_out.
5941 (vect_build_gather_load_calls): Add a mask_dt argument and use
5942 it instead of a call to vect_is_simple_use.
5943 (vectorizable_store): Update calls to vect_check_load_store_mask
5944 and vect_check_store_rhs. Use the dt returned by the latter instead
5945 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
5946 instead of calls to vect_is_simple_use. Pass the scalar rather
5947 than the vector operand to vect_is_simple_use when handling
5948 second and subsequent copies of an rhs value.
5949 (vectorizable_load): Update calls to vect_check_load_store_mask
5950 and vect_build_gather_load_calls. Use the cached mask_dt and
5951 gs_info.offset_dt instead of calls to vect_is_simple_use.
5953 2018-01-20 Jakub Jelinek <jakub@redhat.com>
5955 PR middle-end/83945
5956 * tree-emutls.c: Include gimplify.h.
5957 (lower_emutls_2): New function.
5958 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
5959 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
5960 it before further processing.
5962 PR target/83930
5963 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
5964 UINTVAL (trueop1) instead of INTVAL (op1).
5966 2018-01-19 Jakub Jelinek <jakub@redhat.com>
5968 PR debug/81570
5969 PR debug/83728
5970 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
5971 INCOMING_FRAME_SP_OFFSET if not defined.
5972 (scan_trace): Add ENTRY argument. If true and
5973 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
5974 emit a note to adjust the CFA offset.
5975 (create_cfi_notes): Adjust scan_trace callers.
5976 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
5977 INCOMING_FRAME_SP_OFFSET in the CIE.
5978 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
5979 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
5980 Likewise.
5981 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
5982 * doc/tm.texi: Regenerated.
5984 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5986 PR rtl-optimization/83147
5987 * lra-constraints.c (remove_inheritance_pseudos): Use
5988 lra_substitute_pseudo_within_insn.
5990 2018-01-19 Tom de Vries <tom@codesourcery.com>
5991 Cesar Philippidis <cesar@codesourcery.com>
5993 PR target/83920
5994 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
5996 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
5998 PR target/83790
5999 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6000 spaces for function labels.
6002 2018-01-19 Martin Liska <mliska@suse.cz>
6004 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6005 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6006 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6007 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6008 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6009 (PRED_CONST_RETURN): Change from 69 to 65.
6010 (PRED_NULL_RETURN): Change from 91 to 71.
6011 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6012 (PRED_LOOP_GUARD): Change from 66 to 73.
6014 2018-01-19 Martin Liska <mliska@suse.cz>
6016 * predict.c (predict_insn_def): Add new assert.
6017 (struct branch_predictor): Change type to signed integer.
6018 (test_prediction_value_range): Amend test to cover
6019 PROB_UNINITIALIZED.
6020 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6021 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6022 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6023 (PRED_LOOP_IV_COMPARE): Likewise.
6024 * predict.h (PROB_UNINITIALIZED): Define new constant.
6026 2018-01-19 Martin Liska <mliska@suse.cz>
6028 * predict.c (dump_prediction): Add new format for
6029 analyze_brprob.py script which is enabled with -details
6030 suboption.
6031 * profile-count.h (precise_p): New function.
6033 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6035 PR tree-optimization/83922
6036 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6037 there are no statements that need masking.
6038 (vect_active_double_reduction_p): New function.
6039 (vect_analyze_loop_operations): Use it when handling phis that
6040 are not in the loop header.
6042 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6044 PR tree-optimization/83914
6045 * tree-vect-loop.c (vectorizable_induction): Don't convert
6046 init_expr or apply the peeling adjustment for inductions
6047 that are nested within the vectorized loop.
6049 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6051 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6052 instead of NEG.
6054 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6056 PR sanitizer/81715
6057 PR testsuite/83882
6058 * function.h (gimplify_parameters): Add gimple_seq * argument.
6059 * function.c: Include gimple.h and options.h.
6060 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6061 for the added local temporaries if needed.
6062 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6063 if there are any parameter cleanups, wrap whole body into a
6064 try/finally with the cleanups.
6066 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6068 PR target/82964
6069 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6070 Use GET_MODE_CLASS for scalar floating point.
6072 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6074 PR ipa/82256
6075 patch by PaX Team
6076 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6077 Fix call of call_cgraph_insertion_hooks.
6079 2018-01-18 Martin Sebor <msebor@redhat.com>
6081 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6083 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6085 PR ipa/83619
6086 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6087 frequencies.
6089 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6091 PR other/70268
6092 * common.opt: (-ffile-prefix-map): New option.
6093 * opts.c (common_handle_option): Defer it.
6094 * opts-global.c (handle_common_deferred_options): Handle it.
6095 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6096 * file-prefix-map.h: New file.
6097 (remap_debug_filename, add_debug_prefix_map): ...here.
6098 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6099 * final.c (debug_prefix_map, add_debug_prefix_map
6100 remap_debug_filename): Move to...
6101 * file-prefix-map.c: New file.
6102 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6103 generalize, get rid of alloca(), use strrchr() instead of strchr().
6104 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6105 Implement in terms of add_prefix_map().
6106 (remap_macro_filename, remap_debug_filename): Implement in term of
6107 remap_filename().
6108 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6109 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6110 * dbxout.c: Include file-prefix-map.h.
6111 * varasm.c: Likewise.
6112 * vmsdbgout.c: Likewise.
6113 * xcoffout.c: Likewise.
6114 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6115 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6116 * doc/invoke.texi (-ffile-prefix-map): Document.
6117 (-fdebug-prefix-map): Update description.
6119 2018-01-18 Martin Liska <mliska@suse.cz>
6121 * config/i386/i386.c (indirect_thunk_name): Document that also
6122 lfence is emitted.
6123 (output_indirect_thunk): Document why both instructions
6124 (pause and lfence) are generated.
6126 2018-01-18 Richard Biener <rguenther@suse.de>
6128 PR tree-optimization/83887
6129 * graphite-scop-detection.c
6130 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6131 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6132 (scop_detection::merge_sese): Re-implement with a flood-fill
6133 algorithm that properly finds a SESE region if it exists.
6135 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6137 PR c/61240
6138 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6139 pointer_diff optimizations use view_convert instead of convert.
6141 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6143 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6144 Generate different code for -mno-speculate-indirect-jumps.
6145 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6146 (*call_indirect_aix<mode>): Disable for
6147 -mno-speculate-indirect-jumps.
6148 (*call_indirect_aix<mode>_nospec): New define_insn.
6149 (*call_value_indirect_aix<mode>): Disable for
6150 -mno-speculate-indirect-jumps.
6151 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6152 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6153 -mno-speculate-indirect-jumps.
6154 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6156 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6158 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6159 long double type, set the flags for noting the default long double
6160 type, even if we don't pass or return a long double type.
6162 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6164 PR ipa/83051
6165 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6166 failure.
6168 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6170 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6171 support for merge[hl].
6172 (fold_mergehl_helper): New helper function.
6173 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6174 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6175 (altivec_vmrglw_direct): Add xxmrglw insn.
6177 2018-01-17 Andrew Waterman <andrew@sifive.com>
6179 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6180 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6182 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6184 PR lto/83121
6185 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6186 call the lto_location_cache before reading the
6187 DECL_SOURCE_LOCATION of the types.
6189 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6190 Richard Sandiford <richard.sandiford@linaro.org>
6192 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6193 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6194 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6195 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6196 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6197 Add declaration.
6198 * config/aarch64/constraints.md (aarch64_movti_operand):
6199 Limit immediates.
6200 * config/aarch64/predicates.md (Uti): Add new constraint.
6202 2018-01-17 Carl Love <cel@us.ibm.com>
6204 * config/rs6000/vsx.md (define_expand xl_len_r,
6205 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6206 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6207 lxvll.
6208 (define_expand, define_insn): Move the shift left from the
6209 define_insn to the define_expand for lxvl and stxvl instructions.
6210 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6211 and XL_LEN_R definitions to PURE.
6213 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6215 * config/i386/i386.c (indirect_thunk_name): Declare regno
6216 as unsigned int. Compare regno with INVALID_REGNUM.
6217 (output_indirect_thunk): Ditto.
6218 (output_indirect_thunk_function): Ditto.
6219 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6220 in the call to output_indirect_thunk_function.
6222 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6224 PR middle-end/83884
6225 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6226 rather than the size of inner_type to determine the stack slot size
6227 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6229 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6231 PR target/83546
6232 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6233 to PTA_SILVERMONT.
6235 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6237 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6238 endian Linux systems to optionally enable multilibs for selecting
6239 the long double type if the user configured an explicit type.
6240 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6241 have no long double multilibs if not defined.
6242 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6243 warn if the user used -mabi={ieee,ibm}longdouble and we built
6244 multilibs for long double.
6245 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6246 appropriate multilib option.
6247 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6248 multilib options.
6249 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6250 for building long double multilibs.
6251 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6253 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6255 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6256 copies.
6258 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6259 64 bits.
6260 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6261 128 bits.
6263 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6264 variables.
6266 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6267 return value.
6269 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6271 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6272 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6274 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6276 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6277 different rtl trees depending on TARGET_64BIT.
6278 (rs6000_gen_lvx): Likewise.
6280 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6282 * config/visium/visium.md (nop): Tweak comment.
6283 (hazard_nop): Likewise.
6285 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6287 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6288 -mspeculate-indirect-jumps.
6289 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6290 for -mno-speculate-indirect-jumps.
6291 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6292 (*call_value_indirect_elfv2<mode>): Disable for
6293 -mno-speculate-indirect-jumps.
6294 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6295 (indirect_jump): Emit different RTL for
6296 -mno-speculate-indirect-jumps.
6297 (*indirect_jump<mode>): Disable for
6298 -mno-speculate-indirect-jumps.
6299 (*indirect_jump<mode>_nospec): New define_insn.
6300 (tablejump): Emit different RTL for
6301 -mno-speculate-indirect-jumps.
6302 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6303 (tablejumpsi_nospec): New define_expand.
6304 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6305 (tablejumpdi_nospec): New define_expand.
6306 (*tablejump<mode>_internal1): Disable for
6307 -mno-speculate-indirect-jumps.
6308 (*tablejump<mode>_internal1_nospec): New define_insn.
6309 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6310 option.
6312 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6314 * caller-save.c (insert_save): Drop unnecessary parameter. All
6315 callers updated.
6317 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6318 Richard Biener <rguenth@suse.de>
6320 PR libgomp/83590
6321 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6322 return early, inline manually is_gimple_sizepos. Make sure if we
6323 call gimplify_expr we don't end up with a gimple constant.
6324 * tree.c (variably_modified_type_p): Don't return true for
6325 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6326 * gimplify.h (is_gimple_sizepos): Remove.
6328 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6330 PR tree-optimization/83857
6331 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6332 vectorizable_live_operation for pure SLP statements.
6333 (vectorizable_live_operation): Handle PHIs.
6335 2018-01-16 Richard Biener <rguenther@suse.de>
6337 PR tree-optimization/83867
6338 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6339 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6341 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6343 PR c/83844
6344 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6345 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6346 If off is not INTEGER_CST, issue a may not be aligned warning
6347 rather than isn't aligned. Use isn%'t rather than isn't.
6348 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6349 into MULT_EXPR.
6350 <case MULT_EXPR>: Improve the case when bottom and one of the
6351 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6352 operand, in that case check if the other operand is multiple of
6353 bottom divided by the INTEGER_CST operand.
6355 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6357 PR target/83858
6358 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6359 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6360 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6361 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6362 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6363 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6364 (pa_function_arg_size): New function.
6366 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6368 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6369 in a separate statement.
6371 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6373 PR tree-optimization/83847
6374 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6375 group gathers and scatters.
6377 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6379 PR rtl-optimization/86620
6380 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6382 PR rtl-optimization/83213
6383 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6384 to last if both are JUMP_INSNs.
6386 PR tree-optimization/83843
6387 * gimple-ssa-store-merging.c
6388 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6389 store_immediate_info for bswap/nop orig_stores.
6391 2018-01-15 Andrew Waterman <andrew@sifive.com>
6393 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6394 !TARGET_MUL.
6395 <UDIV>: Increase cost if !TARGET_DIV.
6397 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6399 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6400 (define_attr "cr_logical_3op"): New.
6401 (cceq_ior_compare): Adjust.
6402 (cceq_ior_compare_complement): Adjust.
6403 (*cceq_rev_compare): Adjust.
6404 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6405 (is_cracked_insn): Adjust.
6406 (insn_must_be_first_in_group): Adjust.
6407 * config/rs6000/40x.md: Adjust.
6408 * config/rs6000/440.md: Adjust.
6409 * config/rs6000/476.md: Adjust.
6410 * config/rs6000/601.md: Adjust.
6411 * config/rs6000/603.md: Adjust.
6412 * config/rs6000/6xx.md: Adjust.
6413 * config/rs6000/7450.md: Adjust.
6414 * config/rs6000/7xx.md: Adjust.
6415 * config/rs6000/8540.md: Adjust.
6416 * config/rs6000/cell.md: Adjust.
6417 * config/rs6000/e300c2c3.md: Adjust.
6418 * config/rs6000/e500mc.md: Adjust.
6419 * config/rs6000/e500mc64.md: Adjust.
6420 * config/rs6000/e5500.md: Adjust.
6421 * config/rs6000/e6500.md: Adjust.
6422 * config/rs6000/mpc.md: Adjust.
6423 * config/rs6000/power4.md: Adjust.
6424 * config/rs6000/power5.md: Adjust.
6425 * config/rs6000/power6.md: Adjust.
6426 * config/rs6000/power7.md: Adjust.
6427 * config/rs6000/power8.md: Adjust.
6428 * config/rs6000/power9.md: Adjust.
6429 * config/rs6000/rs64.md: Adjust.
6430 * config/rs6000/titan.md: Adjust.
6432 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6434 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6435 ix86_indirect_branch_register logic.
6437 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6439 * config/i386/constraints.md (Bs): Update
6440 ix86_indirect_branch_register check. Don't check
6441 ix86_indirect_branch_register with GOT_memory_operand.
6442 (Bw): Likewise.
6443 * config/i386/predicates.md (GOT_memory_operand): Don't check
6444 ix86_indirect_branch_register here.
6445 (GOT32_symbol_operand): Likewise.
6447 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6449 * config/i386/predicates.md (constant_call_address_operand):
6450 Rewrite ix86_indirect_branch_register logic.
6451 (sibcall_insn_operand): Likewise.
6453 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6455 * config/i386/constraints.md (Bs): Replace
6456 ix86_indirect_branch_thunk_register with
6457 ix86_indirect_branch_register.
6458 (Bw): Likewise.
6459 * config/i386/i386.md (indirect_jump): Likewise.
6460 (tablejump): Likewise.
6461 (*sibcall_memory): Likewise.
6462 (*sibcall_value_memory): Likewise.
6463 Peepholes of indirect call and jump via memory: Likewise.
6464 * config/i386/i386.opt: Likewise.
6465 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6466 (GOT_memory_operand): Likewise.
6467 (call_insn_operand): Likewise.
6468 (sibcall_insn_operand): Likewise.
6469 (GOT32_symbol_operand): Likewise.
6471 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6473 PR middle-end/83837
6474 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
6475 type rather than type addr's type points to.
6476 (expand_omp_atomic_mutex): Likewise.
6477 (expand_omp_atomic): Likewise.
6479 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6481 PR target/83839
6482 * config/i386/i386.c (output_indirect_thunk_function): Use
6483 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6484 for __x86_return_thunk.
6486 2018-01-15 Richard Biener <rguenther@suse.de>
6488 PR middle-end/83850
6489 * expmed.c (extract_bit_field_1): Fix typo.
6491 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6493 PR target/83687
6494 * config/arm/iterators.md (VF): New mode iterator.
6495 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6496 Remove integer-related logic from pattern.
6497 (neon_vabd<mode>_3): Likewise.
6499 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6501 PR middle-end/82694
6502 * common.opt (fstrict-overflow): No longer an alias.
6503 (fwrapv-pointer): New option.
6504 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6505 also for pointer types based on flag_wrapv_pointer.
6506 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6507 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6508 opts->x_flag_wrapv got set.
6509 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6510 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6511 POINTER_TYPE_OVERFLOW_UNDEFINED.
6512 * match.pd: Likewise in address comparison pattern.
6513 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6515 2018-01-15 Richard Biener <rguenther@suse.de>
6517 PR lto/83804
6518 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6519 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6520 Reset type names to their identifier if their TYPE_DECL doesn't
6521 have linkage (and thus is used for ODR and devirt).
6522 (save_debug_info_for_decl): Remove.
6523 (save_debug_info_for_type): Likewise.
6524 (add_tree_to_fld_list): Adjust.
6525 * tree-pretty-print.c (dump_generic_node): Make dumping of
6526 type names more robust.
6528 2018-01-15 Richard Biener <rguenther@suse.de>
6530 * BASE-VER: Bump to 8.0.1.
6532 2018-01-14 Martin Sebor <msebor@redhat.com>
6534 PR other/83508
6535 * builtins.c (check_access): Avoid warning when the no-warning bit
6536 is set.
6538 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6540 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6541 * ira-color (allocno_hard_regs_compare): Likewise.
6543 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6545 PR target/83013
6546 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6547 Use .pushsection/.popsection.
6549 2018-01-14 Martin Sebor <msebor@redhat.com>
6551 PR c++/81327
6552 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6554 2018-01-14 Jakub Jelinek <jakub@redhat.com>
6556 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6557 entry from extra_headers.
6558 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6559 extra_headers, make the list bitwise identical to the i?86-*-* one.
6561 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6563 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6564 -mcmodel=large with -mindirect-branch=thunk,
6565 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6566 -mfunction-return=thunk-extern.
6567 * doc/invoke.texi: Document -mcmodel=large is incompatible with
6568 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6569 -mfunction-return=thunk and -mfunction-return=thunk-extern.
6571 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6573 * config/i386/i386.c (print_reg): Print the name of the full
6574 integer register without '%'.
6575 (ix86_print_operand): Handle 'V'.
6576 * doc/extend.texi: Document 'V' modifier.
6578 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6580 * config/i386/constraints.md (Bs): Disallow memory operand for
6581 -mindirect-branch-register.
6582 (Bw): Likewise.
6583 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6584 (GOT_memory_operand): Likewise.
6585 (call_insn_operand): Likewise.
6586 (sibcall_insn_operand): Likewise.
6587 (GOT32_symbol_operand): Likewise.
6588 * config/i386/i386.md (indirect_jump): Call convert_memory_address
6589 for -mindirect-branch-register.
6590 (tablejump): Likewise.
6591 (*sibcall_memory): Likewise.
6592 (*sibcall_value_memory): Likewise.
6593 Disallow peepholes of indirect call and jump via memory for
6594 -mindirect-branch-register.
6595 (*call_pop): Replace m with Bw.
6596 (*call_value_pop): Likewise.
6597 (*sibcall_pop_memory): Replace m with Bs.
6598 * config/i386/i386.opt (mindirect-branch-register): New option.
6599 * doc/invoke.texi: Document -mindirect-branch-register option.
6601 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6603 * config/i386/i386-protos.h (ix86_output_function_return): New.
6604 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6605 set function_return_type.
6606 (indirect_thunk_name): Add ret_p to indicate thunk for function
6607 return.
6608 (output_indirect_thunk_function): Pass false to
6609 indirect_thunk_name.
6610 (ix86_output_indirect_branch_via_reg): Likewise.
6611 (ix86_output_indirect_branch_via_push): Likewise.
6612 (output_indirect_thunk_function): Create alias for function
6613 return thunk if regno < 0.
6614 (ix86_output_function_return): New function.
6615 (ix86_handle_fndecl_attribute): Handle function_return.
6616 (ix86_attribute_table): Add function_return.
6617 * config/i386/i386.h (machine_function): Add
6618 function_return_type.
6619 * config/i386/i386.md (simple_return_internal): Use
6620 ix86_output_function_return.
6621 (simple_return_internal_long): Likewise.
6622 * config/i386/i386.opt (mfunction-return=): New option.
6623 (indirect_branch): Mention -mfunction-return=.
6624 * doc/extend.texi: Document function_return function attribute.
6625 * doc/invoke.texi: Document -mfunction-return= option.
6627 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6629 * config/i386/i386-opts.h (indirect_branch): New.
6630 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6631 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6632 with local indirect jump when converting indirect call and jump.
6633 (ix86_set_indirect_branch_type): New.
6634 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6635 (indirectlabelno): New.
6636 (indirect_thunk_needed): Likewise.
6637 (indirect_thunk_bnd_needed): Likewise.
6638 (indirect_thunks_used): Likewise.
6639 (indirect_thunks_bnd_used): Likewise.
6640 (INDIRECT_LABEL): Likewise.
6641 (indirect_thunk_name): Likewise.
6642 (output_indirect_thunk): Likewise.
6643 (output_indirect_thunk_function): Likewise.
6644 (ix86_output_indirect_branch_via_reg): Likewise.
6645 (ix86_output_indirect_branch_via_push): Likewise.
6646 (ix86_output_indirect_branch): Likewise.
6647 (ix86_output_indirect_jmp): Likewise.
6648 (ix86_code_end): Call output_indirect_thunk_function if needed.
6649 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6650 needed.
6651 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6652 (ix86_attribute_table): Add indirect_branch.
6653 * config/i386/i386.h (machine_function): Add indirect_branch_type
6654 and has_local_indirect_jump.
6655 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6656 to true.
6657 (tablejump): Likewise.
6658 (*indirect_jump): Use ix86_output_indirect_jmp.
6659 (*tablejump_1): Likewise.
6660 (simple_return_indirect_internal): Likewise.
6661 * config/i386/i386.opt (mindirect-branch=): New option.
6662 (indirect_branch): New.
6663 (keep): Likewise.
6664 (thunk): Likewise.
6665 (thunk-inline): Likewise.
6666 (thunk-extern): Likewise.
6667 * doc/extend.texi: Document indirect_branch function attribute.
6668 * doc/invoke.texi: Document -mindirect-branch= option.
6670 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6672 PR ipa/83051
6673 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6675 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6677 * ipa-inline.c (want_inline_small_function_p): Return false if
6678 inlining has already failed with CIF_FINAL_ERROR.
6679 (update_caller_keys): Call want_inline_small_function_p before
6680 can_inline_edge_p.
6681 (update_callee_keys): Likewise.
6683 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6685 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6686 New function.
6687 (rs6000_quadword_masked_address_p): Likewise.
6688 (quad_aligned_load_p): Likewise.
6689 (quad_aligned_store_p): Likewise.
6690 (const_load_sequence_p): Add comment to describe the outer-most loop.
6691 (mimic_memory_attributes_and_flags): New function.
6692 (rs6000_gen_stvx): Likewise.
6693 (replace_swapped_aligned_store): Likewise.
6694 (rs6000_gen_lvx): Likewise.
6695 (replace_swapped_aligned_load): Likewise.
6696 (replace_swapped_load_constant): Capitalize argument name in
6697 comment describing this function.
6698 (rs6000_analyze_swaps): Add a third pass to search for vector loads
6699 and stores that access quad-word aligned addresses and replace
6700 with stvx or lvx instructions when appropriate.
6701 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6702 New function prototype.
6703 (rs6000_quadword_masked_address_p): Likewise.
6704 (rs6000_gen_lvx): Likewise.
6705 (rs6000_gen_stvx): Likewise.
6706 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6707 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6708 when memory address is aligned.
6709 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6710 this split to select lvx instruction when memory address is aligned.
6711 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6712 instruction when memory address is aligned.
6713 (*vsx_le_perm_load_v16qi): Likewise.
6714 (four unnamed splitters): Modify to select the stvx instruction
6715 when memory is aligned.
6717 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
6719 * predict.c (determine_unlikely_bbs): Handle correctly BBs
6720 which appears in the queue multiple times.
6722 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6723 Alan Hayward <alan.hayward@arm.com>
6724 David Sherwood <david.sherwood@arm.com>
6726 * tree-vectorizer.h (vec_lower_bound): New structure.
6727 (_loop_vec_info): Add check_nonzero and lower_bounds.
6728 (LOOP_VINFO_CHECK_NONZERO): New macro.
6729 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6730 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6731 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6732 fields. Make seg_len the distance travelled, not including the
6733 access size.
6734 (dr_direction_indicator): Declare.
6735 (dr_zero_step_indicator): Likewise.
6736 (dr_known_forward_stride_p): Likewise.
6737 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
6738 tree-ssanames.h.
6739 (runtime_alias_check_p): Allow runtime alias checks with
6740 variable strides.
6741 (operator ==): Compare access_size and align.
6742 (prune_runtime_alias_test_list): Rework for new distinction between
6743 the access_size and seg_len.
6744 (create_intersect_range_checks_index): Likewise. Cope with polynomial
6745 segment lengths.
6746 (get_segment_min_max): New function.
6747 (create_intersect_range_checks): Use it.
6748 (dr_step_indicator): New function.
6749 (dr_direction_indicator): Likewise.
6750 (dr_zero_step_indicator): Likewise.
6751 (dr_known_forward_stride_p): Likewise.
6752 * tree-loop-distribution.c (data_ref_segment_size): Return
6753 DR_STEP * (niters - 1).
6754 (compute_alias_check_pairs): Update call to the dr_with_seg_len
6755 constructor.
6756 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
6757 (vect_preserves_scalar_order_p): New function, split out from...
6758 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
6759 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
6760 (vect_vfa_access_size): New function.
6761 (vect_vfa_align): Likewise.
6762 (vect_compile_time_alias): Take access_size_a and access_b arguments.
6763 (dump_lower_bound): New function.
6764 (vect_check_lower_bound): Likewise.
6765 (vect_small_gap_p): Likewise.
6766 (vectorizable_with_step_bound_p): Likewise.
6767 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
6768 depencies if the vectorization factor is 1. Convert the checks
6769 for nonzero steps into checks on the bounds of DR_STEP. Try using
6770 a bunds check for variable steps if the minimum required step is
6771 relatively small. Update calls to the dr_with_seg_len
6772 constructor and to vect_compile_time_alias.
6773 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
6774 function.
6775 (vect_loop_versioning): Call it.
6776 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
6777 when retrying.
6778 (vect_estimate_min_profitable_iters): Account for any bounds checks.
6780 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6781 Alan Hayward <alan.hayward@arm.com>
6782 David Sherwood <david.sherwood@arm.com>
6784 * doc/sourcebuild.texi (vect_scatter_store): Document.
6785 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
6786 optabs.
6787 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6788 Document.
6789 * genopinit.c (main): Add supports_vec_scatter_store and
6790 supports_vec_scatter_store_cached to target_optabs.
6791 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6792 IFN_MASK_SCATTER_STORE.
6793 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6794 functions.
6795 * internal-fn.h (internal_store_fn_p): Declare.
6796 (internal_fn_stored_value_index): Likewise.
6797 * internal-fn.c (scatter_store_direct): New macro.
6798 (expand_scatter_store_optab_fn): New function.
6799 (direct_scatter_store_optab_supported_p): New macro.
6800 (internal_store_fn_p): New function.
6801 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6802 IFN_MASK_SCATTER_STORE.
6803 (internal_fn_mask_index): Likewise.
6804 (internal_fn_stored_value_index): New function.
6805 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6806 for scatter stores.
6807 * optabs-query.h (supports_vec_scatter_store_p): Declare.
6808 * optabs-query.c (supports_vec_scatter_store_p): New function.
6809 * tree-vectorizer.h (vect_get_store_rhs): Declare.
6810 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6811 true for scatter stores.
6812 (vect_gather_scatter_fn_p): Handle scatter stores too.
6813 (vect_check_gather_scatter): Consider using scatter stores if
6814 supports_vec_scatter_store_p.
6815 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6816 scatter stores too.
6817 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6818 internal_fn_stored_value_index.
6819 (check_load_store_masking): Handle scatter stores too.
6820 (vect_get_store_rhs): Make public.
6821 (vectorizable_call): Use internal_store_fn_p.
6822 (vectorizable_store): Handle scatter store internal functions.
6823 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6824 when deciding whether the end of the group has been reached.
6825 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6826 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6827 (mask_scatter_store<mode>): New insns.
6829 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6830 Alan Hayward <alan.hayward@arm.com>
6831 David Sherwood <david.sherwood@arm.com>
6833 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6834 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6835 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6836 function.
6837 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6838 Use vect_truncate_gather_scatter_offset if we can't treat the
6839 operation as a normal gather load or scatter store.
6840 (get_group_load_store_type): Take the gather_scatter_info
6841 as argument. Try using a gather load or scatter store for
6842 single-element groups.
6843 (get_load_store_type): Update calls to get_group_load_store_type
6844 and vect_use_strided_gather_scatters_p.
6846 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6847 Alan Hayward <alan.hayward@arm.com>
6848 David Sherwood <david.sherwood@arm.com>
6850 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6851 optional tree argument.
6852 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6853 null target hooks.
6854 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6855 but continue to use the current value as a fallback.
6856 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6857 to compare the updates.
6858 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6859 (get_load_store_type): Use it when handling a strided access.
6860 (vect_get_strided_load_store_ops): New function.
6861 (vect_get_data_ptr_increment): Likewise.
6862 (vectorizable_load): Handle strided gather loads. Always pass
6863 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6865 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6866 Alan Hayward <alan.hayward@arm.com>
6867 David Sherwood <david.sherwood@arm.com>
6869 * doc/md.texi (gather_load@var{m}): Document.
6870 (mask_gather_load@var{m}): Likewise.
6871 * genopinit.c (main): Add supports_vec_gather_load and
6872 supports_vec_gather_load_cached to target_optabs.
6873 * optabs-tree.c (init_tree_optimization_optabs): Use
6874 ggc_cleared_alloc to allocate target_optabs.
6875 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6876 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6877 functions.
6878 * internal-fn.h (internal_load_fn_p): Declare.
6879 (internal_gather_scatter_fn_p): Likewise.
6880 (internal_fn_mask_index): Likewise.
6881 (internal_gather_scatter_fn_supported_p): Likewise.
6882 * internal-fn.c (gather_load_direct): New macro.
6883 (expand_gather_load_optab_fn): New function.
6884 (direct_gather_load_optab_supported_p): New macro.
6885 (direct_internal_fn_optab): New function.
6886 (internal_load_fn_p): Likewise.
6887 (internal_gather_scatter_fn_p): Likewise.
6888 (internal_fn_mask_index): Likewise.
6889 (internal_gather_scatter_fn_supported_p): Likewise.
6890 * optabs-query.c (supports_at_least_one_mode_p): New function.
6891 (supports_vec_gather_load_p): Likewise.
6892 * optabs-query.h (supports_vec_gather_load_p): Declare.
6893 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6894 and memory_type field.
6895 (NUM_PATTERNS): Bump to 15.
6896 * tree-vect-data-refs.c: Include internal-fn.h.
6897 (vect_gather_scatter_fn_p): New function.
6898 (vect_describe_gather_scatter_call): Likewise.
6899 (vect_check_gather_scatter): Try using internal functions for
6900 gather loads. Recognize existing calls to a gather load function.
6901 (vect_analyze_data_refs): Consider using gather loads if
6902 supports_vec_gather_load_p.
6903 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6904 (vect_get_gather_scatter_offset_type): Likewise.
6905 (vect_convert_mask_for_vectype): Likewise.
6906 (vect_add_conversion_to_patterm): Likewise.
6907 (vect_try_gather_scatter_pattern): Likewise.
6908 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6909 (vect_vect_recog_func_ptrs): Add it.
6910 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6911 internal_fn_mask_index and internal_gather_scatter_fn_p.
6912 (check_load_store_masking): Take the gather_scatter_info as an
6913 argument and handle gather loads.
6914 (vect_get_gather_scatter_ops): New function.
6915 (vectorizable_call): Check internal_load_fn_p.
6916 (vectorizable_load): Likewise. Handle gather load internal
6917 functions.
6918 (vectorizable_store): Update call to check_load_store_masking.
6919 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
6920 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
6921 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
6922 (aarch64_gather_scale_operand_d): New predicates.
6923 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
6924 (mask_gather_load<mode>): New insns.
6926 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6927 Alan Hayward <alan.hayward@arm.com>
6928 David Sherwood <david.sherwood@arm.com>
6930 * optabs.def (fold_left_plus_optab): New optab.
6931 * doc/md.texi (fold_left_plus_@var{m}): Document.
6932 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
6933 * internal-fn.c (fold_left_direct): Define.
6934 (expand_fold_left_optab_fn): Likewise.
6935 (direct_fold_left_optab_supported_p): Likewise.
6936 * fold-const-call.c (fold_const_fold_left): New function.
6937 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
6938 * tree-parloops.c (valid_reduction_p): New function.
6939 (gather_scalar_reductions): Use it.
6940 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
6941 (vect_finish_replace_stmt): Declare.
6942 * tree-vect-loop.c (fold_left_reduction_fn): New function.
6943 (needs_fold_left_reduction_p): New function, split out from...
6944 (vect_is_simple_reduction): ...here. Accept reductions that
6945 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
6946 (vect_force_simple_reduction): Also store the reduction type in
6947 the assignment's STMT_VINFO_REDUC_TYPE.
6948 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
6949 (merge_with_identity): New function.
6950 (vect_expand_fold_left): Likewise.
6951 (vectorize_fold_left_reduction): Likewise.
6952 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
6953 scalar phi in place for it. Check for target support and reject
6954 cases that would reassociate the operation. Defer the transform
6955 phase to vectorize_fold_left_reduction.
6956 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
6957 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
6958 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
6960 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6962 * tree-if-conv.c (predicate_mem_writes): Remove redundant
6963 call to ifc_temp_var.
6965 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6966 Alan Hayward <alan.hayward@arm.com>
6967 David Sherwood <david.sherwood@arm.com>
6969 * target.def (legitimize_address_displacement): Take the original
6970 offset as a poly_int.
6971 * targhooks.h (default_legitimize_address_displacement): Update
6972 accordingly.
6973 * targhooks.c (default_legitimize_address_displacement): Likewise.
6974 * doc/tm.texi: Regenerate.
6975 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
6976 as an argument, moving assert of ad->disp == ad->disp_term to...
6977 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
6978 Try calling targetm.legitimize_address_displacement before expanding
6979 the address rather than afterwards, and adjust for the new interface.
6980 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
6981 Match the new hook interface. Handle SVE addresses.
6982 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
6983 new hook interface.
6985 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6987 * Makefile.in (OBJS): Add early-remat.o.
6988 * target.def (select_early_remat_modes): New hook.
6989 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
6990 * doc/tm.texi: Regenerate.
6991 * targhooks.h (default_select_early_remat_modes): Declare.
6992 * targhooks.c (default_select_early_remat_modes): New function.
6993 * timevar.def (TV_EARLY_REMAT): New timevar.
6994 * passes.def (pass_early_remat): New pass.
6995 * tree-pass.h (make_pass_early_remat): Declare.
6996 * early-remat.c: New file.
6997 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
6998 function.
6999 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7001 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7002 Alan Hayward <alan.hayward@arm.com>
7003 David Sherwood <david.sherwood@arm.com>
7005 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7006 vfm1 with a bound_epilog parameter.
7007 (vect_do_peeling): Update calls accordingly, and move the prologue
7008 call earlier in the function. Treat the base bound_epilog as 0 for
7009 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7010 this base when peeling for gaps.
7011 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7012 with fully-masked loops.
7013 (vect_estimate_min_profitable_iters): Handle the single peeled
7014 iteration in that case.
7016 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7017 Alan Hayward <alan.hayward@arm.com>
7018 David Sherwood <david.sherwood@arm.com>
7020 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7021 single-element interleaving even if the size is not a power of 2.
7022 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7023 accesses for single-element interleaving if the group size is
7024 not a power of 2.
7026 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7027 Alan Hayward <alan.hayward@arm.com>
7028 David Sherwood <david.sherwood@arm.com>
7030 * doc/md.texi (fold_extract_last_@var{m}): Document.
7031 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7032 * optabs.def (fold_extract_last_optab): New optab.
7033 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7034 * internal-fn.c (fold_extract_direct): New macro.
7035 (expand_fold_extract_optab_fn): Likewise.
7036 (direct_fold_extract_optab_supported_p): Likewise.
7037 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7038 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7039 EXTRACT_LAST_REDUCTION.
7040 (get_initial_def_for_reduction): Do not create an initial vector
7041 for EXTRACT_LAST_REDUCTION reductions.
7042 (vectorizable_reduction): Leave the scalar phi in place for
7043 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7044 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7045 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7046 transform phase to vectorizable_condition.
7047 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7048 split out from...
7049 (vect_finish_stmt_generation): ...here.
7050 (vect_finish_replace_stmt): New function.
7051 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7052 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7053 pattern.
7054 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7056 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7057 Alan Hayward <alan.hayward@arm.com>
7058 David Sherwood <david.sherwood@arm.com>
7060 * doc/md.texi (extract_last_@var{m}): Document.
7061 * optabs.def (extract_last_optab): New optab.
7062 * internal-fn.def (EXTRACT_LAST): New internal function.
7063 * internal-fn.c (cond_unary_direct): New macro.
7064 (expand_cond_unary_optab_fn): Likewise.
7065 (direct_cond_unary_optab_supported_p): Likewise.
7066 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7067 loops using EXTRACT_LAST.
7068 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7069 (extract_last_<mode>): ...this optab.
7070 (vec_extract<mode><Vel>): Update accordingly.
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 * target.def (empty_mask_is_expensive): New hook.
7077 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7078 * doc/tm.texi: Regenerate.
7079 * targhooks.h (default_empty_mask_is_expensive): Declare.
7080 * targhooks.c (default_empty_mask_is_expensive): New function.
7081 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7082 if the target says that empty masks are expensive.
7083 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7084 New function.
7085 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7087 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7088 Alan Hayward <alan.hayward@arm.com>
7089 David Sherwood <david.sherwood@arm.com>
7091 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7092 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7093 (vect_use_loop_mask_for_alignment_p): New function.
7094 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7095 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7096 niters_skip argument. Make sure that the first niters_skip elements
7097 of the first iteration are inactive.
7098 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7099 Update call to vect_set_loop_masks_directly.
7100 (get_misalign_in_elems): New function, split out from...
7101 (vect_gen_prolog_loop_niters): ...here.
7102 (vect_update_init_of_dr): Take a code argument that specifies whether
7103 the adjustment should be added or subtracted.
7104 (vect_update_init_of_drs): Likewise.
7105 (vect_prepare_for_masked_peels): New function.
7106 (vect_do_peeling): Skip prologue peeling if we're using a mask
7107 instead. Update call to vect_update_inits_of_drs.
7108 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7109 mask_skip_niters.
7110 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7111 alignment. Do not include the number of peeled iterations in
7112 the minimum threshold in that case.
7113 (vectorizable_induction): Adjust the start value down by
7114 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7115 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7116 Take the number of skipped iterations into account when calculating
7117 the loop bounds.
7118 * tree-vect-stmts.c (vect_gen_while_not): New function.
7120 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7121 Alan Hayward <alan.hayward@arm.com>
7122 David Sherwood <david.sherwood@arm.com>
7124 * doc/sourcebuild.texi (vect_fully_masked): Document.
7125 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7126 default value to 0.
7127 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7128 split out from...
7129 (vect_analyze_loop_2): ...here. Don't check the vectorization
7130 factor against the number of loop iterations if the loop is
7131 fully-masked.
7133 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7134 Alan Hayward <alan.hayward@arm.com>
7135 David Sherwood <david.sherwood@arm.com>
7137 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7138 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7139 (dump_groups): Update accordingly.
7140 (iv_use::mem_type): New member variable.
7141 (address_p): New function.
7142 (record_use): Add a mem_type argument and initialize the new
7143 mem_type field.
7144 (record_group_use): Add a mem_type argument. Use address_p.
7145 Remove obsolete null checks of base_object. Update call to record_use.
7146 (find_interesting_uses_op): Update call to record_group_use.
7147 (find_interesting_uses_cond): Likewise.
7148 (find_interesting_uses_address): Likewise.
7149 (get_mem_type_for_internal_fn): New function.
7150 (find_address_like_use): Likewise.
7151 (find_interesting_uses_stmt): Try find_address_like_use before
7152 calling find_interesting_uses_op.
7153 (addr_offset_valid_p): Use the iv mem_type field as the type
7154 of the addressed memory.
7155 (add_autoinc_candidates): Likewise.
7156 (get_address_cost): Likewise.
7157 (split_small_address_groups_p): Use address_p.
7158 (split_address_groups): Likewise.
7159 (add_iv_candidate_for_use): Likewise.
7160 (autoinc_possible_for_pair): Likewise.
7161 (rewrite_groups): Likewise.
7162 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7163 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7164 (get_alias_ptr_type_for_ptr_address): New function.
7165 (rewrite_use_address): Rewrite address uses in calls that were
7166 identified by find_address_like_use.
7168 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7169 Alan Hayward <alan.hayward@arm.com>
7170 David Sherwood <david.sherwood@arm.com>
7172 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7173 TARGET_MEM_REFs.
7174 * gimple-expr.h (is_gimple_addressable: Likewise.
7175 * gimple-expr.c (is_gimple_address): Likewise.
7176 * internal-fn.c (expand_call_mem_ref): New function.
7177 (expand_mask_load_optab_fn): Use it.
7178 (expand_mask_store_optab_fn): Likewise.
7180 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7181 Alan Hayward <alan.hayward@arm.com>
7182 David Sherwood <david.sherwood@arm.com>
7184 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7185 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7186 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7187 (cond_umax@var{mode}): Document.
7188 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7189 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7190 (cond_umin_optab, cond_umax_optab): New optabs.
7191 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7192 (COND_IOR, COND_XOR): New internal functions.
7193 * internal-fn.h (get_conditional_internal_fn): Declare.
7194 * internal-fn.c (cond_binary_direct): New macro.
7195 (expand_cond_binary_optab_fn): Likewise.
7196 (direct_cond_binary_optab_supported_p): Likewise.
7197 (get_conditional_internal_fn): New function.
7198 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7199 Cope with reduction statements that are vectorized as calls rather
7200 than assignments.
7201 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7202 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7203 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7204 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7205 (UNSPEC_COND_EOR): New unspecs.
7206 (optab): Add mappings for them.
7207 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7208 (sve_int_op, sve_fp_op): New int attributes.
7210 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7211 Alan Hayward <alan.hayward@arm.com>
7212 David Sherwood <david.sherwood@arm.com>
7214 * optabs.def (while_ult_optab): New optab.
7215 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7216 * internal-fn.def (WHILE_ULT): New internal function.
7217 * internal-fn.h (direct_internal_fn_supported_p): New override
7218 that takes two types as argument.
7219 * internal-fn.c (while_direct): New macro.
7220 (expand_while_optab_fn): New function.
7221 (convert_optab_supported_p): Likewise.
7222 (direct_while_optab_supported_p): New macro.
7223 * wide-int.h (wi::udiv_ceil): New function.
7224 * tree-vectorizer.h (rgroup_masks): New structure.
7225 (vec_loop_masks): New typedef.
7226 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7227 and fully_masked_p.
7228 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7229 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7230 (vect_max_vf): New function.
7231 (slpeel_make_loop_iterate_ntimes): Delete.
7232 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7233 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7234 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7235 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7236 internal-fn.h, stor-layout.h and optabs-query.h.
7237 (vect_set_loop_mask): New function.
7238 (add_preheader_seq): Likewise.
7239 (add_header_seq): Likewise.
7240 (interleave_supported_p): Likewise.
7241 (vect_maybe_permute_loop_masks): Likewise.
7242 (vect_set_loop_masks_directly): Likewise.
7243 (vect_set_loop_condition_masked): Likewise.
7244 (vect_set_loop_condition_unmasked): New function, split out from
7245 slpeel_make_loop_iterate_ntimes.
7246 (slpeel_make_loop_iterate_ntimes): Rename to..
7247 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7248 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7249 (vect_do_peeling): Update call accordingly.
7250 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7251 loops.
7252 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7253 mask_compare_type, can_fully_mask_p and fully_masked_p.
7254 (release_vec_loop_masks): New function.
7255 (_loop_vec_info): Use it to free the loop masks.
7256 (can_produce_all_loop_masks_p): New function.
7257 (vect_get_max_nscalars_per_iter): Likewise.
7258 (vect_verify_full_masking): Likewise.
7259 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7260 retries, and free the mask rgroups before retrying. Check loop-wide
7261 reasons for disallowing fully-masked loops. Make the final decision
7262 about whether use a fully-masked loop or not.
7263 (vect_estimate_min_profitable_iters): Do not assume that peeling
7264 for the number of iterations will be needed for fully-masked loops.
7265 (vectorizable_reduction): Disable fully-masked loops.
7266 (vectorizable_live_operation): Likewise.
7267 (vect_halve_mask_nunits): New function.
7268 (vect_double_mask_nunits): Likewise.
7269 (vect_record_loop_mask): Likewise.
7270 (vect_get_loop_mask): Likewise.
7271 (vect_transform_loop): Handle the case in which the final loop
7272 iteration might handle a partial vector. Call vect_set_loop_condition
7273 instead of slpeel_make_loop_iterate_ntimes.
7274 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7275 (check_load_store_masking): New function.
7276 (prepare_load_store_mask): Likewise.
7277 (vectorizable_store): Handle fully-masked loops.
7278 (vectorizable_load): Likewise.
7279 (supportable_widening_operation): Use vect_halve_mask_nunits for
7280 booleans.
7281 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7282 (vect_gen_while): New function.
7283 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7284 (aarch64_uqdec<mode>): New insn.
7286 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7287 Alan Hayward <alan.hayward@arm.com>
7288 David Sherwood <david.sherwood@arm.com>
7290 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7291 (reduc_xor_scal_optab): New optabs.
7292 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7293 (reduc_xor_scal_@var{m}): Document.
7294 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7295 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7296 internal functions.
7297 * fold-const-call.c (fold_const_call): Handle them.
7298 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7299 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7300 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7301 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7302 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7303 (UNSPEC_XORV): New unspecs.
7304 (optab): Add entries for them.
7305 (BITWISEV): New int iterator.
7306 (bit_reduc_op): New int attributes.
7308 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7309 Alan Hayward <alan.hayward@arm.com>
7310 David Sherwood <david.sherwood@arm.com>
7312 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7313 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7314 * optabs.def (vec_shl_insert_optab): New optab.
7315 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7316 (duplicate_and_interleave): Likewise.
7317 * tree-vect-loop.c: Include internal-fn.h.
7318 (neutral_op_for_slp_reduction): New function, split out from
7319 get_initial_defs_for_reduction.
7320 (get_initial_def_for_reduction): Handle option 2 for variable-length
7321 vectors by loading the neutral value into a vector and then shifting
7322 the initial value into element 0.
7323 (get_initial_defs_for_reduction): Replace the code argument with
7324 the neutral value calculated by neutral_op_for_slp_reduction.
7325 Use gimple_build_vector for constant-length vectors.
7326 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7327 but the first group_size elements have a neutral value.
7328 Use duplicate_and_interleave otherwise.
7329 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7330 Update call to get_initial_defs_for_reduction. Handle SLP
7331 reductions for variable-length vectors by creating one vector
7332 result for each scalar result, with the elements associated
7333 with other scalar results stubbed out with the neutral value.
7334 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7335 Require IFN_VEC_SHL_INSERT for double reductions on
7336 variable-length vectors, or SLP reductions that have
7337 a neutral value. Require can_duplicate_and_interleave_p
7338 support for variable-length unchained SLP reductions if there
7339 is no neutral value, such as for MIN/MAX reductions. Also require
7340 the number of vector elements to be a multiple of the number of
7341 SLP statements when doing variable-length unchained SLP reductions.
7342 Update call to vect_create_epilog_for_reduction.
7343 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7344 and remove initial values.
7345 (duplicate_and_interleave): Make public.
7346 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7347 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7349 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7350 Alan Hayward <alan.hayward@arm.com>
7351 David Sherwood <david.sherwood@arm.com>
7353 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7354 (can_duplicate_and_interleave_p): New function.
7355 (vect_get_and_check_slp_defs): Take the vector of statements
7356 rather than just the current one. Remove excess parentheses.
7357 Restriction rejectinon of vect_constant_def and vect_external_def
7358 for variable-length vectors to boolean types, or types for which
7359 can_duplicate_and_interleave_p is false.
7360 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7361 (duplicate_and_interleave): New function.
7362 (vect_get_constant_vectors): Use gimple_build_vector for
7363 constant-length vectors and suitable variable-length constant
7364 vectors. Use duplicate_and_interleave for other variable-length
7365 vectors. Don't defer the update when inserting new statements.
7367 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7368 Alan Hayward <alan.hayward@arm.com>
7369 David Sherwood <david.sherwood@arm.com>
7371 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7372 min_profitable_iters doesn't go negative.
7374 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7375 Alan Hayward <alan.hayward@arm.com>
7376 David Sherwood <david.sherwood@arm.com>
7378 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7379 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7380 * optabs.def (vec_mask_load_lanes_optab): New optab.
7381 (vec_mask_store_lanes_optab): Likewise.
7382 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7383 (MASK_STORE_LANES): Likewise.
7384 * internal-fn.c (mask_load_lanes_direct): New macro.
7385 (mask_store_lanes_direct): Likewise.
7386 (expand_mask_load_optab_fn): Handle masked operations.
7387 (expand_mask_load_lanes_optab_fn): New macro.
7388 (expand_mask_store_optab_fn): Handle masked operations.
7389 (expand_mask_store_lanes_optab_fn): New macro.
7390 (direct_mask_load_lanes_optab_supported_p): Likewise.
7391 (direct_mask_store_lanes_optab_supported_p): Likewise.
7392 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7393 parameter.
7394 (vect_load_lanes_supported): Likewise.
7395 * tree-vect-data-refs.c (strip_conversion): New function.
7396 (can_group_stmts_p): Likewise.
7397 (vect_analyze_data_ref_accesses): Use it instead of checking
7398 for a pair of assignments.
7399 (vect_store_lanes_supported): Take a masked_p parameter.
7400 (vect_load_lanes_supported): Likewise.
7401 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7402 vect_store_lanes_supported and vect_load_lanes_supported.
7403 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7404 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7405 parameter. Don't allow gaps for masked accesses.
7406 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7407 and vect_load_lanes_supported.
7408 (get_load_store_type): Take a masked_p parameter and update
7409 call to get_group_load_store_type.
7410 (vectorizable_store): Update call to get_load_store_type.
7411 Handle IFN_MASK_STORE_LANES.
7412 (vectorizable_load): Update call to get_load_store_type.
7413 Handle IFN_MASK_LOAD_LANES.
7415 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7416 Alan Hayward <alan.hayward@arm.com>
7417 David Sherwood <david.sherwood@arm.com>
7419 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7420 modes for SVE.
7421 * config/aarch64/aarch64-protos.h
7422 (aarch64_sve_struct_memory_operand_p): Declare.
7423 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7424 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7425 (VPRED, vpred): Handle SVE structure modes.
7426 * config/aarch64/constraints.md (Utx): New constraint.
7427 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7428 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7429 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7430 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7431 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7432 structure modes. Split into pieces after RA.
7433 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7434 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7435 New patterns.
7436 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7437 SVE structure modes.
7438 (aarch64_classify_address): Likewise.
7439 (sizetochar): Move earlier in file.
7440 (aarch64_print_operand): Handle SVE register lists.
7441 (aarch64_array_mode): New function.
7442 (aarch64_sve_struct_memory_operand_p): Likewise.
7443 (TARGET_ARRAY_MODE): Redefine.
7445 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7446 Alan Hayward <alan.hayward@arm.com>
7447 David Sherwood <david.sherwood@arm.com>
7449 * target.def (array_mode): New target hook.
7450 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
7451 * doc/tm.texi: Regenerate.
7452 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
7453 * hooks.c (hook_optmode_mode_uhwi_none): New function.
7454 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
7455 targetm.array_mode.
7456 * stor-layout.c (mode_for_array): Likewise. Support polynomial
7457 type sizes.
7459 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7460 Alan Hayward <alan.hayward@arm.com>
7461 David Sherwood <david.sherwood@arm.com>
7463 * fold-const.c (fold_binary_loc): Check the argument types
7464 rather than the result type when testing for a vector operation.
7466 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7468 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
7469 * doc/tm.texi: Regenerate.
7471 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7472 Alan Hayward <alan.hayward@arm.com>
7473 David Sherwood <david.sherwood@arm.com>
7475 * doc/invoke.texi (-msve-vector-bits=): Document new option.
7476 (sve): Document new AArch64 extension.
7477 * doc/md.texi (w): Extend the description of the AArch64
7478 constraint to include SVE vectors.
7479 (Upl, Upa): Document new AArch64 predicate constraints.
7480 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7481 enum.
7482 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7483 (msve-vector-bits=): New option.
7484 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7485 SVE when these are disabled.
7486 (sve): New extension.
7487 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7488 modes. Adjust their number of units based on aarch64_sve_vg.
7489 (MAX_BITSIZE_MODE_ANY_MODE): Define.
7490 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7491 aarch64_addr_query_type.
7492 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7493 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7494 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7495 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7496 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7497 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7498 (aarch64_simd_imm_zero_p): Delete.
7499 (aarch64_check_zero_based_sve_index_immediate): Declare.
7500 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7501 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7502 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7503 (aarch64_sve_float_mul_immediate_p): Likewise.
7504 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7505 rather than an rtx.
7506 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7507 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7508 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7509 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7510 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7511 (aarch64_regmode_natural_size): Likewise.
7512 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7513 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7514 left one place.
7515 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7516 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7517 for VG and the SVE predicate registers.
7518 (V_ALIASES): Add a "z"-prefixed alias.
7519 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7520 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7521 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7522 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7523 (REG_CLASS_NAMES): Add entries for them.
7524 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7525 and the predicate registers.
7526 (aarch64_sve_vg): Declare.
7527 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7528 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7529 (REGMODE_NATURAL_SIZE): Define.
7530 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7531 SVE macros.
7532 * config/aarch64/aarch64.c: Include cfgrtl.h.
7533 (simd_immediate_info): Add a constructor for series vectors,
7534 and an associated step field.
7535 (aarch64_sve_vg): New variable.
7536 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7537 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7538 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7539 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7540 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7541 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7542 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7543 (aarch64_get_mask_mode): New functions.
7544 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7545 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7546 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7547 predicate modes and predicate registers. Explicitly restrict
7548 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7549 to store a vector mode if it is recognized by
7550 aarch64_classify_vector_mode.
7551 (aarch64_regmode_natural_size): New function.
7552 (aarch64_hard_regno_caller_save_mode): Return the original mode
7553 for predicates.
7554 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7555 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7556 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7557 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7558 functions.
7559 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
7560 does not overlap dest if the function is frame-related. Handle
7561 SVE constants.
7562 (aarch64_split_add_offset): New function.
7563 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7564 them aarch64_add_offset.
7565 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7566 and update call to aarch64_sub_sp.
7567 (aarch64_add_cfa_expression): New function.
7568 (aarch64_expand_prologue): Pass extra temporary registers to the
7569 functions above. Handle the case in which we need to emit new
7570 DW_CFA_expressions for registers that were originally saved
7571 relative to the stack pointer, but now have to be expressed
7572 relative to the frame pointer.
7573 (aarch64_output_mi_thunk): Pass extra temporary registers to the
7574 functions above.
7575 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
7576 IP0 and IP1 values for SVE frames.
7577 (aarch64_expand_vec_series): New function.
7578 (aarch64_expand_sve_widened_duplicate): Likewise.
7579 (aarch64_expand_sve_const_vector): Likewise.
7580 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7581 Handle SVE constants. Use emit_move_insn to move a force_const_mem
7582 into the register, rather than emitting a SET directly.
7583 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7584 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7585 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7586 (offset_9bit_signed_scaled_p): New functions.
7587 (aarch64_replicate_bitmask_imm): New function.
7588 (aarch64_bitmask_imm): Use it.
7589 (aarch64_cannot_force_const_mem): Reject expressions involving
7590 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
7591 (aarch64_classify_index): Handle SVE indices, by requiring
7592 a plain register index with a scale that matches the element size.
7593 (aarch64_classify_address): Handle SVE addresses. Assert that
7594 the mode of the address is VOIDmode or an integer mode.
7595 Update call to aarch64_classify_symbol.
7596 (aarch64_classify_symbolic_expression): Update call to
7597 aarch64_classify_symbol.
7598 (aarch64_const_vec_all_in_range_p): New function.
7599 (aarch64_print_vector_float_operand): Likewise.
7600 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
7601 "vN" for FP registers with SVE modes. Handle (const ...) vectors
7602 and the FP immediates 1.0 and 0.5.
7603 (aarch64_print_address_internal): Handle SVE addresses.
7604 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7605 (aarch64_regno_regclass): Handle predicate registers.
7606 (aarch64_secondary_reload): Handle big-endian reloads of SVE
7607 data modes.
7608 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7609 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7610 (aarch64_convert_sve_vector_bits): New function.
7611 (aarch64_override_options): Use it to handle -msve-vector-bits=.
7612 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7613 rather than an rtx.
7614 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7615 Handle SVE vector and predicate modes. Accept VL-based constants
7616 that need only one temporary register, and VL offsets that require
7617 no temporary registers.
7618 (aarch64_conditional_register_usage): Mark the predicate registers
7619 as fixed if SVE isn't available.
7620 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7621 Return true for SVE vector and predicate modes.
7622 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7623 rather than an unsigned int. Handle SVE modes.
7624 (aarch64_preferred_simd_mode): Update call accordingly. Handle
7625 SVE modes.
7626 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7627 if SVE is enabled.
7628 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7629 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7630 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7631 (aarch64_sve_float_mul_immediate_p): New functions.
7632 (aarch64_sve_valid_immediate): New function.
7633 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7634 Explicitly reject structure modes. Check for INDEX constants.
7635 Handle PTRUE and PFALSE constants.
7636 (aarch64_check_zero_based_sve_index_immediate): New function.
7637 (aarch64_simd_imm_zero_p): Delete.
7638 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7639 vector modes. Accept constants in the range of CNT[BHWD].
7640 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7641 ask for an Advanced SIMD mode.
7642 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7643 (aarch64_simd_vector_alignment): Handle SVE predicates.
7644 (aarch64_vectorize_preferred_vector_alignment): New function.
7645 (aarch64_simd_vector_alignment_reachable): Use it instead of
7646 the vector size.
7647 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7648 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7649 functions.
7650 (MAX_VECT_LEN): Delete.
7651 (expand_vec_perm_d): Add a vec_flags field.
7652 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7653 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7654 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7655 for SVE modes.
7656 (aarch64_evpc_rev): Rename to...
7657 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7658 (aarch64_evpc_rev_global): New function.
7659 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7660 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7661 MAX_VECT_LEN.
7662 (aarch64_evpc_sve_tbl): New function.
7663 (aarch64_expand_vec_perm_const_1): Update after rename of
7664 aarch64_evpc_rev. Handle SVE permutes too, trying
7665 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7666 than aarch64_evpc_tbl.
7667 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7668 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7669 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7670 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7671 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7672 (aarch64_expand_sve_vcond): New functions.
7673 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7674 of aarch64_vector_mode_p.
7675 (aarch64_dwarf_poly_indeterminate_value): New function.
7676 (aarch64_compute_pressure_classes): Likewise.
7677 (aarch64_can_change_mode_class): Likewise.
7678 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7679 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7680 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7681 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7682 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7683 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7684 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7685 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7686 constraints.
7687 (Dn, Dl, Dr): Accept const as well as const_vector.
7688 (Dz): Likewise. Compare against CONST0_RTX.
7689 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7690 of "vector" where appropriate.
7691 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7692 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7693 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7694 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7695 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7696 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7697 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7698 (v_int_equiv): Extend to SVE modes.
7699 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7700 mode attributes.
7701 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7702 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7703 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7704 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7705 (SVE_COND_FP_CMP): New int iterators.
7706 (perm_hilo): Handle the new unpack unspecs.
7707 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7708 attributes.
7709 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7710 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7711 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7712 (aarch64_equality_operator, aarch64_constant_vector_operand)
7713 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7714 (aarch64_sve_nonimmediate_operand): Likewise.
7715 (aarch64_sve_general_operand): Likewise.
7716 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7717 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7718 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7719 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7720 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7721 (aarch64_sve_float_arith_immediate): Likewise.
7722 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7723 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7724 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7725 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7726 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7727 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7728 (aarch64_sve_float_arith_operand): Likewise.
7729 (aarch64_sve_float_arith_with_sub_operand): Likewise.
7730 (aarch64_sve_float_mul_operand): Likewise.
7731 (aarch64_sve_vec_perm_operand): Likewise.
7732 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7733 (aarch64_mov_operand): Accept const_poly_int and const_vector.
7734 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7735 as well as const_vector.
7736 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
7737 in file. Use CONST0_RTX and CONSTM1_RTX.
7738 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
7739 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
7740 Use aarch64_simd_imm_zero.
7741 * config/aarch64/aarch64-sve.md: New file.
7742 * config/aarch64/aarch64.md: Include it.
7743 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
7744 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
7745 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
7746 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
7747 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
7748 (sve): New attribute.
7749 (enabled): Disable instructions with the sve attribute unless
7750 TARGET_SVE.
7751 (movqi, movhi): Pass CONST_POLY_INT operaneds through
7752 aarch64_expand_mov_immediate.
7753 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
7754 CNT[BHSD] immediates.
7755 (movti): Split CONST_POLY_INT moves into two halves.
7756 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
7757 Split additions that need a temporary here if the destination
7758 is the stack pointer.
7759 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
7760 (*add<mode>3_poly_1): New instruction.
7761 (set_clobber_cc): New expander.
7763 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7765 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
7766 parameter and use it instead of GET_MODE_SIZE (innermode). Use
7767 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
7768 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
7769 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
7770 Change innermode from fixed_mode_size to machine_mode.
7771 (simplify_subreg): Update call accordingly. Handle a constant-sized
7772 subreg of a variable-length CONST_VECTOR.
7774 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7775 Alan Hayward <alan.hayward@arm.com>
7776 David Sherwood <david.sherwood@arm.com>
7778 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
7779 (add_offset_to_base): New function, split out from...
7780 (create_mem_ref): ...here. When handling a scale other than 1,
7781 check first whether the address is valid without the offset.
7782 Add it into the base if so, leaving the index and scale as-is.
7784 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7786 PR c++/83778
7787 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7788 fold_for_warn before checking if arg2 is INTEGER_CST.
7790 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
7792 * config/rs6000/predicates.md (load_multiple_operation): Delete.
7793 (store_multiple_operation): Delete.
7794 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7795 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7796 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7797 guarded by TARGET_STRING.
7798 (rs6000_output_load_multiple): Delete.
7799 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7800 OPTION_MASK_STRING / TARGET_STRING handling.
7801 (print_operand) <'N', 'O'>: Add comment that these are unused now.
7802 (const rs6000_opt_masks) <"string">: Change mask to 0.
7803 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7804 (MASK_STRING): Delete.
7805 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7806 parts. Simplify.
7807 (load_multiple): Delete.
7808 (*ldmsi8): Delete.
7809 (*ldmsi7): Delete.
7810 (*ldmsi6): Delete.
7811 (*ldmsi5): Delete.
7812 (*ldmsi4): Delete.
7813 (*ldmsi3): Delete.
7814 (store_multiple): Delete.
7815 (*stmsi8): Delete.
7816 (*stmsi7): Delete.
7817 (*stmsi6): Delete.
7818 (*stmsi5): Delete.
7819 (*stmsi4): Delete.
7820 (*stmsi3): Delete.
7821 (movmemsi_8reg): Delete.
7822 (corresponding unnamed define_insn): Delete.
7823 (movmemsi_6reg): Delete.
7824 (corresponding unnamed define_insn): Delete.
7825 (movmemsi_4reg): Delete.
7826 (corresponding unnamed define_insn): Delete.
7827 (movmemsi_2reg): Delete.
7828 (corresponding unnamed define_insn): Delete.
7829 (movmemsi_1reg): Delete.
7830 (corresponding unnamed define_insn): Delete.
7831 * config/rs6000/rs6000.opt (mno-string): New.
7832 (mstring): Replace by deprecation warning stub.
7833 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7835 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7837 * regrename.c (regrename_do_replace): If replacing the same
7838 reg multiple times, try to reuse last created gen_raw_REG.
7840 PR debug/81155
7841 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7842 main to workaround a bug in GDB.
7844 2018-01-12 Tom de Vries <tom@codesourcery.com>
7846 PR target/83737
7847 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7849 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7851 PR rtl-optimization/80481
7852 * ira-color.c (get_cap_member): New function.
7853 (allocnos_conflict_by_live_ranges_p): Use it.
7854 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7855 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7857 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7859 PR target/83628
7860 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7861 (*saddl_se_1): Ditto.
7862 (*ssubsi_1): Ditto.
7863 (*ssubl_se_1): Ditto.
7865 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7867 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7868 rather than wi::to_widest for DR_INITs.
7869 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7870 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7871 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7872 INTEGER_CSTs.
7873 (vect_analyze_group_access_1): Note that here.
7875 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7877 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7878 polynomial type sizes.
7880 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7882 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7883 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7884 (gimple_add_tmp_var): Likewise.
7886 2018-01-12 Martin Liska <mliska@suse.cz>
7888 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7889 (gimple_alloc_sizes): Likewise.
7890 (dump_gimple_statistics): Use PRIu64 in printf format.
7891 * gimple.h: Change uint64_t to int.
7893 2018-01-12 Martin Liska <mliska@suse.cz>
7895 * tree-core.h: Use uint64_t instead of int.
7896 * tree.c (tree_node_counts): Likewise.
7897 (tree_node_sizes): Likewise.
7898 (dump_tree_statistics): Use PRIu64 in printf format.
7900 2018-01-12 Martin Liska <mliska@suse.cz>
7902 * Makefile.in: As qsort_chk is implemented in vec.c, add
7903 vec.o to linkage of gencfn-macros.
7904 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7905 passing the info to record_node_allocation_statistics.
7906 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7907 and pass the info.
7908 * ggc-common.c (struct ggc_usage): Add operator== and use
7909 it in operator< and compare function.
7910 * mem-stats.h (struct mem_usage): Likewise.
7911 * vec.c (struct vec_usage): Remove operator< and compare
7912 function. Can be simply inherited.
7914 2018-01-12 Martin Jambor <mjambor@suse.cz>
7916 PR target/81616
7917 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7918 * tree-ssa-math-opts.c: Include domwalk.h.
7919 (convert_mult_to_fma_1): New function.
7920 (fma_transformation_info): New type.
7921 (fma_deferring_state): Likewise.
7922 (cancel_fma_deferring): New function.
7923 (result_of_phi): Likewise.
7924 (last_fma_candidate_feeds_initial_phi): Likewise.
7925 (convert_mult_to_fma): Added deferring logic, split actual
7926 transformation to convert_mult_to_fma_1.
7927 (math_opts_dom_walker): New type.
7928 (math_opts_dom_walker::after_dom_children): New method, body moved
7929 here from pass_optimize_widening_mul::execute, added deferring logic
7930 bits.
7931 (pass_optimize_widening_mul::execute): Moved most of code to
7932 math_opts_dom_walker::after_dom_children.
7933 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
7934 * config/i386/i386.c (ix86_option_override_internal): Added
7935 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
7937 2018-01-12 Richard Biener <rguenther@suse.de>
7939 PR debug/83157
7940 * dwarf2out.c (gen_variable_die): Do not reset old_die for
7941 inline instance vars.
7943 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
7945 PR target/81819
7946 * config/rx/rx.c (rx_is_restricted_memory_address):
7947 Handle SUBREG case.
7949 2018-01-12 Richard Biener <rguenther@suse.de>
7951 PR tree-optimization/80846
7952 * target.def (split_reduction): New target hook.
7953 * targhooks.c (default_split_reduction): New function.
7954 * targhooks.h (default_split_reduction): Declare.
7955 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
7956 target requests first reduce vectors by combining low and high
7957 parts.
7958 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
7959 (get_vectype_for_scalar_type_and_size): Export.
7960 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
7961 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
7962 * doc/tm.texi: Regenerate.
7963 * config/i386/i386.c (ix86_split_reduction): Implement
7964 TARGET_VECTORIZE_SPLIT_REDUCTION.
7966 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7968 PR target/83368
7969 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
7970 in PIC mode except for TARGET_VXWORKS_RTP.
7971 * config/sparc/sparc.c: Include cfgrtl.h.
7972 (TARGET_INIT_PIC_REG): Define.
7973 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
7974 (sparc_pic_register_p): New predicate.
7975 (sparc_legitimate_address_p): Use it.
7976 (sparc_legitimize_pic_address): Likewise.
7977 (sparc_delegitimize_address): Likewise.
7978 (sparc_mode_dependent_address_p): Likewise.
7979 (gen_load_pcrel_sym): Remove 4th parameter.
7980 (load_got_register): Adjust call to above. Remove obsolete stuff.
7981 (sparc_expand_prologue): Do not call load_got_register here.
7982 (sparc_flat_expand_prologue): Likewise.
7983 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
7984 (sparc_use_pseudo_pic_reg): New function.
7985 (sparc_init_pic_reg): Likewise.
7986 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7987 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
7989 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
7991 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
7992 Add item for branch_cost.
7994 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7996 PR rtl-optimization/83565
7997 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
7998 not extend the result to a larger mode for rotate operations.
7999 (num_sign_bit_copies1): Likewise.
8001 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8003 PR target/40411
8004 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8005 -symbolic.
8006 Use values-Xc.o for -pedantic.
8007 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8009 2018-01-12 Martin Liska <mliska@suse.cz>
8011 PR ipa/83054
8012 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8013 New function.
8014 (possible_polymorphic_call_targets): Use it.
8015 (ipa_devirt): Likewise.
8017 2018-01-12 Martin Liska <mliska@suse.cz>
8019 * profile-count.h (enum profile_quality): Use 0 as invalid
8020 enum value of profile_quality.
8022 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8024 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8025 -mext-string options.
8027 2018-01-12 Richard Biener <rguenther@suse.de>
8029 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8030 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8031 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8032 Likewise.
8033 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8035 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8037 * configure.ac (--with-long-double-format): Add support for the
8038 configuration option to change the default long double format on
8039 PowerPC systems.
8040 * config.gcc (powerpc*-linux*-*): Likewise.
8041 * configure: Regenerate.
8042 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8043 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8044 used without modification.
8046 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8048 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8049 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8050 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8051 MISC_BUILTIN_SPEC_BARRIER.
8052 (rs6000_init_builtins): Likewise.
8053 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8054 enum value.
8055 (speculation_barrier): New define_insn.
8056 * doc/extend.texi: Document __builtin_speculation_barrier.
8058 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8060 PR target/83203
8061 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8062 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8063 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8064 iterators.
8065 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8066 integral modes instead of "ss" and "sd".
8067 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8068 vectors with 32-bit and 64-bit elements.
8069 (vecdupssescalarmodesuffix): New mode attribute.
8070 (vec_dup<mode>): Use it.
8072 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8074 PR target/83330
8075 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8076 frame if argument is passed on stack.
8078 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8080 PR target/82682
8081 * ree.c (combine_reaching_defs): Optimize also
8082 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8083 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8085 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8087 PR middle-end/83189
8088 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8090 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8092 PR middle-end/83718
8093 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8094 after they are computed.
8096 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8098 PR tree-optimization/83695
8099 * gimple-loop-linterchange.cc
8100 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8101 reset cached scev information after interchange.
8102 (pass_linterchange::execute): Remove call to scev_reset_htab.
8104 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8106 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8107 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8108 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8109 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8110 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8111 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8112 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8113 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8114 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8115 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8116 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8117 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8118 (V_lane_reg): Likewise.
8119 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8120 New define_expand.
8121 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8122 (vfmal_lane_low<mode>_intrinsic,
8123 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8124 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8125 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8126 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8127 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8128 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8130 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8132 * config/arm/arm-cpus.in (fp16fml): New feature.
8133 (ALL_SIMD): Add fp16fml.
8134 (armv8.2-a): Add fp16fml as an option.
8135 (armv8.3-a): Likewise.
8136 (armv8.4-a): Add fp16fml as part of fp16.
8137 * config/arm/arm.h (TARGET_FP16FML): Define.
8138 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8139 when appropriate.
8140 * config/arm/arm-modes.def (V2HF): Define.
8141 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8142 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8143 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8144 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8145 vfmsl_low, vfmsl_high): New set of builtins.
8146 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8147 (vfml_op): New code attribute.
8148 (VFMLHALVES): New int iterator.
8149 (VFML, VFMLSEL): New mode attributes.
8150 (V_reg): Define mapping for V2HF.
8151 (V_hi, V_lo): New mode attributes.
8152 (VF_constraint): Likewise.
8153 (vfml_half, vfml_half_selector): New int attributes.
8154 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8155 define_expand.
8156 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8157 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8158 New define_insn.
8159 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8160 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8161 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8162 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8163 documentation.
8164 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8165 Document new effective target and option set.
8167 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8169 * config/arm/arm-cpus.in (armv8_4): New feature.
8170 (ARMv8_4a): New fgroup.
8171 (armv8.4-a): New arch.
8172 * config/arm/arm-tables.opt: Regenerate.
8173 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8174 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8175 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8176 Add matching rules for -march=armv8.4-a and extensions.
8177 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8179 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8181 PR target/81821
8182 * config/rx/rx.md (BW): New mode attribute.
8183 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8185 2018-01-11 Richard Biener <rguenther@suse.de>
8187 PR tree-optimization/83435
8188 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8189 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8190 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8192 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8193 Alan Hayward <alan.hayward@arm.com>
8194 David Sherwood <david.sherwood@arm.com>
8196 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8197 field.
8198 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8199 (aarch64_print_address_internal): Use it to check for a zero offset.
8201 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8202 Alan Hayward <alan.hayward@arm.com>
8203 David Sherwood <david.sherwood@arm.com>
8205 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8206 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8207 Return a poly_int64 rather than a HOST_WIDE_INT.
8208 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8209 rather than a HOST_WIDE_INT.
8210 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8211 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8212 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8213 final_offset from HOST_WIDE_INT to poly_int64.
8214 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8215 to_constant when getting the number of units in an Advanced SIMD
8216 mode.
8217 (aarch64_builtin_vectorized_function): Check for a constant number
8218 of units.
8219 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8220 GET_MODE_SIZE.
8221 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8222 attribute instead of GET_MODE_NUNITS.
8223 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8224 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8225 GET_MODE_SIZE for fixed-size registers.
8226 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8227 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8228 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8229 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8230 (aarch64_print_operand, aarch64_print_address_internal)
8231 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8232 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8233 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8234 Handle polynomial GET_MODE_SIZE.
8235 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8236 wider than SImode without modification.
8237 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8238 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8239 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8240 passing and returning SVE modes.
8241 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8242 rather than GEN_INT.
8243 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8244 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8245 (aarch64_allocate_and_probe_stack_space): Likewise.
8246 (aarch64_layout_frame): Cope with polynomial offsets.
8247 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8248 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8249 polynomial offsets.
8250 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8251 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8252 poly_int64 rather than a HOST_WIDE_INT.
8253 (aarch64_get_separate_components, aarch64_process_components)
8254 (aarch64_expand_prologue, aarch64_expand_epilogue)
8255 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8256 (aarch64_anchor_offset): New function, split out from...
8257 (aarch64_legitimize_address): ...here.
8258 (aarch64_builtin_vectorization_cost): Handle polynomial
8259 TYPE_VECTOR_SUBPARTS.
8260 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8261 GET_MODE_NUNITS.
8262 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8263 number of elements from the PARALLEL rather than the mode.
8264 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8265 rather than GET_MODE_BITSIZE.
8266 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8267 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8268 (aarch64_expand_vec_perm_const_1): Handle polynomial
8269 d->perm.length () and d->perm elements.
8270 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8271 Apply to_constant to d->perm elements.
8272 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8273 polynomial CONST_VECTOR_NUNITS.
8274 (aarch64_move_pointer): Take amount as a poly_int64 rather
8275 than an int.
8276 (aarch64_progress_pointer): Avoid temporary variable.
8277 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8278 the mode attribute instead of GET_MODE.
8280 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8281 Alan Hayward <alan.hayward@arm.com>
8282 David Sherwood <david.sherwood@arm.com>
8284 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8285 x exists before using it.
8286 (aarch64_add_constant_internal): Rename to...
8287 (aarch64_add_offset_1): ...this. Replace regnum with separate
8288 src and dest rtxes. Handle the case in which they're different,
8289 including when the offset is zero. Replace scratchreg with an rtx.
8290 Use 2 additions if there is no spare register into which we can
8291 move a 16-bit constant.
8292 (aarch64_add_constant): Delete.
8293 (aarch64_add_offset): Replace reg with separate src and dest
8294 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8295 Use aarch64_add_offset_1.
8296 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8297 an rtx rather than an int. Take the delta as a poly_int64
8298 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8299 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8300 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8301 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8302 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8303 and aarch64_add_sp.
8304 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8305 aarch64_add_constant.
8307 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8309 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8310 Use scalar_float_mode.
8312 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8314 * config/aarch64/aarch64-simd.md
8315 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8316 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8317 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8318 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8319 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8320 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8321 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8322 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8323 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8324 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8326 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8328 PR target/83514
8329 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8330 targ_options->x_arm_arch_string is non NULL.
8332 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8334 * config/aarch64/aarch64.h
8335 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8337 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8339 PR target/82096
8340 * expmed.c (emit_store_flag_force): Swap if const op0
8341 and change VOIDmode to mode of op0.
8343 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8345 PR rtl-optimization/83761
8346 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8347 than bytes to mode_for_size.
8349 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8351 PR middle-end/83189
8352 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8353 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8354 profile.
8356 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8358 PR middle-end/83575
8359 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8360 when in layout mode.
8361 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8362 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8363 partition fixup.
8365 2018-01-10 Michael Collison <michael.collison@arm.com>
8367 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8368 * config/aarch64/aarch64-option-extension.def: Add
8369 AARCH64_OPT_EXTENSION of 'fp16fml'.
8370 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8371 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8372 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8373 * config/aarch64/constraints.md (Ui7): New constraint.
8374 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8375 (VFMLA_SEL_W): Ditto.
8376 (f16quad): Ditto.
8377 (f16mac1): Ditto.
8378 (VFMLA16_LOW): New int iterator.
8379 (VFMLA16_HIGH): Ditto.
8380 (UNSPEC_FMLAL): New unspec.
8381 (UNSPEC_FMLSL): Ditto.
8382 (UNSPEC_FMLAL2): Ditto.
8383 (UNSPEC_FMLSL2): Ditto.
8384 (f16mac): New code attribute.
8385 * config/aarch64/aarch64-simd-builtins.def
8386 (aarch64_fmlal_lowv2sf): Ditto.
8387 (aarch64_fmlsl_lowv2sf): Ditto.
8388 (aarch64_fmlalq_lowv4sf): Ditto.
8389 (aarch64_fmlslq_lowv4sf): Ditto.
8390 (aarch64_fmlal_highv2sf): Ditto.
8391 (aarch64_fmlsl_highv2sf): Ditto.
8392 (aarch64_fmlalq_highv4sf): Ditto.
8393 (aarch64_fmlslq_highv4sf): Ditto.
8394 (aarch64_fmlal_lane_lowv2sf): Ditto.
8395 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8396 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8397 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8398 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8399 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8400 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8401 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8402 (aarch64_fmlal_lane_highv2sf): Ditto.
8403 (aarch64_fmlsl_lane_highv2sf): Ditto.
8404 (aarch64_fmlal_laneq_highv2sf): Ditto.
8405 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8406 (aarch64_fmlalq_lane_highv4sf): Ditto.
8407 (aarch64_fmlsl_lane_highv4sf): Ditto.
8408 (aarch64_fmlalq_laneq_highv4sf): Ditto.
8409 (aarch64_fmlsl_laneq_highv4sf): Ditto.
8410 * config/aarch64/aarch64-simd.md:
8411 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8412 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8413 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8414 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8415 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8416 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8417 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8418 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8419 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8420 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8421 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8422 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8423 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8424 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8425 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8426 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8427 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8428 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8429 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8430 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8431 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8432 (vfmlsl_low_u32): Ditto.
8433 (vfmlalq_low_u32): Ditto.
8434 (vfmlslq_low_u32): Ditto.
8435 (vfmlal_high_u32): Ditto.
8436 (vfmlsl_high_u32): Ditto.
8437 (vfmlalq_high_u32): Ditto.
8438 (vfmlslq_high_u32): Ditto.
8439 (vfmlal_lane_low_u32): Ditto.
8440 (vfmlsl_lane_low_u32): Ditto.
8441 (vfmlal_laneq_low_u32): Ditto.
8442 (vfmlsl_laneq_low_u32): Ditto.
8443 (vfmlalq_lane_low_u32): Ditto.
8444 (vfmlslq_lane_low_u32): Ditto.
8445 (vfmlalq_laneq_low_u32): Ditto.
8446 (vfmlslq_laneq_low_u32): Ditto.
8447 (vfmlal_lane_high_u32): Ditto.
8448 (vfmlsl_lane_high_u32): Ditto.
8449 (vfmlal_laneq_high_u32): Ditto.
8450 (vfmlsl_laneq_high_u32): Ditto.
8451 (vfmlalq_lane_high_u32): Ditto.
8452 (vfmlslq_lane_high_u32): Ditto.
8453 (vfmlalq_laneq_high_u32): Ditto.
8454 (vfmlslq_laneq_high_u32): Ditto.
8455 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
8456 (AARCH64_FL_FOR_ARCH8_4): New.
8457 (AARCH64_ISA_F16FML): New ISA flag.
8458 (TARGET_F16FML): New feature flag for fp16fml.
8459 (doc/invoke.texi): Document new fp16fml option.
8461 2018-01-10 Michael Collison <michael.collison@arm.com>
8463 * config/aarch64/aarch64-builtins.c:
8464 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
8465 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8466 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
8467 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
8468 (AARCH64_ISA_SHA3): New ISA flag.
8469 (TARGET_SHA3): New feature flag for sha3.
8470 * config/aarch64/iterators.md (sha512_op): New int attribute.
8471 (CRYPTO_SHA512): New int iterator.
8472 (UNSPEC_SHA512H): New unspec.
8473 (UNSPEC_SHA512H2): Ditto.
8474 (UNSPEC_SHA512SU0): Ditto.
8475 (UNSPEC_SHA512SU1): Ditto.
8476 * config/aarch64/aarch64-simd-builtins.def
8477 (aarch64_crypto_sha512hqv2di): New builtin.
8478 (aarch64_crypto_sha512h2qv2di): Ditto.
8479 (aarch64_crypto_sha512su0qv2di): Ditto.
8480 (aarch64_crypto_sha512su1qv2di): Ditto.
8481 (aarch64_eor3qv8hi): Ditto.
8482 (aarch64_rax1qv2di): Ditto.
8483 (aarch64_xarqv2di): Ditto.
8484 (aarch64_bcaxqv8hi): Ditto.
8485 * config/aarch64/aarch64-simd.md:
8486 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8487 (aarch64_crypto_sha512su0qv2di): Ditto.
8488 (aarch64_crypto_sha512su1qv2di): Ditto.
8489 (aarch64_eor3qv8hi): Ditto.
8490 (aarch64_rax1qv2di): Ditto.
8491 (aarch64_xarqv2di): Ditto.
8492 (aarch64_bcaxqv8hi): Ditto.
8493 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8494 (vsha512h2q_u64): Ditto.
8495 (vsha512su0q_u64): Ditto.
8496 (vsha512su1q_u64): Ditto.
8497 (veor3q_u16): Ditto.
8498 (vrax1q_u64): Ditto.
8499 (vxarq_u64): Ditto.
8500 (vbcaxq_u16): Ditto.
8501 * config/arm/types.md (crypto_sha512): New type attribute.
8502 (crypto_sha3): Ditto.
8503 (doc/invoke.texi): Document new sha3 option.
8505 2018-01-10 Michael Collison <michael.collison@arm.com>
8507 * config/aarch64/aarch64-builtins.c:
8508 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8509 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8510 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8511 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8512 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8513 (AARCH64_ISA_SM4): New ISA flag.
8514 (TARGET_SM4): New feature flag for sm4.
8515 * config/aarch64/aarch64-simd-builtins.def
8516 (aarch64_sm3ss1qv4si): Ditto.
8517 (aarch64_sm3tt1aq4si): Ditto.
8518 (aarch64_sm3tt1bq4si): Ditto.
8519 (aarch64_sm3tt2aq4si): Ditto.
8520 (aarch64_sm3tt2bq4si): Ditto.
8521 (aarch64_sm3partw1qv4si): Ditto.
8522 (aarch64_sm3partw2qv4si): Ditto.
8523 (aarch64_sm4eqv4si): Ditto.
8524 (aarch64_sm4ekeyqv4si): Ditto.
8525 * config/aarch64/aarch64-simd.md:
8526 (aarch64_sm3ss1qv4si): Ditto.
8527 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8528 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8529 (aarch64_sm4eqv4si): Ditto.
8530 (aarch64_sm4ekeyqv4si): Ditto.
8531 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8532 (sm3part_op): Ditto.
8533 (CRYPTO_SM3TT): Ditto.
8534 (CRYPTO_SM3PART): Ditto.
8535 (UNSPEC_SM3SS1): New unspec.
8536 (UNSPEC_SM3TT1A): Ditto.
8537 (UNSPEC_SM3TT1B): Ditto.
8538 (UNSPEC_SM3TT2A): Ditto.
8539 (UNSPEC_SM3TT2B): Ditto.
8540 (UNSPEC_SM3PARTW1): Ditto.
8541 (UNSPEC_SM3PARTW2): Ditto.
8542 (UNSPEC_SM4E): Ditto.
8543 (UNSPEC_SM4EKEY): Ditto.
8544 * config/aarch64/constraints.md (Ui2): New constraint.
8545 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8546 * config/arm/types.md (crypto_sm3): New type attribute.
8547 (crypto_sm4): Ditto.
8548 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8549 (vsm3tt1aq_u32): Ditto.
8550 (vsm3tt1bq_u32): Ditto.
8551 (vsm3tt2aq_u32): Ditto.
8552 (vsm3tt2bq_u32): Ditto.
8553 (vsm3partw1q_u32): Ditto.
8554 (vsm3partw2q_u32): Ditto.
8555 (vsm4eq_u32): Ditto.
8556 (vsm4ekeyq_u32): Ditto.
8557 (doc/invoke.texi): Document new sm4 option.
8559 2018-01-10 Michael Collison <michael.collison@arm.com>
8561 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8562 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8563 (AARCH64_FL_FOR_ARCH8_4): New.
8564 (AARCH64_FL_V8_4): New flag.
8565 (doc/invoke.texi): Document new armv8.4-a option.
8567 2018-01-10 Michael Collison <michael.collison@arm.com>
8569 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8570 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8571 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8572 * config/aarch64/aarch64-option-extension.def: Add
8573 AARCH64_OPT_EXTENSION of 'sha2'.
8574 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8575 (crypto): Disable sha2 and aes if crypto disabled.
8576 (crypto): Enable aes and sha2 if enabled.
8577 (simd): Disable sha2 and aes if simd disabled.
8578 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8579 New flags.
8580 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8581 (TARGET_SHA2): New feature flag for sha2.
8582 (TARGET_AES): New feature flag for aes.
8583 * config/aarch64/aarch64-simd.md:
8584 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8585 conditional on TARGET_AES.
8586 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8587 (aarch64_crypto_sha1hsi): Make pattern conditional
8588 on TARGET_SHA2.
8589 (aarch64_crypto_sha1hv4si): Ditto.
8590 (aarch64_be_crypto_sha1hv4si): Ditto.
8591 (aarch64_crypto_sha1su1v4si): Ditto.
8592 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8593 (aarch64_crypto_sha1su0v4si): Ditto.
8594 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8595 (aarch64_crypto_sha256su0v4si): Ditto.
8596 (aarch64_crypto_sha256su1v4si): Ditto.
8597 (doc/invoke.texi): Document new aes and sha2 options.
8599 2018-01-10 Martin Sebor <msebor@redhat.com>
8601 PR tree-optimization/83781
8602 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8603 as string arrays.
8605 2018-01-11 Martin Sebor <msebor@gmail.com>
8606 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8608 PR tree-optimization/83501
8609 PR tree-optimization/81703
8611 * tree-ssa-strlen.c (get_string_cst): Rename...
8612 (get_string_len): ...to this. Handle global constants.
8613 (handle_char_store): Adjust.
8615 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
8616 Jim Wilson <jimw@sifive.com>
8618 * config/riscv/riscv-protos.h (riscv_output_return): New.
8619 * config/riscv/riscv.c (struct machine_function): New naked_p field.
8620 (riscv_attribute_table, riscv_output_return),
8621 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8622 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8623 (riscv_compute_frame_info): Only compute frame->mask if not a naked
8624 function.
8625 (riscv_expand_prologue): Add early return for naked function.
8626 (riscv_expand_epilogue): Likewise.
8627 (riscv_function_ok_for_sibcall): Return false for naked function.
8628 (riscv_set_current_function): New.
8629 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8630 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8631 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8632 * doc/extend.texi (RISC-V Function Attributes): New.
8634 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8636 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8637 check for 128-bit long double before checking TCmode.
8638 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8639 128-bit long doubles before checking TFmode or TCmode.
8640 (FLOAT128_IBM_P): Likewise.
8642 2018-01-10 Martin Sebor <msebor@redhat.com>
8644 PR tree-optimization/83671
8645 * builtins.c (c_strlen): Unconditionally return zero for the empty
8646 string.
8647 Use -Warray-bounds for warnings.
8648 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8649 for non-constant array indices with COMPONENT_REF, arrays of
8650 arrays, and pointers to arrays.
8651 (gimple_fold_builtin_strlen): Determine and set length range for
8652 non-constant character arrays.
8654 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8656 PR middle-end/81897
8657 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8658 empty blocks.
8660 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8662 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8664 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8666 PR target/83399
8667 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8668 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8669 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8670 indexed_or_indirect_operand predicate.
8671 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8672 (*vsx_le_perm_load_v8hi): Likewise.
8673 (*vsx_le_perm_load_v16qi): Likewise.
8674 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8675 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8676 (*vsx_le_perm_store_v8hi): Likewise.
8677 (*vsx_le_perm_store_v16qi): Likewise.
8678 (eight unnamed splitters): Likewise.
8680 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8682 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8683 * config/rs6000/emmintrin.h: Likewise.
8684 * config/rs6000/mmintrin.h: Likewise.
8685 * config/rs6000/xmmintrin.h: Likewise.
8687 2018-01-10 David Malcolm <dmalcolm@redhat.com>
8689 PR c++/43486
8690 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8691 "public_flag".
8692 * tree.c (tree_nop_conversion): Return true for location wrapper
8693 nodes.
8694 (maybe_wrap_with_location): New function.
8695 (selftest::check_strip_nops): New function.
8696 (selftest::test_location_wrappers): New function.
8697 (selftest::tree_c_tests): Call it.
8698 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8699 (maybe_wrap_with_location): New decl.
8700 (EXPR_LOCATION_WRAPPER_P): New macro.
8701 (location_wrapper_p): New inline function.
8702 (tree_strip_any_location_wrapper): New inline function.
8704 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
8706 PR target/83735
8707 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8708 stack_realign_offset for the largest alignment of stack slot
8709 actually used.
8710 (ix86_find_max_used_stack_alignment): New function.
8711 (ix86_finalize_stack_frame_flags): Use it. Set
8712 max_used_stack_alignment if we don't realign stack.
8713 * config/i386/i386.h (machine_function): Add
8714 max_used_stack_alignment.
8716 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
8718 * config/arm/arm.opt (-mbranch-cost): New option.
8719 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8720 account.
8722 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
8724 PR target/83629
8725 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8726 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8728 2018-01-10 Richard Biener <rguenther@suse.de>
8730 PR debug/83765
8731 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8732 early out so it also covers the case where we have a non-NULL
8733 origin.
8735 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8737 PR tree-optimization/83753
8738 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
8739 for non-strided grouped accesses if the number of elements is 1.
8741 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8743 PR target/81616
8744 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
8745 * i386.h (TARGET_USE_GATHER): Define.
8746 * x86-tune.def (X86_TUNE_USE_GATHER): New.
8748 2018-01-10 Martin Liska <mliska@suse.cz>
8750 PR bootstrap/82831
8751 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
8752 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
8753 partitioning.
8754 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
8755 CLEANUP_NO_PARTITIONING is not set.
8757 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8759 * doc/rtl.texi: Remove documentation of (const ...) wrappers
8760 for vectors, as a partial revert of r254296.
8761 * rtl.h (const_vec_p): Delete.
8762 (const_vec_duplicate_p): Don't test for vector CONSTs.
8763 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
8764 * expmed.c (make_tree): Likewise.
8766 Revert:
8767 * common.md (E, F): Use CONSTANT_P instead of checking for
8768 CONST_VECTOR.
8769 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
8770 checking for CONST_VECTOR.
8772 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8774 PR middle-end/83575
8775 * predict.c (force_edge_cold): Handle in more sane way edges
8776 with no prediction.
8778 2018-01-09 Carl Love <cel@us.ibm.com>
8780 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
8781 V4SI, V4SF types.
8782 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
8783 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
8784 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
8785 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
8786 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
8787 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
8788 * config/rs6000/rs6000-protos.h: Add extern defition for
8789 rs6000_generate_float2_double_code.
8790 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8791 function.
8792 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8793 (float2_v2df): Add define_expand.
8795 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
8797 PR target/83628
8798 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8799 op_mode in the force_to_mode call.
8801 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8803 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8804 instead of checking each element individually.
8805 (aarch64_evpc_uzp): Likewise.
8806 (aarch64_evpc_zip): Likewise.
8807 (aarch64_evpc_ext): Likewise.
8808 (aarch64_evpc_rev): Likewise.
8809 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8810 instead of checking each element individually. Return true without
8811 generating rtl if
8812 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8813 whether all selected elements come from the same input, instead of
8814 checking each element individually. Remove calls to gen_rtx_REG,
8815 start_sequence and end_sequence and instead assert that no rtl is
8816 generated.
8818 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8820 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8821 order of HIGH and CONST checks.
8823 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8825 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8826 if the destination isn't an SSA_NAME.
8828 2018-01-09 Richard Biener <rguenther@suse.de>
8830 PR tree-optimization/83668
8831 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8832 move prologue...
8833 (canonicalize_loop_form): ... here, renamed from ...
8834 (canonicalize_loop_closed_ssa_form): ... this and amended to
8835 swap successor edges for loop exit blocks to make us use
8836 the RPO order we need for initial schedule generation.
8838 2018-01-09 Joseph Myers <joseph@codesourcery.com>
8840 PR tree-optimization/64811
8841 * match.pd: When optimizing comparisons with Inf, avoid
8842 introducing or losing exceptions from comparisons with NaN.
8844 2018-01-09 Martin Liska <mliska@suse.cz>
8846 PR sanitizer/82517
8847 * asan.c (shadow_mem_size): Add gcc_assert.
8849 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8851 Don't save registers in main().
8853 PR target/83738
8854 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8855 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8856 * config/avr/avr.c (avr_set_current_function): Don't error if
8857 naked, OS_task or OS_main are specified at the same time.
8858 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8859 OS_main.
8860 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8861 attribute.
8862 * common/config/avr/avr-common.c (avr_option_optimization_table):
8863 Switch on -mmain-is-OS_task for optimizing compilations.
8865 2018-01-09 Richard Biener <rguenther@suse.de>
8867 PR tree-optimization/83572
8868 * graphite.c: Include cfganal.h.
8869 (graphite_transform_loops): Connect infinite loops to exit
8870 and remove fake edges at the end.
8872 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8874 * ipa-inline.c (edge_badness): Revert accidental checkin.
8876 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8878 PR ipa/80763
8879 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8880 symbols; not inline clones.
8882 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8884 PR target/83507
8885 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8886 hard registers. Formatting fixes.
8888 PR preprocessor/83722
8889 * gcc.c (try_generate_repro): Pass
8890 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8891 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8892 do_report_bug.
8894 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8895 Kito Cheng <kito.cheng@gmail.com>
8897 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8898 (riscv_leaf_function_p): Delete.
8899 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8901 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8903 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8904 function.
8905 (do_ifelse): New function.
8906 (do_isel): New function.
8907 (do_sub3): New function.
8908 (do_add3): New function.
8909 (do_load_mask_compare): New function.
8910 (do_overlap_load_compare): New function.
8911 (expand_compare_loop): New function.
8912 (expand_block_compare): Call expand_compare_loop() when appropriate.
8913 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8914 option description.
8915 (-mblock-compare-inline-loop-limit): New option.
8917 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8919 PR target/83677
8920 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
8921 Reverse order of second and third operands in first alternative.
8922 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
8923 of first and second elements in UNSPEC_VPERMR vector.
8924 (altivec_expand_vec_perm_le): Likewise.
8926 2018-01-08 Jeff Law <law@redhat.com>
8928 PR rtl-optimizatin/81308
8929 * tree-switch-conversion.c (cfg_altered): New file scoped static.
8930 (process_switch): If group_case_labels makes a change, then set
8931 cfg_altered.
8932 (pass_convert_switch::execute): If a switch is converted, then
8933 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
8935 PR rtl-optimization/81308
8936 * recog.c (split_all_insns): Conditionally cleanup the CFG after
8937 splitting insns.
8939 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
8941 PR target/83663 - Revert r255946
8942 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
8943 generation for cases where splatting a value is not useful.
8944 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8945 across a vec_duplicate and a paradoxical subreg forming a vector
8946 mode to a vec_concat.
8948 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8950 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
8951 -march=armv8.3-a variants.
8952 * config/arm/t-multilib: Likewise.
8953 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
8955 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8957 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
8958 to generate rtl.
8959 (cceq_ior_compare_complement): Give it a name so I can use it, and
8960 change boolean_or_operator predicate to boolean_operator so it can
8961 be used to generate a crand.
8962 (eqne): New code iterator.
8963 (bd/bd_neg): New code_attrs.
8964 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
8965 a single define_insn.
8966 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
8967 decrement (bdnzt/bdnzf/bdzt/bdzf).
8968 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
8969 with the new names of the branch decrement patterns, and added the
8970 names of the branch decrement conditional patterns.
8972 2018-01-08 Richard Biener <rguenther@suse.de>
8974 PR tree-optimization/83563
8975 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
8976 cache.
8978 2018-01-08 Richard Biener <rguenther@suse.de>
8980 PR middle-end/83713
8981 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
8983 2018-01-08 Richard Biener <rguenther@suse.de>
8985 PR tree-optimization/83685
8986 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
8987 references to abnormals.
8989 2018-01-08 Richard Biener <rguenther@suse.de>
8991 PR lto/83719
8992 * dwarf2out.c (output_indirect_strings): Handle empty
8993 skeleton_debug_str_hash.
8994 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
8996 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8998 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
8999 (emit_store_direct): Likewise.
9000 (arc_trampoline_adjust_address): Likewise.
9001 (arc_asm_trampoline_template): New function.
9002 (arc_initialize_trampoline): Use asm_trampoline_template.
9003 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9004 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9005 * config/arc/arc.md (flush_icache): Delete pattern.
9007 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9009 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9010 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9011 munaligned-access.
9013 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9015 PR target/83681
9016 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9017 by not USED_FOR_TARGET.
9018 (make_pass_resolve_sw_modes): Likewise.
9020 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9022 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9023 USED_FOR_TARGET.
9025 2018-01-08 Richard Biener <rguenther@suse.de>
9027 PR middle-end/83580
9028 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9030 2018-01-08 Richard Biener <rguenther@suse.de>
9032 PR middle-end/83517
9033 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9035 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9037 PR middle-end/81897
9038 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9039 basic blocks with a small number of successors.
9040 (convert_control_dep_chain_into_preds): Improve handling of
9041 forwarder blocks.
9042 (dump_predicates): Split apart into...
9043 (dump_pred_chain): ...here...
9044 (dump_pred_info): ...and here.
9045 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9046 (can_chain_union_be_invalidated_p): Improve check for invalidation
9047 of paths.
9048 (uninit_uses_cannot_happen): Avoid unnecessary if
9049 convert_control_dep_chain_into_preds yielded nothing.
9051 2018-01-06 Martin Sebor <msebor@redhat.com>
9053 PR tree-optimization/83640
9054 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9055 subtracting negative offset from size.
9056 (builtin_access::overlap): Adjust offset bounds of the access to fall
9057 within the size of the object if possible.
9059 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9061 PR rtl-optimization/83699
9062 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9063 extract_bit_field_as_subreg to cases in which the extracted
9064 value is also a vector.
9066 * lra-constraints.c (process_alt_operands): Test for the equivalence
9067 substitutions when detecting a possible reload cycle.
9069 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9071 PR debug/83480
9072 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9073 by default if flag_selective_schedling{,2}. Formatting fixes.
9075 PR rtl-optimization/83682
9076 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9077 if it has non-VECTOR_MODE element mode.
9078 (vec_duplicate_p): Likewise.
9080 PR middle-end/83694
9081 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9082 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9084 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9086 PR target/83604
9087 * config/i386/i386-builtin.def
9088 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9089 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9090 Require also OPTION_MASK_ISA_AVX512F in addition to
9091 OPTION_MASK_ISA_GFNI.
9092 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9093 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9094 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9095 to OPTION_MASK_ISA_GFNI.
9096 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9097 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9098 OPTION_MASK_ISA_AVX512BW.
9099 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9100 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9101 addition to OPTION_MASK_ISA_GFNI.
9102 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9103 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9104 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9105 to OPTION_MASK_ISA_GFNI.
9106 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9107 a requirement for all ISAs rather than any of them with a few
9108 exceptions.
9109 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9110 processing.
9111 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9112 bitmasks to be enabled with 3 exceptions, instead of requiring any
9113 enabled ISA with lots of exceptions.
9114 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9115 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9116 Change avx512bw in isa attribute to avx512f.
9117 * config/i386/sgxintrin.h: Add license boilerplate.
9118 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9119 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9120 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9121 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9122 defined.
9123 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9124 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9125 temporarily sse2 rather than sse if not enabled already.
9127 PR target/83604
9128 * config/i386/sse.md (VI248_VLBW): Rename to ...
9129 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9130 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9131 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9132 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9133 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9134 mode iterator instead of VI248_VLBW.
9136 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9138 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9139 (record_modified): Skip clobbers; add debug output.
9140 (param_change_prob): Use sreal frequencies.
9142 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9144 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9145 punt for user-aligned variables.
9147 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9149 * tree-chrec.c (chrec_contains_symbols): Return true for
9150 POLY_INT_CST.
9152 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9154 PR target/82439
9155 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9156 of (x|y) == x for BICS pattern.
9158 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9160 PR tree-optimization/83605
9161 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9162 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9163 can throw.
9165 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9167 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9168 * config/epiphany/rtems.h: New file.
9170 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9171 Uros Bizjak <ubizjak@gmail.com>
9173 PR target/83554
9174 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9175 QIreg_operand instead of register_operand predicate.
9176 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9177 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9178 comments instead of -fmitigate[-_]rop.
9180 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9182 PR bootstrap/81926
9183 * cgraphunit.c (symbol_table::compile): Switch to text_section
9184 before calling assembly_start debug hook.
9185 * run-rtl-passes.c (run_rtl_passes): Likewise.
9186 Include output.h.
9188 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9190 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9191 range_int_cst_p rather than !symbolic_range_p before calling
9192 extract_range_from_multiplicative_op_1.
9194 2018-01-04 Jeff Law <law@redhat.com>
9196 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9197 redundant test in assertion.
9199 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9201 * doc/rtl.texi: Document machine_mode wrapper classes.
9203 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9205 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9206 using tree_to_uhwi.
9208 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9210 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9211 the VEC_PERM_EXPR fold to fail.
9213 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9215 PR debug/83585
9216 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9217 to switched_sections.
9219 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9221 PR target/83680
9222 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9223 test for d.testing.
9225 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9227 PR target/83387
9228 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9229 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9231 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9233 PR debug/83666
9234 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9235 is BLKmode and bitpos not zero or mode change is needed.
9237 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9239 PR target/83675
9240 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9241 TARGET_VIS2.
9243 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9245 PR target/83628
9246 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9247 instead of MULT rtx. Update all corresponding splitters.
9248 (*saddl_se): Ditto.
9249 (*ssub<modesuffix>): Ditto.
9250 (*ssubl_se): Ditto.
9251 (*cmp_sadd_di): Update split patterns.
9252 (*cmp_sadd_si): Ditto.
9253 (*cmp_sadd_sidi): Ditto.
9254 (*cmp_ssub_di): Ditto.
9255 (*cmp_ssub_si): Ditto.
9256 (*cmp_ssub_sidi): Ditto.
9257 * config/alpha/predicates.md (const23_operand): New predicate.
9258 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9259 Look for ASHIFT, not MULT inner operand.
9260 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9262 2018-01-04 Martin Liska <mliska@suse.cz>
9264 PR gcov-profile/83669
9265 * gcov.c (output_intermediate_file): Add version to intermediate
9266 gcov file.
9267 * doc/gcov.texi: Document new field 'version' in intermediate
9268 file format. Fix location of '-k' option of gcov command.
9270 2018-01-04 Martin Liska <mliska@suse.cz>
9272 PR ipa/82352
9273 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9275 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9277 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9279 2018-01-03 Martin Sebor <msebor@redhat.com>
9281 PR tree-optimization/83655
9282 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9283 checking calls with invalid arguments.
9285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9287 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9288 (vectorizable_mask_load_store): Delete.
9289 (vectorizable_call): Return false for masked loads and stores.
9290 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9291 instead of gimple_assign_rhs1.
9292 (vectorizable_load): Handle IFN_MASK_LOAD.
9293 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9297 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9298 split out from..,
9299 (vectorizable_mask_load_store): ...here.
9300 (vectorizable_load): ...and here.
9302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9304 * tree-vect-stmts.c (vect_build_all_ones_mask)
9305 (vect_build_zero_merge_argument): New functions, split out from...
9306 (vectorizable_load): ...here.
9308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9310 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9311 split out from...
9312 (vectorizable_mask_load_store): ...here.
9313 (vectorizable_store): ...and here.
9315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9317 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9318 split out from...
9319 (vectorizable_mask_load_store): ...here.
9321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9323 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9324 (vect_model_store_cost): Take a vec_load_store_type instead of a
9325 vect_def_type.
9326 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9327 (vect_model_store_cost): Take a vec_load_store_type instead of a
9328 vect_def_type.
9329 (vectorizable_mask_load_store): Update accordingly.
9330 (vectorizable_store): Likewise.
9331 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9335 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9336 IFN_MASK_LOAD calls here rather than...
9337 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9340 Alan Hayward <alan.hayward@arm.com>
9341 David Sherwood <david.sherwood@arm.com>
9343 * expmed.c (extract_bit_field_1): For vector extracts,
9344 fall back to extract_bit_field_as_subreg if vec_extract
9345 isn't available.
9347 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9348 Alan Hayward <alan.hayward@arm.com>
9349 David Sherwood <david.sherwood@arm.com>
9351 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9352 they are variable or constant sized.
9353 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9354 slots for constant-sized data.
9356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9357 Alan Hayward <alan.hayward@arm.com>
9358 David Sherwood <david.sherwood@arm.com>
9360 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9361 handling COND_EXPRs with boolean comparisons, try to find a better
9362 basis for the mask type than the boolean itself.
9364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9366 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9367 is calculated and how it can be overridden.
9368 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9369 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9370 if defined.
9371 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9372 if nonzero.
9374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9375 Alan Hayward <alan.hayward@arm.com>
9376 David Sherwood <david.sherwood@arm.com>
9378 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9379 Remove the mode argument.
9380 (aarch64_simd_valid_immediate): Remove the mode and inverse
9381 arguments.
9382 * config/aarch64/iterators.md (bitsize): New iterator.
9383 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9384 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9385 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9386 aarch64_simd_valid_immediate.
9387 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9388 (aarch64_reg_or_bic_imm): Likewise.
9389 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9390 with an insn_type enum and msl with a modifier_type enum.
9391 Replace element_width with a scalar_mode. Change the shift
9392 to unsigned int. Add constructors for scalar_float_mode and
9393 scalar_int_mode elements.
9394 (aarch64_vect_float_const_representable_p): Delete.
9395 (aarch64_can_const_movi_rtx_p)
9396 (aarch64_simd_scalar_immediate_valid_for_move)
9397 (aarch64_simd_make_constant): Update call to
9398 aarch64_simd_valid_immediate.
9399 (aarch64_advsimd_valid_immediate_hs): New function.
9400 (aarch64_advsimd_valid_immediate): Likewise.
9401 (aarch64_simd_valid_immediate): Remove mode and inverse
9402 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9403 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9404 and aarch64_float_const_representable_p on the result.
9405 (aarch64_output_simd_mov_immediate): Remove mode argument.
9406 Update call to aarch64_simd_valid_immediate and use of
9407 simd_immediate_info.
9408 (aarch64_output_scalar_simd_mov_immediate): Update call
9409 accordingly.
9411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9412 Alan Hayward <alan.hayward@arm.com>
9413 David Sherwood <david.sherwood@arm.com>
9415 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9416 (mode_nunits): Likewise CONST_MODE_NUNITS.
9417 * machmode.def (ADJUST_NUNITS): Document.
9418 * genmodes.c (mode_data::need_nunits_adj): New field.
9419 (blank_mode): Update accordingly.
9420 (adj_nunits): New variable.
9421 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9422 parameter.
9423 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9424 listed in adj_nunits.
9425 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9426 listed in adj_nunits. Don't emit case statements for such modes.
9427 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9428 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9429 nothing if adj_nunits is nonnull.
9430 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9431 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9432 (emit_mode_fbit): Update use of print_maybe_const_decl.
9433 (emit_move_size): Likewise. Treat the array as non-const
9434 if adj_nunits.
9435 (emit_mode_adjustments): Handle adj_nunits.
9437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9439 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9440 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9441 (VECTOR_MODES): Use it.
9442 (make_vector_modes): Take the prefix as an argument.
9444 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9445 Alan Hayward <alan.hayward@arm.com>
9446 David Sherwood <david.sherwood@arm.com>
9448 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
9449 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
9450 for MODE_VECTOR_BOOL.
9451 * machmode.def (VECTOR_BOOL_MODE): Document.
9452 * genmodes.c (VECTOR_BOOL_MODE): New macro.
9453 (make_vector_bool_mode): New function.
9454 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
9455 MODE_VECTOR_BOOL.
9456 * lto-streamer-in.c (lto_input_mode_table): Likewise.
9457 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
9458 Likewise.
9459 * stor-layout.c (int_mode_for_mode): Likewise.
9460 * tree.c (build_vector_type_for_mode): Likewise.
9461 * varasm.c (output_constant_pool_2): Likewise.
9462 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
9463 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
9464 for MODE_VECTOR_BOOL.
9465 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
9466 of mode class checks.
9467 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
9468 instead of a list of mode class checks.
9469 (expand_vector_scalar_condition): Likewise.
9470 (type_for_widest_vector_mode): Handle BImode as an inner mode.
9472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9473 Alan Hayward <alan.hayward@arm.com>
9474 David Sherwood <david.sherwood@arm.com>
9476 * machmode.h (mode_size): Change from unsigned short to
9477 poly_uint16_pod.
9478 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
9479 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9480 or if measurement_type is not polynomial.
9481 (fixed_size_mode::includes_p): Check for constant-sized modes.
9482 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9483 return a poly_uint16 rather than an unsigned short.
9484 (emit_mode_size): Change the type of mode_size from unsigned short
9485 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
9486 (emit_mode_adjustments): Cope with polynomial vector sizes.
9487 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9488 for GET_MODE_SIZE.
9489 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9490 for GET_MODE_SIZE.
9491 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9492 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9493 * caller-save.c (setup_save_areas): Likewise.
9494 (replace_reg_with_saved_mem): Likewise.
9495 * calls.c (emit_library_call_value_1): Likewise.
9496 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9497 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9498 (gen_lowpart_for_combine): Likewise.
9499 * convert.c (convert_to_integer_1): Likewise.
9500 * cse.c (equiv_constant, cse_insn): Likewise.
9501 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9502 (cselib_subst_to_values): Likewise.
9503 * dce.c (word_dce_process_block): Likewise.
9504 * df-problems.c (df_word_lr_mark_ref): Likewise.
9505 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9506 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9507 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9508 (rtl_for_decl_location): Likewise.
9509 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9510 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9511 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9512 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9513 (expand_expr_real_1): Likewise.
9514 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9515 (pad_below): Likewise.
9516 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9517 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9518 * ira.c (get_subreg_tracking_sizes): Likewise.
9519 * ira-build.c (ira_create_allocno_objects): Likewise.
9520 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9521 (ira_sort_regnos_for_alter_reg): Likewise.
9522 * ira-costs.c (record_operand_costs): Likewise.
9523 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9524 (resolve_simple_move): Likewise.
9525 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9526 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9527 (lra_constraints): Likewise.
9528 (CONST_POOL_OK_P): Reject variable-sized modes.
9529 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9530 (add_pseudo_to_slot, lra_spill): Likewise.
9531 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9532 * optabs-query.c (get_best_extraction_insn): Likewise.
9533 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9534 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9535 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9536 * recog.c (offsettable_address_addr_space_p): Likewise.
9537 * regcprop.c (maybe_mode_change): Likewise.
9538 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9539 * regrename.c (build_def_use): Likewise.
9540 * regstat.c (dump_reg_info): Likewise.
9541 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9542 (find_reloads, find_reloads_subreg_address): Likewise.
9543 * reload1.c (eliminate_regs_1): Likewise.
9544 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9545 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9546 (simplify_binary_operation_1, simplify_subreg): Likewise.
9547 * targhooks.c (default_function_arg_padding): Likewise.
9548 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9549 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9550 (verify_gimple_assign_ternary): Likewise.
9551 * tree-inline.c (estimate_move_cost): Likewise.
9552 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9553 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9554 (get_address_cost_ainc): Likewise.
9555 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9556 (vect_supportable_dr_alignment): Likewise.
9557 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9558 (vectorizable_reduction): Likewise.
9559 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9560 (vectorizable_operation, vectorizable_load): Likewise.
9561 * tree.c (build_same_sized_truth_vector_type): Likewise.
9562 * valtrack.c (cleanup_auto_inc_dec): Likewise.
9563 * var-tracking.c (emit_note_insn_var_location): Likewise.
9564 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9565 (ADDR_VEC_ALIGN): Likewise.
9567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9568 Alan Hayward <alan.hayward@arm.com>
9569 David Sherwood <david.sherwood@arm.com>
9571 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9572 unsigned short.
9573 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9574 or if measurement_type is polynomial.
9575 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9576 * combine.c (make_extraction): Likewise.
9577 * dse.c (find_shift_sequence): Likewise.
9578 * dwarf2out.c (mem_loc_descriptor): Likewise.
9579 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9580 (extract_bit_field, extract_low_bits): Likewise.
9581 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9582 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9583 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9584 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9585 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9586 * reload.c (find_reloads): Likewise.
9587 * reload1.c (alter_reg): Likewise.
9588 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9589 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9590 * tree-if-conv.c (predicate_mem_writes): Likewise.
9591 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9592 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9593 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9594 * valtrack.c (dead_debug_insert_temp): Likewise.
9595 * varasm.c (mergeable_constant_section): Likewise.
9596 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
9598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9599 Alan Hayward <alan.hayward@arm.com>
9600 David Sherwood <david.sherwood@arm.com>
9602 * expr.c (expand_assignment): Cope with polynomial mode sizes
9603 when assigning to a CONCAT.
9605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9606 Alan Hayward <alan.hayward@arm.com>
9607 David Sherwood <david.sherwood@arm.com>
9609 * machmode.h (mode_precision): Change from unsigned short to
9610 poly_uint16_pod.
9611 (mode_to_precision): Return a poly_uint16 rather than an unsigned
9612 short.
9613 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9614 or if measurement_type is not polynomial.
9615 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
9616 in which the mode is already known to be a scalar_int_mode.
9617 * genmodes.c (emit_mode_precision): Change the type of mode_precision
9618 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
9619 initializer.
9620 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9621 for GET_MODE_PRECISION.
9622 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9623 for GET_MODE_PRECISION.
9624 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9625 as polynomial.
9626 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9627 (expand_field_assignment, make_extraction): Likewise.
9628 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9629 (get_last_value): Likewise.
9630 * convert.c (convert_to_integer_1): Likewise.
9631 * cse.c (cse_insn): Likewise.
9632 * expr.c (expand_expr_real_1): Likewise.
9633 * lra-constraints.c (simplify_operand_subreg): Likewise.
9634 * optabs-query.c (can_atomic_load_p): Likewise.
9635 * optabs.c (expand_atomic_load): Likewise.
9636 (expand_atomic_store): Likewise.
9637 * ree.c (combine_reaching_defs): Likewise.
9638 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9639 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9640 * tree.h (type_has_mode_precision_p): Likewise.
9641 * ubsan.c (instrument_si_overflow): Likewise.
9643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9644 Alan Hayward <alan.hayward@arm.com>
9645 David Sherwood <david.sherwood@arm.com>
9647 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9648 polynomial numbers of units.
9649 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9650 (valid_vector_subparts_p): New function.
9651 (build_vector_type): Remove temporary shim and take the number
9652 of units as a poly_uint64 rather than an int.
9653 (build_opaque_vector_type): Take the number of units as a
9654 poly_uint64 rather than an int.
9655 * tree.c (build_vector_from_ctor): Handle polynomial
9656 TYPE_VECTOR_SUBPARTS.
9657 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9658 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9659 (build_vector_from_val): If the number of units is variable,
9660 use build_vec_duplicate_cst for constant operands and
9661 VEC_DUPLICATE_EXPR otherwise.
9662 (make_vector_type): Remove temporary is_constant ().
9663 (build_vector_type, build_opaque_vector_type): Take the number of
9664 units as a poly_uint64 rather than an int.
9665 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9666 VECTOR_CST_NELTS.
9667 * cfgexpand.c (expand_debug_expr): Likewise.
9668 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9669 (store_constructor, expand_expr_real_1): Likewise.
9670 (const_scalar_mask_from_tree): Likewise.
9671 * fold-const-call.c (fold_const_reduction): Likewise.
9672 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9673 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9674 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9675 (fold_relational_const): Likewise.
9676 (native_interpret_vector): Likewise. Change the size from an
9677 int to an unsigned int.
9678 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9679 TYPE_VECTOR_SUBPARTS.
9680 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9681 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9682 duplicating a non-constant operand into a variable-length vector.
9683 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9684 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9685 * ipa-icf.c (sem_variable::equals): Likewise.
9686 * match.pd: Likewise.
9687 * omp-simd-clone.c (simd_clone_subparts): Likewise.
9688 * print-tree.c (print_node): Likewise.
9689 * stor-layout.c (layout_type): Likewise.
9690 * targhooks.c (default_builtin_vectorization_cost): Likewise.
9691 * tree-cfg.c (verify_gimple_comparison): Likewise.
9692 (verify_gimple_assign_binary): Likewise.
9693 (verify_gimple_assign_ternary): Likewise.
9694 (verify_gimple_assign_single): Likewise.
9695 * tree-pretty-print.c (dump_generic_node): Likewise.
9696 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9697 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9698 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9699 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9700 (vect_shift_permute_load_chain): Likewise.
9701 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9702 (expand_vector_condition, optimize_vector_constructor): Likewise.
9703 (lower_vec_perm, get_compute_type): Likewise.
9704 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9705 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9706 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9707 (vect_recog_mask_conversion_pattern): Likewise.
9708 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9709 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9710 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9711 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9712 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9713 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9714 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9715 (supportable_widening_operation): Likewise.
9716 (supportable_narrowing_operation): Likewise.
9717 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9718 Likewise.
9719 * varasm.c (output_constant): Likewise.
9721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9722 Alan Hayward <alan.hayward@arm.com>
9723 David Sherwood <david.sherwood@arm.com>
9725 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9726 so that both the length == 3 and length != 3 cases set up their
9727 own permute vectors. Add comments explaining why we know the
9728 number of elements is constant.
9729 (vect_permute_load_chain): Likewise.
9731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9732 Alan Hayward <alan.hayward@arm.com>
9733 David Sherwood <david.sherwood@arm.com>
9735 * machmode.h (mode_nunits): Change from unsigned char to
9736 poly_uint16_pod.
9737 (ONLY_FIXED_SIZE_MODES): New macro.
9738 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
9739 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
9740 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
9741 New typedefs.
9742 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
9743 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
9744 or if measurement_type is not polynomial.
9745 * genmodes.c (ZERO_COEFFS): New macro.
9746 (emit_mode_nunits_inline): Make mode_nunits_inline return a
9747 poly_uint16.
9748 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
9749 Use ZERO_COEFFS when emitting initializers.
9750 * data-streamer.h (bp_pack_poly_value): New function.
9751 (bp_unpack_poly_value): Likewise.
9752 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9753 for GET_MODE_NUNITS.
9754 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9755 for GET_MODE_NUNITS.
9756 * tree.c (make_vector_type): Remove temporary shim and make
9757 the real function take the number of units as a poly_uint64
9758 rather than an int.
9759 (build_vector_type_for_mode): Handle polynomial nunits.
9760 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
9761 * emit-rtl.c (const_vec_series_p_1): Likewise.
9762 (gen_rtx_CONST_VECTOR): Likewise.
9763 * fold-const.c (test_vec_duplicate_folding): Likewise.
9764 * genrecog.c (validate_pattern): Likewise.
9765 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
9766 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9767 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
9768 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
9769 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
9770 * rtlanal.c (subreg_get_info): Likewise.
9771 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9772 (vect_grouped_load_supported): Likewise.
9773 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
9774 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
9775 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9776 (simplify_const_unary_operation, simplify_binary_operation_1)
9777 (simplify_const_binary_operation, simplify_ternary_operation)
9778 (test_vector_ops_duplicate, test_vector_ops): Likewise.
9779 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
9780 instead of CONST_VECTOR_NUNITS.
9781 * varasm.c (output_constant_pool_2): Likewise.
9782 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
9783 explicit-encoded elements in the XVEC for variable-length vectors.
9785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9787 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9790 Alan Hayward <alan.hayward@arm.com>
9791 David Sherwood <david.sherwood@arm.com>
9793 * coretypes.h (fixed_size_mode): Declare.
9794 (fixed_size_mode_pod): New typedef.
9795 * builtins.h (target_builtins::x_apply_args_mode)
9796 (target_builtins::x_apply_result_mode): Change type to
9797 fixed_size_mode_pod.
9798 * builtins.c (apply_args_size, apply_result_size, result_vector)
9799 (expand_builtin_apply_args_1, expand_builtin_apply)
9800 (expand_builtin_return): Update accordingly.
9802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9804 * cse.c (hash_rtx_cb): Hash only the encoded elements.
9805 * cselib.c (cselib_hash_rtx): Likewise.
9806 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9807 CONST_VECTOR encoding.
9809 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9810 Jeff Law <law@redhat.com>
9812 PR target/83641
9813 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9814 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9815 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9816 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9818 PR target/83641
9819 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9820 explicitly probe *sp in a noreturn function if there were any callee
9821 register saves or frame pointer is needed.
9823 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9825 PR debug/83621
9826 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9827 BLKmode for ternary, binary or unary expressions.
9829 PR debug/83645
9830 * var-tracking.c (delete_vta_debug_insn): New inline function.
9831 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9832 insns from get_insns () to NULL instead of each bb separately.
9833 Use delete_vta_debug_insn. No longer static.
9834 (vt_debug_insns_local, variable_tracking_main_1): Adjust
9835 delete_vta_debug_insns callers.
9836 * rtl.h (delete_vta_debug_insns): Declare.
9837 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9838 instead of variable_tracking_main.
9840 2018-01-03 Martin Sebor <msebor@redhat.com>
9842 PR tree-optimization/83603
9843 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9844 arguments past the endof the argument list in functions declared
9845 without a prototype.
9846 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9847 Avoid checking when arguments are null.
9849 2018-01-03 Martin Sebor <msebor@redhat.com>
9851 PR c/83559
9852 * doc/extend.texi (attribute const): Fix a typo.
9853 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9854 issuing -Wsuggest-attribute for void functions.
9856 2018-01-03 Martin Sebor <msebor@redhat.com>
9858 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9859 offset_int::from instead of wide_int::to_shwi.
9860 (maybe_diag_overlap): Remove assertion.
9861 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9862 * gimple-ssa-sprintf.c (format_directive): Same.
9863 (parse_directive): Same.
9864 (sprintf_dom_walker::compute_format_length): Same.
9865 (try_substitute_return_value): Same.
9867 2018-01-03 Jeff Law <law@redhat.com>
9869 PR middle-end/83654
9870 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9871 non-constant residual for zero at runtime and avoid probing in
9872 that case. Reorganize code for trailing problem to mirror handling
9873 of the residual.
9875 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9877 PR tree-optimization/83501
9878 * tree-ssa-strlen.c (get_string_cst): New.
9879 (handle_char_store): Call get_string_cst.
9881 2018-01-03 Martin Liska <mliska@suse.cz>
9883 PR tree-optimization/83593
9884 * tree-ssa-strlen.c: Include tree-cfg.h.
9885 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9886 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9887 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9888 to false.
9889 (strlen_dom_walker::before_dom_children): Call
9890 gimple_purge_dead_eh_edges. Dump tranformation with details
9891 dump flags.
9892 (strlen_dom_walker::before_dom_children): Update call by adding
9893 new argument cleanup_eh.
9894 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9896 2018-01-03 Martin Liska <mliska@suse.cz>
9898 PR ipa/83549
9899 * cif-code.def (VARIADIC_THUNK): New enum value.
9900 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9901 thunks.
9903 2018-01-03 Jan Beulich <jbeulich@suse.com>
9905 * sse.md (mov<mode>_internal): Tighten condition for when to use
9906 vmovdqu<ssescalarsize> for TI and OI modes.
9908 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9910 Update copyright years.
9912 2018-01-03 Martin Liska <mliska@suse.cz>
9914 PR ipa/83594
9915 * ipa-visibility.c (function_and_variable_visibility): Skip
9916 functions with noipa attribure.
9918 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9920 * gcc.c (process_command): Update copyright notice dates.
9921 * gcov-dump.c (print_version): Ditto.
9922 * gcov.c (print_version): Ditto.
9923 * gcov-tool.c (print_version): Ditto.
9924 * gengtype.c (create_file): Ditto.
9925 * doc/cpp.texi: Bump @copying's copyright year.
9926 * doc/cppinternals.texi: Ditto.
9927 * doc/gcc.texi: Ditto.
9928 * doc/gccint.texi: Ditto.
9929 * doc/gcov.texi: Ditto.
9930 * doc/install.texi: Ditto.
9931 * doc/invoke.texi: Ditto.
9933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9935 * vector-builder.h (vector_builder::m_full_nelts): Change from
9936 unsigned int to poly_uint64.
9937 (vector_builder::full_nelts): Update prototype accordingly.
9938 (vector_builder::new_vector): Likewise.
9939 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
9940 (vector_builder::operator ==): Likewise.
9941 (vector_builder::finalize): Likewise.
9942 * int-vector-builder.h (int_vector_builder::int_vector_builder):
9943 Take the number of elements as a poly_uint64 rather than an
9944 unsigned int.
9945 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
9946 from unsigned int to poly_uint64.
9947 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
9948 (vec_perm_indices::new_vector): Likewise.
9949 (vec_perm_indices::length): Likewise.
9950 (vec_perm_indices::nelts_per_input): Likewise.
9951 (vec_perm_indices::input_nelts): Likewise.
9952 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
9953 number of elements per input as a poly_uint64 rather than an
9954 unsigned int. Use the original encoding for variable-length
9955 vectors, rather than clamping each individual element.
9956 For the second and subsequent elements in each pattern,
9957 clamp the step and base before clamping their sum.
9958 (vec_perm_indices::series_p): Handle polynomial element counts.
9959 (vec_perm_indices::all_in_range_p): Likewise.
9960 (vec_perm_indices_to_tree): Likewise.
9961 (vec_perm_indices_to_rtx): Likewise.
9962 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
9963 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
9964 (tree_vector_builder::new_binary_operation): Handle polynomial
9965 element counts. Return false if we need to know the number
9966 of elements at compile time.
9967 * fold-const.c (fold_vec_perm): Punt if the number of elements
9968 isn't known at compile time.
9970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9972 * vec-perm-indices.h (vec_perm_builder): Change element type
9973 from HOST_WIDE_INT to poly_int64.
9974 (vec_perm_indices::element_type): Update accordingly.
9975 (vec_perm_indices::clamp): Handle polynomial element_types.
9976 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9977 (vec_perm_indices::all_in_range_p): Likewise.
9978 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
9979 than shwi trees.
9980 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
9981 polynomial vec_perm_indices element types.
9982 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
9983 * fold-const.c (fold_vec_perm): Likewise.
9984 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
9985 * tree-vect-generic.c (lower_vec_perm): Likewise.
9986 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9987 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
9988 element type to HOST_WIDE_INT.
9990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9991 Alan Hayward <alan.hayward@arm.com>
9992 David Sherwood <david.sherwood@arm.com>
9994 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
9995 rather than an int. Use plus_constant.
9996 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
9997 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
9999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10000 Alan Hayward <alan.hayward@arm.com>
10001 David Sherwood <david.sherwood@arm.com>
10003 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10004 a HOST_WIDE_INT to a poly_int64.
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 * calls.c (load_register_parameters): Cope with polynomial
10011 mode sizes. Require a constant size for BLKmode parameters
10012 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10013 forces a parameter to be padded at the lsb end in order to
10014 fill a complete number of words, require the parameter size
10015 to be ordered wrt UNITS_PER_WORD.
10017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10018 Alan Hayward <alan.hayward@arm.com>
10019 David Sherwood <david.sherwood@arm.com>
10021 * reload1.c (spill_stack_slot_width): Change element type
10022 from unsigned int to poly_uint64_pod.
10023 (alter_reg): Treat mode sizes as polynomial.
10025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10026 Alan Hayward <alan.hayward@arm.com>
10027 David Sherwood <david.sherwood@arm.com>
10029 * reload.c (complex_word_subreg_p): New function.
10030 (reload_inner_reg_of_subreg, push_reload): Use it.
10032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10033 Alan Hayward <alan.hayward@arm.com>
10034 David Sherwood <david.sherwood@arm.com>
10036 * lra-constraints.c (process_alt_operands): Reject matched
10037 operands whose sizes aren't ordered.
10038 (match_reload): Refer to this check here.
10040 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10041 Alan Hayward <alan.hayward@arm.com>
10042 David Sherwood <david.sherwood@arm.com>
10044 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10045 that the mode size is in the set {1, 2, 4, 8, 16}.
10047 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10048 Alan Hayward <alan.hayward@arm.com>
10049 David Sherwood <david.sherwood@arm.com>
10051 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10052 Use plus_constant instead of gen_rtx_PLUS.
10054 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10055 Alan Hayward <alan.hayward@arm.com>
10056 David Sherwood <david.sherwood@arm.com>
10058 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10059 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10060 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10061 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10062 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10063 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10064 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10065 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10066 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10067 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10068 a poly_int64.
10069 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10070 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10071 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10072 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10073 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10074 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10075 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10076 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10077 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10078 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10079 function.
10080 * expr.c (emit_move_resolve_push): Treat the input and result
10081 of PUSH_ROUNDING as a poly_int64.
10082 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10083 (emit_push_insn): Likewise.
10084 * lra-eliminations.c (mark_not_eliminable): Likewise.
10085 * recog.c (push_operand): Likewise.
10086 * reload1.c (elimination_effects): Likewise.
10087 * rtlanal.c (nonzero_bits1): Likewise.
10088 * calls.c (store_one_arg): Likewise. Require the padding to be
10089 known at compile time.
10091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10092 Alan Hayward <alan.hayward@arm.com>
10093 David Sherwood <david.sherwood@arm.com>
10095 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10096 Use plus_constant instead of gen_rtx_PLUS.
10098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10099 Alan Hayward <alan.hayward@arm.com>
10100 David Sherwood <david.sherwood@arm.com>
10102 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10103 rather than an int.
10105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10106 Alan Hayward <alan.hayward@arm.com>
10107 David Sherwood <david.sherwood@arm.com>
10109 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10110 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10111 via stack temporaries. Treat the mode size as polynomial too.
10113 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10114 Alan Hayward <alan.hayward@arm.com>
10115 David Sherwood <david.sherwood@arm.com>
10117 * expr.c (expand_expr_real_2): When handling conversions involving
10118 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10119 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10120 as a poly_uint64 too.
10122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10123 Alan Hayward <alan.hayward@arm.com>
10124 David Sherwood <david.sherwood@arm.com>
10126 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10129 Alan Hayward <alan.hayward@arm.com>
10130 David Sherwood <david.sherwood@arm.com>
10132 * combine.c (can_change_dest_mode): Handle polynomial
10133 REGMODE_NATURAL_SIZE.
10134 * expmed.c (store_bit_field_1): Likewise.
10135 * expr.c (store_constructor): Likewise.
10136 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10137 and polynomial REGMODE_NATURAL_SIZE.
10138 (gen_lowpart_common): Likewise.
10139 * reginfo.c (record_subregs_of_mode): Likewise.
10140 * rtlanal.c (read_modify_subreg_p): Likewise.
10142 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10143 Alan Hayward <alan.hayward@arm.com>
10144 David Sherwood <david.sherwood@arm.com>
10146 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10147 numbers of elements.
10149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10150 Alan Hayward <alan.hayward@arm.com>
10151 David Sherwood <david.sherwood@arm.com>
10153 * match.pd: Cope with polynomial numbers of vector elements.
10155 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10156 Alan Hayward <alan.hayward@arm.com>
10157 David Sherwood <david.sherwood@arm.com>
10159 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10160 in a POINTER_PLUS_EXPR.
10162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10163 Alan Hayward <alan.hayward@arm.com>
10164 David Sherwood <david.sherwood@arm.com>
10166 * omp-simd-clone.c (simd_clone_subparts): New function.
10167 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10168 (ipa_simd_modify_function_body): Likewise.
10170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10171 Alan Hayward <alan.hayward@arm.com>
10172 David Sherwood <david.sherwood@arm.com>
10174 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10175 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10176 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10177 (expand_vector_condition, vector_element): Likewise.
10178 (subparts_gt): New function.
10179 (get_compute_type): Use subparts_gt.
10180 (count_type_subparts): Delete.
10181 (expand_vector_operations_1): Use subparts_gt instead of
10182 count_type_subparts.
10184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10185 Alan Hayward <alan.hayward@arm.com>
10186 David Sherwood <david.sherwood@arm.com>
10188 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10189 (vect_compile_time_alias): ...this new function. Do the calculation
10190 on poly_ints rather than trees.
10191 (vect_prune_runtime_alias_test_list): Update call accordingly.
10193 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10194 Alan Hayward <alan.hayward@arm.com>
10195 David Sherwood <david.sherwood@arm.com>
10197 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10198 numbers of units.
10199 (vect_schedule_slp_instance): Likewise.
10201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10202 Alan Hayward <alan.hayward@arm.com>
10203 David Sherwood <david.sherwood@arm.com>
10205 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10206 constant and extern definitions for variable-length vectors.
10207 (vect_get_constant_vectors): Note that the number of units
10208 is known to be constant.
10210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10211 Alan Hayward <alan.hayward@arm.com>
10212 David Sherwood <david.sherwood@arm.com>
10214 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10215 of units as polynomial. Choose between WIDE and NARROW based
10216 on multiple_p.
10218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10219 Alan Hayward <alan.hayward@arm.com>
10220 David Sherwood <david.sherwood@arm.com>
10222 * tree-vect-stmts.c (simd_clone_subparts): New function.
10223 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10226 Alan Hayward <alan.hayward@arm.com>
10227 David Sherwood <david.sherwood@arm.com>
10229 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10230 vectors as polynomial. Use build_index_vector for
10231 IFN_GOMP_SIMD_LANE.
10233 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10234 Alan Hayward <alan.hayward@arm.com>
10235 David Sherwood <david.sherwood@arm.com>
10237 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10238 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10239 for variable-length vectors.
10240 (vectorizable_mask_load_store): Treat the number of units as
10241 polynomial, asserting that it is constant if the condition has
10242 already been enforced.
10243 (vectorizable_store, vectorizable_load): Likewise.
10245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10246 Alan Hayward <alan.hayward@arm.com>
10247 David Sherwood <david.sherwood@arm.com>
10249 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10250 of units as polynomial. Punt if we can't tell at compile time
10251 which vector contains the final result.
10253 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10254 Alan Hayward <alan.hayward@arm.com>
10255 David Sherwood <david.sherwood@arm.com>
10257 * tree-vect-loop.c (vectorizable_induction): Treat the number
10258 of units as polynomial. Punt on SLP inductions. Use an integer
10259 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10260 cast of such a series for variable-length floating-point
10261 reductions.
10263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10264 Alan Hayward <alan.hayward@arm.com>
10265 David Sherwood <david.sherwood@arm.com>
10267 * tree.h (build_index_vector): Declare.
10268 * tree.c (build_index_vector): New function.
10269 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10270 of units as polynomial, forcibly converting it to a constant if
10271 vectorizable_reduction has already enforced the condition.
10272 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10273 to create a {1,2,3,...} vector.
10274 (vectorizable_reduction): Treat the number of units as polynomial.
10275 Choose vectype_in based on the largest scalar element size rather
10276 than the smallest number of units. Enforce the restrictions
10277 relied on above.
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-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10284 number of units as polynomial.
10286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10287 Alan Hayward <alan.hayward@arm.com>
10288 David Sherwood <david.sherwood@arm.com>
10290 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10291 * target.def (autovectorize_vector_sizes): Return the vector sizes
10292 by pointer, using vector_sizes rather than a bitmask.
10293 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10294 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10295 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10296 Likewise.
10297 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10298 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10299 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10300 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10301 * omp-general.c (omp_max_vf): Likewise.
10302 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10303 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10304 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10305 * tree-vect-slp.c (vect_slp_bb): Likewise.
10306 * doc/tm.texi: Regenerate.
10307 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10308 to a poly_uint64.
10309 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10310 the vector size as a poly_uint64 rather than an unsigned int.
10311 (current_vector_size): Change from an unsigned int to a poly_uint64.
10312 (get_vectype_for_scalar_type): Update accordingly.
10313 * tree.h (build_truth_vector_type): Take the size and number of
10314 units as a poly_uint64 rather than an unsigned int.
10315 (build_vector_type): Add a temporary overload that takes
10316 the number of units as a poly_uint64 rather than an unsigned int.
10317 * tree.c (make_vector_type): Likewise.
10318 (build_truth_vector_type): Take the number of units as a poly_uint64
10319 rather than an unsigned int.
10321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10322 Alan Hayward <alan.hayward@arm.com>
10323 David Sherwood <david.sherwood@arm.com>
10325 * target.def (get_mask_mode): Take the number of units and length
10326 as poly_uint64s rather than unsigned ints.
10327 * targhooks.h (default_get_mask_mode): Update accordingly.
10328 * targhooks.c (default_get_mask_mode): Likewise.
10329 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10330 * doc/tm.texi: Regenerate.
10332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10333 Alan Hayward <alan.hayward@arm.com>
10334 David Sherwood <david.sherwood@arm.com>
10336 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10337 * omp-general.c (omp_max_vf): Likewise.
10338 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10339 (expand_omp_simd): Handle polynomial safelen.
10340 * omp-low.c (omplow_simd_context): Add a default constructor.
10341 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10342 (lower_rec_simd_input_clauses): Update accordingly.
10343 (lower_rec_input_clauses): Likewise.
10345 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10346 Alan Hayward <alan.hayward@arm.com>
10347 David Sherwood <david.sherwood@arm.com>
10349 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10350 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10351 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10352 (vect_analyze_slp_cost): Likewise.
10353 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10354 (vect_model_load_cost): Likewise.
10356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10357 Alan Hayward <alan.hayward@arm.com>
10358 David Sherwood <david.sherwood@arm.com>
10360 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10361 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10362 from an unsigned int * to a poly_uint64_pod *.
10363 (calculate_unrolling_factor): New function.
10364 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10367 Alan Hayward <alan.hayward@arm.com>
10368 David Sherwood <david.sherwood@arm.com>
10370 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10371 from an unsigned int to a poly_uint64.
10372 (_loop_vec_info::slp_unrolling_factor): Likewise.
10373 (_loop_vec_info::vectorization_factor): Change from an int
10374 to a poly_uint64.
10375 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10376 (vect_get_num_vectors): New function.
10377 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10378 (vect_get_num_copies): Use vect_get_num_vectors.
10379 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10380 to an unsigned int *.
10381 (vect_analyze_data_refs): Change min_vf from an int * to a
10382 poly_uint64 *.
10383 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10384 than an unsigned HOST_WIDE_INT.
10385 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10386 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10387 to an unsigned int *.
10388 (vect_analyze_data_ref_dependences): Likewise.
10389 (vect_compute_data_ref_alignment): Handle polynomial vf.
10390 (vect_enhance_data_refs_alignment): Likewise.
10391 (vect_prune_runtime_alias_test_list): Likewise.
10392 (vect_shift_permute_load_chain): Likewise.
10393 (vect_supportable_dr_alignment): Likewise.
10394 (dependence_distance_ge_vf): Take the vectorization factor as a
10395 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10396 (vect_analyze_data_refs): Change min_vf from an int * to a
10397 poly_uint64 *.
10398 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10399 vfm1 as a poly_uint64 rather than an int. Make the same change
10400 for the returned bound_scalar.
10401 (vect_gen_vector_loop_niters): Handle polynomial vf.
10402 (vect_do_peeling): Likewise. Update call to
10403 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10404 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10405 be constant.
10406 * tree-vect-loop.c (vect_determine_vectorization_factor)
10407 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10408 (vect_get_known_peeling_cost): Likewise.
10409 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10410 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10411 (vect_transform_loop): Likewise. Use the lowest possible VF when
10412 updating the upper bounds of the loop.
10413 (vect_min_worthwhile_factor): Make static. Return an unsigned int
10414 rather than an int.
10415 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10416 polynomial unroll factors.
10417 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10418 (vect_make_slp_decision): Likewise.
10419 (vect_supported_load_permutation_p): Likewise, and polynomial
10420 vf too.
10421 (vect_analyze_slp_cost): Handle polynomial vf.
10422 (vect_slp_analyze_node_operations): Likewise.
10423 (vect_slp_analyze_bb_1): Likewise.
10424 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10425 than an unsigned HOST_WIDE_INT.
10426 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10427 (vectorizable_load): Handle polynomial vf.
10428 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10429 a poly_uint64.
10430 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10432 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10433 Alan Hayward <alan.hayward@arm.com>
10434 David Sherwood <david.sherwood@arm.com>
10436 * match.pd: Handle bit operations involving three constants
10437 and try to fold one pair.
10439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10441 * tree-vect-loop-manip.c: Include gimple-fold.h.
10442 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10443 niters_maybe_zero parameters. Handle other cases besides a step of 1.
10444 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10445 Add a path that uses a step of VF instead of 1, but disable it
10446 for now.
10447 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
10448 and niters_no_overflow parameters. Update calls to
10449 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
10450 Create a new SSA name if the latter choses to use a ste other
10451 than zero, and return it via niters_vector_mult_vf_var.
10452 * tree-vect-loop.c (vect_transform_loop): Update calls to
10453 vect_do_peeling, vect_gen_vector_loop_niters and
10454 slpeel_make_loop_iterate_ntimes.
10455 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
10456 (vect_gen_vector_loop_niters): Update declarations after above changes.
10458 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
10460 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
10461 128-bit round to integer instructions.
10462 (ceil<mode>2): Likewise.
10463 (btrunc<mode>2): Likewise.
10464 (round<mode>2): Likewise.
10466 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10468 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
10469 unaligned VSX load/store on P8/P9.
10470 (expand_block_clear): Allow the use of unaligned VSX
10471 load/store on P8/P9.
10473 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10475 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
10476 New function.
10477 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
10478 swap associated with both a load and a store.
10480 2018-01-02 Andrew Waterman <andrew@sifive.com>
10482 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10483 * config/riscv/riscv.md (clear_cache): Use it.
10485 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
10487 * web.c: Remove out-of-date comment.
10489 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10491 * expr.c (fixup_args_size_notes): Check that any existing
10492 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10493 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10494 (emit_single_push_insn): ...here.
10496 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10498 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10499 (const_vector_encoded_nelts): New function.
10500 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10501 (const_vector_int_elt, const_vector_elt): Declare.
10502 * emit-rtl.c (const_vector_int_elt_1): New function.
10503 (const_vector_elt): Likewise.
10504 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10505 of CONST_VECTOR_ELT.
10507 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10509 * expr.c: Include rtx-vector-builder.h.
10510 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10511 directly on the tree encoding.
10512 (const_vector_from_tree): Likewise.
10513 * optabs.c: Include rtx-vector-builder.h.
10514 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10515 sequence of "u" values.
10516 * vec-perm-indices.c: Include rtx-vector-builder.h.
10517 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10518 directly on the vec_perm_indices encoding.
10520 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10522 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10523 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10524 * rtx-vector-builder.h: New file.
10525 * rtx-vector-builder.c: Likewise.
10526 * rtl.h (rtx_def::u2): Add a const_vector field.
10527 (CONST_VECTOR_NPATTERNS): New macro.
10528 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10529 (CONST_VECTOR_DUPLICATE_P): Likewise.
10530 (CONST_VECTOR_STEPPED_P): Likewise.
10531 (CONST_VECTOR_ENCODED_ELT): Likewise.
10532 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10533 (unwrap_const_vec_duplicate): Likewise.
10534 (const_vec_series_p): Check for a non-duplicated vector encoding.
10535 Say that the function only returns true for integer vectors.
10536 * emit-rtl.c: Include rtx-vector-builder.h.
10537 (gen_const_vec_duplicate_1): Delete.
10538 (gen_const_vector): Call gen_const_vec_duplicate instead of
10539 gen_const_vec_duplicate_1.
10540 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10541 (gen_const_vec_duplicate): Use rtx_vector_builder.
10542 (gen_const_vec_series): Likewise.
10543 (gen_rtx_CONST_VECTOR): Likewise.
10544 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10545 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10546 Build a new vector rather than modifying a CONST_VECTOR in-place.
10547 (handle_special_swappables): Update call accordingly.
10548 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10549 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10550 Build a new vector rather than modifying a CONST_VECTOR in-place.
10551 (handle_special_swappables): Update call accordingly.
10553 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10555 * simplify-rtx.c (simplify_const_binary_operation): Use
10556 CONST_VECTOR_ELT instead of XVECEXP.
10558 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10560 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10561 the selector elements to be different from the data elements
10562 if the selector is a VECTOR_CST.
10563 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10564 ssizetype for the selector.
10566 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10568 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10569 before testing each element individually.
10570 * tree-vect-generic.c (lower_vec_perm): Likewise.
10572 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10574 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10575 * selftest-run-tests.c (selftest::run_tests): Call it.
10576 * vector-builder.h (vector_builder::operator ==): New function.
10577 (vector_builder::operator !=): Likewise.
10578 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10579 (vec_perm_indices::all_from_input_p): New function.
10580 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10581 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10582 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10583 instead of reading the VECTOR_CST directly. Detect whether both
10584 vector inputs are the same before constructing the vec_perm_indices,
10585 and update the number of inputs argument accordingly. Use the
10586 utility functions added above. Only construct sel2 if we need to.
10588 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10590 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10591 the broadcast of the low byte.
10592 (expand_mult_highpart): Use an explicit encoding for the permutes.
10593 * optabs-query.c (can_mult_highpart_p): Likewise.
10594 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10595 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10596 (vectorizable_bswap): Likewise.
10597 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10598 explicit encoding for the power-of-2 permutes.
10599 (vect_permute_store_chain): Likewise.
10600 (vect_grouped_load_supported): Likewise.
10601 (vect_permute_load_chain): Likewise.
10603 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10605 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10606 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10607 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10608 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10609 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10610 (vect_gen_perm_mask_any): Likewise.
10612 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10614 * int-vector-builder.h: New file.
10615 * vec-perm-indices.h: Include int-vector-builder.h.
10616 (vec_perm_indices): Redefine as an int_vector_builder.
10617 (auto_vec_perm_indices): Delete.
10618 (vec_perm_builder): Redefine as a stand-alone class.
10619 (vec_perm_indices::vec_perm_indices): New function.
10620 (vec_perm_indices::clamp): Likewise.
10621 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10622 (vec_perm_indices::new_vector): New function.
10623 (vec_perm_indices::new_expanded_vector): Update for new
10624 vec_perm_indices class.
10625 (vec_perm_indices::rotate_inputs): New function.
10626 (vec_perm_indices::all_in_range_p): Operate directly on the
10627 encoded form, without computing elided elements.
10628 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10629 encoding. Update for new vec_perm_indices class.
10630 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10631 the given vec_perm_builder.
10632 (expand_vec_perm_var): Update vec_perm_builder constructor.
10633 (expand_mult_highpart): Use vec_perm_builder instead of
10634 auto_vec_perm_indices.
10635 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10636 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10637 or double series encoding as appropriate.
10638 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10639 vec_perm_indices instead of auto_vec_perm_indices.
10640 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10641 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10642 (vect_permute_store_chain): Likewise.
10643 (vect_grouped_load_supported): Likewise.
10644 (vect_permute_load_chain): Likewise.
10645 (vect_shift_permute_load_chain): Likewise.
10646 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10647 (vect_transform_slp_perm_load): Likewise.
10648 (vect_schedule_slp_instance): Likewise.
10649 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10650 (vectorizable_mask_load_store): Likewise.
10651 (vectorizable_bswap): Likewise.
10652 (vectorizable_store): Likewise.
10653 (vectorizable_load): Likewise.
10654 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10655 vec_perm_indices instead of auto_vec_perm_indices. Use
10656 tree_to_vec_perm_builder to read the vector from a tree.
10657 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10658 vec_perm_builder instead of a vec_perm_indices.
10659 (have_whole_vector_shift): Use vec_perm_builder and
10660 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10661 truncation to calc_vec_perm_mask_for_shift.
10662 (vect_create_epilog_for_reduction): Likewise.
10663 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10664 from auto_vec_perm_indices to vec_perm_indices.
10665 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10666 instead of changing individual elements.
10667 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10668 the vector in d.perm.
10669 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10670 from auto_vec_perm_indices to vec_perm_indices.
10671 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10672 instead of changing individual elements.
10673 (arm_vectorize_vec_perm_const): Use new_vector to install
10674 the vector in d.perm.
10675 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10676 Update vec_perm_builder constructor.
10677 (rs6000_expand_interleave): Likewise.
10678 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10679 (rs6000_expand_interleave): Likewise.
10681 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10683 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10684 to qimode could truncate the indices.
10685 * optabs.c (expand_vec_perm_var): Likewise.
10687 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10689 * Makefile.in (OBJS): Add vec-perm-indices.o.
10690 * vec-perm-indices.h: New file.
10691 * vec-perm-indices.c: Likewise.
10692 * target.h (vec_perm_indices): Replace with a forward class
10693 declaration.
10694 (auto_vec_perm_indices): Move to vec-perm-indices.h.
10695 * optabs.h: Include vec-perm-indices.h.
10696 (expand_vec_perm): Delete.
10697 (selector_fits_mode_p, expand_vec_perm_var): Declare.
10698 (expand_vec_perm_const): Declare.
10699 * target.def (vec_perm_const_ok): Replace with...
10700 (vec_perm_const): ...this new hook.
10701 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10702 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10703 * doc/tm.texi: Regenerate.
10704 * optabs.def (vec_perm_const): Delete.
10705 * doc/md.texi (vec_perm_const): Likewise.
10706 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10707 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10708 expand_vec_perm for constant permutation vectors. Assert that
10709 the mode of variable permutation vectors is the integer equivalent
10710 of the mode that is being permuted.
10711 * optabs-query.h (selector_fits_mode_p): Declare.
10712 * optabs-query.c: Include vec-perm-indices.h.
10713 (selector_fits_mode_p): New function.
10714 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10715 is defined, instead of checking whether the vec_perm_const_optab
10716 exists. Use targetm.vectorize.vec_perm_const instead of
10717 targetm.vectorize.vec_perm_const_ok. Check whether the indices
10718 fit in the vector mode before using a variable permute.
10719 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10720 vec_perm_indices instead of an rtx.
10721 (expand_vec_perm): Replace with...
10722 (expand_vec_perm_const): ...this new function. Take the selector
10723 as a vec_perm_indices rather than an rtx. Also take the mode of
10724 the selector. Update call to shift_amt_for_vec_perm_mask.
10725 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10726 Use vec_perm_indices::new_expanded_vector to expand the original
10727 selector into bytes. Check whether the indices fit in the vector
10728 mode before using a variable permute.
10729 (expand_vec_perm_var): Make global.
10730 (expand_mult_highpart): Use expand_vec_perm_const.
10731 * fold-const.c: Includes vec-perm-indices.h.
10732 * tree-ssa-forwprop.c: Likewise.
10733 * tree-vect-data-refs.c: Likewise.
10734 * tree-vect-generic.c: Likewise.
10735 * tree-vect-loop.c: Likewise.
10736 * tree-vect-slp.c: Likewise.
10737 * tree-vect-stmts.c: Likewise.
10738 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
10739 Delete.
10740 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
10741 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
10742 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
10743 (aarch64_vectorize_vec_perm_const): ...this new function.
10744 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10745 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10746 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
10747 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
10748 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10749 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10750 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
10751 into...
10752 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
10753 check for NEON modes.
10754 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
10755 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
10756 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
10757 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
10758 into...
10759 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
10760 the old VEC_PERM_CONST conditions.
10761 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
10762 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
10763 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
10764 (ia64_vectorize_vec_perm_const_ok): Merge into...
10765 (ia64_vectorize_vec_perm_const): ...this new function.
10766 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
10767 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
10768 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
10769 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
10770 * config/mips/mips.c (mips_expand_vec_perm_const)
10771 (mips_vectorize_vec_perm_const_ok): Merge into...
10772 (mips_vectorize_vec_perm_const): ...this new function.
10773 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
10774 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
10775 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
10776 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
10777 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
10778 (rs6000_expand_vec_perm_const): Delete.
10779 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
10780 Delete.
10781 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10782 (altivec_expand_vec_perm_const_le): Take each operand individually.
10783 Operate on constant selectors rather than rtxes.
10784 (altivec_expand_vec_perm_const): Likewise. Update call to
10785 altivec_expand_vec_perm_const_le.
10786 (rs6000_expand_vec_perm_const): Delete.
10787 (rs6000_vectorize_vec_perm_const_ok): Delete.
10788 (rs6000_vectorize_vec_perm_const): New function.
10789 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10790 an element count and rtx array.
10791 (rs6000_expand_extract_even): Update call accordingly.
10792 (rs6000_expand_interleave): Likewise.
10793 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10794 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10795 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10796 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10797 (rs6000_expand_vec_perm_const): Delete.
10798 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10799 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10800 (altivec_expand_vec_perm_const_le): Take each operand individually.
10801 Operate on constant selectors rather than rtxes.
10802 (altivec_expand_vec_perm_const): Likewise. Update call to
10803 altivec_expand_vec_perm_const_le.
10804 (rs6000_expand_vec_perm_const): Delete.
10805 (rs6000_vectorize_vec_perm_const_ok): Delete.
10806 (rs6000_vectorize_vec_perm_const): New function. Remove stray
10807 reference to the SPE evmerge intructions.
10808 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10809 an element count and rtx array.
10810 (rs6000_expand_extract_even): Update call accordingly.
10811 (rs6000_expand_interleave): Likewise.
10812 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10813 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10814 new function.
10815 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10817 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10819 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10820 vector mode and that that mode matches the mode of the data
10821 being permuted.
10822 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10823 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
10824 directly using expand_vec_perm_1 when forcing selectors into
10825 registers.
10826 (expand_vec_perm_var): New function, split out from expand_vec_perm.
10828 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10830 * optabs-query.h (can_vec_perm_p): Delete.
10831 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10832 * optabs-query.c (can_vec_perm_p): Split into...
10833 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10834 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10835 particular selector is valid.
10836 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10837 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10838 (vect_grouped_load_supported): Likewise.
10839 (vect_shift_permute_load_chain): Likewise.
10840 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10841 (vect_transform_slp_perm_load): Likewise.
10842 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10843 (vectorizable_bswap): Likewise.
10844 (vect_gen_perm_mask_checked): Likewise.
10845 * fold-const.c (fold_ternary_loc): Likewise. Don't take
10846 implementations of variable permutation vectors into account
10847 when deciding which selector to use.
10848 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10849 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10850 with a false third argument.
10851 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10852 to test whether the constant selector is valid and can_vec_perm_var_p
10853 to test whether a variable selector is valid.
10855 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10857 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10858 * optabs-query.c (can_vec_perm_p): Likewise.
10859 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10860 instead of vec_perm_indices.
10861 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10862 (vect_gen_perm_mask_checked): Likewise,
10863 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10864 (vect_gen_perm_mask_checked): Likewise,
10866 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10868 * optabs-query.h (qimode_for_vec_perm): Declare.
10869 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10870 (qimode_for_vec_perm): ...this new function.
10871 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10873 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10875 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10876 does not have a conditional at the top.
10878 2018-01-02 Richard Biener <rguenther@suse.de>
10880 * ipa-inline.c (big_speedup_p): Fix expression.
10882 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10884 PR target/81616
10885 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10886 for generic 4->6.
10888 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10890 PR target/81616
10891 Generic tuning.
10892 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10893 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10894 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10895 cond_taken_branch_cost 3->4.
10897 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10899 PR tree-optimization/83581
10900 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10901 TODO_cleanup_cfg if any changes have been made.
10903 PR middle-end/83608
10904 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10905 convert_modes if target mode has the right side, but different mode
10906 class.
10908 PR middle-end/83609
10909 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10910 last argument when extracting from CONCAT. If either from_real or
10911 from_imag is NULL, use expansion through memory. If result is not
10912 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10913 the parts directly to inner mode, if even that fails, use expansion
10914 through memory.
10916 PR middle-end/83623
10917 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10918 check for bswap in mode rather than HImode and use that in expand_unop
10919 too.
10920 \f
10921 Copyright (C) 2018 Free Software Foundation, Inc.
10923 Copying and distribution of this file, with or without modification,
10924 are permitted in any medium without royalty provided the copyright
10925 notice and this notice are preserved.