| blob | 47862a80a0dc62d8337a073fd56f3ca9b2f52448 |
1 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
3 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
4 To improve optimization opportunities.
5 * builtin-types.def: The new needed builtin types for the above.
7 2018-05-04 Richard Biener <rguenther@suse.de>
9 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
10 * gimple-ssa-store-merging.c
11 (imm_store_chain_info::output_merged_store): Remove redundant create,
12 release split_store vector contents on failure.
13 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
14 scalar stmt vector on cache hit.
16 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
18 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
19 Xilinx FP support.
20 * config.gcc (powerpc-xilinx-eabi*): Remove.
21 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
22 support.
23 (fusion_addis_mem_combo_load): Ditto.
24 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
25 FP support.
26 (rs6000_cpu_cpp_builtins): Ditto.
27 * config/rs6000/rs6000-linux.c
28 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
29 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
30 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
31 support.
32 (rs6000_setup_reg_addr_masks): Ditto.
33 (rs6000_init_hard_regno_mode_ok): Ditto.
34 (rs6000_option_override_internal): Ditto.
35 (legitimate_lo_sum_address_p): Ditto.
36 (rs6000_legitimize_address): Ditto.
37 (rs6000_legitimize_reload_address): Ditto.
38 (rs6000_legitimate_address_p): Ditto.
39 (abi_v4_pass_in_fpr): Ditto.
40 (setup_incoming_varargs): Ditto.
41 (rs6000_gimplify_va_arg): Ditto.
42 (rs6000_split_multireg_move): Ditto.
43 (rs6000_savres_strategy): Ditto.
44 (rs6000_emit_prologue_components): Ditto.
45 (rs6000_emit_epilogue_components): Ditto.
46 (rs6000_emit_prologue): Ditto.
47 (rs6000_emit_epilogue): Ditto.
48 (rs6000_elf_file_end): Ditto.
49 (rs6000_function_value): Ditto.
50 (rs6000_libcall_value): Ditto.
51 * config/rs6000/rs6000.h: Ditto.
52 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
53 (TARGET_MINMAX): ... this. New.
54 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
55 * config/rs6000/rs6000.md: Remove Xilinx FP support.
56 (*movsi_internal1_single): Delete.
57 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
58 mfpu=, mxilinx-fpu): Delete.
59 * config/rs6000/singlefp.h: Delete.
60 * config/rs6000/sysv4.h: Remove Xilinx FP support.
61 * config/rs6000/t-rs6000: Ditto.
62 * config/rs6000/t-xilinx: Delete.
63 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
64 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
65 (VStype_simple): Delete.
66 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
67 * config/rs6000/xfpu.h: Delete.
68 * config/rs6000/xfpu.md: Delete.
69 * config/rs6000/xilinx.h: Delete.
70 * config/rs6000/xilinx.opt: Delete.
71 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
72 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
74 2018-05-04 Tom de Vries <tom@codesourcery.com>
76 PR libgomp/85639
77 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
78 if ignore == 0.
80 2018-05-04 Richard Biener <rguenther@suse.de>
82 PR middle-end/85627
83 * tree-complex.c (update_complex_assignment): We are always in SSA form.
84 (expand_complex_div_wide): Likewise.
85 (expand_complex_operations_1): Likewise.
86 (expand_complex_libcall): Preserve EH info of the original stmt.
87 (tree_lower_complex): Handle removed blocks.
88 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
89 on complex multiplication and division libcall builtins.
91 2018-05-04 Richard Biener <rguenther@suse.de>
93 PR middle-end/85574
94 * fold-const.c (negate_expr_p): Restrict negation of operand
95 zero of a division to when we know that can happen without
96 overflow.
97 (fold_negate_expr_1): Likewise.
99 2018-05-04 Jakub Jelinek <jakub@redhat.com>
101 PR libstdc++/85466
102 * real.h (real_nextafter): Declare.
103 * real.c (real_nextafter): New function.
104 * fold-const-call.c (fold_const_nextafter): New function.
105 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
106 CASE_CFN_NEXTTOWARD.
107 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
108 even when arg1_mode is different from arg0_mode.
110 2018-05-03 Nathan Sidwell <nathan@acm.org>
112 * doc/extend.texi (Deprecated Features): Remove
113 -ffriend-injection.
114 (Backwards Compatibility): Likewise.
115 * doc/invoke.texi (C++ Language Options): Likewise.
116 (C++ Dialect Options): Likewise.
118 2018-05-03 Jakub Jelinek <jakub@redhat.com>
120 PR target/85530
121 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
122 _mm512_mask_mullox_epi64): New intrinsics.
124 2018-05-03 Tom de Vries <tom@codesourcery.com>
126 PR testsuite/85106
127 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
128 dump files): Add offload-tree.
130 2018-05-03 Richard Biener <rguenther@suse.de>
132 PR tree-optimization/85615
133 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
134 to loops not nested in BBs loop father to avoid creating multi-entry
135 loops.
137 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
139 PR tree-optimization/70291
140 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
141 arguments. Change return type to tree. Emit libcall as a new
142 statement rather than replacing existing one when inplace_p is true.
143 (expand_complex_multiplication_components): New function.
144 (expand_complex_multiplication): Expand floating-point complex
145 multiplication using the above.
146 (expand_complex_division): Rename inner_type parameter to type.
147 Update expand_complex_libcall call-site.
148 (expand_complex_operations_1): Update expand_complex_multiplication
149 and expand_complex_division call-sites.
151 2018-05-02 Jakub Jelinek <jakub@redhat.com>
153 PR target/85582
154 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
155 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
156 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
157 the highest significant bit of the shift count mask is clear. In
158 check whether and[sq]i3 is needed verify that all significant bits
159 of the shift count other than the highest are set.
161 2018-05-02 Tom de Vries <tom@codesourcery.com>
163 PR libgomp/82428
164 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
165 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
166 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
167 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
168 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
169 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
170 __builtin_goacc_parlevel_size.
172 2018-05-02 Richard Biener <rguenther@suse.de>
174 PR tree-optimization/85597
175 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
176 do not use split vect_get_vec_defs call but call vect_get_slp_defs
177 directly.
179 2018-05-02 Tom de Vries <tom@codesourcery.com>
181 PR testsuite/85106
182 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
183 dump files): Add ltrans-tree.
185 2018-05-02 Tom de Vries <tom@codesourcery.com>
187 PR testsuite/85106
188 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
189 dump files): Add wpa-ipa.
191 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
193 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
194 powerpc*-*-linux*paired* target.
195 * config/rs6000/750cl.h: Delete.
196 * config/rs6000/paired.h: Delete.
197 * config/rs6000/paired.md: Delete.
198 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
199 float support.
200 * config/rs6000/rs6000-builtin.def: Remove paired float support.
201 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
202 comment. Remove paired float support.
203 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
204 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
205 VECTOR_PAIRED.
206 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
207 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
208 declarations.
209 * config/rs6000/rs6000.c: Remove paired float support.
210 (paired_expand_vector_init, paired_expand_vector_move,
211 paired_emit_vector_compare, paired_emit_vector_cond_expr,
212 (paired_expand_lv_builtin, paired_expand_stv_builtin,
213 paired_expand_builtin, paired_expand_predicate_builtin,
214 paired_init_builtins): Delete.
215 * config/rs6000/rs6000.h: Remove paired float support.
216 * config/rs6000/rs6000.md: Remove paired float support.
217 (move_from_CR_ov_bit): Delete.
218 * config/rs6000/rs6000.opt (mpaired): Delete.
219 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
220 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
222 2018-05-02 Richard Biener <rguenther@suse.de>
224 PR middle-end/85567
225 * gimplify.c (gimplify_save_expr): When in SSA form allow
226 SAVE_EXPRs to compute to SSA vars.
228 2018-05-02 Jakub Jelinek <jakub@redhat.com>
230 PR target/85582
231 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
232 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
233 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
234 clobber operands[2], instead use a new pseudo. Formatting fixes.
236 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
238 PR tree-optimization/85586
239 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
240 exit early for statements in the same group if the accesses are
241 not strided.
243 2018-05-02 Tom de Vries <tom@codesourcery.com>
245 PR lto/85451
246 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
247 error message.
249 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
251 PR tree-optimization/85143
252 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
254 2018-05-01 Tom de Vries <tom@codesourcery.com>
256 PR lto/85451
257 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
258 not found" error message.
260 2018-05-01 Tom de Vries <tom@codesourcery.com>
262 PR other/83786
263 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
264 * vec.c (test_ordered_remove_if): New function.
265 (vec_c_tests): Call test_ordered_remove_if.
266 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
267 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
268 * tree-vect-patterns.c (vect_pattern_recog_1): Use
269 VEC_ORDERED_REMOVE_IF.
271 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
273 PR tree-optimization/82665
274 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
275 pointer subtraction where arguments come from a memchr call.
277 2018-05-01 Jakub Jelinek <jakub@redhat.com>
279 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
280 --push-state --as-needed and --pop-state instead of --as-needed and
281 --no-as-needed if ld supports it.
282 * configure: Regenerated.
284 PR web/85578
285 * doc/install.texi2html: Replace _002d with - and _002a with * in
286 generated html files using sed.
288 2018-04-30 David Malcolm <dmalcolm@redhat.com>
290 PR c++/85523
291 * gcc-rich-location.c (blank_line_before_p): New function.
292 (use_new_line): New function.
293 (gcc_rich_location::add_fixit_insert_formatted): New function.
294 * gcc-rich-location.h
295 (gcc_rich_location::add_fixit_insert_formatted): New function.
297 2018-04-30 David Malcolm <dmalcolm@redhat.com>
299 * selftest.c (assert_streq): Rename "expected" and "actual" to
300 "val1" and "val2". Extend NULL-handling to cover both inputs
301 symmetrically, while still requiring both to be non-NULL for a pass.
302 * selftest.h (assert_streq): Rename "expected" and "actual" to
303 "val1" and "val2".
304 (ASSERT_EQ): Likewise.
305 (ASSERT_EQ_AT): Likewise.
306 (ASSERT_KNOWN_EQ): Likewise.
307 (ASSERT_KNOWN_EQ_AT): Likewise.
308 (ASSERT_NE): Likewise.
309 (ASSERT_MAYBE_NE): Likewise.
310 (ASSERT_MAYBE_NE_AT): Likewise.
311 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
312 the assertion to pass.
313 (ASSERT_STREQ_AT): Likewise.
315 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
317 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
318 interaction with -pie.
320 2018-04-30 David Malcolm <dmalcolm@redhat.com>
322 * selftest.h: Fix alphabetization of per-source-file selftest
323 declarations.
325 2018-04-30 Jason Merrill <jason@redhat.com>
327 PR c++/61982 - dead stores to destroyed objects.
328 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
329 of clobber.
331 2018-04-30 Jason Merrill <jason@redhat.com>
333 * tree.c (build_clobber): New.
334 * tree.h: Declare it.
335 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
337 2018-04-30 David Malcolm <dmalcolm@redhat.com>
339 * diagnostic-show-locus.c (layout::layout): Update for
340 location_get_source_line returning a char_span.
341 (struct char_span): Move to input.h.
342 (struct correction): Update for fields in char_span becoming
343 private.
344 (struct source_line): Update for location_get_source_line
345 returning a char_span.
346 (layout::print_line): Likewise.
347 * edit-context.c (edited_file::print_content): Likewise.
348 (edited_file::print_diff_hunk): Likewise.
349 (edited_file::print_run_of_changed_lines): Likewise.
350 (edited_file::get_num_lines): Likewise.
351 (edited_line::edited_line): Likewise.
352 * final.c (asm_show_source): Likewise.
353 * input.c (location_get_source_line): Convert return type
354 from const char * to char_span, losing the final "line_len"
355 param.
356 (dump_location_info): Update for the above.
357 (get_substring_ranges_for_loc): Likewise. Use a char_span
358 when handling the literal within the line.
359 (test_reading_source_line): Update for location_get_source_line
360 returning a char_span.
361 * input.h (class char_span): Move here from
362 diagnostic-show-locus.c, converting from a struct to a class.
363 Make data members private.
364 (char_span::operator bool): New.
365 (char_span::length): New.
366 (char_span::get_buffer): New.
367 (char_span::operator[]): New.
368 (char_span::subspan): Make const.
369 (char_span::xstrdup): New.
370 (location_get_source_line): Convert return type from const char *
371 to char_span, losing the final "line_size" param.
373 2018-04-30 Jan Hubicka <jh@suse.cz>
375 * lto-wrapper.c (ltrans_priorities): New static var.
376 (cmp_priority): New.
377 (run_gcc): Read priorities and if doing parallel build order
378 the Makefile by them.
380 2018-04-30 David Malcolm <dmalcolm@redhat.com>
382 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
384 2018-04-30 Richard Biener <rguenther@suse.de>
386 * tree-cfg.c (verify_address): Remove base argument, add
387 flag whether to check TREE_ADDRESSABLE and do that.
388 (verify_expr): Remove.
389 (verify_types_in_gimple_reference): Add pieces from verify_expr.
390 (verify_gimple_assign_single): Likewise.
391 (verify_gimple_switch): Likewise.
392 (verify_expr_location_1): Dereference tp once. Add (disabled)
393 piece from verify_expr.
394 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
396 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
398 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
400 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
402 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
403 (small_data_pattern): Likewise.
404 (arc_rewrite_small_data): Likewise.
405 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
406 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
407 (get_symbol_alignment): New function.
408 (legitimate_small_data_address_p): Likewise.
409 (legitimate_scaled_address): Update, call
410 legitimate_small_data_address_p.
411 (output_sdata): New static variable.
412 (arc_print_operand): Update how we handle small data operands.
413 (arc_print_operand_address): Likewise.
414 (arc_legitimate_address_p): Update, use
415 legitimate_small_data_address_p.
416 (arc_rewrite_small_data_p): Remove.
417 (arc_rewrite_small_data_1): Likewise.
418 (arc_rewrite_small_data): Likewise.
419 (small_data_pattern): Likewise.
420 (compact_sda_memory_operand): Update to use
421 legitimate_small_data_address_p and get_symbol_alignment.
422 (prepare_move_operands): Don't rewite sdata pattern.
423 (prepare_extend_operands): Remove.
424 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
425 pattern.
426 (zero_extendqisi2): Likewise.
427 (zero_extendhisi2): Likewise.
428 (extendqihi2): Likewise.
429 (extendqisi2): Likewise.
430 (extendhisi2): Likewise.
431 (addsi3): Likewise.
432 (subsi3): Likewise.
433 (andsi3): Likewise.
434 * config/arc/constraints.md (Usd): Change it to memory constraint.
436 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
438 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
439 as source of std instructions.
440 * config/arc/arc.md (movsi_insn): Update pattern predicate to
441 allow 6-bit constants as source for store instructions.
442 (movdi_insn): Update instruction pattern to allow 6-bit constants
443 as source for store instructions.
445 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
447 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
449 2018-04-30 Nathan Sidwell <nathan@acm.org>
450 Sandra Loosemore <sandra@codesourcery.com>
452 * dumpfile.c (dump_open): Allow '-' for stdout.
453 * doc/invoke.texi (Developer Options): Document dump filename
454 determination early. Document stdin/stdout selection.
456 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
458 Microblaze Target: PIC data text relative
460 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
461 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
462 Add declaration.
463 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
464 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
465 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
466 New addressing mode for data-text relative position indepenedent code.
467 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
468 'ADDRESS_SYMBOLIC_TXT_REL'.
469 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
470 (microblaze_legitimate_pic_operand): Exclude function calls from
471 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
472 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
473 addresses cases.
474 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
475 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
476 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
477 for 'address + offset'.
478 (microblaze_expand_prologue): Add new function prologue call for
479 'r20' assignation.
480 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
481 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
482 table in case of TARGET_PIC_DATA_TEXT_REL.
483 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
484 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
485 Add new macros 'UNSPEC_TEXT',
486 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
487 + exclude function calls from 'UNSPEC_PLT' in case of data text
488 relative mode.
489 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
490 new target hook for generating address diff vector tables in case of
491 flag_pic.
492 * doc/tm.texi : Regenerate.
493 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
494 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
495 of addr diff vector generation.
496 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
497 target hook definition.
498 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
499 Add default function for generate_pic_addr_diff_vec -> flag_pic.
500 * doc/invoke.texi (Add new pic option): Add new microblaze pic
501 option for data text relative.
503 2018-04-30 Richard Biener <rguenther@suse.de>
505 * tree-chrec.h (evolution_function_is_constant_p): Remove
506 redundant check.
507 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
509 2018-04-30 Richard Biener <rguenther@suse.de>
511 PR bootstrap/85571
512 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
514 2018-04-30 Richard Biener <rguenther@suse.de>
516 PR tree-optimization/28364
517 PR tree-optimization/85275
518 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
519 copying first exit test.
521 2018-04-28 Mark Wielaard <mark@klomp.org>
523 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
524 dwarf_version >= 5.
525 (dwarf_AT): Handle DW_AT_addr_base.
526 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
528 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
530 PR target/84431
531 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
532 (*ashl<dwi>3_doubleword_mask_1): Ditto.
533 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
534 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
536 2018-04-28 Richard Biener <rguenther@suse.de>
538 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
539 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
540 to reflect use. Only add interesting stmts.
542 2018-04-27 Martin Jambor <mjambor@suse.cz>
544 PR ipa/85549
545 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
546 the jump function allows for passing through aggregate values.
548 2018-04-27 David Malcolm <dmalcolm@redhat.com>
550 * input.h (in_system_header_at): Convert from macro to inline
551 function.
552 (from_macro_expansion_at): Likewise.
553 (from_macro_definition_at): Likewise.
555 2018-04-27 Jeff Law <law@redhat.com>
557 * config.gcc: Mark tile* targets as deprecated/obsolete.
559 2018-04-27 Richard Biener <rguenther@suse.de>
561 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
562 fix for ILP32.
564 2018-04-27 Richard Biener <rguenther@suse.de>
566 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
568 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
570 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
571 with Yd constraint. Set "preferred_for_speed" attribute from
572 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
573 with Yd constraint.
574 (*movdi_internal): Ditto.
575 (movti_interunit splitters): Remove
576 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
577 (movdi_interunit splitters): Ditto.
578 * config/i386/constraints.md (Ye): Remove.
579 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
581 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
583 PR target/85512
584 * config/aarch64/constraints.md (Usg): Limit to 31.
585 (Usj): Limit to 63.
587 2018-04-27 Jakub Jelinek <jakub@redhat.com>
589 PR tree-optimization/85529
590 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
591 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
592 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
593 zero extension or masking of the MSB bit.
594 (optimize_range_tests): Add FIRST_BB argument, pass it through
595 to optimize_range_tests_var_bound.
596 (maybe_optimize_range_tests, reassociate_bb): Adjust
597 optimize_range_tests callers.
599 2018-04-26 Richard Biener <rguenther@suse.de>
600 Jakub Jelinek <jakub@redhat.com>
602 * cgraph.h (symbol_table): Just declare debug method here.
603 * symtab.c (symbol_table::debug): Define.
605 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
607 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
609 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
611 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
612 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
613 (*movdi_internal): Substitute Yi and Yj constraint with x
614 and Ym and Yn constraint with y constraint. Update "isa"
615 attribute and set "preferred_for_speed" attribute from
616 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
617 (*movsi_internal): Ditto.
618 (*movdf_internal): Ditto.
619 (*movsf_internal): Ditto.
620 (*zero_extendsidi2): Ditto.
621 * config/i386/sse.md (vec_set<mode>_0): Ditto.
622 (sse2_loadld): Ditto.
623 (*vec_extract<ssevecmodelower>_0): Ditto.
624 (*vec_extractv4si_0_zext_sse4): Ditto.
625 (vec_concatv2di): Ditto.
626 (*vec_dup<mode>): Ditto.
627 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
628 * config/i386/constraints.md (Yi): Remove.
629 (Yj): Remove.
630 (Ym): Remove.
631 (Yn): Remove.
633 2018-04-26 Nathan Sidwell <nathan@acm.org>
635 * dumpfile.c (dump_open): New.
636 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
637 (dump_finish): Detect stdio/stderr by value not name.
639 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
641 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
643 2018-04-26 Tom de Vries <tom@codesourcery.com>
645 PR target/84952
646 * config/nvptx/nvptx.c (verify_neutering_jumps)
647 (verify_neutering_labels): New function
648 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
650 2018-04-26 Tom de Vries <tom@codesourcery.com>
652 PR target/84025
653 * config/nvptx/nvptx.c (needs_neutering_p): New function.
654 (nvptx_single): Use needs_neutering_p to skip over insns that do not
655 need neutering.
657 2018-04-26 Richard Biener <rguenther@suse.de>
658 Tom de Vries <tom@codesourcery.com>
660 PR lto/85422
661 * lto-streamer-out.c (output_function): Fixup loops if required to match
662 discovery done in the reader.
664 2018-04-26 Richard Biener <rguenther@suse.de>
666 PR tree-optimization/85116
667 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
668 have a loop exit from the single latch predecessor. Remove
669 case of header with just condition.
670 (ch_base::copy_headers): Exclude infinite loops from any
671 processing.
672 (pass_ch::execute): Record exits.
674 2018-04-26 Richard Biener <rguenther@suse.de>
676 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
677 prologue cost vector and pass it to vect_get_load_cost.
678 (vect_get_peeling_costs_all_drs): Likewise.
679 (vect_peeling_hash_get_lowest_cost): Likewise.
680 (vect_enhance_data_refs_alignment): Likewise.
682 2018-04-26 Richard Biener <rguenther@suse.de>
684 PR middle-end/85450
685 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
686 checking of integer<->pointer conversions.
687 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
688 sign-/zero-extending pointer types.
689 (expand_omp_for_static_chunk): Likewise.
691 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
692 Jean Lee <xiaoyur347@gmail.com>
694 * config/mips/mips.c (mips_asan_shadow_offset): New function.
695 (TARGET_ASAN_SHADOW_OFFSET): Define.
696 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
697 true for -fsanitize=address.
699 2018-04-25 Mark Wielaard <mark@klomp.org>
701 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
702 shorter ones.
704 2018-04-25 Jakub Jelinek <jakub@redhat.com>
706 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
707 than "alu", remove explicit "memory" and "imm_disp" attributes.
708 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
710 PR middle-end/85414
711 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
712 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
713 gen_lowpart_no_emit.
715 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
717 PR target/85473
718 * config/i386/i386.c (ix86_expand_builtin): Change memory
719 operand to XI, extend p0 to Pmode.
720 * config/i386/i386.md: Change unspec volatile and operand
721 1 mode to XI, change operand 0 mode to P.
723 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
725 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
726 GET_MODE_MASK before any checking.
727 (nds32_can_use_bset_p): Likewise.
728 (nds32_can_use_btgl_p): Likewise.
730 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
732 * config/nds32/nds32-doubleword.md: New define_split pattern for
733 illegal register number.
735 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
737 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
739 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
741 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
743 2018-04-25 Richard Biener <rguenther@suse.de>
745 * lto-streamer.h (LTO_major_version): Bump to 8.
747 2018-04-25 Jakub Jelinek <jakub@redhat.com>
749 * BASE-VER: Set to 9.0.0.
751 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
753 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
754 in __abskf2 and __powikf2.
756 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
758 PR target/85512
759 * config/aarch64/constraints.md (Usg, Usj): New constraints.
760 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
761 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
762 Use the above on operand 2. Reindent.
763 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
765 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
767 PR target/85485
768 * common/config/i386/i386-common.c (ix86_handle_option): Don't
769 handle OPT_mcet.
770 * config/i386/i386.opt (mcet): Removed.
771 * doc/install.texi: Remove -mcet documentation.
772 * doc/invoke.texi: Likewise.
774 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
776 PR target/85485
777 * doc/install.texi: Remove -mcet from bootstrap-cet.
779 2018-04-24 Jakub Jelinek <jakub@redhat.com>
781 PR target/85511
782 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
783 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
784 if TARGET_64BIT.
786 PR target/85503
787 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
788 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
789 containing a CONST_VECTOR.
791 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
793 * doc/install.texi: Update newlib dependency for nvptx.
795 2018-04-24 Jakub Jelinek <jakub@redhat.com>
797 PR target/85508
798 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
799 instead of INTVAL when shifting x left.
801 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
803 PR tree-optimization/85478
804 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
805 vect_grouped_store_supported for single element vectors.
807 2018-04-24 Richard Biener <rguenther@suse.de>
809 PR target/85491
810 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
811 load cost increase to the case of non-constant step.
813 2018-04-24 Jakub Jelinek <jakub@redhat.com>
815 PR target/84828
816 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
817 destination if any_malformed_asm.
819 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
821 PR middle-end/85496
822 * expr.c (store_field): In the bitfield case, if the value comes from
823 a function call and is returned in registers by means of a PARALLEL,
824 do not change the mode of the temporary unless BLKmode and VOIDmode.
826 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
828 PR rtl-optimization/85423
829 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
830 dependencies to debug insns when the previous insn is non-debug.
832 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
834 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
835 enums into a single definition.
836 (fls): Fix predicates and printing.
837 (seti): Likewise.
839 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
841 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
842 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
843 and short u6 immediate.
844 (check_if_valid_sleep_operand): Remove.
845 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
847 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
849 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
850 flag_always_save_lp condition.
851 * config/nds32/nds32.opt (malways-save-lp): New option.
853 2018-04-22 Shiva Chen <shiva0217@gmail.com>
855 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
856 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
857 * config/nds32/nds32.h
858 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
859 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
861 2018-04-22 Shiva Chen <shiva0217@gmail.com>
863 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
864 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
866 2018-04-22 Shiva Chen <shiva0217@gmail.com>
867 Chung-Ju Wu <jasonwucj@gmail.com>
869 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
870 Declare.
871 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
872 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
874 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
876 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
878 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
880 * config/nds32/nds32-protos.h (nds32_data_alignment,
881 nds32_local_alignment): Declare.
882 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
883 nds32_local_alignment): New functions.
884 (TARGET_CONSTANT_ALIGNMENT): Define.
885 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
887 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
889 * config/nds32/nds32.c
890 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
891 (TARGET_MODES_TIEABLE_P): Likewise.
893 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
895 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
896 level Ofast and Og.
898 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
899 Chung-Ju Wu <jasonwucj@gmail.com>
901 * config/nds32/constants.md (unspec_volatile_element): Add enum values
902 for unaligned access.
903 * config/nds32/nds32-intrinsic.c: Implementation of expanding
904 unaligned access.
905 * config/nds32/nds32-intrinsic.md: Likewise.
906 * config/nds32/nds32_intrinsic.h: Likewise.
907 * config/nds32/nds32.h (nds32_builtins): Likewise.
908 * config/nds32/nds32.opt (munaligned-access): New option.
909 * config/nds32/nds32.c (nds32_asm_file_start): Display
910 flag_unaligned_access status.
912 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
914 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
915 -mno-relax is present.
916 * config/riscv/linux.h (LINK_SPEC): Ditto.
918 2018-04-20 Martin Sebor <msebor@redhat.com>
920 PR c/85365
921 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
922 for null pointers.
923 (gimple_fold_builtin_stxcpy_chk): Same.
924 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
926 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
928 PR target/85456
929 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
930 __powikf2 when long double is IEEE 128-bit.
932 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
934 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
935 step to make sure stack always aligned.
937 2018-04-20 Carl Love <cel@us.ibm.com>
939 PR target/83402
940 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
941 size check for arg0.
943 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
944 Tom de Vries <tom@codesourcery.com>
946 PR target/85445
947 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
948 Emit insns for calls too.
949 (nvptx_find_par): Always look for worker-level predecessor insn.
950 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
951 calls.
952 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
953 (nvptx_process_pars): Propagate frames for calls.
955 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
957 PR target/85469
958 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
959 Removed.
960 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
961 (ix86_handle_option): Don't handle OPT_mibt.
962 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
963 __SHSTK__.
964 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
965 has_ibt and ibt.
966 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
967 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
968 (ix86_target_macros): Define __CET__ with flag_cf_protection
969 for -fcf-protection.
970 * config/i386/i386.c (isa2_opts): Remove -mibt.
971 * config/i386/i386.h (TARGET_IBT): Removed.
972 (TARGET_IBT_P): Likewise.
973 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
974 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
975 * config/i386/i386.opt (mcet): Update help message.
976 (mshstk): Likewise.
977 (mibt): Removed.
978 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
979 -mcet as an alias for -mshstk.
981 2018-04-20 Richard Biener <rguenther@suse.de>
983 PR middle-end/85475
984 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
985 complexity by forcing a single use of the multiply operand.
987 2018-04-20 Martin Jambor <mjambor@suse.cz>
989 ipa/85449
990 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
991 recursion dependency to only apply to non-clones.
993 2018-04-20 Martin Jambor <mjambor@suse.cz>
995 ipa/85447
996 * ipa-cp.c (create_specialized_node): Check that clones of
997 self-recursive edges exist during IPA-CP.
999 2018-04-19 Toon Moene <toon@moene.org>
1001 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1002 by -O3.
1004 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1006 PR tree-optimization/85467
1007 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1008 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1009 VECTOR_CST element to type.
1011 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1013 PR target/85397
1014 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1015 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1016 (builtin_longjmp): Likewise.
1017 (save_stack_nonlocal): New pattern.
1018 (restore_stack_nonlocal): Likewise.
1020 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1022 PR target/85404
1023 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1024 Replace ASM_OUTPUT_LABEL with fprintf.
1026 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1028 PR target/85417
1029 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1030 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1031 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1032 define __IBT__ and __SHSTK__ for -fcf-protection.
1033 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1034 TARGET_IBT.
1035 (ix86_trampoline_init): Likewise.
1036 (x86_output_mi_thunk): Likewise.
1037 (ix86_notrack_prefixed_insn_p): Likewise.
1038 (ix86_option_override_internal): Don't disallow -fcf-protection.
1039 * config/i386/i386.md (rdssp<mode>): Also enable for
1040 -fcf-protection.
1041 (incssp<mode>): Likewise.
1042 (nop_endbr): Likewise.
1043 * config/i386/i386.opt (mcet): Change help message to built-in
1044 functions only.
1045 (mibt): Likewise.
1046 (mshstk): Likewise.
1047 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1048 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1049 enable CET built-in functions.
1051 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1053 * common/config/i386/i386-common.c
1054 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1055 OPTION_MASK_ISA_MOVDIRI_UNSET,
1056 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1057 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1058 * config.gcc (movdirintrin.h): New header.
1059 * config/i386/cpuid.h (bit_MOVDIRI,
1060 bit_MOVDIR64B): New bits.
1061 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1062 and -mmvodir64b.
1063 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1064 (VOID, PVOID, PCVOID)): New function types.
1065 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1066 __builtin_ia32_directstoreu_u64,
1067 __builtin_ia32_movdir64b): New builtins.
1068 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1069 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1070 and -mmovdiri.
1071 (ix86_valid_target_attribute_inner_p): Ditto.
1072 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1073 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1074 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1075 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1076 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1077 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1078 (movdiri<mode>, movdir64b_<mode>): New.
1079 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1080 * config/i386/immintrin.h: Include movdirintrin.h.
1081 * config/i386/movdirintrin.h: New file.
1082 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1084 2018-04-19 Richard Biener <rguenther@suse.de>
1086 PR middle-end/85455
1087 * cfg.c (clear_bb_flags): When loop state says we have
1088 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1090 2018-04-19 Richard Biener <rguenther@suse.de>
1092 PR tree-optimization/84737
1093 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1094 copying restrict info.
1095 (vect_setup_realignment): Use it.
1096 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1097 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1098 the first DR to all generated stores.
1099 (vectorizable_load): Likewise for loads.
1101 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1103 PR tree-optimization/85446
1104 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1105 the integral and pointer types to have the same precision.
1107 * doc/install.texi: Document --disable-cet being the default and
1108 --enable-cet=auto.
1110 2018-04-18 Martin Liska <mliska@suse.cz>
1112 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1113 style.
1115 2018-04-18 Martin Liska <mliska@suse.cz>
1117 Revert
1118 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1120 PR ipa/83983
1121 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1122 arguments if they are comparable.
1124 2018-04-18 Martin Liska <mliska@suse.cz>
1126 Revert
1127 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1129 PR lto/84805
1130 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1131 incomplete types.
1133 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1135 PR target/85388
1136 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1137 ENDBR after calling __morestack.
1139 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1141 PR jit/85384
1142 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1143 by using gcc_base_ver to generate a gcc_driver_version, and use
1144 it when generating GCC_DRIVER_NAME.
1145 * configure: Regenerate.
1147 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1149 PR target/81084
1150 * config.gcc: Obsolete powerpc*-*-*spe*.
1152 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1154 PR debug/84637
1155 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1156 (stabstr_D): Change type of unum from unsigned int to
1157 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1158 type.
1160 2018-04-17 Jim Wilson <jimw@sifive.com>
1162 PR 84856
1163 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1164 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1165 Set arg_pointer_offset after using pretend_args_size.
1167 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1169 PR rtl-optimization/85431
1170 * dse.c (record_store): Ignore zero width stores.
1172 PR sanitizer/85230
1173 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1174 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1175 __builtin_stack_restore rather than after it.
1176 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1177 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1178 argument instead of virtual_dynamic_stack_rtx.
1180 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1182 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1183 New prototype.
1184 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1185 Add note to error message to explain internal mapping of overloaded
1186 built-in function name to non-overloaded built-in function name.
1187 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1188 function.
1190 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1192 PR target/85424
1193 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1194 where the inputs overlap with the output.
1196 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1198 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1199 (=v, v) alternative and explicit "memory" attribute.
1200 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1201 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1202 attributes.
1203 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1204 "sselog1" type instead of "sselog".
1205 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1206 "sselog". Remove explicit "memory" attribute.
1207 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1208 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1209 attributes.
1210 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1211 "sselog1" type instead of "sselog". Remove explicit "memory"
1212 attribute.
1213 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1214 use "sselog1" type instead of "sselog". Remove explicit "memory"
1215 attribute.
1216 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1217 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1218 attributes.
1219 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1220 "sselog1" type instead of "sselog". Remove explicit "memory"
1221 attribute.
1222 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1223 use "sselog1" type instead of "sselog". Remove explicit "memory"
1224 attribute.
1226 PR target/85430
1227 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1229 PR middle-end/85414
1230 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1231 on a SUBREG.
1233 2018-04-17 Martin Jambor <mjambor@suse.cz>
1235 PR ipa/85421
1236 * ipa-cp.c (create_specialized_node): Call
1237 expand_all_artificial_thunks if necessary.
1239 2018-04-17 Martin Liska <mliska@suse.cz>
1241 PR lto/85405
1242 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1243 in message, remote space in between '_G' and '('.
1245 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1247 PR target/85281
1248 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1249 avx512f_vmcmp<mode>3<round_saeonly_name>,
1250 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1251 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1252 avx512f_rndscale<mode><round_saeonly_name>,
1253 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1254 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1255 Use %<iptr>2 instead of %2 for -masm=intel.
1256 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1257 avx512f_vcvttss2usi<round_saeonly_name>,
1258 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1259 -masm=intel.
1260 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1261 avx512f_vcvttsd2usi<round_saeonly_name>,
1262 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1263 Use %q1 instead of %1 for -masm=intel.
1264 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1265 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1266 of %3 for -masm=intel.
1267 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1268 -masm=intel.
1269 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1270 -masm=intel.
1271 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1272 -masm=intel.
1273 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1274 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1275 %g1.
1276 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1277 -masm=intel.
1278 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1279 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1280 %g1 and one with %0 and %1.
1281 (avx512er_vmrcp28<mode><round_saeonly_name>,
1282 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1283 %1 for -masm=intel.
1284 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1285 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1286 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1287 of %0 and %{%4%} for -masm=intel.
1288 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1289 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1290 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1291 order of %0 and %{%5%}%{z%} for -masm=intel.
1293 2018-04-17 Jan Hubicka <jh@suse.cz>
1295 PR lto/85405
1296 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1298 2018-04-17 Martin Liska <mliska@suse.cz>
1300 PR ipa/85329
1301 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1302 for target_clone error message. Make default implementation
1303 clone to be a local declaration.
1304 (separate_attrs): Add new argument and check for an empty
1305 string.
1306 (expand_target_clones): Handle it.
1307 (ipa_target_clone): Make redirection just for target_clones
1308 functions.
1310 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1311 Tom de Vries <tom@codesourcery.com>
1313 PR middle-end/84955
1314 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1315 tiled basic blocks without omp continue statements.
1317 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1319 PR target/83660
1320 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1321 vec_extract expression as having side effects to make sure it gets
1322 a cleanup point.
1324 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1326 PR target/85403
1327 * config/i386/i386.c (get_builtin_code_for_version): Check
1328 error_mark_node.
1330 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1332 PR target/84331
1333 * gcc/config.gcc: Support "skylake".
1334 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1335 PROCESSOR_SKYLAKE.
1336 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1337 (processor_target_table): Add "skylake".
1338 (ix86_option_override_internal): Add "skylake".
1339 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1340 PROCESSOR_CANNONLAKE.
1341 (get_builtin_code_for_version): Fix priority for
1342 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1343 PROCESSOR_SKYLAKE-AVX512.
1344 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1345 (processor_type): Add PROCESSOR_SKYLAKE.
1347 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1348 Jason Merrill <jason@redhat.com>
1350 PR c++/85112
1351 * convert.c (convert_to_integer_1): Use direct recursion for
1352 enumeral types and types with a precision less than the number
1353 of bits in their mode.
1355 2018-04-16 Julia Koval <julia.koval@intel.com>
1357 PR target/84413
1358 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1359 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1361 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1363 PR target/85293
1364 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1365 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1366 and -mno-direct-move.
1368 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1370 PR target/83402
1371 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1372 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1373 Ensure negative shifts result in {0}.
1375 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1377 PR rtl-optimization/79916
1378 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1379 regs (if any) to define how to gnerate SD moves when LRA is in
1380 progress.
1382 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1384 PR rtl-optimization/85393
1385 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1386 * except.c (expand_dw2_landing_pad_for_region): Make static.
1387 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1388 a label and unconditional jump to old_bb, rather than
1389 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1390 basic block.
1392 PR rtl-optimization/85376
1393 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1394 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1395 instead of a specific value.
1397 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1398 Bin Cheng <bin.cheng@arm.com>
1400 PR tree-optimization/82965
1401 PR tree-optimization/83991
1402 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1403 by_profile_only parameter.
1404 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1405 information if the loop was predicted to iterate too many times.
1406 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1408 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1410 PR lto/71991
1411 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1412 always inline.
1414 2018-04-13 Martin Liska <mliska@suse.cz>
1415 Jakub Jelinek <jakub@redhat.com>
1417 PR middle-end/81657
1418 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1419 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1420 * builtins.c (expand_builtin_memory_copy_args): Use
1421 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1422 handle dest_addr == pc_rtx.
1424 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1426 PR target/85291
1427 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1428 asked to not generate direct moves.
1429 (fix_trunc<mode>si2_stfiwx): Similar.
1430 (fix_trunc<mode>si2_internal): Similar.
1432 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1434 PR debug/83157
1435 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1436 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1437 lookup if dest in some wider mode is known to be const0_rtx and
1438 if so, record permanent equivalence for it to be ZERO_EXTEND of
1439 the narrower mode destination.
1441 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1443 * lto-streamer-out.c (output_function): Revert 259346.
1444 * omp-expand.c (expand_oacc_for): Likewise.
1446 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1448 PR rtl-optimization/85354
1449 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1450 * sel-sched.c (sel_global_init): ... here.
1452 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1454 PR target/85238
1455 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1456 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1457 mode for PE-COFF targets.
1458 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1459 (i386_pe_asm_lto_end): Likewise.
1460 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1461 (TARGET_ASM_LTO_END): Likewise.
1462 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1463 (i386_pe_asm_lto_start): New function.
1464 (i386_pe_asm_lto_end): Likewise.
1466 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1467 Richard Biener <rguenther@suse.de>
1469 PR middle-end/84955
1470 * lto-streamer-out.c (output_function): Fix CFG loop state before
1471 streaming out.
1472 * omp-expand.c (expand_oacc_for): Handle calls to internal
1473 functions like regular functions.
1475 2018-04-12 Richard Biener <rguenther@suse.de>
1477 PR lto/85371
1478 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1479 for the early LTO debug to properly generate references to it
1480 during DIE emission. Do not re-use that for the skeleton for
1481 split-dwarf.
1482 (dwarf2out_early_finish): Likewise.
1484 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1486 PR target/85328
1487 * config/i386/sse.md
1488 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1489 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1490 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1491 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1492 and output is a reg, avoid creating invalid lowpart subreg, but
1493 instead split into a 512-bit move. Don't split if not AVX512VL,
1494 input is xmm16+ reg and output is a mem.
1495 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1496 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1497 xmm16+ reg and output is a mem.
1499 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1501 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1502 also for flag_dwarf2_cfi_asm.
1504 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1506 PR rtl-optimization/85342
1507 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1508 a bool scalar var inside of the loop instead. Don't try to update
1509 recog_data.operand after failed apply_change_group.
1511 2018-04-12 Tom de Vries <tom@codesourcery.com>
1513 PR target/85296
1514 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1515 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1516 array with flexible array member as array without given dimension.
1517 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1518 argument for undefined param to true.
1520 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1522 PR target/85321
1523 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1524 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1525 from PowerPC section.
1526 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1527 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1528 help text that is too long.
1529 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1530 help text that is too long.
1531 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1532 help text that is too long.
1534 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1536 * config/alpha/alpha.md (stack_probe_internal): Rename
1537 from "probe_stack". Update all callers.
1539 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1541 PR rtl-optimization/84566
1542 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1543 sched_macro_fuse_insns.
1545 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1547 PR target/84301
1548 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1549 (compute_block_dependences): ... from here.
1551 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1553 PR tree-optimization/85331
1554 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1555 from int to HOST_WIDE_INT.
1557 2018-04-11 Martin Jambor <mjambor@suse.cz>
1559 PR ipa/84149
1560 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1561 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1562 not the same as the source val.
1563 (cgraph_edge_brings_value_p): New parameter.
1564 (gather_edges_for_value): Pass destination value to
1565 cgraph_edge_brings_value_p.
1566 (perhaps_add_new_callers): Likewise.
1567 (get_info_about_necessary_edges): Likewise and exclude values brought
1568 only by self-recursive edges.
1569 (create_specialized_node): Redirect only clones of self-calling edges.
1570 (+self_recursive_pass_through_p): New function.
1571 (find_more_scalar_values_for_callers_subset): Use it.
1572 (find_aggregate_values_for_callers_subset): Likewise.
1573 (known_aggs_to_agg_replacement_list): Removed.
1574 (decide_whether_version_node): Re-calculate known constants for all
1575 remaining context clones.
1577 2018-04-11 Richard Biener <rguenther@suse.de>
1579 PR lto/85339
1580 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1581 from early DWARF output.
1582 (dwarf2out_early_finish): Output line info unconditionally into
1583 early DWARF and add reference to it.
1585 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1587 PR target/85281
1588 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1589 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1590 other than V2DFmode using iptr mode attribute.
1591 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1593 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1595 PR rtl-optimization/84659
1596 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1598 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1600 PR debug/85302
1601 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1602 SIZEP is NULL.
1603 (output_loc_list): Pass address of a dummy size variable even in the
1604 locview handling loop.
1605 (index_location_lists): Add comment on why skip_loc_list_entry can't
1606 call size_of_locs.
1608 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1610 PR target/85261
1611 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1612 into register.
1614 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1616 PR target/85321
1617 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1618 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1619 and -mstring-compare-inline-limit.
1621 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1623 PR target/85287
1624 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1625 for stack clash protection in a register whenever we need it to be in
1626 a register.
1628 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1630 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1631 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1633 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1635 PR target/85321
1636 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1637 the help text.
1638 (mlong-double-): Ditto.
1639 * config/rs6000/sysv4.opt (msdata=): Ditto.
1640 (mtls-size=): Ditto.
1642 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1644 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1645 erroneous entries for
1646 "vector int vec_ldl (int, long int *)", and
1647 "vector unsigned int vec_ldl (int, unsigned long int *)".
1648 Add comments and entries for
1649 "vector bool char vec_ldl (int, bool char *)",
1650 "vector bool short vec_ldl (int, bool short *)",
1651 "vector bool int vec_ldl (int, bool int *)",
1652 "vector bool long long vec_ldl (int, bool long long *)",
1653 "vector pixel vec_ldl (int, pixel *)",
1654 "vector long long vec_ldl (int, long long *)",
1655 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1656 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1657 type tree bool_long_long_type_node and correct definition of
1658 bool_V2DI_type_node to make reference to this new type tree.
1659 (rs6000_mangle_type): Replace erroneous reference to
1660 bool_long_type_node with bool_long_long_type_node.
1661 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1662 comments to emphasize sign distinctions for char and int types and
1663 replace RS6000_BTI_bool_long constant with
1664 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1665 use of RS6000_BTI_pixel.
1666 (bool_long_type_node): Remove this macro definition.
1667 (bool_long_long_type_node): New macro definition
1669 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1671 PR rtl-optimization/85300
1672 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1673 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1674 simplify_unary_operation fails.
1676 2018-04-10 Martin Liska <mliska@suse.cz>
1678 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1679 cgraph_edge and ipa_ref.
1681 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1683 PR target/85177
1684 PR target/85255
1685 * config/i386/sse.md
1686 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1687 computation of the VEC_MERGE selector from mask.
1688 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1689 Fix decoding of the VEC_MERGE selector into mask.
1691 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1693 PR tree-optimization/85286
1694 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1696 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1698 * final.c (final_1): Set insn_last_address as well as
1699 insn_current_address.
1701 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1703 PR target/85173
1704 * explow.c (emit_stack_probe): Call validize_mem on memory location
1705 before passing it to gen_probe_stack. Create address operand and
1706 legitimize it for the probe_stack_address case.
1708 2018-04-09 Jan Hubicka <jh@suse.cz>
1710 PR lto/85078
1711 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1712 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1713 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1714 walk basetypes.
1715 (free_lang_data): Rebuild type inheritance graph.
1717 2018-04-09 Martin Sebor <msebor@redhat.com>
1719 * invoke.texi (-finline-small-functions): Mention other optimization
1720 options.
1721 (-findirect-inlining, -fpartial-inlining): Same.
1722 (-finline-functions-called-once): Same.
1723 (-freorder-blocks-and-partition): Same.
1725 2018-04-09 Jan Hubicka <jh@suse.cz>
1727 PR rtl/84058
1728 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1729 jumps; choose last target that matches the criteria (i.e.
1730 no partition changes for non-crossing jumps).
1731 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1732 support for redirecting crossing jumps to non-crossing.
1734 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1736 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1737 also for naked functions.
1739 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1741 * config/arc/arc.md (add_shift): New pattern.
1742 (add_shift2): Likewise.
1743 (sub_shift): Likewise.
1744 (sub_shift_cmp0_noout): Likewise.
1745 (compare_si_ashiftsi): Likewise.
1746 (xbfu_cmp0_noout): New combine pattern.
1747 (xbfu_cmp0"): Likewise.
1748 (movsi_set_cc_insn): Place the predicable variant first.
1749 (commutative_binary_cmp0_noout): Remove clobber.
1750 (commutative_binary_cmp0): New pattern.
1751 (noncommutative_binary_cmp0): Likewise.
1752 (noncommutative_binary_cmp0_noout): Likewise.
1753 (noncommutative_binary_comparison_result_used): Removed.
1754 (rsub_cmp0): New pattern.
1755 (rsub_cmp0_noout): Likewise.
1756 (extzvsi): Changed, keep only meaningful variants.
1757 (SQH, SEZ): New iterators.
1758 (SQH_postfix): New mode attribute.
1759 (SEZ_prefix): New code attribute.
1760 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1761 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1762 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1763 of numerical value.
1764 (noncommutative_operator): Check the availability of barrel
1765 shifter option.
1767 2018-04-09 Richard Biener <rguenther@suse.de>
1769 PR tree-optimization/85284
1770 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1771 Only use the niter constraining form of simple_iv when the exit
1772 is always executed.
1774 2018-04-09 Tom de Vries <tom@codesourcery.com>
1776 PR target/84041
1777 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1778 (define_expand "*memory_barrier"): New define_expand.
1779 (define_insn "memory_barrier"): New insn.
1781 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1783 PR rtl-optimization/80463
1784 PR rtl-optimization/83972
1785 PR rtl-optimization/83480
1787 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1788 correct producer for the insn.
1789 (tidy_control_flow): Fixup seqnos in case of debug insns.
1791 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1793 PR rtl-optimization/83913
1795 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1796 different sched-times when merging exprs.
1798 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1800 PR rtl-optimization/83962
1802 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1803 tidy_fallthru_edge and tidy_control_flow.
1805 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1807 PR rtl-optimization/83530
1809 * sel-sched.c (force_next_insn): New global variable.
1810 (remove_insn_for_debug): When force_next_insn is true, also leave only
1811 next insn in the ready list.
1812 (sel_sched_region): When the region wasn't scheduled, make another pass
1813 over it with force_next_insn set to 1.
1815 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1817 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1818 into tm_file.
1819 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1820 for interrupt control.
1821 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1822 functions for interrupt control.
1823 * config/nds32/nds32-intrinsic.md: Likewise.
1824 * config/nds32/nds32_intrinsic.h: Likewise.
1825 * config/nds32/nds32.h (nds32_builtins): Likewise.
1827 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1829 * config/nds32/nds32.c (nds32_init_machine_status,
1830 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1831 strict_aligned_p field.
1832 (nds32_expand_to_rtl_hook): New function.
1833 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1834 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1836 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1837 Chung-Ju Wu <jasonwucj@gmail.com>
1839 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1840 * config/nds32/nds32-n7.md: New file.
1841 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1842 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1843 pipeline.
1844 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1845 * config/nds32/nds32.md (pipeline_model): Add n7.
1846 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1847 * config/nds32/pipelines.md: Include n7 settings.
1849 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1850 Chung-Ju Wu <jasonwucj@gmail.com>
1852 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1853 * config/nds32/nds32-e8.md: New file.
1854 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1855 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1856 pipeline.
1857 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1858 * config/nds32/nds32.md (pipeline_model): Add e8.
1859 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1860 * config/nds32/pipelines.md: Include e8 settings.
1862 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1863 Chung-Ju Wu <jasonwucj@gmail.com>
1865 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1866 * config/nds32/nds32-n8.md: New file.
1867 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1868 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1869 pipeline.
1870 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1871 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1872 * config/nds32/nds32.md (pipeline_model): Add n8.
1873 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1874 * config/nds32/pipelines.md: Include n8 settings.
1876 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1877 Chung-Ju Wu <jasonwucj@gmail.com>
1879 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1880 * config/nds32/nds32-n9-2r1w.md: New file.
1881 * config/nds32/nds32-n9-3r2w.md: New file.
1882 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1883 nds32_register_ports): New or modify for cpu n9.
1884 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1885 pipeline.
1886 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1887 * config/nds32/nds32-utils.c: New file.
1888 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1889 TARGET_MUL_SLOW): Define.
1890 * config/nds32/nds32.md (pipeline_model): New attribute.
1891 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1892 New options that support cpu n9.
1893 * config/nds32/pipelines.md: Include n9 settings.
1894 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1896 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1898 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1899 information if necessary.
1900 (output_cond_branch_compare_zero): Likewise.
1901 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1902 (nds32_target_alignment): Refine for alignment.
1903 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1904 (FUNCTION_BOUNDARY): Modify.
1905 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1906 align case.
1907 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1909 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1911 * config/nds32/constants.md (unspec_volatile_element): Add values for
1912 TLB operation and data prefetch.
1913 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1914 functions for TLB operation and data prefetch.
1915 * config/nds32/nds32-intrinsic.md: Likewise.
1916 * config/nds32/nds32_intrinsic.h: Likewise.
1917 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1918 (nds32_print_operand): Likewise.
1919 * config/nds32/nds32.h (nds32_builtins): Likewise.
1921 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1922 Andrew Pinski <pinsika@gcc.gnu.org>
1924 PR middle-end/82976
1925 * match.pd: Use constant_boolean_node of correct type instead of
1926 boolean_true_node or boolean_false_node for simplifying
1927 pointer comparisons to zero.
1929 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1931 PR tree-optimization/80021
1932 * tree.c (verify_type_variant): Make error call in verify_variant_match
1933 translatable and remove final full stop.
1935 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1937 * config/nds32/constants.md (unspec_volatile_element): Add
1938 UNSPEC_VOLATILE_EH_RETURN.
1939 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1940 nds32_output_stack_pop): Support dwarf exception handling process.
1941 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1942 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1943 exception handling process.
1944 (nds32_compute_stack_frame): Likewise.
1945 (nds32_return_addr_rtx): Likewise.
1946 (nds32_initial_elimination_offset): Likewise.
1947 (nds32_expand_prologue): Likewise.
1948 (nds32_expand_epilogue): Likewise.
1949 (nds32_dynamic_chain_address): New function.
1950 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1951 exception handling.
1952 (DYNAMIC_CHAIN_ADDRESS): Define.
1953 (EH_RETURN_DATA_REGNO): Define.
1954 (EH_RETURN_STACKADJ_RTX): Define.
1955 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1956 patterns for dwarf exception handling.
1958 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1960 * config/nds32/nds32.h: Clean up obsolete macros.
1962 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1964 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1965 Add enum values for particular instructions.
1966 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1967 particular intrinsic functions.
1968 * config/nds32/nds32-intrinsic.md: Likewise.
1969 * config/nds32/nds32_intrinsic.h: Likewise.
1970 * config/nds32/nds32.h (nds32_builtins): Likewise.
1971 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1972 (btst, ave): New patterns for particular instructions.
1974 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1976 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1977 Add enum values for atomic load/store and memory sync.
1978 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1979 and memory sync.
1980 * config/nds32/nds32-intrinsic.md: Likewise.
1981 * config/nds32/nds32_intrinsic.h: Likewise.
1982 * config/nds32/nds32.h (nds32_builtins): Likewise.
1984 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1986 PR tree-optimization/85257
1987 * fold-const.c (native_encode_vector): If not all elts could fit
1988 and off is -1, return 0 rather than offset.
1989 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1990 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1991 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1992 adjust buffer in native_interpret_expr call.
1994 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1996 * config/nds32/constants.md (unspec_volatile_element): Add cache
1997 control enum values.
1998 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1999 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2000 * config/nds32/nds32.c (nds32_cctl_names): New.
2001 (nds32_print_operand): Handle cache control register names.
2002 * config/nds32/nds32.h (nds32_builtins): New enum values.
2003 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2004 macros.
2005 * config/nds32/nds32.md (type): Add mmu.
2006 * config/nds32/pipelines.md (simple_insn): Add mmu.
2008 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2010 * config/nds32/nds32.md (type): Remove call.
2011 * config/nds32/pipelines.md (simple_insn): Likewise.
2013 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2015 * config/nds32/constants.md (unspec_volatile_element): Add
2016 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2017 UNSPEC_VOLATILE_FMFCFG.
2018 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2019 description for fmfcfg and fmfcsr.
2020 (bdesc_1arg): Add fmtcsr.
2021 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2022 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2023 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2024 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2025 unspec_fmfcfg): New patterns.
2026 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2027 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2028 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2029 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2030 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2031 __nds32__fmfcfg): Define.
2033 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2035 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2036 intrinsic register names.
2037 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2038 intrinsic register enum values and macros.
2040 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2042 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2043 for load/store addressing form.
2044 (nds32_print_operand_address): Likewise.
2046 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2048 PR target/85196
2049 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2050 based on LABEL_REF. Remove useless assertion.
2051 (pic_address_needs_scratch): Fix formatting.
2052 (sparc_legitimize_pic_address): Minor tweaks.
2053 (sparc_delegitimize_address): Adjust assertion accordingly.
2054 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2055 into symbolic_operand.
2056 (movsi_high_pic_label_ref): Likewise.
2057 (movsi_lo_sum_pic_label_ref): Likewise.
2058 (movdi_pic_label_ref): Likewise.
2059 (movdi_high_pic_label_ref): Likewise.
2060 (movdi_lo_sum_pic_label_ref): Likewise.
2062 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2064 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2065 custom LIB_SPEC setup.
2067 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2068 Kito Cheng <kito.cheng@gmail.com>
2070 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2071 * config/riscv/freebsd.h: New.
2073 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2075 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2076 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2077 file.
2079 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2080 Kito Cheng <kito.cheng@gmail.com>
2082 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2083 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2084 * config/nds32/nds32-protos.h (nds32_output_call,
2085 nds32_output_return): Declare.
2086 * config/nds32/nds32.md: Refine all the call and return patterns.
2088 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2090 PR debug/85252
2091 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2092 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2094 PR rtl-optimization/84872
2095 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2096 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2097 EDGE_CROSSING edge.
2099 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2101 * expr.c (copy_blkmode_to_reg): Revert 254862.
2102 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2104 2018-04-06 Richard Biener <rguenther@suse.de>
2106 PR middle-end/85244
2107 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2108 after seeing a component reference with an adjacent field. Treat
2109 refs to arrays at struct end of external decls similar to
2110 refs to unconstrained commons.
2112 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2114 PR sanitizer/85213
2115 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2116 look through SAVE_EXPRs with non-side-effects argument. Adjust
2117 recursive calls.
2118 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2119 save_p here.
2121 2018-04-06 Richard Biener <rguenther@suse.de>
2123 PR middle-end/85180
2124 * alias.c (find_base_term): New wrapper around find_base_term
2125 unwinding CSELIB_VAL_PTR changes.
2126 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2127 recursion.
2129 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2131 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2132 instructions.
2133 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2134 constant definitions.
2135 ("nop"): lr 0,0 -> nopr r0
2136 ("nop_lr0", "nop_lr1"): New insn definitions.
2138 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2140 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2141 NDS32_V3PUSH_AVAILABLE_P macro.
2143 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2144 Chung-Ju Wu <jasonwucj@gmail.com>
2146 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2147 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2148 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2149 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2150 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2151 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2152 * config/nds32/constraints.md: New constraints and checking for hard
2153 float configuration.
2154 * config/nds32/iterators.md: New mode iterator and attribute for hard
2155 float configuration.
2156 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2157 patterns.
2158 * config/nds32/nds32-fpu.md: New file.
2159 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2160 deal with hard float code generation.
2161 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2162 ARCH_V3S.
2163 (abi_type, float_reg_number): New enum type.
2164 * config/nds32/nds32-predicates.c: New predicates for hard float.
2165 * config/nds32/nds32-protos.h: Declare functions for hard float.
2166 * config/nds32/nds32.c: Implementation for hard float configuration.
2167 * config/nds32/nds32.h: Definitions for hard float configuration.
2168 * config/nds32/nds32.md: Include hard float machine description and
2169 modify patterns for hard float configuration.
2170 * config/nds32/nds32.opt: New options for hard float configuration.
2171 * config/nds32/predicates.md: New predicates for hard float
2172 configuration.
2174 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2176 * common/config/nds32/nds32-common.c
2177 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2179 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2181 PR middle-end/85195
2182 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2183 CONSTRUCTOR_ELT (ctor, ...)->value.
2185 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2187 PR target/85193
2188 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2190 2018-04-05 Tom de Vries <tom@codesourcery.com>
2192 PR target/85204
2193 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2194 cond jump.
2196 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2197 Kito Cheng <kito.cheng@gmail.com>
2199 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2200 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2201 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2202 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2204 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2205 Kito Cheng <kito.cheng@gmail.com>
2207 * config/nds32/constraints.md (Ufe): New memory constraint.
2208 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2209 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2210 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2211 operands.
2212 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2213 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2215 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2217 * config/nds32/nds32.md: Use optimize_size in the condition for
2218 alu-shift instructions.
2220 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2222 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2224 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2226 * config/nds32/nds32.md (negsi2): Refine pattern.
2228 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2229 Chung-Ju Wu <jasonwucj@gmail.com>
2231 * config/nds32/iterators.md (shift_rotate): New code iterator.
2232 (shift): New code attribute.
2233 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2234 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2235 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2236 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2237 bit-wise operations.
2238 (andsi3, *andsi3): Ditto.
2239 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2240 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2241 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2242 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2243 nds32_ior_operand, nds32_xor_operand): New predicates.
2245 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2247 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2248 (addsi3, subsi3): ... this.
2250 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2252 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2254 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2256 * config/nds32/nds32.md: Adjust indention.
2258 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2260 * config/nds32/nds32.md (feature): New attribute.
2262 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2264 * config/nds32/nds32.md (subtype): New attribute.
2266 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2268 PR target/85203
2269 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2270 expansion to perform a bitwise AND of the argument followed by a
2271 boolean negation of the result.
2273 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2275 PR rtl-optimization/84878
2276 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2277 the basic block. Assert the use reference is not artificial and that
2278 it has an associated insn.
2280 2018-04-04 Michael Matz <matz@suse.de>
2282 * builtins.c (compute_objsize): Pass correct operand
2283 to array_at_struct_end_p.
2285 2018-04-04 Richard Biener <rguenther@suse.de>
2287 PR lto/85176
2288 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2289 from contexts for DINFO_LEVEL_TERSE and below.
2291 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2293 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2294 resiter_operand condition.
2295 * config/nds32/nds32.md (*move<mode>): Ditto.
2297 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2298 Monk Chiang <sh.chiang04@gmail.com>
2300 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2302 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2304 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2306 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2307 Kito Cheng <kito.cheng@gmail.com>
2309 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2310 nds32_cond_code_str, output_cond_branch,
2311 output_cond_branch_compare_zero, nds32_expand_cbranch,
2312 nds32_expand_cstore, nds32_expand_movcc,
2313 nds32_output_cbranchsi4_equality_zero,
2314 nds32_output_cbranchsi4_equality_reg,
2315 nds32_output_cbranchsi4_equality_reg_or_const_int,
2316 nds32_output_cbranchsi4_greater_less_zero: New functions.
2317 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2318 nds32_expand_cstore, nds32_expand_movcc,
2319 nds32_output_cbranchsi4_equality_zero,
2320 nds32_output_cbranchsi4_equality_reg,
2321 nds32_output_cbranchsi4_equality_reg_or_const_int,
2322 nds32_output_cbranchsi4_greater_less_zero): Declare.
2323 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2324 nds32_rimm11s_operand): New predicates.
2325 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2326 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2327 patterns.
2329 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2331 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2332 * config/nds32/nds32.md: Ditto.
2333 * config/nds32/pipelines.md: Ditto.
2335 2018-04-04 Richard Biener <rguenther@suse.de>
2337 PR tree-optimization/85168
2338 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2339 propagating abnormals.
2341 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2343 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2345 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2346 Kito Cheng <kito.cheng@gmail.com>
2348 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2349 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2350 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2351 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2352 * config/nds32/nds32.md (sibcall_internal): New.
2353 (sibcall_register): Remove.
2354 (sibcall_immediate): Remove.
2355 (sibcall_value_internal): New.
2356 (sibcall_value_register): Remove.
2357 (sibcall_value_immediate): Remove.
2358 * config/nds32/predicates.md (nds32_general_register_operand): New.
2359 (nds32_call_address_operand): New.
2361 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2363 PR rtl-optimization/85167
2364 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2365 bb_defs if *split_p, instead preinitialize it to NULL.
2367 PR tree-optimization/85156
2368 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2369 evaluating the argument multiple times.
2371 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2373 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2374 than vector.
2375 (_mm_cvtpd_ps): Likewise.
2376 (_mm_cvttpd_epi32): Likewise.
2377 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2378 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2379 vector, pixel, and bool following altivec.h include.
2381 2018-04-03 Martin Sebor <msebor@redhat.com>
2383 * doc/extend.texi (Common Function Attributes): Clarify.
2384 (const attribute): Likewise.
2385 (pure attribute): Likewise.
2387 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2389 PR target/85169
2390 * config/i386/i386.c (ix86_expand_vector_set): Use
2391 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2393 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2395 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2396 instructions when changing rounding bits to preserve precision bits
2397 in the x87 control word.
2399 2018-04-03 Martin Liska <mliska@suse.cz>
2401 PR tree-optimization/82491
2402 * rtl.h (strip_offset_and_add): Replace += suboffset with
2403 poly_uint64 () + suboffset.
2405 2018-03-29 Martin Liska <mliska@suse.cz>
2406 Martin Jambor <mjambor@suse.cz>
2408 PR ipa/84947
2409 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2410 param_type is not an integral or pointer type.
2412 2018-04-03 Richard Biener <rguenther@suse.de>
2414 * sese.h (recompute_all_dominators): Remove.
2416 2018-04-02 Martin Sebor <msebor@redhat.com>
2418 * doc/invoke.texi (-Wrestrict): Fix typos.
2420 2018-04-02 Jim Wilson <jimw@sifive.com>
2422 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2423 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2424 (<optab>di3, <optab>si3_extend): Likewise.
2425 (<optab>si3_mask, <optab>si3_mask_1): New.
2426 (<optab>di3_mask, <optab>di3_mask_1): New.
2427 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2428 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2429 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2431 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2433 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2434 example.
2436 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2438 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2439 (nds32_canonicalize_comparison): New function.
2441 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2442 Kito Cheng <kito.cheng@gmail.com>
2443 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2445 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2446 * config/nds32/constants.md (unspec_volatile_element): Add
2447 UNSPEC_VOLATILE_RELAX_GROUP.
2448 * config/nds32/nds32-relax-opt.c: New file.
2449 * config/nds32/nds32-predicates.c
2450 (nds32_symbol_load_store_p): New function.
2451 * config/nds32/nds32-protos.h
2452 (nds32_symbol_load_store_p): Declare function.
2453 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2454 * config/nds32/nds32.c
2455 (nds32_register_pass): New function to register pass.
2456 (nds32_register_passes): New function to register passes.
2457 * config/nds32/nds32.md (relax_group): New pattern.
2458 * config/nds32/nds32.opt (mrelax-hint): New option.
2459 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2461 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2463 * config/nds32/t-nds32: Modify files dependency.
2465 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2467 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2468 (PROFILE_HOOK): Define its implementation.
2470 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2472 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2473 type and 32-bit size.
2475 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2477 PR middle-end/85090
2478 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2479 (V_128_256): New mode iterator.
2480 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2481 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2482 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2483 of V.
2484 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2485 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2487 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2489 PR target/83315
2490 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2491 NaN inputs correctly.
2493 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2495 PR target/80546
2496 * config/rs6000/vsx.md (??r): New mode attribute.
2497 (*vsx_mov<mode>_64bit): Use it.
2498 (*vsx_mov<mode>_32bit): Likewise.
2500 2018-03-30 Martin Sebor <msebor@redhat.com>
2502 PR tree-optimization/84818
2503 * builtins.c (check_access): Use warning_n.
2505 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2507 PR target/83822
2508 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2509 condition.
2510 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2511 condition.
2513 2018-03-30 Julia Koval <julia.koval@intel.com>
2515 PR target/84413
2516 * x86-tune.def (movx, partial_reg_dependency): Enable for
2517 m_SKYLAKE_AVX512.
2519 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2521 PR inline-asm/84985
2522 * lra-constraints.c (process_alt_operands): Move setting
2523 this_alternative_matches below.
2525 2018-03-29 Martin Liska <mliska@suse.cz>
2527 PR lto/84995.
2528 * doc/invoke.texi: Document how LTO works with debug info.
2529 Describe auto-load support of binutils. Mention 'x86-64'
2530 as valid option value of -march option.
2532 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2534 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2536 PR c/85094
2537 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2538 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2539 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2540 checking.
2542 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2544 PR target/84912
2545 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2546 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2547 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2548 for RS6000_BTM_POWERPC64.
2549 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2550 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2551 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2552 definition.
2553 (DIVDE): Use it.
2554 (DIVDEU): Likewise.
2556 2018-03-28 Carl Love <cel@us.ibm.com>
2558 Revert
2559 2017-09-27 Carl Love <cel@us.ibm.com>
2561 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2562 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2563 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2564 fctiw instruction.
2566 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2568 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2569 instead of __vector bool.
2570 (_mm_max_pu8): Likewise.
2571 (_mm_min_pi16): Likewise.
2573 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2575 PR target/84912
2576 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2577 (DIVWEUO): Likewise.
2578 (DIVDEO): Likewise.
2579 (DIVDEUO): Likewise.
2580 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2581 DIVWEUO and DIVDEUO.
2582 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2583 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2584 (div_extend): Likewise.
2585 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2586 builtin function.
2587 (__builtin_divweuo): Likewise.
2588 (__builtin_divdeo): Likewise.
2589 (__builtin_divdeuo): Likewise.
2591 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2593 PR target/85095
2594 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2595 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2597 PR tree-optimization/82004
2598 * gimple-match-head.c (optimize_pow_to_exp): New function.
2599 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2600 Don't fold to exp if optimize_pow_to_exp is false.
2602 2018-03-28 Martin Liska <mliska@suse.cz>
2604 PR other/84819
2605 * calls.c (initialize_argument_information): Fix trailing space.
2606 * common.opt: Fix typo and provide better explanation for
2607 -fsanitize-coverage option.
2608 * config/i386/i386.opt: Fix typo.
2610 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2611 Martin Liska <mliska@suse.cz>
2613 PR sanitizer/85081
2614 * gimplify.c (asan_poison_variable): Don't do the check for
2615 gimplify_omp_ctxp here.
2616 (gimplify_decl_expr): Do it here.
2617 (gimplify_target_expr): Likewise.
2619 2018-03-28 Martin Liska <mliska@suse.cz>
2621 PR target/84988
2622 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2623 chkp_type_bounds_count if MPX is not enabled.
2625 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2627 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2629 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2631 PR target/84914
2632 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2633 function to create the function decl for complex long double
2634 multiply and divide for -mabi=ieeelongdouble.
2635 (init_float128_ieee): Call it.
2637 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2639 PR target/85044
2640 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2641 -fcf-protection=branch -mibt.
2642 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2644 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2646 PR target/81863
2647 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2649 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2651 PR target/85056
2652 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2653 extern array declarations.
2655 2018-03-27 Richard Biener <rguenther@suse.de>
2657 PR middle-end/84067
2658 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2659 explicit single_use checks.
2661 2018-03-27 Richard Biener <rguenther@suse.de>
2663 PR tree-optimization/85082
2664 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2665 Valueize the VUSE.
2667 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2669 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2670 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2671 Turn on fasynchronous-unwind-tables and funwind-tables.
2673 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2675 PR target/85073
2676 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2677 (*bmi_blsr_<mode>_ccz): Ditto.
2679 2018-03-26 Tom de Vries <tom@codesourcery.com>
2681 PR tree-optimization/85063
2682 * omp-general.c (offloading_function_p): New function. Factor out
2683 of ...
2684 * omp-offload.c (pass_omp_target_link::gate): ... here.
2685 * omp-general.h (offloading_function_p): Declare.
2686 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2687 with attribute omp declare target for offloading functions.
2689 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2691 PR tree-optimization/84005
2692 * tree-data-ref.h (get_base_for_alignment): Declare.
2693 * tree-data-ref.c (get_base_for_alignment_1): New function.
2694 (get_base_for_alignment): Likewise.
2695 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2696 get_base_for_alignment to find a suitable base object, instead
2697 of always using drb->base_address.
2699 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2701 PR inline-asm/85022
2702 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2703 known size by default.
2705 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2707 PR inline-asm/85030
2708 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2709 and non BLKmode operands.
2711 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2713 PR target/85026
2714 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2715 Clean up attributes.
2717 2018-03-23 Richard Biener <rguenther@suse.de>
2719 PR debug/85020
2720 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2721 we are going to emit early debug for LTO.
2723 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2725 PR inline-asm/85034
2726 * function.c (match_asm_constraints_1): Don't optimize if input
2727 doesn't satisfy general_operand predicate for output's mode.
2729 PR inline-asm/85022
2730 * alias.c (write_dependence_p): Don't require for x_canonicalized
2731 non-VOIDmode if x has VOIDmode.
2733 PR sanitizer/85029
2734 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2735 just don't try to optimize it rather than assert it never happens.
2737 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2739 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2740 macro expansions for definition of ST_INTERNAL_<mode> and
2741 LD_INTERNAL_<mode> builtins.
2742 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2743 Remove prototype.
2744 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2745 function.
2746 (altivec_expand_st_builtin): Likewise.
2747 (altivec_expand_builtin): Remove calls to deleted functions.
2748 (rs6000_address_for_altivec): Delete this function.
2749 * config/rs6000/vector.md: Remove expands for
2750 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2752 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2754 PR target/84826
2755 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2756 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2757 re-computing once computed.
2758 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2759 (arm_init_machine_status): Initialize
2760 machine->static_chain_stack_bytes.
2762 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2764 PR target/84760
2765 * doc/extend.texi: Add four new prototypes for vec_ld.
2766 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2767 definitions for more logical presentation.
2768 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2769 entries for V1TI variants of __builtin_altivec_ld builtin.
2770 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2771 handling of V1TI variant of LVX icode pattern.
2772 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2773 (rs6000_gimple_fold_builtin): Likewise.
2774 (altivec_init_builtins): Add code to define
2775 __builtin_altivec_lvx_v1ti function.
2777 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2779 PR inline-asm/84941
2780 * function.c (match_asm_constraints_1): Don't do the optimization
2781 if input isn't a REG, SUBREG, MEM or constant.
2783 2018-03-22 Tom de Vries <tom@codesourcery.com>
2785 PR tree-optimization/84956
2786 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2787 bb_has_abnormal_pred.
2789 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2791 PR sanitizer/85018
2792 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2793 DECL_INITIAL (decl) to decl at the end.
2794 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2795 adjust the comment.
2797 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2799 * doc/extend.texi (__builtin_tgmath): Document when complex
2800 integer types are treated as _Complex _Float64.
2802 2018-03-21 Tom de Vries <tom@codesourcery.com>
2804 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2806 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2808 PR tree-optimization/84960
2809 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2810 if it is ENTRY block, move them into single succ of ENTRY in that case.
2812 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2814 PR tree-optimization/84811
2815 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2816 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2817 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2818 (known_subrange_p): Update accordingly. Cast each value involved
2819 in the size comparison, rather than casting the result of the
2820 subtraction.
2822 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2824 PR tree-optimization/84982
2825 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2826 by flipping the least significant bit rather than all bits from
2827 bitpos to bitpos + bitsize - 1.
2829 2018-03-21 Nathan Sidwell <nathan@acm.org>
2831 * doc/extend.texi (Deprecated Features): Remove mention of
2832 long-deleted deprecations.
2834 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2836 PR jit/84288
2837 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2838 * configure: Regenerate.
2840 2018-03-21 Tom de Vries <tom@codesourcery.com>
2842 PR tree-optimization/83126
2843 * tree-parloops.c (num_phis): New function.
2844 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2846 2018-03-21 Nathan Sidwell <nathan@acm.org>
2848 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2849 mention anon-struct/union members and trailing attributes.
2851 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2853 PR tree-optimization/84969
2854 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2855 builtin memset partitions if they set different rhs values.
2857 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2859 PR rtl-optimization/84989
2860 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2861 VEC_DUPLICATE with scalar result mode.
2863 2018-03-21 Martin Liska <mliska@suse.cz>
2865 PR ipa/84963
2866 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2867 not intended return statement.
2869 2018-03-21 Martin Liska <mliska@suse.cz>
2871 PR target/84988
2872 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2873 (chkp_find_bound_slots_1): Limit number of iterations.
2875 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2877 PR target/84838
2878 * Minor grammar fixes for x86 options.
2880 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2882 PR debug/84875
2883 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2884 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2886 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2888 PR target/83789
2889 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2890 (altivec_lvx_<mode>_1op): Likewise.
2891 (altivec_stvx_<mode>_2op): Likewise.
2892 (altivec_stvx_<mode>_1op): Likewise.
2893 (altivec_lvx_<VM2:mode>): New define_expand.
2894 (altivec_stvx_<VM2:mode>): Likewise.
2895 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2896 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2897 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2898 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2899 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2900 (rs6000_gen_lvx): Likewise.
2901 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2902 (altivec_expand_stv_builtin): Likewise.
2903 (altivec_expand_builtin): Likewise.
2904 * config/rs6000/vector.md: Likewise.
2906 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2908 PR target/82518
2909 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2910 BYTES_BIG_ENDIAN.
2912 2018-03-20 Richard Biener <rguenther@suse.de>
2914 PR target/84986
2915 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2916 sign-conversions as zero, fall back to standard scalar_stmt
2917 cost for the rest.
2919 2018-03-20 Martin Liska <mliska@suse.cz>
2921 PR ipa/84825
2922 * predict.c (rebuild_frequencies): Handle case when we have
2923 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2925 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2927 PR target/84990
2928 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2929 flag_section_anchors.
2930 * varasm.c (use_blocks_for_decl_p): Remove hack for
2931 dw2_force_const_mem.
2933 PR target/84845
2934 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2935 to ...
2936 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2937 be created, use lowpart_subreg of operands[0] rather than operands[0]
2938 itself.
2939 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2940 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2941 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2942 and n constraint instead of aarch64_shift_imm_di and Usd.
2943 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2944 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2946 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2948 PR target/82989
2949 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2950 to favor GPR over NEON registers.
2951 (<shift>di3_neon): Likewise.
2953 2018-03-20 Tom de Vries <tom@codesourcery.com>
2955 PR target/84952
2956 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2957 (nvptx_process_pars): Emit bar.sync asap and alap.
2959 2018-03-20 Tom de Vries <tom@codesourcery.com>
2961 PR target/84954
2962 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2963 seen_label if seen_label is already set.
2965 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2967 PR target/84945
2968 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2969 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2970 Use 1U instead of 1. Formatting fixes.
2972 PR c/84953
2973 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2974 instead of TREE_TYPE (s1) for the return value.
2976 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2978 PR tree-optimization/84946
2979 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2980 bitsize + bitsize in poly_uint64 rather than poly_int64.
2982 PR sanitizer/78651
2983 * dwarf2asm.c: Include fold-const.c.
2984 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2985 of decl rather than decl itself.
2987 PR rtl-optimization/84643
2988 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2990 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2992 PR sanitizer/78651
2993 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2994 calling assemble_variable.
2996 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2998 PR target/81647
2999 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3000 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3002 2018-03-19 Jim Wilson <jimw@sifive.com>
3004 PR bootstrap/84856
3005 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3006 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3007 (riscv_first_stack_step): Likewise.
3008 (riscv_option_override): Use STACK_BOUNDARY instead of
3009 MIN_STACK_BOUNDARY.
3010 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3011 MIN_STACK_BOUNDARY.
3012 (BIGGEST_ALIGNMENT): Set to 128.
3013 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3014 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3015 STACK_BOUNDARY.
3017 2018-03-19 Richard Biener <rguenther@suse.de>
3019 PR tree-optimization/84933
3020 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3021 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3023 2018-03-19 Richard Biener <rguenther@suse.de>
3025 PR tree-optimization/84859
3026 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3027 (cond_if_else_store_replacement): Perform sinking operation on
3028 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3029 Generalize what a BB with a single eligible store is.
3031 2018-03-19 Richard Biener <rguenther@suse.de>
3033 PR tree-optimization/84929
3034 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3035 chrec_is_positive against non-chrec arg.
3037 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3039 PR target/84711
3040 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3042 2018-03-18 Martin Liska <mliska@suse.cz>
3044 PR rtl-optimization/84635
3045 * regrename.c (build_def_use): Use matches_mode only when
3046 matches >= 0.
3048 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3050 PR tree-optimization/84913
3051 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3052 vectorize chains of COND_EXPRs.
3054 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3056 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3058 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3060 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3062 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3064 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3066 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3067 Kito Cheng <kito.cheng@gmail.com>
3069 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3070 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3071 (nds32_adjust_reg_alloc_order): New function.
3072 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3074 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3076 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3077 nds32_print_operand, nds32_print_operand_address): Use
3078 HOST_WIDE_INT_PRINT_DEC instead.
3080 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3082 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3084 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3086 PR target/84902
3087 * config/i386/i386.c (initial_ix86_tune_features,
3088 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3089 unsigned long long.
3090 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3091 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3092 rather than 1u << ix86_tune. Formatting fix.
3093 (ix86_option_override_internal): Change ix86_arch_mask from
3094 unsigned int to unsigned HOST_WIDE_INT, initialize to
3095 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3096 (ix86_function_specific_restore): Likewise.
3098 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3100 PR target/84899
3101 * postreload.c (reload_combine_recognize_pattern): Perform
3102 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3103 truncate_int_for_mode the result for the destination's mode.
3105 PR c/84909
3106 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3107 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3108 Likewise.
3109 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3110 Likewise.
3112 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3114 PR target/84876
3115 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3116 regno_allocno_class_array and sorted_pseudos.
3117 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3118 insns where regno is used.
3120 2018-03-16 Martin Liska <mliska@suse.cz>
3122 PR ipa/84833
3123 * multiple_target.c (create_dispatcher_calls): Redirect
3124 reference in the symbol table.
3126 2018-03-16 Martin Liska <mliska@suse.cz>
3128 PR ipa/84722
3129 * multiple_target.c (create_dispatcher_calls): Redirect also
3130 an alias.
3132 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3134 PR c++/79937
3135 PR c++/82410
3136 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3137 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3138 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3140 2018-03-16 Julia Koval <julia.koval@intel.com>
3142 * doc/invoke.texi (Skylake Server): Add CLWB.
3143 Cannonlake): Remove CLWB.
3145 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3147 PR tree-optimization/84841
3148 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3149 1 << 3.
3150 (FLOAT_ONE_CONST_TYPE): Define.
3151 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3152 (sort_by_operand_rank): Put entries with higher constant_type last
3153 rather than first to match comments.
3155 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3157 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3158 split predicate.
3160 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3162 PR c++/79085
3163 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3164 check and use address of target always.
3166 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3168 PR target/84574
3169 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3170 (indirect_thunk_bnd_needed): Likewise.
3171 (indirect_thunks_used): Likewise.
3172 (indirect_thunks_bnd_used): Likewise.
3173 (indirect_return_needed): New.
3174 (indirect_return_bnd_needed): Likewise.
3175 (output_indirect_thunk_function): Add a bool argument for
3176 function return.
3177 (output_indirect_thunk_function): Don't generate alias for
3178 function return thunk.
3179 (ix86_code_end): Call output_indirect_thunk_function to generate
3180 function return thunks.
3181 (ix86_output_function_return): Set indirect_return_bnd_needed
3182 and indirect_return_needed instead of indirect_thunk_bnd_needed
3183 and indirect_thunk_needed.
3185 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3187 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3188 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3189 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3191 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3192 Paul Hua <paul.hua.gm@gmail.com>
3194 PR c/84852
3195 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3197 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3199 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3200 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3201 resp. SFmode cases.
3203 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3205 PR target/84711
3206 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3207 instead of GET_MODE_SIZE when comparing Units.
3209 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3211 PR target/68256
3212 * varasm.c (hash_section): Return an unchangeble hash value
3213 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3214 Return !aarch64_can_use_per_function_literal_pools_p ().
3216 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3218 PR target/84860
3219 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3220 rather than address of cmode as last argument to prepare_cmp_insn.
3222 2018-03-15 Julia Koval <julia.koval@intel.com>
3224 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3225 F_AVX512VNNI, F_AVX512BITALG): New.
3227 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3229 PR target/83451
3230 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3231 insn for floating-point loads and stores.
3233 2018-03-14 Carl Love <cel@us.ibm.com>
3235 * config/rs6000/rs6000-c.c: Add macro definitions for
3236 ALTIVEC_BUILTIN_VEC_PERMXOR.
3237 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3238 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3239 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3240 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3241 UNSPEC_VPERMXOR.
3242 * config/doc/extend.texi: Add prototypes for vec_permxor.
3244 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3246 PR c/84852
3247 * diagnostic-show-locus.c (class layout_point): Convert m_line
3248 from int to linenum_type.
3249 (line_span::comparator): Use linenum "compare" function when
3250 comparing line numbers.
3251 (test_line_span): New function.
3252 (layout_range::contains_point): Convert param "row" from int to
3253 linenum_type.
3254 (layout_range::intersects_line_p): Likewise.
3255 (layout::will_show_line_p): Likewise.
3256 (layout::print_source_line): Likewise.
3257 (layout::should_print_annotation_line_p): Likewise.
3258 (layout::print_annotation_line): Likewise.
3259 (layout::print_leading_fixits): Likewise.
3260 (layout::annotation_line_showed_range_p): Likewise.
3261 (struct line_corrections): Likewise for field m_row.
3262 (line_corrections::line_corrections): Likewise for param "row".
3263 (layout::print_trailing_fixits): Likewise.
3264 (layout::get_state_at_point): Likewise.
3265 (layout::get_x_bound_for_row): Likewise.
3266 (layout::print_line): Likewise.
3267 (diagnostic_show_locus): Likewise for locals "last_line" and
3268 "row".
3269 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3270 * input.c (selftest::test_linenum_comparisons): New function.
3271 (selftest::input_c_tests): Call it.
3272 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3273 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3274 * selftest.h (ASSERT_GT): New macro.
3275 (ASSERT_GT_AT): New macro.
3276 (ASSERT_LT): New macro.
3277 (ASSERT_LT_AT): New macro.
3279 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3281 PR rtl-optimization/84780
3282 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3284 2018-03-14 Martin Liska <mliska@suse.cz>
3286 * tree.c (record_node_allocation_statistics): Use
3287 get_stats_node_kind.
3288 (get_stats_node_kind): New function extracted from
3289 record_node_allocation_statistics.
3290 (free_node): Use get_stats_node_kind.
3292 2018-03-14 Richard Biener <rguenther@suse.de>
3294 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3295 that the value-set of ANTIC_IN doesn't grow.
3297 Revert
3298 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3299 member.
3300 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3301 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3303 2018-03-14 Julia Koval <julia.koval@intel.com>
3305 * config.gcc (icelake-client, icelake-server): New.
3306 (icelake): Remove.
3307 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3308 (initial_ix86_arch_features): Ditto.
3309 (PTA_SKYLAKE): Add SGX.
3310 (PTA_ICELAKE): Remove.
3311 (PTA_ICELAKE_CLIENT): New.
3312 (PTA_ICELAKE_SERVER): New.
3313 (ix86_option_override_internal): Split up icelake on icelake client and
3314 icelake server.
3315 (get_builtin_code_for_version): Ditto.
3316 (fold_builtin_cpu): Ditto.
3317 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3318 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3319 * config/i386/i386.h (processor_type): Ditto.
3320 * doc/invoke.texi: Ditto.
3322 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3324 PR sanitizer/83392
3325 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3326 INTEGER_CST offset, add it together with bitpos / 8 and
3327 sign extend based on POINTER_SIZE.
3329 PR target/84844
3330 Revert
3331 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3333 PR target/78090
3334 * config/i386/constraints.md (Yc): New register constraint.
3335 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3336 Use Yc constraint for alternative 2 of operand 0. Remove
3337 preferred_for_speed attribute.
3339 2018-03-14 Richard Biener <rguenther@suse.de>
3341 PR tree-optimization/84830
3342 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3343 with the old one to avoid oscillations.
3345 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3347 PR target/83712
3348 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3349 pseudos.
3350 (assign_by_spills): Return a flag of reload assignment failure.
3351 Do not process the reload assignment failures. Do not spill other
3352 reload pseudos if they has the same reg class. Update n if
3353 necessary.
3354 (lra_assign): Add a return arg. Set up from the result of
3355 assign_by_spills call.
3356 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3357 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3358 usage_insns if it is not NULL.
3359 (spill_hard_reg_in_range): New function.
3360 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3361 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3362 function prototypes.
3363 (lra_assign): Change prototype.
3364 * lra.c (lra): Add code to deal with fails by splitting hard reg
3365 live ranges.
3367 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3369 * config/riscv/riscv.opt (mrelax): New option.
3370 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3371 "norelax" when riscv_mrelax is disabled.
3372 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3374 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3376 PR target/84743
3377 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3378 reassociation for int modes.
3380 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3382 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3383 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3384 for big-endian.
3385 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3386 * config/aarch64/aarch64-sve.md
3387 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3388 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3389 (*extend<mode><Vwide>2): Rename to...
3390 (aarch64_sve_extend<mode><Vwide>2): ...this.
3391 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3392 renaming the old pattern to...
3393 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3394 unsigned packs.
3395 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3396 define_expand, renaming the old pattern to...
3397 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3398 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3399 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3400 account when deciding which SVE instruction the optab should use.
3401 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3403 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3405 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3406 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3407 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3408 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3409 (tlsdesc_small_<mode>): Turn a define_expand and use
3410 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3411 (tlsdesc_small_advsimd_<mode>): ...this.
3412 (tlsdesc_small_sve_<mode>): New pattern.
3414 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3416 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3417 (UNSPEC_UMUL_HIGHPART): New constants.
3418 (MUL_HIGHPART): New int iteraor.
3419 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3420 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3421 define_expand.
3422 (*<su>mul<mode>3_highpart): New define_insn.
3424 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3426 PR lto/84805
3427 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3428 incomplete types.
3430 2018-03-13 Martin Liska <mliska@suse.cz>
3432 PR ipa/84658.
3433 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3434 vector.
3435 (sem_item_optimizer::~sem_item_optimizer): Release it.
3436 (sem_item_optimizer::merge_classes): Register variable aliases.
3437 (sem_item_optimizer::fixup_pt_set): New function.
3438 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3439 * ipa-icf.h: Declare new variables and functions.
3441 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3443 PR middle-end/84834
3444 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3445 integer_pow2p@2 and test integer_pow2p in condition.
3446 (A < 0 ? C : 0): Similarly for @1.
3448 PR middle-end/84831
3449 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3450 characters starting at p contain '\0' character, don't look beyond
3451 that.
3453 PR target/84827
3454 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3455 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3457 PR target/84828
3458 * reg-stack.c (change_stack): Change update_end var from int to
3459 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3460 also call set_block_for_insn on the newly added insns and rescan.
3462 PR target/84786
3463 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3464 on the last operand.
3466 PR c++/84704
3467 * tree.c (stabilize_reference_1): Return save_expr (e) for
3468 STATEMENT_LIST even if it doesn't have side-effects.
3470 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3472 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3474 2018-03-12 Renlin Li <renlin.li@arm.com>
3476 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3477 aarch64_output_scalar_simd_mov_immediate.
3479 2018-03-12 Martin Sebor <msebor@redhat.com>
3481 PR tree-optimization/83456
3482 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3483 for perfectly overlapping calls to memcpy.
3484 (gimple_fold_builtin_memory_chk): Same.
3485 (gimple_fold_builtin_strcpy): Handle no-warning.
3486 (gimple_fold_builtin_stxcpy_chk): Same.
3487 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3489 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3491 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3492 parameter. Use it for SFmode.
3493 (rs6000_function_arg_advance_1): Adjust.
3494 (rs6000_function_arg): Adjust.
3495 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3497 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3499 PR rtl-optimization/84169
3500 PR rtl-optimization/84780
3501 * combine.c (can_combine_p): Check for a 2-insn combination whether
3502 the destination register is used between the two insns, too.
3504 2018-03-12 Richard Biener <rguenther@suse.de>
3506 PR tree-optimization/84803
3507 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3508 for refs DR analysis didn't process.
3510 2018-03-12 Richard Biener <rguenther@suse.de>
3512 PR tree-optimization/84777
3513 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3514 force-vectorize loops ignore whether we are optimizing for size.
3516 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3518 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3519 (TARGET_MD_ASM_ADJUST): Define.
3521 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3522 Kito Cheng <kito.cheng@gmail.com>
3523 Chung-Ju Wu <jasonwucj@gmail.com>
3525 * config/nds32/nds32.c (nds32_compute_stack_frame,
3526 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3527 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3528 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3529 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3530 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3531 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3532 * config/nds32/nds32.md (prologue, epilogue): Use macro
3533 NDS32_V3PUSH_AVAILABLE_P to do checking.
3535 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3537 PR debug/58150
3538 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3539 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3540 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3541 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3542 addition of most attributes on !orig_type_die or the attribute not
3543 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3545 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3546 Chung-Ju Wu <jasonwucj@gmail.com>
3548 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3549 __NDS32_VH__ macro.
3550 * config/nds32/nds32.opt (mvh): New option.
3552 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3553 Chung-Ju Wu <jasonwucj@gmail.com>
3555 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3556 function.
3557 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3558 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3559 definition.
3561 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3562 Chung-Ju Wu <jasonwucj@gmail.com>
3564 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3565 function.
3566 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3567 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3569 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3570 Kito Cheng <kito.cheng@gmail.com>
3571 Chung-Ju Wu <jasonwucj@gmail.com>
3573 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3574 UNSPEC_FFMISM and UNSPEC_FLMISM.
3575 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3576 for ffb, ffmism and flmism.
3577 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3578 (unspec_ffmism): Ditto.
3579 (unspec_flmism): Ditto.
3580 (nds32_expand_builtin_impl): Check if string extension is available.
3581 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3582 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3584 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3586 Reverting patch:
3587 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3589 PR target/83712
3590 * lra-assigns.c (assign_by_spills): Return a flag of reload
3591 assignment failure. Do not process the reload assignment
3592 failures. Do not spill other reload pseudos if they has the same
3593 reg class.
3594 (lra_assign): Add a return arg. Set up from the result of
3595 assign_by_spills call.
3596 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3597 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3598 usage_insns if it is not NULL.
3599 (spill_hard_reg_in_range): New function.
3600 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3601 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3602 function prototypes.
3603 (lra_assign): Change prototype.
3604 * lra.c (lra): Add code to deal with fails by splitting hard reg
3605 live ranges.
3607 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3609 PR target/84807
3610 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3612 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3614 PR debug/84620
3615 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3616 (dw_val_node): Add val_symbolic_view.
3617 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3618 (symview_upper_bound): New.
3619 (new_line_info_table): Initialize symviews_since_reset.
3620 (dwarf2out_source_line): Count symviews_since_reset and set
3621 symview_upper_bound.
3622 (dw_val_equal_p): Handle symview.
3623 (add_AT_symview): New.
3624 (print_dw_val): Handle symview.
3625 (attr_checksum, attr_checksum_ordered): Likewise.
3626 (same_dw_val_p, size_of_die): Likewise.
3627 (value_format, output_die): Likewise.
3628 (add_high_low_attributes): Use add_AT_symview for entry_view.
3629 (dwarf2out_finish): Reset symview_upper_bound, clear
3630 zero_view_p.
3632 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3634 PR target/83969
3635 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3636 Add strict argument and use it.
3637 (rs6000_split_multireg_move): Update for new strict argument.
3638 (mem_operand_gpr): Disallow all non-offsettable addresses.
3639 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3641 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3643 PR target/84772
3644 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3645 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3646 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3648 PR c++/84767
3649 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3650 decl, use remap_type if we want to use the type.
3652 2018-03-09 Martin Sebor <msebor@redhat.com>
3654 PR tree-optimization/84526
3655 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3656 Remove dead code.
3657 (builtin_access::generic_overlap): Be prepared to handle non-array
3658 base objects.
3660 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3662 PR rtl-optimization/84682
3663 * lra-constraints.c (process_address_1): Check is_address flag
3664 for address constraints.
3665 (process_alt_operands): Likewise.
3666 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3667 preprocess_constraints.
3668 * recog.h (preprocess_constraints): Add oploc parameter.
3669 Adjust callers.
3670 * recog.c (preprocess_constraints): Test address_operand for
3671 CT_ADDRESS constraints.
3673 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3675 PR target/83712
3676 * lra-assigns.c (assign_by_spills): Return a flag of reload
3677 assignment failure. Do not process the reload assignment
3678 failures. Do not spill other reload pseudos if they has the same
3679 reg class.
3680 (lra_assign): Add a return arg. Set up from the result of
3681 assign_by_spills call.
3682 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3683 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3684 usage_insns if it is not NULL.
3685 (spill_hard_reg_in_range): New function.
3686 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3687 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3688 function prototypes.
3689 (lra_assign): Change prototype.
3690 * lra.c (lra): Add code to deal with fails by splitting hard reg
3691 live ranges.
3693 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3695 PR target/83193
3696 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3697 Accept complain bool parameter. Only emit errors if it is true.
3698 (arm_parse_cpu_option_name): Likewise.
3699 (arm_target_thumb_only): Adjust callers of the above.
3700 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3701 prototype to take a default true bool parameter.
3702 (arm_parse_arch_option_name): Likewise.
3704 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3705 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3707 PR jit/64089
3708 PR jit/84288
3709 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3710 * configure: Regenerate.
3711 * configure.ac ("linker --version-script option"): New.
3712 ("linker soname option"): New.
3714 2018-03-09 Richard Biener <rguenther@suse.de>
3716 PR tree-optimization/84775
3717 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3718 immediate uses of predicate stmts and mark them modified.
3720 Revert
3721 PR tree-optimization/84178
3722 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3723 to caller.
3724 (version_loop_for_if_conversion): Delay update_ssa call.
3725 (tree_if_conversion): Delay update_ssa until after predicate
3726 insertion.
3728 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3730 PR target/84763
3731 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3732 when the function accesses prior frames.
3734 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3736 PR debug/84456
3737 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3738 gen_llsym, otherwise call maybe_gen_llsym.
3740 PR inline-asm/84742
3741 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3742 has ',' character inside of it.
3744 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3746 PR target/84748
3747 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3748 as clobbering CC_REGNUM.
3750 2018-03-08 Richard Biener <rguenther@suse.de>
3752 PR middle-end/84552
3753 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3754 (follow_copies_to_constant): Do not follow SSA names registered
3755 for update.
3757 2018-03-08 Richard Biener <rguenther@suse.de>
3759 PR tree-optimization/84178
3760 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3761 to caller.
3762 (version_loop_for_if_conversion): Delay update_ssa call.
3763 (tree_if_conversion): Delay update_ssa until after predicate
3764 insertion.
3766 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3768 PR tree-optimization/84178
3769 * tree-if-conv.c (release_bb_predicate): Remove the
3770 the assertion that the stmts have NULL use_ops.
3771 Discard the statements, asserting that they haven't
3772 yet been added to a BB.
3774 2018-03-08 Richard Biener <rguenther@suse.de>
3776 PR tree-optimization/84746
3777 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3778 (phi_translate): Pass in destination ANTIC_OUT set.
3779 (phi_translate_1): Likewise. For a simplified result lookup
3780 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3781 (phi_translate_set): Adjust.
3782 (do_pre_regular_insertion): Likewise.
3783 (do_pre_partial_partial_insertion): Likewise.
3785 2018-03-08 Martin Liska <mliska@suse.cz>
3787 PR gcov-profile/84735
3788 * doc/gcov.texi: Document usage of profile files.
3789 * gcov-io.h: Document changes in the format.
3791 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3793 PR debug/84404
3794 PR debug/84408
3795 * dwarf2out.c (struct dw_line_info_table): Update comments for
3796 view == -1.
3797 (FORCE_RESET_NEXT_VIEW): New.
3798 (FORCE_RESETTING_VIEW_P): New.
3799 (RESETTING_VIEW_P): Check for -1 too.
3800 (ZERO_VIEW_P): Likewise.
3801 (new_line_info_table): Force-reset next view.
3802 (dwarf2out_begin_function): Likewise.
3803 (dwarf2out_source_line): Simplify zero_view_p initialization.
3804 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3805 view directly. Omit view when omitting .loc at line 0.
3807 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3809 PR tree-optimization/84740
3810 * tree-switch-conversion.c (process_switch): Call build_constructors
3811 only if info.phi_count is non-zero.
3813 PR tree-optimization/84739
3814 * tree-tailcall.c (find_tail_calls): Check call arguments against
3815 DECL_ARGUMENTS (current_function_decl) rather than
3816 DECL_ARGUMENTS (func) when checking for tail recursion.
3818 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3820 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3821 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3822 Volker Reichelt's entry and add entries for people that perform
3823 GCC fuzzy testing and report numerous bugs.
3825 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3827 PR target/82411
3828 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3829 readonly data in sdata, if that is disabled.
3830 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3831 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3832 -mreadonly-in-sdata option.
3834 2018-03-07 Martin Sebor <msebor@redhat.com>
3836 PR tree-optimization/84468
3837 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3838 basic block when looking for nul assignment.
3840 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3842 PR target/84277
3843 * except.h (output_function_exception_table): Adjust prototype.
3844 * except.c (output_function_exception_table): Remove FNNAME parameter
3845 and add SECTION parameter. Ouput one part of the table at a time.
3846 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3847 the first part of the exception table and emit unwind directives.
3848 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3849 (i386_pe_seh_cold_init): Likewise.
3850 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3851 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3852 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3853 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3854 * config/i386/winnt.c: Include except.h.
3855 (struct seh_frame_state): Add reg_offset, after_prologue and
3856 in_cold_section fields.
3857 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3858 (i386_pe_seh_cold_init): New function.
3859 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3860 to seh->in_cold_section.
3861 (seh_emit_push): Record the offset of the push.
3862 (seh_emit_save): Record the offet of the save.
3863 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3864 Test seh->after_prologue to disregard the epilogue.
3865 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3866 (i386_pe_end_cold_function): New function.
3868 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3870 PR fortran/84565
3871 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3872 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3874 PR c++/84704
3875 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3876 on tmp_var.
3877 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3878 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3880 PR middle-end/84723
3881 * multiple_target.c: Include tree-inline.h and intl.h.
3882 (expand_target_clones): Diagnose and fail if node->definition and
3883 !tree_versionable_function_p (node->decl).
3885 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3887 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3888 sprint_ul.
3889 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3890 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3891 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3893 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3895 PR target/84710
3896 * combine.c (try_combine): Use reg_or_subregno instead of handling
3897 just paradoxical SUBREGs and REGs.
3899 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3901 * config/arc/arc.c (arc_finalize_pic): Remove function.
3902 (arc_must_save_register): We use single base PIC register, remove
3903 checks to save/restore the PIC register.
3904 (arc_expand_prologue): Likewise.
3905 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3906 Remove.
3907 (arc_verify_short): Likewise.
3908 (arc_attr_type): Likewise.
3909 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3910 (walk_stores): Likewise.
3911 (arc_address_cost): Make it static.
3912 (arc_verify_short): Likewise.
3913 (branch_dest): Likewise.
3914 (arc_attr_type): Likewise.
3915 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3916 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3917 (arc_final_prescan_insn): Remove inserting the nops due to
3918 hardware hazards. It is done in reorg step.
3919 (insn_length_variant_t): Remove.
3920 (insn_length_parameters_t): Likewise.
3921 (arc_insn_length_parameters): Likewise.
3922 (arc_get_insn_variants): Likewise.
3923 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3925 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3927 PR inline-asm/84683
3928 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3929 assertion failure.
3931 PR tree-optimization/84687
3932 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3933 on new_node->decl.
3934 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3936 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3938 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3939 Rename to ppc_speculation_barrier.
3940 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3941 __builtin_ppc_speculation_barrier.
3943 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3945 PR target/84700
3946 * combine.c (combine_simplify_rtx): Don't try to simplify if
3947 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3948 are equal to x.
3950 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3952 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3953 to 32 bytes when compiling for POWER9.
3955 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3957 PR target/84564
3958 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3959 regparm >= 3 with no arg reg available also for calls with
3960 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3962 PR target/84524
3963 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3964 orig,vex.
3965 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3967 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3969 PR target/84264
3970 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3972 2018-03-05 Richard Biener <rguenther@suse.de>
3974 PR tree-optimization/84486
3975 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3976 When inserting a __builtin_assume_aligned call set the LHS
3977 SSA name alignment info accordingly.
3979 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3981 PR tree-optimization/84114
3982 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3983 Avoid reassociation of FLOAT_MODE addition.
3985 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3987 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3988 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3989 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3990 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3991 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3992 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3993 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3994 and -mwbnoinvd.
3995 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3996 __builtin_ia32_wbinvd): New builtins.
3997 (SPECIAL_ARGS2): New.
3998 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3999 (SPECIAL_ARGS2): New.
4000 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4001 (ix86_valid_target_attribute_inner_p): Ditto.
4002 (ix86_init_mmx_sse_builtins): Add special_args2.
4003 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4004 TARGET_WBNOINVD_P): New.
4005 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4006 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4007 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4008 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4009 * config/i386/pconfigintrin.h: New file.
4010 * config/i386/wbnoinvdintrin.h: Ditto.
4011 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4012 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4014 2018-03-05 Richard Biener <rguenther@suse.de>
4016 PR tree-optimization/84670
4017 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4018 member.
4019 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4020 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4021 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4022 doesn't grow if all successors (recursively) were visited at least
4023 once.
4025 2018-03-05 Richard Biener <rguenther@suse.de>
4027 PR tree-optimization/84650
4028 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4029 if executed in the loop pipeline.
4031 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4033 * doc/configfiles.texi (Configuration Files): Move info about
4034 conditionalizing $target-protos.h to...
4035 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4036 differs from $target-protos.h.
4038 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4039 Chung-Ju Wu <jasonwucj@gmail.com>
4041 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4042 * config/nds32/nds32-multiple.md (setmemsi): Define.
4043 * config/nds32/nds32-memory-manipulation.c
4044 (nds32_gen_dup_4_byte_to_word_value): New.
4045 (emit_setmem_word_loop): New.
4046 (emit_setmem_byte_loop): New.
4047 (nds32_expand_setmem_loop): New.
4048 (nds32_expand_setmem_loop_v3m): New.
4049 (nds32_expand_setmem_unroll): New.
4050 (nds32_expand_setmem): New.
4052 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4053 Chung-Ju Wu <jasonwucj@gmail.com>
4055 * config/nds32/nds32-memory-manipulation.c
4056 (nds32_emit_load_store): New.
4057 (nds32_emit_post_inc_load_store): New.
4058 (nds32_emit_mem_move): New.
4059 (nds32_emit_mem_move_block): New.
4060 (nds32_expand_movmemsi_loop_unknown_size): New.
4061 (nds32_expand_movmemsi_loop_known_size): New.
4062 (nds32_expand_movmemsi_loop): New.
4063 (nds32_expand_movmemsi_unroll): New.
4064 (nds32_expand_movmemqi): Rename ...
4065 (nds32_expand_movmemsi): ... to this.
4066 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4067 (movmemsi): ... to this.
4068 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4069 (nds32_expand_movmemsi): ... to this.
4071 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4072 Monk Chiang <sh.chiang04@gmail.com>
4073 Chung-Ju Wu <jasonwucj@gmail.com>
4075 * config/nds32/nds32-protos.h
4076 (nds32_expand_load_multiple): New arguments.
4077 (nds32_expand_store_multiple): Ditto.
4078 (nds32_valid_multiple_load_store): Rename ...
4079 (nds32_valid_multiple_load_store_p): ... to this.
4080 * config/nds32/nds32-memory-manipulation.c
4081 (nds32_expand_load_multiple): Refine implementation.
4082 (nds32_expand_store_multiple): Ditto.
4083 * config/nds32/nds32-multiple.md
4084 (load_multiple): Update nds32_expand_load_multiple interface.
4085 (store_multiple): Update nds32_expand_store_multiple interface.
4086 * config/nds32/nds32-predicates.c
4087 (nds32_valid_multiple_load_store): Rename ...
4088 (nds32_valid_multiple_load_store_p): ... to this and refine
4089 implementation.
4090 * config/nds32/predicates.md
4091 (nds32_load_multiple_and_update_address_operation): New predicate.
4092 (nds32_store_multiple_and_update_address_operation): New predicate.
4094 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4095 Chung-Ju Wu <jasonwucj@gmail.com>
4097 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4098 (combo): New attribute.
4099 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4101 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4103 * config/nds32/nds32.opt: Change -mcmodel= default value.
4105 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4106 Monk Chiang <sh.chiang04@gmail.com>
4107 Chung-Ju Wu <jasonwucj@gmail.com>
4109 * config/nds32/constants.md (unspec_element): New enum.
4110 * config/nds32/constraints.md (Umw): New constraint.
4111 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4112 * config/nds32/nds32-intrinsic.md: Likewise.
4113 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4114 (nds32_valid_smw_lwm_base_p): New.
4115 (nds32_output_smw_single_word): New.
4116 (nds32_output_lmw_single_word): New.
4117 (nds32_expand_unaligned_load): New.
4118 (nds32_expand_unaligned_store): New.
4119 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4120 (nds32_output_smw_single_word): Declare.
4121 (nds32_output_lmw_single_word): Declare.
4122 (nds32_expand_unaligned_load): Declare.
4123 (nds32_expand_unaligned_store): Declare.
4124 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4125 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4126 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4127 NDS32_BUILTIN_UASTORE_DW.
4128 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4129 predicate.
4131 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4132 Kito Cheng <kito.cheng@gmail.com>
4133 Chung-Ju Wu <jasonwucj@gmail.com>
4135 * config/nds32/nds32-intrinsic.c
4136 (nds32_expand_builtin_null_ftype_reg): Delete.
4137 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4138 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4139 (nds32_read_argument): New.
4140 (nds32_legitimize_target): Ditto.
4141 (nds32_legitimize_argument): Ditto.
4142 (nds32_check_constant_argument): Ditto.
4143 (nds32_expand_unop_builtin): Ditto.
4144 (nds32_expand_unopimm_builtin): Ditto.
4145 (nds32_expand_binop_builtin): Ditto.
4146 (nds32_builtin_decl_impl): Ditto.
4147 (builtin_description): Ditto.
4148 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4149 (nds32_init_builtins_impl): Ditto.
4150 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4151 (nds32_builtin_decl): New.
4152 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4153 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4155 2018-03-02 Jeff Law <law@redhat.com>
4157 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4158 (redundant_insn, fill_simple_delay_slots): Likewise.
4159 (fill_slots_from_thread): Likewise.
4160 * resource.c (mark_referenced_resources): Likewise.
4161 (mark_set_resources, find_dead_or_set_registers): Likewise.
4163 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4165 * substring-locations.h (format_warning_va): Formatting fix for
4166 ATTRIBUTE_GCC_DIAG.
4167 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4168 argument.
4169 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4170 * substring-locations.c: Include intl.h.
4171 (format_warning_va): Turned into small wrapper around
4172 format_warning_n_va, renamed to ...
4173 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4174 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4175 use ngettext.
4176 (format_warning_at_substring_n): New function.
4177 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4178 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4179 format_warning_at_substring with just a shorter name instead of
4180 const function pointer.
4181 (fmtwarn_n): New function.
4182 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4183 appropriate, get rid of all the fmtstr temporaries, move conditionals
4184 with G_() wrapped string literals directly into fmtwarn arguments,
4185 cast dir.len to (int), formatting fixes.
4187 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4189 * doc/invoke.texi: Remove "Cilk Plus" references.
4191 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4192 Richard Biener <rguenther@suse.de>
4194 PR ipa/84628
4195 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4196 for error or warning attributes if CALL_FROM_THUNK_P is set.
4197 Formatting fixes.
4199 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4201 PR target/56540
4202 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4203 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4205 PR target/56540
4206 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4207 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4209 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4210 instead of -1U in last predictors element's probability member.
4212 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4214 PR ipa/83983
4215 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4216 arguments if they are comparable.
4218 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4220 PR tree-optimization/84634
4221 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4222 masks and masked_loop_p with a single loop_masks, making sure it's
4223 null for bb vectorization.
4225 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4227 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4228 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4229 loop->force_vectorize to check whether there is no alias.
4231 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4233 PR target/84614
4234 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4235 prototypes.
4236 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4237 comments.
4238 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4239 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4240 instead of a loop around prev_real_insn.
4241 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4242 prev_real_insn.
4244 PR inline-asm/84625
4245 * config/i386/i386.c (ix86_print_operand): Use conditional
4246 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4247 zero vector.
4249 2018-03-02 Richard Biener <rguenther@suse.de>
4251 PR tree-optimization/84427
4252 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4253 (bitmap_set_subtract_values): Rewrite to handle multiple
4254 exprs per value.
4255 (clean): Likewise.
4256 (prune_clobbered_mems): Likewise.
4257 (phi_translate): Take edge instead of pred/phiblock.
4258 (phi_translate_1): Likewise.
4259 (phi_translate_set): Likewise. Insert all translated
4260 exprs for a value into the set, keeping possibly multiple
4261 expressions per value.
4262 (compute_antic_aux): Adjust for phi_translate changes.
4263 When intersecting union the expressions and prune those
4264 not in the final value set, keeping possibly multiple
4265 expressions per value. Do not use value-insertion
4266 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4267 all expressions. Add verification that the value-sets
4268 only shrink during iteration.
4269 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4270 (do_pre_regular_insertion): Likewise.
4271 (do_pre_partial_partial_insertion): Likewise.
4273 2018-03-02 Richard Biener <rguenther@suse.de>
4275 PR target/82005
4276 * config/darwin.c (saved_debug_info_level): New static global.
4277 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4278 (darwin_asm_lto_end): Restore debug info generation settings.
4280 2018-03-01 Martin Liska <mliska@suse.cz>
4282 PR sanitizer/82484
4283 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4284 volatile arguments.
4286 2018-03-01 Richard Biener <rguenther@suse.de>
4288 PR debug/84645
4289 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4290 type annotation with LTO when debug was disabled at compile-time.
4292 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4294 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4295 XINT with INTVAL.
4296 (mips_final_postscan_insn): Likewise.
4298 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4300 PR rtl-optimization/84528
4301 * alias.c (init_alias_target): Add commentary.
4302 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4303 a unique base value if the frame pointer is not eliminated
4304 to the stack pointer.
4306 2018-03-01 Tom de Vries <tom@codesourcery.com>
4308 PR rtl-optimization/83327
4309 * lra-int.h (hard_regs_spilled_into): Declare.
4310 * lra.c (hard_regs_spilled_into): Define.
4311 (init_reg_info): Init hard_regs_spilled_into.
4312 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4313 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4314 (process_bb_lives): Handle hard_regs_spilled_into.
4315 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4316 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4318 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4320 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4321 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4322 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4323 * config/rs6000/aix72.h: New file.
4325 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4327 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4328 instead of warning_at with conditional singular and plural messages
4329 where possible.
4331 PR target/52991
4332 * stor-layout.c (update_alignment_for_field): For
4333 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4334 && !DECL_PACKED (field), do the alignment update, just use
4335 only desired_align instead of MAX (type_align, desired_align)
4336 as the alignment.
4337 (place_field): Don't do known_align < desired_align handling
4338 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4339 is non-NULL, instead do it after rli->prev_field handling and
4340 only if not within a bitfield word. For DECL_PACKED (field)
4341 use type_align of BITS_PER_UNIT.
4343 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4345 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4346 superfluous parentheses and trailing spaces.
4348 2018-02-28 Richard Biener <rguenther@suse.de>
4350 PR tree-optimization/84584
4351 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4352 SCoPs with fake exit edge.
4354 2018-02-28 Martin Liska <mliska@suse.cz>
4356 PR testsuite/84597
4357 * timevar.c (timer::print): Fix format to properly print 100%
4358 values.
4360 2018-02-28 Richard Biener <rguenther@suse.de>
4362 PR middle-end/84607
4363 * genmatch.c (capture_info::walk_match): Do not mark
4364 captured expressions without operands as expr_p given
4365 they act more like predicates and should be subject to
4366 "lost tail" side-effect preserving.
4368 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4370 PR rtl-optimization/81611
4371 * auto-inc-dec.c (attempt_change): Move dead note from
4372 mem_insn if it's the next use of regno
4373 (find_address): Take address use of reg holding
4374 non-incremented value. Add parm to limit search to the named
4375 reg only.
4376 (merge_in_block): Attempt to use a mem insn that is the next
4377 use of the original regno.
4379 2018-02-27 Martin Sebor <msebor@redhat.com>
4381 PR c++/83871
4382 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4383 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4385 2018-02-27 Martin Sebor <msebor@redhat.com>
4387 PR translation/84207
4388 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4389 n argument to unsigned HOST_WIDE_INT.
4390 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4391 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4392 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4393 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4395 2018-02-27 Richard Biener <rguenther@suse.de>
4397 PR tree-optimization/84512
4398 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4399 Do not use the estimate returned from record_stmt_cost for
4400 the scalar iteration cost but sum properly using add_stmt_cost.
4402 2018-02-27 Richard Biener <rguenther@suse.de>
4404 PR tree-optimization/84466
4405 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4406 Adjust last change to less strictly validate use operands.
4408 2018-02-27 Martin Liska <mliska@suse.cz>
4410 PR gcov-profile/84548
4411 * gcov.c (process_file): Allow partial overlap and consider it
4412 also as group functions.
4413 (output_lines): Properly calculate range of lines for a group.
4415 2018-02-27 Martin Liska <mliska@suse.cz>
4417 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4418 'ggc' suffixes. Change first column width.
4419 (timer::print): Fix formatting of the column.
4421 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4423 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4424 preserve inline entry blocks for the sake of debug inline
4425 entry point markers alone.
4426 (remove_unused_locals): Suggest in comments a better place to
4427 force the preservation of inline entry blocks that are
4428 otherwise unused, but do not preserve them.
4430 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4432 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4434 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4436 PR target/84039
4437 * config/i386/constraints.md (Bs): Replace
4438 ix86_indirect_branch_register with
4439 TARGET_INDIRECT_BRANCH_REGISTER.
4440 (Bw): Likewise.
4441 * config/i386/i386.md (indirect_jump): Likewise.
4442 (tablejump): Likewise.
4443 (*sibcall_memory): Likewise.
4444 (*sibcall_value_memory): Likewise.
4445 Peepholes of indirect call and jump via memory: Likewise.
4446 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4447 (*sibcall_value_GOT_32): Likewise.
4448 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4449 (GOT_memory_operand): Likewise.
4450 (call_insn_operand): Likewise.
4451 (sibcall_insn_operand): Likewise.
4452 (GOT32_symbol_operand): Likewise.
4453 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4455 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4457 PR rtl-optimization/83496
4458 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4459 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4460 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4461 redundant insn, if any.
4462 (relax_delay_slots): Likewise.
4463 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4465 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4467 PR tree-optimization/83965
4468 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4469 that grouped statements are part of a reduction chain. Return
4470 true if the statement is not marked as a reduction itself but
4471 is part of a group.
4472 (vect_recog_dot_prod_pattern): Don't check whether the statement
4473 is part of a group here.
4474 (vect_recog_sad_pattern): Likewise.
4475 (vect_recog_widen_sum_pattern): Likewise.
4477 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4479 PR debug/84545
4480 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4482 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4484 PR target/84530
4485 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4486 the bool argument.
4487 (ix86_output_indirect_function_return): New prototype.
4488 (ix86_split_simple_return_pop_internal): Likewise.
4489 * config/i386/i386.c (indirect_return_via_cx): New.
4490 (indirect_return_via_cx_bnd): Likewise.
4491 (indirect_thunk_name): Handle return va CX_REG.
4492 (output_indirect_thunk_function): Create alias for
4493 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4494 (ix86_output_indirect_jmp): Remove the bool argument.
4495 (ix86_output_indirect_function_return): New function.
4496 (ix86_split_simple_return_pop_internal): Likewise.
4497 * config/i386/i386.md (*indirect_jump): Don't pass false
4498 to ix86_output_indirect_jmp.
4499 (*tablejump_1): Likewise.
4500 (simple_return_pop_internal): Change it to define_insn_and_split.
4501 Call ix86_split_simple_return_pop_internal to split it for
4502 -mfunction-return=.
4503 (simple_return_indirect_internal): Call
4504 ix86_output_indirect_function_return instead of
4505 ix86_output_indirect_jmp.
4507 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4509 PR bootstrap/84405
4510 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4511 memset and value initialization afterwards.
4513 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4515 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4517 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4519 PR target/84521
4520 * common/config/aarch64/aarch64-common.c
4521 (aarch_option_optimization_table[]): Switch
4522 off fomit-frame-pointer
4524 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4525 Chung-Ju Wu <jasonwucj@gmail.com>
4527 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4528 volatile memory.
4529 (store_multiple): Ditto.
4531 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4533 * config.gcc: Add --with-cpu support for nds32 target.
4534 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4535 * config/nds32/nds32.opt: Add -mcpu= option.
4537 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4539 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4540 isel=yes): Warn for these deprecated options.
4542 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4544 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4545 ISA_2_5_MASKS_EMBEDDED.
4547 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4549 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4550 p->max as pointers rather than using iterative_hash_expr.
4552 2018-02-23 Carl Love <cel@us.ibm.com>
4554 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4555 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4556 BU_P8V_OVERLOAD_2.
4557 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4558 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4559 P8V_BUILTIN_VEC_VUNSIGNED2.
4561 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4563 PR target/81572
4564 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4565 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4566 LRA_UNKNOWN_ALT.
4567 * lra-constraints.c (curr_insn_transform): Set up
4568 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4569 LRA_UNKNOWN_ALT.
4570 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4571 * lra-eliminations.c (spill_pseudos): Ditto.
4572 (process_insn_for_elimination): Ditto.
4573 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4574 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4575 LRA_NON_CLOBBERED_ALT.
4577 2018-02-22 Martin Sebor <msebor@redhat.com>
4579 PR tree-optimization/84480
4580 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4581 to maybe_diag_stxncpy_trunc. Call it.
4582 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4583 from gimple_fold_builtin_strcpy. Print inlining stack.
4584 (handle_builtin_stxncpy): Print inlining stack.
4585 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4587 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4589 PR target/84176
4590 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4591 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4592 and -fcheck-pointer-bounds are used together.
4593 (indirect_thunk_prefix): New enum.
4594 (indirect_thunk_need_prefix): New function.
4595 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4596 "_nt" instead of "_bnd" for NOTRACK prefix.
4597 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4598 (output_indirect_thunk_function): Likewise.
4599 (): Likewise.
4600 (ix86_code_end): Update output_indirect_thunk_function calls.
4601 (ix86_output_indirect_branch_via_reg): Replace
4602 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4603 (ix86_output_indirect_branch_via_push): Likewise.
4604 (ix86_output_function_return): Likewise.
4605 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4606 incompatible with -fcf-protection=branch and
4607 -fcheck-pointer-bounds.
4609 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4611 PR target/83335
4612 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4613 Change gcc_assert call to output_operand_lossage.
4615 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4617 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4619 2018-02-22 DJ Delorie <dj@redhat.com>
4620 Sebastian Perta <sebastian.perta@renesas.com>
4621 Oleg Endo <olegendo@gcc.gnu.org>
4623 * config/rx/rx.c (rx_rtx_costs): New function.
4624 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4626 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4628 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4630 2018-02-22 Martin Liska <mliska@suse.cz>
4632 PR driver/83193
4633 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4634 Add "native" as a possible value.
4636 2018-02-22 Martin Liska <mliska@suse.cz>
4638 PR driver/83193
4639 * config/i386/i386.c (ix86_option_override_internal):
4640 Add "native" as a possible value for -march and -mtune.
4642 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4644 PR target/84502
4645 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4646 to all type variants.
4648 PR tree-optimization/84503
4649 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4650 width as info->bitpos + info->bitsize - start.
4651 (merged_store_group::merge_overlapping): Simplify width computation.
4652 (check_no_overlap): New function.
4653 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4654 start + width and last_order of the group, fail if check_no_overlap
4655 fails.
4656 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4657 to group if check_no_overlap fails.
4659 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4661 * config/rs6000/altivec.md: Delete contraint arguments to
4662 define_expand, define_split, and define_peephole2, and in
4663 define_insn_and_split if always unused.
4664 * config/rs6000/darwin.md: Ditto.
4665 * config/rs6000/dfp.md: Ditto.
4666 * config/rs6000/rs6000.md: Ditto.
4667 * config/rs6000/sync.md: Ditto.
4668 * config/rs6000/vector.md: Ditto.
4669 * config/rs6000/vsx.md: Ditto.
4671 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4673 * config/rs6000/altivec.md: Write output control strings as braced
4674 blocks instead of double-quoted strings.
4675 * config/rs6000/darwin.md: Ditto.
4676 * config/rs6000/rs6000.md: Ditto.
4677 * config/rs6000/vector.md: Ditto.
4678 * config/rs6000/vsx.md: Ditto.
4680 2018-02-21 Jason Merrill <jason@redhat.com>
4682 PR c++/84314 - ICE with templates and fastcall attribute.
4683 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4685 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4687 * ipa-cp.c (determine_versionability): Fix comment typos.
4689 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4691 PR c/84229
4692 * ipa-cp.c (determine_versionability): Do not version functions caling
4693 va_arg_pack.
4695 2018-02-21 Martin Liska <mliska@suse.cz>
4697 PR driver/83193
4698 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4699 Add "native" as a possible value.
4700 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4701 the macro when native cpu detection is available.
4703 2018-02-21 Martin Liska <mliska@suse.cz>
4705 PR driver/83193
4706 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4707 Add "native" as a possible value.
4708 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4709 when native cpu detection is available.
4711 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4712 Martin Sebor <msebor@redhat.com>
4714 PR tree-optimization/84478
4715 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4716 false.
4717 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4718 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4719 support which is conservatively correct, for 2 only stay conservative
4720 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4721 argument to the 2 argument get_range_strlen, adjust 6 arg
4722 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4723 false.
4724 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4725 (gimple_fold_builtin_strlen): Pass true as last argument to
4726 get_range_strlen.
4728 2018-02-20 Martin Sebor <msebor@redhat.com>
4730 PR middle-end/84095
4731 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4732 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4733 (builtin_memref::builtin_memref): Factor out parts into
4734 set_base_and_offset and call it.
4736 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4738 PR middle-end/84406
4739 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4740 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4741 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4742 search at the associated MODE_INT.
4744 2018-02-20 Jeff Law <law@redhat.com>
4746 PR middle-end/82123
4747 PR tree-optimization/81592
4748 PR middle-end/79257
4749 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4750 for range data rather than using global data.
4751 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4752 range data rather than using global data.
4753 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4754 pass it to children as needed.
4755 (struct directive::fmtresult): Similarly.
4756 (struct directive::set_width): Similarly.
4757 (struct directive::set_precision): Similarly.
4758 (format_integer, format_directive, parse_directive): Similarly.
4759 (format_none): Accept unnamed vr_values parameter.
4760 (format_percent, format_floating, format_character): Similarly.
4761 (format_string, format_plain): Similarly.
4762 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4763 the EVRP range analyzer for range data rather than using global data.
4764 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4765 gimple-ssa-evrp-analyze.h
4766 (class sprintf_dom_walker): Add after_dom_children member function.
4767 Add evrp_range_analyzer member.
4768 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4769 range analyzer as needed.
4770 (sprintf_dom_walker::after_dom_children): New member function.
4771 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4772 if not optimizing.
4773 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4774 (evrp_range_analyzer::pop_to_marker): Likewise.
4776 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4778 PR tree-optimization/84419
4779 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4780 with the required type if its current type is compatible but
4781 different.
4783 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4785 PR middle-end/82004
4786 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4787 after vectorization.
4789 2018-02-20 Martin Liska <mliska@suse.cz>
4791 PR driver/83193
4792 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4793 possible values if we don't have a hint.
4795 2018-02-20 Martin Liska <mliska@suse.cz>
4797 PR c/84310
4798 PR target/79747
4799 * final.c (shorten_branches): Build align_tab array with one
4800 more element.
4801 * opts.c (finish_options): Add alignment option limit check.
4802 (MAX_CODE_ALIGN): Likewise.
4803 (MAX_CODE_ALIGN_VALUE): Likewise.
4804 * doc/invoke.texi: Document maximum allowed option value for
4805 all -falign-* options.
4807 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4809 PR target/84146
4810 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4811 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4812 * var-tracking.c (emit_note_insn_var_location): Remove all references
4813 to NOTE_INSN_CALL_ARG_LOCATION.
4814 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4815 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4816 Use copy_rtx_if_shared.
4817 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4818 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4819 (dwarf2out_var_location): Remove handling of
4820 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4821 on call_insn.
4822 * final.c (final_scan_insn): Remove all references to
4823 NOTE_INSN_CALL_ARG_LOCATION.
4824 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4825 before dumping final insns.
4826 * except.c (emit_note_eh_region_end): Remove all references to
4827 NOTE_INSN_CALL_ARG_LOCATION.
4828 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4829 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4830 * config/arc/arc.c (hwloop_optimize): Likewise.
4831 * config/arm/arm.c (create_fix_barrier): Likewise.
4832 * config/s390/s390.c (s390_chunkify_start): Likewise.
4833 * config/sh/sh.c (find_barrier): Likewise.
4834 * config/i386/i386.c (rest_of_insert_endbranch,
4835 ix86_seh_fixup_eh_fallthru): Likewise.
4836 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4837 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4838 * config/frv/frv.c (frv_function_prologue): Likewise.
4839 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4840 reg note.
4841 (note_outside_basic_block_p): Remove all references to
4842 NOTE_INSN_CALL_ARG_LOCATION.
4843 * gengtype.c (adjust_field_rtx_def): Likewise.
4844 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4845 Likewise.
4846 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4847 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4849 PR c++/84444
4850 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4851 is ADDR_EXPR.
4853 PR tree-optimization/84452
4854 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4855 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4856 is NULL.
4858 2018-02-19 Martin Liska <mliska@suse.cz>
4860 PR sanitizer/82183
4861 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4863 2018-02-19 Martin Liska <mliska@suse.cz>
4864 Richard Sandiford <richard.sandiford@linaro.org>
4866 PR tree-optimization/82491
4867 * gimple-fold.c (get_base_constructor): Make earlier bail out
4868 to prevent ubsan.
4870 2018-02-19 Carl Love <cel@us.ibm.com>
4872 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4873 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4874 BU_P8V_OVERLOAD_1.
4875 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4876 P8V_BUILTIN_VEC_NEG.
4878 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4880 * config/rl78/rl78.md (movdf): New define expand.
4882 2018-02-19 Martin Liska <mliska@suse.cz>
4884 PR other/80589
4885 * doc/invoke.texi: Fix typo.
4886 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4888 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4890 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4891 handle rs6000_single_float and rs6000_double_float specially for
4892 e500 family CPUs.
4894 2018-02-16 Jeff Law <law@redhat.com>
4896 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4897 (pop_regs): Use it.
4899 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4901 PR ipa/84425
4902 * ipa-inline.c (inline_small_functions): Fix a typo.
4904 2018-02-16 Nathan Sidwell <nathan@acm.org>
4906 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4908 2018-02-16 Carl Love <cel@us.ibm.com>
4910 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4911 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4912 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4913 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4914 expansion to P8V_BUILTIN_VEC_FLOAT2.
4916 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4918 PR rtl-optimization/70023
4919 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4920 src_regno into account.
4922 2018-02-16 Carl Love <cel@us.ibm.com>
4924 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4925 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4926 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4927 * config/rs6000/rs6000.c: Remove case statements for
4928 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4929 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4930 and P9V_BUILTIN_VEC_VINSERT4B.
4931 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4932 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4933 * config/rs6000/vsx.md:
4934 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4935 vec_insert4b.
4937 2018-02-16 Carl Love <cel@us.ibm.com>
4939 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4940 vec_insert4b.
4941 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4942 definitions.
4943 * config/rs6000/rs6000-c.c: Add the definitions for
4944 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4945 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4946 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4947 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4948 definition for insert4b and define insn *insert3b_internal.
4949 * doc/extend.texi: Add documentation for vec_extract4b.
4951 2018-02-16 Nathan Sidwell <nathan@acm.org>
4953 * doc/extend.texi (Backwards Compatibility): Mention friend
4954 injection. Note for-scope is deprecated.
4955 * doc/invoke.texi (-ffriend-injection): Deprecate.
4957 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4959 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4960 that moved to I2, also allow destinations that are a paradoxical
4961 subreg (instead of a normal reg).
4963 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4965 PR target/83831
4966 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4967 to QImode.
4969 2018-02-16 Richard Biener <rguenther@suse.de>
4971 PR tree-optimization/84037
4972 PR tree-optimization/84016
4973 PR target/82862
4974 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4975 Adjust vec_construct for the fact we need additional higher latency
4976 128bit inserts for AVX256 and AVX512 vector builds.
4977 (ix86_add_stmt_cost): Scale vector construction cost for
4978 elementwise loads.
4980 2018-02-16 Richard Biener <rguenther@suse.de>
4982 PR tree-optimization/84417
4983 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4984 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4985 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4987 2018-02-16 Martin Liska <mliska@suse.cz>
4989 PR sanitizer/84307
4990 * internal-fn.def (ASAN_CHECK): Set proper flags.
4991 (ASAN_MARK): Likewise.
4993 2018-02-16 Julia Koval <julia.koval@intel.com>
4995 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4996 from PTA_CANNONLAKE.
4998 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5000 PR target/84272
5001 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5002 Use ++iter rather than iter++ for std::list iterators.
5003 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5004 defer deleting them until all nodes in the forest are processed. Do
5005 free even leaf nodes. Change to_process into auto_vec.
5007 PR bootstrap/84405
5008 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5009 * vec.h (vec_default_construct): Use memset instead of placement new
5010 if BROKEN_VALUE_INITIALIZATION is defined.
5011 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5012 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5013 is defined.
5015 PR rtl-optimization/83723
5016 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5017 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5018 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5019 recursive calls.
5020 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5021 callers.
5022 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5024 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5026 PR rtl-optimization/81443
5027 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5028 from inner REGs to paradoxical SUBREGs.
5030 2018-02-16 Richard Biener <rguenther@suse.de>
5032 PR tree-optimization/84399
5033 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5034 For operands we can analyze at their definition make sure we can
5035 analyze them at each use as well.
5037 2018-02-16 Richard Biener <rguenther@suse.de>
5039 PR tree-optimization/84190
5040 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5041 volatile accesses if the decl isn't volatile.
5043 2018-02-15 Jason Merrill <jason@redhat.com>
5045 PR c++/84314 - ICE with templates and fastcall attribute.
5046 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5047 TYPE_CANONICAL on an existing type.
5049 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5051 PR tree-optimization/84383
5052 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5053 dstoff nor call operand_equal_p if dstbase is NULL.
5055 PR tree-optimization/84334
5056 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5057 also a CONSTANT_CLASS_P, punt.
5059 2018-02-14 Jim Wilson <jimw@sifive.com>
5061 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5062 first SMALL_OPERAND check. New local min_second_step. Move assert
5063 to where locals are set. Add TARGET_RVC support.
5064 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5066 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5068 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5070 2018-02-14 Martin Sebor <msebor@redhat.com>
5072 PR tree-optimization/83698
5073 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5074 arrays constrain the offset range to their bounds.
5075 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5076 (builtin_access::overlap): Avoid setting the size of overlap if it's
5077 already been set.
5078 (maybe_diag_overlap): Also consider arrays when deciding what values
5079 of offsets to include in diagnostics.
5081 2018-02-14 Martin Sebor <msebor@redhat.com>
5083 PR c/84108
5084 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5085 that correspond to the kind of a declaration.
5087 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5089 PR target/83984
5090 * config/pa/pa.md: Load address of PIC label using the linkage table
5091 if the label is nonlocal.
5093 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5095 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5096 warning message if user requests -maltivec=be.
5097 * doc/invoke.texi: Document deprecation of -maltivec=be.
5099 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5101 PR target/84220
5102 * config/rs6000/rs6000-c.c: Update definitions for
5103 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5104 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5106 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5108 PR target/84239
5109 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5110 add _get_ssp intrinsics. Remove argument from
5111 __builtin_ia32_rdssp[d|q].
5112 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5113 * config/i386/i386-builtin.def: Remove argument from
5114 __builtin_ia32_rdssp[d|q].
5115 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5116 ix86_expand_special_args_builtin for _rdssp[d|q].
5117 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5118 Clear register before usage.
5119 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5120 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5122 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5124 PR tree-optimization/84357
5125 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5126 operand 1 of an ARRAY_REF too.
5128 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5130 PR target/83831
5131 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5132 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5133 declarations.
5134 (set_of_reg): New struct.
5135 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5136 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5137 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5138 functions.
5139 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5140 Split into bitclr, bitset, bitinvert patterns if appropriate.
5141 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5142 use rx_fuse_in_memory_bitop.
5143 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5144 to named insn, correct maximum insn length.
5146 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5148 PR target/79242
5149 * machmode.def: Define a complex mode for PARTIAL_INT.
5150 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5151 MODE_PARTIAL_INT.
5152 * doc/rtl.texi: Document CSPImode.
5153 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5154 handling.
5155 (msp430_hard_regno_nregs_with_padding): Likewise.
5157 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5159 PR target/84279
5160 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5162 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5164 PR rtl-optimization/84169
5165 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5166 we generated a parallel as new i3 and we split that to new i2 and i3
5167 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5168 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5169 those to i2, not i1. Partially rewrite this scan code.
5171 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5173 PR c/82210
5174 * stor-layout.c (place_field): For variable length fields, adjust
5175 offset_align afterwards not just based on the field's alignment,
5176 but also on the size.
5178 PR middle-end/84309
5179 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5180 of exps and logs in the use_exp2 case.
5182 2018-02-13 Jeff Law <law@redhat.com>
5184 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5185 entry for "vector".
5187 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5188 ARGS as unused.
5190 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5192 PR debug/84342
5193 PR debug/84319
5194 * common.opt (gas-loc-support, gas-locview-support): New.
5195 (ginline-points, ginternal-reset-location-views): New.
5196 * doc/invoke.texi: Document them. Use @itemx where intended.
5197 (gvariable-location-views): Adjust.
5198 * target.def (reset_location_view): New.
5199 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5200 (TARGET_RESET_LOCATION_VIEW): New.
5201 * doc/tm.texi: Rebuilt.
5202 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5203 (dwarf2out_default_as_locview_support): New.
5204 (output_asm_line_debug_info): Use option variables.
5205 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5206 (output_loc_list): Likewise.
5207 (add_high_low_attributes): Check option variables.
5208 Don't output entry view attribute in strict mode.
5209 (gen_inlined_subroutine_die): Check option variables.
5210 (dwarf2out_inline_entry): Likewise.
5211 (init_sections_and_labels): Likewise.
5212 (dwarf2out_early_finish): Likewise.
5213 (maybe_reset_location_view): New, from...
5214 (dwarf2out_var_location): ... here. Call it.
5215 * debug.h (dwarf2out_default_as_loc_support): Declare.
5216 (dwarf2out_default_as_locview_support): Declare.
5217 * hooks.c (hook_int_rtx_insn_0): New.
5218 * hooks.h (hook_int_rtx_insn_0): Declare.
5219 * toplev.c (process_options): Take -gas-loc-support and
5220 -gas-locview-support from dwarf2out. Enable
5221 -gvariable-location-views by default only with locview
5222 assembler support. Enable -ginternal-reset-location-views by
5223 default only if the target defines the corresponding hook.
5224 Enable -ginline-points by default if location views are
5225 enabled; force it disabled if statement frontiers are
5226 disabled.
5227 * tree-inline.c (expand_call_inline): Check option variables.
5228 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5230 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5232 PR tree-optimization/84321
5233 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5234 handling. Also check whether the anti-range contains any values
5235 that satisfy the mask; switch to a VR_RANGE if not.
5237 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5239 PR sanitizer/84340
5240 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5242 2018-02-13 Martin Jambor <mjambor@suse.cz>
5244 PR c++/83990
5245 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5246 of call statements, also set location of a load to a temporary.
5248 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5250 * config/rl78/rl78.c (add_vector_labels): New function.
5251 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5252 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5253 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5254 which checks that no arguments are passed.
5255 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5256 * doc/extend.texi: Documentation for the new attribute.
5258 2018-02-13 Andreas Schwab <schwab@suse.de>
5260 * config/riscv/linux.h (CPP_SPEC): Define.
5262 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5264 PR target/84335
5265 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5266 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5267 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5268 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5269 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5270 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5271 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5272 temporarily for AES and PCLMUL builtins.
5274 PR tree-optimization/84339
5275 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5276 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5277 Formatting fixes.
5279 PR middle-end/84309
5280 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5281 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5282 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5283 inline function.
5284 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5285 inline function.
5286 * omp-simd-clone.h: New file.
5287 * omp-simd-clone.c: Include omp-simd-clone.h.
5288 (expand_simd_clones): No longer static.
5289 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5290 cgraph.h and omp-simd-clone.h.
5291 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5292 (vect_recog_widen_shift_pattern): Formatting fix.
5293 (vect_pattern_recog_1): Don't check optab for calls.
5295 PR target/84336
5296 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5297 operands[2] into a REG before using gen_lowpart on it.
5299 2018-02-12 Jeff Law <law@redhat.com>
5301 PR target/83760
5302 * config/sh/sh.c (find_barrier): Consider a sibling call
5303 a barrier as well.
5305 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5306 successfully back substituting a reg.
5308 2018-02-12 Richard Biener <rguenther@suse.de>
5310 PR tree-optimization/84037
5311 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5312 parameter, move visited init to caller.
5313 (vect_slp_analyze_operations): Separate cost from validity
5314 check, initialize visited once for all instances.
5315 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5316 for all instances.
5317 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5318 out an assert.
5319 (vect_model_promotion_demotion_cost): Likewise.
5320 (vectorizable_bswap): Guard cost modeling with !slp_node
5321 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5322 SLP stmts.
5323 (vectorizable_call): Likewise.
5324 (vectorizable_conversion): Likewise.
5325 (vectorizable_assignment): Likewise.
5326 (vectorizable_shift): Likewise.
5327 (vectorizable_operation): Likewise.
5328 (vectorizable_store): Likewise.
5329 (vectorizable_load): Likewise.
5330 (vectorizable_condition): Likewise.
5331 (vectorizable_comparison): Likewise.
5333 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5335 PR sanitizer/84307
5336 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5337 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5338 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5340 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5342 PR middle-end/83665
5343 * params.def (inline-min-speedup): Increase from 8 to 15.
5344 (max-inline-insns-auto): Decrease from 40 to 30.
5345 * ipa-split.c (consider_split): Add some buffer for function to
5346 be considered inlining candidate.
5347 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5348 default values.
5350 2018-02-12 Richard Biener <rguenther@suse.de>
5352 PR tree-optimization/84037
5353 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5354 matched stmts if we cannot swap the non-matched ones.
5356 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5358 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5359 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5360 _mm_maskz_scalef_round_ss): New intrinsics.
5361 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5362 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5363 __builtin_ia32_scalefss_round): Remove.
5364 (__builtin_ia32_scalefsd_mask_round,
5365 __builtin_ia32_scalefss_mask_round): New intrinsics.
5366 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5367 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5368 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5369 "<round_constraint>")): Changed to ...
5370 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5371 "<round_scalar_constraint>")): ... this.
5372 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5373 %0, %1, %2<round_op3>}"): Changed to ...
5374 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5375 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5376 %2<round_scalar_mask_op3>}"): ... this.
5377 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5379 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5381 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5382 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5383 (_mm_maskz_sqrt_round_ss): New intrinsics.
5384 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5385 (__builtin_ia32_sqrtsd_mask_round)
5386 (__builtin_ia32_sqrtss_mask_round): New builtins.
5387 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5388 (__builtin_ia32_sqrtss_round): Remove.
5389 (__builtin_ia32_sqrtsd_mask_round)
5390 (__builtin_ia32_sqrtss_mask_round): New builtins.
5391 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5392 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5393 ((match_operand:VF_128 1 "vector_operand"
5394 "xBm,<round_constraint>")): Changed to ...
5395 ((match_operand:VF_128 1 "vector_operand"
5396 "xBm,<round_scalar_constraint>")): ... this.
5397 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5398 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5399 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5400 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5401 %<iptr>1<round_scalar_mask_op3>}): ... this.
5402 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5403 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5405 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5407 PR target/84266
5408 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5409 Cast vec_cmpeq result to correct type.
5410 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5411 Cast vec_cmpgt result to correct type.
5413 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5415 * final.c (final_scan_insn_1): Renamed from...
5416 (final_scan_insn): ... this. New wrapper, to recover
5417 seen from the outermost call in recursive ones.
5418 * config/sparc/sparc.c (output_return): Drop seen from call.
5419 (output_sibcall): Likewise.
5420 * config/visium/visium.c (output_branch): Likewise.
5422 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5424 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5425 function label.
5427 2018-02-10 Alan Modra <amodra@gmail.com>
5429 PR target/84300
5430 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5431 Specify LR as an input.
5433 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5435 PR sanitizer/83987
5436 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5437 remove_member_access_dummy_vars): New functions.
5438 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5439 lower_omp_1, execute_lower_omp): Use them.
5441 PR rtl-optimization/84308
5442 * shrink-wrap.c (spread_components): Release todo vector.
5444 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5446 PR rtl-optimization/57193
5447 * ira-color.c (struct allocno_color_data): Add member
5448 conflict_allocno_hard_prefs.
5449 (update_conflict_allocno_hard_prefs): New.
5450 (bucket_allocno_compare_func): Add a preference based on
5451 conflict_allocno_hard_prefs.
5452 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5453 (color_allocnos): Remove a dead code. Initiate
5454 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5456 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5458 PR target/84226
5459 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5460 constraint from =wa to wa. Avoid a subreg on the output operand,
5461 instead use a pseudo and subreg it in a move.
5462 (p9_xxbrd_<mode>): Changed to ...
5463 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5464 (p9_xxbrd_v2df): New expander.
5465 (p9_xxbrw_<mode>): Changed to ...
5466 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5467 (p9_xxbrw_v4sf): New expander.
5469 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5471 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5473 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5475 PR target/83926
5476 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5477 multiply in 32-bit mode.
5478 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5479 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5480 mode.
5482 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5484 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5485 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5486 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5487 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5489 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5491 PR lto/84213
5492 * dwarf2out.c (is_trivial_indirect_ref): New function.
5493 (dwarf2out_late_global_decl): Do not generate a location
5494 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5495 and that are not defined in the current unit.
5497 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5499 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5500 instead of a libcall for UNORDERED.
5502 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5504 PR target/82641
5505 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5506 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5508 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5510 PR target/PR84295
5511 * config/s390/s390.c (s390_set_current_function): Invoke
5512 s390_indirect_branch_settings also if fndecl didn't change.
5514 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5516 * config/rs6000/rs6000.md (blockage): Set length to zero.
5518 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5520 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5522 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5524 PR sanitizer/84285
5525 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5526 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5527 -static-lib*san.
5529 PR debug/84252
5530 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5531 PARALLEL incoming that failed vt_get_decl_and_offset check.
5533 PR middle-end/84237
5534 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5535 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5536 TREE_READONLY bit.
5537 (get_variable_section): For decls in named .bss* sections pass true as
5538 second argument to bss_initializer_p.
5540 2018-02-09 Marek Polacek <polacek@redhat.com>
5541 Jakub Jelinek <jakub@redhat.com>
5543 PR c++/83659
5544 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5545 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5546 Sync some changes from cxx_fold_indirect_ref.
5548 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5550 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5551 markers.
5552 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5553 (BLOCK_INLINE_ENTRY_LABEL): New.
5554 (dwarf2out_var_location): Disregard inline entry markers.
5555 (inline_entry_data): New struct.
5556 (inline_entry_data_hasher): New hashtable type.
5557 (inline_entry_data_hasher::hash): New.
5558 (inline_entry_data_hasher::equal): New.
5559 (inline_entry_data_table): New variable.
5560 (add_high_low_attributes): Add DW_AT_entry_pc and
5561 DW_AT_GNU_entry_view attributes if a pending entry is found
5562 in inline_entry_data_table. Add old entry_pc attribute only
5563 if debug nonbinding markers are disabled.
5564 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5565 markers are enabled.
5566 (block_within_block_p, dwarf2out_inline_entry): New.
5567 (dwarf2out_finish): Check that no entries remained in
5568 inline_entry_data_table.
5569 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5570 (final_scan_insn, notice_source_line): Likewise.
5571 (rest_of_clean_state): Skip inline entry markers.
5572 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5573 markers.
5574 * gimple.c (gimple_build_debug_inline_entry): New.
5575 * gimple.h (enum gimple_debug_subcode): Add
5576 GIMPLE_DEBUG_INLINE_ENTRY.
5577 (gimple_build_debug_inline_entry): Declare.
5578 (gimple_debug_inline_entry_p): New.
5579 (gimple_debug_nonbind_marker_p): Adjust.
5580 * insn-notes.def (INLINE_ENTRY): New.
5581 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5582 inline entry marker notes.
5583 (print_insn): Likewise.
5584 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5585 (INSN_DEBUG_MARKER_KIND): Likewise.
5586 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5587 * tree-inline.c (expand_call_inline): Build and insert
5588 debug_inline_entry stmt.
5589 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5590 inline entry blocks early, if nonbind markers are enabled.
5591 (dump_scope_block): Dump fragment info.
5592 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5593 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5594 (gimple_build_debug_inline_entry): New.
5595 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5596 Enable/disable inline entry points too.
5597 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5598 (DEBUG_INSN): Describe inline entry markers.
5600 * common.opt (gvariable-location-views): New.
5601 (gvariable-location-views=incompat5): New.
5602 * config.in: Rebuilt.
5603 * configure: Rebuilt.
5604 * configure.ac: Test assembler for view support.
5605 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5606 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5607 * dwarf2out.c (var_loc_view): New typedef.
5608 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5609 (dwarf2out_locviews_in_attribute): New.
5610 (dwarf2out_locviews_in_loclist): New.
5611 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5612 (enum dw_line_info_opcode): Add LI_adv_address.
5613 (struct dw_line_info_table): Add view.
5614 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5615 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5616 (zero_view_p): New variable.
5617 (ZERO_VIEW_P): New macro.
5618 (output_asm_line_debug_info): New.
5619 (struct var_loc_node): Add view.
5620 (add_AT_view_list, AT_loc_list): New.
5621 (add_var_loc_to_decl): Add view param. Test it against last.
5622 (new_loc_list): Add view params. Record them.
5623 (AT_loc_list_ptr): Handle loc and view lists.
5624 (view_list_to_loc_list_val_node): New.
5625 (print_dw_val): Handle dw_val_class_view_list.
5626 (size_of_die): Likewise.
5627 (value_format): Likewise.
5628 (loc_list_has_views): New.
5629 (gen_llsym): Set vl_symbol too.
5630 (maybe_gen_llsym, skip_loc_list_entry): New.
5631 (dwarf2out_maybe_output_loclist_view_pair): New.
5632 (output_loc_list): Output view list or entries too.
5633 (output_view_list_offset): New.
5634 (output_die): Handle dw_val_class_view_list.
5635 (output_dwarf_version): New.
5636 (output_compilation_unit_header): Use it.
5637 (output_skeleton_debug_sections): Likewise.
5638 (output_rnglists, output_line_info): Likewise.
5639 (output_pubnames, output_aranges): Update version comments.
5640 (output_one_line_info_table): Output view numbers in asm comments.
5641 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5642 Call maybe_gen_llsym.
5643 (loc_list_from_tree_1): Adjust.
5644 (add_AT_location_description): Create view list attribute if
5645 needed, check it's absent otherwise.
5646 (convert_cfa_to_fb_loc_list): Adjust.
5647 (maybe_emit_file): Call output_asm_line_debug_info for test.
5648 (dwarf2out_var_location): Reset views as needed. Precompute
5649 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5650 attribute. Set view.
5651 (new_line_info_table): Reset next view.
5652 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5653 (dwarf2out_source_line): Likewise. Output view resets and labels to
5654 the assembler, or select appropriate line info opcodes.
5655 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5656 (optimize_string_length): Catch it. Adjust.
5657 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5658 dw_val_class_view_list, and remove it if no longer needed.
5659 (hash_loc_list): Hash view numbers.
5660 (loc_list_hasher::equal): Compare them.
5661 (optimize_location_lists): Check whether a view list symbol is
5662 needed, and whether the locview attribute is present, and
5663 whether they match. Remove the locview attribute if no longer
5664 needed.
5665 (index_location_lists): Call skip_loc_list_entry for test.
5666 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5667 Use output_dwarf_version.
5668 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5669 (struct dw_val_node): Add val_view_list.
5670 * final.c (SEEN_NEXT_VIEW): New.
5671 (set_next_view_needed): New.
5672 (clear_next_view_needed): New.
5673 (maybe_output_next_view): New.
5674 (final_start_function): Rename to...
5675 (final_start_function_1): ... this. Take pointer to FIRST,
5676 add SEEN parameter. Emit param bindings in the initial view.
5677 (final_start_function): Reintroduce SEEN-less interface.
5678 (final): Rename to...
5679 (final_1): ... this. Take SEEN parameter. Output final pending
5680 next view at the end.
5681 (final): Reintroduce seen-less interface.
5682 (final_scan_insn): Output pending next view before switching
5683 sections or ending a block. Mark the next view as needed when
5684 outputting variable locations. Notify debug backend of section
5685 changes, and of location view changes.
5686 (rest_of_handle_final): Adjust.
5687 * toplev.c (process_options): Autodetect value for debug variable
5688 location views option. Warn on incompat5 without -gdwarf-5.
5689 * doc/invoke.texi (gvariable-location-views): New.
5690 (gvariable-location-views=incompat5): New.
5691 (gno-variable-location-views): New.
5693 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5695 PR tree-optimization/84136
5696 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5697 that the result of find_edge is non-NULL.
5699 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5701 PR target/83008
5702 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5703 storing integer register in SImode. Fix cost of 256 and 512
5704 byte aligned SSE register store.
5706 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5708 * config/i386/i386.c (ix86_multiplication_cost): Fix
5709 multiplication cost for TARGET_AVX512DQ.
5711 2018-02-08 Marek Polacek <polacek@redhat.com>
5713 PR tree-optimization/84238
5714 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5715 get_range_strlen.
5717 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5719 PR tree-optimization/84265
5720 * tree-vect-stmts.c (vectorizable_store): Don't treat
5721 VMAT_CONTIGUOUS accesses as grouped.
5722 (vectorizable_load): Likewise.
5724 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5726 PR tree-optimization/81635
5727 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5728 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5729 (test_round_for_mask): New functions.
5730 (wide_int_cc_tests): Call test_round_for_mask.
5731 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5732 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5733 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5734 range returned by get_range_info.
5736 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5738 PR ipa/81360
5739 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5740 * symtab.c: Include builtins.h
5741 (symtab_node::output_to_lto_symbol_table_p): Move here
5742 from lto-streamer-out.c:output_symbol_p.
5743 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5744 (output_symbol_p): Move all logic to symtab.c
5745 (produce_symtab): Update.
5747 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5749 * config/s390/s390-opts.h (enum indirect_branch): Define.
5750 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5751 (s390_indirect_branch_via_thunk)
5752 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5753 (enum s390_indirect_branch_type): Define.
5754 * config/s390/s390.c (struct s390_frame_layout, struct
5755 machine_function): Remove.
5756 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5757 (indirect_branch_table_label_no, indirect_branch_table_name):
5758 Define variables.
5759 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5760 (enum s390_indirect_branch_option): Define.
5761 (s390_return_addr_from_memory): New function.
5762 (s390_handle_string_attribute): New function.
5763 (s390_attribute_table): Add new attribute handler.
5764 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5765 (s390_indirect_branch_via_thunk): New function.
5766 (s390_indirect_branch_via_inline_thunk): New function.
5767 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5768 sibling call optimization for non z10 compiles.
5769 (s390_emit_call): Force indirect branch target to be a single
5770 register. Add r1 clobber for non-z10 compiles.
5771 (s390_emit_epilogue): Emit return jump via return_use expander.
5772 (s390_reorg): Handle JUMP_INSNs as execute targets.
5773 (s390_option_override_internal): Perform validity checks for the
5774 new command line options.
5775 (s390_indirect_branch_attrvalue): New function.
5776 (s390_indirect_branch_settings): New function.
5777 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5778 (s390_output_indirect_thunk_function): New function.
5779 (s390_code_end): Implement target hook.
5780 (s390_case_values_threshold): Implement target hook.
5781 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5782 macros.
5783 * config/s390/s390.h (struct s390_frame_layout)
5784 (struct machine_function): Move here from s390.c.
5785 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5786 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5787 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5788 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5789 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5790 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5791 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5792 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5793 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5794 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5795 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5796 (mnemonic attribute): Add values which aren't recognized
5797 automatically.
5798 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5799 pattern for branch conversion. Fix mnemonic attribute.
5800 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5801 indirect branch via thunk if requested.
5802 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5803 ("*indirect_jump"): Disable for branch conversion using out of
5804 line thunks.
5805 ("indirect_jump_via_thunk<mode>_z10")
5806 ("indirect_jump_via_thunk<mode>")
5807 ("indirect_jump_via_inlinethunk<mode>_z10")
5808 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5809 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5810 ("casesi_jump_via_inlinethunk<mode>_z10")
5811 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5812 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5813 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5814 ("*indirect2_jump"): Disable for branch conversion.
5815 ("casesi_jump"): Turn into expander and expand patterns for branch
5816 conversion.
5817 ("return_use"): New expander.
5818 ("*return"): Emit return via thunk and rename it to ...
5819 ("*return<mode>"): ... this one.
5820 * config/s390/s390.opt: Add new options and and enum for the
5821 option values.
5823 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5825 * lra-constraints.c (match_reload): Unconditionally use
5826 gen_lowpart_SUBREG, rather than selecting between that
5827 and equivalent gen_rtx_SUBREG code.
5829 2018-02-08 Richard Biener <rguenther@suse.de>
5831 PR tree-optimization/84233
5832 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5833 changed flag instead of boguously re-using phi_inserted.
5835 2018-02-08 Martin Jambor <mjambor@suse.cz>
5837 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5838 static local variables.
5840 2018-02-08 Richard Biener <rguenther@suse.de>
5842 PR tree-optimization/84278
5843 * tree-vect-stmts.c (vectorizable_store): When looking for
5844 smaller vector types to perform grouped strided loads/stores
5845 make sure the mode is supported by the target.
5846 (vectorizable_load): Likewise.
5848 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5850 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5851 Increase LDP/STP opportunities by adding adjacent callee-saves.
5853 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5855 PR rtl-optimization/84068
5856 PR rtl-optimization/83459
5857 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5859 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5861 PR tree-optimization/84224
5862 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5863 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5864 non-zero arguments.
5866 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5868 PR target/84113
5869 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5870 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5871 count, remove one USE.
5873 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5875 * doc/install.texi (Configuration): Document the
5876 --with-long-double-format={ibm,ieee} PowerPC configuration
5877 options.
5879 PR target/84154
5880 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5881 Convert from define_expand to be define_insn_and_split. Rework
5882 float/double/_Float128 conversions to QI/HI/SImode to work with
5883 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5884 conversions to QI/HImode types did a store and then a load to
5885 truncate the value. For conversions to VSX registers, don't split
5886 the insn, instead emit the code directly. Use the code iterator
5887 any_fix to combine signed and unsigned conversions.
5888 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5889 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5890 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5891 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5892 (fix_<mode>di2_hw): Likewise.
5893 (fixuns_<mode>di2_hw): Likewise.
5894 (fix_<mode>si2_hw): Likewise.
5895 (fixuns_<mode>si2_hw): Likewise.
5896 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5897 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5898 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5899 fix<uns>_trunc<SFDF:mode>si2_p8.
5900 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5901 used.
5902 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5903 (fix<uns>_<mode>_mem): Likewise.
5904 (fctiw<u>z_<mode>_mem): Likewise.
5905 (fix<uns>_<mode>_mem): Likewise.
5906 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5907 the register allocator from doing a direct move to the GPRs to do
5908 a store, and instead use the ISA 3.0 store byte/half-word from
5909 vector register instruction. For IEEE 128-bit floating point,
5910 also optimize stores of 32-bit ints.
5911 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5913 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5915 * genextract.c (push_pathstr_operand): New function to support
5916 [a-zA-Z].
5917 (walk_rtx): Call push_pathstr_operand.
5918 (print_path): Support [a-zA-Z].
5920 2018-02-07 Richard Biener <rguenther@suse.de>
5922 PR tree-optimization/84037
5923 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5924 (cse_and_gimplify_to_preheader): Declare.
5925 (vect_get_place_in_interleaving_chain): Likewise.
5926 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5927 ivexpr_map.
5928 (_loop_vec_info::~_loop_vec_info): Delete it.
5929 (cse_and_gimplify_to_preheader): New function.
5930 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5931 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5932 (vectorizable_load): Likewise. For grouped stores always base
5933 the IV on the first element.
5934 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5935 condition before gimplifying.
5937 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5939 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5940 *DIV_EXPR and *MOD_EXPR.
5942 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5944 PR target/84248
5945 * config/i386/i386.c (ix86_option_override_internal): Mask out
5946 the CF_SET bit when checking -fcf-protection.
5948 2018-02-07 Tom de Vries <tom@codesourcery.com>
5950 PR libgomp/84217
5951 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5952 enough.
5954 2018-02-07 Richard Biener <rguenther@suse.de>
5956 PR tree-optimization/84204
5957 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5958 this place.
5960 PR tree-optimization/84205
5961 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5962 special-case isl_ast_op_zdiv_r.
5964 PR tree-optimization/84223
5965 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5966 Only add conditions from within the region.
5967 (gather_bbs::after_dom_children): Adjust.
5969 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5971 PR target/84209
5972 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5973 * config/avr/avr.md: Only post-reload split REG-REG moves if
5974 either register is GENERAL_REG_P.
5976 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5978 PR tree-optimization/84235
5979 * tree-ssa-scopedtables.c
5980 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5981 if the subtraction is performed in floating point type where NaNs are
5982 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5983 build 1. Formatting fix.
5985 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5987 PR target/84146
5988 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5989 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5990 and skip it regardless of bb boundaries. Use CALL_P macro,
5991 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5992 unnecessarily, formatting fix.
5994 2018-02-06 Michael Collison <michael.collison@arm.com>
5996 * config/arm/thumb2.md:
5997 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5998 (*thumb_mov_notscc): Ditto.
6000 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6002 PR target/84154
6003 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6004 unsigned_fix, not "s".
6006 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6008 * configure.ac (gcc_fn_eh_frame_ro): New function.
6009 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6010 correct .eh_frame permissions.
6011 * configure: Regenerate.
6013 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6015 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6016 irrelevant options.
6018 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6020 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6021 Display warning message for -mno-speculate-indirect-jumps.
6023 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6025 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6026 Undocumented.
6027 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6029 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6031 PR tree-optimization/84225
6032 * tree-eh.c (find_trapping_overflow): Only call
6033 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6035 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6037 PR target/84145
6038 * config/i386/i386.c: Reimplement the check of possible options
6039 -mibt/-mshstk conbination. Change error messages.
6040 * doc/invoke.texi: Fix a typo: remove extra '='.
6042 2018-02-06 Marek Polacek <polacek@redhat.com>
6044 PR tree-optimization/84228
6045 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6047 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6049 PR target/82641
6050 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6051 emitted arch directives.
6052 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6053 __ARM_FEATURE_COPROC before changing architectures.
6055 2018-02-06 Richard Biener <rguenther@suse.de>
6057 * config/i386/i386.c (print_reg): Fix typo.
6058 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6060 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6062 * configure: Regenerate.
6064 2018-02-05 Martin Sebor <msebor@redhat.com>
6066 PR tree-optimization/83369
6067 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6068 inlining context.
6070 2018-02-05 Martin Liska <mliska@suse.cz>
6072 * doc/invoke.texi: Cherry-pick upstream r323995.
6074 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6076 * ira.c (ira_init_register_move_cost): Adjust comment.
6078 2018-02-05 Martin Liska <mliska@suse.cz>
6080 PR gcov-profile/84137
6081 * doc/gcov.texi: Fix typo in documentation.
6083 2018-02-05 Martin Liska <mliska@suse.cz>
6085 PR gcov-profile/83879
6086 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6087 using dlopen functionality.
6089 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6091 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6092 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6093 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6094 _mm_maskz_range_round_ss): New intrinsics.
6095 (__builtin_ia32_rangesd128_round)
6096 (__builtin_ia32_rangess128_round): Remove.
6097 (__builtin_ia32_rangesd128_mask_round,
6098 __builtin_ia32_rangess128_mask_round): New builtins.
6099 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6100 __builtin_ia32_rangess128_round): Remove.
6101 (__builtin_ia32_rangesd128_mask_round,
6102 __builtin_ia32_rangess128_mask_round): New builtins.
6103 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6104 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6105 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6106 "<round_saeonly_constraint>")): Changed to ...
6107 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6108 "<round_saeonly_scalar_constraint>")): ... this.
6109 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6110 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6111 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6112 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6113 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6115 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6117 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6118 options.
6119 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6120 Remove all values except native, 8540 and 8548.
6122 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6124 * config/i386/i386.c (ix86_output_function_return): Pass
6125 INVALID_REGNUM, instead of -1, as invalid register number to
6126 indirect_thunk_name and output_indirect_thunk.
6128 2018-02-02 Julia Koval <julia.koval@intel.com>
6130 * config.gcc: Add -march=icelake.
6131 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6132 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6133 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6134 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6135 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6136 (processor_target_table): Add icelake.
6137 (ix86_option_override_internal): Handle new PTAs.
6138 (get_builtin_code_for_version): Handle icelake.
6139 (M_INTEL_COREI7_ICELAKE): New.
6140 (fold_builtin_cpu): Handle icelake.
6141 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6142 * doc/invoke.texi: Add -march=icelake.
6144 2018-02-02 Julia Koval <julia.koval@intel.com>
6146 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6147 to wide_int_bitmask.
6148 * wide-int-bitmask.h: New.
6150 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6152 PR target/84066
6153 * config/i386/i386.md: Replace Pmode with word_mode in
6154 builtin_setjmp_setup and builtin_longjmp to support x32.
6156 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6158 PR target/56010
6159 PR target/83743
6160 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6161 #include "opts.h".
6162 (rs6000_supported_cpu_names): New static variable.
6163 (linux_cpu_translation_table): Likewise.
6164 (elf_platform) <cpu>: Define new static variable and use it.
6165 Translate kernel AT_PLATFORM name to canonical name if needed.
6166 Error if platform name is unknown.
6168 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6170 PR target/84089
6171 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6173 2018-02-01 Jeff Law <law@redhat.com>
6175 PR target/84128
6176 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6177 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6178 the scratch if RELEASE_VIA_POP is false.
6179 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6180 If we have to save a temporary register, decrement SIZE appropriately.
6181 Pass new arguments to release_scratch_register_on_entry.
6182 (ix86_adjust_stack_and_probe): Likewise.
6183 (ix86_emit_probe_stack_range): Pass new arguments to
6184 release_scratch_register_on_entry.
6186 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6188 PR rtl-optimization/84157
6189 * combine.c (change_zero_ext): Use REG_P predicate in
6190 front of HARD_REGISTER_P predicate.
6192 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6194 * config/avr/avr.c (avr_option_override): Move disabling of
6195 -fdelete-null-pointer-checks to...
6196 * common/config/avr/avr-common.c (avr_option_optimization_table):
6197 ...here.
6199 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6201 PR tree-optimization/81635
6202 * tree-data-ref.c (split_constant_offset_1): For types that
6203 wrap on overflow, try to use range info to prove that wrapping
6204 cannot occur.
6206 2018-02-01 Renlin Li <renlin.li@arm.com>
6208 PR target/83370
6209 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6210 TAILCALL_ADDR_REGS.
6211 (aarch64_register_move_cost): Likewise.
6212 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6213 TAILCALL_ADDR_REGS.
6214 (REG_CLASS_NAMES): Likewise.
6215 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6216 TAILCALL_ADDR_REGS. Remove IP registers.
6217 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6219 2018-02-01 Richard Biener <rguenther@suse.de>
6221 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6222 for specifying RPO order and allow NULL for that.
6223 * domwalk.c (dom_walker::dom_walker): Likewise.
6224 (dom_walker::walk): Handle NULL RPO order.
6225 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6226 in RPO order.
6227 (rewrite_update_dom_walker): Likewise.
6228 (mark_def_dom_walker): Likewise.
6230 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6232 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6233 (aarch64_maybe_expand_sve_subreg_move): Declare.
6234 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6235 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6236 predicate.
6237 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6238 that are semantically a reverse operation.
6239 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6240 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6241 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6242 functions.
6243 (aarch64_can_change_mode_class): For big-endian, forbid changes
6244 between two SVE modes if they have different element sizes.
6246 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6248 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6249 the TImode handling for big-endian targets.
6251 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6253 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6254 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6255 not just bytes.
6256 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6257 Remove BSWAP handing for big-endian targets and use the form of
6258 LD1RQ appropariate for the mode.
6260 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6262 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6263 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6264 duplicated element.
6266 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6268 PR tearget/83845
6269 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6270 check for operands that need to go through aarch64_sve_reload_be.
6272 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6274 PR tree-optimization/81661
6275 PR tree-optimization/84117
6276 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6277 * tree-eh.c: Include gimplify.h.
6278 (find_trapping_overflow, replace_trapping_overflow,
6279 rewrite_to_non_trapping_overflow): New functions.
6280 * tree-vect-loop.c: Include tree-eh.h.
6281 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6282 * tree-data-ref.c: Include tree-eh.h.
6283 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6285 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6287 PR rtl-optimization/84123
6288 * combine.c (change_zero_ext): Check if hard register satisfies
6289 can_change_dest_mode before calling gen_lowpart_SUBREG.
6291 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6293 PR target/82444
6294 * ira.c (ira_init_register_move_cost): Remove assert.
6296 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6298 PR rtl-optimization/84071
6299 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6300 * doc/tm.texi: Regenerate.
6302 2018-01-31 Richard Biener <rguenther@suse.de>
6304 PR tree-optimization/84132
6305 * tree-data-ref.c (analyze_miv_subscript): Properly
6306 check whether evolution_function_is_affine_multivariate_p
6307 before calling gcd_of_steps_may_divide_p.
6309 2018-01-31 Julia Koval <julia.koval@intel.com>
6311 PR target/83618
6312 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6313 * config/i386/i386.md (rdpid_rex64) New.
6314 (rdpid): Make 32bit only.
6316 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6318 PR lto/84105
6319 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6320 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6322 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6324 Revert
6325 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6327 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6329 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6331 PR rtl-optimization/84071
6332 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6333 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6335 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6337 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6338 (arc_attribute_table): Add 'aux' attribute.
6339 (arc_in_small_data_p): Consider aux like variables.
6340 (arc_is_aux_reg_p): New function.
6341 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6342 (arc_get_aux_arg): New function.
6343 (prepare_move_operands): Handle aux-register access.
6344 (arc_handle_aux_attribute): New function.
6345 * doc/extend.texi (ARC Variable attributes): Add subsection.
6347 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6349 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6350 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6351 (arc_attribute_table): Add 'uncached' attribute.
6352 (arc_print_operand): Print '.di' flag for uncached memory
6353 accesses.
6354 (arc_in_small_data_p): Do not consider for small data the uncached
6355 types.
6356 (arc_is_uncached_mem_p): New function.
6357 * config/arc/predicates.md (compact_store_memory_operand): Check
6358 for uncached memory accesses.
6359 (nonvol_nonimm_operand): Likewise.
6360 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6362 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6364 PR c/84100
6365 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6366 falign-loops=): Add Optimization flag.
6368 2018-01-30 Jeff Law <law@redhat.com>
6370 PR target/84064
6371 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6372 INT_REGISTERS_SAVED. Check it prior to calling
6373 get_scratch_register_on_entry.
6374 (ix86_adjust_stack_and_probe): Similarly.
6375 (ix86_emit_probe_stack_range): Similarly.
6376 (ix86_expand_prologue): Corresponding changes.
6378 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6380 PR target/40411
6381 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6382 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6384 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6386 PR target/84112
6387 * lra-constraints.c (curr_insn_transform): Process AND in the
6388 address.
6390 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6392 PR rtl-optimization/83986
6393 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6394 dependence against last_pending_memory_flush in addition to
6395 pending_jump_insns.
6397 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6399 PR tree-optimization/81611
6400 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6401 copies.
6403 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6405 PR target/83758
6406 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6407 a reg rtx.
6409 2018-01-30 Richard Biener <rguenther@suse.de>
6410 Jakub Jelinek <jakub@redhat.com>
6412 PR tree-optimization/84111
6413 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6414 inner loops added during recursion, as they don't have up-to-date
6415 SSA form.
6417 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6419 PR ipa/81360
6420 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6421 (can_inline_edge_by_limits_p): ... here.
6422 (can_early_inline_edge_p, check_callers,
6423 update_caller_keys, update_callee_keys, recursive_inlining,
6424 add_new_edges_to_heap, speculation_useful_p,
6425 inline_small_functions,
6426 inline_small_functions, flatten_function,
6427 inline_to_all_callers_1): Update.
6429 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6431 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6432 zeros correctly.
6434 2018-01-30 Richard Biener <rguenther@suse.de>
6436 PR tree-optimization/83008
6437 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6438 invariant and constant vector uses in stmts when they need
6439 more than one stmt.
6441 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6443 PR bootstrap/84017
6444 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6445 * configure: Regenerate.
6447 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6449 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6450 pattern.
6451 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6452 Use gen_rtx_REG rather than gen_lowpart.
6454 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6456 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6457 rather than 0 when creating partial subregs.
6459 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6461 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6462 of usage.
6464 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6466 PR target/81550
6467 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6468 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6469 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6470 flags. This restores the settings used before the 2017-07-24.
6471 Turning off pre increment/decrement/modify allows IVOPTS to
6472 optimize DF/SF loops where the index is an int.
6474 2018-01-29 Richard Biener <rguenther@suse.de>
6475 Kelvin Nilsen <kelvin@gcc.gnu.org>
6477 PR bootstrap/80867
6478 * tree-vect-stmts.c (vectorizable_call): Don't call
6479 targetm.vectorize_builtin_md_vectorized_function if callee is
6480 NULL.
6482 2018-01-22 Carl Love <cel@us.ibm.com>
6484 * doc/extend.tex: Fix typo in second arg in
6485 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6487 2018-01-29 Richard Biener <rguenther@suse.de>
6489 PR tree-optimization/84086
6490 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6491 (flush_ssaname_freelist): When SSA names were released reset
6492 the SCEV hash table.
6494 2018-01-29 Richard Biener <rguenther@suse.de>
6496 PR tree-optimization/84057
6497 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6498 removed paths when removing edges.
6500 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6502 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6503 -mfunction-return=@var{choice}.
6505 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6507 PR diagnostic/84034
6508 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6509 Handle CR like TAB.
6510 (layout::print_source_line): Likewise.
6511 (test_get_line_width_without_trailing_whitespace): Add test cases.
6513 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6515 PR middle-end/84040
6516 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6517 debug insns.
6519 2018-01-26 Jim Wilson <jimw@sifive.com>
6521 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6523 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6524 specified.
6526 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6528 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6529 and CMP + SUB-immediate -> SUBS.
6531 2018-01-26 Martin Sebor <msebor@redhat.com>
6533 PR tree-optimization/83896
6534 * tree-ssa-strlen.c (get_string_len): Rename...
6535 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6536 Avoid assuming length is constant.
6537 (handle_char_store): Use HOST_WIDE_INT for string length.
6539 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6541 PR target/81763
6542 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6543 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6545 2018-01-26 Richard Biener <rguenther@suse.de>
6547 PR rtl-optimization/84003
6548 * dse.c (record_store): Only record redundant stores when
6549 the earlier store aliases at least all accesses the later one does.
6551 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6553 PR rtl-optimization/83985
6554 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6555 REG_CFA_RESTORE insns.
6556 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6557 REG_CFA_RESTORE insns here.
6559 PR c/83989
6560 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6561 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6563 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6565 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6566 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6567 (arc_init): Likewise.
6568 (arc_override_options): Likewise.
6569 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6570 value.
6571 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6572 support.
6573 * config/arc/arc.h (TARGET_DBNZ): Define.
6574 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6575 properly set the tune attribute.
6576 (dbnz): Use TARGET_DBNZ guard.
6577 * config/arc/arc.opt (mtune): Add core3 option.
6579 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6581 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6582 recognize new pic like addresses.
6583 (arc_delegitimize_address): Clean up.
6585 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6587 * config/arc/arc-arches.def: Option mrf16 valid for all
6588 architectures.
6589 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6590 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6591 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6592 * config/arc/arc-tables.opt: Regenerate.
6593 * config/arc/arc.c (arc_conditional_register_usage): Handle
6594 reduced register file case.
6595 (arc_file_start): Set must have build attributes.
6596 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6597 mrf16 option value.
6598 * config/arc/arc.opt (mrf16): Add new option.
6599 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6600 * config/arc/genmultilib.awk: Handle new mrf16 option.
6601 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6602 * config/arc/t-multilib: Regenerate.
6603 * doc/invoke.texi (ARC Options): Document mrf16 option.
6605 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6607 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6608 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6609 (arc_attribute_table): Add 'secure_call' attribute.
6610 (arc_print_operand): Print secure call operand.
6611 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6612 secure.
6613 (arc_is_secure_call_p): New function. * config/arc/arc.md
6614 (call_i): Add support for sjli instruction.
6615 (call_value_i): Likewise.
6616 * config/arc/constraints.md (Csc): New constraint.
6618 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6619 John Eric Martin <John.Martin@emmicro-us.com>
6621 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6622 * config/arc/arc.c (_arc_jli_section): New struct.
6623 (arc_jli_section): New type.
6624 (rc_jli_sections): New static variable.
6625 (arc_handle_jli_attribute): New function.
6626 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6627 (arc_file_end): New function.
6628 (TARGET_ASM_FILE_END): Define.
6629 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6630 (arc_add_jli_section): New function.
6631 (jli_call_scan): Likewise.
6632 (arc_reorg): Call jli_call_scan.
6633 (arc_output_addsi): Remove 'S' from printing asm operand.
6634 (arc_is_jli_call_p): New function.
6635 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6636 operand.
6637 (movhi_insn): Likewise.
6638 (movsi_insn): Likewise.
6639 (movsi_set_cc_insn): Likewise.
6640 (loadqi_update): Likewise.
6641 (load_zeroextendqisi_update): Likewise.
6642 (load_signextendqisi_update): Likewise.
6643 (loadhi_update): Likewise.
6644 (load_zeroextendhisi_update): Likewise.
6645 (load_signextendhisi_update): Likewise.
6646 (loadsi_update): Likewise.
6647 (loadsf_update): Likewise.
6648 (movsicc_insn): Likewise.
6649 (bset_insn): Likewise.
6650 (bxor_insn): Likewise.
6651 (bclr_insn): Likewise.
6652 (bmsk_insn): Likewise.
6653 (bicsi3_insn): Likewise.
6654 (cmpsi_cc_c_insn): Likewise.
6655 (movsi_ne): Likewise.
6656 (movsi_cond_exec): Likewise.
6657 (clrsbsi2): Likewise.
6658 (norm_f): Likewise.
6659 (normw): Likewise.
6660 (swap): Likewise.
6661 (divaw): Likewise.
6662 (flag): Likewise.
6663 (sr): Likewise.
6664 (kflag): Likewise.
6665 (ffs): Likewise.
6666 (ffs_f): Likewise.
6667 (fls): Likewise.
6668 (call_i): Remove 'S' asm letter, add jli instruction.
6669 (call_value_i): Likewise.
6670 * config/arc/arc.op (mjli-always): New option.
6671 * config/arc/constraints.md (Cji): New constraint.
6672 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6673 operand.
6674 (subsf3_fpx): Likewise.
6675 (mulsf3_fpx): Likewise.
6676 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6677 asm operand.
6678 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6679 function attrbutes.
6680 * doc/invoke.texi (ARC): Document mjli-always option.
6682 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6684 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6685 avoid addition with 0 and use incw and decw where possible.
6687 2018-01-26 Richard Biener <rguenther@suse.de>
6689 PR tree-optimization/81082
6690 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6691 association if it requires casting to unsigned.
6692 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6693 from fold_plusminus_mult_expr to catch important cases late when
6694 range info is available.
6696 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6698 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6699 * configure.ac (hidden_linkonce): New test.
6700 * configure: Regenerate.
6701 * config.in: Regenerate.
6703 2018-01-26 Julia Koval <julia.koval@intel.com>
6705 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6706 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6707 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6708 _mm_mask_bitshuffle_epi64_mask): Fix type.
6709 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6710 USI_FTYPE_V4DI_V4DI_USI): Remove.
6711 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6712 __builtin_ia32_vpshufbitqmb256_mask,
6713 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6714 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6715 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6717 2018-01-26 Alan Modra <amodra@gmail.com>
6719 PR target/84033
6720 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6721 UNSPEC_VBPERMQ. Sort other unspecs.
6723 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6725 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6727 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6729 PR middle-end/83055
6730 * predict.c (drop_profile): Do not push/pop cfun; update also
6731 node->count.
6732 (handle_missing_profiles): Fix logic looking for zero profiles.
6734 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6736 PR middle-end/83977
6737 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6738 on functions with #pragma omp declare simd or functions with simd
6739 attribute.
6740 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6741 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6742 Remove trailing \n from warning_at calls.
6744 2018-01-25 Tom de Vries <tom@codesourcery.com>
6746 PR target/84028
6747 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6748 for neutered workers.
6750 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6752 PR target/68467
6753 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6754 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6756 2018-01-24 Jeff Law <law@redhat.com>
6758 PR target/83994
6759 * i386.c (get_probe_interval): Move to earlier point.
6760 (ix86_compute_frame_layout): If -fstack-clash-protection and
6761 the frame is larger than the probe interval, then use pushes
6762 to save registers rather than reg->mem moves.
6763 (ix86_expand_prologue): Remove conditional for int_registers_saved
6764 assertion.
6766 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6768 PR target/84014
6769 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6770 min/max for never referenced object.
6772 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6774 PR middle-end/83977
6775 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6776 here.
6777 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6778 attributes from DECL_ATTRIBUTES (decl) without affecting
6779 DECL_ATTRIBUTES (current_function_decl).
6780 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6781 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6783 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6785 PR tree-optimization/83979
6786 * fold-const.c (fold_comparison): Use constant_boolean_node
6787 instead of boolean_{true,false}_node.
6789 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6791 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6792 with zero counts.
6794 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6796 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6797 Simplify the clause that sets the length attribute.
6798 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6799 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6800 clause that sets the length attribute.
6801 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6803 2018-01-24 Tom de Vries <tom@codesourcery.com>
6805 PR target/83589
6806 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6807 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6808 Add strict parameter.
6809 (prevent_branch_around_nothing): Insert dummy insn between branch to
6810 label and label with no ptx insn inbetween.
6811 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6813 2018-01-24 Tom de Vries <tom@codesourcery.com>
6815 PR target/81352
6816 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6817 for neutered threads in warp.
6818 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6820 2018-01-24 Richard Biener <rguenther@suse.de>
6822 PR tree-optimization/83176
6823 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6824 operands.
6826 2018-01-24 Richard Biener <rguenther@suse.de>
6828 PR tree-optimization/82819
6829 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6830 code generating pluses that are no-ops in the target precision.
6832 2018-01-24 Richard Biener <rguenther@suse.de>
6834 PR middle-end/84000
6835 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6837 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6839 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6840 to merge probabilities.
6841 * predict.c (probably_never_executed): Also mark as cold functions
6842 with global 0 profile and guessed local profile.
6843 * profile-count.c (profile_probability::combine_with_count): New
6844 member function.
6845 * profile-count.h (profile_probability::operator*,
6846 profile_probability::operator*=, profile_probability::operator/,
6847 profile_probability::operator/=): Reduce precision to adjusted
6848 and set value to guessed on contradictory divisions.
6849 (profile_probability::combine_with_freq): Remove.
6850 (profile_probability::combine_wiht_count): Declare.
6851 (profile_count::force_nonzero):: Set to adjusted.
6852 (profile_count::probability_in):: Set quality to adjusted.
6853 * tree-ssa-tail-merge.c (replace_block_by): Use
6854 combine_with_count.
6856 2018-01-23 Andrew Waterman <andrew@sifive.com>
6857 Jim Wilson <jimw@sifive.com>
6859 * config/riscv/riscv.c (riscv_stack_boundary): New.
6860 (riscv_option_override): Set riscv_stack_boundary. Handle
6861 riscv_preferred_stack_boundary_arg.
6862 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6863 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6864 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6865 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6866 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6867 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6869 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6871 PR target/83905
6872 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6873 of struct ix86_frame.
6874 (ix86_expand_epilogue): Likewise. Add a local variable for
6875 the reg_save_offset field in struct ix86_frame.
6877 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6879 PR tree-optimization/82604
6880 * tree-loop-distribution.c (enum partition_kind): New enum item
6881 PKIND_PARTIAL_MEMSET.
6882 (partition_builtin_p): Support above new enum item.
6883 (generate_code_for_partition): Ditto.
6884 (compute_access_range): Differentiate cases that equality can be
6885 proven at all loops, the innermost loops or no loops.
6886 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6887 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6888 (finalize_partitions, distribute_loop): Don't fuse partition of
6889 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6890 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6891 parloop is enabled.
6893 2018-01-23 Martin Liska <mliska@suse.cz>
6895 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6896 order to ignore the predictor.
6897 (PRED_POLYMORPHIC_CALL): Likewise.
6898 (PRED_RECURSIVE_CALL): Likewise.
6900 2018-01-23 Martin Liska <mliska@suse.cz>
6902 * tree-profile.c (tree_profiling): Print function header to
6903 aware reader which function we are working on.
6904 * value-prof.c (gimple_find_values_to_profile): Do not print
6905 not interesting value histograms.
6907 2018-01-23 Martin Liska <mliska@suse.cz>
6909 * profile-count.h (enum profile_quality): Add
6910 profile_uninitialized as the first value. Do not number values
6911 as they are zero based.
6912 (profile_count::verify): Update sanity check.
6913 (profile_probability::verify): Likewise.
6915 2018-01-23 Nathan Sidwell <nathan@acm.org>
6917 * doc/invoke.texi (ffor-scope): Deprecate.
6919 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6921 PR tree-optimization/83510
6922 * domwalk.c (set_all_edges_as_executable): New function.
6923 (dom_walker::dom_walker): Convert bool param
6924 "skip_unreachable_blocks" to enum reachability. Move setup of
6925 edge flags to set_all_edges_as_executable and only do it when
6926 reachability is REACHABLE_BLOCKS.
6927 * domwalk.h (enum dom_walker::reachability): New enum.
6928 (dom_walker::dom_walker): Convert bool param
6929 "skip_unreachable_blocks" to enum reachability.
6930 (set_all_edges_as_executable): New decl.
6931 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6932 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6933 "reachability".
6934 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6935 but converting true to REACHABLE_BLOCKS.
6936 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6937 * tree-vrp.c
6938 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6939 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6940 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6941 REACHABLE_BLOCKS.
6942 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6943 if check_all_array_refs will be called.
6945 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6947 * tree.c (selftest::test_location_wrappers): Add more test
6948 coverage.
6950 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6952 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6953 (selftest::test_bit_in_range): Likewise.
6955 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6957 PR testsuite/83888
6958 * doc/sourcebuild.texi (vect_float): Say that the selector
6959 only describes the situation when -funsafe-math-optimizations is on.
6960 (vect_float_strict): Document.
6962 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6964 PR tree-optimization/83965
6965 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6966 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6967 instead of checking only for a reduction.
6968 (vect_recog_widen_sum_pattern): Likewise.
6970 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6972 * predict.c (probably_never_executed): Only use precise profile info.
6973 (compute_function_frequency): Skip after inlining hack since we now
6974 have quality checking.
6976 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6978 * profile-count.h (profile_probability::very_unlikely,
6979 profile_probability::unlikely, profile_probability::even): Set
6980 precision to guessed.
6982 2018-01-23 Richard Biener <rguenther@suse.de>
6984 PR tree-optimization/83963
6985 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6986 Properly terminate dominator walk when crossing the exit edge not
6987 when visiting its source block.
6989 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6991 PR c++/83918
6992 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6993 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6995 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6997 PR tree-optimization/83957
6998 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6999 semicolon after for body surrounded by braces.
7001 PR tree-optimization/83081
7002 * profile-count.h (profile_probability::split): New method.
7003 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7004 Use profile_probability::split.
7005 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7006 when splitting a single conditional jump into 2.
7008 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7010 PR tree-optimization/69452
7011 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7012 decl.
7014 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7016 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7017 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7018 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7020 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7022 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7023 * config/rl78/rl78.md (movdi): New define_expand.
7024 * config/rl78/rl78.c (rl78_split_movdi): New function.
7026 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7028 PR target/83862
7029 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7030 no longer used.
7031 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7032 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7033 128-bit to produce an UNSPEC move to get the double word with the
7034 signbit and then a shift directly to do signbit.
7035 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7036 implementation with a new version that just does either a direct
7037 move or a regular move. Move memory interface to separate insns.
7038 Move insns so they are next to the expander.
7039 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7040 with signbit move. Split big and little endian case.
7041 (signbit<mode>2_dm_mem_le): Likewise.
7042 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7043 (signbit<mode>2_dm2): Likewise.
7045 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7047 * config/rl78/rl78.md (anddi3): New define_expand.
7049 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7051 * config/rl78/rl78.md (umindi3): New define_expand.
7053 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7055 * config/rl78/rl78.md (smindi3): New define_expand.
7057 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7059 * config/rl78/rl78.md (smaxdi3): New define_expand.
7061 2018-01-22 Carl Love <cel@us.ibm.com>
7063 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7064 LVX_V1TI): Add macro expansion.
7065 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7066 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7067 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7068 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7069 Change check to determine if the instruction is a byte reversing
7070 entry. Fix typo in comment.
7071 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7072 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7073 Add def_builtin calls for new builtins.
7074 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7075 Add define_insn expansion.
7077 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7079 * config/rl78/rl78.md (umaxdi3): New define_expand.
7081 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7083 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7084 for non-QImode registers.
7086 2018-01-22 Richard Biener <rguenther@suse.de>
7088 PR tree-optimization/83963
7089 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7090 including the loop exit block.
7091 (scop_detection::merge_sese): Likewise.
7092 (scop_detection::add_scop): Do it here instead.
7094 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7096 * doc/sourcebuild.texi (arm_softfloat): Document.
7098 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7100 PR gcc/77734
7101 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7102 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7103 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7105 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7106 David Edelsohn <dje.gcc@gmail.com>
7108 PR target/83946
7109 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7110 Change "crset eq" to "crset 2".
7111 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7112 (*call_indirect_aix<mode>_nospec): Likewise.
7113 (*call_value_indirect_aix<mode>_nospec): Likewise.
7114 (*call_indirect_elfv2<mode>_nospec): Likewise.
7115 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7116 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7117 change assembly output from . to $.
7118 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7119 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7120 (*tablejump<mode>_internal1_nospec): Likewise.
7122 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7124 PR target/80870
7125 * config/sh/sh_optimize_sett_clrt.cc:
7126 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7128 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7130 PR tree-optimization/83940
7131 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7132 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7133 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7134 use it to pass back the definition type.
7135 (vect_check_store_rhs): Likewise rhs_dt_out.
7136 (vect_build_gather_load_calls): Add a mask_dt argument and use
7137 it instead of a call to vect_is_simple_use.
7138 (vectorizable_store): Update calls to vect_check_load_store_mask
7139 and vect_check_store_rhs. Use the dt returned by the latter instead
7140 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7141 instead of calls to vect_is_simple_use. Pass the scalar rather
7142 than the vector operand to vect_is_simple_use when handling
7143 second and subsequent copies of an rhs value.
7144 (vectorizable_load): Update calls to vect_check_load_store_mask
7145 and vect_build_gather_load_calls. Use the cached mask_dt and
7146 gs_info.offset_dt instead of calls to vect_is_simple_use.
7148 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7150 PR middle-end/83945
7151 * tree-emutls.c: Include gimplify.h.
7152 (lower_emutls_2): New function.
7153 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7154 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7155 it before further processing.
7157 PR target/83930
7158 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7159 UINTVAL (trueop1) instead of INTVAL (op1).
7161 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7163 PR debug/81570
7164 PR debug/83728
7165 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7166 INCOMING_FRAME_SP_OFFSET if not defined.
7167 (scan_trace): Add ENTRY argument. If true and
7168 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7169 emit a note to adjust the CFA offset.
7170 (create_cfi_notes): Adjust scan_trace callers.
7171 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7172 INCOMING_FRAME_SP_OFFSET in the CIE.
7173 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7174 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7175 Likewise.
7176 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7177 * doc/tm.texi: Regenerated.
7179 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7181 PR rtl-optimization/83147
7182 * lra-constraints.c (remove_inheritance_pseudos): Use
7183 lra_substitute_pseudo_within_insn.
7185 2018-01-19 Tom de Vries <tom@codesourcery.com>
7186 Cesar Philippidis <cesar@codesourcery.com>
7188 PR target/83920
7189 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7191 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7193 PR target/83790
7194 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7195 spaces for function labels.
7197 2018-01-19 Martin Liska <mliska@suse.cz>
7199 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7200 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7201 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7202 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7203 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7204 (PRED_CONST_RETURN): Change from 69 to 65.
7205 (PRED_NULL_RETURN): Change from 91 to 71.
7206 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7207 (PRED_LOOP_GUARD): Change from 66 to 73.
7209 2018-01-19 Martin Liska <mliska@suse.cz>
7211 * predict.c (predict_insn_def): Add new assert.
7212 (struct branch_predictor): Change type to signed integer.
7213 (test_prediction_value_range): Amend test to cover
7214 PROB_UNINITIALIZED.
7215 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7216 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7217 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7218 (PRED_LOOP_IV_COMPARE): Likewise.
7219 * predict.h (PROB_UNINITIALIZED): Define new constant.
7221 2018-01-19 Martin Liska <mliska@suse.cz>
7223 * predict.c (dump_prediction): Add new format for
7224 analyze_brprob.py script which is enabled with -details
7225 suboption.
7226 * profile-count.h (precise_p): New function.
7228 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7230 PR tree-optimization/83922
7231 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7232 there are no statements that need masking.
7233 (vect_active_double_reduction_p): New function.
7234 (vect_analyze_loop_operations): Use it when handling phis that
7235 are not in the loop header.
7237 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7239 PR tree-optimization/83914
7240 * tree-vect-loop.c (vectorizable_induction): Don't convert
7241 init_expr or apply the peeling adjustment for inductions
7242 that are nested within the vectorized loop.
7244 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7246 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7247 instead of NEG.
7249 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7251 PR sanitizer/81715
7252 PR testsuite/83882
7253 * function.h (gimplify_parameters): Add gimple_seq * argument.
7254 * function.c: Include gimple.h and options.h.
7255 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7256 for the added local temporaries if needed.
7257 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7258 if there are any parameter cleanups, wrap whole body into a
7259 try/finally with the cleanups.
7261 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7263 PR target/82964
7264 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7265 Use GET_MODE_CLASS for scalar floating point.
7267 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7269 PR ipa/82256
7270 patch by PaX Team
7271 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7272 Fix call of call_cgraph_insertion_hooks.
7274 2018-01-18 Martin Sebor <msebor@redhat.com>
7276 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7278 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7280 PR ipa/83619
7281 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7282 frequencies.
7284 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7286 PR other/70268
7287 * common.opt: (-ffile-prefix-map): New option.
7288 * opts.c (common_handle_option): Defer it.
7289 * opts-global.c (handle_common_deferred_options): Handle it.
7290 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7291 * file-prefix-map.h: New file.
7292 (remap_debug_filename, add_debug_prefix_map): ...here.
7293 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7294 * final.c (debug_prefix_map, add_debug_prefix_map
7295 remap_debug_filename): Move to...
7296 * file-prefix-map.c: New file.
7297 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7298 generalize, get rid of alloca(), use strrchr() instead of strchr().
7299 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7300 Implement in terms of add_prefix_map().
7301 (remap_macro_filename, remap_debug_filename): Implement in term of
7302 remap_filename().
7303 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7304 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7305 * dbxout.c: Include file-prefix-map.h.
7306 * varasm.c: Likewise.
7307 * vmsdbgout.c: Likewise.
7308 * xcoffout.c: Likewise.
7309 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7310 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7311 * doc/invoke.texi (-ffile-prefix-map): Document.
7312 (-fdebug-prefix-map): Update description.
7314 2018-01-18 Martin Liska <mliska@suse.cz>
7316 * config/i386/i386.c (indirect_thunk_name): Document that also
7317 lfence is emitted.
7318 (output_indirect_thunk): Document why both instructions
7319 (pause and lfence) are generated.
7321 2018-01-18 Richard Biener <rguenther@suse.de>
7323 PR tree-optimization/83887
7324 * graphite-scop-detection.c
7325 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7326 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7327 (scop_detection::merge_sese): Re-implement with a flood-fill
7328 algorithm that properly finds a SESE region if it exists.
7330 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7332 PR c/61240
7333 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7334 pointer_diff optimizations use view_convert instead of convert.
7336 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7338 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7339 Generate different code for -mno-speculate-indirect-jumps.
7340 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7341 (*call_indirect_aix<mode>): Disable for
7342 -mno-speculate-indirect-jumps.
7343 (*call_indirect_aix<mode>_nospec): New define_insn.
7344 (*call_value_indirect_aix<mode>): Disable for
7345 -mno-speculate-indirect-jumps.
7346 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7347 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7348 -mno-speculate-indirect-jumps.
7349 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7351 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7353 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7354 long double type, set the flags for noting the default long double
7355 type, even if we don't pass or return a long double type.
7357 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7359 PR ipa/83051
7360 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7361 failure.
7363 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7365 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7366 support for merge[hl].
7367 (fold_mergehl_helper): New helper function.
7368 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7369 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7370 (altivec_vmrglw_direct): Add xxmrglw insn.
7372 2018-01-17 Andrew Waterman <andrew@sifive.com>
7374 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7375 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7377 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7379 PR lto/83121
7380 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7381 call the lto_location_cache before reading the
7382 DECL_SOURCE_LOCATION of the types.
7384 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7385 Richard Sandiford <richard.sandiford@linaro.org>
7387 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7388 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7389 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7390 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7391 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7392 Add declaration.
7393 * config/aarch64/constraints.md (aarch64_movti_operand):
7394 Limit immediates.
7395 * config/aarch64/predicates.md (Uti): Add new constraint.
7397 2018-01-17 Carl Love <cel@us.ibm.com>
7399 * config/rs6000/vsx.md (define_expand xl_len_r,
7400 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7401 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7402 lxvll.
7403 (define_expand, define_insn): Move the shift left from the
7404 define_insn to the define_expand for lxvl and stxvl instructions.
7405 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7406 and XL_LEN_R definitions to PURE.
7408 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7410 * config/i386/i386.c (indirect_thunk_name): Declare regno
7411 as unsigned int. Compare regno with INVALID_REGNUM.
7412 (output_indirect_thunk): Ditto.
7413 (output_indirect_thunk_function): Ditto.
7414 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7415 in the call to output_indirect_thunk_function.
7417 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7419 PR middle-end/83884
7420 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7421 rather than the size of inner_type to determine the stack slot size
7422 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7424 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7426 PR target/83546
7427 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7428 to PTA_SILVERMONT.
7430 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7432 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7433 endian Linux systems to optionally enable multilibs for selecting
7434 the long double type if the user configured an explicit type.
7435 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7436 have no long double multilibs if not defined.
7437 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7438 warn if the user used -mabi={ieee,ibm}longdouble and we built
7439 multilibs for long double.
7440 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7441 appropriate multilib option.
7442 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7443 multilib options.
7444 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7445 for building long double multilibs.
7446 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7448 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7450 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7451 copies.
7453 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7454 64 bits.
7455 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7456 128 bits.
7458 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7459 variables.
7461 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7462 return value.
7464 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7466 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7467 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7469 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7471 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7472 different rtl trees depending on TARGET_64BIT.
7473 (rs6000_gen_lvx): Likewise.
7475 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7477 * config/visium/visium.md (nop): Tweak comment.
7478 (hazard_nop): Likewise.
7480 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7482 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7483 -mspeculate-indirect-jumps.
7484 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7485 for -mno-speculate-indirect-jumps.
7486 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7487 (*call_value_indirect_elfv2<mode>): Disable for
7488 -mno-speculate-indirect-jumps.
7489 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7490 (indirect_jump): Emit different RTL for
7491 -mno-speculate-indirect-jumps.
7492 (*indirect_jump<mode>): Disable for
7493 -mno-speculate-indirect-jumps.
7494 (*indirect_jump<mode>_nospec): New define_insn.
7495 (tablejump): Emit different RTL for
7496 -mno-speculate-indirect-jumps.
7497 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7498 (tablejumpsi_nospec): New define_expand.
7499 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7500 (tablejumpdi_nospec): New define_expand.
7501 (*tablejump<mode>_internal1): Disable for
7502 -mno-speculate-indirect-jumps.
7503 (*tablejump<mode>_internal1_nospec): New define_insn.
7504 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7505 option.
7507 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7509 * caller-save.c (insert_save): Drop unnecessary parameter. All
7510 callers updated.
7512 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7513 Richard Biener <rguenth@suse.de>
7515 PR libgomp/83590
7516 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7517 return early, inline manually is_gimple_sizepos. Make sure if we
7518 call gimplify_expr we don't end up with a gimple constant.
7519 * tree.c (variably_modified_type_p): Don't return true for
7520 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7521 * gimplify.h (is_gimple_sizepos): Remove.
7523 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7525 PR tree-optimization/83857
7526 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7527 vectorizable_live_operation for pure SLP statements.
7528 (vectorizable_live_operation): Handle PHIs.
7530 2018-01-16 Richard Biener <rguenther@suse.de>
7532 PR tree-optimization/83867
7533 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7534 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7536 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7538 PR c/83844
7539 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7540 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7541 If off is not INTEGER_CST, issue a may not be aligned warning
7542 rather than isn't aligned. Use isn%'t rather than isn't.
7543 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7544 into MULT_EXPR.
7545 <case MULT_EXPR>: Improve the case when bottom and one of the
7546 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7547 operand, in that case check if the other operand is multiple of
7548 bottom divided by the INTEGER_CST operand.
7550 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7552 PR target/83858
7553 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7554 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7555 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7556 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7557 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7558 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7559 (pa_function_arg_size): New function.
7561 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7563 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7564 in a separate statement.
7566 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7568 PR tree-optimization/83847
7569 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7570 group gathers and scatters.
7572 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7574 PR rtl-optimization/86620
7575 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7577 PR rtl-optimization/83213
7578 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7579 to last if both are JUMP_INSNs.
7581 PR tree-optimization/83843
7582 * gimple-ssa-store-merging.c
7583 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7584 store_immediate_info for bswap/nop orig_stores.
7586 2018-01-15 Andrew Waterman <andrew@sifive.com>
7588 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7589 !TARGET_MUL.
7590 <UDIV>: Increase cost if !TARGET_DIV.
7592 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7594 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7595 (define_attr "cr_logical_3op"): New.
7596 (cceq_ior_compare): Adjust.
7597 (cceq_ior_compare_complement): Adjust.
7598 (*cceq_rev_compare): Adjust.
7599 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7600 (is_cracked_insn): Adjust.
7601 (insn_must_be_first_in_group): Adjust.
7602 * config/rs6000/40x.md: Adjust.
7603 * config/rs6000/440.md: Adjust.
7604 * config/rs6000/476.md: Adjust.
7605 * config/rs6000/601.md: Adjust.
7606 * config/rs6000/603.md: Adjust.
7607 * config/rs6000/6xx.md: Adjust.
7608 * config/rs6000/7450.md: Adjust.
7609 * config/rs6000/7xx.md: Adjust.
7610 * config/rs6000/8540.md: Adjust.
7611 * config/rs6000/cell.md: Adjust.
7612 * config/rs6000/e300c2c3.md: Adjust.
7613 * config/rs6000/e500mc.md: Adjust.
7614 * config/rs6000/e500mc64.md: Adjust.
7615 * config/rs6000/e5500.md: Adjust.
7616 * config/rs6000/e6500.md: Adjust.
7617 * config/rs6000/mpc.md: Adjust.
7618 * config/rs6000/power4.md: Adjust.
7619 * config/rs6000/power5.md: Adjust.
7620 * config/rs6000/power6.md: Adjust.
7621 * config/rs6000/power7.md: Adjust.
7622 * config/rs6000/power8.md: Adjust.
7623 * config/rs6000/power9.md: Adjust.
7624 * config/rs6000/rs64.md: Adjust.
7625 * config/rs6000/titan.md: Adjust.
7627 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7629 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7630 ix86_indirect_branch_register logic.
7632 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7634 * config/i386/constraints.md (Bs): Update
7635 ix86_indirect_branch_register check. Don't check
7636 ix86_indirect_branch_register with GOT_memory_operand.
7637 (Bw): Likewise.
7638 * config/i386/predicates.md (GOT_memory_operand): Don't check
7639 ix86_indirect_branch_register here.
7640 (GOT32_symbol_operand): Likewise.
7642 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7644 * config/i386/predicates.md (constant_call_address_operand):
7645 Rewrite ix86_indirect_branch_register logic.
7646 (sibcall_insn_operand): Likewise.
7648 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7650 * config/i386/constraints.md (Bs): Replace
7651 ix86_indirect_branch_thunk_register with
7652 ix86_indirect_branch_register.
7653 (Bw): Likewise.
7654 * config/i386/i386.md (indirect_jump): Likewise.
7655 (tablejump): Likewise.
7656 (*sibcall_memory): Likewise.
7657 (*sibcall_value_memory): Likewise.
7658 Peepholes of indirect call and jump via memory: Likewise.
7659 * config/i386/i386.opt: Likewise.
7660 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7661 (GOT_memory_operand): Likewise.
7662 (call_insn_operand): Likewise.
7663 (sibcall_insn_operand): Likewise.
7664 (GOT32_symbol_operand): Likewise.
7666 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7668 PR middle-end/83837
7669 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7670 type rather than type addr's type points to.
7671 (expand_omp_atomic_mutex): Likewise.
7672 (expand_omp_atomic): Likewise.
7674 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7676 PR target/83839
7677 * config/i386/i386.c (output_indirect_thunk_function): Use
7678 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7679 for __x86_return_thunk.
7681 2018-01-15 Richard Biener <rguenther@suse.de>
7683 PR middle-end/83850
7684 * expmed.c (extract_bit_field_1): Fix typo.
7686 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7688 PR target/83687
7689 * config/arm/iterators.md (VF): New mode iterator.
7690 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7691 Remove integer-related logic from pattern.
7692 (neon_vabd<mode>_3): Likewise.
7694 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7696 PR middle-end/82694
7697 * common.opt (fstrict-overflow): No longer an alias.
7698 (fwrapv-pointer): New option.
7699 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7700 also for pointer types based on flag_wrapv_pointer.
7701 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7702 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7703 opts->x_flag_wrapv got set.
7704 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7705 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7706 POINTER_TYPE_OVERFLOW_UNDEFINED.
7707 * match.pd: Likewise in address comparison pattern.
7708 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7710 2018-01-15 Richard Biener <rguenther@suse.de>
7712 PR lto/83804
7713 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7714 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7715 Reset type names to their identifier if their TYPE_DECL doesn't
7716 have linkage (and thus is used for ODR and devirt).
7717 (save_debug_info_for_decl): Remove.
7718 (save_debug_info_for_type): Likewise.
7719 (add_tree_to_fld_list): Adjust.
7720 * tree-pretty-print.c (dump_generic_node): Make dumping of
7721 type names more robust.
7723 2018-01-15 Richard Biener <rguenther@suse.de>
7725 * BASE-VER: Bump to 8.0.1.
7727 2018-01-14 Martin Sebor <msebor@redhat.com>
7729 PR other/83508
7730 * builtins.c (check_access): Avoid warning when the no-warning bit
7731 is set.
7733 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7735 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7736 * ira-color (allocno_hard_regs_compare): Likewise.
7738 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7740 PR target/83013
7741 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7742 Use .pushsection/.popsection.
7744 2018-01-14 Martin Sebor <msebor@redhat.com>
7746 PR c++/81327
7747 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7749 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7751 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7752 entry from extra_headers.
7753 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7754 extra_headers, make the list bitwise identical to the i?86-*-* one.
7756 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7758 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7759 -mcmodel=large with -mindirect-branch=thunk,
7760 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7761 -mfunction-return=thunk-extern.
7762 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7763 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7764 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7766 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7768 * config/i386/i386.c (print_reg): Print the name of the full
7769 integer register without '%'.
7770 (ix86_print_operand): Handle 'V'.
7771 * doc/extend.texi: Document 'V' modifier.
7773 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7775 * config/i386/constraints.md (Bs): Disallow memory operand for
7776 -mindirect-branch-register.
7777 (Bw): Likewise.
7778 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7779 (GOT_memory_operand): Likewise.
7780 (call_insn_operand): Likewise.
7781 (sibcall_insn_operand): Likewise.
7782 (GOT32_symbol_operand): Likewise.
7783 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7784 for -mindirect-branch-register.
7785 (tablejump): Likewise.
7786 (*sibcall_memory): Likewise.
7787 (*sibcall_value_memory): Likewise.
7788 Disallow peepholes of indirect call and jump via memory for
7789 -mindirect-branch-register.
7790 (*call_pop): Replace m with Bw.
7791 (*call_value_pop): Likewise.
7792 (*sibcall_pop_memory): Replace m with Bs.
7793 * config/i386/i386.opt (mindirect-branch-register): New option.
7794 * doc/invoke.texi: Document -mindirect-branch-register option.
7796 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7798 * config/i386/i386-protos.h (ix86_output_function_return): New.
7799 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7800 set function_return_type.
7801 (indirect_thunk_name): Add ret_p to indicate thunk for function
7802 return.
7803 (output_indirect_thunk_function): Pass false to
7804 indirect_thunk_name.
7805 (ix86_output_indirect_branch_via_reg): Likewise.
7806 (ix86_output_indirect_branch_via_push): Likewise.
7807 (output_indirect_thunk_function): Create alias for function
7808 return thunk if regno < 0.
7809 (ix86_output_function_return): New function.
7810 (ix86_handle_fndecl_attribute): Handle function_return.
7811 (ix86_attribute_table): Add function_return.
7812 * config/i386/i386.h (machine_function): Add
7813 function_return_type.
7814 * config/i386/i386.md (simple_return_internal): Use
7815 ix86_output_function_return.
7816 (simple_return_internal_long): Likewise.
7817 * config/i386/i386.opt (mfunction-return=): New option.
7818 (indirect_branch): Mention -mfunction-return=.
7819 * doc/extend.texi: Document function_return function attribute.
7820 * doc/invoke.texi: Document -mfunction-return= option.
7822 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7824 * config/i386/i386-opts.h (indirect_branch): New.
7825 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7826 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7827 with local indirect jump when converting indirect call and jump.
7828 (ix86_set_indirect_branch_type): New.
7829 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7830 (indirectlabelno): New.
7831 (indirect_thunk_needed): Likewise.
7832 (indirect_thunk_bnd_needed): Likewise.
7833 (indirect_thunks_used): Likewise.
7834 (indirect_thunks_bnd_used): Likewise.
7835 (INDIRECT_LABEL): Likewise.
7836 (indirect_thunk_name): Likewise.
7837 (output_indirect_thunk): Likewise.
7838 (output_indirect_thunk_function): Likewise.
7839 (ix86_output_indirect_branch_via_reg): Likewise.
7840 (ix86_output_indirect_branch_via_push): Likewise.
7841 (ix86_output_indirect_branch): Likewise.
7842 (ix86_output_indirect_jmp): Likewise.
7843 (ix86_code_end): Call output_indirect_thunk_function if needed.
7844 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7845 needed.
7846 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7847 (ix86_attribute_table): Add indirect_branch.
7848 * config/i386/i386.h (machine_function): Add indirect_branch_type
7849 and has_local_indirect_jump.
7850 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7851 to true.
7852 (tablejump): Likewise.
7853 (*indirect_jump): Use ix86_output_indirect_jmp.
7854 (*tablejump_1): Likewise.
7855 (simple_return_indirect_internal): Likewise.
7856 * config/i386/i386.opt (mindirect-branch=): New option.
7857 (indirect_branch): New.
7858 (keep): Likewise.
7859 (thunk): Likewise.
7860 (thunk-inline): Likewise.
7861 (thunk-extern): Likewise.
7862 * doc/extend.texi: Document indirect_branch function attribute.
7863 * doc/invoke.texi: Document -mindirect-branch= option.
7865 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7867 PR ipa/83051
7868 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7870 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7872 * ipa-inline.c (want_inline_small_function_p): Return false if
7873 inlining has already failed with CIF_FINAL_ERROR.
7874 (update_caller_keys): Call want_inline_small_function_p before
7875 can_inline_edge_p.
7876 (update_callee_keys): Likewise.
7878 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7880 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7881 New function.
7882 (rs6000_quadword_masked_address_p): Likewise.
7883 (quad_aligned_load_p): Likewise.
7884 (quad_aligned_store_p): Likewise.
7885 (const_load_sequence_p): Add comment to describe the outer-most loop.
7886 (mimic_memory_attributes_and_flags): New function.
7887 (rs6000_gen_stvx): Likewise.
7888 (replace_swapped_aligned_store): Likewise.
7889 (rs6000_gen_lvx): Likewise.
7890 (replace_swapped_aligned_load): Likewise.
7891 (replace_swapped_load_constant): Capitalize argument name in
7892 comment describing this function.
7893 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7894 and stores that access quad-word aligned addresses and replace
7895 with stvx or lvx instructions when appropriate.
7896 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7897 New function prototype.
7898 (rs6000_quadword_masked_address_p): Likewise.
7899 (rs6000_gen_lvx): Likewise.
7900 (rs6000_gen_stvx): Likewise.
7901 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7902 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7903 when memory address is aligned.
7904 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7905 this split to select lvx instruction when memory address is aligned.
7906 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7907 instruction when memory address is aligned.
7908 (*vsx_le_perm_load_v16qi): Likewise.
7909 (four unnamed splitters): Modify to select the stvx instruction
7910 when memory is aligned.
7912 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7914 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7915 which appears in the queue multiple times.
7917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7918 Alan Hayward <alan.hayward@arm.com>
7919 David Sherwood <david.sherwood@arm.com>
7921 * tree-vectorizer.h (vec_lower_bound): New structure.
7922 (_loop_vec_info): Add check_nonzero and lower_bounds.
7923 (LOOP_VINFO_CHECK_NONZERO): New macro.
7924 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7925 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7926 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7927 fields. Make seg_len the distance travelled, not including the
7928 access size.
7929 (dr_direction_indicator): Declare.
7930 (dr_zero_step_indicator): Likewise.
7931 (dr_known_forward_stride_p): Likewise.
7932 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7933 tree-ssanames.h.
7934 (runtime_alias_check_p): Allow runtime alias checks with
7935 variable strides.
7936 (operator ==): Compare access_size and align.
7937 (prune_runtime_alias_test_list): Rework for new distinction between
7938 the access_size and seg_len.
7939 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7940 segment lengths.
7941 (get_segment_min_max): New function.
7942 (create_intersect_range_checks): Use it.
7943 (dr_step_indicator): New function.
7944 (dr_direction_indicator): Likewise.
7945 (dr_zero_step_indicator): Likewise.
7946 (dr_known_forward_stride_p): Likewise.
7947 * tree-loop-distribution.c (data_ref_segment_size): Return
7948 DR_STEP * (niters - 1).
7949 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7950 constructor.
7951 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7952 (vect_preserves_scalar_order_p): New function, split out from...
7953 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7954 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7955 (vect_vfa_access_size): New function.
7956 (vect_vfa_align): Likewise.
7957 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7958 (dump_lower_bound): New function.
7959 (vect_check_lower_bound): Likewise.
7960 (vect_small_gap_p): Likewise.
7961 (vectorizable_with_step_bound_p): Likewise.
7962 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7963 depencies if the vectorization factor is 1. Convert the checks
7964 for nonzero steps into checks on the bounds of DR_STEP. Try using
7965 a bunds check for variable steps if the minimum required step is
7966 relatively small. Update calls to the dr_with_seg_len
7967 constructor and to vect_compile_time_alias.
7968 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7969 function.
7970 (vect_loop_versioning): Call it.
7971 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7972 when retrying.
7973 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7975 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7976 Alan Hayward <alan.hayward@arm.com>
7977 David Sherwood <david.sherwood@arm.com>
7979 * doc/sourcebuild.texi (vect_scatter_store): Document.
7980 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7981 optabs.
7982 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7983 Document.
7984 * genopinit.c (main): Add supports_vec_scatter_store and
7985 supports_vec_scatter_store_cached to target_optabs.
7986 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7987 IFN_MASK_SCATTER_STORE.
7988 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7989 functions.
7990 * internal-fn.h (internal_store_fn_p): Declare.
7991 (internal_fn_stored_value_index): Likewise.
7992 * internal-fn.c (scatter_store_direct): New macro.
7993 (expand_scatter_store_optab_fn): New function.
7994 (direct_scatter_store_optab_supported_p): New macro.
7995 (internal_store_fn_p): New function.
7996 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7997 IFN_MASK_SCATTER_STORE.
7998 (internal_fn_mask_index): Likewise.
7999 (internal_fn_stored_value_index): New function.
8000 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8001 for scatter stores.
8002 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8003 * optabs-query.c (supports_vec_scatter_store_p): New function.
8004 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8005 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8006 true for scatter stores.
8007 (vect_gather_scatter_fn_p): Handle scatter stores too.
8008 (vect_check_gather_scatter): Consider using scatter stores if
8009 supports_vec_scatter_store_p.
8010 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8011 scatter stores too.
8012 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8013 internal_fn_stored_value_index.
8014 (check_load_store_masking): Handle scatter stores too.
8015 (vect_get_store_rhs): Make public.
8016 (vectorizable_call): Use internal_store_fn_p.
8017 (vectorizable_store): Handle scatter store internal functions.
8018 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8019 when deciding whether the end of the group has been reached.
8020 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8021 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8022 (mask_scatter_store<mode>): New insns.
8024 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8025 Alan Hayward <alan.hayward@arm.com>
8026 David Sherwood <david.sherwood@arm.com>
8028 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8029 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8030 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8031 function.
8032 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8033 Use vect_truncate_gather_scatter_offset if we can't treat the
8034 operation as a normal gather load or scatter store.
8035 (get_group_load_store_type): Take the gather_scatter_info
8036 as argument. Try using a gather load or scatter store for
8037 single-element groups.
8038 (get_load_store_type): Update calls to get_group_load_store_type
8039 and vect_use_strided_gather_scatters_p.
8041 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8042 Alan Hayward <alan.hayward@arm.com>
8043 David Sherwood <david.sherwood@arm.com>
8045 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8046 optional tree argument.
8047 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8048 null target hooks.
8049 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8050 but continue to use the current value as a fallback.
8051 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8052 to compare the updates.
8053 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8054 (get_load_store_type): Use it when handling a strided access.
8055 (vect_get_strided_load_store_ops): New function.
8056 (vect_get_data_ptr_increment): Likewise.
8057 (vectorizable_load): Handle strided gather loads. Always pass
8058 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8060 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8061 Alan Hayward <alan.hayward@arm.com>
8062 David Sherwood <david.sherwood@arm.com>
8064 * doc/md.texi (gather_load@var{m}): Document.
8065 (mask_gather_load@var{m}): Likewise.
8066 * genopinit.c (main): Add supports_vec_gather_load and
8067 supports_vec_gather_load_cached to target_optabs.
8068 * optabs-tree.c (init_tree_optimization_optabs): Use
8069 ggc_cleared_alloc to allocate target_optabs.
8070 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8071 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8072 functions.
8073 * internal-fn.h (internal_load_fn_p): Declare.
8074 (internal_gather_scatter_fn_p): Likewise.
8075 (internal_fn_mask_index): Likewise.
8076 (internal_gather_scatter_fn_supported_p): Likewise.
8077 * internal-fn.c (gather_load_direct): New macro.
8078 (expand_gather_load_optab_fn): New function.
8079 (direct_gather_load_optab_supported_p): New macro.
8080 (direct_internal_fn_optab): New function.
8081 (internal_load_fn_p): Likewise.
8082 (internal_gather_scatter_fn_p): Likewise.
8083 (internal_fn_mask_index): Likewise.
8084 (internal_gather_scatter_fn_supported_p): Likewise.
8085 * optabs-query.c (supports_at_least_one_mode_p): New function.
8086 (supports_vec_gather_load_p): Likewise.
8087 * optabs-query.h (supports_vec_gather_load_p): Declare.
8088 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8089 and memory_type field.
8090 (NUM_PATTERNS): Bump to 15.
8091 * tree-vect-data-refs.c: Include internal-fn.h.
8092 (vect_gather_scatter_fn_p): New function.
8093 (vect_describe_gather_scatter_call): Likewise.
8094 (vect_check_gather_scatter): Try using internal functions for
8095 gather loads. Recognize existing calls to a gather load function.
8096 (vect_analyze_data_refs): Consider using gather loads if
8097 supports_vec_gather_load_p.
8098 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8099 (vect_get_gather_scatter_offset_type): Likewise.
8100 (vect_convert_mask_for_vectype): Likewise.
8101 (vect_add_conversion_to_patterm): Likewise.
8102 (vect_try_gather_scatter_pattern): Likewise.
8103 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8104 (vect_vect_recog_func_ptrs): Add it.
8105 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8106 internal_fn_mask_index and internal_gather_scatter_fn_p.
8107 (check_load_store_masking): Take the gather_scatter_info as an
8108 argument and handle gather loads.
8109 (vect_get_gather_scatter_ops): New function.
8110 (vectorizable_call): Check internal_load_fn_p.
8111 (vectorizable_load): Likewise. Handle gather load internal
8112 functions.
8113 (vectorizable_store): Update call to check_load_store_masking.
8114 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8115 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8116 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8117 (aarch64_gather_scale_operand_d): New predicates.
8118 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8119 (mask_gather_load<mode>): New insns.
8121 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8122 Alan Hayward <alan.hayward@arm.com>
8123 David Sherwood <david.sherwood@arm.com>
8125 * optabs.def (fold_left_plus_optab): New optab.
8126 * doc/md.texi (fold_left_plus_@var{m}): Document.
8127 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8128 * internal-fn.c (fold_left_direct): Define.
8129 (expand_fold_left_optab_fn): Likewise.
8130 (direct_fold_left_optab_supported_p): Likewise.
8131 * fold-const-call.c (fold_const_fold_left): New function.
8132 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8133 * tree-parloops.c (valid_reduction_p): New function.
8134 (gather_scalar_reductions): Use it.
8135 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8136 (vect_finish_replace_stmt): Declare.
8137 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8138 (needs_fold_left_reduction_p): New function, split out from...
8139 (vect_is_simple_reduction): ...here. Accept reductions that
8140 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8141 (vect_force_simple_reduction): Also store the reduction type in
8142 the assignment's STMT_VINFO_REDUC_TYPE.
8143 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8144 (merge_with_identity): New function.
8145 (vect_expand_fold_left): Likewise.
8146 (vectorize_fold_left_reduction): Likewise.
8147 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8148 scalar phi in place for it. Check for target support and reject
8149 cases that would reassociate the operation. Defer the transform
8150 phase to vectorize_fold_left_reduction.
8151 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8152 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8153 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8155 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8157 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8158 call to ifc_temp_var.
8160 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8161 Alan Hayward <alan.hayward@arm.com>
8162 David Sherwood <david.sherwood@arm.com>
8164 * target.def (legitimize_address_displacement): Take the original
8165 offset as a poly_int.
8166 * targhooks.h (default_legitimize_address_displacement): Update
8167 accordingly.
8168 * targhooks.c (default_legitimize_address_displacement): Likewise.
8169 * doc/tm.texi: Regenerate.
8170 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8171 as an argument, moving assert of ad->disp == ad->disp_term to...
8172 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8173 Try calling targetm.legitimize_address_displacement before expanding
8174 the address rather than afterwards, and adjust for the new interface.
8175 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8176 Match the new hook interface. Handle SVE addresses.
8177 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8178 new hook interface.
8180 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8182 * Makefile.in (OBJS): Add early-remat.o.
8183 * target.def (select_early_remat_modes): New hook.
8184 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8185 * doc/tm.texi: Regenerate.
8186 * targhooks.h (default_select_early_remat_modes): Declare.
8187 * targhooks.c (default_select_early_remat_modes): New function.
8188 * timevar.def (TV_EARLY_REMAT): New timevar.
8189 * passes.def (pass_early_remat): New pass.
8190 * tree-pass.h (make_pass_early_remat): Declare.
8191 * early-remat.c: New file.
8192 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8193 function.
8194 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8196 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8197 Alan Hayward <alan.hayward@arm.com>
8198 David Sherwood <david.sherwood@arm.com>
8200 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8201 vfm1 with a bound_epilog parameter.
8202 (vect_do_peeling): Update calls accordingly, and move the prologue
8203 call earlier in the function. Treat the base bound_epilog as 0 for
8204 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8205 this base when peeling for gaps.
8206 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8207 with fully-masked loops.
8208 (vect_estimate_min_profitable_iters): Handle the single peeled
8209 iteration in that case.
8211 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8212 Alan Hayward <alan.hayward@arm.com>
8213 David Sherwood <david.sherwood@arm.com>
8215 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8216 single-element interleaving even if the size is not a power of 2.
8217 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8218 accesses for single-element interleaving if the group size is
8219 not a power of 2.
8221 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8222 Alan Hayward <alan.hayward@arm.com>
8223 David Sherwood <david.sherwood@arm.com>
8225 * doc/md.texi (fold_extract_last_@var{m}): Document.
8226 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8227 * optabs.def (fold_extract_last_optab): New optab.
8228 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8229 * internal-fn.c (fold_extract_direct): New macro.
8230 (expand_fold_extract_optab_fn): Likewise.
8231 (direct_fold_extract_optab_supported_p): Likewise.
8232 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8233 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8234 EXTRACT_LAST_REDUCTION.
8235 (get_initial_def_for_reduction): Do not create an initial vector
8236 for EXTRACT_LAST_REDUCTION reductions.
8237 (vectorizable_reduction): Leave the scalar phi in place for
8238 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8239 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8240 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8241 transform phase to vectorizable_condition.
8242 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8243 split out from...
8244 (vect_finish_stmt_generation): ...here.
8245 (vect_finish_replace_stmt): New function.
8246 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8247 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8248 pattern.
8249 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8251 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8252 Alan Hayward <alan.hayward@arm.com>
8253 David Sherwood <david.sherwood@arm.com>
8255 * doc/md.texi (extract_last_@var{m}): Document.
8256 * optabs.def (extract_last_optab): New optab.
8257 * internal-fn.def (EXTRACT_LAST): New internal function.
8258 * internal-fn.c (cond_unary_direct): New macro.
8259 (expand_cond_unary_optab_fn): Likewise.
8260 (direct_cond_unary_optab_supported_p): Likewise.
8261 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8262 loops using EXTRACT_LAST.
8263 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8264 (extract_last_<mode>): ...this optab.
8265 (vec_extract<mode><Vel>): Update accordingly.
8267 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8268 Alan Hayward <alan.hayward@arm.com>
8269 David Sherwood <david.sherwood@arm.com>
8271 * target.def (empty_mask_is_expensive): New hook.
8272 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8273 * doc/tm.texi: Regenerate.
8274 * targhooks.h (default_empty_mask_is_expensive): Declare.
8275 * targhooks.c (default_empty_mask_is_expensive): New function.
8276 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8277 if the target says that empty masks are expensive.
8278 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8279 New function.
8280 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8282 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8283 Alan Hayward <alan.hayward@arm.com>
8284 David Sherwood <david.sherwood@arm.com>
8286 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8287 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8288 (vect_use_loop_mask_for_alignment_p): New function.
8289 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8290 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8291 niters_skip argument. Make sure that the first niters_skip elements
8292 of the first iteration are inactive.
8293 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8294 Update call to vect_set_loop_masks_directly.
8295 (get_misalign_in_elems): New function, split out from...
8296 (vect_gen_prolog_loop_niters): ...here.
8297 (vect_update_init_of_dr): Take a code argument that specifies whether
8298 the adjustment should be added or subtracted.
8299 (vect_update_init_of_drs): Likewise.
8300 (vect_prepare_for_masked_peels): New function.
8301 (vect_do_peeling): Skip prologue peeling if we're using a mask
8302 instead. Update call to vect_update_inits_of_drs.
8303 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8304 mask_skip_niters.
8305 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8306 alignment. Do not include the number of peeled iterations in
8307 the minimum threshold in that case.
8308 (vectorizable_induction): Adjust the start value down by
8309 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8310 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8311 Take the number of skipped iterations into account when calculating
8312 the loop bounds.
8313 * tree-vect-stmts.c (vect_gen_while_not): New function.
8315 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8316 Alan Hayward <alan.hayward@arm.com>
8317 David Sherwood <david.sherwood@arm.com>
8319 * doc/sourcebuild.texi (vect_fully_masked): Document.
8320 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8321 default value to 0.
8322 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8323 split out from...
8324 (vect_analyze_loop_2): ...here. Don't check the vectorization
8325 factor against the number of loop iterations if the loop is
8326 fully-masked.
8328 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8329 Alan Hayward <alan.hayward@arm.com>
8330 David Sherwood <david.sherwood@arm.com>
8332 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8333 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8334 (dump_groups): Update accordingly.
8335 (iv_use::mem_type): New member variable.
8336 (address_p): New function.
8337 (record_use): Add a mem_type argument and initialize the new
8338 mem_type field.
8339 (record_group_use): Add a mem_type argument. Use address_p.
8340 Remove obsolete null checks of base_object. Update call to record_use.
8341 (find_interesting_uses_op): Update call to record_group_use.
8342 (find_interesting_uses_cond): Likewise.
8343 (find_interesting_uses_address): Likewise.
8344 (get_mem_type_for_internal_fn): New function.
8345 (find_address_like_use): Likewise.
8346 (find_interesting_uses_stmt): Try find_address_like_use before
8347 calling find_interesting_uses_op.
8348 (addr_offset_valid_p): Use the iv mem_type field as the type
8349 of the addressed memory.
8350 (add_autoinc_candidates): Likewise.
8351 (get_address_cost): Likewise.
8352 (split_small_address_groups_p): Use address_p.
8353 (split_address_groups): Likewise.
8354 (add_iv_candidate_for_use): Likewise.
8355 (autoinc_possible_for_pair): Likewise.
8356 (rewrite_groups): Likewise.
8357 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8358 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8359 (get_alias_ptr_type_for_ptr_address): New function.
8360 (rewrite_use_address): Rewrite address uses in calls that were
8361 identified by find_address_like_use.
8363 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8364 Alan Hayward <alan.hayward@arm.com>
8365 David Sherwood <david.sherwood@arm.com>
8367 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8368 TARGET_MEM_REFs.
8369 * gimple-expr.h (is_gimple_addressable: Likewise.
8370 * gimple-expr.c (is_gimple_address): Likewise.
8371 * internal-fn.c (expand_call_mem_ref): New function.
8372 (expand_mask_load_optab_fn): Use it.
8373 (expand_mask_store_optab_fn): Likewise.
8375 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8376 Alan Hayward <alan.hayward@arm.com>
8377 David Sherwood <david.sherwood@arm.com>
8379 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8380 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8381 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8382 (cond_umax@var{mode}): Document.
8383 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8384 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8385 (cond_umin_optab, cond_umax_optab): New optabs.
8386 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8387 (COND_IOR, COND_XOR): New internal functions.
8388 * internal-fn.h (get_conditional_internal_fn): Declare.
8389 * internal-fn.c (cond_binary_direct): New macro.
8390 (expand_cond_binary_optab_fn): Likewise.
8391 (direct_cond_binary_optab_supported_p): Likewise.
8392 (get_conditional_internal_fn): New function.
8393 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8394 Cope with reduction statements that are vectorized as calls rather
8395 than assignments.
8396 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8397 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8398 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8399 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8400 (UNSPEC_COND_EOR): New unspecs.
8401 (optab): Add mappings for them.
8402 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8403 (sve_int_op, sve_fp_op): New int attributes.
8405 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8406 Alan Hayward <alan.hayward@arm.com>
8407 David Sherwood <david.sherwood@arm.com>
8409 * optabs.def (while_ult_optab): New optab.
8410 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8411 * internal-fn.def (WHILE_ULT): New internal function.
8412 * internal-fn.h (direct_internal_fn_supported_p): New override
8413 that takes two types as argument.
8414 * internal-fn.c (while_direct): New macro.
8415 (expand_while_optab_fn): New function.
8416 (convert_optab_supported_p): Likewise.
8417 (direct_while_optab_supported_p): New macro.
8418 * wide-int.h (wi::udiv_ceil): New function.
8419 * tree-vectorizer.h (rgroup_masks): New structure.
8420 (vec_loop_masks): New typedef.
8421 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8422 and fully_masked_p.
8423 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8424 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8425 (vect_max_vf): New function.
8426 (slpeel_make_loop_iterate_ntimes): Delete.
8427 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8428 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8429 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8430 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8431 internal-fn.h, stor-layout.h and optabs-query.h.
8432 (vect_set_loop_mask): New function.
8433 (add_preheader_seq): Likewise.
8434 (add_header_seq): Likewise.
8435 (interleave_supported_p): Likewise.
8436 (vect_maybe_permute_loop_masks): Likewise.
8437 (vect_set_loop_masks_directly): Likewise.
8438 (vect_set_loop_condition_masked): Likewise.
8439 (vect_set_loop_condition_unmasked): New function, split out from
8440 slpeel_make_loop_iterate_ntimes.
8441 (slpeel_make_loop_iterate_ntimes): Rename to..
8442 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8443 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8444 (vect_do_peeling): Update call accordingly.
8445 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8446 loops.
8447 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8448 mask_compare_type, can_fully_mask_p and fully_masked_p.
8449 (release_vec_loop_masks): New function.
8450 (_loop_vec_info): Use it to free the loop masks.
8451 (can_produce_all_loop_masks_p): New function.
8452 (vect_get_max_nscalars_per_iter): Likewise.
8453 (vect_verify_full_masking): Likewise.
8454 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8455 retries, and free the mask rgroups before retrying. Check loop-wide
8456 reasons for disallowing fully-masked loops. Make the final decision
8457 about whether use a fully-masked loop or not.
8458 (vect_estimate_min_profitable_iters): Do not assume that peeling
8459 for the number of iterations will be needed for fully-masked loops.
8460 (vectorizable_reduction): Disable fully-masked loops.
8461 (vectorizable_live_operation): Likewise.
8462 (vect_halve_mask_nunits): New function.
8463 (vect_double_mask_nunits): Likewise.
8464 (vect_record_loop_mask): Likewise.
8465 (vect_get_loop_mask): Likewise.
8466 (vect_transform_loop): Handle the case in which the final loop
8467 iteration might handle a partial vector. Call vect_set_loop_condition
8468 instead of slpeel_make_loop_iterate_ntimes.
8469 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8470 (check_load_store_masking): New function.
8471 (prepare_load_store_mask): Likewise.
8472 (vectorizable_store): Handle fully-masked loops.
8473 (vectorizable_load): Likewise.
8474 (supportable_widening_operation): Use vect_halve_mask_nunits for
8475 booleans.
8476 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8477 (vect_gen_while): New function.
8478 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8479 (aarch64_uqdec<mode>): New insn.
8481 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8482 Alan Hayward <alan.hayward@arm.com>
8483 David Sherwood <david.sherwood@arm.com>
8485 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8486 (reduc_xor_scal_optab): New optabs.
8487 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8488 (reduc_xor_scal_@var{m}): Document.
8489 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8490 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8491 internal functions.
8492 * fold-const-call.c (fold_const_call): Handle them.
8493 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8494 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8495 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8496 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8497 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8498 (UNSPEC_XORV): New unspecs.
8499 (optab): Add entries for them.
8500 (BITWISEV): New int iterator.
8501 (bit_reduc_op): New int attributes.
8503 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8504 Alan Hayward <alan.hayward@arm.com>
8505 David Sherwood <david.sherwood@arm.com>
8507 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8508 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8509 * optabs.def (vec_shl_insert_optab): New optab.
8510 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8511 (duplicate_and_interleave): Likewise.
8512 * tree-vect-loop.c: Include internal-fn.h.
8513 (neutral_op_for_slp_reduction): New function, split out from
8514 get_initial_defs_for_reduction.
8515 (get_initial_def_for_reduction): Handle option 2 for variable-length
8516 vectors by loading the neutral value into a vector and then shifting
8517 the initial value into element 0.
8518 (get_initial_defs_for_reduction): Replace the code argument with
8519 the neutral value calculated by neutral_op_for_slp_reduction.
8520 Use gimple_build_vector for constant-length vectors.
8521 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8522 but the first group_size elements have a neutral value.
8523 Use duplicate_and_interleave otherwise.
8524 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8525 Update call to get_initial_defs_for_reduction. Handle SLP
8526 reductions for variable-length vectors by creating one vector
8527 result for each scalar result, with the elements associated
8528 with other scalar results stubbed out with the neutral value.
8529 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8530 Require IFN_VEC_SHL_INSERT for double reductions on
8531 variable-length vectors, or SLP reductions that have
8532 a neutral value. Require can_duplicate_and_interleave_p
8533 support for variable-length unchained SLP reductions if there
8534 is no neutral value, such as for MIN/MAX reductions. Also require
8535 the number of vector elements to be a multiple of the number of
8536 SLP statements when doing variable-length unchained SLP reductions.
8537 Update call to vect_create_epilog_for_reduction.
8538 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8539 and remove initial values.
8540 (duplicate_and_interleave): Make public.
8541 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8542 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8544 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8545 Alan Hayward <alan.hayward@arm.com>
8546 David Sherwood <david.sherwood@arm.com>
8548 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8549 (can_duplicate_and_interleave_p): New function.
8550 (vect_get_and_check_slp_defs): Take the vector of statements
8551 rather than just the current one. Remove excess parentheses.
8552 Restriction rejectinon of vect_constant_def and vect_external_def
8553 for variable-length vectors to boolean types, or types for which
8554 can_duplicate_and_interleave_p is false.
8555 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8556 (duplicate_and_interleave): New function.
8557 (vect_get_constant_vectors): Use gimple_build_vector for
8558 constant-length vectors and suitable variable-length constant
8559 vectors. Use duplicate_and_interleave for other variable-length
8560 vectors. Don't defer the update when inserting new statements.
8562 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8563 Alan Hayward <alan.hayward@arm.com>
8564 David Sherwood <david.sherwood@arm.com>
8566 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8567 min_profitable_iters doesn't go negative.
8569 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8570 Alan Hayward <alan.hayward@arm.com>
8571 David Sherwood <david.sherwood@arm.com>
8573 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8574 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8575 * optabs.def (vec_mask_load_lanes_optab): New optab.
8576 (vec_mask_store_lanes_optab): Likewise.
8577 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8578 (MASK_STORE_LANES): Likewise.
8579 * internal-fn.c (mask_load_lanes_direct): New macro.
8580 (mask_store_lanes_direct): Likewise.
8581 (expand_mask_load_optab_fn): Handle masked operations.
8582 (expand_mask_load_lanes_optab_fn): New macro.
8583 (expand_mask_store_optab_fn): Handle masked operations.
8584 (expand_mask_store_lanes_optab_fn): New macro.
8585 (direct_mask_load_lanes_optab_supported_p): Likewise.
8586 (direct_mask_store_lanes_optab_supported_p): Likewise.
8587 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8588 parameter.
8589 (vect_load_lanes_supported): Likewise.
8590 * tree-vect-data-refs.c (strip_conversion): New function.
8591 (can_group_stmts_p): Likewise.
8592 (vect_analyze_data_ref_accesses): Use it instead of checking
8593 for a pair of assignments.
8594 (vect_store_lanes_supported): Take a masked_p parameter.
8595 (vect_load_lanes_supported): Likewise.
8596 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8597 vect_store_lanes_supported and vect_load_lanes_supported.
8598 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8599 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8600 parameter. Don't allow gaps for masked accesses.
8601 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8602 and vect_load_lanes_supported.
8603 (get_load_store_type): Take a masked_p parameter and update
8604 call to get_group_load_store_type.
8605 (vectorizable_store): Update call to get_load_store_type.
8606 Handle IFN_MASK_STORE_LANES.
8607 (vectorizable_load): Update call to get_load_store_type.
8608 Handle IFN_MASK_LOAD_LANES.
8610 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8611 Alan Hayward <alan.hayward@arm.com>
8612 David Sherwood <david.sherwood@arm.com>
8614 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8615 modes for SVE.
8616 * config/aarch64/aarch64-protos.h
8617 (aarch64_sve_struct_memory_operand_p): Declare.
8618 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8619 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8620 (VPRED, vpred): Handle SVE structure modes.
8621 * config/aarch64/constraints.md (Utx): New constraint.
8622 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8623 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8624 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8625 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8626 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8627 structure modes. Split into pieces after RA.
8628 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8629 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8630 New patterns.
8631 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8632 SVE structure modes.
8633 (aarch64_classify_address): Likewise.
8634 (sizetochar): Move earlier in file.
8635 (aarch64_print_operand): Handle SVE register lists.
8636 (aarch64_array_mode): New function.
8637 (aarch64_sve_struct_memory_operand_p): Likewise.
8638 (TARGET_ARRAY_MODE): Redefine.
8640 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8641 Alan Hayward <alan.hayward@arm.com>
8642 David Sherwood <david.sherwood@arm.com>
8644 * target.def (array_mode): New target hook.
8645 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8646 * doc/tm.texi: Regenerate.
8647 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8648 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8649 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8650 targetm.array_mode.
8651 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8652 type sizes.
8654 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8655 Alan Hayward <alan.hayward@arm.com>
8656 David Sherwood <david.sherwood@arm.com>
8658 * fold-const.c (fold_binary_loc): Check the argument types
8659 rather than the result type when testing for a vector operation.
8661 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8663 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8664 * doc/tm.texi: Regenerate.
8666 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8667 Alan Hayward <alan.hayward@arm.com>
8668 David Sherwood <david.sherwood@arm.com>
8670 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8671 (sve): Document new AArch64 extension.
8672 * doc/md.texi (w): Extend the description of the AArch64
8673 constraint to include SVE vectors.
8674 (Upl, Upa): Document new AArch64 predicate constraints.
8675 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8676 enum.
8677 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8678 (msve-vector-bits=): New option.
8679 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8680 SVE when these are disabled.
8681 (sve): New extension.
8682 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8683 modes. Adjust their number of units based on aarch64_sve_vg.
8684 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8685 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8686 aarch64_addr_query_type.
8687 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8688 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8689 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8690 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8691 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8692 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8693 (aarch64_simd_imm_zero_p): Delete.
8694 (aarch64_check_zero_based_sve_index_immediate): Declare.
8695 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8696 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8697 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8698 (aarch64_sve_float_mul_immediate_p): Likewise.
8699 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8700 rather than an rtx.
8701 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8702 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8703 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8704 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8705 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8706 (aarch64_regmode_natural_size): Likewise.
8707 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8708 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8709 left one place.
8710 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8711 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8712 for VG and the SVE predicate registers.
8713 (V_ALIASES): Add a "z"-prefixed alias.
8714 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8715 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8716 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8717 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8718 (REG_CLASS_NAMES): Add entries for them.
8719 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8720 and the predicate registers.
8721 (aarch64_sve_vg): Declare.
8722 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8723 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8724 (REGMODE_NATURAL_SIZE): Define.
8725 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8726 SVE macros.
8727 * config/aarch64/aarch64.c: Include cfgrtl.h.
8728 (simd_immediate_info): Add a constructor for series vectors,
8729 and an associated step field.
8730 (aarch64_sve_vg): New variable.
8731 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8732 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8733 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8734 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8735 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8736 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8737 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8738 (aarch64_get_mask_mode): New functions.
8739 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8740 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8741 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8742 predicate modes and predicate registers. Explicitly restrict
8743 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8744 to store a vector mode if it is recognized by
8745 aarch64_classify_vector_mode.
8746 (aarch64_regmode_natural_size): New function.
8747 (aarch64_hard_regno_caller_save_mode): Return the original mode
8748 for predicates.
8749 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8750 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8751 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8752 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8753 functions.
8754 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8755 does not overlap dest if the function is frame-related. Handle
8756 SVE constants.
8757 (aarch64_split_add_offset): New function.
8758 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8759 them aarch64_add_offset.
8760 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8761 and update call to aarch64_sub_sp.
8762 (aarch64_add_cfa_expression): New function.
8763 (aarch64_expand_prologue): Pass extra temporary registers to the
8764 functions above. Handle the case in which we need to emit new
8765 DW_CFA_expressions for registers that were originally saved
8766 relative to the stack pointer, but now have to be expressed
8767 relative to the frame pointer.
8768 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8769 functions above.
8770 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8771 IP0 and IP1 values for SVE frames.
8772 (aarch64_expand_vec_series): New function.
8773 (aarch64_expand_sve_widened_duplicate): Likewise.
8774 (aarch64_expand_sve_const_vector): Likewise.
8775 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8776 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8777 into the register, rather than emitting a SET directly.
8778 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8779 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8780 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8781 (offset_9bit_signed_scaled_p): New functions.
8782 (aarch64_replicate_bitmask_imm): New function.
8783 (aarch64_bitmask_imm): Use it.
8784 (aarch64_cannot_force_const_mem): Reject expressions involving
8785 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8786 (aarch64_classify_index): Handle SVE indices, by requiring
8787 a plain register index with a scale that matches the element size.
8788 (aarch64_classify_address): Handle SVE addresses. Assert that
8789 the mode of the address is VOIDmode or an integer mode.
8790 Update call to aarch64_classify_symbol.
8791 (aarch64_classify_symbolic_expression): Update call to
8792 aarch64_classify_symbol.
8793 (aarch64_const_vec_all_in_range_p): New function.
8794 (aarch64_print_vector_float_operand): Likewise.
8795 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8796 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8797 and the FP immediates 1.0 and 0.5.
8798 (aarch64_print_address_internal): Handle SVE addresses.
8799 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8800 (aarch64_regno_regclass): Handle predicate registers.
8801 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8802 data modes.
8803 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8804 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8805 (aarch64_convert_sve_vector_bits): New function.
8806 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8807 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8808 rather than an rtx.
8809 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8810 Handle SVE vector and predicate modes. Accept VL-based constants
8811 that need only one temporary register, and VL offsets that require
8812 no temporary registers.
8813 (aarch64_conditional_register_usage): Mark the predicate registers
8814 as fixed if SVE isn't available.
8815 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8816 Return true for SVE vector and predicate modes.
8817 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8818 rather than an unsigned int. Handle SVE modes.
8819 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8820 SVE modes.
8821 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8822 if SVE is enabled.
8823 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8824 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8825 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8826 (aarch64_sve_float_mul_immediate_p): New functions.
8827 (aarch64_sve_valid_immediate): New function.
8828 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8829 Explicitly reject structure modes. Check for INDEX constants.
8830 Handle PTRUE and PFALSE constants.
8831 (aarch64_check_zero_based_sve_index_immediate): New function.
8832 (aarch64_simd_imm_zero_p): Delete.
8833 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8834 vector modes. Accept constants in the range of CNT[BHWD].
8835 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8836 ask for an Advanced SIMD mode.
8837 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8838 (aarch64_simd_vector_alignment): Handle SVE predicates.
8839 (aarch64_vectorize_preferred_vector_alignment): New function.
8840 (aarch64_simd_vector_alignment_reachable): Use it instead of
8841 the vector size.
8842 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8843 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8844 functions.
8845 (MAX_VECT_LEN): Delete.
8846 (expand_vec_perm_d): Add a vec_flags field.
8847 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8848 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8849 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8850 for SVE modes.
8851 (aarch64_evpc_rev): Rename to...
8852 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8853 (aarch64_evpc_rev_global): New function.
8854 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8855 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8856 MAX_VECT_LEN.
8857 (aarch64_evpc_sve_tbl): New function.
8858 (aarch64_expand_vec_perm_const_1): Update after rename of
8859 aarch64_evpc_rev. Handle SVE permutes too, trying
8860 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8861 than aarch64_evpc_tbl.
8862 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8863 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8864 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8865 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8866 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8867 (aarch64_expand_sve_vcond): New functions.
8868 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8869 of aarch64_vector_mode_p.
8870 (aarch64_dwarf_poly_indeterminate_value): New function.
8871 (aarch64_compute_pressure_classes): Likewise.
8872 (aarch64_can_change_mode_class): Likewise.
8873 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8874 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8875 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8876 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8877 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8878 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8879 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8880 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8881 constraints.
8882 (Dn, Dl, Dr): Accept const as well as const_vector.
8883 (Dz): Likewise. Compare against CONST0_RTX.
8884 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8885 of "vector" where appropriate.
8886 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8887 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8888 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8889 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8890 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8891 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8892 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8893 (v_int_equiv): Extend to SVE modes.
8894 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8895 mode attributes.
8896 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8897 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8898 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8899 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8900 (SVE_COND_FP_CMP): New int iterators.
8901 (perm_hilo): Handle the new unpack unspecs.
8902 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8903 attributes.
8904 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8905 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8906 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8907 (aarch64_equality_operator, aarch64_constant_vector_operand)
8908 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8909 (aarch64_sve_nonimmediate_operand): Likewise.
8910 (aarch64_sve_general_operand): Likewise.
8911 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8912 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8913 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8914 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8915 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8916 (aarch64_sve_float_arith_immediate): Likewise.
8917 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8918 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8919 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8920 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8921 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8922 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8923 (aarch64_sve_float_arith_operand): Likewise.
8924 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8925 (aarch64_sve_float_mul_operand): Likewise.
8926 (aarch64_sve_vec_perm_operand): Likewise.
8927 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8928 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8929 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8930 as well as const_vector.
8931 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8932 in file. Use CONST0_RTX and CONSTM1_RTX.
8933 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8934 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8935 Use aarch64_simd_imm_zero.
8936 * config/aarch64/aarch64-sve.md: New file.
8937 * config/aarch64/aarch64.md: Include it.
8938 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8939 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8940 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8941 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8942 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8943 (sve): New attribute.
8944 (enabled): Disable instructions with the sve attribute unless
8945 TARGET_SVE.
8946 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8947 aarch64_expand_mov_immediate.
8948 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8949 CNT[BHSD] immediates.
8950 (movti): Split CONST_POLY_INT moves into two halves.
8951 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8952 Split additions that need a temporary here if the destination
8953 is the stack pointer.
8954 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8955 (*add<mode>3_poly_1): New instruction.
8956 (set_clobber_cc): New expander.
8958 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8960 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8961 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8962 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8963 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8964 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8965 Change innermode from fixed_mode_size to machine_mode.
8966 (simplify_subreg): Update call accordingly. Handle a constant-sized
8967 subreg of a variable-length CONST_VECTOR.
8969 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8970 Alan Hayward <alan.hayward@arm.com>
8971 David Sherwood <david.sherwood@arm.com>
8973 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8974 (add_offset_to_base): New function, split out from...
8975 (create_mem_ref): ...here. When handling a scale other than 1,
8976 check first whether the address is valid without the offset.
8977 Add it into the base if so, leaving the index and scale as-is.
8979 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8981 PR c++/83778
8982 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8983 fold_for_warn before checking if arg2 is INTEGER_CST.
8985 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8987 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8988 (store_multiple_operation): Delete.
8989 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8990 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8991 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8992 guarded by TARGET_STRING.
8993 (rs6000_output_load_multiple): Delete.
8994 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8995 OPTION_MASK_STRING / TARGET_STRING handling.
8996 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8997 (const rs6000_opt_masks) <"string">: Change mask to 0.
8998 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8999 (MASK_STRING): Delete.
9000 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9001 parts. Simplify.
9002 (load_multiple): Delete.
9003 (*ldmsi8): Delete.
9004 (*ldmsi7): Delete.
9005 (*ldmsi6): Delete.
9006 (*ldmsi5): Delete.
9007 (*ldmsi4): Delete.
9008 (*ldmsi3): Delete.
9009 (store_multiple): Delete.
9010 (*stmsi8): Delete.
9011 (*stmsi7): Delete.
9012 (*stmsi6): Delete.
9013 (*stmsi5): Delete.
9014 (*stmsi4): Delete.
9015 (*stmsi3): Delete.
9016 (movmemsi_8reg): Delete.
9017 (corresponding unnamed define_insn): Delete.
9018 (movmemsi_6reg): Delete.
9019 (corresponding unnamed define_insn): Delete.
9020 (movmemsi_4reg): Delete.
9021 (corresponding unnamed define_insn): Delete.
9022 (movmemsi_2reg): Delete.
9023 (corresponding unnamed define_insn): Delete.
9024 (movmemsi_1reg): Delete.
9025 (corresponding unnamed define_insn): Delete.
9026 * config/rs6000/rs6000.opt (mno-string): New.
9027 (mstring): Replace by deprecation warning stub.
9028 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9030 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9032 * regrename.c (regrename_do_replace): If replacing the same
9033 reg multiple times, try to reuse last created gen_raw_REG.
9035 PR debug/81155
9036 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9037 main to workaround a bug in GDB.
9039 2018-01-12 Tom de Vries <tom@codesourcery.com>
9041 PR target/83737
9042 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9044 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9046 PR rtl-optimization/80481
9047 * ira-color.c (get_cap_member): New function.
9048 (allocnos_conflict_by_live_ranges_p): Use it.
9049 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9050 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9052 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9054 PR target/83628
9055 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9056 (*saddl_se_1): Ditto.
9057 (*ssubsi_1): Ditto.
9058 (*ssubl_se_1): Ditto.
9060 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9062 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9063 rather than wi::to_widest for DR_INITs.
9064 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9065 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9066 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9067 INTEGER_CSTs.
9068 (vect_analyze_group_access_1): Note that here.
9070 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9072 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9073 polynomial type sizes.
9075 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9077 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9078 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9079 (gimple_add_tmp_var): Likewise.
9081 2018-01-12 Martin Liska <mliska@suse.cz>
9083 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9084 (gimple_alloc_sizes): Likewise.
9085 (dump_gimple_statistics): Use PRIu64 in printf format.
9086 * gimple.h: Change uint64_t to int.
9088 2018-01-12 Martin Liska <mliska@suse.cz>
9090 * tree-core.h: Use uint64_t instead of int.
9091 * tree.c (tree_node_counts): Likewise.
9092 (tree_node_sizes): Likewise.
9093 (dump_tree_statistics): Use PRIu64 in printf format.
9095 2018-01-12 Martin Liska <mliska@suse.cz>
9097 * Makefile.in: As qsort_chk is implemented in vec.c, add
9098 vec.o to linkage of gencfn-macros.
9099 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9100 passing the info to record_node_allocation_statistics.
9101 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9102 and pass the info.
9103 * ggc-common.c (struct ggc_usage): Add operator== and use
9104 it in operator< and compare function.
9105 * mem-stats.h (struct mem_usage): Likewise.
9106 * vec.c (struct vec_usage): Remove operator< and compare
9107 function. Can be simply inherited.
9109 2018-01-12 Martin Jambor <mjambor@suse.cz>
9111 PR target/81616
9112 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9113 * tree-ssa-math-opts.c: Include domwalk.h.
9114 (convert_mult_to_fma_1): New function.
9115 (fma_transformation_info): New type.
9116 (fma_deferring_state): Likewise.
9117 (cancel_fma_deferring): New function.
9118 (result_of_phi): Likewise.
9119 (last_fma_candidate_feeds_initial_phi): Likewise.
9120 (convert_mult_to_fma): Added deferring logic, split actual
9121 transformation to convert_mult_to_fma_1.
9122 (math_opts_dom_walker): New type.
9123 (math_opts_dom_walker::after_dom_children): New method, body moved
9124 here from pass_optimize_widening_mul::execute, added deferring logic
9125 bits.
9126 (pass_optimize_widening_mul::execute): Moved most of code to
9127 math_opts_dom_walker::after_dom_children.
9128 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9129 * config/i386/i386.c (ix86_option_override_internal): Added
9130 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9132 2018-01-12 Richard Biener <rguenther@suse.de>
9134 PR debug/83157
9135 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9136 inline instance vars.
9138 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9140 PR target/81819
9141 * config/rx/rx.c (rx_is_restricted_memory_address):
9142 Handle SUBREG case.
9144 2018-01-12 Richard Biener <rguenther@suse.de>
9146 PR tree-optimization/80846
9147 * target.def (split_reduction): New target hook.
9148 * targhooks.c (default_split_reduction): New function.
9149 * targhooks.h (default_split_reduction): Declare.
9150 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9151 target requests first reduce vectors by combining low and high
9152 parts.
9153 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9154 (get_vectype_for_scalar_type_and_size): Export.
9155 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9156 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9157 * doc/tm.texi: Regenerate.
9158 * config/i386/i386.c (ix86_split_reduction): Implement
9159 TARGET_VECTORIZE_SPLIT_REDUCTION.
9161 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9163 PR target/83368
9164 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9165 in PIC mode except for TARGET_VXWORKS_RTP.
9166 * config/sparc/sparc.c: Include cfgrtl.h.
9167 (TARGET_INIT_PIC_REG): Define.
9168 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9169 (sparc_pic_register_p): New predicate.
9170 (sparc_legitimate_address_p): Use it.
9171 (sparc_legitimize_pic_address): Likewise.
9172 (sparc_delegitimize_address): Likewise.
9173 (sparc_mode_dependent_address_p): Likewise.
9174 (gen_load_pcrel_sym): Remove 4th parameter.
9175 (load_got_register): Adjust call to above. Remove obsolete stuff.
9176 (sparc_expand_prologue): Do not call load_got_register here.
9177 (sparc_flat_expand_prologue): Likewise.
9178 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9179 (sparc_use_pseudo_pic_reg): New function.
9180 (sparc_init_pic_reg): Likewise.
9181 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9182 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9184 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9186 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9187 Add item for branch_cost.
9189 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9191 PR rtl-optimization/83565
9192 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9193 not extend the result to a larger mode for rotate operations.
9194 (num_sign_bit_copies1): Likewise.
9196 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9198 PR target/40411
9199 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9200 -symbolic.
9201 Use values-Xc.o for -pedantic.
9202 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9204 2018-01-12 Martin Liska <mliska@suse.cz>
9206 PR ipa/83054
9207 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9208 New function.
9209 (possible_polymorphic_call_targets): Use it.
9210 (ipa_devirt): Likewise.
9212 2018-01-12 Martin Liska <mliska@suse.cz>
9214 * profile-count.h (enum profile_quality): Use 0 as invalid
9215 enum value of profile_quality.
9217 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9219 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9220 -mext-string options.
9222 2018-01-12 Richard Biener <rguenther@suse.de>
9224 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9225 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9226 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9227 Likewise.
9228 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9230 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9232 * configure.ac (--with-long-double-format): Add support for the
9233 configuration option to change the default long double format on
9234 PowerPC systems.
9235 * config.gcc (powerpc*-linux*-*): Likewise.
9236 * configure: Regenerate.
9237 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9238 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9239 used without modification.
9241 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9243 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9244 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9245 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9246 MISC_BUILTIN_SPEC_BARRIER.
9247 (rs6000_init_builtins): Likewise.
9248 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9249 enum value.
9250 (speculation_barrier): New define_insn.
9251 * doc/extend.texi: Document __builtin_speculation_barrier.
9253 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9255 PR target/83203
9256 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9257 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9258 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9259 iterators.
9260 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9261 integral modes instead of "ss" and "sd".
9262 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9263 vectors with 32-bit and 64-bit elements.
9264 (vecdupssescalarmodesuffix): New mode attribute.
9265 (vec_dup<mode>): Use it.
9267 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9269 PR target/83330
9270 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9271 frame if argument is passed on stack.
9273 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9275 PR target/82682
9276 * ree.c (combine_reaching_defs): Optimize also
9277 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9278 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9280 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9282 PR middle-end/83189
9283 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9285 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9287 PR middle-end/83718
9288 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9289 after they are computed.
9291 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9293 PR tree-optimization/83695
9294 * gimple-loop-linterchange.cc
9295 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9296 reset cached scev information after interchange.
9297 (pass_linterchange::execute): Remove call to scev_reset_htab.
9299 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9301 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9302 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9303 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9304 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9305 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9306 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9307 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9308 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9309 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9310 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9311 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9312 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9313 (V_lane_reg): Likewise.
9314 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9315 New define_expand.
9316 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9317 (vfmal_lane_low<mode>_intrinsic,
9318 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9319 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9320 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9321 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9322 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9323 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9325 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9327 * config/arm/arm-cpus.in (fp16fml): New feature.
9328 (ALL_SIMD): Add fp16fml.
9329 (armv8.2-a): Add fp16fml as an option.
9330 (armv8.3-a): Likewise.
9331 (armv8.4-a): Add fp16fml as part of fp16.
9332 * config/arm/arm.h (TARGET_FP16FML): Define.
9333 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9334 when appropriate.
9335 * config/arm/arm-modes.def (V2HF): Define.
9336 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9337 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9338 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9339 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9340 vfmsl_low, vfmsl_high): New set of builtins.
9341 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9342 (vfml_op): New code attribute.
9343 (VFMLHALVES): New int iterator.
9344 (VFML, VFMLSEL): New mode attributes.
9345 (V_reg): Define mapping for V2HF.
9346 (V_hi, V_lo): New mode attributes.
9347 (VF_constraint): Likewise.
9348 (vfml_half, vfml_half_selector): New int attributes.
9349 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9350 define_expand.
9351 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9352 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9353 New define_insn.
9354 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9355 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9356 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9357 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9358 documentation.
9359 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9360 Document new effective target and option set.
9362 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9364 * config/arm/arm-cpus.in (armv8_4): New feature.
9365 (ARMv8_4a): New fgroup.
9366 (armv8.4-a): New arch.
9367 * config/arm/arm-tables.opt: Regenerate.
9368 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9369 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9370 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9371 Add matching rules for -march=armv8.4-a and extensions.
9372 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9374 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9376 PR target/81821
9377 * config/rx/rx.md (BW): New mode attribute.
9378 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9380 2018-01-11 Richard Biener <rguenther@suse.de>
9382 PR tree-optimization/83435
9383 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9384 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9385 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9387 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9388 Alan Hayward <alan.hayward@arm.com>
9389 David Sherwood <david.sherwood@arm.com>
9391 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9392 field.
9393 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9394 (aarch64_print_address_internal): Use it to check for a zero offset.
9396 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9397 Alan Hayward <alan.hayward@arm.com>
9398 David Sherwood <david.sherwood@arm.com>
9400 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9401 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9402 Return a poly_int64 rather than a HOST_WIDE_INT.
9403 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9404 rather than a HOST_WIDE_INT.
9405 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9406 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9407 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9408 final_offset from HOST_WIDE_INT to poly_int64.
9409 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9410 to_constant when getting the number of units in an Advanced SIMD
9411 mode.
9412 (aarch64_builtin_vectorized_function): Check for a constant number
9413 of units.
9414 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9415 GET_MODE_SIZE.
9416 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9417 attribute instead of GET_MODE_NUNITS.
9418 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9419 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9420 GET_MODE_SIZE for fixed-size registers.
9421 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9422 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9423 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9424 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9425 (aarch64_print_operand, aarch64_print_address_internal)
9426 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9427 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9428 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9429 Handle polynomial GET_MODE_SIZE.
9430 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9431 wider than SImode without modification.
9432 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9433 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9434 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9435 passing and returning SVE modes.
9436 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9437 rather than GEN_INT.
9438 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9439 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9440 (aarch64_allocate_and_probe_stack_space): Likewise.
9441 (aarch64_layout_frame): Cope with polynomial offsets.
9442 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9443 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9444 polynomial offsets.
9445 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9446 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9447 poly_int64 rather than a HOST_WIDE_INT.
9448 (aarch64_get_separate_components, aarch64_process_components)
9449 (aarch64_expand_prologue, aarch64_expand_epilogue)
9450 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9451 (aarch64_anchor_offset): New function, split out from...
9452 (aarch64_legitimize_address): ...here.
9453 (aarch64_builtin_vectorization_cost): Handle polynomial
9454 TYPE_VECTOR_SUBPARTS.
9455 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9456 GET_MODE_NUNITS.
9457 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9458 number of elements from the PARALLEL rather than the mode.
9459 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9460 rather than GET_MODE_BITSIZE.
9461 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9462 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9463 (aarch64_expand_vec_perm_const_1): Handle polynomial
9464 d->perm.length () and d->perm elements.
9465 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9466 Apply to_constant to d->perm elements.
9467 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9468 polynomial CONST_VECTOR_NUNITS.
9469 (aarch64_move_pointer): Take amount as a poly_int64 rather
9470 than an int.
9471 (aarch64_progress_pointer): Avoid temporary variable.
9472 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9473 the mode attribute instead of GET_MODE.
9475 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9476 Alan Hayward <alan.hayward@arm.com>
9477 David Sherwood <david.sherwood@arm.com>
9479 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9480 x exists before using it.
9481 (aarch64_add_constant_internal): Rename to...
9482 (aarch64_add_offset_1): ...this. Replace regnum with separate
9483 src and dest rtxes. Handle the case in which they're different,
9484 including when the offset is zero. Replace scratchreg with an rtx.
9485 Use 2 additions if there is no spare register into which we can
9486 move a 16-bit constant.
9487 (aarch64_add_constant): Delete.
9488 (aarch64_add_offset): Replace reg with separate src and dest
9489 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9490 Use aarch64_add_offset_1.
9491 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9492 an rtx rather than an int. Take the delta as a poly_int64
9493 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9494 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9495 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9496 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9497 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9498 and aarch64_add_sp.
9499 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9500 aarch64_add_constant.
9502 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9504 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9505 Use scalar_float_mode.
9507 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9509 * config/aarch64/aarch64-simd.md
9510 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9511 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9512 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9513 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9514 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9515 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9516 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9517 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9518 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9519 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9521 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9523 PR target/83514
9524 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9525 targ_options->x_arm_arch_string is non NULL.
9527 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9529 * config/aarch64/aarch64.h
9530 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9532 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9534 PR target/82096
9535 * expmed.c (emit_store_flag_force): Swap if const op0
9536 and change VOIDmode to mode of op0.
9538 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9540 PR rtl-optimization/83761
9541 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9542 than bytes to mode_for_size.
9544 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9546 PR middle-end/83189
9547 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9548 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9549 profile.
9551 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9553 PR middle-end/83575
9554 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9555 when in layout mode.
9556 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9557 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9558 partition fixup.
9560 2018-01-10 Michael Collison <michael.collison@arm.com>
9562 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9563 * config/aarch64/aarch64-option-extension.def: Add
9564 AARCH64_OPT_EXTENSION of 'fp16fml'.
9565 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9566 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9567 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9568 * config/aarch64/constraints.md (Ui7): New constraint.
9569 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9570 (VFMLA_SEL_W): Ditto.
9571 (f16quad): Ditto.
9572 (f16mac1): Ditto.
9573 (VFMLA16_LOW): New int iterator.
9574 (VFMLA16_HIGH): Ditto.
9575 (UNSPEC_FMLAL): New unspec.
9576 (UNSPEC_FMLSL): Ditto.
9577 (UNSPEC_FMLAL2): Ditto.
9578 (UNSPEC_FMLSL2): Ditto.
9579 (f16mac): New code attribute.
9580 * config/aarch64/aarch64-simd-builtins.def
9581 (aarch64_fmlal_lowv2sf): Ditto.
9582 (aarch64_fmlsl_lowv2sf): Ditto.
9583 (aarch64_fmlalq_lowv4sf): Ditto.
9584 (aarch64_fmlslq_lowv4sf): Ditto.
9585 (aarch64_fmlal_highv2sf): Ditto.
9586 (aarch64_fmlsl_highv2sf): Ditto.
9587 (aarch64_fmlalq_highv4sf): Ditto.
9588 (aarch64_fmlslq_highv4sf): Ditto.
9589 (aarch64_fmlal_lane_lowv2sf): Ditto.
9590 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9591 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9592 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9593 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9594 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9595 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9596 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9597 (aarch64_fmlal_lane_highv2sf): Ditto.
9598 (aarch64_fmlsl_lane_highv2sf): Ditto.
9599 (aarch64_fmlal_laneq_highv2sf): Ditto.
9600 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9601 (aarch64_fmlalq_lane_highv4sf): Ditto.
9602 (aarch64_fmlsl_lane_highv4sf): Ditto.
9603 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9604 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9605 * config/aarch64/aarch64-simd.md:
9606 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9607 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9608 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9609 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9610 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9611 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9612 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9613 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9614 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9615 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9616 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9617 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9618 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9619 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9620 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9621 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9622 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9623 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9624 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9625 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9626 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9627 (vfmlsl_low_u32): Ditto.
9628 (vfmlalq_low_u32): Ditto.
9629 (vfmlslq_low_u32): Ditto.
9630 (vfmlal_high_u32): Ditto.
9631 (vfmlsl_high_u32): Ditto.
9632 (vfmlalq_high_u32): Ditto.
9633 (vfmlslq_high_u32): Ditto.
9634 (vfmlal_lane_low_u32): Ditto.
9635 (vfmlsl_lane_low_u32): Ditto.
9636 (vfmlal_laneq_low_u32): Ditto.
9637 (vfmlsl_laneq_low_u32): Ditto.
9638 (vfmlalq_lane_low_u32): Ditto.
9639 (vfmlslq_lane_low_u32): Ditto.
9640 (vfmlalq_laneq_low_u32): Ditto.
9641 (vfmlslq_laneq_low_u32): Ditto.
9642 (vfmlal_lane_high_u32): Ditto.
9643 (vfmlsl_lane_high_u32): Ditto.
9644 (vfmlal_laneq_high_u32): Ditto.
9645 (vfmlsl_laneq_high_u32): Ditto.
9646 (vfmlalq_lane_high_u32): Ditto.
9647 (vfmlslq_lane_high_u32): Ditto.
9648 (vfmlalq_laneq_high_u32): Ditto.
9649 (vfmlslq_laneq_high_u32): Ditto.
9650 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9651 (AARCH64_FL_FOR_ARCH8_4): New.
9652 (AARCH64_ISA_F16FML): New ISA flag.
9653 (TARGET_F16FML): New feature flag for fp16fml.
9654 (doc/invoke.texi): Document new fp16fml option.
9656 2018-01-10 Michael Collison <michael.collison@arm.com>
9658 * config/aarch64/aarch64-builtins.c:
9659 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9660 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9661 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9662 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9663 (AARCH64_ISA_SHA3): New ISA flag.
9664 (TARGET_SHA3): New feature flag for sha3.
9665 * config/aarch64/iterators.md (sha512_op): New int attribute.
9666 (CRYPTO_SHA512): New int iterator.
9667 (UNSPEC_SHA512H): New unspec.
9668 (UNSPEC_SHA512H2): Ditto.
9669 (UNSPEC_SHA512SU0): Ditto.
9670 (UNSPEC_SHA512SU1): Ditto.
9671 * config/aarch64/aarch64-simd-builtins.def
9672 (aarch64_crypto_sha512hqv2di): New builtin.
9673 (aarch64_crypto_sha512h2qv2di): Ditto.
9674 (aarch64_crypto_sha512su0qv2di): Ditto.
9675 (aarch64_crypto_sha512su1qv2di): Ditto.
9676 (aarch64_eor3qv8hi): Ditto.
9677 (aarch64_rax1qv2di): Ditto.
9678 (aarch64_xarqv2di): Ditto.
9679 (aarch64_bcaxqv8hi): Ditto.
9680 * config/aarch64/aarch64-simd.md:
9681 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9682 (aarch64_crypto_sha512su0qv2di): Ditto.
9683 (aarch64_crypto_sha512su1qv2di): Ditto.
9684 (aarch64_eor3qv8hi): Ditto.
9685 (aarch64_rax1qv2di): Ditto.
9686 (aarch64_xarqv2di): Ditto.
9687 (aarch64_bcaxqv8hi): Ditto.
9688 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9689 (vsha512h2q_u64): Ditto.
9690 (vsha512su0q_u64): Ditto.
9691 (vsha512su1q_u64): Ditto.
9692 (veor3q_u16): Ditto.
9693 (vrax1q_u64): Ditto.
9694 (vxarq_u64): Ditto.
9695 (vbcaxq_u16): Ditto.
9696 * config/arm/types.md (crypto_sha512): New type attribute.
9697 (crypto_sha3): Ditto.
9698 (doc/invoke.texi): Document new sha3 option.
9700 2018-01-10 Michael Collison <michael.collison@arm.com>
9702 * config/aarch64/aarch64-builtins.c:
9703 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9704 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9705 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9706 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9707 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9708 (AARCH64_ISA_SM4): New ISA flag.
9709 (TARGET_SM4): New feature flag for sm4.
9710 * config/aarch64/aarch64-simd-builtins.def
9711 (aarch64_sm3ss1qv4si): Ditto.
9712 (aarch64_sm3tt1aq4si): Ditto.
9713 (aarch64_sm3tt1bq4si): Ditto.
9714 (aarch64_sm3tt2aq4si): Ditto.
9715 (aarch64_sm3tt2bq4si): Ditto.
9716 (aarch64_sm3partw1qv4si): Ditto.
9717 (aarch64_sm3partw2qv4si): Ditto.
9718 (aarch64_sm4eqv4si): Ditto.
9719 (aarch64_sm4ekeyqv4si): Ditto.
9720 * config/aarch64/aarch64-simd.md:
9721 (aarch64_sm3ss1qv4si): Ditto.
9722 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9723 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9724 (aarch64_sm4eqv4si): Ditto.
9725 (aarch64_sm4ekeyqv4si): Ditto.
9726 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9727 (sm3part_op): Ditto.
9728 (CRYPTO_SM3TT): Ditto.
9729 (CRYPTO_SM3PART): Ditto.
9730 (UNSPEC_SM3SS1): New unspec.
9731 (UNSPEC_SM3TT1A): Ditto.
9732 (UNSPEC_SM3TT1B): Ditto.
9733 (UNSPEC_SM3TT2A): Ditto.
9734 (UNSPEC_SM3TT2B): Ditto.
9735 (UNSPEC_SM3PARTW1): Ditto.
9736 (UNSPEC_SM3PARTW2): Ditto.
9737 (UNSPEC_SM4E): Ditto.
9738 (UNSPEC_SM4EKEY): Ditto.
9739 * config/aarch64/constraints.md (Ui2): New constraint.
9740 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9741 * config/arm/types.md (crypto_sm3): New type attribute.
9742 (crypto_sm4): Ditto.
9743 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9744 (vsm3tt1aq_u32): Ditto.
9745 (vsm3tt1bq_u32): Ditto.
9746 (vsm3tt2aq_u32): Ditto.
9747 (vsm3tt2bq_u32): Ditto.
9748 (vsm3partw1q_u32): Ditto.
9749 (vsm3partw2q_u32): Ditto.
9750 (vsm4eq_u32): Ditto.
9751 (vsm4ekeyq_u32): Ditto.
9752 (doc/invoke.texi): Document new sm4 option.
9754 2018-01-10 Michael Collison <michael.collison@arm.com>
9756 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9757 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9758 (AARCH64_FL_FOR_ARCH8_4): New.
9759 (AARCH64_FL_V8_4): New flag.
9760 (doc/invoke.texi): Document new armv8.4-a option.
9762 2018-01-10 Michael Collison <michael.collison@arm.com>
9764 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9765 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9766 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9767 * config/aarch64/aarch64-option-extension.def: Add
9768 AARCH64_OPT_EXTENSION of 'sha2'.
9769 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9770 (crypto): Disable sha2 and aes if crypto disabled.
9771 (crypto): Enable aes and sha2 if enabled.
9772 (simd): Disable sha2 and aes if simd disabled.
9773 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9774 New flags.
9775 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9776 (TARGET_SHA2): New feature flag for sha2.
9777 (TARGET_AES): New feature flag for aes.
9778 * config/aarch64/aarch64-simd.md:
9779 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9780 conditional on TARGET_AES.
9781 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9782 (aarch64_crypto_sha1hsi): Make pattern conditional
9783 on TARGET_SHA2.
9784 (aarch64_crypto_sha1hv4si): Ditto.
9785 (aarch64_be_crypto_sha1hv4si): Ditto.
9786 (aarch64_crypto_sha1su1v4si): Ditto.
9787 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9788 (aarch64_crypto_sha1su0v4si): Ditto.
9789 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9790 (aarch64_crypto_sha256su0v4si): Ditto.
9791 (aarch64_crypto_sha256su1v4si): Ditto.
9792 (doc/invoke.texi): Document new aes and sha2 options.
9794 2018-01-10 Martin Sebor <msebor@redhat.com>
9796 PR tree-optimization/83781
9797 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9798 as string arrays.
9800 2018-01-11 Martin Sebor <msebor@gmail.com>
9801 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9803 PR tree-optimization/83501
9804 PR tree-optimization/81703
9806 * tree-ssa-strlen.c (get_string_cst): Rename...
9807 (get_string_len): ...to this. Handle global constants.
9808 (handle_char_store): Adjust.
9810 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9811 Jim Wilson <jimw@sifive.com>
9813 * config/riscv/riscv-protos.h (riscv_output_return): New.
9814 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9815 (riscv_attribute_table, riscv_output_return),
9816 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9817 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9818 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9819 function.
9820 (riscv_expand_prologue): Add early return for naked function.
9821 (riscv_expand_epilogue): Likewise.
9822 (riscv_function_ok_for_sibcall): Return false for naked function.
9823 (riscv_set_current_function): New.
9824 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9825 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9826 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9827 * doc/extend.texi (RISC-V Function Attributes): New.
9829 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9831 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9832 check for 128-bit long double before checking TCmode.
9833 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9834 128-bit long doubles before checking TFmode or TCmode.
9835 (FLOAT128_IBM_P): Likewise.
9837 2018-01-10 Martin Sebor <msebor@redhat.com>
9839 PR tree-optimization/83671
9840 * builtins.c (c_strlen): Unconditionally return zero for the empty
9841 string.
9842 Use -Warray-bounds for warnings.
9843 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9844 for non-constant array indices with COMPONENT_REF, arrays of
9845 arrays, and pointers to arrays.
9846 (gimple_fold_builtin_strlen): Determine and set length range for
9847 non-constant character arrays.
9849 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9851 PR middle-end/81897
9852 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9853 empty blocks.
9855 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9857 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9859 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9861 PR target/83399
9862 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9863 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9864 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9865 indexed_or_indirect_operand predicate.
9866 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9867 (*vsx_le_perm_load_v8hi): Likewise.
9868 (*vsx_le_perm_load_v16qi): Likewise.
9869 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9870 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9871 (*vsx_le_perm_store_v8hi): Likewise.
9872 (*vsx_le_perm_store_v16qi): Likewise.
9873 (eight unnamed splitters): Likewise.
9875 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9877 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9878 * config/rs6000/emmintrin.h: Likewise.
9879 * config/rs6000/mmintrin.h: Likewise.
9880 * config/rs6000/xmmintrin.h: Likewise.
9882 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9884 PR c++/43486
9885 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9886 "public_flag".
9887 * tree.c (tree_nop_conversion): Return true for location wrapper
9888 nodes.
9889 (maybe_wrap_with_location): New function.
9890 (selftest::check_strip_nops): New function.
9891 (selftest::test_location_wrappers): New function.
9892 (selftest::tree_c_tests): Call it.
9893 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9894 (maybe_wrap_with_location): New decl.
9895 (EXPR_LOCATION_WRAPPER_P): New macro.
9896 (location_wrapper_p): New inline function.
9897 (tree_strip_any_location_wrapper): New inline function.
9899 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9901 PR target/83735
9902 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9903 stack_realign_offset for the largest alignment of stack slot
9904 actually used.
9905 (ix86_find_max_used_stack_alignment): New function.
9906 (ix86_finalize_stack_frame_flags): Use it. Set
9907 max_used_stack_alignment if we don't realign stack.
9908 * config/i386/i386.h (machine_function): Add
9909 max_used_stack_alignment.
9911 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9913 * config/arm/arm.opt (-mbranch-cost): New option.
9914 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9915 account.
9917 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9919 PR target/83629
9920 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9921 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9923 2018-01-10 Richard Biener <rguenther@suse.de>
9925 PR debug/83765
9926 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9927 early out so it also covers the case where we have a non-NULL
9928 origin.
9930 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9932 PR tree-optimization/83753
9933 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9934 for non-strided grouped accesses if the number of elements is 1.
9936 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9938 PR target/81616
9939 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9940 * i386.h (TARGET_USE_GATHER): Define.
9941 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9943 2018-01-10 Martin Liska <mliska@suse.cz>
9945 PR bootstrap/82831
9946 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9947 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9948 partitioning.
9949 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9950 CLEANUP_NO_PARTITIONING is not set.
9952 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9954 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9955 for vectors, as a partial revert of r254296.
9956 * rtl.h (const_vec_p): Delete.
9957 (const_vec_duplicate_p): Don't test for vector CONSTs.
9958 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9959 * expmed.c (make_tree): Likewise.
9961 Revert:
9962 * common.md (E, F): Use CONSTANT_P instead of checking for
9963 CONST_VECTOR.
9964 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9965 checking for CONST_VECTOR.
9967 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9969 PR middle-end/83575
9970 * predict.c (force_edge_cold): Handle in more sane way edges
9971 with no prediction.
9973 2018-01-09 Carl Love <cel@us.ibm.com>
9975 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9976 V4SI, V4SF types.
9977 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9978 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9979 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9980 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9981 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9982 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9983 * config/rs6000/rs6000-protos.h: Add extern defition for
9984 rs6000_generate_float2_double_code.
9985 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9986 function.
9987 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9988 (float2_v2df): Add define_expand.
9990 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9992 PR target/83628
9993 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9994 op_mode in the force_to_mode call.
9996 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9998 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9999 instead of checking each element individually.
10000 (aarch64_evpc_uzp): Likewise.
10001 (aarch64_evpc_zip): Likewise.
10002 (aarch64_evpc_ext): Likewise.
10003 (aarch64_evpc_rev): Likewise.
10004 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10005 instead of checking each element individually. Return true without
10006 generating rtl if
10007 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10008 whether all selected elements come from the same input, instead of
10009 checking each element individually. Remove calls to gen_rtx_REG,
10010 start_sequence and end_sequence and instead assert that no rtl is
10011 generated.
10013 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10015 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10016 order of HIGH and CONST checks.
10018 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10020 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10021 if the destination isn't an SSA_NAME.
10023 2018-01-09 Richard Biener <rguenther@suse.de>
10025 PR tree-optimization/83668
10026 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10027 move prologue...
10028 (canonicalize_loop_form): ... here, renamed from ...
10029 (canonicalize_loop_closed_ssa_form): ... this and amended to
10030 swap successor edges for loop exit blocks to make us use
10031 the RPO order we need for initial schedule generation.
10033 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10035 PR tree-optimization/64811
10036 * match.pd: When optimizing comparisons with Inf, avoid
10037 introducing or losing exceptions from comparisons with NaN.
10039 2018-01-09 Martin Liska <mliska@suse.cz>
10041 PR sanitizer/82517
10042 * asan.c (shadow_mem_size): Add gcc_assert.
10044 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10046 Don't save registers in main().
10048 PR target/83738
10049 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10050 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10051 * config/avr/avr.c (avr_set_current_function): Don't error if
10052 naked, OS_task or OS_main are specified at the same time.
10053 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10054 OS_main.
10055 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10056 attribute.
10057 * common/config/avr/avr-common.c (avr_option_optimization_table):
10058 Switch on -mmain-is-OS_task for optimizing compilations.
10060 2018-01-09 Richard Biener <rguenther@suse.de>
10062 PR tree-optimization/83572
10063 * graphite.c: Include cfganal.h.
10064 (graphite_transform_loops): Connect infinite loops to exit
10065 and remove fake edges at the end.
10067 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10069 * ipa-inline.c (edge_badness): Revert accidental checkin.
10071 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10073 PR ipa/80763
10074 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10075 symbols; not inline clones.
10077 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10079 PR target/83507
10080 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10081 hard registers. Formatting fixes.
10083 PR preprocessor/83722
10084 * gcc.c (try_generate_repro): Pass
10085 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10086 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10087 do_report_bug.
10089 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10090 Kito Cheng <kito.cheng@gmail.com>
10092 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10093 (riscv_leaf_function_p): Delete.
10094 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10096 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10098 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10099 function.
10100 (do_ifelse): New function.
10101 (do_isel): New function.
10102 (do_sub3): New function.
10103 (do_add3): New function.
10104 (do_load_mask_compare): New function.
10105 (do_overlap_load_compare): New function.
10106 (expand_compare_loop): New function.
10107 (expand_block_compare): Call expand_compare_loop() when appropriate.
10108 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10109 option description.
10110 (-mblock-compare-inline-loop-limit): New option.
10112 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10114 PR target/83677
10115 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10116 Reverse order of second and third operands in first alternative.
10117 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10118 of first and second elements in UNSPEC_VPERMR vector.
10119 (altivec_expand_vec_perm_le): Likewise.
10121 2018-01-08 Jeff Law <law@redhat.com>
10123 PR rtl-optimizatin/81308
10124 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10125 (process_switch): If group_case_labels makes a change, then set
10126 cfg_altered.
10127 (pass_convert_switch::execute): If a switch is converted, then
10128 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10130 PR rtl-optimization/81308
10131 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10132 splitting insns.
10134 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10136 PR target/83663 - Revert r255946
10137 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10138 generation for cases where splatting a value is not useful.
10139 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10140 across a vec_duplicate and a paradoxical subreg forming a vector
10141 mode to a vec_concat.
10143 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10145 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10146 -march=armv8.3-a variants.
10147 * config/arm/t-multilib: Likewise.
10148 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10150 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10152 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10153 to generate rtl.
10154 (cceq_ior_compare_complement): Give it a name so I can use it, and
10155 change boolean_or_operator predicate to boolean_operator so it can
10156 be used to generate a crand.
10157 (eqne): New code iterator.
10158 (bd/bd_neg): New code_attrs.
10159 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10160 a single define_insn.
10161 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10162 decrement (bdnzt/bdnzf/bdzt/bdzf).
10163 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10164 with the new names of the branch decrement patterns, and added the
10165 names of the branch decrement conditional patterns.
10167 2018-01-08 Richard Biener <rguenther@suse.de>
10169 PR tree-optimization/83563
10170 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10171 cache.
10173 2018-01-08 Richard Biener <rguenther@suse.de>
10175 PR middle-end/83713
10176 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10178 2018-01-08 Richard Biener <rguenther@suse.de>
10180 PR tree-optimization/83685
10181 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10182 references to abnormals.
10184 2018-01-08 Richard Biener <rguenther@suse.de>
10186 PR lto/83719
10187 * dwarf2out.c (output_indirect_strings): Handle empty
10188 skeleton_debug_str_hash.
10189 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10191 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10193 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10194 (emit_store_direct): Likewise.
10195 (arc_trampoline_adjust_address): Likewise.
10196 (arc_asm_trampoline_template): New function.
10197 (arc_initialize_trampoline): Use asm_trampoline_template.
10198 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10199 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10200 * config/arc/arc.md (flush_icache): Delete pattern.
10202 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10204 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10205 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10206 munaligned-access.
10208 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10210 PR target/83681
10211 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10212 by not USED_FOR_TARGET.
10213 (make_pass_resolve_sw_modes): Likewise.
10215 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10217 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10218 USED_FOR_TARGET.
10220 2018-01-08 Richard Biener <rguenther@suse.de>
10222 PR middle-end/83580
10223 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10225 2018-01-08 Richard Biener <rguenther@suse.de>
10227 PR middle-end/83517
10228 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10230 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10232 PR middle-end/81897
10233 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10234 basic blocks with a small number of successors.
10235 (convert_control_dep_chain_into_preds): Improve handling of
10236 forwarder blocks.
10237 (dump_predicates): Split apart into...
10238 (dump_pred_chain): ...here...
10239 (dump_pred_info): ...and here.
10240 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10241 (can_chain_union_be_invalidated_p): Improve check for invalidation
10242 of paths.
10243 (uninit_uses_cannot_happen): Avoid unnecessary if
10244 convert_control_dep_chain_into_preds yielded nothing.
10246 2018-01-06 Martin Sebor <msebor@redhat.com>
10248 PR tree-optimization/83640
10249 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10250 subtracting negative offset from size.
10251 (builtin_access::overlap): Adjust offset bounds of the access to fall
10252 within the size of the object if possible.
10254 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10256 PR rtl-optimization/83699
10257 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10258 extract_bit_field_as_subreg to cases in which the extracted
10259 value is also a vector.
10261 * lra-constraints.c (process_alt_operands): Test for the equivalence
10262 substitutions when detecting a possible reload cycle.
10264 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10266 PR debug/83480
10267 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10268 by default if flag_selective_schedling{,2}. Formatting fixes.
10270 PR rtl-optimization/83682
10271 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10272 if it has non-VECTOR_MODE element mode.
10273 (vec_duplicate_p): Likewise.
10275 PR middle-end/83694
10276 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10277 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10279 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10281 PR target/83604
10282 * config/i386/i386-builtin.def
10283 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10284 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10285 Require also OPTION_MASK_ISA_AVX512F in addition to
10286 OPTION_MASK_ISA_GFNI.
10287 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10288 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10289 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10290 to OPTION_MASK_ISA_GFNI.
10291 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10292 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10293 OPTION_MASK_ISA_AVX512BW.
10294 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10295 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10296 addition to OPTION_MASK_ISA_GFNI.
10297 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10298 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10299 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10300 to OPTION_MASK_ISA_GFNI.
10301 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10302 a requirement for all ISAs rather than any of them with a few
10303 exceptions.
10304 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10305 processing.
10306 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10307 bitmasks to be enabled with 3 exceptions, instead of requiring any
10308 enabled ISA with lots of exceptions.
10309 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10310 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10311 Change avx512bw in isa attribute to avx512f.
10312 * config/i386/sgxintrin.h: Add license boilerplate.
10313 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10314 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10315 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10316 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10317 defined.
10318 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10319 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10320 temporarily sse2 rather than sse if not enabled already.
10322 PR target/83604
10323 * config/i386/sse.md (VI248_VLBW): Rename to ...
10324 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10325 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10326 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10327 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10328 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10329 mode iterator instead of VI248_VLBW.
10331 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10333 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10334 (record_modified): Skip clobbers; add debug output.
10335 (param_change_prob): Use sreal frequencies.
10337 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10339 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10340 punt for user-aligned variables.
10342 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10344 * tree-chrec.c (chrec_contains_symbols): Return true for
10345 POLY_INT_CST.
10347 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10349 PR target/82439
10350 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10351 of (x|y) == x for BICS pattern.
10353 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10355 PR tree-optimization/83605
10356 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10357 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10358 can throw.
10360 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10362 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10363 * config/epiphany/rtems.h: New file.
10365 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10366 Uros Bizjak <ubizjak@gmail.com>
10368 PR target/83554
10369 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10370 QIreg_operand instead of register_operand predicate.
10371 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10372 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10373 comments instead of -fmitigate[-_]rop.
10375 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10377 PR bootstrap/81926
10378 * cgraphunit.c (symbol_table::compile): Switch to text_section
10379 before calling assembly_start debug hook.
10380 * run-rtl-passes.c (run_rtl_passes): Likewise.
10381 Include output.h.
10383 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10385 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10386 range_int_cst_p rather than !symbolic_range_p before calling
10387 extract_range_from_multiplicative_op_1.
10389 2018-01-04 Jeff Law <law@redhat.com>
10391 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10392 redundant test in assertion.
10394 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10396 * doc/rtl.texi: Document machine_mode wrapper classes.
10398 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10400 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10401 using tree_to_uhwi.
10403 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10405 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10406 the VEC_PERM_EXPR fold to fail.
10408 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10410 PR debug/83585
10411 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10412 to switched_sections.
10414 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10416 PR target/83680
10417 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10418 test for d.testing.
10420 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10422 PR target/83387
10423 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10424 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10426 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10428 PR debug/83666
10429 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10430 is BLKmode and bitpos not zero or mode change is needed.
10432 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10434 PR target/83675
10435 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10436 TARGET_VIS2.
10438 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10440 PR target/83628
10441 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10442 instead of MULT rtx. Update all corresponding splitters.
10443 (*saddl_se): Ditto.
10444 (*ssub<modesuffix>): Ditto.
10445 (*ssubl_se): Ditto.
10446 (*cmp_sadd_di): Update split patterns.
10447 (*cmp_sadd_si): Ditto.
10448 (*cmp_sadd_sidi): Ditto.
10449 (*cmp_ssub_di): Ditto.
10450 (*cmp_ssub_si): Ditto.
10451 (*cmp_ssub_sidi): Ditto.
10452 * config/alpha/predicates.md (const23_operand): New predicate.
10453 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10454 Look for ASHIFT, not MULT inner operand.
10455 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10457 2018-01-04 Martin Liska <mliska@suse.cz>
10459 PR gcov-profile/83669
10460 * gcov.c (output_intermediate_file): Add version to intermediate
10461 gcov file.
10462 * doc/gcov.texi: Document new field 'version' in intermediate
10463 file format. Fix location of '-k' option of gcov command.
10465 2018-01-04 Martin Liska <mliska@suse.cz>
10467 PR ipa/82352
10468 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10470 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10472 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10474 2018-01-03 Martin Sebor <msebor@redhat.com>
10476 PR tree-optimization/83655
10477 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10478 checking calls with invalid arguments.
10480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10482 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10483 (vectorizable_mask_load_store): Delete.
10484 (vectorizable_call): Return false for masked loads and stores.
10485 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10486 instead of gimple_assign_rhs1.
10487 (vectorizable_load): Handle IFN_MASK_LOAD.
10488 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10492 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10493 split out from..,
10494 (vectorizable_mask_load_store): ...here.
10495 (vectorizable_load): ...and here.
10497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10499 * tree-vect-stmts.c (vect_build_all_ones_mask)
10500 (vect_build_zero_merge_argument): New functions, split out from...
10501 (vectorizable_load): ...here.
10503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10505 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10506 split out from...
10507 (vectorizable_mask_load_store): ...here.
10508 (vectorizable_store): ...and here.
10510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10512 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10513 split out from...
10514 (vectorizable_mask_load_store): ...here.
10516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10518 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10519 (vect_model_store_cost): Take a vec_load_store_type instead of a
10520 vect_def_type.
10521 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10522 (vect_model_store_cost): Take a vec_load_store_type instead of a
10523 vect_def_type.
10524 (vectorizable_mask_load_store): Update accordingly.
10525 (vectorizable_store): Likewise.
10526 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10530 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10531 IFN_MASK_LOAD calls here rather than...
10532 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10534 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10535 Alan Hayward <alan.hayward@arm.com>
10536 David Sherwood <david.sherwood@arm.com>
10538 * expmed.c (extract_bit_field_1): For vector extracts,
10539 fall back to extract_bit_field_as_subreg if vec_extract
10540 isn't available.
10542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10543 Alan Hayward <alan.hayward@arm.com>
10544 David Sherwood <david.sherwood@arm.com>
10546 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10547 they are variable or constant sized.
10548 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10549 slots for constant-sized data.
10551 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10552 Alan Hayward <alan.hayward@arm.com>
10553 David Sherwood <david.sherwood@arm.com>
10555 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10556 handling COND_EXPRs with boolean comparisons, try to find a better
10557 basis for the mask type than the boolean itself.
10559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10561 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10562 is calculated and how it can be overridden.
10563 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10564 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10565 if defined.
10566 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10567 if nonzero.
10569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10570 Alan Hayward <alan.hayward@arm.com>
10571 David Sherwood <david.sherwood@arm.com>
10573 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10574 Remove the mode argument.
10575 (aarch64_simd_valid_immediate): Remove the mode and inverse
10576 arguments.
10577 * config/aarch64/iterators.md (bitsize): New iterator.
10578 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10579 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10580 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10581 aarch64_simd_valid_immediate.
10582 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10583 (aarch64_reg_or_bic_imm): Likewise.
10584 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10585 with an insn_type enum and msl with a modifier_type enum.
10586 Replace element_width with a scalar_mode. Change the shift
10587 to unsigned int. Add constructors for scalar_float_mode and
10588 scalar_int_mode elements.
10589 (aarch64_vect_float_const_representable_p): Delete.
10590 (aarch64_can_const_movi_rtx_p)
10591 (aarch64_simd_scalar_immediate_valid_for_move)
10592 (aarch64_simd_make_constant): Update call to
10593 aarch64_simd_valid_immediate.
10594 (aarch64_advsimd_valid_immediate_hs): New function.
10595 (aarch64_advsimd_valid_immediate): Likewise.
10596 (aarch64_simd_valid_immediate): Remove mode and inverse
10597 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10598 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10599 and aarch64_float_const_representable_p on the result.
10600 (aarch64_output_simd_mov_immediate): Remove mode argument.
10601 Update call to aarch64_simd_valid_immediate and use of
10602 simd_immediate_info.
10603 (aarch64_output_scalar_simd_mov_immediate): Update call
10604 accordingly.
10606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10607 Alan Hayward <alan.hayward@arm.com>
10608 David Sherwood <david.sherwood@arm.com>
10610 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10611 (mode_nunits): Likewise CONST_MODE_NUNITS.
10612 * machmode.def (ADJUST_NUNITS): Document.
10613 * genmodes.c (mode_data::need_nunits_adj): New field.
10614 (blank_mode): Update accordingly.
10615 (adj_nunits): New variable.
10616 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10617 parameter.
10618 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10619 listed in adj_nunits.
10620 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10621 listed in adj_nunits. Don't emit case statements for such modes.
10622 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10623 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10624 nothing if adj_nunits is nonnull.
10625 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10626 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10627 (emit_mode_fbit): Update use of print_maybe_const_decl.
10628 (emit_move_size): Likewise. Treat the array as non-const
10629 if adj_nunits.
10630 (emit_mode_adjustments): Handle adj_nunits.
10632 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10634 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10635 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10636 (VECTOR_MODES): Use it.
10637 (make_vector_modes): Take the prefix as an argument.
10639 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10640 Alan Hayward <alan.hayward@arm.com>
10641 David Sherwood <david.sherwood@arm.com>
10643 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10644 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10645 for MODE_VECTOR_BOOL.
10646 * machmode.def (VECTOR_BOOL_MODE): Document.
10647 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10648 (make_vector_bool_mode): New function.
10649 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10650 MODE_VECTOR_BOOL.
10651 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10652 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10653 Likewise.
10654 * stor-layout.c (int_mode_for_mode): Likewise.
10655 * tree.c (build_vector_type_for_mode): Likewise.
10656 * varasm.c (output_constant_pool_2): Likewise.
10657 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10658 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10659 for MODE_VECTOR_BOOL.
10660 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10661 of mode class checks.
10662 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10663 instead of a list of mode class checks.
10664 (expand_vector_scalar_condition): Likewise.
10665 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10668 Alan Hayward <alan.hayward@arm.com>
10669 David Sherwood <david.sherwood@arm.com>
10671 * machmode.h (mode_size): Change from unsigned short to
10672 poly_uint16_pod.
10673 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10674 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10675 or if measurement_type is not polynomial.
10676 (fixed_size_mode::includes_p): Check for constant-sized modes.
10677 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10678 return a poly_uint16 rather than an unsigned short.
10679 (emit_mode_size): Change the type of mode_size from unsigned short
10680 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10681 (emit_mode_adjustments): Cope with polynomial vector sizes.
10682 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10683 for GET_MODE_SIZE.
10684 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10685 for GET_MODE_SIZE.
10686 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10687 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10688 * caller-save.c (setup_save_areas): Likewise.
10689 (replace_reg_with_saved_mem): Likewise.
10690 * calls.c (emit_library_call_value_1): Likewise.
10691 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10692 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10693 (gen_lowpart_for_combine): Likewise.
10694 * convert.c (convert_to_integer_1): Likewise.
10695 * cse.c (equiv_constant, cse_insn): Likewise.
10696 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10697 (cselib_subst_to_values): Likewise.
10698 * dce.c (word_dce_process_block): Likewise.
10699 * df-problems.c (df_word_lr_mark_ref): Likewise.
10700 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10701 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10702 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10703 (rtl_for_decl_location): Likewise.
10704 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10705 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10706 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10707 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10708 (expand_expr_real_1): Likewise.
10709 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10710 (pad_below): Likewise.
10711 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10712 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10713 * ira.c (get_subreg_tracking_sizes): Likewise.
10714 * ira-build.c (ira_create_allocno_objects): Likewise.
10715 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10716 (ira_sort_regnos_for_alter_reg): Likewise.
10717 * ira-costs.c (record_operand_costs): Likewise.
10718 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10719 (resolve_simple_move): Likewise.
10720 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10721 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10722 (lra_constraints): Likewise.
10723 (CONST_POOL_OK_P): Reject variable-sized modes.
10724 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10725 (add_pseudo_to_slot, lra_spill): Likewise.
10726 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10727 * optabs-query.c (get_best_extraction_insn): Likewise.
10728 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10729 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10730 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10731 * recog.c (offsettable_address_addr_space_p): Likewise.
10732 * regcprop.c (maybe_mode_change): Likewise.
10733 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10734 * regrename.c (build_def_use): Likewise.
10735 * regstat.c (dump_reg_info): Likewise.
10736 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10737 (find_reloads, find_reloads_subreg_address): Likewise.
10738 * reload1.c (eliminate_regs_1): Likewise.
10739 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10740 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10741 (simplify_binary_operation_1, simplify_subreg): Likewise.
10742 * targhooks.c (default_function_arg_padding): Likewise.
10743 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10744 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10745 (verify_gimple_assign_ternary): Likewise.
10746 * tree-inline.c (estimate_move_cost): Likewise.
10747 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10748 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10749 (get_address_cost_ainc): Likewise.
10750 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10751 (vect_supportable_dr_alignment): Likewise.
10752 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10753 (vectorizable_reduction): Likewise.
10754 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10755 (vectorizable_operation, vectorizable_load): Likewise.
10756 * tree.c (build_same_sized_truth_vector_type): Likewise.
10757 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10758 * var-tracking.c (emit_note_insn_var_location): Likewise.
10759 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10760 (ADDR_VEC_ALIGN): Likewise.
10762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10763 Alan Hayward <alan.hayward@arm.com>
10764 David Sherwood <david.sherwood@arm.com>
10766 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10767 unsigned short.
10768 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10769 or if measurement_type is polynomial.
10770 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10771 * combine.c (make_extraction): Likewise.
10772 * dse.c (find_shift_sequence): Likewise.
10773 * dwarf2out.c (mem_loc_descriptor): Likewise.
10774 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10775 (extract_bit_field, extract_low_bits): Likewise.
10776 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10777 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10778 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10779 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10780 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10781 * reload.c (find_reloads): Likewise.
10782 * reload1.c (alter_reg): Likewise.
10783 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10784 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10785 * tree-if-conv.c (predicate_mem_writes): Likewise.
10786 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10787 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10788 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10789 * valtrack.c (dead_debug_insert_temp): Likewise.
10790 * varasm.c (mergeable_constant_section): Likewise.
10791 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10794 Alan Hayward <alan.hayward@arm.com>
10795 David Sherwood <david.sherwood@arm.com>
10797 * expr.c (expand_assignment): Cope with polynomial mode sizes
10798 when assigning to a CONCAT.
10800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10801 Alan Hayward <alan.hayward@arm.com>
10802 David Sherwood <david.sherwood@arm.com>
10804 * machmode.h (mode_precision): Change from unsigned short to
10805 poly_uint16_pod.
10806 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10807 short.
10808 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10809 or if measurement_type is not polynomial.
10810 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10811 in which the mode is already known to be a scalar_int_mode.
10812 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10813 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10814 initializer.
10815 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10816 for GET_MODE_PRECISION.
10817 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10818 for GET_MODE_PRECISION.
10819 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10820 as polynomial.
10821 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10822 (expand_field_assignment, make_extraction): Likewise.
10823 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10824 (get_last_value): Likewise.
10825 * convert.c (convert_to_integer_1): Likewise.
10826 * cse.c (cse_insn): Likewise.
10827 * expr.c (expand_expr_real_1): Likewise.
10828 * lra-constraints.c (simplify_operand_subreg): Likewise.
10829 * optabs-query.c (can_atomic_load_p): Likewise.
10830 * optabs.c (expand_atomic_load): Likewise.
10831 (expand_atomic_store): Likewise.
10832 * ree.c (combine_reaching_defs): Likewise.
10833 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10834 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10835 * tree.h (type_has_mode_precision_p): Likewise.
10836 * ubsan.c (instrument_si_overflow): Likewise.
10838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10839 Alan Hayward <alan.hayward@arm.com>
10840 David Sherwood <david.sherwood@arm.com>
10842 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10843 polynomial numbers of units.
10844 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10845 (valid_vector_subparts_p): New function.
10846 (build_vector_type): Remove temporary shim and take the number
10847 of units as a poly_uint64 rather than an int.
10848 (build_opaque_vector_type): Take the number of units as a
10849 poly_uint64 rather than an int.
10850 * tree.c (build_vector_from_ctor): Handle polynomial
10851 TYPE_VECTOR_SUBPARTS.
10852 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10853 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10854 (build_vector_from_val): If the number of units is variable,
10855 use build_vec_duplicate_cst for constant operands and
10856 VEC_DUPLICATE_EXPR otherwise.
10857 (make_vector_type): Remove temporary is_constant ().
10858 (build_vector_type, build_opaque_vector_type): Take the number of
10859 units as a poly_uint64 rather than an int.
10860 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10861 VECTOR_CST_NELTS.
10862 * cfgexpand.c (expand_debug_expr): Likewise.
10863 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10864 (store_constructor, expand_expr_real_1): Likewise.
10865 (const_scalar_mask_from_tree): Likewise.
10866 * fold-const-call.c (fold_const_reduction): Likewise.
10867 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10868 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10869 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10870 (fold_relational_const): Likewise.
10871 (native_interpret_vector): Likewise. Change the size from an
10872 int to an unsigned int.
10873 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10874 TYPE_VECTOR_SUBPARTS.
10875 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10876 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10877 duplicating a non-constant operand into a variable-length vector.
10878 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10879 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10880 * ipa-icf.c (sem_variable::equals): Likewise.
10881 * match.pd: Likewise.
10882 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10883 * print-tree.c (print_node): Likewise.
10884 * stor-layout.c (layout_type): Likewise.
10885 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10886 * tree-cfg.c (verify_gimple_comparison): Likewise.
10887 (verify_gimple_assign_binary): Likewise.
10888 (verify_gimple_assign_ternary): Likewise.
10889 (verify_gimple_assign_single): Likewise.
10890 * tree-pretty-print.c (dump_generic_node): Likewise.
10891 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10892 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10893 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10894 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10895 (vect_shift_permute_load_chain): Likewise.
10896 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10897 (expand_vector_condition, optimize_vector_constructor): Likewise.
10898 (lower_vec_perm, get_compute_type): Likewise.
10899 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10900 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10901 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10902 (vect_recog_mask_conversion_pattern): Likewise.
10903 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10904 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10905 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10906 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10907 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10908 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10909 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10910 (supportable_widening_operation): Likewise.
10911 (supportable_narrowing_operation): Likewise.
10912 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10913 Likewise.
10914 * varasm.c (output_constant): Likewise.
10916 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10917 Alan Hayward <alan.hayward@arm.com>
10918 David Sherwood <david.sherwood@arm.com>
10920 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10921 so that both the length == 3 and length != 3 cases set up their
10922 own permute vectors. Add comments explaining why we know the
10923 number of elements is constant.
10924 (vect_permute_load_chain): Likewise.
10926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10927 Alan Hayward <alan.hayward@arm.com>
10928 David Sherwood <david.sherwood@arm.com>
10930 * machmode.h (mode_nunits): Change from unsigned char to
10931 poly_uint16_pod.
10932 (ONLY_FIXED_SIZE_MODES): New macro.
10933 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10934 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10935 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10936 New typedefs.
10937 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10938 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10939 or if measurement_type is not polynomial.
10940 * genmodes.c (ZERO_COEFFS): New macro.
10941 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10942 poly_uint16.
10943 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10944 Use ZERO_COEFFS when emitting initializers.
10945 * data-streamer.h (bp_pack_poly_value): New function.
10946 (bp_unpack_poly_value): Likewise.
10947 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10948 for GET_MODE_NUNITS.
10949 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10950 for GET_MODE_NUNITS.
10951 * tree.c (make_vector_type): Remove temporary shim and make
10952 the real function take the number of units as a poly_uint64
10953 rather than an int.
10954 (build_vector_type_for_mode): Handle polynomial nunits.
10955 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10956 * emit-rtl.c (const_vec_series_p_1): Likewise.
10957 (gen_rtx_CONST_VECTOR): Likewise.
10958 * fold-const.c (test_vec_duplicate_folding): Likewise.
10959 * genrecog.c (validate_pattern): Likewise.
10960 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10961 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10962 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10963 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10964 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10965 * rtlanal.c (subreg_get_info): Likewise.
10966 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10967 (vect_grouped_load_supported): Likewise.
10968 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10969 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10970 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10971 (simplify_const_unary_operation, simplify_binary_operation_1)
10972 (simplify_const_binary_operation, simplify_ternary_operation)
10973 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10974 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10975 instead of CONST_VECTOR_NUNITS.
10976 * varasm.c (output_constant_pool_2): Likewise.
10977 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10978 explicit-encoded elements in the XVEC for variable-length vectors.
10980 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10982 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10985 Alan Hayward <alan.hayward@arm.com>
10986 David Sherwood <david.sherwood@arm.com>
10988 * coretypes.h (fixed_size_mode): Declare.
10989 (fixed_size_mode_pod): New typedef.
10990 * builtins.h (target_builtins::x_apply_args_mode)
10991 (target_builtins::x_apply_result_mode): Change type to
10992 fixed_size_mode_pod.
10993 * builtins.c (apply_args_size, apply_result_size, result_vector)
10994 (expand_builtin_apply_args_1, expand_builtin_apply)
10995 (expand_builtin_return): Update accordingly.
10997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10999 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11000 * cselib.c (cselib_hash_rtx): Likewise.
11001 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11002 CONST_VECTOR encoding.
11004 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11005 Jeff Law <law@redhat.com>
11007 PR target/83641
11008 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11009 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11010 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11011 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11013 PR target/83641
11014 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11015 explicitly probe *sp in a noreturn function if there were any callee
11016 register saves or frame pointer is needed.
11018 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11020 PR debug/83621
11021 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11022 BLKmode for ternary, binary or unary expressions.
11024 PR debug/83645
11025 * var-tracking.c (delete_vta_debug_insn): New inline function.
11026 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11027 insns from get_insns () to NULL instead of each bb separately.
11028 Use delete_vta_debug_insn. No longer static.
11029 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11030 delete_vta_debug_insns callers.
11031 * rtl.h (delete_vta_debug_insns): Declare.
11032 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11033 instead of variable_tracking_main.
11035 2018-01-03 Martin Sebor <msebor@redhat.com>
11037 PR tree-optimization/83603
11038 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11039 arguments past the endof the argument list in functions declared
11040 without a prototype.
11041 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11042 Avoid checking when arguments are null.
11044 2018-01-03 Martin Sebor <msebor@redhat.com>
11046 PR c/83559
11047 * doc/extend.texi (attribute const): Fix a typo.
11048 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11049 issuing -Wsuggest-attribute for void functions.
11051 2018-01-03 Martin Sebor <msebor@redhat.com>
11053 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11054 offset_int::from instead of wide_int::to_shwi.
11055 (maybe_diag_overlap): Remove assertion.
11056 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11057 * gimple-ssa-sprintf.c (format_directive): Same.
11058 (parse_directive): Same.
11059 (sprintf_dom_walker::compute_format_length): Same.
11060 (try_substitute_return_value): Same.
11062 2018-01-03 Jeff Law <law@redhat.com>
11064 PR middle-end/83654
11065 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11066 non-constant residual for zero at runtime and avoid probing in
11067 that case. Reorganize code for trailing problem to mirror handling
11068 of the residual.
11070 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11072 PR tree-optimization/83501
11073 * tree-ssa-strlen.c (get_string_cst): New.
11074 (handle_char_store): Call get_string_cst.
11076 2018-01-03 Martin Liska <mliska@suse.cz>
11078 PR tree-optimization/83593
11079 * tree-ssa-strlen.c: Include tree-cfg.h.
11080 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11081 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11082 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11083 to false.
11084 (strlen_dom_walker::before_dom_children): Call
11085 gimple_purge_dead_eh_edges. Dump tranformation with details
11086 dump flags.
11087 (strlen_dom_walker::before_dom_children): Update call by adding
11088 new argument cleanup_eh.
11089 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11091 2018-01-03 Martin Liska <mliska@suse.cz>
11093 PR ipa/83549
11094 * cif-code.def (VARIADIC_THUNK): New enum value.
11095 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11096 thunks.
11098 2018-01-03 Jan Beulich <jbeulich@suse.com>
11100 * sse.md (mov<mode>_internal): Tighten condition for when to use
11101 vmovdqu<ssescalarsize> for TI and OI modes.
11103 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11105 Update copyright years.
11107 2018-01-03 Martin Liska <mliska@suse.cz>
11109 PR ipa/83594
11110 * ipa-visibility.c (function_and_variable_visibility): Skip
11111 functions with noipa attribure.
11113 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11115 * gcc.c (process_command): Update copyright notice dates.
11116 * gcov-dump.c (print_version): Ditto.
11117 * gcov.c (print_version): Ditto.
11118 * gcov-tool.c (print_version): Ditto.
11119 * gengtype.c (create_file): Ditto.
11120 * doc/cpp.texi: Bump @copying's copyright year.
11121 * doc/cppinternals.texi: Ditto.
11122 * doc/gcc.texi: Ditto.
11123 * doc/gccint.texi: Ditto.
11124 * doc/gcov.texi: Ditto.
11125 * doc/install.texi: Ditto.
11126 * doc/invoke.texi: Ditto.
11128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11130 * vector-builder.h (vector_builder::m_full_nelts): Change from
11131 unsigned int to poly_uint64.
11132 (vector_builder::full_nelts): Update prototype accordingly.
11133 (vector_builder::new_vector): Likewise.
11134 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11135 (vector_builder::operator ==): Likewise.
11136 (vector_builder::finalize): Likewise.
11137 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11138 Take the number of elements as a poly_uint64 rather than an
11139 unsigned int.
11140 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11141 from unsigned int to poly_uint64.
11142 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11143 (vec_perm_indices::new_vector): Likewise.
11144 (vec_perm_indices::length): Likewise.
11145 (vec_perm_indices::nelts_per_input): Likewise.
11146 (vec_perm_indices::input_nelts): Likewise.
11147 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11148 number of elements per input as a poly_uint64 rather than an
11149 unsigned int. Use the original encoding for variable-length
11150 vectors, rather than clamping each individual element.
11151 For the second and subsequent elements in each pattern,
11152 clamp the step and base before clamping their sum.
11153 (vec_perm_indices::series_p): Handle polynomial element counts.
11154 (vec_perm_indices::all_in_range_p): Likewise.
11155 (vec_perm_indices_to_tree): Likewise.
11156 (vec_perm_indices_to_rtx): Likewise.
11157 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11158 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11159 (tree_vector_builder::new_binary_operation): Handle polynomial
11160 element counts. Return false if we need to know the number
11161 of elements at compile time.
11162 * fold-const.c (fold_vec_perm): Punt if the number of elements
11163 isn't known at compile time.
11165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11167 * vec-perm-indices.h (vec_perm_builder): Change element type
11168 from HOST_WIDE_INT to poly_int64.
11169 (vec_perm_indices::element_type): Update accordingly.
11170 (vec_perm_indices::clamp): Handle polynomial element_types.
11171 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11172 (vec_perm_indices::all_in_range_p): Likewise.
11173 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11174 than shwi trees.
11175 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11176 polynomial vec_perm_indices element types.
11177 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11178 * fold-const.c (fold_vec_perm): Likewise.
11179 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11180 * tree-vect-generic.c (lower_vec_perm): Likewise.
11181 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11182 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11183 element type to HOST_WIDE_INT.
11185 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11186 Alan Hayward <alan.hayward@arm.com>
11187 David Sherwood <david.sherwood@arm.com>
11189 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11190 rather than an int. Use plus_constant.
11191 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11192 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11195 Alan Hayward <alan.hayward@arm.com>
11196 David Sherwood <david.sherwood@arm.com>
11198 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11199 a HOST_WIDE_INT to a poly_int64.
11201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11202 Alan Hayward <alan.hayward@arm.com>
11203 David Sherwood <david.sherwood@arm.com>
11205 * calls.c (load_register_parameters): Cope with polynomial
11206 mode sizes. Require a constant size for BLKmode parameters
11207 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11208 forces a parameter to be padded at the lsb end in order to
11209 fill a complete number of words, require the parameter size
11210 to be ordered wrt UNITS_PER_WORD.
11212 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11213 Alan Hayward <alan.hayward@arm.com>
11214 David Sherwood <david.sherwood@arm.com>
11216 * reload1.c (spill_stack_slot_width): Change element type
11217 from unsigned int to poly_uint64_pod.
11218 (alter_reg): Treat mode sizes as polynomial.
11220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11221 Alan Hayward <alan.hayward@arm.com>
11222 David Sherwood <david.sherwood@arm.com>
11224 * reload.c (complex_word_subreg_p): New function.
11225 (reload_inner_reg_of_subreg, push_reload): Use it.
11227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11228 Alan Hayward <alan.hayward@arm.com>
11229 David Sherwood <david.sherwood@arm.com>
11231 * lra-constraints.c (process_alt_operands): Reject matched
11232 operands whose sizes aren't ordered.
11233 (match_reload): Refer to this check here.
11235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11236 Alan Hayward <alan.hayward@arm.com>
11237 David Sherwood <david.sherwood@arm.com>
11239 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11240 that the mode size is in the set {1, 2, 4, 8, 16}.
11242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11243 Alan Hayward <alan.hayward@arm.com>
11244 David Sherwood <david.sherwood@arm.com>
11246 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11247 Use plus_constant instead of gen_rtx_PLUS.
11249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11250 Alan Hayward <alan.hayward@arm.com>
11251 David Sherwood <david.sherwood@arm.com>
11253 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11254 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11255 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11256 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11257 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11258 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11259 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11260 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11261 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11262 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11263 a poly_int64.
11264 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11265 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11266 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11267 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11268 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11269 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11270 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11271 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11272 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11273 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11274 function.
11275 * expr.c (emit_move_resolve_push): Treat the input and result
11276 of PUSH_ROUNDING as a poly_int64.
11277 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11278 (emit_push_insn): Likewise.
11279 * lra-eliminations.c (mark_not_eliminable): Likewise.
11280 * recog.c (push_operand): Likewise.
11281 * reload1.c (elimination_effects): Likewise.
11282 * rtlanal.c (nonzero_bits1): Likewise.
11283 * calls.c (store_one_arg): Likewise. Require the padding to be
11284 known at compile time.
11286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11287 Alan Hayward <alan.hayward@arm.com>
11288 David Sherwood <david.sherwood@arm.com>
11290 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11291 Use plus_constant instead of gen_rtx_PLUS.
11293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11294 Alan Hayward <alan.hayward@arm.com>
11295 David Sherwood <david.sherwood@arm.com>
11297 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11298 rather than an int.
11300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11301 Alan Hayward <alan.hayward@arm.com>
11302 David Sherwood <david.sherwood@arm.com>
11304 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11305 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11306 via stack temporaries. Treat the mode size as polynomial too.
11308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11309 Alan Hayward <alan.hayward@arm.com>
11310 David Sherwood <david.sherwood@arm.com>
11312 * expr.c (expand_expr_real_2): When handling conversions involving
11313 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11314 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11315 as a poly_uint64 too.
11317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11318 Alan Hayward <alan.hayward@arm.com>
11319 David Sherwood <david.sherwood@arm.com>
11321 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
11323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11324 Alan Hayward <alan.hayward@arm.com>
11325 David Sherwood <david.sherwood@arm.com>
11327 * combine.c (can_change_dest_mode): Handle polynomial
11328 REGMODE_NATURAL_SIZE.
11329 * expmed.c (store_bit_field_1): Likewise.
11330 * expr.c (store_constructor): Likewise.
11331 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11332 and polynomial REGMODE_NATURAL_SIZE.
11333 (gen_lowpart_common): Likewise.
11334 * reginfo.c (record_subregs_of_mode): Likewise.
11335 * rtlanal.c (read_modify_subreg_p): Likewise.
11337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11338 Alan Hayward <alan.hayward@arm.com>
11339 David Sherwood <david.sherwood@arm.com>
11341 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11342 numbers of elements.
11344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11345 Alan Hayward <alan.hayward@arm.com>
11346 David Sherwood <david.sherwood@arm.com>
11348 * match.pd: Cope with polynomial numbers of vector elements.
11350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11351 Alan Hayward <alan.hayward@arm.com>
11352 David Sherwood <david.sherwood@arm.com>
11354 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11355 in a POINTER_PLUS_EXPR.
11357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11358 Alan Hayward <alan.hayward@arm.com>
11359 David Sherwood <david.sherwood@arm.com>
11361 * omp-simd-clone.c (simd_clone_subparts): New function.
11362 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11363 (ipa_simd_modify_function_body): Likewise.
11365 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11366 Alan Hayward <alan.hayward@arm.com>
11367 David Sherwood <david.sherwood@arm.com>
11369 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11370 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11371 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11372 (expand_vector_condition, vector_element): Likewise.
11373 (subparts_gt): New function.
11374 (get_compute_type): Use subparts_gt.
11375 (count_type_subparts): Delete.
11376 (expand_vector_operations_1): Use subparts_gt instead of
11377 count_type_subparts.
11379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11380 Alan Hayward <alan.hayward@arm.com>
11381 David Sherwood <david.sherwood@arm.com>
11383 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11384 (vect_compile_time_alias): ...this new function. Do the calculation
11385 on poly_ints rather than trees.
11386 (vect_prune_runtime_alias_test_list): Update call accordingly.
11388 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11389 Alan Hayward <alan.hayward@arm.com>
11390 David Sherwood <david.sherwood@arm.com>
11392 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11393 numbers of units.
11394 (vect_schedule_slp_instance): Likewise.
11396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11397 Alan Hayward <alan.hayward@arm.com>
11398 David Sherwood <david.sherwood@arm.com>
11400 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11401 constant and extern definitions for variable-length vectors.
11402 (vect_get_constant_vectors): Note that the number of units
11403 is known to be constant.
11405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11406 Alan Hayward <alan.hayward@arm.com>
11407 David Sherwood <david.sherwood@arm.com>
11409 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11410 of units as polynomial. Choose between WIDE and NARROW based
11411 on multiple_p.
11413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11414 Alan Hayward <alan.hayward@arm.com>
11415 David Sherwood <david.sherwood@arm.com>
11417 * tree-vect-stmts.c (simd_clone_subparts): New function.
11418 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11421 Alan Hayward <alan.hayward@arm.com>
11422 David Sherwood <david.sherwood@arm.com>
11424 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11425 vectors as polynomial. Use build_index_vector for
11426 IFN_GOMP_SIMD_LANE.
11428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11429 Alan Hayward <alan.hayward@arm.com>
11430 David Sherwood <david.sherwood@arm.com>
11432 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11433 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11434 for variable-length vectors.
11435 (vectorizable_mask_load_store): Treat the number of units as
11436 polynomial, asserting that it is constant if the condition has
11437 already been enforced.
11438 (vectorizable_store, vectorizable_load): Likewise.
11440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11441 Alan Hayward <alan.hayward@arm.com>
11442 David Sherwood <david.sherwood@arm.com>
11444 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11445 of units as polynomial. Punt if we can't tell at compile time
11446 which vector contains the final result.
11448 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11449 Alan Hayward <alan.hayward@arm.com>
11450 David Sherwood <david.sherwood@arm.com>
11452 * tree-vect-loop.c (vectorizable_induction): Treat the number
11453 of units as polynomial. Punt on SLP inductions. Use an integer
11454 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11455 cast of such a series for variable-length floating-point
11456 reductions.
11458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11459 Alan Hayward <alan.hayward@arm.com>
11460 David Sherwood <david.sherwood@arm.com>
11462 * tree.h (build_index_vector): Declare.
11463 * tree.c (build_index_vector): New function.
11464 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11465 of units as polynomial, forcibly converting it to a constant if
11466 vectorizable_reduction has already enforced the condition.
11467 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11468 to create a {1,2,3,...} vector.
11469 (vectorizable_reduction): Treat the number of units as polynomial.
11470 Choose vectype_in based on the largest scalar element size rather
11471 than the smallest number of units. Enforce the restrictions
11472 relied on above.
11474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11475 Alan Hayward <alan.hayward@arm.com>
11476 David Sherwood <david.sherwood@arm.com>
11478 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11479 number of units as polynomial.
11481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11482 Alan Hayward <alan.hayward@arm.com>
11483 David Sherwood <david.sherwood@arm.com>
11485 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11486 * target.def (autovectorize_vector_sizes): Return the vector sizes
11487 by pointer, using vector_sizes rather than a bitmask.
11488 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11489 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11490 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11491 Likewise.
11492 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11493 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11494 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11495 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11496 * omp-general.c (omp_max_vf): Likewise.
11497 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11498 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11499 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11500 * tree-vect-slp.c (vect_slp_bb): Likewise.
11501 * doc/tm.texi: Regenerate.
11502 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11503 to a poly_uint64.
11504 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11505 the vector size as a poly_uint64 rather than an unsigned int.
11506 (current_vector_size): Change from an unsigned int to a poly_uint64.
11507 (get_vectype_for_scalar_type): Update accordingly.
11508 * tree.h (build_truth_vector_type): Take the size and number of
11509 units as a poly_uint64 rather than an unsigned int.
11510 (build_vector_type): Add a temporary overload that takes
11511 the number of units as a poly_uint64 rather than an unsigned int.
11512 * tree.c (make_vector_type): Likewise.
11513 (build_truth_vector_type): Take the number of units as a poly_uint64
11514 rather than an unsigned int.
11516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11517 Alan Hayward <alan.hayward@arm.com>
11518 David Sherwood <david.sherwood@arm.com>
11520 * target.def (get_mask_mode): Take the number of units and length
11521 as poly_uint64s rather than unsigned ints.
11522 * targhooks.h (default_get_mask_mode): Update accordingly.
11523 * targhooks.c (default_get_mask_mode): Likewise.
11524 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11525 * doc/tm.texi: Regenerate.
11527 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11528 Alan Hayward <alan.hayward@arm.com>
11529 David Sherwood <david.sherwood@arm.com>
11531 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11532 * omp-general.c (omp_max_vf): Likewise.
11533 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11534 (expand_omp_simd): Handle polynomial safelen.
11535 * omp-low.c (omplow_simd_context): Add a default constructor.
11536 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11537 (lower_rec_simd_input_clauses): Update accordingly.
11538 (lower_rec_input_clauses): Likewise.
11540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11541 Alan Hayward <alan.hayward@arm.com>
11542 David Sherwood <david.sherwood@arm.com>
11544 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11545 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11546 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11547 (vect_analyze_slp_cost): Likewise.
11548 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11549 (vect_model_load_cost): Likewise.
11551 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11552 Alan Hayward <alan.hayward@arm.com>
11553 David Sherwood <david.sherwood@arm.com>
11555 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11556 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11557 from an unsigned int * to a poly_uint64_pod *.
11558 (calculate_unrolling_factor): New function.
11559 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11562 Alan Hayward <alan.hayward@arm.com>
11563 David Sherwood <david.sherwood@arm.com>
11565 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11566 from an unsigned int to a poly_uint64.
11567 (_loop_vec_info::slp_unrolling_factor): Likewise.
11568 (_loop_vec_info::vectorization_factor): Change from an int
11569 to a poly_uint64.
11570 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11571 (vect_get_num_vectors): New function.
11572 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11573 (vect_get_num_copies): Use vect_get_num_vectors.
11574 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11575 to an unsigned int *.
11576 (vect_analyze_data_refs): Change min_vf from an int * to a
11577 poly_uint64 *.
11578 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11579 than an unsigned HOST_WIDE_INT.
11580 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11581 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11582 to an unsigned int *.
11583 (vect_analyze_data_ref_dependences): Likewise.
11584 (vect_compute_data_ref_alignment): Handle polynomial vf.
11585 (vect_enhance_data_refs_alignment): Likewise.
11586 (vect_prune_runtime_alias_test_list): Likewise.
11587 (vect_shift_permute_load_chain): Likewise.
11588 (vect_supportable_dr_alignment): Likewise.
11589 (dependence_distance_ge_vf): Take the vectorization factor as a
11590 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11591 (vect_analyze_data_refs): Change min_vf from an int * to a
11592 poly_uint64 *.
11593 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11594 vfm1 as a poly_uint64 rather than an int. Make the same change
11595 for the returned bound_scalar.
11596 (vect_gen_vector_loop_niters): Handle polynomial vf.
11597 (vect_do_peeling): Likewise. Update call to
11598 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11599 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11600 be constant.
11601 * tree-vect-loop.c (vect_determine_vectorization_factor)
11602 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11603 (vect_get_known_peeling_cost): Likewise.
11604 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11605 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11606 (vect_transform_loop): Likewise. Use the lowest possible VF when
11607 updating the upper bounds of the loop.
11608 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11609 rather than an int.
11610 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11611 polynomial unroll factors.
11612 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11613 (vect_make_slp_decision): Likewise.
11614 (vect_supported_load_permutation_p): Likewise, and polynomial
11615 vf too.
11616 (vect_analyze_slp_cost): Handle polynomial vf.
11617 (vect_slp_analyze_node_operations): Likewise.
11618 (vect_slp_analyze_bb_1): Likewise.
11619 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11620 than an unsigned HOST_WIDE_INT.
11621 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11622 (vectorizable_load): Handle polynomial vf.
11623 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11624 a poly_uint64.
11625 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11628 Alan Hayward <alan.hayward@arm.com>
11629 David Sherwood <david.sherwood@arm.com>
11631 * match.pd: Handle bit operations involving three constants
11632 and try to fold one pair.
11634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11636 * tree-vect-loop-manip.c: Include gimple-fold.h.
11637 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11638 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11639 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11640 Add a path that uses a step of VF instead of 1, but disable it
11641 for now.
11642 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11643 and niters_no_overflow parameters. Update calls to
11644 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11645 Create a new SSA name if the latter choses to use a ste other
11646 than zero, and return it via niters_vector_mult_vf_var.
11647 * tree-vect-loop.c (vect_transform_loop): Update calls to
11648 vect_do_peeling, vect_gen_vector_loop_niters and
11649 slpeel_make_loop_iterate_ntimes.
11650 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11651 (vect_gen_vector_loop_niters): Update declarations after above changes.
11653 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11655 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11656 128-bit round to integer instructions.
11657 (ceil<mode>2): Likewise.
11658 (btrunc<mode>2): Likewise.
11659 (round<mode>2): Likewise.
11661 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11663 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11664 unaligned VSX load/store on P8/P9.
11665 (expand_block_clear): Allow the use of unaligned VSX
11666 load/store on P8/P9.
11668 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11670 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11671 New function.
11672 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11673 swap associated with both a load and a store.
11675 2018-01-02 Andrew Waterman <andrew@sifive.com>
11677 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11678 * config/riscv/riscv.md (clear_cache): Use it.
11680 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11682 * web.c: Remove out-of-date comment.
11684 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11686 * expr.c (fixup_args_size_notes): Check that any existing
11687 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11688 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11689 (emit_single_push_insn): ...here.
11691 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11693 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11694 (const_vector_encoded_nelts): New function.
11695 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11696 (const_vector_int_elt, const_vector_elt): Declare.
11697 * emit-rtl.c (const_vector_int_elt_1): New function.
11698 (const_vector_elt): Likewise.
11699 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11700 of CONST_VECTOR_ELT.
11702 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11704 * expr.c: Include rtx-vector-builder.h.
11705 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11706 directly on the tree encoding.
11707 (const_vector_from_tree): Likewise.
11708 * optabs.c: Include rtx-vector-builder.h.
11709 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11710 sequence of "u" values.
11711 * vec-perm-indices.c: Include rtx-vector-builder.h.
11712 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11713 directly on the vec_perm_indices encoding.
11715 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11717 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11718 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11719 * rtx-vector-builder.h: New file.
11720 * rtx-vector-builder.c: Likewise.
11721 * rtl.h (rtx_def::u2): Add a const_vector field.
11722 (CONST_VECTOR_NPATTERNS): New macro.
11723 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11724 (CONST_VECTOR_DUPLICATE_P): Likewise.
11725 (CONST_VECTOR_STEPPED_P): Likewise.
11726 (CONST_VECTOR_ENCODED_ELT): Likewise.
11727 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11728 (unwrap_const_vec_duplicate): Likewise.
11729 (const_vec_series_p): Check for a non-duplicated vector encoding.
11730 Say that the function only returns true for integer vectors.
11731 * emit-rtl.c: Include rtx-vector-builder.h.
11732 (gen_const_vec_duplicate_1): Delete.
11733 (gen_const_vector): Call gen_const_vec_duplicate instead of
11734 gen_const_vec_duplicate_1.
11735 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11736 (gen_const_vec_duplicate): Use rtx_vector_builder.
11737 (gen_const_vec_series): Likewise.
11738 (gen_rtx_CONST_VECTOR): Likewise.
11739 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11740 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11741 Build a new vector rather than modifying a CONST_VECTOR in-place.
11742 (handle_special_swappables): Update call accordingly.
11743 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11744 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11745 Build a new vector rather than modifying a CONST_VECTOR in-place.
11746 (handle_special_swappables): Update call accordingly.
11748 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11750 * simplify-rtx.c (simplify_const_binary_operation): Use
11751 CONST_VECTOR_ELT instead of XVECEXP.
11753 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11755 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11756 the selector elements to be different from the data elements
11757 if the selector is a VECTOR_CST.
11758 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11759 ssizetype for the selector.
11761 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11763 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11764 before testing each element individually.
11765 * tree-vect-generic.c (lower_vec_perm): Likewise.
11767 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11769 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11770 * selftest-run-tests.c (selftest::run_tests): Call it.
11771 * vector-builder.h (vector_builder::operator ==): New function.
11772 (vector_builder::operator !=): Likewise.
11773 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11774 (vec_perm_indices::all_from_input_p): New function.
11775 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11776 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11777 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11778 instead of reading the VECTOR_CST directly. Detect whether both
11779 vector inputs are the same before constructing the vec_perm_indices,
11780 and update the number of inputs argument accordingly. Use the
11781 utility functions added above. Only construct sel2 if we need to.
11783 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11785 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11786 the broadcast of the low byte.
11787 (expand_mult_highpart): Use an explicit encoding for the permutes.
11788 * optabs-query.c (can_mult_highpart_p): Likewise.
11789 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11790 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11791 (vectorizable_bswap): Likewise.
11792 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11793 explicit encoding for the power-of-2 permutes.
11794 (vect_permute_store_chain): Likewise.
11795 (vect_grouped_load_supported): Likewise.
11796 (vect_permute_load_chain): Likewise.
11798 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11800 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11801 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11802 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11803 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11804 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11805 (vect_gen_perm_mask_any): Likewise.
11807 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11809 * int-vector-builder.h: New file.
11810 * vec-perm-indices.h: Include int-vector-builder.h.
11811 (vec_perm_indices): Redefine as an int_vector_builder.
11812 (auto_vec_perm_indices): Delete.
11813 (vec_perm_builder): Redefine as a stand-alone class.
11814 (vec_perm_indices::vec_perm_indices): New function.
11815 (vec_perm_indices::clamp): Likewise.
11816 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11817 (vec_perm_indices::new_vector): New function.
11818 (vec_perm_indices::new_expanded_vector): Update for new
11819 vec_perm_indices class.
11820 (vec_perm_indices::rotate_inputs): New function.
11821 (vec_perm_indices::all_in_range_p): Operate directly on the
11822 encoded form, without computing elided elements.
11823 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11824 encoding. Update for new vec_perm_indices class.
11825 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11826 the given vec_perm_builder.
11827 (expand_vec_perm_var): Update vec_perm_builder constructor.
11828 (expand_mult_highpart): Use vec_perm_builder instead of
11829 auto_vec_perm_indices.
11830 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11831 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11832 or double series encoding as appropriate.
11833 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11834 vec_perm_indices instead of auto_vec_perm_indices.
11835 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11836 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11837 (vect_permute_store_chain): Likewise.
11838 (vect_grouped_load_supported): Likewise.
11839 (vect_permute_load_chain): Likewise.
11840 (vect_shift_permute_load_chain): Likewise.
11841 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11842 (vect_transform_slp_perm_load): Likewise.
11843 (vect_schedule_slp_instance): Likewise.
11844 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11845 (vectorizable_mask_load_store): Likewise.
11846 (vectorizable_bswap): Likewise.
11847 (vectorizable_store): Likewise.
11848 (vectorizable_load): Likewise.
11849 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11850 vec_perm_indices instead of auto_vec_perm_indices. Use
11851 tree_to_vec_perm_builder to read the vector from a tree.
11852 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11853 vec_perm_builder instead of a vec_perm_indices.
11854 (have_whole_vector_shift): Use vec_perm_builder and
11855 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11856 truncation to calc_vec_perm_mask_for_shift.
11857 (vect_create_epilog_for_reduction): Likewise.
11858 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11859 from auto_vec_perm_indices to vec_perm_indices.
11860 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11861 instead of changing individual elements.
11862 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11863 the vector in d.perm.
11864 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11865 from auto_vec_perm_indices to vec_perm_indices.
11866 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11867 instead of changing individual elements.
11868 (arm_vectorize_vec_perm_const): Use new_vector to install
11869 the vector in d.perm.
11870 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11871 Update vec_perm_builder constructor.
11872 (rs6000_expand_interleave): Likewise.
11873 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11874 (rs6000_expand_interleave): Likewise.
11876 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11878 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11879 to qimode could truncate the indices.
11880 * optabs.c (expand_vec_perm_var): Likewise.
11882 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11884 * Makefile.in (OBJS): Add vec-perm-indices.o.
11885 * vec-perm-indices.h: New file.
11886 * vec-perm-indices.c: Likewise.
11887 * target.h (vec_perm_indices): Replace with a forward class
11888 declaration.
11889 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11890 * optabs.h: Include vec-perm-indices.h.
11891 (expand_vec_perm): Delete.
11892 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11893 (expand_vec_perm_const): Declare.
11894 * target.def (vec_perm_const_ok): Replace with...
11895 (vec_perm_const): ...this new hook.
11896 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11897 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11898 * doc/tm.texi: Regenerate.
11899 * optabs.def (vec_perm_const): Delete.
11900 * doc/md.texi (vec_perm_const): Likewise.
11901 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11902 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11903 expand_vec_perm for constant permutation vectors. Assert that
11904 the mode of variable permutation vectors is the integer equivalent
11905 of the mode that is being permuted.
11906 * optabs-query.h (selector_fits_mode_p): Declare.
11907 * optabs-query.c: Include vec-perm-indices.h.
11908 (selector_fits_mode_p): New function.
11909 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11910 is defined, instead of checking whether the vec_perm_const_optab
11911 exists. Use targetm.vectorize.vec_perm_const instead of
11912 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11913 fit in the vector mode before using a variable permute.
11914 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11915 vec_perm_indices instead of an rtx.
11916 (expand_vec_perm): Replace with...
11917 (expand_vec_perm_const): ...this new function. Take the selector
11918 as a vec_perm_indices rather than an rtx. Also take the mode of
11919 the selector. Update call to shift_amt_for_vec_perm_mask.
11920 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11921 Use vec_perm_indices::new_expanded_vector to expand the original
11922 selector into bytes. Check whether the indices fit in the vector
11923 mode before using a variable permute.
11924 (expand_vec_perm_var): Make global.
11925 (expand_mult_highpart): Use expand_vec_perm_const.
11926 * fold-const.c: Includes vec-perm-indices.h.
11927 * tree-ssa-forwprop.c: Likewise.
11928 * tree-vect-data-refs.c: Likewise.
11929 * tree-vect-generic.c: Likewise.
11930 * tree-vect-loop.c: Likewise.
11931 * tree-vect-slp.c: Likewise.
11932 * tree-vect-stmts.c: Likewise.
11933 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11934 Delete.
11935 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11936 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11937 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11938 (aarch64_vectorize_vec_perm_const): ...this new function.
11939 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11940 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11941 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11942 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11943 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11944 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11945 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11946 into...
11947 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11948 check for NEON modes.
11949 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11950 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11951 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11952 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11953 into...
11954 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11955 the old VEC_PERM_CONST conditions.
11956 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11957 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11958 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11959 (ia64_vectorize_vec_perm_const_ok): Merge into...
11960 (ia64_vectorize_vec_perm_const): ...this new function.
11961 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11962 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11963 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11964 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11965 * config/mips/mips.c (mips_expand_vec_perm_const)
11966 (mips_vectorize_vec_perm_const_ok): Merge into...
11967 (mips_vectorize_vec_perm_const): ...this new function.
11968 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11969 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11970 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11971 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11972 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11973 (rs6000_expand_vec_perm_const): Delete.
11974 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11975 Delete.
11976 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11977 (altivec_expand_vec_perm_const_le): Take each operand individually.
11978 Operate on constant selectors rather than rtxes.
11979 (altivec_expand_vec_perm_const): Likewise. Update call to
11980 altivec_expand_vec_perm_const_le.
11981 (rs6000_expand_vec_perm_const): Delete.
11982 (rs6000_vectorize_vec_perm_const_ok): Delete.
11983 (rs6000_vectorize_vec_perm_const): New function.
11984 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11985 an element count and rtx array.
11986 (rs6000_expand_extract_even): Update call accordingly.
11987 (rs6000_expand_interleave): Likewise.
11988 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11989 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11990 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11991 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11992 (rs6000_expand_vec_perm_const): Delete.
11993 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11994 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11995 (altivec_expand_vec_perm_const_le): Take each operand individually.
11996 Operate on constant selectors rather than rtxes.
11997 (altivec_expand_vec_perm_const): Likewise. Update call to
11998 altivec_expand_vec_perm_const_le.
11999 (rs6000_expand_vec_perm_const): Delete.
12000 (rs6000_vectorize_vec_perm_const_ok): Delete.
12001 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12002 reference to the SPE evmerge intructions.
12003 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12004 an element count and rtx array.
12005 (rs6000_expand_extract_even): Update call accordingly.
12006 (rs6000_expand_interleave): Likewise.
12007 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12008 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12009 new function.
12010 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12012 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12014 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12015 vector mode and that that mode matches the mode of the data
12016 being permuted.
12017 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12018 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12019 directly using expand_vec_perm_1 when forcing selectors into
12020 registers.
12021 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12023 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12025 * optabs-query.h (can_vec_perm_p): Delete.
12026 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12027 * optabs-query.c (can_vec_perm_p): Split into...
12028 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12029 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12030 particular selector is valid.
12031 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12032 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12033 (vect_grouped_load_supported): Likewise.
12034 (vect_shift_permute_load_chain): Likewise.
12035 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12036 (vect_transform_slp_perm_load): Likewise.
12037 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12038 (vectorizable_bswap): Likewise.
12039 (vect_gen_perm_mask_checked): Likewise.
12040 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12041 implementations of variable permutation vectors into account
12042 when deciding which selector to use.
12043 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12044 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12045 with a false third argument.
12046 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12047 to test whether the constant selector is valid and can_vec_perm_var_p
12048 to test whether a variable selector is valid.
12050 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12052 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12053 * optabs-query.c (can_vec_perm_p): Likewise.
12054 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12055 instead of vec_perm_indices.
12056 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12057 (vect_gen_perm_mask_checked): Likewise,
12058 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12059 (vect_gen_perm_mask_checked): Likewise,
12061 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12063 * optabs-query.h (qimode_for_vec_perm): Declare.
12064 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12065 (qimode_for_vec_perm): ...this new function.
12066 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12068 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12070 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12071 does not have a conditional at the top.
12073 2018-01-02 Richard Biener <rguenther@suse.de>
12075 * ipa-inline.c (big_speedup_p): Fix expression.
12077 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12079 PR target/81616
12080 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12081 for generic 4->6.
12083 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12085 PR target/81616
12086 Generic tuning.
12087 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12088 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12089 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12090 cond_taken_branch_cost 3->4.
12092 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12094 PR tree-optimization/83581
12095 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12096 TODO_cleanup_cfg if any changes have been made.
12098 PR middle-end/83608
12099 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12100 convert_modes if target mode has the right side, but different mode
12101 class.
12103 PR middle-end/83609
12104 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12105 last argument when extracting from CONCAT. If either from_real or
12106 from_imag is NULL, use expansion through memory. If result is not
12107 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12108 the parts directly to inner mode, if even that fails, use expansion
12109 through memory.
12111 PR middle-end/83623
12112 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12113 check for bswap in mode rather than HImode and use that in expand_unop
12114 too.
12115 \f
12116 Copyright (C) 2018 Free Software Foundation, Inc.
12118 Copying and distribution of this file, with or without modification,
12119 are permitted in any medium without royalty provided the copyright
12120 notice and this notice are preserved.