1 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
4 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
6 (cmp_op, sve_imm_con): New code attributes.
7 (SVE_COND_INT_CMP, imm_con): Delete.
8 (cmp_op): Remove above unspecs from int attribute.
9 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
11 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
12 comparison-specific unspecs.
13 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
14 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
15 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
16 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
17 (*vec_fcm<cmp_op><mode>): Rename to...
18 (*fcm<cmp_op><mode>): ...this and adjust likewise.
19 (*vec_fcmuo<mode>): Rename to...
20 (*fcmuo<mode>): ...this and adjust likewise.
21 (*pred_fcm<cmp_op><mode>): New pattern.
22 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
23 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
25 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
27 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
28 (aarch64_emit_sve_predicated_cond): New function.
29 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
30 (aarch64_emit_unspec_cond_or): Replace with...
31 (aarch64_emit_sve_or_conds): ...this new function. Use
32 aarch64_emit_sve_ptrue_op for the individual comparisons and
33 aarch64_emit_binop to OR them together.
34 (aarch64_emit_inverted_unspec_cond): Replace with...
35 (aarch64_emit_sve_inverted_cond): ...this new function. Use
36 aarch64_emit_sve_ptrue_op for the comparison and
37 aarch64_emit_unop to invert the result.
38 (aarch64_expand_sve_vec_cmp_float): Update after the above
39 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
41 2018-05-07 Nathan Sidwell <nathan@acm.org>
43 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
44 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
45 (Backwards Compatibility): Likewise.
47 2018-05-07 Luis Machado <luis.machado@linaro.org>
51 2018-05-07 Luis Machado <luis.machado@linaro.org>
53 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
54 <prefetch_dynamic_strides>: New const bool field.
55 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
56 prefetch_dynamic_strides.
57 (exynosm1_prefetch_tune): Likewise.
58 (thunderxt88_prefetch_tune): Likewise.
59 (thunderx_prefetch_tune): Likewise.
60 (thunderx2t99_prefetch_tune): Likewise.
61 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
62 (aarch64_override_options_internal): Update to set
63 PARAM_PREFETCH_DYNAMIC_STRIDES.
64 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
65 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
66 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
67 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
68 prefetch-dynamic-strides setting.
70 2018-05-07 Luis Machado <luis.machado@linaro.org>
72 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
73 <minimum_stride>: New const int field.
74 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
76 (exynosm1_prefetch_tune): Likewise.
77 (thunderxt88_prefetch_tune): Likewise.
78 (thunderx_prefetch_tune): Likewise.
79 (thunderx2t99_prefetch_tune): Likewise.
80 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
81 (aarch64_override_options_internal): Update to set
82 PARAM_PREFETCH_MINIMUM_STRIDE.
83 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
84 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
85 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
86 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
87 stride is constant and is below the minimum stride threshold.
89 2018-05-07 Luis Machado <luis.machado@linaro.org>
91 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
94 2018-05-07 Luis Machado <luis.machado@linaro.org>
96 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
97 <prefetch_dynamic_strides>: New const bool field.
98 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
99 prefetch_dynamic_strides.
100 (exynosm1_prefetch_tune): Likewise.
101 (thunderxt88_prefetch_tune): Likewise.
102 (thunderx_prefetch_tune): Likewise.
103 (thunderx2t99_prefetch_tune): Likewise.
104 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
105 (aarch64_override_options_internal): Update to set
106 PARAM_PREFETCH_DYNAMIC_STRIDES.
107 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
108 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
109 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
110 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
111 prefetch-dynamic-strides setting.
113 2018-05-07 Luis Machado <luis.machado@linaro.org>
115 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
116 <minimum_stride>: New const int field.
117 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
118 minimum_stride field.
119 (exynosm1_prefetch_tune): Likewise.
120 (thunderxt88_prefetch_tune): Likewise.
121 (thunderx_prefetch_tune): Likewise.
122 (thunderx2t99_prefetch_tune): Likewise.
123 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
124 (aarch64_override_options_internal): Update to set
125 PARAM_PREFETCH_MINIMUM_STRIDE.
126 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
127 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
128 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
129 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
130 stride is constant and is below the minimum stride threshold.
132 2018-05-06 Jakub Jelinek <jakub@redhat.com>
135 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
136 the type is addressable. Don't force op into register if it has
139 2018-05-05 Roland McGrath <mcgrathr@google.com>
142 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
143 any section for which we don't know a specific type it should have,
144 regardless of name. Previously this was done only for the exact
145 names ".init_array", ".fini_array", and ".preinit_array".
146 (default_elf_asm_named_section): Add comment about
147 relationship with default_section_type_flags and SECTION_NOTYPE.
148 (get_section): Don't consider it a type conflict if one side has
149 SECTION_NOTYPE and the other doesn't, as long as neither has the
150 SECTION_BSS et al used in the default_section_type_flags logic.
152 2018-05-05 Tom de Vries <tom@codesourcery.com>
155 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
156 (workaround_barsyncs): New function.
157 (nvptx_reorg): Use workaround_barsyncs.
158 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
159 (define_expand "nvptx_membar_cta"): New define_expand.
160 (define_insn "*nvptx_membar_cta"): New insn.
162 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
164 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
165 To improve optimization opportunities.
166 * builtin-types.def: The new needed builtin types for the above.
168 2018-05-04 Richard Biener <rguenther@suse.de>
170 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
171 * gimple-ssa-store-merging.c
172 (imm_store_chain_info::output_merged_store): Remove redundant create,
173 release split_store vector contents on failure.
174 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
175 scalar stmt vector on cache hit.
177 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
179 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
181 * config.gcc (powerpc-xilinx-eabi*): Remove.
182 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
184 (fusion_addis_mem_combo_load): Ditto.
185 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
187 (rs6000_cpu_cpp_builtins): Ditto.
188 * config/rs6000/rs6000-linux.c
189 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
190 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
191 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
193 (rs6000_setup_reg_addr_masks): Ditto.
194 (rs6000_init_hard_regno_mode_ok): Ditto.
195 (rs6000_option_override_internal): Ditto.
196 (legitimate_lo_sum_address_p): Ditto.
197 (rs6000_legitimize_address): Ditto.
198 (rs6000_legitimize_reload_address): Ditto.
199 (rs6000_legitimate_address_p): Ditto.
200 (abi_v4_pass_in_fpr): Ditto.
201 (setup_incoming_varargs): Ditto.
202 (rs6000_gimplify_va_arg): Ditto.
203 (rs6000_split_multireg_move): Ditto.
204 (rs6000_savres_strategy): Ditto.
205 (rs6000_emit_prologue_components): Ditto.
206 (rs6000_emit_epilogue_components): Ditto.
207 (rs6000_emit_prologue): Ditto.
208 (rs6000_emit_epilogue): Ditto.
209 (rs6000_elf_file_end): Ditto.
210 (rs6000_function_value): Ditto.
211 (rs6000_libcall_value): Ditto.
212 * config/rs6000/rs6000.h: Ditto.
213 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
214 (TARGET_MINMAX): ... this. New.
215 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
216 * config/rs6000/rs6000.md: Remove Xilinx FP support.
217 (*movsi_internal1_single): Delete.
218 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
219 mfpu=, mxilinx-fpu): Delete.
220 * config/rs6000/singlefp.h: Delete.
221 * config/rs6000/sysv4.h: Remove Xilinx FP support.
222 * config/rs6000/t-rs6000: Ditto.
223 * config/rs6000/t-xilinx: Delete.
224 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
225 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
226 (VStype_simple): Delete.
227 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
228 * config/rs6000/xfpu.h: Delete.
229 * config/rs6000/xfpu.md: Delete.
230 * config/rs6000/xilinx.h: Delete.
231 * config/rs6000/xilinx.opt: Delete.
232 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
233 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
235 2018-05-04 Tom de Vries <tom@codesourcery.com>
238 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
241 2018-05-04 Richard Biener <rguenther@suse.de>
244 * tree-complex.c (update_complex_assignment): We are always in SSA form.
245 (expand_complex_div_wide): Likewise.
246 (expand_complex_operations_1): Likewise.
247 (expand_complex_libcall): Preserve EH info of the original stmt.
248 (tree_lower_complex): Handle removed blocks.
249 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
250 on complex multiplication and division libcall builtins.
252 2018-05-04 Richard Biener <rguenther@suse.de>
255 * fold-const.c (negate_expr_p): Restrict negation of operand
256 zero of a division to when we know that can happen without
258 (fold_negate_expr_1): Likewise.
260 2018-05-04 Jakub Jelinek <jakub@redhat.com>
263 * real.h (real_nextafter): Declare.
264 * real.c (real_nextafter): New function.
265 * fold-const-call.c (fold_const_nextafter): New function.
266 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
268 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
269 even when arg1_mode is different from arg0_mode.
271 2018-05-03 Nathan Sidwell <nathan@acm.org>
273 * doc/extend.texi (Deprecated Features): Remove
275 (Backwards Compatibility): Likewise.
276 * doc/invoke.texi (C++ Language Options): Likewise.
277 (C++ Dialect Options): Likewise.
279 2018-05-03 Jakub Jelinek <jakub@redhat.com>
282 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
283 _mm512_mask_mullox_epi64): New intrinsics.
285 2018-05-03 Tom de Vries <tom@codesourcery.com>
288 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
289 dump files): Add offload-tree.
291 2018-05-03 Richard Biener <rguenther@suse.de>
293 PR tree-optimization/85615
294 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
295 to loops not nested in BBs loop father to avoid creating multi-entry
298 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
300 PR tree-optimization/70291
301 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
302 arguments. Change return type to tree. Emit libcall as a new
303 statement rather than replacing existing one when inplace_p is true.
304 (expand_complex_multiplication_components): New function.
305 (expand_complex_multiplication): Expand floating-point complex
306 multiplication using the above.
307 (expand_complex_division): Rename inner_type parameter to type.
308 Update expand_complex_libcall call-site.
309 (expand_complex_operations_1): Update expand_complex_multiplication
310 and expand_complex_division call-sites.
312 2018-05-02 Jakub Jelinek <jakub@redhat.com>
315 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
316 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
317 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
318 the highest significant bit of the shift count mask is clear. In
319 check whether and[sq]i3 is needed verify that all significant bits
320 of the shift count other than the highest are set.
322 2018-05-02 Tom de Vries <tom@codesourcery.com>
325 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
326 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
327 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
328 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
329 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
330 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
331 __builtin_goacc_parlevel_size.
333 2018-05-02 Richard Biener <rguenther@suse.de>
335 PR tree-optimization/85597
336 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
337 do not use split vect_get_vec_defs call but call vect_get_slp_defs
340 2018-05-02 Tom de Vries <tom@codesourcery.com>
343 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
344 dump files): Add ltrans-tree.
346 2018-05-02 Tom de Vries <tom@codesourcery.com>
349 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
350 dump files): Add wpa-ipa.
352 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
354 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
355 powerpc*-*-linux*paired* target.
356 * config/rs6000/750cl.h: Delete.
357 * config/rs6000/paired.h: Delete.
358 * config/rs6000/paired.md: Delete.
359 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
361 * config/rs6000/rs6000-builtin.def: Remove paired float support.
362 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
363 comment. Remove paired float support.
364 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
365 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
367 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
368 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
370 * config/rs6000/rs6000.c: Remove paired float support.
371 (paired_expand_vector_init, paired_expand_vector_move,
372 paired_emit_vector_compare, paired_emit_vector_cond_expr,
373 (paired_expand_lv_builtin, paired_expand_stv_builtin,
374 paired_expand_builtin, paired_expand_predicate_builtin,
375 paired_init_builtins): Delete.
376 * config/rs6000/rs6000.h: Remove paired float support.
377 * config/rs6000/rs6000.md: Remove paired float support.
378 (move_from_CR_ov_bit): Delete.
379 * config/rs6000/rs6000.opt (mpaired): Delete.
380 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
381 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
383 2018-05-02 Richard Biener <rguenther@suse.de>
386 * gimplify.c (gimplify_save_expr): When in SSA form allow
387 SAVE_EXPRs to compute to SSA vars.
389 2018-05-02 Jakub Jelinek <jakub@redhat.com>
392 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
393 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
394 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
395 clobber operands[2], instead use a new pseudo. Formatting fixes.
397 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
399 PR tree-optimization/85586
400 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
401 exit early for statements in the same group if the accesses are
404 2018-05-02 Tom de Vries <tom@codesourcery.com>
407 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
410 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
412 PR tree-optimization/85143
413 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
415 2018-05-01 Tom de Vries <tom@codesourcery.com>
418 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
419 not found" error message.
421 2018-05-01 Tom de Vries <tom@codesourcery.com>
424 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
425 * vec.c (test_ordered_remove_if): New function.
426 (vec_c_tests): Call test_ordered_remove_if.
427 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
428 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
429 * tree-vect-patterns.c (vect_pattern_recog_1): Use
430 VEC_ORDERED_REMOVE_IF.
432 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
434 PR tree-optimization/82665
435 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
436 pointer subtraction where arguments come from a memchr call.
438 2018-05-01 Jakub Jelinek <jakub@redhat.com>
440 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
441 --push-state --as-needed and --pop-state instead of --as-needed and
442 --no-as-needed if ld supports it.
443 * configure: Regenerated.
446 * doc/install.texi2html: Replace _002d with - and _002a with * in
447 generated html files using sed.
449 2018-04-30 David Malcolm <dmalcolm@redhat.com>
452 * gcc-rich-location.c (blank_line_before_p): New function.
453 (use_new_line): New function.
454 (gcc_rich_location::add_fixit_insert_formatted): New function.
455 * gcc-rich-location.h
456 (gcc_rich_location::add_fixit_insert_formatted): New function.
458 2018-04-30 David Malcolm <dmalcolm@redhat.com>
460 * selftest.c (assert_streq): Rename "expected" and "actual" to
461 "val1" and "val2". Extend NULL-handling to cover both inputs
462 symmetrically, while still requiring both to be non-NULL for a pass.
463 * selftest.h (assert_streq): Rename "expected" and "actual" to
465 (ASSERT_EQ): Likewise.
466 (ASSERT_EQ_AT): Likewise.
467 (ASSERT_KNOWN_EQ): Likewise.
468 (ASSERT_KNOWN_EQ_AT): Likewise.
469 (ASSERT_NE): Likewise.
470 (ASSERT_MAYBE_NE): Likewise.
471 (ASSERT_MAYBE_NE_AT): Likewise.
472 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
473 the assertion to pass.
474 (ASSERT_STREQ_AT): Likewise.
476 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
478 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
479 interaction with -pie.
481 2018-04-30 David Malcolm <dmalcolm@redhat.com>
483 * selftest.h: Fix alphabetization of per-source-file selftest
486 2018-04-30 Jason Merrill <jason@redhat.com>
488 PR c++/61982 - dead stores to destroyed objects.
489 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
492 2018-04-30 Jason Merrill <jason@redhat.com>
494 * tree.c (build_clobber): New.
495 * tree.h: Declare it.
496 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
498 2018-04-30 David Malcolm <dmalcolm@redhat.com>
500 * diagnostic-show-locus.c (layout::layout): Update for
501 location_get_source_line returning a char_span.
502 (struct char_span): Move to input.h.
503 (struct correction): Update for fields in char_span becoming
505 (struct source_line): Update for location_get_source_line
506 returning a char_span.
507 (layout::print_line): Likewise.
508 * edit-context.c (edited_file::print_content): Likewise.
509 (edited_file::print_diff_hunk): Likewise.
510 (edited_file::print_run_of_changed_lines): Likewise.
511 (edited_file::get_num_lines): Likewise.
512 (edited_line::edited_line): Likewise.
513 * final.c (asm_show_source): Likewise.
514 * input.c (location_get_source_line): Convert return type
515 from const char * to char_span, losing the final "line_len"
517 (dump_location_info): Update for the above.
518 (get_substring_ranges_for_loc): Likewise. Use a char_span
519 when handling the literal within the line.
520 (test_reading_source_line): Update for location_get_source_line
521 returning a char_span.
522 * input.h (class char_span): Move here from
523 diagnostic-show-locus.c, converting from a struct to a class.
524 Make data members private.
525 (char_span::operator bool): New.
526 (char_span::length): New.
527 (char_span::get_buffer): New.
528 (char_span::operator[]): New.
529 (char_span::subspan): Make const.
530 (char_span::xstrdup): New.
531 (location_get_source_line): Convert return type from const char *
532 to char_span, losing the final "line_size" param.
534 2018-04-30 Jan Hubicka <jh@suse.cz>
536 * lto-wrapper.c (ltrans_priorities): New static var.
538 (run_gcc): Read priorities and if doing parallel build order
539 the Makefile by them.
541 2018-04-30 David Malcolm <dmalcolm@redhat.com>
543 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
545 2018-04-30 Richard Biener <rguenther@suse.de>
547 * tree-cfg.c (verify_address): Remove base argument, add
548 flag whether to check TREE_ADDRESSABLE and do that.
549 (verify_expr): Remove.
550 (verify_types_in_gimple_reference): Add pieces from verify_expr.
551 (verify_gimple_assign_single): Likewise.
552 (verify_gimple_switch): Likewise.
553 (verify_expr_location_1): Dereference tp once. Add (disabled)
554 piece from verify_expr.
555 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
557 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
559 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
561 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
563 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
564 (small_data_pattern): Likewise.
565 (arc_rewrite_small_data): Likewise.
566 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
567 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
568 (get_symbol_alignment): New function.
569 (legitimate_small_data_address_p): Likewise.
570 (legitimate_scaled_address): Update, call
571 legitimate_small_data_address_p.
572 (output_sdata): New static variable.
573 (arc_print_operand): Update how we handle small data operands.
574 (arc_print_operand_address): Likewise.
575 (arc_legitimate_address_p): Update, use
576 legitimate_small_data_address_p.
577 (arc_rewrite_small_data_p): Remove.
578 (arc_rewrite_small_data_1): Likewise.
579 (arc_rewrite_small_data): Likewise.
580 (small_data_pattern): Likewise.
581 (compact_sda_memory_operand): Update to use
582 legitimate_small_data_address_p and get_symbol_alignment.
583 (prepare_move_operands): Don't rewite sdata pattern.
584 (prepare_extend_operands): Remove.
585 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
587 (zero_extendqisi2): Likewise.
588 (zero_extendhisi2): Likewise.
589 (extendqihi2): Likewise.
590 (extendqisi2): Likewise.
591 (extendhisi2): Likewise.
595 * config/arc/constraints.md (Usd): Change it to memory constraint.
597 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
599 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
600 as source of std instructions.
601 * config/arc/arc.md (movsi_insn): Update pattern predicate to
602 allow 6-bit constants as source for store instructions.
603 (movdi_insn): Update instruction pattern to allow 6-bit constants
604 as source for store instructions.
606 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
608 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
610 2018-04-30 Nathan Sidwell <nathan@acm.org>
611 Sandra Loosemore <sandra@codesourcery.com>
613 * dumpfile.c (dump_open): Allow '-' for stdout.
614 * doc/invoke.texi (Developer Options): Document dump filename
615 determination early. Document stdin/stdout selection.
617 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
619 Microblaze Target: PIC data text relative
621 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
622 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
624 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
625 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
626 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
627 New addressing mode for data-text relative position indepenedent code.
628 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
629 'ADDRESS_SYMBOLIC_TXT_REL'.
630 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
631 (microblaze_legitimate_pic_operand): Exclude function calls from
632 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
633 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
635 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
636 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
637 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
638 for 'address + offset'.
639 (microblaze_expand_prologue): Add new function prologue call for
641 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
642 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
643 table in case of TARGET_PIC_DATA_TEXT_REL.
644 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
645 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
646 Add new macros 'UNSPEC_TEXT',
647 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
648 + exclude function calls from 'UNSPEC_PLT' in case of data text
650 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
651 new target hook for generating address diff vector tables in case of
653 * doc/tm.texi : Regenerate.
654 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
655 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
656 of addr diff vector generation.
657 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
658 target hook definition.
659 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
660 Add default function for generate_pic_addr_diff_vec -> flag_pic.
661 * doc/invoke.texi (Add new pic option): Add new microblaze pic
662 option for data text relative.
664 2018-04-30 Richard Biener <rguenther@suse.de>
666 * tree-chrec.h (evolution_function_is_constant_p): Remove
668 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
670 2018-04-30 Richard Biener <rguenther@suse.de>
673 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
675 2018-04-30 Richard Biener <rguenther@suse.de>
677 PR tree-optimization/28364
678 PR tree-optimization/85275
679 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
680 copying first exit test.
682 2018-04-28 Mark Wielaard <mark@klomp.org>
684 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
686 (dwarf_AT): Handle DW_AT_addr_base.
687 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
689 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
692 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
693 (*ashl<dwi>3_doubleword_mask_1): Ditto.
694 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
695 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
697 2018-04-28 Richard Biener <rguenther@suse.de>
699 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
700 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
701 to reflect use. Only add interesting stmts.
703 2018-04-27 Martin Jambor <mjambor@suse.cz>
706 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
707 the jump function allows for passing through aggregate values.
709 2018-04-27 David Malcolm <dmalcolm@redhat.com>
711 * input.h (in_system_header_at): Convert from macro to inline
713 (from_macro_expansion_at): Likewise.
714 (from_macro_definition_at): Likewise.
716 2018-04-27 Jeff Law <law@redhat.com>
718 * config.gcc: Mark tile* targets as deprecated/obsolete.
720 2018-04-27 Richard Biener <rguenther@suse.de>
722 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
725 2018-04-27 Richard Biener <rguenther@suse.de>
727 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
729 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
731 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
732 with Yd constraint. Set "preferred_for_speed" attribute from
733 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
735 (*movdi_internal): Ditto.
736 (movti_interunit splitters): Remove
737 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
738 (movdi_interunit splitters): Ditto.
739 * config/i386/constraints.md (Ye): Remove.
740 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
742 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
745 * config/aarch64/constraints.md (Usg): Limit to 31.
748 2018-04-27 Jakub Jelinek <jakub@redhat.com>
750 PR tree-optimization/85529
751 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
752 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
753 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
754 zero extension or masking of the MSB bit.
755 (optimize_range_tests): Add FIRST_BB argument, pass it through
756 to optimize_range_tests_var_bound.
757 (maybe_optimize_range_tests, reassociate_bb): Adjust
758 optimize_range_tests callers.
760 2018-04-26 Richard Biener <rguenther@suse.de>
761 Jakub Jelinek <jakub@redhat.com>
763 * cgraph.h (symbol_table): Just declare debug method here.
764 * symtab.c (symbol_table::debug): Define.
766 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
768 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
770 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
772 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
773 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
774 (*movdi_internal): Substitute Yi and Yj constraint with x
775 and Ym and Yn constraint with y constraint. Update "isa"
776 attribute and set "preferred_for_speed" attribute from
777 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
778 (*movsi_internal): Ditto.
779 (*movdf_internal): Ditto.
780 (*movsf_internal): Ditto.
781 (*zero_extendsidi2): Ditto.
782 * config/i386/sse.md (vec_set<mode>_0): Ditto.
783 (sse2_loadld): Ditto.
784 (*vec_extract<ssevecmodelower>_0): Ditto.
785 (*vec_extractv4si_0_zext_sse4): Ditto.
786 (vec_concatv2di): Ditto.
787 (*vec_dup<mode>): Ditto.
788 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
789 * config/i386/constraints.md (Yi): Remove.
794 2018-04-26 Nathan Sidwell <nathan@acm.org>
796 * dumpfile.c (dump_open): New.
797 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
798 (dump_finish): Detect stdio/stderr by value not name.
800 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
802 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
804 2018-04-26 Tom de Vries <tom@codesourcery.com>
807 * config/nvptx/nvptx.c (verify_neutering_jumps)
808 (verify_neutering_labels): New function
809 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
811 2018-04-26 Tom de Vries <tom@codesourcery.com>
814 * config/nvptx/nvptx.c (needs_neutering_p): New function.
815 (nvptx_single): Use needs_neutering_p to skip over insns that do not
818 2018-04-26 Richard Biener <rguenther@suse.de>
819 Tom de Vries <tom@codesourcery.com>
822 * lto-streamer-out.c (output_function): Fixup loops if required to match
823 discovery done in the reader.
825 2018-04-26 Richard Biener <rguenther@suse.de>
827 PR tree-optimization/85116
828 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
829 have a loop exit from the single latch predecessor. Remove
830 case of header with just condition.
831 (ch_base::copy_headers): Exclude infinite loops from any
833 (pass_ch::execute): Record exits.
835 2018-04-26 Richard Biener <rguenther@suse.de>
837 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
838 prologue cost vector and pass it to vect_get_load_cost.
839 (vect_get_peeling_costs_all_drs): Likewise.
840 (vect_peeling_hash_get_lowest_cost): Likewise.
841 (vect_enhance_data_refs_alignment): Likewise.
843 2018-04-26 Richard Biener <rguenther@suse.de>
846 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
847 checking of integer<->pointer conversions.
848 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
849 sign-/zero-extending pointer types.
850 (expand_omp_for_static_chunk): Likewise.
852 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
853 Jean Lee <xiaoyur347@gmail.com>
855 * config/mips/mips.c (mips_asan_shadow_offset): New function.
856 (TARGET_ASAN_SHADOW_OFFSET): Define.
857 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
858 true for -fsanitize=address.
860 2018-04-25 Mark Wielaard <mark@klomp.org>
862 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
865 2018-04-25 Jakub Jelinek <jakub@redhat.com>
867 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
868 than "alu", remove explicit "memory" and "imm_disp" attributes.
869 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
872 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
873 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
876 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
879 * config/i386/i386.c (ix86_expand_builtin): Change memory
880 operand to XI, extend p0 to Pmode.
881 * config/i386/i386.md: Change unspec volatile and operand
882 1 mode to XI, change operand 0 mode to P.
884 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
886 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
887 GET_MODE_MASK before any checking.
888 (nds32_can_use_bset_p): Likewise.
889 (nds32_can_use_btgl_p): Likewise.
891 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
893 * config/nds32/nds32-doubleword.md: New define_split pattern for
894 illegal register number.
896 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
898 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
900 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
902 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
904 2018-04-25 Richard Biener <rguenther@suse.de>
906 * lto-streamer.h (LTO_major_version): Bump to 8.
908 2018-04-25 Jakub Jelinek <jakub@redhat.com>
910 * BASE-VER: Set to 9.0.0.
912 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
914 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
915 in __abskf2 and __powikf2.
917 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
920 * config/aarch64/constraints.md (Usg, Usj): New constraints.
921 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
922 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
923 Use the above on operand 2. Reindent.
924 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
926 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
929 * common/config/i386/i386-common.c (ix86_handle_option): Don't
931 * config/i386/i386.opt (mcet): Removed.
932 * doc/install.texi: Remove -mcet documentation.
933 * doc/invoke.texi: Likewise.
935 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
938 * doc/install.texi: Remove -mcet from bootstrap-cet.
940 2018-04-24 Jakub Jelinek <jakub@redhat.com>
943 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
944 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
948 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
949 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
950 containing a CONST_VECTOR.
952 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
954 * doc/install.texi: Update newlib dependency for nvptx.
956 2018-04-24 Jakub Jelinek <jakub@redhat.com>
959 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
960 instead of INTVAL when shifting x left.
962 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
964 PR tree-optimization/85478
965 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
966 vect_grouped_store_supported for single element vectors.
968 2018-04-24 Richard Biener <rguenther@suse.de>
971 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
972 load cost increase to the case of non-constant step.
974 2018-04-24 Jakub Jelinek <jakub@redhat.com>
977 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
978 destination if any_malformed_asm.
980 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
983 * expr.c (store_field): In the bitfield case, if the value comes from
984 a function call and is returned in registers by means of a PARALLEL,
985 do not change the mode of the temporary unless BLKmode and VOIDmode.
987 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
989 PR rtl-optimization/85423
990 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
991 dependencies to debug insns when the previous insn is non-debug.
993 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
995 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
996 enums into a single definition.
997 (fls): Fix predicates and printing.
1000 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1002 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1003 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1004 and short u6 immediate.
1005 (check_if_valid_sleep_operand): Remove.
1006 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1008 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1010 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1011 flag_always_save_lp condition.
1012 * config/nds32/nds32.opt (malways-save-lp): New option.
1014 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1016 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1017 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1018 * config/nds32/nds32.h
1019 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1020 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1022 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1024 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1025 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1027 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1028 Chung-Ju Wu <jasonwucj@gmail.com>
1030 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1032 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1033 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1035 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1037 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1039 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1041 * config/nds32/nds32-protos.h (nds32_data_alignment,
1042 nds32_local_alignment): Declare.
1043 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1044 nds32_local_alignment): New functions.
1045 (TARGET_CONSTANT_ALIGNMENT): Define.
1046 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1048 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1050 * config/nds32/nds32.c
1051 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1052 (TARGET_MODES_TIEABLE_P): Likewise.
1054 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1056 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1059 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1060 Chung-Ju Wu <jasonwucj@gmail.com>
1062 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1063 for unaligned access.
1064 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1066 * config/nds32/nds32-intrinsic.md: Likewise.
1067 * config/nds32/nds32_intrinsic.h: Likewise.
1068 * config/nds32/nds32.h (nds32_builtins): Likewise.
1069 * config/nds32/nds32.opt (munaligned-access): New option.
1070 * config/nds32/nds32.c (nds32_asm_file_start): Display
1071 flag_unaligned_access status.
1073 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1075 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1076 -mno-relax is present.
1077 * config/riscv/linux.h (LINK_SPEC): Ditto.
1079 2018-04-20 Martin Sebor <msebor@redhat.com>
1082 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1084 (gimple_fold_builtin_stxcpy_chk): Same.
1085 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1087 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1090 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1091 __powikf2 when long double is IEEE 128-bit.
1093 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1095 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1096 step to make sure stack always aligned.
1098 2018-04-20 Carl Love <cel@us.ibm.com>
1101 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1102 size check for arg0.
1104 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1105 Tom de Vries <tom@codesourcery.com>
1108 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1109 Emit insns for calls too.
1110 (nvptx_find_par): Always look for worker-level predecessor insn.
1111 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1113 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1114 (nvptx_process_pars): Propagate frames for calls.
1116 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1119 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1121 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1122 (ix86_handle_option): Don't handle OPT_mibt.
1123 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1125 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1127 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1128 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1129 (ix86_target_macros): Define __CET__ with flag_cf_protection
1130 for -fcf-protection.
1131 * config/i386/i386.c (isa2_opts): Remove -mibt.
1132 * config/i386/i386.h (TARGET_IBT): Removed.
1133 (TARGET_IBT_P): Likewise.
1134 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1135 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1136 * config/i386/i386.opt (mcet): Update help message.
1139 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1140 -mcet as an alias for -mshstk.
1142 2018-04-20 Richard Biener <rguenther@suse.de>
1145 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1146 complexity by forcing a single use of the multiply operand.
1148 2018-04-20 Martin Jambor <mjambor@suse.cz>
1151 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1152 recursion dependency to only apply to non-clones.
1154 2018-04-20 Martin Jambor <mjambor@suse.cz>
1157 * ipa-cp.c (create_specialized_node): Check that clones of
1158 self-recursive edges exist during IPA-CP.
1160 2018-04-19 Toon Moene <toon@moene.org>
1162 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1165 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1167 PR tree-optimization/85467
1168 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1169 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1170 VECTOR_CST element to type.
1172 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1175 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1176 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1177 (builtin_longjmp): Likewise.
1178 (save_stack_nonlocal): New pattern.
1179 (restore_stack_nonlocal): Likewise.
1181 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1184 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1185 Replace ASM_OUTPUT_LABEL with fprintf.
1187 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1190 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1191 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1192 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1193 define __IBT__ and __SHSTK__ for -fcf-protection.
1194 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1196 (ix86_trampoline_init): Likewise.
1197 (x86_output_mi_thunk): Likewise.
1198 (ix86_notrack_prefixed_insn_p): Likewise.
1199 (ix86_option_override_internal): Don't disallow -fcf-protection.
1200 * config/i386/i386.md (rdssp<mode>): Also enable for
1202 (incssp<mode>): Likewise.
1203 (nop_endbr): Likewise.
1204 * config/i386/i386.opt (mcet): Change help message to built-in
1208 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1209 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1210 enable CET built-in functions.
1212 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1214 * common/config/i386/i386-common.c
1215 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1216 OPTION_MASK_ISA_MOVDIRI_UNSET,
1217 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1218 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1219 * config.gcc (movdirintrin.h): New header.
1220 * config/i386/cpuid.h (bit_MOVDIRI,
1221 bit_MOVDIR64B): New bits.
1222 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1224 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1225 (VOID, PVOID, PCVOID)): New function types.
1226 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1227 __builtin_ia32_directstoreu_u64,
1228 __builtin_ia32_movdir64b): New builtins.
1229 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1230 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1232 (ix86_valid_target_attribute_inner_p): Ditto.
1233 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1234 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1235 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1236 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1237 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1238 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1239 (movdiri<mode>, movdir64b_<mode>): New.
1240 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1241 * config/i386/immintrin.h: Include movdirintrin.h.
1242 * config/i386/movdirintrin.h: New file.
1243 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1245 2018-04-19 Richard Biener <rguenther@suse.de>
1248 * cfg.c (clear_bb_flags): When loop state says we have
1249 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1251 2018-04-19 Richard Biener <rguenther@suse.de>
1253 PR tree-optimization/84737
1254 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1255 copying restrict info.
1256 (vect_setup_realignment): Use it.
1257 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1258 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1259 the first DR to all generated stores.
1260 (vectorizable_load): Likewise for loads.
1262 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1264 PR tree-optimization/85446
1265 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1266 the integral and pointer types to have the same precision.
1268 * doc/install.texi: Document --disable-cet being the default and
1271 2018-04-18 Martin Liska <mliska@suse.cz>
1273 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1276 2018-04-18 Martin Liska <mliska@suse.cz>
1279 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1282 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1283 arguments if they are comparable.
1285 2018-04-18 Martin Liska <mliska@suse.cz>
1288 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1291 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1294 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1297 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1298 ENDBR after calling __morestack.
1300 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1303 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1304 by using gcc_base_ver to generate a gcc_driver_version, and use
1305 it when generating GCC_DRIVER_NAME.
1306 * configure: Regenerate.
1308 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1311 * config.gcc: Obsolete powerpc*-*-*spe*.
1313 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1316 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1317 (stabstr_D): Change type of unum from unsigned int to
1318 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1321 2018-04-17 Jim Wilson <jimw@sifive.com>
1324 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1325 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1326 Set arg_pointer_offset after using pretend_args_size.
1328 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1330 PR rtl-optimization/85431
1331 * dse.c (record_store): Ignore zero width stores.
1334 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1335 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1336 __builtin_stack_restore rather than after it.
1337 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1338 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1339 argument instead of virtual_dynamic_stack_rtx.
1341 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1343 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1345 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1346 Add note to error message to explain internal mapping of overloaded
1347 built-in function name to non-overloaded built-in function name.
1348 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1351 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1354 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1355 where the inputs overlap with the output.
1357 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1359 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1360 (=v, v) alternative and explicit "memory" attribute.
1361 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1362 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1364 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1365 "sselog1" type instead of "sselog".
1366 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1367 "sselog". Remove explicit "memory" attribute.
1368 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1369 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1371 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1372 "sselog1" type instead of "sselog". Remove explicit "memory"
1374 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1375 use "sselog1" type instead of "sselog". Remove explicit "memory"
1377 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1378 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1380 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1381 "sselog1" type instead of "sselog". Remove explicit "memory"
1383 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1384 use "sselog1" type instead of "sselog". Remove explicit "memory"
1388 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1391 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1394 2018-04-17 Martin Jambor <mjambor@suse.cz>
1397 * ipa-cp.c (create_specialized_node): Call
1398 expand_all_artificial_thunks if necessary.
1400 2018-04-17 Martin Liska <mliska@suse.cz>
1403 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1404 in message, remote space in between '_G' and '('.
1406 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1409 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1410 avx512f_vmcmp<mode>3<round_saeonly_name>,
1411 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1412 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1413 avx512f_rndscale<mode><round_saeonly_name>,
1414 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1415 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1416 Use %<iptr>2 instead of %2 for -masm=intel.
1417 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1418 avx512f_vcvttss2usi<round_saeonly_name>,
1419 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1421 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1422 avx512f_vcvttsd2usi<round_saeonly_name>,
1423 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1424 Use %q1 instead of %1 for -masm=intel.
1425 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1426 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1427 of %3 for -masm=intel.
1428 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1430 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1432 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1434 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1435 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1437 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1439 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1440 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1441 %g1 and one with %0 and %1.
1442 (avx512er_vmrcp28<mode><round_saeonly_name>,
1443 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1445 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1446 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1447 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1448 of %0 and %{%4%} for -masm=intel.
1449 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1450 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1451 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1452 order of %0 and %{%5%}%{z%} for -masm=intel.
1454 2018-04-17 Jan Hubicka <jh@suse.cz>
1457 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1459 2018-04-17 Martin Liska <mliska@suse.cz>
1462 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1463 for target_clone error message. Make default implementation
1464 clone to be a local declaration.
1465 (separate_attrs): Add new argument and check for an empty
1467 (expand_target_clones): Handle it.
1468 (ipa_target_clone): Make redirection just for target_clones
1471 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1472 Tom de Vries <tom@codesourcery.com>
1475 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1476 tiled basic blocks without omp continue statements.
1478 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1481 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1482 vec_extract expression as having side effects to make sure it gets
1485 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1488 * config/i386/i386.c (get_builtin_code_for_version): Check
1491 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1494 * gcc/config.gcc: Support "skylake".
1495 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1497 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1498 (processor_target_table): Add "skylake".
1499 (ix86_option_override_internal): Add "skylake".
1500 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1501 PROCESSOR_CANNONLAKE.
1502 (get_builtin_code_for_version): Fix priority for
1503 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1504 PROCESSOR_SKYLAKE-AVX512.
1505 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1506 (processor_type): Add PROCESSOR_SKYLAKE.
1508 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1509 Jason Merrill <jason@redhat.com>
1512 * convert.c (convert_to_integer_1): Use direct recursion for
1513 enumeral types and types with a precision less than the number
1514 of bits in their mode.
1516 2018-04-16 Julia Koval <julia.koval@intel.com>
1519 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1520 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1522 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1525 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1526 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1527 and -mno-direct-move.
1529 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1532 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1533 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1534 Ensure negative shifts result in {0}.
1536 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1538 PR rtl-optimization/79916
1539 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1540 regs (if any) to define how to gnerate SD moves when LRA is in
1543 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1545 PR rtl-optimization/85393
1546 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1547 * except.c (expand_dw2_landing_pad_for_region): Make static.
1548 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1549 a label and unconditional jump to old_bb, rather than
1550 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1553 PR rtl-optimization/85376
1554 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1555 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1556 instead of a specific value.
1558 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1559 Bin Cheng <bin.cheng@arm.com>
1561 PR tree-optimization/82965
1562 PR tree-optimization/83991
1563 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1564 by_profile_only parameter.
1565 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1566 information if the loop was predicted to iterate too many times.
1567 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1569 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1572 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1575 2018-04-13 Martin Liska <mliska@suse.cz>
1576 Jakub Jelinek <jakub@redhat.com>
1579 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1580 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1581 * builtins.c (expand_builtin_memory_copy_args): Use
1582 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1583 handle dest_addr == pc_rtx.
1585 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1588 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1589 asked to not generate direct moves.
1590 (fix_trunc<mode>si2_stfiwx): Similar.
1591 (fix_trunc<mode>si2_internal): Similar.
1593 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1596 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1597 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1598 lookup if dest in some wider mode is known to be const0_rtx and
1599 if so, record permanent equivalence for it to be ZERO_EXTEND of
1600 the narrower mode destination.
1602 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1604 * lto-streamer-out.c (output_function): Revert 259346.
1605 * omp-expand.c (expand_oacc_for): Likewise.
1607 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1609 PR rtl-optimization/85354
1610 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1611 * sel-sched.c (sel_global_init): ... here.
1613 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1616 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1617 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1618 mode for PE-COFF targets.
1619 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1620 (i386_pe_asm_lto_end): Likewise.
1621 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1622 (TARGET_ASM_LTO_END): Likewise.
1623 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1624 (i386_pe_asm_lto_start): New function.
1625 (i386_pe_asm_lto_end): Likewise.
1627 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1628 Richard Biener <rguenther@suse.de>
1631 * lto-streamer-out.c (output_function): Fix CFG loop state before
1633 * omp-expand.c (expand_oacc_for): Handle calls to internal
1634 functions like regular functions.
1636 2018-04-12 Richard Biener <rguenther@suse.de>
1639 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1640 for the early LTO debug to properly generate references to it
1641 during DIE emission. Do not re-use that for the skeleton for
1643 (dwarf2out_early_finish): Likewise.
1645 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1648 * config/i386/sse.md
1649 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1650 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1651 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1652 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1653 and output is a reg, avoid creating invalid lowpart subreg, but
1654 instead split into a 512-bit move. Don't split if not AVX512VL,
1655 input is xmm16+ reg and output is a mem.
1656 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1657 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1658 xmm16+ reg and output is a mem.
1660 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1662 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1663 also for flag_dwarf2_cfi_asm.
1665 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1667 PR rtl-optimization/85342
1668 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1669 a bool scalar var inside of the loop instead. Don't try to update
1670 recog_data.operand after failed apply_change_group.
1672 2018-04-12 Tom de Vries <tom@codesourcery.com>
1675 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1676 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1677 array with flexible array member as array without given dimension.
1678 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1679 argument for undefined param to true.
1681 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1684 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1685 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1686 from PowerPC section.
1687 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1688 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1689 help text that is too long.
1690 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1691 help text that is too long.
1692 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1693 help text that is too long.
1695 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1697 * config/alpha/alpha.md (stack_probe_internal): Rename
1698 from "probe_stack". Update all callers.
1700 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1702 PR rtl-optimization/84566
1703 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1704 sched_macro_fuse_insns.
1706 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1709 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1710 (compute_block_dependences): ... from here.
1712 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1714 PR tree-optimization/85331
1715 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1716 from int to HOST_WIDE_INT.
1718 2018-04-11 Martin Jambor <mjambor@suse.cz>
1721 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1722 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1723 not the same as the source val.
1724 (cgraph_edge_brings_value_p): New parameter.
1725 (gather_edges_for_value): Pass destination value to
1726 cgraph_edge_brings_value_p.
1727 (perhaps_add_new_callers): Likewise.
1728 (get_info_about_necessary_edges): Likewise and exclude values brought
1729 only by self-recursive edges.
1730 (create_specialized_node): Redirect only clones of self-calling edges.
1731 (+self_recursive_pass_through_p): New function.
1732 (find_more_scalar_values_for_callers_subset): Use it.
1733 (find_aggregate_values_for_callers_subset): Likewise.
1734 (known_aggs_to_agg_replacement_list): Removed.
1735 (decide_whether_version_node): Re-calculate known constants for all
1736 remaining context clones.
1738 2018-04-11 Richard Biener <rguenther@suse.de>
1741 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1742 from early DWARF output.
1743 (dwarf2out_early_finish): Output line info unconditionally into
1744 early DWARF and add reference to it.
1746 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1749 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1750 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1751 other than V2DFmode using iptr mode attribute.
1752 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1754 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1756 PR rtl-optimization/84659
1757 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1759 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1762 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1764 (output_loc_list): Pass address of a dummy size variable even in the
1765 locview handling loop.
1766 (index_location_lists): Add comment on why skip_loc_list_entry can't
1769 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1772 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1775 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1778 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1779 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1780 and -mstring-compare-inline-limit.
1782 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1785 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1786 for stack clash protection in a register whenever we need it to be in
1789 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1791 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1792 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1794 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1797 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1799 (mlong-double-): Ditto.
1800 * config/rs6000/sysv4.opt (msdata=): Ditto.
1801 (mtls-size=): Ditto.
1803 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1805 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1806 erroneous entries for
1807 "vector int vec_ldl (int, long int *)", and
1808 "vector unsigned int vec_ldl (int, unsigned long int *)".
1809 Add comments and entries for
1810 "vector bool char vec_ldl (int, bool char *)",
1811 "vector bool short vec_ldl (int, bool short *)",
1812 "vector bool int vec_ldl (int, bool int *)",
1813 "vector bool long long vec_ldl (int, bool long long *)",
1814 "vector pixel vec_ldl (int, pixel *)",
1815 "vector long long vec_ldl (int, long long *)",
1816 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1817 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1818 type tree bool_long_long_type_node and correct definition of
1819 bool_V2DI_type_node to make reference to this new type tree.
1820 (rs6000_mangle_type): Replace erroneous reference to
1821 bool_long_type_node with bool_long_long_type_node.
1822 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1823 comments to emphasize sign distinctions for char and int types and
1824 replace RS6000_BTI_bool_long constant with
1825 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1826 use of RS6000_BTI_pixel.
1827 (bool_long_type_node): Remove this macro definition.
1828 (bool_long_long_type_node): New macro definition
1830 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1832 PR rtl-optimization/85300
1833 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1834 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1835 simplify_unary_operation fails.
1837 2018-04-10 Martin Liska <mliska@suse.cz>
1839 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1840 cgraph_edge and ipa_ref.
1842 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1846 * config/i386/sse.md
1847 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1848 computation of the VEC_MERGE selector from mask.
1849 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1850 Fix decoding of the VEC_MERGE selector into mask.
1852 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1854 PR tree-optimization/85286
1855 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1857 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1859 * final.c (final_1): Set insn_last_address as well as
1860 insn_current_address.
1862 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1865 * explow.c (emit_stack_probe): Call validize_mem on memory location
1866 before passing it to gen_probe_stack. Create address operand and
1867 legitimize it for the probe_stack_address case.
1869 2018-04-09 Jan Hubicka <jh@suse.cz>
1872 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1873 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1874 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1876 (free_lang_data): Rebuild type inheritance graph.
1878 2018-04-09 Martin Sebor <msebor@redhat.com>
1880 * invoke.texi (-finline-small-functions): Mention other optimization
1882 (-findirect-inlining, -fpartial-inlining): Same.
1883 (-finline-functions-called-once): Same.
1884 (-freorder-blocks-and-partition): Same.
1886 2018-04-09 Jan Hubicka <jh@suse.cz>
1889 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1890 jumps; choose last target that matches the criteria (i.e.
1891 no partition changes for non-crossing jumps).
1892 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1893 support for redirecting crossing jumps to non-crossing.
1895 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1897 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1898 also for naked functions.
1900 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1902 * config/arc/arc.md (add_shift): New pattern.
1903 (add_shift2): Likewise.
1904 (sub_shift): Likewise.
1905 (sub_shift_cmp0_noout): Likewise.
1906 (compare_si_ashiftsi): Likewise.
1907 (xbfu_cmp0_noout): New combine pattern.
1908 (xbfu_cmp0"): Likewise.
1909 (movsi_set_cc_insn): Place the predicable variant first.
1910 (commutative_binary_cmp0_noout): Remove clobber.
1911 (commutative_binary_cmp0): New pattern.
1912 (noncommutative_binary_cmp0): Likewise.
1913 (noncommutative_binary_cmp0_noout): Likewise.
1914 (noncommutative_binary_comparison_result_used): Removed.
1915 (rsub_cmp0): New pattern.
1916 (rsub_cmp0_noout): Likewise.
1917 (extzvsi): Changed, keep only meaningful variants.
1918 (SQH, SEZ): New iterators.
1919 (SQH_postfix): New mode attribute.
1920 (SEZ_prefix): New code attribute.
1921 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1922 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1923 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1925 (noncommutative_operator): Check the availability of barrel
1928 2018-04-09 Richard Biener <rguenther@suse.de>
1930 PR tree-optimization/85284
1931 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1932 Only use the niter constraining form of simple_iv when the exit
1935 2018-04-09 Tom de Vries <tom@codesourcery.com>
1938 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1939 (define_expand "*memory_barrier"): New define_expand.
1940 (define_insn "memory_barrier"): New insn.
1942 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1944 PR rtl-optimization/80463
1945 PR rtl-optimization/83972
1946 PR rtl-optimization/83480
1948 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1949 correct producer for the insn.
1950 (tidy_control_flow): Fixup seqnos in case of debug insns.
1952 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1954 PR rtl-optimization/83913
1956 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1957 different sched-times when merging exprs.
1959 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1961 PR rtl-optimization/83962
1963 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1964 tidy_fallthru_edge and tidy_control_flow.
1966 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1968 PR rtl-optimization/83530
1970 * sel-sched.c (force_next_insn): New global variable.
1971 (remove_insn_for_debug): When force_next_insn is true, also leave only
1972 next insn in the ready list.
1973 (sel_sched_region): When the region wasn't scheduled, make another pass
1974 over it with force_next_insn set to 1.
1976 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1978 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1980 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1981 for interrupt control.
1982 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1983 functions for interrupt control.
1984 * config/nds32/nds32-intrinsic.md: Likewise.
1985 * config/nds32/nds32_intrinsic.h: Likewise.
1986 * config/nds32/nds32.h (nds32_builtins): Likewise.
1988 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1990 * config/nds32/nds32.c (nds32_init_machine_status,
1991 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1992 strict_aligned_p field.
1993 (nds32_expand_to_rtl_hook): New function.
1994 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1995 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1997 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1998 Chung-Ju Wu <jasonwucj@gmail.com>
2000 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2001 * config/nds32/nds32-n7.md: New file.
2002 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2003 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2005 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2006 * config/nds32/nds32.md (pipeline_model): Add n7.
2007 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2008 * config/nds32/pipelines.md: Include n7 settings.
2010 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2011 Chung-Ju Wu <jasonwucj@gmail.com>
2013 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2014 * config/nds32/nds32-e8.md: New file.
2015 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2016 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2018 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2019 * config/nds32/nds32.md (pipeline_model): Add e8.
2020 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2021 * config/nds32/pipelines.md: Include e8 settings.
2023 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2024 Chung-Ju Wu <jasonwucj@gmail.com>
2026 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2027 * config/nds32/nds32-n8.md: New file.
2028 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2029 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2031 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2032 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2033 * config/nds32/nds32.md (pipeline_model): Add n8.
2034 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2035 * config/nds32/pipelines.md: Include n8 settings.
2037 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2038 Chung-Ju Wu <jasonwucj@gmail.com>
2040 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2041 * config/nds32/nds32-n9-2r1w.md: New file.
2042 * config/nds32/nds32-n9-3r2w.md: New file.
2043 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2044 nds32_register_ports): New or modify for cpu n9.
2045 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2047 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2048 * config/nds32/nds32-utils.c: New file.
2049 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2050 TARGET_MUL_SLOW): Define.
2051 * config/nds32/nds32.md (pipeline_model): New attribute.
2052 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2053 New options that support cpu n9.
2054 * config/nds32/pipelines.md: Include n9 settings.
2055 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2057 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2059 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2060 information if necessary.
2061 (output_cond_branch_compare_zero): Likewise.
2062 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2063 (nds32_target_alignment): Refine for alignment.
2064 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2065 (FUNCTION_BOUNDARY): Modify.
2066 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2068 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2070 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2072 * config/nds32/constants.md (unspec_volatile_element): Add values for
2073 TLB operation and data prefetch.
2074 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2075 functions for TLB operation and data prefetch.
2076 * config/nds32/nds32-intrinsic.md: Likewise.
2077 * config/nds32/nds32_intrinsic.h: Likewise.
2078 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2079 (nds32_print_operand): Likewise.
2080 * config/nds32/nds32.h (nds32_builtins): Likewise.
2082 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2083 Andrew Pinski <pinsika@gcc.gnu.org>
2086 * match.pd: Use constant_boolean_node of correct type instead of
2087 boolean_true_node or boolean_false_node for simplifying
2088 pointer comparisons to zero.
2090 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2092 PR tree-optimization/80021
2093 * tree.c (verify_type_variant): Make error call in verify_variant_match
2094 translatable and remove final full stop.
2096 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2098 * config/nds32/constants.md (unspec_volatile_element): Add
2099 UNSPEC_VOLATILE_EH_RETURN.
2100 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2101 nds32_output_stack_pop): Support dwarf exception handling process.
2102 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2103 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2104 exception handling process.
2105 (nds32_compute_stack_frame): Likewise.
2106 (nds32_return_addr_rtx): Likewise.
2107 (nds32_initial_elimination_offset): Likewise.
2108 (nds32_expand_prologue): Likewise.
2109 (nds32_expand_epilogue): Likewise.
2110 (nds32_dynamic_chain_address): New function.
2111 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2113 (DYNAMIC_CHAIN_ADDRESS): Define.
2114 (EH_RETURN_DATA_REGNO): Define.
2115 (EH_RETURN_STACKADJ_RTX): Define.
2116 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2117 patterns for dwarf exception handling.
2119 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2121 * config/nds32/nds32.h: Clean up obsolete macros.
2123 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2125 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2126 Add enum values for particular instructions.
2127 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2128 particular intrinsic functions.
2129 * config/nds32/nds32-intrinsic.md: Likewise.
2130 * config/nds32/nds32_intrinsic.h: Likewise.
2131 * config/nds32/nds32.h (nds32_builtins): Likewise.
2132 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2133 (btst, ave): New patterns for particular instructions.
2135 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2137 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2138 Add enum values for atomic load/store and memory sync.
2139 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2141 * config/nds32/nds32-intrinsic.md: Likewise.
2142 * config/nds32/nds32_intrinsic.h: Likewise.
2143 * config/nds32/nds32.h (nds32_builtins): Likewise.
2145 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2147 PR tree-optimization/85257
2148 * fold-const.c (native_encode_vector): If not all elts could fit
2149 and off is -1, return 0 rather than offset.
2150 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2151 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2152 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2153 adjust buffer in native_interpret_expr call.
2155 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2157 * config/nds32/constants.md (unspec_volatile_element): Add cache
2158 control enum values.
2159 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2160 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2161 * config/nds32/nds32.c (nds32_cctl_names): New.
2162 (nds32_print_operand): Handle cache control register names.
2163 * config/nds32/nds32.h (nds32_builtins): New enum values.
2164 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2166 * config/nds32/nds32.md (type): Add mmu.
2167 * config/nds32/pipelines.md (simple_insn): Add mmu.
2169 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2171 * config/nds32/nds32.md (type): Remove call.
2172 * config/nds32/pipelines.md (simple_insn): Likewise.
2174 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2176 * config/nds32/constants.md (unspec_volatile_element): Add
2177 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2178 UNSPEC_VOLATILE_FMFCFG.
2179 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2180 description for fmfcfg and fmfcsr.
2181 (bdesc_1arg): Add fmtcsr.
2182 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2183 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2184 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2185 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2186 unspec_fmfcfg): New patterns.
2187 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2188 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2189 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2190 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2191 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2192 __nds32__fmfcfg): Define.
2194 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2196 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2197 intrinsic register names.
2198 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2199 intrinsic register enum values and macros.
2201 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2203 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2204 for load/store addressing form.
2205 (nds32_print_operand_address): Likewise.
2207 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2210 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2211 based on LABEL_REF. Remove useless assertion.
2212 (pic_address_needs_scratch): Fix formatting.
2213 (sparc_legitimize_pic_address): Minor tweaks.
2214 (sparc_delegitimize_address): Adjust assertion accordingly.
2215 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2216 into symbolic_operand.
2217 (movsi_high_pic_label_ref): Likewise.
2218 (movsi_lo_sum_pic_label_ref): Likewise.
2219 (movdi_pic_label_ref): Likewise.
2220 (movdi_high_pic_label_ref): Likewise.
2221 (movdi_lo_sum_pic_label_ref): Likewise.
2223 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2225 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2226 custom LIB_SPEC setup.
2228 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2229 Kito Cheng <kito.cheng@gmail.com>
2231 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2232 * config/riscv/freebsd.h: New.
2234 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2236 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2237 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2240 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2241 Kito Cheng <kito.cheng@gmail.com>
2243 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2244 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2245 * config/nds32/nds32-protos.h (nds32_output_call,
2246 nds32_output_return): Declare.
2247 * config/nds32/nds32.md: Refine all the call and return patterns.
2249 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2252 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2253 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2255 PR rtl-optimization/84872
2256 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2257 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2260 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2262 * expr.c (copy_blkmode_to_reg): Revert 254862.
2263 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2265 2018-04-06 Richard Biener <rguenther@suse.de>
2268 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2269 after seeing a component reference with an adjacent field. Treat
2270 refs to arrays at struct end of external decls similar to
2271 refs to unconstrained commons.
2273 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2276 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2277 look through SAVE_EXPRs with non-side-effects argument. Adjust
2279 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2282 2018-04-06 Richard Biener <rguenther@suse.de>
2285 * alias.c (find_base_term): New wrapper around find_base_term
2286 unwinding CSELIB_VAL_PTR changes.
2287 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2290 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2292 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2294 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2295 constant definitions.
2296 ("nop"): lr 0,0 -> nopr r0
2297 ("nop_lr0", "nop_lr1"): New insn definitions.
2299 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2301 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2302 NDS32_V3PUSH_AVAILABLE_P macro.
2304 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2305 Chung-Ju Wu <jasonwucj@gmail.com>
2307 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2308 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2309 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2310 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2311 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2312 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2313 * config/nds32/constraints.md: New constraints and checking for hard
2314 float configuration.
2315 * config/nds32/iterators.md: New mode iterator and attribute for hard
2316 float configuration.
2317 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2319 * config/nds32/nds32-fpu.md: New file.
2320 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2321 deal with hard float code generation.
2322 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2324 (abi_type, float_reg_number): New enum type.
2325 * config/nds32/nds32-predicates.c: New predicates for hard float.
2326 * config/nds32/nds32-protos.h: Declare functions for hard float.
2327 * config/nds32/nds32.c: Implementation for hard float configuration.
2328 * config/nds32/nds32.h: Definitions for hard float configuration.
2329 * config/nds32/nds32.md: Include hard float machine description and
2330 modify patterns for hard float configuration.
2331 * config/nds32/nds32.opt: New options for hard float configuration.
2332 * config/nds32/predicates.md: New predicates for hard float
2335 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2337 * common/config/nds32/nds32-common.c
2338 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2340 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2343 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2344 CONSTRUCTOR_ELT (ctor, ...)->value.
2346 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2349 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2351 2018-04-05 Tom de Vries <tom@codesourcery.com>
2354 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2357 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2358 Kito Cheng <kito.cheng@gmail.com>
2360 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2361 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2362 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2363 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2365 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2366 Kito Cheng <kito.cheng@gmail.com>
2368 * config/nds32/constraints.md (Ufe): New memory constraint.
2369 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2370 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2371 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2373 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2374 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2376 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2378 * config/nds32/nds32.md: Use optimize_size in the condition for
2379 alu-shift instructions.
2381 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2383 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2385 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2387 * config/nds32/nds32.md (negsi2): Refine pattern.
2389 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2390 Chung-Ju Wu <jasonwucj@gmail.com>
2392 * config/nds32/iterators.md (shift_rotate): New code iterator.
2393 (shift): New code attribute.
2394 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2395 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2396 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2397 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2398 bit-wise operations.
2399 (andsi3, *andsi3): Ditto.
2400 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2401 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2402 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2403 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2404 nds32_ior_operand, nds32_xor_operand): New predicates.
2406 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2408 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2409 (addsi3, subsi3): ... this.
2411 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2413 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2415 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2417 * config/nds32/nds32.md: Adjust indention.
2419 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2421 * config/nds32/nds32.md (feature): New attribute.
2423 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2425 * config/nds32/nds32.md (subtype): New attribute.
2427 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2430 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2431 expansion to perform a bitwise AND of the argument followed by a
2432 boolean negation of the result.
2434 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2436 PR rtl-optimization/84878
2437 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2438 the basic block. Assert the use reference is not artificial and that
2439 it has an associated insn.
2441 2018-04-04 Michael Matz <matz@suse.de>
2443 * builtins.c (compute_objsize): Pass correct operand
2444 to array_at_struct_end_p.
2446 2018-04-04 Richard Biener <rguenther@suse.de>
2449 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2450 from contexts for DINFO_LEVEL_TERSE and below.
2452 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2454 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2455 resiter_operand condition.
2456 * config/nds32/nds32.md (*move<mode>): Ditto.
2458 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2459 Monk Chiang <sh.chiang04@gmail.com>
2461 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2463 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2465 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2467 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2468 Kito Cheng <kito.cheng@gmail.com>
2470 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2471 nds32_cond_code_str, output_cond_branch,
2472 output_cond_branch_compare_zero, nds32_expand_cbranch,
2473 nds32_expand_cstore, nds32_expand_movcc,
2474 nds32_output_cbranchsi4_equality_zero,
2475 nds32_output_cbranchsi4_equality_reg,
2476 nds32_output_cbranchsi4_equality_reg_or_const_int,
2477 nds32_output_cbranchsi4_greater_less_zero: New functions.
2478 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2479 nds32_expand_cstore, nds32_expand_movcc,
2480 nds32_output_cbranchsi4_equality_zero,
2481 nds32_output_cbranchsi4_equality_reg,
2482 nds32_output_cbranchsi4_equality_reg_or_const_int,
2483 nds32_output_cbranchsi4_greater_less_zero): Declare.
2484 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2485 nds32_rimm11s_operand): New predicates.
2486 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2487 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2490 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2492 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2493 * config/nds32/nds32.md: Ditto.
2494 * config/nds32/pipelines.md: Ditto.
2496 2018-04-04 Richard Biener <rguenther@suse.de>
2498 PR tree-optimization/85168
2499 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2500 propagating abnormals.
2502 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2504 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2506 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2507 Kito Cheng <kito.cheng@gmail.com>
2509 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2510 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2511 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2512 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2513 * config/nds32/nds32.md (sibcall_internal): New.
2514 (sibcall_register): Remove.
2515 (sibcall_immediate): Remove.
2516 (sibcall_value_internal): New.
2517 (sibcall_value_register): Remove.
2518 (sibcall_value_immediate): Remove.
2519 * config/nds32/predicates.md (nds32_general_register_operand): New.
2520 (nds32_call_address_operand): New.
2522 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2524 PR rtl-optimization/85167
2525 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2526 bb_defs if *split_p, instead preinitialize it to NULL.
2528 PR tree-optimization/85156
2529 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2530 evaluating the argument multiple times.
2532 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2534 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2536 (_mm_cvtpd_ps): Likewise.
2537 (_mm_cvttpd_epi32): Likewise.
2538 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2539 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2540 vector, pixel, and bool following altivec.h include.
2542 2018-04-03 Martin Sebor <msebor@redhat.com>
2544 * doc/extend.texi (Common Function Attributes): Clarify.
2545 (const attribute): Likewise.
2546 (pure attribute): Likewise.
2548 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2551 * config/i386/i386.c (ix86_expand_vector_set): Use
2552 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2554 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2556 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2557 instructions when changing rounding bits to preserve precision bits
2558 in the x87 control word.
2560 2018-04-03 Martin Liska <mliska@suse.cz>
2562 PR tree-optimization/82491
2563 * rtl.h (strip_offset_and_add): Replace += suboffset with
2564 poly_uint64 () + suboffset.
2566 2018-03-29 Martin Liska <mliska@suse.cz>
2567 Martin Jambor <mjambor@suse.cz>
2570 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2571 param_type is not an integral or pointer type.
2573 2018-04-03 Richard Biener <rguenther@suse.de>
2575 * sese.h (recompute_all_dominators): Remove.
2577 2018-04-02 Martin Sebor <msebor@redhat.com>
2579 * doc/invoke.texi (-Wrestrict): Fix typos.
2581 2018-04-02 Jim Wilson <jimw@sifive.com>
2583 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2584 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2585 (<optab>di3, <optab>si3_extend): Likewise.
2586 (<optab>si3_mask, <optab>si3_mask_1): New.
2587 (<optab>di3_mask, <optab>di3_mask_1): New.
2588 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2589 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2590 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2592 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2594 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2597 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2599 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2600 (nds32_canonicalize_comparison): New function.
2602 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2603 Kito Cheng <kito.cheng@gmail.com>
2604 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2606 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2607 * config/nds32/constants.md (unspec_volatile_element): Add
2608 UNSPEC_VOLATILE_RELAX_GROUP.
2609 * config/nds32/nds32-relax-opt.c: New file.
2610 * config/nds32/nds32-predicates.c
2611 (nds32_symbol_load_store_p): New function.
2612 * config/nds32/nds32-protos.h
2613 (nds32_symbol_load_store_p): Declare function.
2614 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2615 * config/nds32/nds32.c
2616 (nds32_register_pass): New function to register pass.
2617 (nds32_register_passes): New function to register passes.
2618 * config/nds32/nds32.md (relax_group): New pattern.
2619 * config/nds32/nds32.opt (mrelax-hint): New option.
2620 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2622 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2624 * config/nds32/t-nds32: Modify files dependency.
2626 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2628 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2629 (PROFILE_HOOK): Define its implementation.
2631 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2633 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2634 type and 32-bit size.
2636 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2639 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2640 (V_128_256): New mode iterator.
2641 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2642 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2643 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2645 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2646 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2648 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2651 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2652 NaN inputs correctly.
2654 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2657 * config/rs6000/vsx.md (??r): New mode attribute.
2658 (*vsx_mov<mode>_64bit): Use it.
2659 (*vsx_mov<mode>_32bit): Likewise.
2661 2018-03-30 Martin Sebor <msebor@redhat.com>
2663 PR tree-optimization/84818
2664 * builtins.c (check_access): Use warning_n.
2666 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2669 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2671 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2674 2018-03-30 Julia Koval <julia.koval@intel.com>
2677 * x86-tune.def (movx, partial_reg_dependency): Enable for
2680 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2683 * lra-constraints.c (process_alt_operands): Move setting
2684 this_alternative_matches below.
2686 2018-03-29 Martin Liska <mliska@suse.cz>
2689 * doc/invoke.texi: Document how LTO works with debug info.
2690 Describe auto-load support of binutils. Mention 'x86-64'
2691 as valid option value of -march option.
2693 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2695 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2698 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2699 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2700 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2703 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2706 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2707 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2708 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2709 for RS6000_BTM_POWERPC64.
2710 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2711 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2712 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2717 2018-03-28 Carl Love <cel@us.ibm.com>
2720 2017-09-27 Carl Love <cel@us.ibm.com>
2722 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2723 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2724 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2727 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2729 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2730 instead of __vector bool.
2731 (_mm_max_pu8): Likewise.
2732 (_mm_min_pi16): Likewise.
2734 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2737 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2738 (DIVWEUO): Likewise.
2740 (DIVDEUO): Likewise.
2741 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2742 DIVWEUO and DIVDEUO.
2743 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2744 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2745 (div_extend): Likewise.
2746 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2748 (__builtin_divweuo): Likewise.
2749 (__builtin_divdeo): Likewise.
2750 (__builtin_divdeuo): Likewise.
2752 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2755 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2756 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2758 PR tree-optimization/82004
2759 * gimple-match-head.c (optimize_pow_to_exp): New function.
2760 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2761 Don't fold to exp if optimize_pow_to_exp is false.
2763 2018-03-28 Martin Liska <mliska@suse.cz>
2766 * calls.c (initialize_argument_information): Fix trailing space.
2767 * common.opt: Fix typo and provide better explanation for
2768 -fsanitize-coverage option.
2769 * config/i386/i386.opt: Fix typo.
2771 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2772 Martin Liska <mliska@suse.cz>
2775 * gimplify.c (asan_poison_variable): Don't do the check for
2776 gimplify_omp_ctxp here.
2777 (gimplify_decl_expr): Do it here.
2778 (gimplify_target_expr): Likewise.
2780 2018-03-28 Martin Liska <mliska@suse.cz>
2783 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2784 chkp_type_bounds_count if MPX is not enabled.
2786 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2788 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2790 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2793 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2794 function to create the function decl for complex long double
2795 multiply and divide for -mabi=ieeelongdouble.
2796 (init_float128_ieee): Call it.
2798 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2801 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2802 -fcf-protection=branch -mibt.
2803 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2805 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2808 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2810 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2813 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2814 extern array declarations.
2816 2018-03-27 Richard Biener <rguenther@suse.de>
2819 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2820 explicit single_use checks.
2822 2018-03-27 Richard Biener <rguenther@suse.de>
2824 PR tree-optimization/85082
2825 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2828 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2830 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2831 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2832 Turn on fasynchronous-unwind-tables and funwind-tables.
2834 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2837 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2838 (*bmi_blsr_<mode>_ccz): Ditto.
2840 2018-03-26 Tom de Vries <tom@codesourcery.com>
2842 PR tree-optimization/85063
2843 * omp-general.c (offloading_function_p): New function. Factor out
2845 * omp-offload.c (pass_omp_target_link::gate): ... here.
2846 * omp-general.h (offloading_function_p): Declare.
2847 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2848 with attribute omp declare target for offloading functions.
2850 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2852 PR tree-optimization/84005
2853 * tree-data-ref.h (get_base_for_alignment): Declare.
2854 * tree-data-ref.c (get_base_for_alignment_1): New function.
2855 (get_base_for_alignment): Likewise.
2856 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2857 get_base_for_alignment to find a suitable base object, instead
2858 of always using drb->base_address.
2860 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2863 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2864 known size by default.
2866 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2869 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2870 and non BLKmode operands.
2872 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2875 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2876 Clean up attributes.
2878 2018-03-23 Richard Biener <rguenther@suse.de>
2881 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2882 we are going to emit early debug for LTO.
2884 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2887 * function.c (match_asm_constraints_1): Don't optimize if input
2888 doesn't satisfy general_operand predicate for output's mode.
2891 * alias.c (write_dependence_p): Don't require for x_canonicalized
2892 non-VOIDmode if x has VOIDmode.
2895 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2896 just don't try to optimize it rather than assert it never happens.
2898 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2900 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2901 macro expansions for definition of ST_INTERNAL_<mode> and
2902 LD_INTERNAL_<mode> builtins.
2903 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2905 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2907 (altivec_expand_st_builtin): Likewise.
2908 (altivec_expand_builtin): Remove calls to deleted functions.
2909 (rs6000_address_for_altivec): Delete this function.
2910 * config/rs6000/vector.md: Remove expands for
2911 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2913 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2916 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2917 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2918 re-computing once computed.
2919 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2920 (arm_init_machine_status): Initialize
2921 machine->static_chain_stack_bytes.
2923 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2926 * doc/extend.texi: Add four new prototypes for vec_ld.
2927 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2928 definitions for more logical presentation.
2929 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2930 entries for V1TI variants of __builtin_altivec_ld builtin.
2931 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2932 handling of V1TI variant of LVX icode pattern.
2933 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2934 (rs6000_gimple_fold_builtin): Likewise.
2935 (altivec_init_builtins): Add code to define
2936 __builtin_altivec_lvx_v1ti function.
2938 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2941 * function.c (match_asm_constraints_1): Don't do the optimization
2942 if input isn't a REG, SUBREG, MEM or constant.
2944 2018-03-22 Tom de Vries <tom@codesourcery.com>
2946 PR tree-optimization/84956
2947 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2948 bb_has_abnormal_pred.
2950 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2953 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2954 DECL_INITIAL (decl) to decl at the end.
2955 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2958 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2960 * doc/extend.texi (__builtin_tgmath): Document when complex
2961 integer types are treated as _Complex _Float64.
2963 2018-03-21 Tom de Vries <tom@codesourcery.com>
2965 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2967 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2969 PR tree-optimization/84960
2970 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2971 if it is ENTRY block, move them into single succ of ENTRY in that case.
2973 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2975 PR tree-optimization/84811
2976 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2977 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2978 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2979 (known_subrange_p): Update accordingly. Cast each value involved
2980 in the size comparison, rather than casting the result of the
2983 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2985 PR tree-optimization/84982
2986 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2987 by flipping the least significant bit rather than all bits from
2988 bitpos to bitpos + bitsize - 1.
2990 2018-03-21 Nathan Sidwell <nathan@acm.org>
2992 * doc/extend.texi (Deprecated Features): Remove mention of
2993 long-deleted deprecations.
2995 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2998 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2999 * configure: Regenerate.
3001 2018-03-21 Tom de Vries <tom@codesourcery.com>
3003 PR tree-optimization/83126
3004 * tree-parloops.c (num_phis): New function.
3005 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3007 2018-03-21 Nathan Sidwell <nathan@acm.org>
3009 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3010 mention anon-struct/union members and trailing attributes.
3012 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3014 PR tree-optimization/84969
3015 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3016 builtin memset partitions if they set different rhs values.
3018 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3020 PR rtl-optimization/84989
3021 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3022 VEC_DUPLICATE with scalar result mode.
3024 2018-03-21 Martin Liska <mliska@suse.cz>
3027 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3028 not intended return statement.
3030 2018-03-21 Martin Liska <mliska@suse.cz>
3033 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3034 (chkp_find_bound_slots_1): Limit number of iterations.
3036 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3039 * Minor grammar fixes for x86 options.
3041 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3044 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3045 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3047 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3050 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3051 (altivec_lvx_<mode>_1op): Likewise.
3052 (altivec_stvx_<mode>_2op): Likewise.
3053 (altivec_stvx_<mode>_1op): Likewise.
3054 (altivec_lvx_<VM2:mode>): New define_expand.
3055 (altivec_stvx_<VM2:mode>): Likewise.
3056 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3057 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3058 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3059 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3060 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3061 (rs6000_gen_lvx): Likewise.
3062 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3063 (altivec_expand_stv_builtin): Likewise.
3064 (altivec_expand_builtin): Likewise.
3065 * config/rs6000/vector.md: Likewise.
3067 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3070 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3073 2018-03-20 Richard Biener <rguenther@suse.de>
3076 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3077 sign-conversions as zero, fall back to standard scalar_stmt
3080 2018-03-20 Martin Liska <mliska@suse.cz>
3083 * predict.c (rebuild_frequencies): Handle case when we have
3084 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3086 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3089 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3090 flag_section_anchors.
3091 * varasm.c (use_blocks_for_decl_p): Remove hack for
3092 dw2_force_const_mem.
3095 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3097 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3098 be created, use lowpart_subreg of operands[0] rather than operands[0]
3100 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3101 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3102 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3103 and n constraint instead of aarch64_shift_imm_di and Usd.
3104 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3105 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3107 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3110 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3111 to favor GPR over NEON registers.
3112 (<shift>di3_neon): Likewise.
3114 2018-03-20 Tom de Vries <tom@codesourcery.com>
3117 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3118 (nvptx_process_pars): Emit bar.sync asap and alap.
3120 2018-03-20 Tom de Vries <tom@codesourcery.com>
3123 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3124 seen_label if seen_label is already set.
3126 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3129 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3130 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3131 Use 1U instead of 1. Formatting fixes.
3134 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3135 instead of TREE_TYPE (s1) for the return value.
3137 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3139 PR tree-optimization/84946
3140 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3141 bitsize + bitsize in poly_uint64 rather than poly_int64.
3144 * dwarf2asm.c: Include fold-const.c.
3145 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3146 of decl rather than decl itself.
3148 PR rtl-optimization/84643
3149 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3151 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3154 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3155 calling assemble_variable.
3157 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3160 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3161 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3163 2018-03-19 Jim Wilson <jimw@sifive.com>
3166 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3167 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3168 (riscv_first_stack_step): Likewise.
3169 (riscv_option_override): Use STACK_BOUNDARY instead of
3171 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3173 (BIGGEST_ALIGNMENT): Set to 128.
3174 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3175 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3178 2018-03-19 Richard Biener <rguenther@suse.de>
3180 PR tree-optimization/84933
3181 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3182 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3184 2018-03-19 Richard Biener <rguenther@suse.de>
3186 PR tree-optimization/84859
3187 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3188 (cond_if_else_store_replacement): Perform sinking operation on
3189 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3190 Generalize what a BB with a single eligible store is.
3192 2018-03-19 Richard Biener <rguenther@suse.de>
3194 PR tree-optimization/84929
3195 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3196 chrec_is_positive against non-chrec arg.
3198 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3201 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3203 2018-03-18 Martin Liska <mliska@suse.cz>
3205 PR rtl-optimization/84635
3206 * regrename.c (build_def_use): Use matches_mode only when
3209 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3211 PR tree-optimization/84913
3212 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3213 vectorize chains of COND_EXPRs.
3215 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3217 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3219 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3221 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3223 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3225 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3227 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3228 Kito Cheng <kito.cheng@gmail.com>
3230 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3231 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3232 (nds32_adjust_reg_alloc_order): New function.
3233 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3235 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3237 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3238 nds32_print_operand, nds32_print_operand_address): Use
3239 HOST_WIDE_INT_PRINT_DEC instead.
3241 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3243 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3245 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3248 * config/i386/i386.c (initial_ix86_tune_features,
3249 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3251 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3252 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3253 rather than 1u << ix86_tune. Formatting fix.
3254 (ix86_option_override_internal): Change ix86_arch_mask from
3255 unsigned int to unsigned HOST_WIDE_INT, initialize to
3256 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3257 (ix86_function_specific_restore): Likewise.
3259 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3262 * postreload.c (reload_combine_recognize_pattern): Perform
3263 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3264 truncate_int_for_mode the result for the destination's mode.
3267 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3268 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3270 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3273 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3276 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3277 regno_allocno_class_array and sorted_pseudos.
3278 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3279 insns where regno is used.
3281 2018-03-16 Martin Liska <mliska@suse.cz>
3284 * multiple_target.c (create_dispatcher_calls): Redirect
3285 reference in the symbol table.
3287 2018-03-16 Martin Liska <mliska@suse.cz>
3290 * multiple_target.c (create_dispatcher_calls): Redirect also
3293 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3297 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3298 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3299 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3301 2018-03-16 Julia Koval <julia.koval@intel.com>
3303 * doc/invoke.texi (Skylake Server): Add CLWB.
3304 Cannonlake): Remove CLWB.
3306 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3308 PR tree-optimization/84841
3309 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3311 (FLOAT_ONE_CONST_TYPE): Define.
3312 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3313 (sort_by_operand_rank): Put entries with higher constant_type last
3314 rather than first to match comments.
3316 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3318 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3321 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3324 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3325 check and use address of target always.
3327 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3330 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3331 (indirect_thunk_bnd_needed): Likewise.
3332 (indirect_thunks_used): Likewise.
3333 (indirect_thunks_bnd_used): Likewise.
3334 (indirect_return_needed): New.
3335 (indirect_return_bnd_needed): Likewise.
3336 (output_indirect_thunk_function): Add a bool argument for
3338 (output_indirect_thunk_function): Don't generate alias for
3339 function return thunk.
3340 (ix86_code_end): Call output_indirect_thunk_function to generate
3341 function return thunks.
3342 (ix86_output_function_return): Set indirect_return_bnd_needed
3343 and indirect_return_needed instead of indirect_thunk_bnd_needed
3344 and indirect_thunk_needed.
3346 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3348 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3349 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3350 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3352 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3353 Paul Hua <paul.hua.gm@gmail.com>
3356 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3358 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3360 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3361 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3364 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3367 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3368 instead of GET_MODE_SIZE when comparing Units.
3370 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3373 * varasm.c (hash_section): Return an unchangeble hash value
3374 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3375 Return !aarch64_can_use_per_function_literal_pools_p ().
3377 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3380 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3381 rather than address of cmode as last argument to prepare_cmp_insn.
3383 2018-03-15 Julia Koval <julia.koval@intel.com>
3385 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3386 F_AVX512VNNI, F_AVX512BITALG): New.
3388 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3391 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3392 insn for floating-point loads and stores.
3394 2018-03-14 Carl Love <cel@us.ibm.com>
3396 * config/rs6000/rs6000-c.c: Add macro definitions for
3397 ALTIVEC_BUILTIN_VEC_PERMXOR.
3398 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3399 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3400 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3401 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3403 * config/doc/extend.texi: Add prototypes for vec_permxor.
3405 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3408 * diagnostic-show-locus.c (class layout_point): Convert m_line
3409 from int to linenum_type.
3410 (line_span::comparator): Use linenum "compare" function when
3411 comparing line numbers.
3412 (test_line_span): New function.
3413 (layout_range::contains_point): Convert param "row" from int to
3415 (layout_range::intersects_line_p): Likewise.
3416 (layout::will_show_line_p): Likewise.
3417 (layout::print_source_line): Likewise.
3418 (layout::should_print_annotation_line_p): Likewise.
3419 (layout::print_annotation_line): Likewise.
3420 (layout::print_leading_fixits): Likewise.
3421 (layout::annotation_line_showed_range_p): Likewise.
3422 (struct line_corrections): Likewise for field m_row.
3423 (line_corrections::line_corrections): Likewise for param "row".
3424 (layout::print_trailing_fixits): Likewise.
3425 (layout::get_state_at_point): Likewise.
3426 (layout::get_x_bound_for_row): Likewise.
3427 (layout::print_line): Likewise.
3428 (diagnostic_show_locus): Likewise for locals "last_line" and
3430 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3431 * input.c (selftest::test_linenum_comparisons): New function.
3432 (selftest::input_c_tests): Call it.
3433 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3434 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3435 * selftest.h (ASSERT_GT): New macro.
3436 (ASSERT_GT_AT): New macro.
3437 (ASSERT_LT): New macro.
3438 (ASSERT_LT_AT): New macro.
3440 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3442 PR rtl-optimization/84780
3443 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3445 2018-03-14 Martin Liska <mliska@suse.cz>
3447 * tree.c (record_node_allocation_statistics): Use
3448 get_stats_node_kind.
3449 (get_stats_node_kind): New function extracted from
3450 record_node_allocation_statistics.
3451 (free_node): Use get_stats_node_kind.
3453 2018-03-14 Richard Biener <rguenther@suse.de>
3455 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3456 that the value-set of ANTIC_IN doesn't grow.
3459 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3461 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3462 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3464 2018-03-14 Julia Koval <julia.koval@intel.com>
3466 * config.gcc (icelake-client, icelake-server): New.
3468 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3469 (initial_ix86_arch_features): Ditto.
3470 (PTA_SKYLAKE): Add SGX.
3471 (PTA_ICELAKE): Remove.
3472 (PTA_ICELAKE_CLIENT): New.
3473 (PTA_ICELAKE_SERVER): New.
3474 (ix86_option_override_internal): Split up icelake on icelake client and
3476 (get_builtin_code_for_version): Ditto.
3477 (fold_builtin_cpu): Ditto.
3478 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3479 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3480 * config/i386/i386.h (processor_type): Ditto.
3481 * doc/invoke.texi: Ditto.
3483 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3486 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3487 INTEGER_CST offset, add it together with bitpos / 8 and
3488 sign extend based on POINTER_SIZE.
3492 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3495 * config/i386/constraints.md (Yc): New register constraint.
3496 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3497 Use Yc constraint for alternative 2 of operand 0. Remove
3498 preferred_for_speed attribute.
3500 2018-03-14 Richard Biener <rguenther@suse.de>
3502 PR tree-optimization/84830
3503 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3504 with the old one to avoid oscillations.
3506 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3509 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3511 (assign_by_spills): Return a flag of reload assignment failure.
3512 Do not process the reload assignment failures. Do not spill other
3513 reload pseudos if they has the same reg class. Update n if
3515 (lra_assign): Add a return arg. Set up from the result of
3516 assign_by_spills call.
3517 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3518 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3519 usage_insns if it is not NULL.
3520 (spill_hard_reg_in_range): New function.
3521 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3522 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3523 function prototypes.
3524 (lra_assign): Change prototype.
3525 * lra.c (lra): Add code to deal with fails by splitting hard reg
3528 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3530 * config/riscv/riscv.opt (mrelax): New option.
3531 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3532 "norelax" when riscv_mrelax is disabled.
3533 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3535 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3538 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3539 reassociation for int modes.
3541 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3543 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3544 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3546 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3547 * config/aarch64/aarch64-sve.md
3548 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3549 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3550 (*extend<mode><Vwide>2): Rename to...
3551 (aarch64_sve_extend<mode><Vwide>2): ...this.
3552 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3553 renaming the old pattern to...
3554 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3556 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3557 define_expand, renaming the old pattern to...
3558 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3559 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3560 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3561 account when deciding which SVE instruction the optab should use.
3562 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3564 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3566 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3567 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3568 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3569 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3570 (tlsdesc_small_<mode>): Turn a define_expand and use
3571 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3572 (tlsdesc_small_advsimd_<mode>): ...this.
3573 (tlsdesc_small_sve_<mode>): New pattern.
3575 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3577 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3578 (UNSPEC_UMUL_HIGHPART): New constants.
3579 (MUL_HIGHPART): New int iteraor.
3580 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3581 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3583 (*<su>mul<mode>3_highpart): New define_insn.
3585 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3588 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3591 2018-03-13 Martin Liska <mliska@suse.cz>
3594 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3596 (sem_item_optimizer::~sem_item_optimizer): Release it.
3597 (sem_item_optimizer::merge_classes): Register variable aliases.
3598 (sem_item_optimizer::fixup_pt_set): New function.
3599 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3600 * ipa-icf.h: Declare new variables and functions.
3602 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3605 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3606 integer_pow2p@2 and test integer_pow2p in condition.
3607 (A < 0 ? C : 0): Similarly for @1.
3610 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3611 characters starting at p contain '\0' character, don't look beyond
3615 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3616 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3619 * reg-stack.c (change_stack): Change update_end var from int to
3620 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3621 also call set_block_for_insn on the newly added insns and rescan.
3624 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3625 on the last operand.
3628 * tree.c (stabilize_reference_1): Return save_expr (e) for
3629 STATEMENT_LIST even if it doesn't have side-effects.
3631 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3633 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3635 2018-03-12 Renlin Li <renlin.li@arm.com>
3637 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3638 aarch64_output_scalar_simd_mov_immediate.
3640 2018-03-12 Martin Sebor <msebor@redhat.com>
3642 PR tree-optimization/83456
3643 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3644 for perfectly overlapping calls to memcpy.
3645 (gimple_fold_builtin_memory_chk): Same.
3646 (gimple_fold_builtin_strcpy): Handle no-warning.
3647 (gimple_fold_builtin_stxcpy_chk): Same.
3648 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3650 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3652 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3653 parameter. Use it for SFmode.
3654 (rs6000_function_arg_advance_1): Adjust.
3655 (rs6000_function_arg): Adjust.
3656 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3658 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3660 PR rtl-optimization/84169
3661 PR rtl-optimization/84780
3662 * combine.c (can_combine_p): Check for a 2-insn combination whether
3663 the destination register is used between the two insns, too.
3665 2018-03-12 Richard Biener <rguenther@suse.de>
3667 PR tree-optimization/84803
3668 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3669 for refs DR analysis didn't process.
3671 2018-03-12 Richard Biener <rguenther@suse.de>
3673 PR tree-optimization/84777
3674 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3675 force-vectorize loops ignore whether we are optimizing for size.
3677 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3679 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3680 (TARGET_MD_ASM_ADJUST): Define.
3682 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3683 Kito Cheng <kito.cheng@gmail.com>
3684 Chung-Ju Wu <jasonwucj@gmail.com>
3686 * config/nds32/nds32.c (nds32_compute_stack_frame,
3687 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3688 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3689 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3690 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3691 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3692 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3693 * config/nds32/nds32.md (prologue, epilogue): Use macro
3694 NDS32_V3PUSH_AVAILABLE_P to do checking.
3696 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3699 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3700 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3701 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3702 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3703 addition of most attributes on !orig_type_die or the attribute not
3704 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3706 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3707 Chung-Ju Wu <jasonwucj@gmail.com>
3709 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3711 * config/nds32/nds32.opt (mvh): New option.
3713 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3714 Chung-Ju Wu <jasonwucj@gmail.com>
3716 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3718 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3719 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3722 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3723 Chung-Ju Wu <jasonwucj@gmail.com>
3725 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3727 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3728 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3730 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3731 Kito Cheng <kito.cheng@gmail.com>
3732 Chung-Ju Wu <jasonwucj@gmail.com>
3734 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3735 UNSPEC_FFMISM and UNSPEC_FLMISM.
3736 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3737 for ffb, ffmism and flmism.
3738 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3739 (unspec_ffmism): Ditto.
3740 (unspec_flmism): Ditto.
3741 (nds32_expand_builtin_impl): Check if string extension is available.
3742 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3743 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3745 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3748 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3751 * lra-assigns.c (assign_by_spills): Return a flag of reload
3752 assignment failure. Do not process the reload assignment
3753 failures. Do not spill other reload pseudos if they has the same
3755 (lra_assign): Add a return arg. Set up from the result of
3756 assign_by_spills call.
3757 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3758 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3759 usage_insns if it is not NULL.
3760 (spill_hard_reg_in_range): New function.
3761 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3762 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3763 function prototypes.
3764 (lra_assign): Change prototype.
3765 * lra.c (lra): Add code to deal with fails by splitting hard reg
3768 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3771 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3773 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3776 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3777 (dw_val_node): Add val_symbolic_view.
3778 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3779 (symview_upper_bound): New.
3780 (new_line_info_table): Initialize symviews_since_reset.
3781 (dwarf2out_source_line): Count symviews_since_reset and set
3782 symview_upper_bound.
3783 (dw_val_equal_p): Handle symview.
3784 (add_AT_symview): New.
3785 (print_dw_val): Handle symview.
3786 (attr_checksum, attr_checksum_ordered): Likewise.
3787 (same_dw_val_p, size_of_die): Likewise.
3788 (value_format, output_die): Likewise.
3789 (add_high_low_attributes): Use add_AT_symview for entry_view.
3790 (dwarf2out_finish): Reset symview_upper_bound, clear
3793 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3796 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3797 Add strict argument and use it.
3798 (rs6000_split_multireg_move): Update for new strict argument.
3799 (mem_operand_gpr): Disallow all non-offsettable addresses.
3800 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3802 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3805 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3806 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3807 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3810 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3811 decl, use remap_type if we want to use the type.
3813 2018-03-09 Martin Sebor <msebor@redhat.com>
3815 PR tree-optimization/84526
3816 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3818 (builtin_access::generic_overlap): Be prepared to handle non-array
3821 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3823 PR rtl-optimization/84682
3824 * lra-constraints.c (process_address_1): Check is_address flag
3825 for address constraints.
3826 (process_alt_operands): Likewise.
3827 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3828 preprocess_constraints.
3829 * recog.h (preprocess_constraints): Add oploc parameter.
3831 * recog.c (preprocess_constraints): Test address_operand for
3832 CT_ADDRESS constraints.
3834 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3837 * lra-assigns.c (assign_by_spills): Return a flag of reload
3838 assignment failure. Do not process the reload assignment
3839 failures. Do not spill other reload pseudos if they has the same
3841 (lra_assign): Add a return arg. Set up from the result of
3842 assign_by_spills call.
3843 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3844 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3845 usage_insns if it is not NULL.
3846 (spill_hard_reg_in_range): New function.
3847 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3848 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3849 function prototypes.
3850 (lra_assign): Change prototype.
3851 * lra.c (lra): Add code to deal with fails by splitting hard reg
3854 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3857 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3858 Accept complain bool parameter. Only emit errors if it is true.
3859 (arm_parse_cpu_option_name): Likewise.
3860 (arm_target_thumb_only): Adjust callers of the above.
3861 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3862 prototype to take a default true bool parameter.
3863 (arm_parse_arch_option_name): Likewise.
3865 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3866 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3870 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3871 * configure: Regenerate.
3872 * configure.ac ("linker --version-script option"): New.
3873 ("linker soname option"): New.
3875 2018-03-09 Richard Biener <rguenther@suse.de>
3877 PR tree-optimization/84775
3878 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3879 immediate uses of predicate stmts and mark them modified.
3882 PR tree-optimization/84178
3883 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3885 (version_loop_for_if_conversion): Delay update_ssa call.
3886 (tree_if_conversion): Delay update_ssa until after predicate
3889 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3892 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3893 when the function accesses prior frames.
3895 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3898 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3899 gen_llsym, otherwise call maybe_gen_llsym.
3902 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3903 has ',' character inside of it.
3905 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3908 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3909 as clobbering CC_REGNUM.
3911 2018-03-08 Richard Biener <rguenther@suse.de>
3914 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3915 (follow_copies_to_constant): Do not follow SSA names registered
3918 2018-03-08 Richard Biener <rguenther@suse.de>
3920 PR tree-optimization/84178
3921 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3923 (version_loop_for_if_conversion): Delay update_ssa call.
3924 (tree_if_conversion): Delay update_ssa until after predicate
3927 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3929 PR tree-optimization/84178
3930 * tree-if-conv.c (release_bb_predicate): Remove the
3931 the assertion that the stmts have NULL use_ops.
3932 Discard the statements, asserting that they haven't
3933 yet been added to a BB.
3935 2018-03-08 Richard Biener <rguenther@suse.de>
3937 PR tree-optimization/84746
3938 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3939 (phi_translate): Pass in destination ANTIC_OUT set.
3940 (phi_translate_1): Likewise. For a simplified result lookup
3941 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3942 (phi_translate_set): Adjust.
3943 (do_pre_regular_insertion): Likewise.
3944 (do_pre_partial_partial_insertion): Likewise.
3946 2018-03-08 Martin Liska <mliska@suse.cz>
3948 PR gcov-profile/84735
3949 * doc/gcov.texi: Document usage of profile files.
3950 * gcov-io.h: Document changes in the format.
3952 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3956 * dwarf2out.c (struct dw_line_info_table): Update comments for
3958 (FORCE_RESET_NEXT_VIEW): New.
3959 (FORCE_RESETTING_VIEW_P): New.
3960 (RESETTING_VIEW_P): Check for -1 too.
3961 (ZERO_VIEW_P): Likewise.
3962 (new_line_info_table): Force-reset next view.
3963 (dwarf2out_begin_function): Likewise.
3964 (dwarf2out_source_line): Simplify zero_view_p initialization.
3965 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3966 view directly. Omit view when omitting .loc at line 0.
3968 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3970 PR tree-optimization/84740
3971 * tree-switch-conversion.c (process_switch): Call build_constructors
3972 only if info.phi_count is non-zero.
3974 PR tree-optimization/84739
3975 * tree-tailcall.c (find_tail_calls): Check call arguments against
3976 DECL_ARGUMENTS (current_function_decl) rather than
3977 DECL_ARGUMENTS (func) when checking for tail recursion.
3979 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3981 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3982 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3983 Volker Reichelt's entry and add entries for people that perform
3984 GCC fuzzy testing and report numerous bugs.
3986 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3989 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3990 readonly data in sdata, if that is disabled.
3991 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3992 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3993 -mreadonly-in-sdata option.
3995 2018-03-07 Martin Sebor <msebor@redhat.com>
3997 PR tree-optimization/84468
3998 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3999 basic block when looking for nul assignment.
4001 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4004 * except.h (output_function_exception_table): Adjust prototype.
4005 * except.c (output_function_exception_table): Remove FNNAME parameter
4006 and add SECTION parameter. Ouput one part of the table at a time.
4007 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4008 the first part of the exception table and emit unwind directives.
4009 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4010 (i386_pe_seh_cold_init): Likewise.
4011 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4012 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4013 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4014 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4015 * config/i386/winnt.c: Include except.h.
4016 (struct seh_frame_state): Add reg_offset, after_prologue and
4017 in_cold_section fields.
4018 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4019 (i386_pe_seh_cold_init): New function.
4020 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4021 to seh->in_cold_section.
4022 (seh_emit_push): Record the offset of the push.
4023 (seh_emit_save): Record the offet of the save.
4024 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4025 Test seh->after_prologue to disregard the epilogue.
4026 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4027 (i386_pe_end_cold_function): New function.
4029 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4032 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4033 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4036 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4038 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4039 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4042 * multiple_target.c: Include tree-inline.h and intl.h.
4043 (expand_target_clones): Diagnose and fail if node->definition and
4044 !tree_versionable_function_p (node->decl).
4046 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4048 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4050 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4051 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4052 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4054 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4057 * combine.c (try_combine): Use reg_or_subregno instead of handling
4058 just paradoxical SUBREGs and REGs.
4060 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4062 * config/arc/arc.c (arc_finalize_pic): Remove function.
4063 (arc_must_save_register): We use single base PIC register, remove
4064 checks to save/restore the PIC register.
4065 (arc_expand_prologue): Likewise.
4066 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4068 (arc_verify_short): Likewise.
4069 (arc_attr_type): Likewise.
4070 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4071 (walk_stores): Likewise.
4072 (arc_address_cost): Make it static.
4073 (arc_verify_short): Likewise.
4074 (branch_dest): Likewise.
4075 (arc_attr_type): Likewise.
4076 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4077 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4078 (arc_final_prescan_insn): Remove inserting the nops due to
4079 hardware hazards. It is done in reorg step.
4080 (insn_length_variant_t): Remove.
4081 (insn_length_parameters_t): Likewise.
4082 (arc_insn_length_parameters): Likewise.
4083 (arc_get_insn_variants): Likewise.
4084 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4086 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4089 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4092 PR tree-optimization/84687
4093 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4095 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4097 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4099 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4100 Rename to ppc_speculation_barrier.
4101 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4102 __builtin_ppc_speculation_barrier.
4104 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4107 * combine.c (combine_simplify_rtx): Don't try to simplify if
4108 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4111 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4113 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4114 to 32 bytes when compiling for POWER9.
4116 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4119 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4120 regparm >= 3 with no arg reg available also for calls with
4121 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4124 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4126 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4128 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4131 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4133 2018-03-05 Richard Biener <rguenther@suse.de>
4135 PR tree-optimization/84486
4136 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4137 When inserting a __builtin_assume_aligned call set the LHS
4138 SSA name alignment info accordingly.
4140 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4142 PR tree-optimization/84114
4143 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4144 Avoid reassociation of FLOAT_MODE addition.
4146 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4148 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4149 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4150 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4151 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4152 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4153 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4154 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4156 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4157 __builtin_ia32_wbinvd): New builtins.
4158 (SPECIAL_ARGS2): New.
4159 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4160 (SPECIAL_ARGS2): New.
4161 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4162 (ix86_valid_target_attribute_inner_p): Ditto.
4163 (ix86_init_mmx_sse_builtins): Add special_args2.
4164 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4165 TARGET_WBNOINVD_P): New.
4166 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4167 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4168 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4169 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4170 * config/i386/pconfigintrin.h: New file.
4171 * config/i386/wbnoinvdintrin.h: Ditto.
4172 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4173 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4175 2018-03-05 Richard Biener <rguenther@suse.de>
4177 PR tree-optimization/84670
4178 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4180 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4181 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4182 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4183 doesn't grow if all successors (recursively) were visited at least
4186 2018-03-05 Richard Biener <rguenther@suse.de>
4188 PR tree-optimization/84650
4189 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4190 if executed in the loop pipeline.
4192 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4194 * doc/configfiles.texi (Configuration Files): Move info about
4195 conditionalizing $target-protos.h to...
4196 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4197 differs from $target-protos.h.
4199 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4200 Chung-Ju Wu <jasonwucj@gmail.com>
4202 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4203 * config/nds32/nds32-multiple.md (setmemsi): Define.
4204 * config/nds32/nds32-memory-manipulation.c
4205 (nds32_gen_dup_4_byte_to_word_value): New.
4206 (emit_setmem_word_loop): New.
4207 (emit_setmem_byte_loop): New.
4208 (nds32_expand_setmem_loop): New.
4209 (nds32_expand_setmem_loop_v3m): New.
4210 (nds32_expand_setmem_unroll): New.
4211 (nds32_expand_setmem): New.
4213 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4214 Chung-Ju Wu <jasonwucj@gmail.com>
4216 * config/nds32/nds32-memory-manipulation.c
4217 (nds32_emit_load_store): New.
4218 (nds32_emit_post_inc_load_store): New.
4219 (nds32_emit_mem_move): New.
4220 (nds32_emit_mem_move_block): New.
4221 (nds32_expand_movmemsi_loop_unknown_size): New.
4222 (nds32_expand_movmemsi_loop_known_size): New.
4223 (nds32_expand_movmemsi_loop): New.
4224 (nds32_expand_movmemsi_unroll): New.
4225 (nds32_expand_movmemqi): Rename ...
4226 (nds32_expand_movmemsi): ... to this.
4227 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4228 (movmemsi): ... to this.
4229 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4230 (nds32_expand_movmemsi): ... to this.
4232 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4233 Monk Chiang <sh.chiang04@gmail.com>
4234 Chung-Ju Wu <jasonwucj@gmail.com>
4236 * config/nds32/nds32-protos.h
4237 (nds32_expand_load_multiple): New arguments.
4238 (nds32_expand_store_multiple): Ditto.
4239 (nds32_valid_multiple_load_store): Rename ...
4240 (nds32_valid_multiple_load_store_p): ... to this.
4241 * config/nds32/nds32-memory-manipulation.c
4242 (nds32_expand_load_multiple): Refine implementation.
4243 (nds32_expand_store_multiple): Ditto.
4244 * config/nds32/nds32-multiple.md
4245 (load_multiple): Update nds32_expand_load_multiple interface.
4246 (store_multiple): Update nds32_expand_store_multiple interface.
4247 * config/nds32/nds32-predicates.c
4248 (nds32_valid_multiple_load_store): Rename ...
4249 (nds32_valid_multiple_load_store_p): ... to this and refine
4251 * config/nds32/predicates.md
4252 (nds32_load_multiple_and_update_address_operation): New predicate.
4253 (nds32_store_multiple_and_update_address_operation): New predicate.
4255 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4256 Chung-Ju Wu <jasonwucj@gmail.com>
4258 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4259 (combo): New attribute.
4260 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4262 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4264 * config/nds32/nds32.opt: Change -mcmodel= default value.
4266 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4267 Monk Chiang <sh.chiang04@gmail.com>
4268 Chung-Ju Wu <jasonwucj@gmail.com>
4270 * config/nds32/constants.md (unspec_element): New enum.
4271 * config/nds32/constraints.md (Umw): New constraint.
4272 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4273 * config/nds32/nds32-intrinsic.md: Likewise.
4274 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4275 (nds32_valid_smw_lwm_base_p): New.
4276 (nds32_output_smw_single_word): New.
4277 (nds32_output_lmw_single_word): New.
4278 (nds32_expand_unaligned_load): New.
4279 (nds32_expand_unaligned_store): New.
4280 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4281 (nds32_output_smw_single_word): Declare.
4282 (nds32_output_lmw_single_word): Declare.
4283 (nds32_expand_unaligned_load): Declare.
4284 (nds32_expand_unaligned_store): Declare.
4285 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4286 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4287 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4288 NDS32_BUILTIN_UASTORE_DW.
4289 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4292 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4293 Kito Cheng <kito.cheng@gmail.com>
4294 Chung-Ju Wu <jasonwucj@gmail.com>
4296 * config/nds32/nds32-intrinsic.c
4297 (nds32_expand_builtin_null_ftype_reg): Delete.
4298 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4299 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4300 (nds32_read_argument): New.
4301 (nds32_legitimize_target): Ditto.
4302 (nds32_legitimize_argument): Ditto.
4303 (nds32_check_constant_argument): Ditto.
4304 (nds32_expand_unop_builtin): Ditto.
4305 (nds32_expand_unopimm_builtin): Ditto.
4306 (nds32_expand_binop_builtin): Ditto.
4307 (nds32_builtin_decl_impl): Ditto.
4308 (builtin_description): Ditto.
4309 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4310 (nds32_init_builtins_impl): Ditto.
4311 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4312 (nds32_builtin_decl): New.
4313 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4314 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4316 2018-03-02 Jeff Law <law@redhat.com>
4318 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4319 (redundant_insn, fill_simple_delay_slots): Likewise.
4320 (fill_slots_from_thread): Likewise.
4321 * resource.c (mark_referenced_resources): Likewise.
4322 (mark_set_resources, find_dead_or_set_registers): Likewise.
4324 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4326 * substring-locations.h (format_warning_va): Formatting fix for
4328 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4330 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4331 * substring-locations.c: Include intl.h.
4332 (format_warning_va): Turned into small wrapper around
4333 format_warning_n_va, renamed to ...
4334 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4335 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4337 (format_warning_at_substring_n): New function.
4338 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4339 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4340 format_warning_at_substring with just a shorter name instead of
4341 const function pointer.
4342 (fmtwarn_n): New function.
4343 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4344 appropriate, get rid of all the fmtstr temporaries, move conditionals
4345 with G_() wrapped string literals directly into fmtwarn arguments,
4346 cast dir.len to (int), formatting fixes.
4348 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4350 * doc/invoke.texi: Remove "Cilk Plus" references.
4352 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4353 Richard Biener <rguenther@suse.de>
4356 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4357 for error or warning attributes if CALL_FROM_THUNK_P is set.
4360 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4363 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4364 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4367 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4368 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4370 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4371 instead of -1U in last predictors element's probability member.
4373 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4376 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4377 arguments if they are comparable.
4379 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4381 PR tree-optimization/84634
4382 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4383 masks and masked_loop_p with a single loop_masks, making sure it's
4384 null for bb vectorization.
4386 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4388 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4389 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4390 loop->force_vectorize to check whether there is no alias.
4392 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4395 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4397 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4399 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4400 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4401 instead of a loop around prev_real_insn.
4402 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4406 * config/i386/i386.c (ix86_print_operand): Use conditional
4407 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4410 2018-03-02 Richard Biener <rguenther@suse.de>
4412 PR tree-optimization/84427
4413 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4414 (bitmap_set_subtract_values): Rewrite to handle multiple
4417 (prune_clobbered_mems): Likewise.
4418 (phi_translate): Take edge instead of pred/phiblock.
4419 (phi_translate_1): Likewise.
4420 (phi_translate_set): Likewise. Insert all translated
4421 exprs for a value into the set, keeping possibly multiple
4422 expressions per value.
4423 (compute_antic_aux): Adjust for phi_translate changes.
4424 When intersecting union the expressions and prune those
4425 not in the final value set, keeping possibly multiple
4426 expressions per value. Do not use value-insertion
4427 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4428 all expressions. Add verification that the value-sets
4429 only shrink during iteration.
4430 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4431 (do_pre_regular_insertion): Likewise.
4432 (do_pre_partial_partial_insertion): Likewise.
4434 2018-03-02 Richard Biener <rguenther@suse.de>
4437 * config/darwin.c (saved_debug_info_level): New static global.
4438 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4439 (darwin_asm_lto_end): Restore debug info generation settings.
4441 2018-03-01 Martin Liska <mliska@suse.cz>
4444 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4447 2018-03-01 Richard Biener <rguenther@suse.de>
4450 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4451 type annotation with LTO when debug was disabled at compile-time.
4453 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4455 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4457 (mips_final_postscan_insn): Likewise.
4459 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4461 PR rtl-optimization/84528
4462 * alias.c (init_alias_target): Add commentary.
4463 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4464 a unique base value if the frame pointer is not eliminated
4465 to the stack pointer.
4467 2018-03-01 Tom de Vries <tom@codesourcery.com>
4469 PR rtl-optimization/83327
4470 * lra-int.h (hard_regs_spilled_into): Declare.
4471 * lra.c (hard_regs_spilled_into): Define.
4472 (init_reg_info): Init hard_regs_spilled_into.
4473 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4474 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4475 (process_bb_lives): Handle hard_regs_spilled_into.
4476 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4477 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4479 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4481 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4482 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4483 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4484 * config/rs6000/aix72.h: New file.
4486 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4488 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4489 instead of warning_at with conditional singular and plural messages
4493 * stor-layout.c (update_alignment_for_field): For
4494 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4495 && !DECL_PACKED (field), do the alignment update, just use
4496 only desired_align instead of MAX (type_align, desired_align)
4498 (place_field): Don't do known_align < desired_align handling
4499 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4500 is non-NULL, instead do it after rli->prev_field handling and
4501 only if not within a bitfield word. For DECL_PACKED (field)
4502 use type_align of BITS_PER_UNIT.
4504 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4506 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4507 superfluous parentheses and trailing spaces.
4509 2018-02-28 Richard Biener <rguenther@suse.de>
4511 PR tree-optimization/84584
4512 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4513 SCoPs with fake exit edge.
4515 2018-02-28 Martin Liska <mliska@suse.cz>
4518 * timevar.c (timer::print): Fix format to properly print 100%
4521 2018-02-28 Richard Biener <rguenther@suse.de>
4524 * genmatch.c (capture_info::walk_match): Do not mark
4525 captured expressions without operands as expr_p given
4526 they act more like predicates and should be subject to
4527 "lost tail" side-effect preserving.
4529 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4531 PR rtl-optimization/81611
4532 * auto-inc-dec.c (attempt_change): Move dead note from
4533 mem_insn if it's the next use of regno
4534 (find_address): Take address use of reg holding
4535 non-incremented value. Add parm to limit search to the named
4537 (merge_in_block): Attempt to use a mem insn that is the next
4538 use of the original regno.
4540 2018-02-27 Martin Sebor <msebor@redhat.com>
4543 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4544 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4546 2018-02-27 Martin Sebor <msebor@redhat.com>
4548 PR translation/84207
4549 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4550 n argument to unsigned HOST_WIDE_INT.
4551 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4552 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4553 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4554 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4556 2018-02-27 Richard Biener <rguenther@suse.de>
4558 PR tree-optimization/84512
4559 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4560 Do not use the estimate returned from record_stmt_cost for
4561 the scalar iteration cost but sum properly using add_stmt_cost.
4563 2018-02-27 Richard Biener <rguenther@suse.de>
4565 PR tree-optimization/84466
4566 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4567 Adjust last change to less strictly validate use operands.
4569 2018-02-27 Martin Liska <mliska@suse.cz>
4571 PR gcov-profile/84548
4572 * gcov.c (process_file): Allow partial overlap and consider it
4573 also as group functions.
4574 (output_lines): Properly calculate range of lines for a group.
4576 2018-02-27 Martin Liska <mliska@suse.cz>
4578 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4579 'ggc' suffixes. Change first column width.
4580 (timer::print): Fix formatting of the column.
4582 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4584 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4585 preserve inline entry blocks for the sake of debug inline
4586 entry point markers alone.
4587 (remove_unused_locals): Suggest in comments a better place to
4588 force the preservation of inline entry blocks that are
4589 otherwise unused, but do not preserve them.
4591 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4593 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4595 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4598 * config/i386/constraints.md (Bs): Replace
4599 ix86_indirect_branch_register with
4600 TARGET_INDIRECT_BRANCH_REGISTER.
4602 * config/i386/i386.md (indirect_jump): Likewise.
4603 (tablejump): Likewise.
4604 (*sibcall_memory): Likewise.
4605 (*sibcall_value_memory): Likewise.
4606 Peepholes of indirect call and jump via memory: Likewise.
4607 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4608 (*sibcall_value_GOT_32): Likewise.
4609 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4610 (GOT_memory_operand): Likewise.
4611 (call_insn_operand): Likewise.
4612 (sibcall_insn_operand): Likewise.
4613 (GOT32_symbol_operand): Likewise.
4614 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4616 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4618 PR rtl-optimization/83496
4619 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4620 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4621 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4622 redundant insn, if any.
4623 (relax_delay_slots): Likewise.
4624 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4626 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4628 PR tree-optimization/83965
4629 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4630 that grouped statements are part of a reduction chain. Return
4631 true if the statement is not marked as a reduction itself but
4633 (vect_recog_dot_prod_pattern): Don't check whether the statement
4634 is part of a group here.
4635 (vect_recog_sad_pattern): Likewise.
4636 (vect_recog_widen_sum_pattern): Likewise.
4638 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4641 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4643 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4646 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4648 (ix86_output_indirect_function_return): New prototype.
4649 (ix86_split_simple_return_pop_internal): Likewise.
4650 * config/i386/i386.c (indirect_return_via_cx): New.
4651 (indirect_return_via_cx_bnd): Likewise.
4652 (indirect_thunk_name): Handle return va CX_REG.
4653 (output_indirect_thunk_function): Create alias for
4654 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4655 (ix86_output_indirect_jmp): Remove the bool argument.
4656 (ix86_output_indirect_function_return): New function.
4657 (ix86_split_simple_return_pop_internal): Likewise.
4658 * config/i386/i386.md (*indirect_jump): Don't pass false
4659 to ix86_output_indirect_jmp.
4660 (*tablejump_1): Likewise.
4661 (simple_return_pop_internal): Change it to define_insn_and_split.
4662 Call ix86_split_simple_return_pop_internal to split it for
4664 (simple_return_indirect_internal): Call
4665 ix86_output_indirect_function_return instead of
4666 ix86_output_indirect_jmp.
4668 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4671 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4672 memset and value initialization afterwards.
4674 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4676 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4678 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4681 * common/config/aarch64/aarch64-common.c
4682 (aarch_option_optimization_table[]): Switch
4683 off fomit-frame-pointer
4685 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4686 Chung-Ju Wu <jasonwucj@gmail.com>
4688 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4690 (store_multiple): Ditto.
4692 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4694 * config.gcc: Add --with-cpu support for nds32 target.
4695 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4696 * config/nds32/nds32.opt: Add -mcpu= option.
4698 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4700 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4701 isel=yes): Warn for these deprecated options.
4703 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4705 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4706 ISA_2_5_MASKS_EMBEDDED.
4708 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4710 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4711 p->max as pointers rather than using iterative_hash_expr.
4713 2018-02-23 Carl Love <cel@us.ibm.com>
4715 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4716 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4718 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4719 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4720 P8V_BUILTIN_VEC_VUNSIGNED2.
4722 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4725 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4726 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4728 * lra-constraints.c (curr_insn_transform): Set up
4729 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4731 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4732 * lra-eliminations.c (spill_pseudos): Ditto.
4733 (process_insn_for_elimination): Ditto.
4734 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4735 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4736 LRA_NON_CLOBBERED_ALT.
4738 2018-02-22 Martin Sebor <msebor@redhat.com>
4740 PR tree-optimization/84480
4741 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4742 to maybe_diag_stxncpy_trunc. Call it.
4743 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4744 from gimple_fold_builtin_strcpy. Print inlining stack.
4745 (handle_builtin_stxncpy): Print inlining stack.
4746 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4748 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4751 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4752 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4753 and -fcheck-pointer-bounds are used together.
4754 (indirect_thunk_prefix): New enum.
4755 (indirect_thunk_need_prefix): New function.
4756 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4757 "_nt" instead of "_bnd" for NOTRACK prefix.
4758 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4759 (output_indirect_thunk_function): Likewise.
4761 (ix86_code_end): Update output_indirect_thunk_function calls.
4762 (ix86_output_indirect_branch_via_reg): Replace
4763 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4764 (ix86_output_indirect_branch_via_push): Likewise.
4765 (ix86_output_function_return): Likewise.
4766 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4767 incompatible with -fcf-protection=branch and
4768 -fcheck-pointer-bounds.
4770 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4773 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4774 Change gcc_assert call to output_operand_lossage.
4776 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4778 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4780 2018-02-22 DJ Delorie <dj@redhat.com>
4781 Sebastian Perta <sebastian.perta@renesas.com>
4782 Oleg Endo <olegendo@gcc.gnu.org>
4784 * config/rx/rx.c (rx_rtx_costs): New function.
4785 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4787 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4789 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4791 2018-02-22 Martin Liska <mliska@suse.cz>
4794 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4795 Add "native" as a possible value.
4797 2018-02-22 Martin Liska <mliska@suse.cz>
4800 * config/i386/i386.c (ix86_option_override_internal):
4801 Add "native" as a possible value for -march and -mtune.
4803 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4806 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4807 to all type variants.
4809 PR tree-optimization/84503
4810 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4811 width as info->bitpos + info->bitsize - start.
4812 (merged_store_group::merge_overlapping): Simplify width computation.
4813 (check_no_overlap): New function.
4814 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4815 start + width and last_order of the group, fail if check_no_overlap
4817 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4818 to group if check_no_overlap fails.
4820 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4822 * config/rs6000/altivec.md: Delete contraint arguments to
4823 define_expand, define_split, and define_peephole2, and in
4824 define_insn_and_split if always unused.
4825 * config/rs6000/darwin.md: Ditto.
4826 * config/rs6000/dfp.md: Ditto.
4827 * config/rs6000/rs6000.md: Ditto.
4828 * config/rs6000/sync.md: Ditto.
4829 * config/rs6000/vector.md: Ditto.
4830 * config/rs6000/vsx.md: Ditto.
4832 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4834 * config/rs6000/altivec.md: Write output control strings as braced
4835 blocks instead of double-quoted strings.
4836 * config/rs6000/darwin.md: Ditto.
4837 * config/rs6000/rs6000.md: Ditto.
4838 * config/rs6000/vector.md: Ditto.
4839 * config/rs6000/vsx.md: Ditto.
4841 2018-02-21 Jason Merrill <jason@redhat.com>
4843 PR c++/84314 - ICE with templates and fastcall attribute.
4844 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4846 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4848 * ipa-cp.c (determine_versionability): Fix comment typos.
4850 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4853 * ipa-cp.c (determine_versionability): Do not version functions caling
4856 2018-02-21 Martin Liska <mliska@suse.cz>
4859 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4860 Add "native" as a possible value.
4861 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4862 the macro when native cpu detection is available.
4864 2018-02-21 Martin Liska <mliska@suse.cz>
4867 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4868 Add "native" as a possible value.
4869 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4870 when native cpu detection is available.
4872 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4873 Martin Sebor <msebor@redhat.com>
4875 PR tree-optimization/84478
4876 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4878 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4879 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4880 support which is conservatively correct, for 2 only stay conservative
4881 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4882 argument to the 2 argument get_range_strlen, adjust 6 arg
4883 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4885 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4886 (gimple_fold_builtin_strlen): Pass true as last argument to
4889 2018-02-20 Martin Sebor <msebor@redhat.com>
4892 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4893 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4894 (builtin_memref::builtin_memref): Factor out parts into
4895 set_base_and_offset and call it.
4897 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4900 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4901 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4902 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4903 search at the associated MODE_INT.
4905 2018-02-20 Jeff Law <law@redhat.com>
4908 PR tree-optimization/81592
4910 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4911 for range data rather than using global data.
4912 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4913 range data rather than using global data.
4914 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4915 pass it to children as needed.
4916 (struct directive::fmtresult): Similarly.
4917 (struct directive::set_width): Similarly.
4918 (struct directive::set_precision): Similarly.
4919 (format_integer, format_directive, parse_directive): Similarly.
4920 (format_none): Accept unnamed vr_values parameter.
4921 (format_percent, format_floating, format_character): Similarly.
4922 (format_string, format_plain): Similarly.
4923 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4924 the EVRP range analyzer for range data rather than using global data.
4925 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4926 gimple-ssa-evrp-analyze.h
4927 (class sprintf_dom_walker): Add after_dom_children member function.
4928 Add evrp_range_analyzer member.
4929 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4930 range analyzer as needed.
4931 (sprintf_dom_walker::after_dom_children): New member function.
4932 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4934 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4935 (evrp_range_analyzer::pop_to_marker): Likewise.
4937 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4939 PR tree-optimization/84419
4940 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4941 with the required type if its current type is compatible but
4944 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4947 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4948 after vectorization.
4950 2018-02-20 Martin Liska <mliska@suse.cz>
4953 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4954 possible values if we don't have a hint.
4956 2018-02-20 Martin Liska <mliska@suse.cz>
4960 * final.c (shorten_branches): Build align_tab array with one
4962 * opts.c (finish_options): Add alignment option limit check.
4963 (MAX_CODE_ALIGN): Likewise.
4964 (MAX_CODE_ALIGN_VALUE): Likewise.
4965 * doc/invoke.texi: Document maximum allowed option value for
4966 all -falign-* options.
4968 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4971 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4972 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4973 * var-tracking.c (emit_note_insn_var_location): Remove all references
4974 to NOTE_INSN_CALL_ARG_LOCATION.
4975 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4976 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4977 Use copy_rtx_if_shared.
4978 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4979 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4980 (dwarf2out_var_location): Remove handling of
4981 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4983 * final.c (final_scan_insn): Remove all references to
4984 NOTE_INSN_CALL_ARG_LOCATION.
4985 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4986 before dumping final insns.
4987 * except.c (emit_note_eh_region_end): Remove all references to
4988 NOTE_INSN_CALL_ARG_LOCATION.
4989 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4990 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4991 * config/arc/arc.c (hwloop_optimize): Likewise.
4992 * config/arm/arm.c (create_fix_barrier): Likewise.
4993 * config/s390/s390.c (s390_chunkify_start): Likewise.
4994 * config/sh/sh.c (find_barrier): Likewise.
4995 * config/i386/i386.c (rest_of_insert_endbranch,
4996 ix86_seh_fixup_eh_fallthru): Likewise.
4997 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4998 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4999 * config/frv/frv.c (frv_function_prologue): Likewise.
5000 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5002 (note_outside_basic_block_p): Remove all references to
5003 NOTE_INSN_CALL_ARG_LOCATION.
5004 * gengtype.c (adjust_field_rtx_def): Likewise.
5005 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5007 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5008 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5011 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5014 PR tree-optimization/84452
5015 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5016 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5019 2018-02-19 Martin Liska <mliska@suse.cz>
5022 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5024 2018-02-19 Martin Liska <mliska@suse.cz>
5025 Richard Sandiford <richard.sandiford@linaro.org>
5027 PR tree-optimization/82491
5028 * gimple-fold.c (get_base_constructor): Make earlier bail out
5031 2018-02-19 Carl Love <cel@us.ibm.com>
5033 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5034 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5036 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5037 P8V_BUILTIN_VEC_NEG.
5039 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5041 * config/rl78/rl78.md (movdf): New define expand.
5043 2018-02-19 Martin Liska <mliska@suse.cz>
5046 * doc/invoke.texi: Fix typo.
5047 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5049 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5051 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5052 handle rs6000_single_float and rs6000_double_float specially for
5055 2018-02-16 Jeff Law <law@redhat.com>
5057 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5060 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5063 * ipa-inline.c (inline_small_functions): Fix a typo.
5065 2018-02-16 Nathan Sidwell <nathan@acm.org>
5067 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5069 2018-02-16 Carl Love <cel@us.ibm.com>
5071 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5072 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5073 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5074 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5075 expansion to P8V_BUILTIN_VEC_FLOAT2.
5077 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5079 PR rtl-optimization/70023
5080 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5081 src_regno into account.
5083 2018-02-16 Carl Love <cel@us.ibm.com>
5085 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5086 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5087 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5088 * config/rs6000/rs6000.c: Remove case statements for
5089 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5090 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5091 and P9V_BUILTIN_VEC_VINSERT4B.
5092 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5093 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5094 * config/rs6000/vsx.md:
5095 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5098 2018-02-16 Carl Love <cel@us.ibm.com>
5100 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5102 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5104 * config/rs6000/rs6000-c.c: Add the definitions for
5105 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5106 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5107 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5108 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5109 definition for insert4b and define insn *insert3b_internal.
5110 * doc/extend.texi: Add documentation for vec_extract4b.
5112 2018-02-16 Nathan Sidwell <nathan@acm.org>
5114 * doc/extend.texi (Backwards Compatibility): Mention friend
5115 injection. Note for-scope is deprecated.
5116 * doc/invoke.texi (-ffriend-injection): Deprecate.
5118 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5120 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5121 that moved to I2, also allow destinations that are a paradoxical
5122 subreg (instead of a normal reg).
5124 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5127 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5130 2018-02-16 Richard Biener <rguenther@suse.de>
5132 PR tree-optimization/84037
5133 PR tree-optimization/84016
5135 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5136 Adjust vec_construct for the fact we need additional higher latency
5137 128bit inserts for AVX256 and AVX512 vector builds.
5138 (ix86_add_stmt_cost): Scale vector construction cost for
5141 2018-02-16 Richard Biener <rguenther@suse.de>
5143 PR tree-optimization/84417
5144 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5145 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5146 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5148 2018-02-16 Martin Liska <mliska@suse.cz>
5151 * internal-fn.def (ASAN_CHECK): Set proper flags.
5152 (ASAN_MARK): Likewise.
5154 2018-02-16 Julia Koval <julia.koval@intel.com>
5156 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5157 from PTA_CANNONLAKE.
5159 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5162 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5163 Use ++iter rather than iter++ for std::list iterators.
5164 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5165 defer deleting them until all nodes in the forest are processed. Do
5166 free even leaf nodes. Change to_process into auto_vec.
5169 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5170 * vec.h (vec_default_construct): Use memset instead of placement new
5171 if BROKEN_VALUE_INITIALIZATION is defined.
5172 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5173 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5176 PR rtl-optimization/83723
5177 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5178 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5179 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5181 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5183 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5185 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5187 PR rtl-optimization/81443
5188 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5189 from inner REGs to paradoxical SUBREGs.
5191 2018-02-16 Richard Biener <rguenther@suse.de>
5193 PR tree-optimization/84399
5194 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5195 For operands we can analyze at their definition make sure we can
5196 analyze them at each use as well.
5198 2018-02-16 Richard Biener <rguenther@suse.de>
5200 PR tree-optimization/84190
5201 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5202 volatile accesses if the decl isn't volatile.
5204 2018-02-15 Jason Merrill <jason@redhat.com>
5206 PR c++/84314 - ICE with templates and fastcall attribute.
5207 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5208 TYPE_CANONICAL on an existing type.
5210 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5212 PR tree-optimization/84383
5213 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5214 dstoff nor call operand_equal_p if dstbase is NULL.
5216 PR tree-optimization/84334
5217 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5218 also a CONSTANT_CLASS_P, punt.
5220 2018-02-14 Jim Wilson <jimw@sifive.com>
5222 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5223 first SMALL_OPERAND check. New local min_second_step. Move assert
5224 to where locals are set. Add TARGET_RVC support.
5225 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5227 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5229 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5231 2018-02-14 Martin Sebor <msebor@redhat.com>
5233 PR tree-optimization/83698
5234 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5235 arrays constrain the offset range to their bounds.
5236 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5237 (builtin_access::overlap): Avoid setting the size of overlap if it's
5239 (maybe_diag_overlap): Also consider arrays when deciding what values
5240 of offsets to include in diagnostics.
5242 2018-02-14 Martin Sebor <msebor@redhat.com>
5245 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5246 that correspond to the kind of a declaration.
5248 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5251 * config/pa/pa.md: Load address of PIC label using the linkage table
5252 if the label is nonlocal.
5254 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5256 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5257 warning message if user requests -maltivec=be.
5258 * doc/invoke.texi: Document deprecation of -maltivec=be.
5260 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5263 * config/rs6000/rs6000-c.c: Update definitions for
5264 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5265 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5267 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5270 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5271 add _get_ssp intrinsics. Remove argument from
5272 __builtin_ia32_rdssp[d|q].
5273 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5274 * config/i386/i386-builtin.def: Remove argument from
5275 __builtin_ia32_rdssp[d|q].
5276 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5277 ix86_expand_special_args_builtin for _rdssp[d|q].
5278 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5279 Clear register before usage.
5280 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5281 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5283 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5285 PR tree-optimization/84357
5286 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5287 operand 1 of an ARRAY_REF too.
5289 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5292 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5293 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5295 (set_of_reg): New struct.
5296 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5297 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5298 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5300 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5301 Split into bitclr, bitset, bitinvert patterns if appropriate.
5302 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5303 use rx_fuse_in_memory_bitop.
5304 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5305 to named insn, correct maximum insn length.
5307 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5310 * machmode.def: Define a complex mode for PARTIAL_INT.
5311 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5313 * doc/rtl.texi: Document CSPImode.
5314 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5316 (msp430_hard_regno_nregs_with_padding): Likewise.
5318 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5321 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5323 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5325 PR rtl-optimization/84169
5326 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5327 we generated a parallel as new i3 and we split that to new i2 and i3
5328 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5329 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5330 those to i2, not i1. Partially rewrite this scan code.
5332 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5335 * stor-layout.c (place_field): For variable length fields, adjust
5336 offset_align afterwards not just based on the field's alignment,
5337 but also on the size.
5340 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5341 of exps and logs in the use_exp2 case.
5343 2018-02-13 Jeff Law <law@redhat.com>
5345 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5348 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5351 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5355 * common.opt (gas-loc-support, gas-locview-support): New.
5356 (ginline-points, ginternal-reset-location-views): New.
5357 * doc/invoke.texi: Document them. Use @itemx where intended.
5358 (gvariable-location-views): Adjust.
5359 * target.def (reset_location_view): New.
5360 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5361 (TARGET_RESET_LOCATION_VIEW): New.
5362 * doc/tm.texi: Rebuilt.
5363 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5364 (dwarf2out_default_as_locview_support): New.
5365 (output_asm_line_debug_info): Use option variables.
5366 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5367 (output_loc_list): Likewise.
5368 (add_high_low_attributes): Check option variables.
5369 Don't output entry view attribute in strict mode.
5370 (gen_inlined_subroutine_die): Check option variables.
5371 (dwarf2out_inline_entry): Likewise.
5372 (init_sections_and_labels): Likewise.
5373 (dwarf2out_early_finish): Likewise.
5374 (maybe_reset_location_view): New, from...
5375 (dwarf2out_var_location): ... here. Call it.
5376 * debug.h (dwarf2out_default_as_loc_support): Declare.
5377 (dwarf2out_default_as_locview_support): Declare.
5378 * hooks.c (hook_int_rtx_insn_0): New.
5379 * hooks.h (hook_int_rtx_insn_0): Declare.
5380 * toplev.c (process_options): Take -gas-loc-support and
5381 -gas-locview-support from dwarf2out. Enable
5382 -gvariable-location-views by default only with locview
5383 assembler support. Enable -ginternal-reset-location-views by
5384 default only if the target defines the corresponding hook.
5385 Enable -ginline-points by default if location views are
5386 enabled; force it disabled if statement frontiers are
5388 * tree-inline.c (expand_call_inline): Check option variables.
5389 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5391 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5393 PR tree-optimization/84321
5394 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5395 handling. Also check whether the anti-range contains any values
5396 that satisfy the mask; switch to a VR_RANGE if not.
5398 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5401 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5403 2018-02-13 Martin Jambor <mjambor@suse.cz>
5406 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5407 of call statements, also set location of a load to a temporary.
5409 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5411 * config/rl78/rl78.c (add_vector_labels): New function.
5412 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5413 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5414 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5415 which checks that no arguments are passed.
5416 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5417 * doc/extend.texi: Documentation for the new attribute.
5419 2018-02-13 Andreas Schwab <schwab@suse.de>
5421 * config/riscv/linux.h (CPP_SPEC): Define.
5423 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5426 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5427 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5428 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5429 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5430 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5431 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5432 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5433 temporarily for AES and PCLMUL builtins.
5435 PR tree-optimization/84339
5436 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5437 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5441 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5442 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5443 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5445 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5447 * omp-simd-clone.h: New file.
5448 * omp-simd-clone.c: Include omp-simd-clone.h.
5449 (expand_simd_clones): No longer static.
5450 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5451 cgraph.h and omp-simd-clone.h.
5452 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5453 (vect_recog_widen_shift_pattern): Formatting fix.
5454 (vect_pattern_recog_1): Don't check optab for calls.
5457 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5458 operands[2] into a REG before using gen_lowpart on it.
5460 2018-02-12 Jeff Law <law@redhat.com>
5463 * config/sh/sh.c (find_barrier): Consider a sibling call
5466 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5467 successfully back substituting a reg.
5469 2018-02-12 Richard Biener <rguenther@suse.de>
5471 PR tree-optimization/84037
5472 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5473 parameter, move visited init to caller.
5474 (vect_slp_analyze_operations): Separate cost from validity
5475 check, initialize visited once for all instances.
5476 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5478 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5480 (vect_model_promotion_demotion_cost): Likewise.
5481 (vectorizable_bswap): Guard cost modeling with !slp_node
5482 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5484 (vectorizable_call): Likewise.
5485 (vectorizable_conversion): Likewise.
5486 (vectorizable_assignment): Likewise.
5487 (vectorizable_shift): Likewise.
5488 (vectorizable_operation): Likewise.
5489 (vectorizable_store): Likewise.
5490 (vectorizable_load): Likewise.
5491 (vectorizable_condition): Likewise.
5492 (vectorizable_comparison): Likewise.
5494 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5497 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5498 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5499 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5501 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5504 * params.def (inline-min-speedup): Increase from 8 to 15.
5505 (max-inline-insns-auto): Decrease from 40 to 30.
5506 * ipa-split.c (consider_split): Add some buffer for function to
5507 be considered inlining candidate.
5508 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5511 2018-02-12 Richard Biener <rguenther@suse.de>
5513 PR tree-optimization/84037
5514 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5515 matched stmts if we cannot swap the non-matched ones.
5517 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5519 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5520 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5521 _mm_maskz_scalef_round_ss): New intrinsics.
5522 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5523 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5524 __builtin_ia32_scalefss_round): Remove.
5525 (__builtin_ia32_scalefsd_mask_round,
5526 __builtin_ia32_scalefss_mask_round): New intrinsics.
5527 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5528 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5529 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5530 "<round_constraint>")): Changed to ...
5531 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5532 "<round_scalar_constraint>")): ... this.
5533 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5534 %0, %1, %2<round_op3>}"): Changed to ...
5535 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5536 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5537 %2<round_scalar_mask_op3>}"): ... this.
5538 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5540 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5542 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5543 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5544 (_mm_maskz_sqrt_round_ss): New intrinsics.
5545 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5546 (__builtin_ia32_sqrtsd_mask_round)
5547 (__builtin_ia32_sqrtss_mask_round): New builtins.
5548 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5549 (__builtin_ia32_sqrtss_round): Remove.
5550 (__builtin_ia32_sqrtsd_mask_round)
5551 (__builtin_ia32_sqrtss_mask_round): New builtins.
5552 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5553 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5554 ((match_operand:VF_128 1 "vector_operand"
5555 "xBm,<round_constraint>")): Changed to ...
5556 ((match_operand:VF_128 1 "vector_operand"
5557 "xBm,<round_scalar_constraint>")): ... this.
5558 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5559 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5560 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5561 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5562 %<iptr>1<round_scalar_mask_op3>}): ... this.
5563 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5564 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5566 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5569 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5570 Cast vec_cmpeq result to correct type.
5571 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5572 Cast vec_cmpgt result to correct type.
5574 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5576 * final.c (final_scan_insn_1): Renamed from...
5577 (final_scan_insn): ... this. New wrapper, to recover
5578 seen from the outermost call in recursive ones.
5579 * config/sparc/sparc.c (output_return): Drop seen from call.
5580 (output_sibcall): Likewise.
5581 * config/visium/visium.c (output_branch): Likewise.
5583 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5585 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5588 2018-02-10 Alan Modra <amodra@gmail.com>
5591 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5592 Specify LR as an input.
5594 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5597 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5598 remove_member_access_dummy_vars): New functions.
5599 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5600 lower_omp_1, execute_lower_omp): Use them.
5602 PR rtl-optimization/84308
5603 * shrink-wrap.c (spread_components): Release todo vector.
5605 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5607 PR rtl-optimization/57193
5608 * ira-color.c (struct allocno_color_data): Add member
5609 conflict_allocno_hard_prefs.
5610 (update_conflict_allocno_hard_prefs): New.
5611 (bucket_allocno_compare_func): Add a preference based on
5612 conflict_allocno_hard_prefs.
5613 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5614 (color_allocnos): Remove a dead code. Initiate
5615 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5617 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5620 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5621 constraint from =wa to wa. Avoid a subreg on the output operand,
5622 instead use a pseudo and subreg it in a move.
5623 (p9_xxbrd_<mode>): Changed to ...
5624 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5625 (p9_xxbrd_v2df): New expander.
5626 (p9_xxbrw_<mode>): Changed to ...
5627 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5628 (p9_xxbrw_v4sf): New expander.
5630 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5632 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5634 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5637 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5638 multiply in 32-bit mode.
5639 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5640 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5643 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5645 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5646 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5647 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5648 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5650 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5653 * dwarf2out.c (is_trivial_indirect_ref): New function.
5654 (dwarf2out_late_global_decl): Do not generate a location
5655 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5656 and that are not defined in the current unit.
5658 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5660 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5661 instead of a libcall for UNORDERED.
5663 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5666 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5667 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5669 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5672 * config/s390/s390.c (s390_set_current_function): Invoke
5673 s390_indirect_branch_settings also if fndecl didn't change.
5675 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5677 * config/rs6000/rs6000.md (blockage): Set length to zero.
5679 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5681 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5683 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5686 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5687 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5691 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5692 PARALLEL incoming that failed vt_get_decl_and_offset check.
5695 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5696 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5698 (get_variable_section): For decls in named .bss* sections pass true as
5699 second argument to bss_initializer_p.
5701 2018-02-09 Marek Polacek <polacek@redhat.com>
5702 Jakub Jelinek <jakub@redhat.com>
5705 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5706 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5707 Sync some changes from cxx_fold_indirect_ref.
5709 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5711 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5713 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5714 (BLOCK_INLINE_ENTRY_LABEL): New.
5715 (dwarf2out_var_location): Disregard inline entry markers.
5716 (inline_entry_data): New struct.
5717 (inline_entry_data_hasher): New hashtable type.
5718 (inline_entry_data_hasher::hash): New.
5719 (inline_entry_data_hasher::equal): New.
5720 (inline_entry_data_table): New variable.
5721 (add_high_low_attributes): Add DW_AT_entry_pc and
5722 DW_AT_GNU_entry_view attributes if a pending entry is found
5723 in inline_entry_data_table. Add old entry_pc attribute only
5724 if debug nonbinding markers are disabled.
5725 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5726 markers are enabled.
5727 (block_within_block_p, dwarf2out_inline_entry): New.
5728 (dwarf2out_finish): Check that no entries remained in
5729 inline_entry_data_table.
5730 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5731 (final_scan_insn, notice_source_line): Likewise.
5732 (rest_of_clean_state): Skip inline entry markers.
5733 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5735 * gimple.c (gimple_build_debug_inline_entry): New.
5736 * gimple.h (enum gimple_debug_subcode): Add
5737 GIMPLE_DEBUG_INLINE_ENTRY.
5738 (gimple_build_debug_inline_entry): Declare.
5739 (gimple_debug_inline_entry_p): New.
5740 (gimple_debug_nonbind_marker_p): Adjust.
5741 * insn-notes.def (INLINE_ENTRY): New.
5742 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5743 inline entry marker notes.
5744 (print_insn): Likewise.
5745 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5746 (INSN_DEBUG_MARKER_KIND): Likewise.
5747 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5748 * tree-inline.c (expand_call_inline): Build and insert
5749 debug_inline_entry stmt.
5750 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5751 inline entry blocks early, if nonbind markers are enabled.
5752 (dump_scope_block): Dump fragment info.
5753 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5754 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5755 (gimple_build_debug_inline_entry): New.
5756 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5757 Enable/disable inline entry points too.
5758 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5759 (DEBUG_INSN): Describe inline entry markers.
5761 * common.opt (gvariable-location-views): New.
5762 (gvariable-location-views=incompat5): New.
5763 * config.in: Rebuilt.
5764 * configure: Rebuilt.
5765 * configure.ac: Test assembler for view support.
5766 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5767 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5768 * dwarf2out.c (var_loc_view): New typedef.
5769 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5770 (dwarf2out_locviews_in_attribute): New.
5771 (dwarf2out_locviews_in_loclist): New.
5772 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5773 (enum dw_line_info_opcode): Add LI_adv_address.
5774 (struct dw_line_info_table): Add view.
5775 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5776 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5777 (zero_view_p): New variable.
5778 (ZERO_VIEW_P): New macro.
5779 (output_asm_line_debug_info): New.
5780 (struct var_loc_node): Add view.
5781 (add_AT_view_list, AT_loc_list): New.
5782 (add_var_loc_to_decl): Add view param. Test it against last.
5783 (new_loc_list): Add view params. Record them.
5784 (AT_loc_list_ptr): Handle loc and view lists.
5785 (view_list_to_loc_list_val_node): New.
5786 (print_dw_val): Handle dw_val_class_view_list.
5787 (size_of_die): Likewise.
5788 (value_format): Likewise.
5789 (loc_list_has_views): New.
5790 (gen_llsym): Set vl_symbol too.
5791 (maybe_gen_llsym, skip_loc_list_entry): New.
5792 (dwarf2out_maybe_output_loclist_view_pair): New.
5793 (output_loc_list): Output view list or entries too.
5794 (output_view_list_offset): New.
5795 (output_die): Handle dw_val_class_view_list.
5796 (output_dwarf_version): New.
5797 (output_compilation_unit_header): Use it.
5798 (output_skeleton_debug_sections): Likewise.
5799 (output_rnglists, output_line_info): Likewise.
5800 (output_pubnames, output_aranges): Update version comments.
5801 (output_one_line_info_table): Output view numbers in asm comments.
5802 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5803 Call maybe_gen_llsym.
5804 (loc_list_from_tree_1): Adjust.
5805 (add_AT_location_description): Create view list attribute if
5806 needed, check it's absent otherwise.
5807 (convert_cfa_to_fb_loc_list): Adjust.
5808 (maybe_emit_file): Call output_asm_line_debug_info for test.
5809 (dwarf2out_var_location): Reset views as needed. Precompute
5810 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5811 attribute. Set view.
5812 (new_line_info_table): Reset next view.
5813 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5814 (dwarf2out_source_line): Likewise. Output view resets and labels to
5815 the assembler, or select appropriate line info opcodes.
5816 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5817 (optimize_string_length): Catch it. Adjust.
5818 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5819 dw_val_class_view_list, and remove it if no longer needed.
5820 (hash_loc_list): Hash view numbers.
5821 (loc_list_hasher::equal): Compare them.
5822 (optimize_location_lists): Check whether a view list symbol is
5823 needed, and whether the locview attribute is present, and
5824 whether they match. Remove the locview attribute if no longer
5826 (index_location_lists): Call skip_loc_list_entry for test.
5827 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5828 Use output_dwarf_version.
5829 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5830 (struct dw_val_node): Add val_view_list.
5831 * final.c (SEEN_NEXT_VIEW): New.
5832 (set_next_view_needed): New.
5833 (clear_next_view_needed): New.
5834 (maybe_output_next_view): New.
5835 (final_start_function): Rename to...
5836 (final_start_function_1): ... this. Take pointer to FIRST,
5837 add SEEN parameter. Emit param bindings in the initial view.
5838 (final_start_function): Reintroduce SEEN-less interface.
5839 (final): Rename to...
5840 (final_1): ... this. Take SEEN parameter. Output final pending
5841 next view at the end.
5842 (final): Reintroduce seen-less interface.
5843 (final_scan_insn): Output pending next view before switching
5844 sections or ending a block. Mark the next view as needed when
5845 outputting variable locations. Notify debug backend of section
5846 changes, and of location view changes.
5847 (rest_of_handle_final): Adjust.
5848 * toplev.c (process_options): Autodetect value for debug variable
5849 location views option. Warn on incompat5 without -gdwarf-5.
5850 * doc/invoke.texi (gvariable-location-views): New.
5851 (gvariable-location-views=incompat5): New.
5852 (gno-variable-location-views): New.
5854 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5856 PR tree-optimization/84136
5857 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5858 that the result of find_edge is non-NULL.
5860 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5863 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5864 storing integer register in SImode. Fix cost of 256 and 512
5865 byte aligned SSE register store.
5867 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5869 * config/i386/i386.c (ix86_multiplication_cost): Fix
5870 multiplication cost for TARGET_AVX512DQ.
5872 2018-02-08 Marek Polacek <polacek@redhat.com>
5874 PR tree-optimization/84238
5875 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5878 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5880 PR tree-optimization/84265
5881 * tree-vect-stmts.c (vectorizable_store): Don't treat
5882 VMAT_CONTIGUOUS accesses as grouped.
5883 (vectorizable_load): Likewise.
5885 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5887 PR tree-optimization/81635
5888 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5889 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5890 (test_round_for_mask): New functions.
5891 (wide_int_cc_tests): Call test_round_for_mask.
5892 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5893 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5894 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5895 range returned by get_range_info.
5897 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5900 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5901 * symtab.c: Include builtins.h
5902 (symtab_node::output_to_lto_symbol_table_p): Move here
5903 from lto-streamer-out.c:output_symbol_p.
5904 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5905 (output_symbol_p): Move all logic to symtab.c
5906 (produce_symtab): Update.
5908 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5910 * config/s390/s390-opts.h (enum indirect_branch): Define.
5911 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5912 (s390_indirect_branch_via_thunk)
5913 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5914 (enum s390_indirect_branch_type): Define.
5915 * config/s390/s390.c (struct s390_frame_layout, struct
5916 machine_function): Remove.
5917 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5918 (indirect_branch_table_label_no, indirect_branch_table_name):
5920 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5921 (enum s390_indirect_branch_option): Define.
5922 (s390_return_addr_from_memory): New function.
5923 (s390_handle_string_attribute): New function.
5924 (s390_attribute_table): Add new attribute handler.
5925 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5926 (s390_indirect_branch_via_thunk): New function.
5927 (s390_indirect_branch_via_inline_thunk): New function.
5928 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5929 sibling call optimization for non z10 compiles.
5930 (s390_emit_call): Force indirect branch target to be a single
5931 register. Add r1 clobber for non-z10 compiles.
5932 (s390_emit_epilogue): Emit return jump via return_use expander.
5933 (s390_reorg): Handle JUMP_INSNs as execute targets.
5934 (s390_option_override_internal): Perform validity checks for the
5935 new command line options.
5936 (s390_indirect_branch_attrvalue): New function.
5937 (s390_indirect_branch_settings): New function.
5938 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5939 (s390_output_indirect_thunk_function): New function.
5940 (s390_code_end): Implement target hook.
5941 (s390_case_values_threshold): Implement target hook.
5942 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5944 * config/s390/s390.h (struct s390_frame_layout)
5945 (struct machine_function): Move here from s390.c.
5946 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5947 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5948 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5949 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5950 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5951 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5952 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5953 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5954 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5955 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5956 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5957 (mnemonic attribute): Add values which aren't recognized
5959 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5960 pattern for branch conversion. Fix mnemonic attribute.
5961 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5962 indirect branch via thunk if requested.
5963 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5964 ("*indirect_jump"): Disable for branch conversion using out of
5966 ("indirect_jump_via_thunk<mode>_z10")
5967 ("indirect_jump_via_thunk<mode>")
5968 ("indirect_jump_via_inlinethunk<mode>_z10")
5969 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5970 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5971 ("casesi_jump_via_inlinethunk<mode>_z10")
5972 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5973 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5974 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5975 ("*indirect2_jump"): Disable for branch conversion.
5976 ("casesi_jump"): Turn into expander and expand patterns for branch
5978 ("return_use"): New expander.
5979 ("*return"): Emit return via thunk and rename it to ...
5980 ("*return<mode>"): ... this one.
5981 * config/s390/s390.opt: Add new options and and enum for the
5984 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5986 * lra-constraints.c (match_reload): Unconditionally use
5987 gen_lowpart_SUBREG, rather than selecting between that
5988 and equivalent gen_rtx_SUBREG code.
5990 2018-02-08 Richard Biener <rguenther@suse.de>
5992 PR tree-optimization/84233
5993 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5994 changed flag instead of boguously re-using phi_inserted.
5996 2018-02-08 Martin Jambor <mjambor@suse.cz>
5998 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5999 static local variables.
6001 2018-02-08 Richard Biener <rguenther@suse.de>
6003 PR tree-optimization/84278
6004 * tree-vect-stmts.c (vectorizable_store): When looking for
6005 smaller vector types to perform grouped strided loads/stores
6006 make sure the mode is supported by the target.
6007 (vectorizable_load): Likewise.
6009 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6011 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6012 Increase LDP/STP opportunities by adding adjacent callee-saves.
6014 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6016 PR rtl-optimization/84068
6017 PR rtl-optimization/83459
6018 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6020 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6022 PR tree-optimization/84224
6023 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6024 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6027 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6030 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6031 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6032 count, remove one USE.
6034 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6036 * doc/install.texi (Configuration): Document the
6037 --with-long-double-format={ibm,ieee} PowerPC configuration
6041 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6042 Convert from define_expand to be define_insn_and_split. Rework
6043 float/double/_Float128 conversions to QI/HI/SImode to work with
6044 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6045 conversions to QI/HImode types did a store and then a load to
6046 truncate the value. For conversions to VSX registers, don't split
6047 the insn, instead emit the code directly. Use the code iterator
6048 any_fix to combine signed and unsigned conversions.
6049 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6050 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6051 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6052 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6053 (fix_<mode>di2_hw): Likewise.
6054 (fixuns_<mode>di2_hw): Likewise.
6055 (fix_<mode>si2_hw): Likewise.
6056 (fixuns_<mode>si2_hw): Likewise.
6057 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6058 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6059 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6060 fix<uns>_trunc<SFDF:mode>si2_p8.
6061 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6063 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6064 (fix<uns>_<mode>_mem): Likewise.
6065 (fctiw<u>z_<mode>_mem): Likewise.
6066 (fix<uns>_<mode>_mem): Likewise.
6067 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6068 the register allocator from doing a direct move to the GPRs to do
6069 a store, and instead use the ISA 3.0 store byte/half-word from
6070 vector register instruction. For IEEE 128-bit floating point,
6071 also optimize stores of 32-bit ints.
6072 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6074 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6076 * genextract.c (push_pathstr_operand): New function to support
6078 (walk_rtx): Call push_pathstr_operand.
6079 (print_path): Support [a-zA-Z].
6081 2018-02-07 Richard Biener <rguenther@suse.de>
6083 PR tree-optimization/84037
6084 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6085 (cse_and_gimplify_to_preheader): Declare.
6086 (vect_get_place_in_interleaving_chain): Likewise.
6087 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6089 (_loop_vec_info::~_loop_vec_info): Delete it.
6090 (cse_and_gimplify_to_preheader): New function.
6091 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6092 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6093 (vectorizable_load): Likewise. For grouped stores always base
6094 the IV on the first element.
6095 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6096 condition before gimplifying.
6098 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6100 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6101 *DIV_EXPR and *MOD_EXPR.
6103 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6106 * config/i386/i386.c (ix86_option_override_internal): Mask out
6107 the CF_SET bit when checking -fcf-protection.
6109 2018-02-07 Tom de Vries <tom@codesourcery.com>
6112 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6115 2018-02-07 Richard Biener <rguenther@suse.de>
6117 PR tree-optimization/84204
6118 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6121 PR tree-optimization/84205
6122 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6123 special-case isl_ast_op_zdiv_r.
6125 PR tree-optimization/84223
6126 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6127 Only add conditions from within the region.
6128 (gather_bbs::after_dom_children): Adjust.
6130 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6133 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6134 * config/avr/avr.md: Only post-reload split REG-REG moves if
6135 either register is GENERAL_REG_P.
6137 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6139 PR tree-optimization/84235
6140 * tree-ssa-scopedtables.c
6141 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6142 if the subtraction is performed in floating point type where NaNs are
6143 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6144 build 1. Formatting fix.
6146 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6149 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6150 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6151 and skip it regardless of bb boundaries. Use CALL_P macro,
6152 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6153 unnecessarily, formatting fix.
6155 2018-02-06 Michael Collison <michael.collison@arm.com>
6157 * config/arm/thumb2.md:
6158 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6159 (*thumb_mov_notscc): Ditto.
6161 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6164 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6165 unsigned_fix, not "s".
6167 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6169 * configure.ac (gcc_fn_eh_frame_ro): New function.
6170 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6171 correct .eh_frame permissions.
6172 * configure: Regenerate.
6174 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6176 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6179 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6181 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6182 Display warning message for -mno-speculate-indirect-jumps.
6184 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6186 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6188 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6190 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6192 PR tree-optimization/84225
6193 * tree-eh.c (find_trapping_overflow): Only call
6194 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6196 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6199 * config/i386/i386.c: Reimplement the check of possible options
6200 -mibt/-mshstk conbination. Change error messages.
6201 * doc/invoke.texi: Fix a typo: remove extra '='.
6203 2018-02-06 Marek Polacek <polacek@redhat.com>
6205 PR tree-optimization/84228
6206 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6208 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6211 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6212 emitted arch directives.
6213 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6214 __ARM_FEATURE_COPROC before changing architectures.
6216 2018-02-06 Richard Biener <rguenther@suse.de>
6218 * config/i386/i386.c (print_reg): Fix typo.
6219 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6221 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6223 * configure: Regenerate.
6225 2018-02-05 Martin Sebor <msebor@redhat.com>
6227 PR tree-optimization/83369
6228 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6231 2018-02-05 Martin Liska <mliska@suse.cz>
6233 * doc/invoke.texi: Cherry-pick upstream r323995.
6235 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6237 * ira.c (ira_init_register_move_cost): Adjust comment.
6239 2018-02-05 Martin Liska <mliska@suse.cz>
6241 PR gcov-profile/84137
6242 * doc/gcov.texi: Fix typo in documentation.
6244 2018-02-05 Martin Liska <mliska@suse.cz>
6246 PR gcov-profile/83879
6247 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6248 using dlopen functionality.
6250 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6252 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6253 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6254 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6255 _mm_maskz_range_round_ss): New intrinsics.
6256 (__builtin_ia32_rangesd128_round)
6257 (__builtin_ia32_rangess128_round): Remove.
6258 (__builtin_ia32_rangesd128_mask_round,
6259 __builtin_ia32_rangess128_mask_round): New builtins.
6260 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6261 __builtin_ia32_rangess128_round): Remove.
6262 (__builtin_ia32_rangesd128_mask_round,
6263 __builtin_ia32_rangess128_mask_round): New builtins.
6264 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6265 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6266 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6267 "<round_saeonly_constraint>")): Changed to ...
6268 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6269 "<round_saeonly_scalar_constraint>")): ... this.
6270 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6271 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6272 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6273 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6274 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6276 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6278 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6280 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6281 Remove all values except native, 8540 and 8548.
6283 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6285 * config/i386/i386.c (ix86_output_function_return): Pass
6286 INVALID_REGNUM, instead of -1, as invalid register number to
6287 indirect_thunk_name and output_indirect_thunk.
6289 2018-02-02 Julia Koval <julia.koval@intel.com>
6291 * config.gcc: Add -march=icelake.
6292 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6293 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6294 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6295 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6296 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6297 (processor_target_table): Add icelake.
6298 (ix86_option_override_internal): Handle new PTAs.
6299 (get_builtin_code_for_version): Handle icelake.
6300 (M_INTEL_COREI7_ICELAKE): New.
6301 (fold_builtin_cpu): Handle icelake.
6302 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6303 * doc/invoke.texi: Add -march=icelake.
6305 2018-02-02 Julia Koval <julia.koval@intel.com>
6307 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6308 to wide_int_bitmask.
6309 * wide-int-bitmask.h: New.
6311 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6314 * config/i386/i386.md: Replace Pmode with word_mode in
6315 builtin_setjmp_setup and builtin_longjmp to support x32.
6317 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6321 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6323 (rs6000_supported_cpu_names): New static variable.
6324 (linux_cpu_translation_table): Likewise.
6325 (elf_platform) <cpu>: Define new static variable and use it.
6326 Translate kernel AT_PLATFORM name to canonical name if needed.
6327 Error if platform name is unknown.
6329 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6332 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6334 2018-02-01 Jeff Law <law@redhat.com>
6337 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6338 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6339 the scratch if RELEASE_VIA_POP is false.
6340 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6341 If we have to save a temporary register, decrement SIZE appropriately.
6342 Pass new arguments to release_scratch_register_on_entry.
6343 (ix86_adjust_stack_and_probe): Likewise.
6344 (ix86_emit_probe_stack_range): Pass new arguments to
6345 release_scratch_register_on_entry.
6347 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6349 PR rtl-optimization/84157
6350 * combine.c (change_zero_ext): Use REG_P predicate in
6351 front of HARD_REGISTER_P predicate.
6353 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6355 * config/avr/avr.c (avr_option_override): Move disabling of
6356 -fdelete-null-pointer-checks to...
6357 * common/config/avr/avr-common.c (avr_option_optimization_table):
6360 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6362 PR tree-optimization/81635
6363 * tree-data-ref.c (split_constant_offset_1): For types that
6364 wrap on overflow, try to use range info to prove that wrapping
6367 2018-02-01 Renlin Li <renlin.li@arm.com>
6370 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6372 (aarch64_register_move_cost): Likewise.
6373 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6375 (REG_CLASS_NAMES): Likewise.
6376 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6377 TAILCALL_ADDR_REGS. Remove IP registers.
6378 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6380 2018-02-01 Richard Biener <rguenther@suse.de>
6382 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6383 for specifying RPO order and allow NULL for that.
6384 * domwalk.c (dom_walker::dom_walker): Likewise.
6385 (dom_walker::walk): Handle NULL RPO order.
6386 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6388 (rewrite_update_dom_walker): Likewise.
6389 (mark_def_dom_walker): Likewise.
6391 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6393 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6394 (aarch64_maybe_expand_sve_subreg_move): Declare.
6395 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6396 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6398 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6399 that are semantically a reverse operation.
6400 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6401 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6402 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6404 (aarch64_can_change_mode_class): For big-endian, forbid changes
6405 between two SVE modes if they have different element sizes.
6407 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6409 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6410 the TImode handling for big-endian targets.
6412 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6414 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6415 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6417 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6418 Remove BSWAP handing for big-endian targets and use the form of
6419 LD1RQ appropariate for the mode.
6421 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6423 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6424 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6427 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6430 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6431 check for operands that need to go through aarch64_sve_reload_be.
6433 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6435 PR tree-optimization/81661
6436 PR tree-optimization/84117
6437 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6438 * tree-eh.c: Include gimplify.h.
6439 (find_trapping_overflow, replace_trapping_overflow,
6440 rewrite_to_non_trapping_overflow): New functions.
6441 * tree-vect-loop.c: Include tree-eh.h.
6442 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6443 * tree-data-ref.c: Include tree-eh.h.
6444 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6446 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6448 PR rtl-optimization/84123
6449 * combine.c (change_zero_ext): Check if hard register satisfies
6450 can_change_dest_mode before calling gen_lowpart_SUBREG.
6452 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6455 * ira.c (ira_init_register_move_cost): Remove assert.
6457 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6459 PR rtl-optimization/84071
6460 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6461 * doc/tm.texi: Regenerate.
6463 2018-01-31 Richard Biener <rguenther@suse.de>
6465 PR tree-optimization/84132
6466 * tree-data-ref.c (analyze_miv_subscript): Properly
6467 check whether evolution_function_is_affine_multivariate_p
6468 before calling gcd_of_steps_may_divide_p.
6470 2018-01-31 Julia Koval <julia.koval@intel.com>
6473 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6474 * config/i386/i386.md (rdpid_rex64) New.
6475 (rdpid): Make 32bit only.
6477 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6480 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6481 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6483 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6486 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6488 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6490 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6492 PR rtl-optimization/84071
6493 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6494 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6496 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6498 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6499 (arc_attribute_table): Add 'aux' attribute.
6500 (arc_in_small_data_p): Consider aux like variables.
6501 (arc_is_aux_reg_p): New function.
6502 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6503 (arc_get_aux_arg): New function.
6504 (prepare_move_operands): Handle aux-register access.
6505 (arc_handle_aux_attribute): New function.
6506 * doc/extend.texi (ARC Variable attributes): Add subsection.
6508 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6510 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6511 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6512 (arc_attribute_table): Add 'uncached' attribute.
6513 (arc_print_operand): Print '.di' flag for uncached memory
6515 (arc_in_small_data_p): Do not consider for small data the uncached
6517 (arc_is_uncached_mem_p): New function.
6518 * config/arc/predicates.md (compact_store_memory_operand): Check
6519 for uncached memory accesses.
6520 (nonvol_nonimm_operand): Likewise.
6521 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6523 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6526 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6527 falign-loops=): Add Optimization flag.
6529 2018-01-30 Jeff Law <law@redhat.com>
6532 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6533 INT_REGISTERS_SAVED. Check it prior to calling
6534 get_scratch_register_on_entry.
6535 (ix86_adjust_stack_and_probe): Similarly.
6536 (ix86_emit_probe_stack_range): Similarly.
6537 (ix86_expand_prologue): Corresponding changes.
6539 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6542 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6543 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6545 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6548 * lra-constraints.c (curr_insn_transform): Process AND in the
6551 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6553 PR rtl-optimization/83986
6554 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6555 dependence against last_pending_memory_flush in addition to
6558 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6560 PR tree-optimization/81611
6561 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6564 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6567 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6570 2018-01-30 Richard Biener <rguenther@suse.de>
6571 Jakub Jelinek <jakub@redhat.com>
6573 PR tree-optimization/84111
6574 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6575 inner loops added during recursion, as they don't have up-to-date
6578 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6581 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6582 (can_inline_edge_by_limits_p): ... here.
6583 (can_early_inline_edge_p, check_callers,
6584 update_caller_keys, update_callee_keys, recursive_inlining,
6585 add_new_edges_to_heap, speculation_useful_p,
6586 inline_small_functions,
6587 inline_small_functions, flatten_function,
6588 inline_to_all_callers_1): Update.
6590 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6592 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6595 2018-01-30 Richard Biener <rguenther@suse.de>
6597 PR tree-optimization/83008
6598 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6599 invariant and constant vector uses in stmts when they need
6602 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6605 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6606 * configure: Regenerate.
6608 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6610 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6612 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6613 Use gen_rtx_REG rather than gen_lowpart.
6615 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6617 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6618 rather than 0 when creating partial subregs.
6620 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6622 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6625 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6628 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6629 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6630 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6631 flags. This restores the settings used before the 2017-07-24.
6632 Turning off pre increment/decrement/modify allows IVOPTS to
6633 optimize DF/SF loops where the index is an int.
6635 2018-01-29 Richard Biener <rguenther@suse.de>
6636 Kelvin Nilsen <kelvin@gcc.gnu.org>
6639 * tree-vect-stmts.c (vectorizable_call): Don't call
6640 targetm.vectorize_builtin_md_vectorized_function if callee is
6643 2018-01-22 Carl Love <cel@us.ibm.com>
6645 * doc/extend.tex: Fix typo in second arg in
6646 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6648 2018-01-29 Richard Biener <rguenther@suse.de>
6650 PR tree-optimization/84086
6651 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6652 (flush_ssaname_freelist): When SSA names were released reset
6653 the SCEV hash table.
6655 2018-01-29 Richard Biener <rguenther@suse.de>
6657 PR tree-optimization/84057
6658 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6659 removed paths when removing edges.
6661 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6663 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6664 -mfunction-return=@var{choice}.
6666 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6669 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6671 (layout::print_source_line): Likewise.
6672 (test_get_line_width_without_trailing_whitespace): Add test cases.
6674 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6677 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6680 2018-01-26 Jim Wilson <jimw@sifive.com>
6682 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6684 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6687 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6689 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6690 and CMP + SUB-immediate -> SUBS.
6692 2018-01-26 Martin Sebor <msebor@redhat.com>
6694 PR tree-optimization/83896
6695 * tree-ssa-strlen.c (get_string_len): Rename...
6696 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6697 Avoid assuming length is constant.
6698 (handle_char_store): Use HOST_WIDE_INT for string length.
6700 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6703 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6704 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6706 2018-01-26 Richard Biener <rguenther@suse.de>
6708 PR rtl-optimization/84003
6709 * dse.c (record_store): Only record redundant stores when
6710 the earlier store aliases at least all accesses the later one does.
6712 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6714 PR rtl-optimization/83985
6715 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6716 REG_CFA_RESTORE insns.
6717 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6718 REG_CFA_RESTORE insns here.
6721 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6722 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6724 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6726 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6727 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6728 (arc_init): Likewise.
6729 (arc_override_options): Likewise.
6730 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6732 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6734 * config/arc/arc.h (TARGET_DBNZ): Define.
6735 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6736 properly set the tune attribute.
6737 (dbnz): Use TARGET_DBNZ guard.
6738 * config/arc/arc.opt (mtune): Add core3 option.
6740 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6742 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6743 recognize new pic like addresses.
6744 (arc_delegitimize_address): Clean up.
6746 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6748 * config/arc/arc-arches.def: Option mrf16 valid for all
6750 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6751 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6752 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6753 * config/arc/arc-tables.opt: Regenerate.
6754 * config/arc/arc.c (arc_conditional_register_usage): Handle
6755 reduced register file case.
6756 (arc_file_start): Set must have build attributes.
6757 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6759 * config/arc/arc.opt (mrf16): Add new option.
6760 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6761 * config/arc/genmultilib.awk: Handle new mrf16 option.
6762 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6763 * config/arc/t-multilib: Regenerate.
6764 * doc/invoke.texi (ARC Options): Document mrf16 option.
6766 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6768 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6769 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6770 (arc_attribute_table): Add 'secure_call' attribute.
6771 (arc_print_operand): Print secure call operand.
6772 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6774 (arc_is_secure_call_p): New function. * config/arc/arc.md
6775 (call_i): Add support for sjli instruction.
6776 (call_value_i): Likewise.
6777 * config/arc/constraints.md (Csc): New constraint.
6779 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6780 John Eric Martin <John.Martin@emmicro-us.com>
6782 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6783 * config/arc/arc.c (_arc_jli_section): New struct.
6784 (arc_jli_section): New type.
6785 (rc_jli_sections): New static variable.
6786 (arc_handle_jli_attribute): New function.
6787 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6788 (arc_file_end): New function.
6789 (TARGET_ASM_FILE_END): Define.
6790 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6791 (arc_add_jli_section): New function.
6792 (jli_call_scan): Likewise.
6793 (arc_reorg): Call jli_call_scan.
6794 (arc_output_addsi): Remove 'S' from printing asm operand.
6795 (arc_is_jli_call_p): New function.
6796 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6798 (movhi_insn): Likewise.
6799 (movsi_insn): Likewise.
6800 (movsi_set_cc_insn): Likewise.
6801 (loadqi_update): Likewise.
6802 (load_zeroextendqisi_update): Likewise.
6803 (load_signextendqisi_update): Likewise.
6804 (loadhi_update): Likewise.
6805 (load_zeroextendhisi_update): Likewise.
6806 (load_signextendhisi_update): Likewise.
6807 (loadsi_update): Likewise.
6808 (loadsf_update): Likewise.
6809 (movsicc_insn): Likewise.
6810 (bset_insn): Likewise.
6811 (bxor_insn): Likewise.
6812 (bclr_insn): Likewise.
6813 (bmsk_insn): Likewise.
6814 (bicsi3_insn): Likewise.
6815 (cmpsi_cc_c_insn): Likewise.
6816 (movsi_ne): Likewise.
6817 (movsi_cond_exec): Likewise.
6818 (clrsbsi2): Likewise.
6829 (call_i): Remove 'S' asm letter, add jli instruction.
6830 (call_value_i): Likewise.
6831 * config/arc/arc.op (mjli-always): New option.
6832 * config/arc/constraints.md (Cji): New constraint.
6833 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6835 (subsf3_fpx): Likewise.
6836 (mulsf3_fpx): Likewise.
6837 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6839 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6841 * doc/invoke.texi (ARC): Document mjli-always option.
6843 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6845 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6846 avoid addition with 0 and use incw and decw where possible.
6848 2018-01-26 Richard Biener <rguenther@suse.de>
6850 PR tree-optimization/81082
6851 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6852 association if it requires casting to unsigned.
6853 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6854 from fold_plusminus_mult_expr to catch important cases late when
6855 range info is available.
6857 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6859 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6860 * configure.ac (hidden_linkonce): New test.
6861 * configure: Regenerate.
6862 * config.in: Regenerate.
6864 2018-01-26 Julia Koval <julia.koval@intel.com>
6866 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6867 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6868 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6869 _mm_mask_bitshuffle_epi64_mask): Fix type.
6870 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6871 USI_FTYPE_V4DI_V4DI_USI): Remove.
6872 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6873 __builtin_ia32_vpshufbitqmb256_mask,
6874 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6875 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6876 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6878 2018-01-26 Alan Modra <amodra@gmail.com>
6881 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6882 UNSPEC_VBPERMQ. Sort other unspecs.
6884 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6886 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6888 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6891 * predict.c (drop_profile): Do not push/pop cfun; update also
6893 (handle_missing_profiles): Fix logic looking for zero profiles.
6895 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6898 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6899 on functions with #pragma omp declare simd or functions with simd
6901 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6902 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6903 Remove trailing \n from warning_at calls.
6905 2018-01-25 Tom de Vries <tom@codesourcery.com>
6908 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6909 for neutered workers.
6911 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6914 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6915 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6917 2018-01-24 Jeff Law <law@redhat.com>
6920 * i386.c (get_probe_interval): Move to earlier point.
6921 (ix86_compute_frame_layout): If -fstack-clash-protection and
6922 the frame is larger than the probe interval, then use pushes
6923 to save registers rather than reg->mem moves.
6924 (ix86_expand_prologue): Remove conditional for int_registers_saved
6927 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6930 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6931 min/max for never referenced object.
6933 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6936 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6938 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6939 attributes from DECL_ATTRIBUTES (decl) without affecting
6940 DECL_ATTRIBUTES (current_function_decl).
6941 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6942 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6944 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6946 PR tree-optimization/83979
6947 * fold-const.c (fold_comparison): Use constant_boolean_node
6948 instead of boolean_{true,false}_node.
6950 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6952 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6955 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6957 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6958 Simplify the clause that sets the length attribute.
6959 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6960 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6961 clause that sets the length attribute.
6962 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6964 2018-01-24 Tom de Vries <tom@codesourcery.com>
6967 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6968 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6969 Add strict parameter.
6970 (prevent_branch_around_nothing): Insert dummy insn between branch to
6971 label and label with no ptx insn inbetween.
6972 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6974 2018-01-24 Tom de Vries <tom@codesourcery.com>
6977 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6978 for neutered threads in warp.
6979 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6981 2018-01-24 Richard Biener <rguenther@suse.de>
6983 PR tree-optimization/83176
6984 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6987 2018-01-24 Richard Biener <rguenther@suse.de>
6989 PR tree-optimization/82819
6990 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6991 code generating pluses that are no-ops in the target precision.
6993 2018-01-24 Richard Biener <rguenther@suse.de>
6996 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6998 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7000 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7001 to merge probabilities.
7002 * predict.c (probably_never_executed): Also mark as cold functions
7003 with global 0 profile and guessed local profile.
7004 * profile-count.c (profile_probability::combine_with_count): New
7006 * profile-count.h (profile_probability::operator*,
7007 profile_probability::operator*=, profile_probability::operator/,
7008 profile_probability::operator/=): Reduce precision to adjusted
7009 and set value to guessed on contradictory divisions.
7010 (profile_probability::combine_with_freq): Remove.
7011 (profile_probability::combine_wiht_count): Declare.
7012 (profile_count::force_nonzero):: Set to adjusted.
7013 (profile_count::probability_in):: Set quality to adjusted.
7014 * tree-ssa-tail-merge.c (replace_block_by): Use
7017 2018-01-23 Andrew Waterman <andrew@sifive.com>
7018 Jim Wilson <jimw@sifive.com>
7020 * config/riscv/riscv.c (riscv_stack_boundary): New.
7021 (riscv_option_override): Set riscv_stack_boundary. Handle
7022 riscv_preferred_stack_boundary_arg.
7023 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7024 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7025 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7026 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7027 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7028 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7030 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7033 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7034 of struct ix86_frame.
7035 (ix86_expand_epilogue): Likewise. Add a local variable for
7036 the reg_save_offset field in struct ix86_frame.
7038 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7040 PR tree-optimization/82604
7041 * tree-loop-distribution.c (enum partition_kind): New enum item
7042 PKIND_PARTIAL_MEMSET.
7043 (partition_builtin_p): Support above new enum item.
7044 (generate_code_for_partition): Ditto.
7045 (compute_access_range): Differentiate cases that equality can be
7046 proven at all loops, the innermost loops or no loops.
7047 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7048 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7049 (finalize_partitions, distribute_loop): Don't fuse partition of
7050 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7051 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7054 2018-01-23 Martin Liska <mliska@suse.cz>
7056 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7057 order to ignore the predictor.
7058 (PRED_POLYMORPHIC_CALL): Likewise.
7059 (PRED_RECURSIVE_CALL): Likewise.
7061 2018-01-23 Martin Liska <mliska@suse.cz>
7063 * tree-profile.c (tree_profiling): Print function header to
7064 aware reader which function we are working on.
7065 * value-prof.c (gimple_find_values_to_profile): Do not print
7066 not interesting value histograms.
7068 2018-01-23 Martin Liska <mliska@suse.cz>
7070 * profile-count.h (enum profile_quality): Add
7071 profile_uninitialized as the first value. Do not number values
7072 as they are zero based.
7073 (profile_count::verify): Update sanity check.
7074 (profile_probability::verify): Likewise.
7076 2018-01-23 Nathan Sidwell <nathan@acm.org>
7078 * doc/invoke.texi (ffor-scope): Deprecate.
7080 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7082 PR tree-optimization/83510
7083 * domwalk.c (set_all_edges_as_executable): New function.
7084 (dom_walker::dom_walker): Convert bool param
7085 "skip_unreachable_blocks" to enum reachability. Move setup of
7086 edge flags to set_all_edges_as_executable and only do it when
7087 reachability is REACHABLE_BLOCKS.
7088 * domwalk.h (enum dom_walker::reachability): New enum.
7089 (dom_walker::dom_walker): Convert bool param
7090 "skip_unreachable_blocks" to enum reachability.
7091 (set_all_edges_as_executable): New decl.
7092 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7093 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7095 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7096 but converting true to REACHABLE_BLOCKS.
7097 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7099 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7100 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7101 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7103 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7104 if check_all_array_refs will be called.
7106 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7108 * tree.c (selftest::test_location_wrappers): Add more test
7111 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7113 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7114 (selftest::test_bit_in_range): Likewise.
7116 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7119 * doc/sourcebuild.texi (vect_float): Say that the selector
7120 only describes the situation when -funsafe-math-optimizations is on.
7121 (vect_float_strict): Document.
7123 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7125 PR tree-optimization/83965
7126 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7127 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7128 instead of checking only for a reduction.
7129 (vect_recog_widen_sum_pattern): Likewise.
7131 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7133 * predict.c (probably_never_executed): Only use precise profile info.
7134 (compute_function_frequency): Skip after inlining hack since we now
7135 have quality checking.
7137 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7139 * profile-count.h (profile_probability::very_unlikely,
7140 profile_probability::unlikely, profile_probability::even): Set
7141 precision to guessed.
7143 2018-01-23 Richard Biener <rguenther@suse.de>
7145 PR tree-optimization/83963
7146 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7147 Properly terminate dominator walk when crossing the exit edge not
7148 when visiting its source block.
7150 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7153 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7154 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7156 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7158 PR tree-optimization/83957
7159 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7160 semicolon after for body surrounded by braces.
7162 PR tree-optimization/83081
7163 * profile-count.h (profile_probability::split): New method.
7164 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7165 Use profile_probability::split.
7166 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7167 when splitting a single conditional jump into 2.
7169 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7171 PR tree-optimization/69452
7172 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7175 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7177 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7178 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7179 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7181 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7183 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7184 * config/rl78/rl78.md (movdi): New define_expand.
7185 * config/rl78/rl78.c (rl78_split_movdi): New function.
7187 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7190 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7192 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7193 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7194 128-bit to produce an UNSPEC move to get the double word with the
7195 signbit and then a shift directly to do signbit.
7196 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7197 implementation with a new version that just does either a direct
7198 move or a regular move. Move memory interface to separate insns.
7199 Move insns so they are next to the expander.
7200 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7201 with signbit move. Split big and little endian case.
7202 (signbit<mode>2_dm_mem_le): Likewise.
7203 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7204 (signbit<mode>2_dm2): Likewise.
7206 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7208 * config/rl78/rl78.md (anddi3): New define_expand.
7210 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7212 * config/rl78/rl78.md (umindi3): New define_expand.
7214 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7216 * config/rl78/rl78.md (smindi3): New define_expand.
7218 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7220 * config/rl78/rl78.md (smaxdi3): New define_expand.
7222 2018-01-22 Carl Love <cel@us.ibm.com>
7224 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7225 LVX_V1TI): Add macro expansion.
7226 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7227 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7228 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7229 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7230 Change check to determine if the instruction is a byte reversing
7231 entry. Fix typo in comment.
7232 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7233 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7234 Add def_builtin calls for new builtins.
7235 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7236 Add define_insn expansion.
7238 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7240 * config/rl78/rl78.md (umaxdi3): New define_expand.
7242 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7244 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7245 for non-QImode registers.
7247 2018-01-22 Richard Biener <rguenther@suse.de>
7249 PR tree-optimization/83963
7250 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7251 including the loop exit block.
7252 (scop_detection::merge_sese): Likewise.
7253 (scop_detection::add_scop): Do it here instead.
7255 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7257 * doc/sourcebuild.texi (arm_softfloat): Document.
7259 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7262 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7263 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7264 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7266 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7267 David Edelsohn <dje.gcc@gmail.com>
7270 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7271 Change "crset eq" to "crset 2".
7272 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7273 (*call_indirect_aix<mode>_nospec): Likewise.
7274 (*call_value_indirect_aix<mode>_nospec): Likewise.
7275 (*call_indirect_elfv2<mode>_nospec): Likewise.
7276 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7277 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7278 change assembly output from . to $.
7279 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7280 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7281 (*tablejump<mode>_internal1_nospec): Likewise.
7283 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7286 * config/sh/sh_optimize_sett_clrt.cc:
7287 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7289 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7291 PR tree-optimization/83940
7292 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7293 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7294 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7295 use it to pass back the definition type.
7296 (vect_check_store_rhs): Likewise rhs_dt_out.
7297 (vect_build_gather_load_calls): Add a mask_dt argument and use
7298 it instead of a call to vect_is_simple_use.
7299 (vectorizable_store): Update calls to vect_check_load_store_mask
7300 and vect_check_store_rhs. Use the dt returned by the latter instead
7301 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7302 instead of calls to vect_is_simple_use. Pass the scalar rather
7303 than the vector operand to vect_is_simple_use when handling
7304 second and subsequent copies of an rhs value.
7305 (vectorizable_load): Update calls to vect_check_load_store_mask
7306 and vect_build_gather_load_calls. Use the cached mask_dt and
7307 gs_info.offset_dt instead of calls to vect_is_simple_use.
7309 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7312 * tree-emutls.c: Include gimplify.h.
7313 (lower_emutls_2): New function.
7314 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7315 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7316 it before further processing.
7319 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7320 UINTVAL (trueop1) instead of INTVAL (op1).
7322 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7326 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7327 INCOMING_FRAME_SP_OFFSET if not defined.
7328 (scan_trace): Add ENTRY argument. If true and
7329 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7330 emit a note to adjust the CFA offset.
7331 (create_cfi_notes): Adjust scan_trace callers.
7332 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7333 INCOMING_FRAME_SP_OFFSET in the CIE.
7334 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7335 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7337 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7338 * doc/tm.texi: Regenerated.
7340 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7342 PR rtl-optimization/83147
7343 * lra-constraints.c (remove_inheritance_pseudos): Use
7344 lra_substitute_pseudo_within_insn.
7346 2018-01-19 Tom de Vries <tom@codesourcery.com>
7347 Cesar Philippidis <cesar@codesourcery.com>
7350 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7352 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7355 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7356 spaces for function labels.
7358 2018-01-19 Martin Liska <mliska@suse.cz>
7360 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7361 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7362 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7363 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7364 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7365 (PRED_CONST_RETURN): Change from 69 to 65.
7366 (PRED_NULL_RETURN): Change from 91 to 71.
7367 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7368 (PRED_LOOP_GUARD): Change from 66 to 73.
7370 2018-01-19 Martin Liska <mliska@suse.cz>
7372 * predict.c (predict_insn_def): Add new assert.
7373 (struct branch_predictor): Change type to signed integer.
7374 (test_prediction_value_range): Amend test to cover
7376 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7377 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7378 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7379 (PRED_LOOP_IV_COMPARE): Likewise.
7380 * predict.h (PROB_UNINITIALIZED): Define new constant.
7382 2018-01-19 Martin Liska <mliska@suse.cz>
7384 * predict.c (dump_prediction): Add new format for
7385 analyze_brprob.py script which is enabled with -details
7387 * profile-count.h (precise_p): New function.
7389 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7391 PR tree-optimization/83922
7392 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7393 there are no statements that need masking.
7394 (vect_active_double_reduction_p): New function.
7395 (vect_analyze_loop_operations): Use it when handling phis that
7396 are not in the loop header.
7398 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7400 PR tree-optimization/83914
7401 * tree-vect-loop.c (vectorizable_induction): Don't convert
7402 init_expr or apply the peeling adjustment for inductions
7403 that are nested within the vectorized loop.
7405 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7407 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7410 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7414 * function.h (gimplify_parameters): Add gimple_seq * argument.
7415 * function.c: Include gimple.h and options.h.
7416 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7417 for the added local temporaries if needed.
7418 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7419 if there are any parameter cleanups, wrap whole body into a
7420 try/finally with the cleanups.
7422 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7425 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7426 Use GET_MODE_CLASS for scalar floating point.
7428 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7432 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7433 Fix call of call_cgraph_insertion_hooks.
7435 2018-01-18 Martin Sebor <msebor@redhat.com>
7437 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7439 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7442 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7445 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7448 * common.opt: (-ffile-prefix-map): New option.
7449 * opts.c (common_handle_option): Defer it.
7450 * opts-global.c (handle_common_deferred_options): Handle it.
7451 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7452 * file-prefix-map.h: New file.
7453 (remap_debug_filename, add_debug_prefix_map): ...here.
7454 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7455 * final.c (debug_prefix_map, add_debug_prefix_map
7456 remap_debug_filename): Move to...
7457 * file-prefix-map.c: New file.
7458 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7459 generalize, get rid of alloca(), use strrchr() instead of strchr().
7460 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7461 Implement in terms of add_prefix_map().
7462 (remap_macro_filename, remap_debug_filename): Implement in term of
7464 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7465 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7466 * dbxout.c: Include file-prefix-map.h.
7467 * varasm.c: Likewise.
7468 * vmsdbgout.c: Likewise.
7469 * xcoffout.c: Likewise.
7470 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7471 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7472 * doc/invoke.texi (-ffile-prefix-map): Document.
7473 (-fdebug-prefix-map): Update description.
7475 2018-01-18 Martin Liska <mliska@suse.cz>
7477 * config/i386/i386.c (indirect_thunk_name): Document that also
7479 (output_indirect_thunk): Document why both instructions
7480 (pause and lfence) are generated.
7482 2018-01-18 Richard Biener <rguenther@suse.de>
7484 PR tree-optimization/83887
7485 * graphite-scop-detection.c
7486 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7487 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7488 (scop_detection::merge_sese): Re-implement with a flood-fill
7489 algorithm that properly finds a SESE region if it exists.
7491 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7494 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7495 pointer_diff optimizations use view_convert instead of convert.
7497 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7499 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7500 Generate different code for -mno-speculate-indirect-jumps.
7501 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7502 (*call_indirect_aix<mode>): Disable for
7503 -mno-speculate-indirect-jumps.
7504 (*call_indirect_aix<mode>_nospec): New define_insn.
7505 (*call_value_indirect_aix<mode>): Disable for
7506 -mno-speculate-indirect-jumps.
7507 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7508 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7509 -mno-speculate-indirect-jumps.
7510 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7512 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7514 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7515 long double type, set the flags for noting the default long double
7516 type, even if we don't pass or return a long double type.
7518 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7521 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7524 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7526 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7527 support for merge[hl].
7528 (fold_mergehl_helper): New helper function.
7529 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7530 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7531 (altivec_vmrglw_direct): Add xxmrglw insn.
7533 2018-01-17 Andrew Waterman <andrew@sifive.com>
7535 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7536 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7538 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7541 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7542 call the lto_location_cache before reading the
7543 DECL_SOURCE_LOCATION of the types.
7545 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7546 Richard Sandiford <richard.sandiford@linaro.org>
7548 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7549 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7550 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7551 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7552 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7554 * config/aarch64/constraints.md (aarch64_movti_operand):
7556 * config/aarch64/predicates.md (Uti): Add new constraint.
7558 2018-01-17 Carl Love <cel@us.ibm.com>
7560 * config/rs6000/vsx.md (define_expand xl_len_r,
7561 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7562 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7564 (define_expand, define_insn): Move the shift left from the
7565 define_insn to the define_expand for lxvl and stxvl instructions.
7566 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7567 and XL_LEN_R definitions to PURE.
7569 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7571 * config/i386/i386.c (indirect_thunk_name): Declare regno
7572 as unsigned int. Compare regno with INVALID_REGNUM.
7573 (output_indirect_thunk): Ditto.
7574 (output_indirect_thunk_function): Ditto.
7575 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7576 in the call to output_indirect_thunk_function.
7578 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7581 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7582 rather than the size of inner_type to determine the stack slot size
7583 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7585 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7588 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7591 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7593 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7594 endian Linux systems to optionally enable multilibs for selecting
7595 the long double type if the user configured an explicit type.
7596 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7597 have no long double multilibs if not defined.
7598 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7599 warn if the user used -mabi={ieee,ibm}longdouble and we built
7600 multilibs for long double.
7601 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7602 appropriate multilib option.
7603 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7605 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7606 for building long double multilibs.
7607 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7609 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7611 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7614 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7616 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7619 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7622 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7625 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7627 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7628 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7630 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7632 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7633 different rtl trees depending on TARGET_64BIT.
7634 (rs6000_gen_lvx): Likewise.
7636 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7638 * config/visium/visium.md (nop): Tweak comment.
7639 (hazard_nop): Likewise.
7641 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7643 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7644 -mspeculate-indirect-jumps.
7645 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7646 for -mno-speculate-indirect-jumps.
7647 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7648 (*call_value_indirect_elfv2<mode>): Disable for
7649 -mno-speculate-indirect-jumps.
7650 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7651 (indirect_jump): Emit different RTL for
7652 -mno-speculate-indirect-jumps.
7653 (*indirect_jump<mode>): Disable for
7654 -mno-speculate-indirect-jumps.
7655 (*indirect_jump<mode>_nospec): New define_insn.
7656 (tablejump): Emit different RTL for
7657 -mno-speculate-indirect-jumps.
7658 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7659 (tablejumpsi_nospec): New define_expand.
7660 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7661 (tablejumpdi_nospec): New define_expand.
7662 (*tablejump<mode>_internal1): Disable for
7663 -mno-speculate-indirect-jumps.
7664 (*tablejump<mode>_internal1_nospec): New define_insn.
7665 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7668 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7670 * caller-save.c (insert_save): Drop unnecessary parameter. All
7673 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7674 Richard Biener <rguenth@suse.de>
7677 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7678 return early, inline manually is_gimple_sizepos. Make sure if we
7679 call gimplify_expr we don't end up with a gimple constant.
7680 * tree.c (variably_modified_type_p): Don't return true for
7681 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7682 * gimplify.h (is_gimple_sizepos): Remove.
7684 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7686 PR tree-optimization/83857
7687 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7688 vectorizable_live_operation for pure SLP statements.
7689 (vectorizable_live_operation): Handle PHIs.
7691 2018-01-16 Richard Biener <rguenther@suse.de>
7693 PR tree-optimization/83867
7694 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7695 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7697 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7700 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7701 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7702 If off is not INTEGER_CST, issue a may not be aligned warning
7703 rather than isn't aligned. Use isn%'t rather than isn't.
7704 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7706 <case MULT_EXPR>: Improve the case when bottom and one of the
7707 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7708 operand, in that case check if the other operand is multiple of
7709 bottom divided by the INTEGER_CST operand.
7711 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7714 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7715 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7716 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7717 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7718 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7719 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7720 (pa_function_arg_size): New function.
7722 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7724 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7725 in a separate statement.
7727 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7729 PR tree-optimization/83847
7730 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7731 group gathers and scatters.
7733 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7735 PR rtl-optimization/86620
7736 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7738 PR rtl-optimization/83213
7739 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7740 to last if both are JUMP_INSNs.
7742 PR tree-optimization/83843
7743 * gimple-ssa-store-merging.c
7744 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7745 store_immediate_info for bswap/nop orig_stores.
7747 2018-01-15 Andrew Waterman <andrew@sifive.com>
7749 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7751 <UDIV>: Increase cost if !TARGET_DIV.
7753 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7755 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7756 (define_attr "cr_logical_3op"): New.
7757 (cceq_ior_compare): Adjust.
7758 (cceq_ior_compare_complement): Adjust.
7759 (*cceq_rev_compare): Adjust.
7760 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7761 (is_cracked_insn): Adjust.
7762 (insn_must_be_first_in_group): Adjust.
7763 * config/rs6000/40x.md: Adjust.
7764 * config/rs6000/440.md: Adjust.
7765 * config/rs6000/476.md: Adjust.
7766 * config/rs6000/601.md: Adjust.
7767 * config/rs6000/603.md: Adjust.
7768 * config/rs6000/6xx.md: Adjust.
7769 * config/rs6000/7450.md: Adjust.
7770 * config/rs6000/7xx.md: Adjust.
7771 * config/rs6000/8540.md: Adjust.
7772 * config/rs6000/cell.md: Adjust.
7773 * config/rs6000/e300c2c3.md: Adjust.
7774 * config/rs6000/e500mc.md: Adjust.
7775 * config/rs6000/e500mc64.md: Adjust.
7776 * config/rs6000/e5500.md: Adjust.
7777 * config/rs6000/e6500.md: Adjust.
7778 * config/rs6000/mpc.md: Adjust.
7779 * config/rs6000/power4.md: Adjust.
7780 * config/rs6000/power5.md: Adjust.
7781 * config/rs6000/power6.md: Adjust.
7782 * config/rs6000/power7.md: Adjust.
7783 * config/rs6000/power8.md: Adjust.
7784 * config/rs6000/power9.md: Adjust.
7785 * config/rs6000/rs64.md: Adjust.
7786 * config/rs6000/titan.md: Adjust.
7788 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7790 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7791 ix86_indirect_branch_register logic.
7793 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7795 * config/i386/constraints.md (Bs): Update
7796 ix86_indirect_branch_register check. Don't check
7797 ix86_indirect_branch_register with GOT_memory_operand.
7799 * config/i386/predicates.md (GOT_memory_operand): Don't check
7800 ix86_indirect_branch_register here.
7801 (GOT32_symbol_operand): Likewise.
7803 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7805 * config/i386/predicates.md (constant_call_address_operand):
7806 Rewrite ix86_indirect_branch_register logic.
7807 (sibcall_insn_operand): Likewise.
7809 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7811 * config/i386/constraints.md (Bs): Replace
7812 ix86_indirect_branch_thunk_register with
7813 ix86_indirect_branch_register.
7815 * config/i386/i386.md (indirect_jump): Likewise.
7816 (tablejump): Likewise.
7817 (*sibcall_memory): Likewise.
7818 (*sibcall_value_memory): Likewise.
7819 Peepholes of indirect call and jump via memory: Likewise.
7820 * config/i386/i386.opt: Likewise.
7821 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7822 (GOT_memory_operand): Likewise.
7823 (call_insn_operand): Likewise.
7824 (sibcall_insn_operand): Likewise.
7825 (GOT32_symbol_operand): Likewise.
7827 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7830 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7831 type rather than type addr's type points to.
7832 (expand_omp_atomic_mutex): Likewise.
7833 (expand_omp_atomic): Likewise.
7835 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7838 * config/i386/i386.c (output_indirect_thunk_function): Use
7839 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7840 for __x86_return_thunk.
7842 2018-01-15 Richard Biener <rguenther@suse.de>
7845 * expmed.c (extract_bit_field_1): Fix typo.
7847 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7850 * config/arm/iterators.md (VF): New mode iterator.
7851 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7852 Remove integer-related logic from pattern.
7853 (neon_vabd<mode>_3): Likewise.
7855 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7858 * common.opt (fstrict-overflow): No longer an alias.
7859 (fwrapv-pointer): New option.
7860 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7861 also for pointer types based on flag_wrapv_pointer.
7862 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7863 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7864 opts->x_flag_wrapv got set.
7865 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7866 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7867 POINTER_TYPE_OVERFLOW_UNDEFINED.
7868 * match.pd: Likewise in address comparison pattern.
7869 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7871 2018-01-15 Richard Biener <rguenther@suse.de>
7874 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7875 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7876 Reset type names to their identifier if their TYPE_DECL doesn't
7877 have linkage (and thus is used for ODR and devirt).
7878 (save_debug_info_for_decl): Remove.
7879 (save_debug_info_for_type): Likewise.
7880 (add_tree_to_fld_list): Adjust.
7881 * tree-pretty-print.c (dump_generic_node): Make dumping of
7882 type names more robust.
7884 2018-01-15 Richard Biener <rguenther@suse.de>
7886 * BASE-VER: Bump to 8.0.1.
7888 2018-01-14 Martin Sebor <msebor@redhat.com>
7891 * builtins.c (check_access): Avoid warning when the no-warning bit
7894 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7896 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7897 * ira-color (allocno_hard_regs_compare): Likewise.
7899 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7902 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7903 Use .pushsection/.popsection.
7905 2018-01-14 Martin Sebor <msebor@redhat.com>
7908 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7910 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7912 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7913 entry from extra_headers.
7914 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7915 extra_headers, make the list bitwise identical to the i?86-*-* one.
7917 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7919 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7920 -mcmodel=large with -mindirect-branch=thunk,
7921 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7922 -mfunction-return=thunk-extern.
7923 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7924 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7925 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7927 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7929 * config/i386/i386.c (print_reg): Print the name of the full
7930 integer register without '%'.
7931 (ix86_print_operand): Handle 'V'.
7932 * doc/extend.texi: Document 'V' modifier.
7934 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7936 * config/i386/constraints.md (Bs): Disallow memory operand for
7937 -mindirect-branch-register.
7939 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7940 (GOT_memory_operand): Likewise.
7941 (call_insn_operand): Likewise.
7942 (sibcall_insn_operand): Likewise.
7943 (GOT32_symbol_operand): Likewise.
7944 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7945 for -mindirect-branch-register.
7946 (tablejump): Likewise.
7947 (*sibcall_memory): Likewise.
7948 (*sibcall_value_memory): Likewise.
7949 Disallow peepholes of indirect call and jump via memory for
7950 -mindirect-branch-register.
7951 (*call_pop): Replace m with Bw.
7952 (*call_value_pop): Likewise.
7953 (*sibcall_pop_memory): Replace m with Bs.
7954 * config/i386/i386.opt (mindirect-branch-register): New option.
7955 * doc/invoke.texi: Document -mindirect-branch-register option.
7957 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7959 * config/i386/i386-protos.h (ix86_output_function_return): New.
7960 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7961 set function_return_type.
7962 (indirect_thunk_name): Add ret_p to indicate thunk for function
7964 (output_indirect_thunk_function): Pass false to
7965 indirect_thunk_name.
7966 (ix86_output_indirect_branch_via_reg): Likewise.
7967 (ix86_output_indirect_branch_via_push): Likewise.
7968 (output_indirect_thunk_function): Create alias for function
7969 return thunk if regno < 0.
7970 (ix86_output_function_return): New function.
7971 (ix86_handle_fndecl_attribute): Handle function_return.
7972 (ix86_attribute_table): Add function_return.
7973 * config/i386/i386.h (machine_function): Add
7974 function_return_type.
7975 * config/i386/i386.md (simple_return_internal): Use
7976 ix86_output_function_return.
7977 (simple_return_internal_long): Likewise.
7978 * config/i386/i386.opt (mfunction-return=): New option.
7979 (indirect_branch): Mention -mfunction-return=.
7980 * doc/extend.texi: Document function_return function attribute.
7981 * doc/invoke.texi: Document -mfunction-return= option.
7983 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7985 * config/i386/i386-opts.h (indirect_branch): New.
7986 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7987 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7988 with local indirect jump when converting indirect call and jump.
7989 (ix86_set_indirect_branch_type): New.
7990 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7991 (indirectlabelno): New.
7992 (indirect_thunk_needed): Likewise.
7993 (indirect_thunk_bnd_needed): Likewise.
7994 (indirect_thunks_used): Likewise.
7995 (indirect_thunks_bnd_used): Likewise.
7996 (INDIRECT_LABEL): Likewise.
7997 (indirect_thunk_name): Likewise.
7998 (output_indirect_thunk): Likewise.
7999 (output_indirect_thunk_function): Likewise.
8000 (ix86_output_indirect_branch_via_reg): Likewise.
8001 (ix86_output_indirect_branch_via_push): Likewise.
8002 (ix86_output_indirect_branch): Likewise.
8003 (ix86_output_indirect_jmp): Likewise.
8004 (ix86_code_end): Call output_indirect_thunk_function if needed.
8005 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8007 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8008 (ix86_attribute_table): Add indirect_branch.
8009 * config/i386/i386.h (machine_function): Add indirect_branch_type
8010 and has_local_indirect_jump.
8011 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8013 (tablejump): Likewise.
8014 (*indirect_jump): Use ix86_output_indirect_jmp.
8015 (*tablejump_1): Likewise.
8016 (simple_return_indirect_internal): Likewise.
8017 * config/i386/i386.opt (mindirect-branch=): New option.
8018 (indirect_branch): New.
8021 (thunk-inline): Likewise.
8022 (thunk-extern): Likewise.
8023 * doc/extend.texi: Document indirect_branch function attribute.
8024 * doc/invoke.texi: Document -mindirect-branch= option.
8026 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8029 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8031 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8033 * ipa-inline.c (want_inline_small_function_p): Return false if
8034 inlining has already failed with CIF_FINAL_ERROR.
8035 (update_caller_keys): Call want_inline_small_function_p before
8037 (update_callee_keys): Likewise.
8039 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8041 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8043 (rs6000_quadword_masked_address_p): Likewise.
8044 (quad_aligned_load_p): Likewise.
8045 (quad_aligned_store_p): Likewise.
8046 (const_load_sequence_p): Add comment to describe the outer-most loop.
8047 (mimic_memory_attributes_and_flags): New function.
8048 (rs6000_gen_stvx): Likewise.
8049 (replace_swapped_aligned_store): Likewise.
8050 (rs6000_gen_lvx): Likewise.
8051 (replace_swapped_aligned_load): Likewise.
8052 (replace_swapped_load_constant): Capitalize argument name in
8053 comment describing this function.
8054 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8055 and stores that access quad-word aligned addresses and replace
8056 with stvx or lvx instructions when appropriate.
8057 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8058 New function prototype.
8059 (rs6000_quadword_masked_address_p): Likewise.
8060 (rs6000_gen_lvx): Likewise.
8061 (rs6000_gen_stvx): Likewise.
8062 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8063 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8064 when memory address is aligned.
8065 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8066 this split to select lvx instruction when memory address is aligned.
8067 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8068 instruction when memory address is aligned.
8069 (*vsx_le_perm_load_v16qi): Likewise.
8070 (four unnamed splitters): Modify to select the stvx instruction
8071 when memory is aligned.
8073 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8075 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8076 which appears in the queue multiple times.
8078 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8079 Alan Hayward <alan.hayward@arm.com>
8080 David Sherwood <david.sherwood@arm.com>
8082 * tree-vectorizer.h (vec_lower_bound): New structure.
8083 (_loop_vec_info): Add check_nonzero and lower_bounds.
8084 (LOOP_VINFO_CHECK_NONZERO): New macro.
8085 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8086 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8087 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8088 fields. Make seg_len the distance travelled, not including the
8090 (dr_direction_indicator): Declare.
8091 (dr_zero_step_indicator): Likewise.
8092 (dr_known_forward_stride_p): Likewise.
8093 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8095 (runtime_alias_check_p): Allow runtime alias checks with
8097 (operator ==): Compare access_size and align.
8098 (prune_runtime_alias_test_list): Rework for new distinction between
8099 the access_size and seg_len.
8100 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8102 (get_segment_min_max): New function.
8103 (create_intersect_range_checks): Use it.
8104 (dr_step_indicator): New function.
8105 (dr_direction_indicator): Likewise.
8106 (dr_zero_step_indicator): Likewise.
8107 (dr_known_forward_stride_p): Likewise.
8108 * tree-loop-distribution.c (data_ref_segment_size): Return
8109 DR_STEP * (niters - 1).
8110 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8112 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8113 (vect_preserves_scalar_order_p): New function, split out from...
8114 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8115 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8116 (vect_vfa_access_size): New function.
8117 (vect_vfa_align): Likewise.
8118 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8119 (dump_lower_bound): New function.
8120 (vect_check_lower_bound): Likewise.
8121 (vect_small_gap_p): Likewise.
8122 (vectorizable_with_step_bound_p): Likewise.
8123 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8124 depencies if the vectorization factor is 1. Convert the checks
8125 for nonzero steps into checks on the bounds of DR_STEP. Try using
8126 a bunds check for variable steps if the minimum required step is
8127 relatively small. Update calls to the dr_with_seg_len
8128 constructor and to vect_compile_time_alias.
8129 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8131 (vect_loop_versioning): Call it.
8132 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8134 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8136 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8137 Alan Hayward <alan.hayward@arm.com>
8138 David Sherwood <david.sherwood@arm.com>
8140 * doc/sourcebuild.texi (vect_scatter_store): Document.
8141 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8143 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8145 * genopinit.c (main): Add supports_vec_scatter_store and
8146 supports_vec_scatter_store_cached to target_optabs.
8147 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8148 IFN_MASK_SCATTER_STORE.
8149 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8151 * internal-fn.h (internal_store_fn_p): Declare.
8152 (internal_fn_stored_value_index): Likewise.
8153 * internal-fn.c (scatter_store_direct): New macro.
8154 (expand_scatter_store_optab_fn): New function.
8155 (direct_scatter_store_optab_supported_p): New macro.
8156 (internal_store_fn_p): New function.
8157 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8158 IFN_MASK_SCATTER_STORE.
8159 (internal_fn_mask_index): Likewise.
8160 (internal_fn_stored_value_index): New function.
8161 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8163 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8164 * optabs-query.c (supports_vec_scatter_store_p): New function.
8165 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8166 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8167 true for scatter stores.
8168 (vect_gather_scatter_fn_p): Handle scatter stores too.
8169 (vect_check_gather_scatter): Consider using scatter stores if
8170 supports_vec_scatter_store_p.
8171 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8173 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8174 internal_fn_stored_value_index.
8175 (check_load_store_masking): Handle scatter stores too.
8176 (vect_get_store_rhs): Make public.
8177 (vectorizable_call): Use internal_store_fn_p.
8178 (vectorizable_store): Handle scatter store internal functions.
8179 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8180 when deciding whether the end of the group has been reached.
8181 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8182 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8183 (mask_scatter_store<mode>): New insns.
8185 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8186 Alan Hayward <alan.hayward@arm.com>
8187 David Sherwood <david.sherwood@arm.com>
8189 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8190 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8191 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8193 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8194 Use vect_truncate_gather_scatter_offset if we can't treat the
8195 operation as a normal gather load or scatter store.
8196 (get_group_load_store_type): Take the gather_scatter_info
8197 as argument. Try using a gather load or scatter store for
8198 single-element groups.
8199 (get_load_store_type): Update calls to get_group_load_store_type
8200 and vect_use_strided_gather_scatters_p.
8202 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8203 Alan Hayward <alan.hayward@arm.com>
8204 David Sherwood <david.sherwood@arm.com>
8206 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8207 optional tree argument.
8208 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8210 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8211 but continue to use the current value as a fallback.
8212 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8213 to compare the updates.
8214 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8215 (get_load_store_type): Use it when handling a strided access.
8216 (vect_get_strided_load_store_ops): New function.
8217 (vect_get_data_ptr_increment): Likewise.
8218 (vectorizable_load): Handle strided gather loads. Always pass
8219 a step to vect_create_data_ref_ptr and bump_vector_ptr.
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 (gather_load@var{m}): Document.
8226 (mask_gather_load@var{m}): Likewise.
8227 * genopinit.c (main): Add supports_vec_gather_load and
8228 supports_vec_gather_load_cached to target_optabs.
8229 * optabs-tree.c (init_tree_optimization_optabs): Use
8230 ggc_cleared_alloc to allocate target_optabs.
8231 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8232 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8234 * internal-fn.h (internal_load_fn_p): Declare.
8235 (internal_gather_scatter_fn_p): Likewise.
8236 (internal_fn_mask_index): Likewise.
8237 (internal_gather_scatter_fn_supported_p): Likewise.
8238 * internal-fn.c (gather_load_direct): New macro.
8239 (expand_gather_load_optab_fn): New function.
8240 (direct_gather_load_optab_supported_p): New macro.
8241 (direct_internal_fn_optab): New function.
8242 (internal_load_fn_p): Likewise.
8243 (internal_gather_scatter_fn_p): Likewise.
8244 (internal_fn_mask_index): Likewise.
8245 (internal_gather_scatter_fn_supported_p): Likewise.
8246 * optabs-query.c (supports_at_least_one_mode_p): New function.
8247 (supports_vec_gather_load_p): Likewise.
8248 * optabs-query.h (supports_vec_gather_load_p): Declare.
8249 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8250 and memory_type field.
8251 (NUM_PATTERNS): Bump to 15.
8252 * tree-vect-data-refs.c: Include internal-fn.h.
8253 (vect_gather_scatter_fn_p): New function.
8254 (vect_describe_gather_scatter_call): Likewise.
8255 (vect_check_gather_scatter): Try using internal functions for
8256 gather loads. Recognize existing calls to a gather load function.
8257 (vect_analyze_data_refs): Consider using gather loads if
8258 supports_vec_gather_load_p.
8259 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8260 (vect_get_gather_scatter_offset_type): Likewise.
8261 (vect_convert_mask_for_vectype): Likewise.
8262 (vect_add_conversion_to_patterm): Likewise.
8263 (vect_try_gather_scatter_pattern): Likewise.
8264 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8265 (vect_vect_recog_func_ptrs): Add it.
8266 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8267 internal_fn_mask_index and internal_gather_scatter_fn_p.
8268 (check_load_store_masking): Take the gather_scatter_info as an
8269 argument and handle gather loads.
8270 (vect_get_gather_scatter_ops): New function.
8271 (vectorizable_call): Check internal_load_fn_p.
8272 (vectorizable_load): Likewise. Handle gather load internal
8274 (vectorizable_store): Update call to check_load_store_masking.
8275 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8276 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8277 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8278 (aarch64_gather_scale_operand_d): New predicates.
8279 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8280 (mask_gather_load<mode>): New insns.
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 * optabs.def (fold_left_plus_optab): New optab.
8287 * doc/md.texi (fold_left_plus_@var{m}): Document.
8288 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8289 * internal-fn.c (fold_left_direct): Define.
8290 (expand_fold_left_optab_fn): Likewise.
8291 (direct_fold_left_optab_supported_p): Likewise.
8292 * fold-const-call.c (fold_const_fold_left): New function.
8293 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8294 * tree-parloops.c (valid_reduction_p): New function.
8295 (gather_scalar_reductions): Use it.
8296 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8297 (vect_finish_replace_stmt): Declare.
8298 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8299 (needs_fold_left_reduction_p): New function, split out from...
8300 (vect_is_simple_reduction): ...here. Accept reductions that
8301 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8302 (vect_force_simple_reduction): Also store the reduction type in
8303 the assignment's STMT_VINFO_REDUC_TYPE.
8304 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8305 (merge_with_identity): New function.
8306 (vect_expand_fold_left): Likewise.
8307 (vectorize_fold_left_reduction): Likewise.
8308 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8309 scalar phi in place for it. Check for target support and reject
8310 cases that would reassociate the operation. Defer the transform
8311 phase to vectorize_fold_left_reduction.
8312 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8313 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8314 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8316 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8318 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8319 call to ifc_temp_var.
8321 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8322 Alan Hayward <alan.hayward@arm.com>
8323 David Sherwood <david.sherwood@arm.com>
8325 * target.def (legitimize_address_displacement): Take the original
8326 offset as a poly_int.
8327 * targhooks.h (default_legitimize_address_displacement): Update
8329 * targhooks.c (default_legitimize_address_displacement): Likewise.
8330 * doc/tm.texi: Regenerate.
8331 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8332 as an argument, moving assert of ad->disp == ad->disp_term to...
8333 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8334 Try calling targetm.legitimize_address_displacement before expanding
8335 the address rather than afterwards, and adjust for the new interface.
8336 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8337 Match the new hook interface. Handle SVE addresses.
8338 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8341 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8343 * Makefile.in (OBJS): Add early-remat.o.
8344 * target.def (select_early_remat_modes): New hook.
8345 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8346 * doc/tm.texi: Regenerate.
8347 * targhooks.h (default_select_early_remat_modes): Declare.
8348 * targhooks.c (default_select_early_remat_modes): New function.
8349 * timevar.def (TV_EARLY_REMAT): New timevar.
8350 * passes.def (pass_early_remat): New pass.
8351 * tree-pass.h (make_pass_early_remat): Declare.
8352 * early-remat.c: New file.
8353 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8355 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8357 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8358 Alan Hayward <alan.hayward@arm.com>
8359 David Sherwood <david.sherwood@arm.com>
8361 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8362 vfm1 with a bound_epilog parameter.
8363 (vect_do_peeling): Update calls accordingly, and move the prologue
8364 call earlier in the function. Treat the base bound_epilog as 0 for
8365 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8366 this base when peeling for gaps.
8367 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8368 with fully-masked loops.
8369 (vect_estimate_min_profitable_iters): Handle the single peeled
8370 iteration in that case.
8372 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8373 Alan Hayward <alan.hayward@arm.com>
8374 David Sherwood <david.sherwood@arm.com>
8376 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8377 single-element interleaving even if the size is not a power of 2.
8378 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8379 accesses for single-element interleaving if the group size is
8382 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8383 Alan Hayward <alan.hayward@arm.com>
8384 David Sherwood <david.sherwood@arm.com>
8386 * doc/md.texi (fold_extract_last_@var{m}): Document.
8387 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8388 * optabs.def (fold_extract_last_optab): New optab.
8389 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8390 * internal-fn.c (fold_extract_direct): New macro.
8391 (expand_fold_extract_optab_fn): Likewise.
8392 (direct_fold_extract_optab_supported_p): Likewise.
8393 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8394 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8395 EXTRACT_LAST_REDUCTION.
8396 (get_initial_def_for_reduction): Do not create an initial vector
8397 for EXTRACT_LAST_REDUCTION reductions.
8398 (vectorizable_reduction): Leave the scalar phi in place for
8399 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8400 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8401 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8402 transform phase to vectorizable_condition.
8403 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8405 (vect_finish_stmt_generation): ...here.
8406 (vect_finish_replace_stmt): New function.
8407 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8408 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8410 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8412 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8413 Alan Hayward <alan.hayward@arm.com>
8414 David Sherwood <david.sherwood@arm.com>
8416 * doc/md.texi (extract_last_@var{m}): Document.
8417 * optabs.def (extract_last_optab): New optab.
8418 * internal-fn.def (EXTRACT_LAST): New internal function.
8419 * internal-fn.c (cond_unary_direct): New macro.
8420 (expand_cond_unary_optab_fn): Likewise.
8421 (direct_cond_unary_optab_supported_p): Likewise.
8422 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8423 loops using EXTRACT_LAST.
8424 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8425 (extract_last_<mode>): ...this optab.
8426 (vec_extract<mode><Vel>): Update accordingly.
8428 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8429 Alan Hayward <alan.hayward@arm.com>
8430 David Sherwood <david.sherwood@arm.com>
8432 * target.def (empty_mask_is_expensive): New hook.
8433 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8434 * doc/tm.texi: Regenerate.
8435 * targhooks.h (default_empty_mask_is_expensive): Declare.
8436 * targhooks.c (default_empty_mask_is_expensive): New function.
8437 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8438 if the target says that empty masks are expensive.
8439 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8441 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8443 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8444 Alan Hayward <alan.hayward@arm.com>
8445 David Sherwood <david.sherwood@arm.com>
8447 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8448 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8449 (vect_use_loop_mask_for_alignment_p): New function.
8450 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8451 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8452 niters_skip argument. Make sure that the first niters_skip elements
8453 of the first iteration are inactive.
8454 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8455 Update call to vect_set_loop_masks_directly.
8456 (get_misalign_in_elems): New function, split out from...
8457 (vect_gen_prolog_loop_niters): ...here.
8458 (vect_update_init_of_dr): Take a code argument that specifies whether
8459 the adjustment should be added or subtracted.
8460 (vect_update_init_of_drs): Likewise.
8461 (vect_prepare_for_masked_peels): New function.
8462 (vect_do_peeling): Skip prologue peeling if we're using a mask
8463 instead. Update call to vect_update_inits_of_drs.
8464 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8466 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8467 alignment. Do not include the number of peeled iterations in
8468 the minimum threshold in that case.
8469 (vectorizable_induction): Adjust the start value down by
8470 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8471 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8472 Take the number of skipped iterations into account when calculating
8474 * tree-vect-stmts.c (vect_gen_while_not): New function.
8476 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8477 Alan Hayward <alan.hayward@arm.com>
8478 David Sherwood <david.sherwood@arm.com>
8480 * doc/sourcebuild.texi (vect_fully_masked): Document.
8481 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8483 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8485 (vect_analyze_loop_2): ...here. Don't check the vectorization
8486 factor against the number of loop iterations if the loop is
8489 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8490 Alan Hayward <alan.hayward@arm.com>
8491 David Sherwood <david.sherwood@arm.com>
8493 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8494 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8495 (dump_groups): Update accordingly.
8496 (iv_use::mem_type): New member variable.
8497 (address_p): New function.
8498 (record_use): Add a mem_type argument and initialize the new
8500 (record_group_use): Add a mem_type argument. Use address_p.
8501 Remove obsolete null checks of base_object. Update call to record_use.
8502 (find_interesting_uses_op): Update call to record_group_use.
8503 (find_interesting_uses_cond): Likewise.
8504 (find_interesting_uses_address): Likewise.
8505 (get_mem_type_for_internal_fn): New function.
8506 (find_address_like_use): Likewise.
8507 (find_interesting_uses_stmt): Try find_address_like_use before
8508 calling find_interesting_uses_op.
8509 (addr_offset_valid_p): Use the iv mem_type field as the type
8510 of the addressed memory.
8511 (add_autoinc_candidates): Likewise.
8512 (get_address_cost): Likewise.
8513 (split_small_address_groups_p): Use address_p.
8514 (split_address_groups): Likewise.
8515 (add_iv_candidate_for_use): Likewise.
8516 (autoinc_possible_for_pair): Likewise.
8517 (rewrite_groups): Likewise.
8518 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8519 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8520 (get_alias_ptr_type_for_ptr_address): New function.
8521 (rewrite_use_address): Rewrite address uses in calls that were
8522 identified by find_address_like_use.
8524 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8525 Alan Hayward <alan.hayward@arm.com>
8526 David Sherwood <david.sherwood@arm.com>
8528 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8530 * gimple-expr.h (is_gimple_addressable: Likewise.
8531 * gimple-expr.c (is_gimple_address): Likewise.
8532 * internal-fn.c (expand_call_mem_ref): New function.
8533 (expand_mask_load_optab_fn): Use it.
8534 (expand_mask_store_optab_fn): Likewise.
8536 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8537 Alan Hayward <alan.hayward@arm.com>
8538 David Sherwood <david.sherwood@arm.com>
8540 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8541 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8542 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8543 (cond_umax@var{mode}): Document.
8544 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8545 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8546 (cond_umin_optab, cond_umax_optab): New optabs.
8547 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8548 (COND_IOR, COND_XOR): New internal functions.
8549 * internal-fn.h (get_conditional_internal_fn): Declare.
8550 * internal-fn.c (cond_binary_direct): New macro.
8551 (expand_cond_binary_optab_fn): Likewise.
8552 (direct_cond_binary_optab_supported_p): Likewise.
8553 (get_conditional_internal_fn): New function.
8554 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8555 Cope with reduction statements that are vectorized as calls rather
8557 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8558 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8559 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8560 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8561 (UNSPEC_COND_EOR): New unspecs.
8562 (optab): Add mappings for them.
8563 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8564 (sve_int_op, sve_fp_op): New int attributes.
8566 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8567 Alan Hayward <alan.hayward@arm.com>
8568 David Sherwood <david.sherwood@arm.com>
8570 * optabs.def (while_ult_optab): New optab.
8571 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8572 * internal-fn.def (WHILE_ULT): New internal function.
8573 * internal-fn.h (direct_internal_fn_supported_p): New override
8574 that takes two types as argument.
8575 * internal-fn.c (while_direct): New macro.
8576 (expand_while_optab_fn): New function.
8577 (convert_optab_supported_p): Likewise.
8578 (direct_while_optab_supported_p): New macro.
8579 * wide-int.h (wi::udiv_ceil): New function.
8580 * tree-vectorizer.h (rgroup_masks): New structure.
8581 (vec_loop_masks): New typedef.
8582 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8584 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8585 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8586 (vect_max_vf): New function.
8587 (slpeel_make_loop_iterate_ntimes): Delete.
8588 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8589 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8590 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8591 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8592 internal-fn.h, stor-layout.h and optabs-query.h.
8593 (vect_set_loop_mask): New function.
8594 (add_preheader_seq): Likewise.
8595 (add_header_seq): Likewise.
8596 (interleave_supported_p): Likewise.
8597 (vect_maybe_permute_loop_masks): Likewise.
8598 (vect_set_loop_masks_directly): Likewise.
8599 (vect_set_loop_condition_masked): Likewise.
8600 (vect_set_loop_condition_unmasked): New function, split out from
8601 slpeel_make_loop_iterate_ntimes.
8602 (slpeel_make_loop_iterate_ntimes): Rename to..
8603 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8604 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8605 (vect_do_peeling): Update call accordingly.
8606 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8608 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8609 mask_compare_type, can_fully_mask_p and fully_masked_p.
8610 (release_vec_loop_masks): New function.
8611 (_loop_vec_info): Use it to free the loop masks.
8612 (can_produce_all_loop_masks_p): New function.
8613 (vect_get_max_nscalars_per_iter): Likewise.
8614 (vect_verify_full_masking): Likewise.
8615 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8616 retries, and free the mask rgroups before retrying. Check loop-wide
8617 reasons for disallowing fully-masked loops. Make the final decision
8618 about whether use a fully-masked loop or not.
8619 (vect_estimate_min_profitable_iters): Do not assume that peeling
8620 for the number of iterations will be needed for fully-masked loops.
8621 (vectorizable_reduction): Disable fully-masked loops.
8622 (vectorizable_live_operation): Likewise.
8623 (vect_halve_mask_nunits): New function.
8624 (vect_double_mask_nunits): Likewise.
8625 (vect_record_loop_mask): Likewise.
8626 (vect_get_loop_mask): Likewise.
8627 (vect_transform_loop): Handle the case in which the final loop
8628 iteration might handle a partial vector. Call vect_set_loop_condition
8629 instead of slpeel_make_loop_iterate_ntimes.
8630 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8631 (check_load_store_masking): New function.
8632 (prepare_load_store_mask): Likewise.
8633 (vectorizable_store): Handle fully-masked loops.
8634 (vectorizable_load): Likewise.
8635 (supportable_widening_operation): Use vect_halve_mask_nunits for
8637 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8638 (vect_gen_while): New function.
8639 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8640 (aarch64_uqdec<mode>): New insn.
8642 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8643 Alan Hayward <alan.hayward@arm.com>
8644 David Sherwood <david.sherwood@arm.com>
8646 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8647 (reduc_xor_scal_optab): New optabs.
8648 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8649 (reduc_xor_scal_@var{m}): Document.
8650 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8651 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8653 * fold-const-call.c (fold_const_call): Handle them.
8654 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8655 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8656 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8657 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8658 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8659 (UNSPEC_XORV): New unspecs.
8660 (optab): Add entries for them.
8661 (BITWISEV): New int iterator.
8662 (bit_reduc_op): New int attributes.
8664 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8665 Alan Hayward <alan.hayward@arm.com>
8666 David Sherwood <david.sherwood@arm.com>
8668 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8669 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8670 * optabs.def (vec_shl_insert_optab): New optab.
8671 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8672 (duplicate_and_interleave): Likewise.
8673 * tree-vect-loop.c: Include internal-fn.h.
8674 (neutral_op_for_slp_reduction): New function, split out from
8675 get_initial_defs_for_reduction.
8676 (get_initial_def_for_reduction): Handle option 2 for variable-length
8677 vectors by loading the neutral value into a vector and then shifting
8678 the initial value into element 0.
8679 (get_initial_defs_for_reduction): Replace the code argument with
8680 the neutral value calculated by neutral_op_for_slp_reduction.
8681 Use gimple_build_vector for constant-length vectors.
8682 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8683 but the first group_size elements have a neutral value.
8684 Use duplicate_and_interleave otherwise.
8685 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8686 Update call to get_initial_defs_for_reduction. Handle SLP
8687 reductions for variable-length vectors by creating one vector
8688 result for each scalar result, with the elements associated
8689 with other scalar results stubbed out with the neutral value.
8690 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8691 Require IFN_VEC_SHL_INSERT for double reductions on
8692 variable-length vectors, or SLP reductions that have
8693 a neutral value. Require can_duplicate_and_interleave_p
8694 support for variable-length unchained SLP reductions if there
8695 is no neutral value, such as for MIN/MAX reductions. Also require
8696 the number of vector elements to be a multiple of the number of
8697 SLP statements when doing variable-length unchained SLP reductions.
8698 Update call to vect_create_epilog_for_reduction.
8699 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8700 and remove initial values.
8701 (duplicate_and_interleave): Make public.
8702 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8703 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8705 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8706 Alan Hayward <alan.hayward@arm.com>
8707 David Sherwood <david.sherwood@arm.com>
8709 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8710 (can_duplicate_and_interleave_p): New function.
8711 (vect_get_and_check_slp_defs): Take the vector of statements
8712 rather than just the current one. Remove excess parentheses.
8713 Restriction rejectinon of vect_constant_def and vect_external_def
8714 for variable-length vectors to boolean types, or types for which
8715 can_duplicate_and_interleave_p is false.
8716 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8717 (duplicate_and_interleave): New function.
8718 (vect_get_constant_vectors): Use gimple_build_vector for
8719 constant-length vectors and suitable variable-length constant
8720 vectors. Use duplicate_and_interleave for other variable-length
8721 vectors. Don't defer the update when inserting new statements.
8723 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8724 Alan Hayward <alan.hayward@arm.com>
8725 David Sherwood <david.sherwood@arm.com>
8727 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8728 min_profitable_iters doesn't go negative.
8730 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8731 Alan Hayward <alan.hayward@arm.com>
8732 David Sherwood <david.sherwood@arm.com>
8734 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8735 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8736 * optabs.def (vec_mask_load_lanes_optab): New optab.
8737 (vec_mask_store_lanes_optab): Likewise.
8738 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8739 (MASK_STORE_LANES): Likewise.
8740 * internal-fn.c (mask_load_lanes_direct): New macro.
8741 (mask_store_lanes_direct): Likewise.
8742 (expand_mask_load_optab_fn): Handle masked operations.
8743 (expand_mask_load_lanes_optab_fn): New macro.
8744 (expand_mask_store_optab_fn): Handle masked operations.
8745 (expand_mask_store_lanes_optab_fn): New macro.
8746 (direct_mask_load_lanes_optab_supported_p): Likewise.
8747 (direct_mask_store_lanes_optab_supported_p): Likewise.
8748 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8750 (vect_load_lanes_supported): Likewise.
8751 * tree-vect-data-refs.c (strip_conversion): New function.
8752 (can_group_stmts_p): Likewise.
8753 (vect_analyze_data_ref_accesses): Use it instead of checking
8754 for a pair of assignments.
8755 (vect_store_lanes_supported): Take a masked_p parameter.
8756 (vect_load_lanes_supported): Likewise.
8757 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8758 vect_store_lanes_supported and vect_load_lanes_supported.
8759 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8760 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8761 parameter. Don't allow gaps for masked accesses.
8762 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8763 and vect_load_lanes_supported.
8764 (get_load_store_type): Take a masked_p parameter and update
8765 call to get_group_load_store_type.
8766 (vectorizable_store): Update call to get_load_store_type.
8767 Handle IFN_MASK_STORE_LANES.
8768 (vectorizable_load): Update call to get_load_store_type.
8769 Handle IFN_MASK_LOAD_LANES.
8771 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8772 Alan Hayward <alan.hayward@arm.com>
8773 David Sherwood <david.sherwood@arm.com>
8775 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8777 * config/aarch64/aarch64-protos.h
8778 (aarch64_sve_struct_memory_operand_p): Declare.
8779 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8780 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8781 (VPRED, vpred): Handle SVE structure modes.
8782 * config/aarch64/constraints.md (Utx): New constraint.
8783 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8784 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8785 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8786 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8787 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8788 structure modes. Split into pieces after RA.
8789 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8790 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8792 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8793 SVE structure modes.
8794 (aarch64_classify_address): Likewise.
8795 (sizetochar): Move earlier in file.
8796 (aarch64_print_operand): Handle SVE register lists.
8797 (aarch64_array_mode): New function.
8798 (aarch64_sve_struct_memory_operand_p): Likewise.
8799 (TARGET_ARRAY_MODE): Redefine.
8801 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8802 Alan Hayward <alan.hayward@arm.com>
8803 David Sherwood <david.sherwood@arm.com>
8805 * target.def (array_mode): New target hook.
8806 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8807 * doc/tm.texi: Regenerate.
8808 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8809 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8810 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8812 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8815 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8816 Alan Hayward <alan.hayward@arm.com>
8817 David Sherwood <david.sherwood@arm.com>
8819 * fold-const.c (fold_binary_loc): Check the argument types
8820 rather than the result type when testing for a vector operation.
8822 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8824 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8825 * doc/tm.texi: Regenerate.
8827 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8828 Alan Hayward <alan.hayward@arm.com>
8829 David Sherwood <david.sherwood@arm.com>
8831 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8832 (sve): Document new AArch64 extension.
8833 * doc/md.texi (w): Extend the description of the AArch64
8834 constraint to include SVE vectors.
8835 (Upl, Upa): Document new AArch64 predicate constraints.
8836 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8838 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8839 (msve-vector-bits=): New option.
8840 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8841 SVE when these are disabled.
8842 (sve): New extension.
8843 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8844 modes. Adjust their number of units based on aarch64_sve_vg.
8845 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8846 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8847 aarch64_addr_query_type.
8848 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8849 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8850 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8851 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8852 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8853 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8854 (aarch64_simd_imm_zero_p): Delete.
8855 (aarch64_check_zero_based_sve_index_immediate): Declare.
8856 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8857 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8858 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8859 (aarch64_sve_float_mul_immediate_p): Likewise.
8860 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8862 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8863 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8864 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8865 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8866 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8867 (aarch64_regmode_natural_size): Likewise.
8868 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8869 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8871 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8872 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8873 for VG and the SVE predicate registers.
8874 (V_ALIASES): Add a "z"-prefixed alias.
8875 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8876 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8877 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8878 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8879 (REG_CLASS_NAMES): Add entries for them.
8880 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8881 and the predicate registers.
8882 (aarch64_sve_vg): Declare.
8883 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8884 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8885 (REGMODE_NATURAL_SIZE): Define.
8886 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8888 * config/aarch64/aarch64.c: Include cfgrtl.h.
8889 (simd_immediate_info): Add a constructor for series vectors,
8890 and an associated step field.
8891 (aarch64_sve_vg): New variable.
8892 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8893 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8894 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8895 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8896 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8897 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8898 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8899 (aarch64_get_mask_mode): New functions.
8900 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8901 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8902 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8903 predicate modes and predicate registers. Explicitly restrict
8904 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8905 to store a vector mode if it is recognized by
8906 aarch64_classify_vector_mode.
8907 (aarch64_regmode_natural_size): New function.
8908 (aarch64_hard_regno_caller_save_mode): Return the original mode
8910 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8911 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8912 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8913 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8915 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8916 does not overlap dest if the function is frame-related. Handle
8918 (aarch64_split_add_offset): New function.
8919 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8920 them aarch64_add_offset.
8921 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8922 and update call to aarch64_sub_sp.
8923 (aarch64_add_cfa_expression): New function.
8924 (aarch64_expand_prologue): Pass extra temporary registers to the
8925 functions above. Handle the case in which we need to emit new
8926 DW_CFA_expressions for registers that were originally saved
8927 relative to the stack pointer, but now have to be expressed
8928 relative to the frame pointer.
8929 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8931 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8932 IP0 and IP1 values for SVE frames.
8933 (aarch64_expand_vec_series): New function.
8934 (aarch64_expand_sve_widened_duplicate): Likewise.
8935 (aarch64_expand_sve_const_vector): Likewise.
8936 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8937 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8938 into the register, rather than emitting a SET directly.
8939 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8940 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8941 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8942 (offset_9bit_signed_scaled_p): New functions.
8943 (aarch64_replicate_bitmask_imm): New function.
8944 (aarch64_bitmask_imm): Use it.
8945 (aarch64_cannot_force_const_mem): Reject expressions involving
8946 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8947 (aarch64_classify_index): Handle SVE indices, by requiring
8948 a plain register index with a scale that matches the element size.
8949 (aarch64_classify_address): Handle SVE addresses. Assert that
8950 the mode of the address is VOIDmode or an integer mode.
8951 Update call to aarch64_classify_symbol.
8952 (aarch64_classify_symbolic_expression): Update call to
8953 aarch64_classify_symbol.
8954 (aarch64_const_vec_all_in_range_p): New function.
8955 (aarch64_print_vector_float_operand): Likewise.
8956 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8957 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8958 and the FP immediates 1.0 and 0.5.
8959 (aarch64_print_address_internal): Handle SVE addresses.
8960 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8961 (aarch64_regno_regclass): Handle predicate registers.
8962 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8964 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8965 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8966 (aarch64_convert_sve_vector_bits): New function.
8967 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8968 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8970 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8971 Handle SVE vector and predicate modes. Accept VL-based constants
8972 that need only one temporary register, and VL offsets that require
8973 no temporary registers.
8974 (aarch64_conditional_register_usage): Mark the predicate registers
8975 as fixed if SVE isn't available.
8976 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8977 Return true for SVE vector and predicate modes.
8978 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8979 rather than an unsigned int. Handle SVE modes.
8980 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8982 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8984 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8985 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8986 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8987 (aarch64_sve_float_mul_immediate_p): New functions.
8988 (aarch64_sve_valid_immediate): New function.
8989 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8990 Explicitly reject structure modes. Check for INDEX constants.
8991 Handle PTRUE and PFALSE constants.
8992 (aarch64_check_zero_based_sve_index_immediate): New function.
8993 (aarch64_simd_imm_zero_p): Delete.
8994 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8995 vector modes. Accept constants in the range of CNT[BHWD].
8996 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8997 ask for an Advanced SIMD mode.
8998 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8999 (aarch64_simd_vector_alignment): Handle SVE predicates.
9000 (aarch64_vectorize_preferred_vector_alignment): New function.
9001 (aarch64_simd_vector_alignment_reachable): Use it instead of
9003 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9004 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9006 (MAX_VECT_LEN): Delete.
9007 (expand_vec_perm_d): Add a vec_flags field.
9008 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9009 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9010 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9012 (aarch64_evpc_rev): Rename to...
9013 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9014 (aarch64_evpc_rev_global): New function.
9015 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9016 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9018 (aarch64_evpc_sve_tbl): New function.
9019 (aarch64_expand_vec_perm_const_1): Update after rename of
9020 aarch64_evpc_rev. Handle SVE permutes too, trying
9021 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9022 than aarch64_evpc_tbl.
9023 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9024 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9025 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9026 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9027 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9028 (aarch64_expand_sve_vcond): New functions.
9029 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9030 of aarch64_vector_mode_p.
9031 (aarch64_dwarf_poly_indeterminate_value): New function.
9032 (aarch64_compute_pressure_classes): Likewise.
9033 (aarch64_can_change_mode_class): Likewise.
9034 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9035 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9036 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9037 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9038 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9039 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9040 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9041 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9043 (Dn, Dl, Dr): Accept const as well as const_vector.
9044 (Dz): Likewise. Compare against CONST0_RTX.
9045 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9046 of "vector" where appropriate.
9047 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9048 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9049 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9050 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9051 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9052 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9053 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9054 (v_int_equiv): Extend to SVE modes.
9055 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9057 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9058 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9059 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9060 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9061 (SVE_COND_FP_CMP): New int iterators.
9062 (perm_hilo): Handle the new unpack unspecs.
9063 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9065 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9066 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9067 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9068 (aarch64_equality_operator, aarch64_constant_vector_operand)
9069 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9070 (aarch64_sve_nonimmediate_operand): Likewise.
9071 (aarch64_sve_general_operand): Likewise.
9072 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9073 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9074 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9075 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9076 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9077 (aarch64_sve_float_arith_immediate): Likewise.
9078 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9079 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9080 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9081 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9082 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9083 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9084 (aarch64_sve_float_arith_operand): Likewise.
9085 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9086 (aarch64_sve_float_mul_operand): Likewise.
9087 (aarch64_sve_vec_perm_operand): Likewise.
9088 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9089 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9090 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9091 as well as const_vector.
9092 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9093 in file. Use CONST0_RTX and CONSTM1_RTX.
9094 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9095 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9096 Use aarch64_simd_imm_zero.
9097 * config/aarch64/aarch64-sve.md: New file.
9098 * config/aarch64/aarch64.md: Include it.
9099 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9100 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9101 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9102 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9103 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9104 (sve): New attribute.
9105 (enabled): Disable instructions with the sve attribute unless
9107 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9108 aarch64_expand_mov_immediate.
9109 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9110 CNT[BHSD] immediates.
9111 (movti): Split CONST_POLY_INT moves into two halves.
9112 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9113 Split additions that need a temporary here if the destination
9114 is the stack pointer.
9115 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9116 (*add<mode>3_poly_1): New instruction.
9117 (set_clobber_cc): New expander.
9119 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9121 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9122 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9123 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9124 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9125 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9126 Change innermode from fixed_mode_size to machine_mode.
9127 (simplify_subreg): Update call accordingly. Handle a constant-sized
9128 subreg of a variable-length CONST_VECTOR.
9130 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9131 Alan Hayward <alan.hayward@arm.com>
9132 David Sherwood <david.sherwood@arm.com>
9134 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9135 (add_offset_to_base): New function, split out from...
9136 (create_mem_ref): ...here. When handling a scale other than 1,
9137 check first whether the address is valid without the offset.
9138 Add it into the base if so, leaving the index and scale as-is.
9140 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9143 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9144 fold_for_warn before checking if arg2 is INTEGER_CST.
9146 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9148 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9149 (store_multiple_operation): Delete.
9150 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9151 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9152 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9153 guarded by TARGET_STRING.
9154 (rs6000_output_load_multiple): Delete.
9155 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9156 OPTION_MASK_STRING / TARGET_STRING handling.
9157 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9158 (const rs6000_opt_masks) <"string">: Change mask to 0.
9159 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9160 (MASK_STRING): Delete.
9161 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9163 (load_multiple): Delete.
9170 (store_multiple): Delete.
9177 (movmemsi_8reg): Delete.
9178 (corresponding unnamed define_insn): Delete.
9179 (movmemsi_6reg): Delete.
9180 (corresponding unnamed define_insn): Delete.
9181 (movmemsi_4reg): Delete.
9182 (corresponding unnamed define_insn): Delete.
9183 (movmemsi_2reg): Delete.
9184 (corresponding unnamed define_insn): Delete.
9185 (movmemsi_1reg): Delete.
9186 (corresponding unnamed define_insn): Delete.
9187 * config/rs6000/rs6000.opt (mno-string): New.
9188 (mstring): Replace by deprecation warning stub.
9189 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9191 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9193 * regrename.c (regrename_do_replace): If replacing the same
9194 reg multiple times, try to reuse last created gen_raw_REG.
9197 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9198 main to workaround a bug in GDB.
9200 2018-01-12 Tom de Vries <tom@codesourcery.com>
9203 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9205 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9207 PR rtl-optimization/80481
9208 * ira-color.c (get_cap_member): New function.
9209 (allocnos_conflict_by_live_ranges_p): Use it.
9210 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9211 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9213 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9216 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9217 (*saddl_se_1): Ditto.
9219 (*ssubl_se_1): Ditto.
9221 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9223 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9224 rather than wi::to_widest for DR_INITs.
9225 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9226 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9227 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9229 (vect_analyze_group_access_1): Note that here.
9231 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9233 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9234 polynomial type sizes.
9236 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9238 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9239 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9240 (gimple_add_tmp_var): Likewise.
9242 2018-01-12 Martin Liska <mliska@suse.cz>
9244 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9245 (gimple_alloc_sizes): Likewise.
9246 (dump_gimple_statistics): Use PRIu64 in printf format.
9247 * gimple.h: Change uint64_t to int.
9249 2018-01-12 Martin Liska <mliska@suse.cz>
9251 * tree-core.h: Use uint64_t instead of int.
9252 * tree.c (tree_node_counts): Likewise.
9253 (tree_node_sizes): Likewise.
9254 (dump_tree_statistics): Use PRIu64 in printf format.
9256 2018-01-12 Martin Liska <mliska@suse.cz>
9258 * Makefile.in: As qsort_chk is implemented in vec.c, add
9259 vec.o to linkage of gencfn-macros.
9260 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9261 passing the info to record_node_allocation_statistics.
9262 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9264 * ggc-common.c (struct ggc_usage): Add operator== and use
9265 it in operator< and compare function.
9266 * mem-stats.h (struct mem_usage): Likewise.
9267 * vec.c (struct vec_usage): Remove operator< and compare
9268 function. Can be simply inherited.
9270 2018-01-12 Martin Jambor <mjambor@suse.cz>
9273 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9274 * tree-ssa-math-opts.c: Include domwalk.h.
9275 (convert_mult_to_fma_1): New function.
9276 (fma_transformation_info): New type.
9277 (fma_deferring_state): Likewise.
9278 (cancel_fma_deferring): New function.
9279 (result_of_phi): Likewise.
9280 (last_fma_candidate_feeds_initial_phi): Likewise.
9281 (convert_mult_to_fma): Added deferring logic, split actual
9282 transformation to convert_mult_to_fma_1.
9283 (math_opts_dom_walker): New type.
9284 (math_opts_dom_walker::after_dom_children): New method, body moved
9285 here from pass_optimize_widening_mul::execute, added deferring logic
9287 (pass_optimize_widening_mul::execute): Moved most of code to
9288 math_opts_dom_walker::after_dom_children.
9289 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9290 * config/i386/i386.c (ix86_option_override_internal): Added
9291 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9293 2018-01-12 Richard Biener <rguenther@suse.de>
9296 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9297 inline instance vars.
9299 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9302 * config/rx/rx.c (rx_is_restricted_memory_address):
9305 2018-01-12 Richard Biener <rguenther@suse.de>
9307 PR tree-optimization/80846
9308 * target.def (split_reduction): New target hook.
9309 * targhooks.c (default_split_reduction): New function.
9310 * targhooks.h (default_split_reduction): Declare.
9311 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9312 target requests first reduce vectors by combining low and high
9314 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9315 (get_vectype_for_scalar_type_and_size): Export.
9316 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9317 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9318 * doc/tm.texi: Regenerate.
9319 * config/i386/i386.c (ix86_split_reduction): Implement
9320 TARGET_VECTORIZE_SPLIT_REDUCTION.
9322 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9325 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9326 in PIC mode except for TARGET_VXWORKS_RTP.
9327 * config/sparc/sparc.c: Include cfgrtl.h.
9328 (TARGET_INIT_PIC_REG): Define.
9329 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9330 (sparc_pic_register_p): New predicate.
9331 (sparc_legitimate_address_p): Use it.
9332 (sparc_legitimize_pic_address): Likewise.
9333 (sparc_delegitimize_address): Likewise.
9334 (sparc_mode_dependent_address_p): Likewise.
9335 (gen_load_pcrel_sym): Remove 4th parameter.
9336 (load_got_register): Adjust call to above. Remove obsolete stuff.
9337 (sparc_expand_prologue): Do not call load_got_register here.
9338 (sparc_flat_expand_prologue): Likewise.
9339 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9340 (sparc_use_pseudo_pic_reg): New function.
9341 (sparc_init_pic_reg): Likewise.
9342 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9343 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9345 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9347 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9348 Add item for branch_cost.
9350 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9352 PR rtl-optimization/83565
9353 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9354 not extend the result to a larger mode for rotate operations.
9355 (num_sign_bit_copies1): Likewise.
9357 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9360 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9362 Use values-Xc.o for -pedantic.
9363 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9365 2018-01-12 Martin Liska <mliska@suse.cz>
9368 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9370 (possible_polymorphic_call_targets): Use it.
9371 (ipa_devirt): Likewise.
9373 2018-01-12 Martin Liska <mliska@suse.cz>
9375 * profile-count.h (enum profile_quality): Use 0 as invalid
9376 enum value of profile_quality.
9378 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9380 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9381 -mext-string options.
9383 2018-01-12 Richard Biener <rguenther@suse.de>
9385 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9386 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9387 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9389 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9391 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9393 * configure.ac (--with-long-double-format): Add support for the
9394 configuration option to change the default long double format on
9396 * config.gcc (powerpc*-linux*-*): Likewise.
9397 * configure: Regenerate.
9398 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9399 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9400 used without modification.
9402 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9404 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9405 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9406 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9407 MISC_BUILTIN_SPEC_BARRIER.
9408 (rs6000_init_builtins): Likewise.
9409 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9411 (speculation_barrier): New define_insn.
9412 * doc/extend.texi: Document __builtin_speculation_barrier.
9414 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9417 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9418 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9419 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9421 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9422 integral modes instead of "ss" and "sd".
9423 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9424 vectors with 32-bit and 64-bit elements.
9425 (vecdupssescalarmodesuffix): New mode attribute.
9426 (vec_dup<mode>): Use it.
9428 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9431 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9432 frame if argument is passed on stack.
9434 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9437 * ree.c (combine_reaching_defs): Optimize also
9438 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9439 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9441 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9444 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9446 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9449 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9450 after they are computed.
9452 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9454 PR tree-optimization/83695
9455 * gimple-loop-linterchange.cc
9456 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9457 reset cached scev information after interchange.
9458 (pass_linterchange::execute): Remove call to scev_reset_htab.
9460 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9462 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9463 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9464 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9465 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9466 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9467 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9468 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9469 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9470 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9471 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9472 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9473 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9474 (V_lane_reg): Likewise.
9475 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9477 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9478 (vfmal_lane_low<mode>_intrinsic,
9479 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9480 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9481 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9482 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9483 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9484 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9486 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9488 * config/arm/arm-cpus.in (fp16fml): New feature.
9489 (ALL_SIMD): Add fp16fml.
9490 (armv8.2-a): Add fp16fml as an option.
9491 (armv8.3-a): Likewise.
9492 (armv8.4-a): Add fp16fml as part of fp16.
9493 * config/arm/arm.h (TARGET_FP16FML): Define.
9494 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9496 * config/arm/arm-modes.def (V2HF): Define.
9497 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9498 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9499 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9500 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9501 vfmsl_low, vfmsl_high): New set of builtins.
9502 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9503 (vfml_op): New code attribute.
9504 (VFMLHALVES): New int iterator.
9505 (VFML, VFMLSEL): New mode attributes.
9506 (V_reg): Define mapping for V2HF.
9507 (V_hi, V_lo): New mode attributes.
9508 (VF_constraint): Likewise.
9509 (vfml_half, vfml_half_selector): New int attributes.
9510 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9512 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9513 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9515 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9516 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9517 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9518 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9520 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9521 Document new effective target and option set.
9523 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9525 * config/arm/arm-cpus.in (armv8_4): New feature.
9526 (ARMv8_4a): New fgroup.
9527 (armv8.4-a): New arch.
9528 * config/arm/arm-tables.opt: Regenerate.
9529 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9530 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9531 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9532 Add matching rules for -march=armv8.4-a and extensions.
9533 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9535 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9538 * config/rx/rx.md (BW): New mode attribute.
9539 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9541 2018-01-11 Richard Biener <rguenther@suse.de>
9543 PR tree-optimization/83435
9544 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9545 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9546 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9548 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9549 Alan Hayward <alan.hayward@arm.com>
9550 David Sherwood <david.sherwood@arm.com>
9552 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9554 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9555 (aarch64_print_address_internal): Use it to check for a zero offset.
9557 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9558 Alan Hayward <alan.hayward@arm.com>
9559 David Sherwood <david.sherwood@arm.com>
9561 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9562 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9563 Return a poly_int64 rather than a HOST_WIDE_INT.
9564 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9565 rather than a HOST_WIDE_INT.
9566 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9567 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9568 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9569 final_offset from HOST_WIDE_INT to poly_int64.
9570 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9571 to_constant when getting the number of units in an Advanced SIMD
9573 (aarch64_builtin_vectorized_function): Check for a constant number
9575 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9577 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9578 attribute instead of GET_MODE_NUNITS.
9579 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9580 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9581 GET_MODE_SIZE for fixed-size registers.
9582 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9583 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9584 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9585 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9586 (aarch64_print_operand, aarch64_print_address_internal)
9587 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9588 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9589 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9590 Handle polynomial GET_MODE_SIZE.
9591 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9592 wider than SImode without modification.
9593 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9594 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9595 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9596 passing and returning SVE modes.
9597 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9598 rather than GEN_INT.
9599 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9600 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9601 (aarch64_allocate_and_probe_stack_space): Likewise.
9602 (aarch64_layout_frame): Cope with polynomial offsets.
9603 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9604 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9606 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9607 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9608 poly_int64 rather than a HOST_WIDE_INT.
9609 (aarch64_get_separate_components, aarch64_process_components)
9610 (aarch64_expand_prologue, aarch64_expand_epilogue)
9611 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9612 (aarch64_anchor_offset): New function, split out from...
9613 (aarch64_legitimize_address): ...here.
9614 (aarch64_builtin_vectorization_cost): Handle polynomial
9615 TYPE_VECTOR_SUBPARTS.
9616 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9618 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9619 number of elements from the PARALLEL rather than the mode.
9620 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9621 rather than GET_MODE_BITSIZE.
9622 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9623 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9624 (aarch64_expand_vec_perm_const_1): Handle polynomial
9625 d->perm.length () and d->perm elements.
9626 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9627 Apply to_constant to d->perm elements.
9628 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9629 polynomial CONST_VECTOR_NUNITS.
9630 (aarch64_move_pointer): Take amount as a poly_int64 rather
9632 (aarch64_progress_pointer): Avoid temporary variable.
9633 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9634 the mode attribute instead of GET_MODE.
9636 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9637 Alan Hayward <alan.hayward@arm.com>
9638 David Sherwood <david.sherwood@arm.com>
9640 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9641 x exists before using it.
9642 (aarch64_add_constant_internal): Rename to...
9643 (aarch64_add_offset_1): ...this. Replace regnum with separate
9644 src and dest rtxes. Handle the case in which they're different,
9645 including when the offset is zero. Replace scratchreg with an rtx.
9646 Use 2 additions if there is no spare register into which we can
9647 move a 16-bit constant.
9648 (aarch64_add_constant): Delete.
9649 (aarch64_add_offset): Replace reg with separate src and dest
9650 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9651 Use aarch64_add_offset_1.
9652 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9653 an rtx rather than an int. Take the delta as a poly_int64
9654 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9655 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9656 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9657 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9658 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9660 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9661 aarch64_add_constant.
9663 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9665 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9666 Use scalar_float_mode.
9668 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9670 * config/aarch64/aarch64-simd.md
9671 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9672 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9673 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9674 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9675 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9676 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9677 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9678 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9679 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9680 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9682 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9685 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9686 targ_options->x_arm_arch_string is non NULL.
9688 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9690 * config/aarch64/aarch64.h
9691 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9693 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9696 * expmed.c (emit_store_flag_force): Swap if const op0
9697 and change VOIDmode to mode of op0.
9699 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9701 PR rtl-optimization/83761
9702 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9703 than bytes to mode_for_size.
9705 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9708 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9709 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9712 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9715 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9716 when in layout mode.
9717 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9718 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9721 2018-01-10 Michael Collison <michael.collison@arm.com>
9723 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9724 * config/aarch64/aarch64-option-extension.def: Add
9725 AARCH64_OPT_EXTENSION of 'fp16fml'.
9726 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9727 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9728 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9729 * config/aarch64/constraints.md (Ui7): New constraint.
9730 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9731 (VFMLA_SEL_W): Ditto.
9734 (VFMLA16_LOW): New int iterator.
9735 (VFMLA16_HIGH): Ditto.
9736 (UNSPEC_FMLAL): New unspec.
9737 (UNSPEC_FMLSL): Ditto.
9738 (UNSPEC_FMLAL2): Ditto.
9739 (UNSPEC_FMLSL2): Ditto.
9740 (f16mac): New code attribute.
9741 * config/aarch64/aarch64-simd-builtins.def
9742 (aarch64_fmlal_lowv2sf): Ditto.
9743 (aarch64_fmlsl_lowv2sf): Ditto.
9744 (aarch64_fmlalq_lowv4sf): Ditto.
9745 (aarch64_fmlslq_lowv4sf): Ditto.
9746 (aarch64_fmlal_highv2sf): Ditto.
9747 (aarch64_fmlsl_highv2sf): Ditto.
9748 (aarch64_fmlalq_highv4sf): Ditto.
9749 (aarch64_fmlslq_highv4sf): Ditto.
9750 (aarch64_fmlal_lane_lowv2sf): Ditto.
9751 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9752 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9753 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9754 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9755 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9756 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9757 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9758 (aarch64_fmlal_lane_highv2sf): Ditto.
9759 (aarch64_fmlsl_lane_highv2sf): Ditto.
9760 (aarch64_fmlal_laneq_highv2sf): Ditto.
9761 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9762 (aarch64_fmlalq_lane_highv4sf): Ditto.
9763 (aarch64_fmlsl_lane_highv4sf): Ditto.
9764 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9765 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9766 * config/aarch64/aarch64-simd.md:
9767 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9768 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9769 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9770 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9771 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9772 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9773 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9774 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9775 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9776 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9777 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9778 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9779 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9780 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9781 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9782 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9783 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9784 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9785 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9786 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9787 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9788 (vfmlsl_low_u32): Ditto.
9789 (vfmlalq_low_u32): Ditto.
9790 (vfmlslq_low_u32): Ditto.
9791 (vfmlal_high_u32): Ditto.
9792 (vfmlsl_high_u32): Ditto.
9793 (vfmlalq_high_u32): Ditto.
9794 (vfmlslq_high_u32): Ditto.
9795 (vfmlal_lane_low_u32): Ditto.
9796 (vfmlsl_lane_low_u32): Ditto.
9797 (vfmlal_laneq_low_u32): Ditto.
9798 (vfmlsl_laneq_low_u32): Ditto.
9799 (vfmlalq_lane_low_u32): Ditto.
9800 (vfmlslq_lane_low_u32): Ditto.
9801 (vfmlalq_laneq_low_u32): Ditto.
9802 (vfmlslq_laneq_low_u32): Ditto.
9803 (vfmlal_lane_high_u32): Ditto.
9804 (vfmlsl_lane_high_u32): Ditto.
9805 (vfmlal_laneq_high_u32): Ditto.
9806 (vfmlsl_laneq_high_u32): Ditto.
9807 (vfmlalq_lane_high_u32): Ditto.
9808 (vfmlslq_lane_high_u32): Ditto.
9809 (vfmlalq_laneq_high_u32): Ditto.
9810 (vfmlslq_laneq_high_u32): Ditto.
9811 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9812 (AARCH64_FL_FOR_ARCH8_4): New.
9813 (AARCH64_ISA_F16FML): New ISA flag.
9814 (TARGET_F16FML): New feature flag for fp16fml.
9815 (doc/invoke.texi): Document new fp16fml option.
9817 2018-01-10 Michael Collison <michael.collison@arm.com>
9819 * config/aarch64/aarch64-builtins.c:
9820 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9821 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9822 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9823 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9824 (AARCH64_ISA_SHA3): New ISA flag.
9825 (TARGET_SHA3): New feature flag for sha3.
9826 * config/aarch64/iterators.md (sha512_op): New int attribute.
9827 (CRYPTO_SHA512): New int iterator.
9828 (UNSPEC_SHA512H): New unspec.
9829 (UNSPEC_SHA512H2): Ditto.
9830 (UNSPEC_SHA512SU0): Ditto.
9831 (UNSPEC_SHA512SU1): Ditto.
9832 * config/aarch64/aarch64-simd-builtins.def
9833 (aarch64_crypto_sha512hqv2di): New builtin.
9834 (aarch64_crypto_sha512h2qv2di): Ditto.
9835 (aarch64_crypto_sha512su0qv2di): Ditto.
9836 (aarch64_crypto_sha512su1qv2di): Ditto.
9837 (aarch64_eor3qv8hi): Ditto.
9838 (aarch64_rax1qv2di): Ditto.
9839 (aarch64_xarqv2di): Ditto.
9840 (aarch64_bcaxqv8hi): Ditto.
9841 * config/aarch64/aarch64-simd.md:
9842 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9843 (aarch64_crypto_sha512su0qv2di): Ditto.
9844 (aarch64_crypto_sha512su1qv2di): Ditto.
9845 (aarch64_eor3qv8hi): Ditto.
9846 (aarch64_rax1qv2di): Ditto.
9847 (aarch64_xarqv2di): Ditto.
9848 (aarch64_bcaxqv8hi): Ditto.
9849 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9850 (vsha512h2q_u64): Ditto.
9851 (vsha512su0q_u64): Ditto.
9852 (vsha512su1q_u64): Ditto.
9853 (veor3q_u16): Ditto.
9854 (vrax1q_u64): Ditto.
9856 (vbcaxq_u16): Ditto.
9857 * config/arm/types.md (crypto_sha512): New type attribute.
9858 (crypto_sha3): Ditto.
9859 (doc/invoke.texi): Document new sha3 option.
9861 2018-01-10 Michael Collison <michael.collison@arm.com>
9863 * config/aarch64/aarch64-builtins.c:
9864 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9865 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9866 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9867 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9868 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9869 (AARCH64_ISA_SM4): New ISA flag.
9870 (TARGET_SM4): New feature flag for sm4.
9871 * config/aarch64/aarch64-simd-builtins.def
9872 (aarch64_sm3ss1qv4si): Ditto.
9873 (aarch64_sm3tt1aq4si): Ditto.
9874 (aarch64_sm3tt1bq4si): Ditto.
9875 (aarch64_sm3tt2aq4si): Ditto.
9876 (aarch64_sm3tt2bq4si): Ditto.
9877 (aarch64_sm3partw1qv4si): Ditto.
9878 (aarch64_sm3partw2qv4si): Ditto.
9879 (aarch64_sm4eqv4si): Ditto.
9880 (aarch64_sm4ekeyqv4si): Ditto.
9881 * config/aarch64/aarch64-simd.md:
9882 (aarch64_sm3ss1qv4si): Ditto.
9883 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9884 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9885 (aarch64_sm4eqv4si): Ditto.
9886 (aarch64_sm4ekeyqv4si): Ditto.
9887 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9888 (sm3part_op): Ditto.
9889 (CRYPTO_SM3TT): Ditto.
9890 (CRYPTO_SM3PART): Ditto.
9891 (UNSPEC_SM3SS1): New unspec.
9892 (UNSPEC_SM3TT1A): Ditto.
9893 (UNSPEC_SM3TT1B): Ditto.
9894 (UNSPEC_SM3TT2A): Ditto.
9895 (UNSPEC_SM3TT2B): Ditto.
9896 (UNSPEC_SM3PARTW1): Ditto.
9897 (UNSPEC_SM3PARTW2): Ditto.
9898 (UNSPEC_SM4E): Ditto.
9899 (UNSPEC_SM4EKEY): Ditto.
9900 * config/aarch64/constraints.md (Ui2): New constraint.
9901 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9902 * config/arm/types.md (crypto_sm3): New type attribute.
9903 (crypto_sm4): Ditto.
9904 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9905 (vsm3tt1aq_u32): Ditto.
9906 (vsm3tt1bq_u32): Ditto.
9907 (vsm3tt2aq_u32): Ditto.
9908 (vsm3tt2bq_u32): Ditto.
9909 (vsm3partw1q_u32): Ditto.
9910 (vsm3partw2q_u32): Ditto.
9911 (vsm4eq_u32): Ditto.
9912 (vsm4ekeyq_u32): Ditto.
9913 (doc/invoke.texi): Document new sm4 option.
9915 2018-01-10 Michael Collison <michael.collison@arm.com>
9917 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9918 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9919 (AARCH64_FL_FOR_ARCH8_4): New.
9920 (AARCH64_FL_V8_4): New flag.
9921 (doc/invoke.texi): Document new armv8.4-a option.
9923 2018-01-10 Michael Collison <michael.collison@arm.com>
9925 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9926 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9927 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9928 * config/aarch64/aarch64-option-extension.def: Add
9929 AARCH64_OPT_EXTENSION of 'sha2'.
9930 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9931 (crypto): Disable sha2 and aes if crypto disabled.
9932 (crypto): Enable aes and sha2 if enabled.
9933 (simd): Disable sha2 and aes if simd disabled.
9934 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9936 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9937 (TARGET_SHA2): New feature flag for sha2.
9938 (TARGET_AES): New feature flag for aes.
9939 * config/aarch64/aarch64-simd.md:
9940 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9941 conditional on TARGET_AES.
9942 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9943 (aarch64_crypto_sha1hsi): Make pattern conditional
9945 (aarch64_crypto_sha1hv4si): Ditto.
9946 (aarch64_be_crypto_sha1hv4si): Ditto.
9947 (aarch64_crypto_sha1su1v4si): Ditto.
9948 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9949 (aarch64_crypto_sha1su0v4si): Ditto.
9950 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9951 (aarch64_crypto_sha256su0v4si): Ditto.
9952 (aarch64_crypto_sha256su1v4si): Ditto.
9953 (doc/invoke.texi): Document new aes and sha2 options.
9955 2018-01-10 Martin Sebor <msebor@redhat.com>
9957 PR tree-optimization/83781
9958 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9961 2018-01-11 Martin Sebor <msebor@gmail.com>
9962 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9964 PR tree-optimization/83501
9965 PR tree-optimization/81703
9967 * tree-ssa-strlen.c (get_string_cst): Rename...
9968 (get_string_len): ...to this. Handle global constants.
9969 (handle_char_store): Adjust.
9971 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9972 Jim Wilson <jimw@sifive.com>
9974 * config/riscv/riscv-protos.h (riscv_output_return): New.
9975 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9976 (riscv_attribute_table, riscv_output_return),
9977 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9978 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9979 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9981 (riscv_expand_prologue): Add early return for naked function.
9982 (riscv_expand_epilogue): Likewise.
9983 (riscv_function_ok_for_sibcall): Return false for naked function.
9984 (riscv_set_current_function): New.
9985 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9986 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9987 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9988 * doc/extend.texi (RISC-V Function Attributes): New.
9990 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9992 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9993 check for 128-bit long double before checking TCmode.
9994 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9995 128-bit long doubles before checking TFmode or TCmode.
9996 (FLOAT128_IBM_P): Likewise.
9998 2018-01-10 Martin Sebor <msebor@redhat.com>
10000 PR tree-optimization/83671
10001 * builtins.c (c_strlen): Unconditionally return zero for the empty
10003 Use -Warray-bounds for warnings.
10004 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10005 for non-constant array indices with COMPONENT_REF, arrays of
10006 arrays, and pointers to arrays.
10007 (gimple_fold_builtin_strlen): Determine and set length range for
10008 non-constant character arrays.
10010 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10012 PR middle-end/81897
10013 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10016 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10018 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10020 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10023 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10024 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10025 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10026 indexed_or_indirect_operand predicate.
10027 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10028 (*vsx_le_perm_load_v8hi): Likewise.
10029 (*vsx_le_perm_load_v16qi): Likewise.
10030 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10031 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10032 (*vsx_le_perm_store_v8hi): Likewise.
10033 (*vsx_le_perm_store_v16qi): Likewise.
10034 (eight unnamed splitters): Likewise.
10036 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10038 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10039 * config/rs6000/emmintrin.h: Likewise.
10040 * config/rs6000/mmintrin.h: Likewise.
10041 * config/rs6000/xmmintrin.h: Likewise.
10043 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10046 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10048 * tree.c (tree_nop_conversion): Return true for location wrapper
10050 (maybe_wrap_with_location): New function.
10051 (selftest::check_strip_nops): New function.
10052 (selftest::test_location_wrappers): New function.
10053 (selftest::tree_c_tests): Call it.
10054 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10055 (maybe_wrap_with_location): New decl.
10056 (EXPR_LOCATION_WRAPPER_P): New macro.
10057 (location_wrapper_p): New inline function.
10058 (tree_strip_any_location_wrapper): New inline function.
10060 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10063 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10064 stack_realign_offset for the largest alignment of stack slot
10066 (ix86_find_max_used_stack_alignment): New function.
10067 (ix86_finalize_stack_frame_flags): Use it. Set
10068 max_used_stack_alignment if we don't realign stack.
10069 * config/i386/i386.h (machine_function): Add
10070 max_used_stack_alignment.
10072 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10074 * config/arm/arm.opt (-mbranch-cost): New option.
10075 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10078 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10081 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10082 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10084 2018-01-10 Richard Biener <rguenther@suse.de>
10087 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10088 early out so it also covers the case where we have a non-NULL
10091 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10093 PR tree-optimization/83753
10094 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10095 for non-strided grouped accesses if the number of elements is 1.
10097 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10100 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10101 * i386.h (TARGET_USE_GATHER): Define.
10102 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10104 2018-01-10 Martin Liska <mliska@suse.cz>
10107 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10108 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10110 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10111 CLEANUP_NO_PARTITIONING is not set.
10113 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10115 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10116 for vectors, as a partial revert of r254296.
10117 * rtl.h (const_vec_p): Delete.
10118 (const_vec_duplicate_p): Don't test for vector CONSTs.
10119 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10120 * expmed.c (make_tree): Likewise.
10123 * common.md (E, F): Use CONSTANT_P instead of checking for
10125 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10126 checking for CONST_VECTOR.
10128 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10130 PR middle-end/83575
10131 * predict.c (force_edge_cold): Handle in more sane way edges
10132 with no prediction.
10134 2018-01-09 Carl Love <cel@us.ibm.com>
10136 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10138 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10139 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10140 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10141 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10142 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10143 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10144 * config/rs6000/rs6000-protos.h: Add extern defition for
10145 rs6000_generate_float2_double_code.
10146 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10148 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10149 (float2_v2df): Add define_expand.
10151 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10154 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10155 op_mode in the force_to_mode call.
10157 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10159 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10160 instead of checking each element individually.
10161 (aarch64_evpc_uzp): Likewise.
10162 (aarch64_evpc_zip): Likewise.
10163 (aarch64_evpc_ext): Likewise.
10164 (aarch64_evpc_rev): Likewise.
10165 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10166 instead of checking each element individually. Return true without
10168 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10169 whether all selected elements come from the same input, instead of
10170 checking each element individually. Remove calls to gen_rtx_REG,
10171 start_sequence and end_sequence and instead assert that no rtl is
10174 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10176 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10177 order of HIGH and CONST checks.
10179 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10181 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10182 if the destination isn't an SSA_NAME.
10184 2018-01-09 Richard Biener <rguenther@suse.de>
10186 PR tree-optimization/83668
10187 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10189 (canonicalize_loop_form): ... here, renamed from ...
10190 (canonicalize_loop_closed_ssa_form): ... this and amended to
10191 swap successor edges for loop exit blocks to make us use
10192 the RPO order we need for initial schedule generation.
10194 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10196 PR tree-optimization/64811
10197 * match.pd: When optimizing comparisons with Inf, avoid
10198 introducing or losing exceptions from comparisons with NaN.
10200 2018-01-09 Martin Liska <mliska@suse.cz>
10203 * asan.c (shadow_mem_size): Add gcc_assert.
10205 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10207 Don't save registers in main().
10210 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10211 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10212 * config/avr/avr.c (avr_set_current_function): Don't error if
10213 naked, OS_task or OS_main are specified at the same time.
10214 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10216 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10218 * common/config/avr/avr-common.c (avr_option_optimization_table):
10219 Switch on -mmain-is-OS_task for optimizing compilations.
10221 2018-01-09 Richard Biener <rguenther@suse.de>
10223 PR tree-optimization/83572
10224 * graphite.c: Include cfganal.h.
10225 (graphite_transform_loops): Connect infinite loops to exit
10226 and remove fake edges at the end.
10228 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10230 * ipa-inline.c (edge_badness): Revert accidental checkin.
10232 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10235 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10236 symbols; not inline clones.
10238 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10241 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10242 hard registers. Formatting fixes.
10244 PR preprocessor/83722
10245 * gcc.c (try_generate_repro): Pass
10246 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10247 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10250 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10251 Kito Cheng <kito.cheng@gmail.com>
10253 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10254 (riscv_leaf_function_p): Delete.
10255 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10257 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10259 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10261 (do_ifelse): New function.
10262 (do_isel): New function.
10263 (do_sub3): New function.
10264 (do_add3): New function.
10265 (do_load_mask_compare): New function.
10266 (do_overlap_load_compare): New function.
10267 (expand_compare_loop): New function.
10268 (expand_block_compare): Call expand_compare_loop() when appropriate.
10269 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10270 option description.
10271 (-mblock-compare-inline-loop-limit): New option.
10273 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10276 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10277 Reverse order of second and third operands in first alternative.
10278 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10279 of first and second elements in UNSPEC_VPERMR vector.
10280 (altivec_expand_vec_perm_le): Likewise.
10282 2018-01-08 Jeff Law <law@redhat.com>
10284 PR rtl-optimizatin/81308
10285 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10286 (process_switch): If group_case_labels makes a change, then set
10288 (pass_convert_switch::execute): If a switch is converted, then
10289 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10291 PR rtl-optimization/81308
10292 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10295 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10297 PR target/83663 - Revert r255946
10298 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10299 generation for cases where splatting a value is not useful.
10300 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10301 across a vec_duplicate and a paradoxical subreg forming a vector
10302 mode to a vec_concat.
10304 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10306 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10307 -march=armv8.3-a variants.
10308 * config/arm/t-multilib: Likewise.
10309 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10311 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10313 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10315 (cceq_ior_compare_complement): Give it a name so I can use it, and
10316 change boolean_or_operator predicate to boolean_operator so it can
10317 be used to generate a crand.
10318 (eqne): New code iterator.
10319 (bd/bd_neg): New code_attrs.
10320 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10321 a single define_insn.
10322 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10323 decrement (bdnzt/bdnzf/bdzt/bdzf).
10324 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10325 with the new names of the branch decrement patterns, and added the
10326 names of the branch decrement conditional patterns.
10328 2018-01-08 Richard Biener <rguenther@suse.de>
10330 PR tree-optimization/83563
10331 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10334 2018-01-08 Richard Biener <rguenther@suse.de>
10336 PR middle-end/83713
10337 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10339 2018-01-08 Richard Biener <rguenther@suse.de>
10341 PR tree-optimization/83685
10342 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10343 references to abnormals.
10345 2018-01-08 Richard Biener <rguenther@suse.de>
10348 * dwarf2out.c (output_indirect_strings): Handle empty
10349 skeleton_debug_str_hash.
10350 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10352 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10354 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10355 (emit_store_direct): Likewise.
10356 (arc_trampoline_adjust_address): Likewise.
10357 (arc_asm_trampoline_template): New function.
10358 (arc_initialize_trampoline): Use asm_trampoline_template.
10359 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10360 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10361 * config/arc/arc.md (flush_icache): Delete pattern.
10363 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10365 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10366 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10369 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10372 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10373 by not USED_FOR_TARGET.
10374 (make_pass_resolve_sw_modes): Likewise.
10376 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10378 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10381 2018-01-08 Richard Biener <rguenther@suse.de>
10383 PR middle-end/83580
10384 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10386 2018-01-08 Richard Biener <rguenther@suse.de>
10388 PR middle-end/83517
10389 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10391 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10393 PR middle-end/81897
10394 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10395 basic blocks with a small number of successors.
10396 (convert_control_dep_chain_into_preds): Improve handling of
10398 (dump_predicates): Split apart into...
10399 (dump_pred_chain): ...here...
10400 (dump_pred_info): ...and here.
10401 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10402 (can_chain_union_be_invalidated_p): Improve check for invalidation
10404 (uninit_uses_cannot_happen): Avoid unnecessary if
10405 convert_control_dep_chain_into_preds yielded nothing.
10407 2018-01-06 Martin Sebor <msebor@redhat.com>
10409 PR tree-optimization/83640
10410 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10411 subtracting negative offset from size.
10412 (builtin_access::overlap): Adjust offset bounds of the access to fall
10413 within the size of the object if possible.
10415 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10417 PR rtl-optimization/83699
10418 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10419 extract_bit_field_as_subreg to cases in which the extracted
10420 value is also a vector.
10422 * lra-constraints.c (process_alt_operands): Test for the equivalence
10423 substitutions when detecting a possible reload cycle.
10425 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10428 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10429 by default if flag_selective_schedling{,2}. Formatting fixes.
10431 PR rtl-optimization/83682
10432 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10433 if it has non-VECTOR_MODE element mode.
10434 (vec_duplicate_p): Likewise.
10436 PR middle-end/83694
10437 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10438 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10440 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10443 * config/i386/i386-builtin.def
10444 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10445 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10446 Require also OPTION_MASK_ISA_AVX512F in addition to
10447 OPTION_MASK_ISA_GFNI.
10448 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10449 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10450 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10451 to OPTION_MASK_ISA_GFNI.
10452 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10453 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10454 OPTION_MASK_ISA_AVX512BW.
10455 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10456 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10457 addition to OPTION_MASK_ISA_GFNI.
10458 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10459 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10460 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10461 to OPTION_MASK_ISA_GFNI.
10462 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10463 a requirement for all ISAs rather than any of them with a few
10465 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10467 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10468 bitmasks to be enabled with 3 exceptions, instead of requiring any
10469 enabled ISA with lots of exceptions.
10470 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10471 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10472 Change avx512bw in isa attribute to avx512f.
10473 * config/i386/sgxintrin.h: Add license boilerplate.
10474 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10475 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10476 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10477 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10479 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10480 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10481 temporarily sse2 rather than sse if not enabled already.
10484 * config/i386/sse.md (VI248_VLBW): Rename to ...
10485 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10486 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10487 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10488 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10489 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10490 mode iterator instead of VI248_VLBW.
10492 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10494 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10495 (record_modified): Skip clobbers; add debug output.
10496 (param_change_prob): Use sreal frequencies.
10498 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10500 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10501 punt for user-aligned variables.
10503 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10505 * tree-chrec.c (chrec_contains_symbols): Return true for
10508 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10511 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10512 of (x|y) == x for BICS pattern.
10514 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10516 PR tree-optimization/83605
10517 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10518 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10521 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10523 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10524 * config/epiphany/rtems.h: New file.
10526 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10527 Uros Bizjak <ubizjak@gmail.com>
10530 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10531 QIreg_operand instead of register_operand predicate.
10532 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10533 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10534 comments instead of -fmitigate[-_]rop.
10536 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10539 * cgraphunit.c (symbol_table::compile): Switch to text_section
10540 before calling assembly_start debug hook.
10541 * run-rtl-passes.c (run_rtl_passes): Likewise.
10544 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10546 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10547 range_int_cst_p rather than !symbolic_range_p before calling
10548 extract_range_from_multiplicative_op_1.
10550 2018-01-04 Jeff Law <law@redhat.com>
10552 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10553 redundant test in assertion.
10555 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10557 * doc/rtl.texi: Document machine_mode wrapper classes.
10559 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10561 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10562 using tree_to_uhwi.
10564 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10566 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10567 the VEC_PERM_EXPR fold to fail.
10569 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10572 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10573 to switched_sections.
10575 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10578 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10579 test for d.testing.
10581 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10584 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10585 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10587 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10590 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10591 is BLKmode and bitpos not zero or mode change is needed.
10593 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10596 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10599 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10602 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10603 instead of MULT rtx. Update all corresponding splitters.
10604 (*saddl_se): Ditto.
10605 (*ssub<modesuffix>): Ditto.
10606 (*ssubl_se): Ditto.
10607 (*cmp_sadd_di): Update split patterns.
10608 (*cmp_sadd_si): Ditto.
10609 (*cmp_sadd_sidi): Ditto.
10610 (*cmp_ssub_di): Ditto.
10611 (*cmp_ssub_si): Ditto.
10612 (*cmp_ssub_sidi): Ditto.
10613 * config/alpha/predicates.md (const23_operand): New predicate.
10614 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10615 Look for ASHIFT, not MULT inner operand.
10616 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10618 2018-01-04 Martin Liska <mliska@suse.cz>
10620 PR gcov-profile/83669
10621 * gcov.c (output_intermediate_file): Add version to intermediate
10623 * doc/gcov.texi: Document new field 'version' in intermediate
10624 file format. Fix location of '-k' option of gcov command.
10626 2018-01-04 Martin Liska <mliska@suse.cz>
10629 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10631 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10633 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10635 2018-01-03 Martin Sebor <msebor@redhat.com>
10637 PR tree-optimization/83655
10638 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10639 checking calls with invalid arguments.
10641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10643 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10644 (vectorizable_mask_load_store): Delete.
10645 (vectorizable_call): Return false for masked loads and stores.
10646 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10647 instead of gimple_assign_rhs1.
10648 (vectorizable_load): Handle IFN_MASK_LOAD.
10649 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10653 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10655 (vectorizable_mask_load_store): ...here.
10656 (vectorizable_load): ...and here.
10658 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10660 * tree-vect-stmts.c (vect_build_all_ones_mask)
10661 (vect_build_zero_merge_argument): New functions, split out from...
10662 (vectorizable_load): ...here.
10664 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10666 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10668 (vectorizable_mask_load_store): ...here.
10669 (vectorizable_store): ...and here.
10671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10673 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10675 (vectorizable_mask_load_store): ...here.
10677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10679 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10680 (vect_model_store_cost): Take a vec_load_store_type instead of a
10682 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10683 (vect_model_store_cost): Take a vec_load_store_type instead of a
10685 (vectorizable_mask_load_store): Update accordingly.
10686 (vectorizable_store): Likewise.
10687 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10691 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10692 IFN_MASK_LOAD calls here rather than...
10693 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10696 Alan Hayward <alan.hayward@arm.com>
10697 David Sherwood <david.sherwood@arm.com>
10699 * expmed.c (extract_bit_field_1): For vector extracts,
10700 fall back to extract_bit_field_as_subreg if vec_extract
10703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10704 Alan Hayward <alan.hayward@arm.com>
10705 David Sherwood <david.sherwood@arm.com>
10707 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10708 they are variable or constant sized.
10709 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10710 slots for constant-sized data.
10712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10713 Alan Hayward <alan.hayward@arm.com>
10714 David Sherwood <david.sherwood@arm.com>
10716 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10717 handling COND_EXPRs with boolean comparisons, try to find a better
10718 basis for the mask type than the boolean itself.
10720 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10722 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10723 is calculated and how it can be overridden.
10724 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10725 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10727 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10731 Alan Hayward <alan.hayward@arm.com>
10732 David Sherwood <david.sherwood@arm.com>
10734 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10735 Remove the mode argument.
10736 (aarch64_simd_valid_immediate): Remove the mode and inverse
10738 * config/aarch64/iterators.md (bitsize): New iterator.
10739 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10740 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10741 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10742 aarch64_simd_valid_immediate.
10743 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10744 (aarch64_reg_or_bic_imm): Likewise.
10745 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10746 with an insn_type enum and msl with a modifier_type enum.
10747 Replace element_width with a scalar_mode. Change the shift
10748 to unsigned int. Add constructors for scalar_float_mode and
10749 scalar_int_mode elements.
10750 (aarch64_vect_float_const_representable_p): Delete.
10751 (aarch64_can_const_movi_rtx_p)
10752 (aarch64_simd_scalar_immediate_valid_for_move)
10753 (aarch64_simd_make_constant): Update call to
10754 aarch64_simd_valid_immediate.
10755 (aarch64_advsimd_valid_immediate_hs): New function.
10756 (aarch64_advsimd_valid_immediate): Likewise.
10757 (aarch64_simd_valid_immediate): Remove mode and inverse
10758 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10759 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10760 and aarch64_float_const_representable_p on the result.
10761 (aarch64_output_simd_mov_immediate): Remove mode argument.
10762 Update call to aarch64_simd_valid_immediate and use of
10763 simd_immediate_info.
10764 (aarch64_output_scalar_simd_mov_immediate): Update call
10767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10768 Alan Hayward <alan.hayward@arm.com>
10769 David Sherwood <david.sherwood@arm.com>
10771 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10772 (mode_nunits): Likewise CONST_MODE_NUNITS.
10773 * machmode.def (ADJUST_NUNITS): Document.
10774 * genmodes.c (mode_data::need_nunits_adj): New field.
10775 (blank_mode): Update accordingly.
10776 (adj_nunits): New variable.
10777 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10779 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10780 listed in adj_nunits.
10781 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10782 listed in adj_nunits. Don't emit case statements for such modes.
10783 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10784 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10785 nothing if adj_nunits is nonnull.
10786 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10787 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10788 (emit_mode_fbit): Update use of print_maybe_const_decl.
10789 (emit_move_size): Likewise. Treat the array as non-const
10791 (emit_mode_adjustments): Handle adj_nunits.
10793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10795 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10796 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10797 (VECTOR_MODES): Use it.
10798 (make_vector_modes): Take the prefix as an argument.
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 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10805 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10806 for MODE_VECTOR_BOOL.
10807 * machmode.def (VECTOR_BOOL_MODE): Document.
10808 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10809 (make_vector_bool_mode): New function.
10810 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10812 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10813 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10815 * stor-layout.c (int_mode_for_mode): Likewise.
10816 * tree.c (build_vector_type_for_mode): Likewise.
10817 * varasm.c (output_constant_pool_2): Likewise.
10818 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10819 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10820 for MODE_VECTOR_BOOL.
10821 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10822 of mode class checks.
10823 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10824 instead of a list of mode class checks.
10825 (expand_vector_scalar_condition): Likewise.
10826 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10828 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10829 Alan Hayward <alan.hayward@arm.com>
10830 David Sherwood <david.sherwood@arm.com>
10832 * machmode.h (mode_size): Change from unsigned short to
10834 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10835 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10836 or if measurement_type is not polynomial.
10837 (fixed_size_mode::includes_p): Check for constant-sized modes.
10838 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10839 return a poly_uint16 rather than an unsigned short.
10840 (emit_mode_size): Change the type of mode_size from unsigned short
10841 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10842 (emit_mode_adjustments): Cope with polynomial vector sizes.
10843 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10845 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10847 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10848 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10849 * caller-save.c (setup_save_areas): Likewise.
10850 (replace_reg_with_saved_mem): Likewise.
10851 * calls.c (emit_library_call_value_1): Likewise.
10852 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10853 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10854 (gen_lowpart_for_combine): Likewise.
10855 * convert.c (convert_to_integer_1): Likewise.
10856 * cse.c (equiv_constant, cse_insn): Likewise.
10857 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10858 (cselib_subst_to_values): Likewise.
10859 * dce.c (word_dce_process_block): Likewise.
10860 * df-problems.c (df_word_lr_mark_ref): Likewise.
10861 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10862 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10863 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10864 (rtl_for_decl_location): Likewise.
10865 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10866 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10867 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10868 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10869 (expand_expr_real_1): Likewise.
10870 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10871 (pad_below): Likewise.
10872 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10873 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10874 * ira.c (get_subreg_tracking_sizes): Likewise.
10875 * ira-build.c (ira_create_allocno_objects): Likewise.
10876 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10877 (ira_sort_regnos_for_alter_reg): Likewise.
10878 * ira-costs.c (record_operand_costs): Likewise.
10879 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10880 (resolve_simple_move): Likewise.
10881 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10882 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10883 (lra_constraints): Likewise.
10884 (CONST_POOL_OK_P): Reject variable-sized modes.
10885 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10886 (add_pseudo_to_slot, lra_spill): Likewise.
10887 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10888 * optabs-query.c (get_best_extraction_insn): Likewise.
10889 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10890 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10891 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10892 * recog.c (offsettable_address_addr_space_p): Likewise.
10893 * regcprop.c (maybe_mode_change): Likewise.
10894 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10895 * regrename.c (build_def_use): Likewise.
10896 * regstat.c (dump_reg_info): Likewise.
10897 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10898 (find_reloads, find_reloads_subreg_address): Likewise.
10899 * reload1.c (eliminate_regs_1): Likewise.
10900 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10901 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10902 (simplify_binary_operation_1, simplify_subreg): Likewise.
10903 * targhooks.c (default_function_arg_padding): Likewise.
10904 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10905 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10906 (verify_gimple_assign_ternary): Likewise.
10907 * tree-inline.c (estimate_move_cost): Likewise.
10908 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10909 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10910 (get_address_cost_ainc): Likewise.
10911 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10912 (vect_supportable_dr_alignment): Likewise.
10913 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10914 (vectorizable_reduction): Likewise.
10915 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10916 (vectorizable_operation, vectorizable_load): Likewise.
10917 * tree.c (build_same_sized_truth_vector_type): Likewise.
10918 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10919 * var-tracking.c (emit_note_insn_var_location): Likewise.
10920 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10921 (ADDR_VEC_ALIGN): Likewise.
10923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10924 Alan Hayward <alan.hayward@arm.com>
10925 David Sherwood <david.sherwood@arm.com>
10927 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10929 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10930 or if measurement_type is polynomial.
10931 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10932 * combine.c (make_extraction): Likewise.
10933 * dse.c (find_shift_sequence): Likewise.
10934 * dwarf2out.c (mem_loc_descriptor): Likewise.
10935 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10936 (extract_bit_field, extract_low_bits): Likewise.
10937 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10938 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10939 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10940 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10941 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10942 * reload.c (find_reloads): Likewise.
10943 * reload1.c (alter_reg): Likewise.
10944 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10945 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10946 * tree-if-conv.c (predicate_mem_writes): Likewise.
10947 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10948 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10949 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10950 * valtrack.c (dead_debug_insert_temp): Likewise.
10951 * varasm.c (mergeable_constant_section): Likewise.
10952 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10954 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10955 Alan Hayward <alan.hayward@arm.com>
10956 David Sherwood <david.sherwood@arm.com>
10958 * expr.c (expand_assignment): Cope with polynomial mode sizes
10959 when assigning to a CONCAT.
10961 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10962 Alan Hayward <alan.hayward@arm.com>
10963 David Sherwood <david.sherwood@arm.com>
10965 * machmode.h (mode_precision): Change from unsigned short to
10967 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10969 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10970 or if measurement_type is not polynomial.
10971 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10972 in which the mode is already known to be a scalar_int_mode.
10973 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10974 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10976 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10977 for GET_MODE_PRECISION.
10978 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10979 for GET_MODE_PRECISION.
10980 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10982 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10983 (expand_field_assignment, make_extraction): Likewise.
10984 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10985 (get_last_value): Likewise.
10986 * convert.c (convert_to_integer_1): Likewise.
10987 * cse.c (cse_insn): Likewise.
10988 * expr.c (expand_expr_real_1): Likewise.
10989 * lra-constraints.c (simplify_operand_subreg): Likewise.
10990 * optabs-query.c (can_atomic_load_p): Likewise.
10991 * optabs.c (expand_atomic_load): Likewise.
10992 (expand_atomic_store): Likewise.
10993 * ree.c (combine_reaching_defs): Likewise.
10994 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10995 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10996 * tree.h (type_has_mode_precision_p): Likewise.
10997 * ubsan.c (instrument_si_overflow): Likewise.
10999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11000 Alan Hayward <alan.hayward@arm.com>
11001 David Sherwood <david.sherwood@arm.com>
11003 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11004 polynomial numbers of units.
11005 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11006 (valid_vector_subparts_p): New function.
11007 (build_vector_type): Remove temporary shim and take the number
11008 of units as a poly_uint64 rather than an int.
11009 (build_opaque_vector_type): Take the number of units as a
11010 poly_uint64 rather than an int.
11011 * tree.c (build_vector_from_ctor): Handle polynomial
11012 TYPE_VECTOR_SUBPARTS.
11013 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11014 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11015 (build_vector_from_val): If the number of units is variable,
11016 use build_vec_duplicate_cst for constant operands and
11017 VEC_DUPLICATE_EXPR otherwise.
11018 (make_vector_type): Remove temporary is_constant ().
11019 (build_vector_type, build_opaque_vector_type): Take the number of
11020 units as a poly_uint64 rather than an int.
11021 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11023 * cfgexpand.c (expand_debug_expr): Likewise.
11024 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11025 (store_constructor, expand_expr_real_1): Likewise.
11026 (const_scalar_mask_from_tree): Likewise.
11027 * fold-const-call.c (fold_const_reduction): Likewise.
11028 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11029 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11030 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11031 (fold_relational_const): Likewise.
11032 (native_interpret_vector): Likewise. Change the size from an
11033 int to an unsigned int.
11034 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11035 TYPE_VECTOR_SUBPARTS.
11036 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11037 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11038 duplicating a non-constant operand into a variable-length vector.
11039 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11040 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11041 * ipa-icf.c (sem_variable::equals): Likewise.
11042 * match.pd: Likewise.
11043 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11044 * print-tree.c (print_node): Likewise.
11045 * stor-layout.c (layout_type): Likewise.
11046 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11047 * tree-cfg.c (verify_gimple_comparison): Likewise.
11048 (verify_gimple_assign_binary): Likewise.
11049 (verify_gimple_assign_ternary): Likewise.
11050 (verify_gimple_assign_single): Likewise.
11051 * tree-pretty-print.c (dump_generic_node): Likewise.
11052 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11053 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11054 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11055 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11056 (vect_shift_permute_load_chain): Likewise.
11057 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11058 (expand_vector_condition, optimize_vector_constructor): Likewise.
11059 (lower_vec_perm, get_compute_type): Likewise.
11060 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11061 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11062 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11063 (vect_recog_mask_conversion_pattern): Likewise.
11064 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11065 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11066 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11067 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11068 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11069 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11070 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11071 (supportable_widening_operation): Likewise.
11072 (supportable_narrowing_operation): Likewise.
11073 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11075 * varasm.c (output_constant): Likewise.
11077 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11078 Alan Hayward <alan.hayward@arm.com>
11079 David Sherwood <david.sherwood@arm.com>
11081 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11082 so that both the length == 3 and length != 3 cases set up their
11083 own permute vectors. Add comments explaining why we know the
11084 number of elements is constant.
11085 (vect_permute_load_chain): Likewise.
11087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11088 Alan Hayward <alan.hayward@arm.com>
11089 David Sherwood <david.sherwood@arm.com>
11091 * machmode.h (mode_nunits): Change from unsigned char to
11093 (ONLY_FIXED_SIZE_MODES): New macro.
11094 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11095 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11096 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11098 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11099 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11100 or if measurement_type is not polynomial.
11101 * genmodes.c (ZERO_COEFFS): New macro.
11102 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11104 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11105 Use ZERO_COEFFS when emitting initializers.
11106 * data-streamer.h (bp_pack_poly_value): New function.
11107 (bp_unpack_poly_value): Likewise.
11108 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11109 for GET_MODE_NUNITS.
11110 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11111 for GET_MODE_NUNITS.
11112 * tree.c (make_vector_type): Remove temporary shim and make
11113 the real function take the number of units as a poly_uint64
11114 rather than an int.
11115 (build_vector_type_for_mode): Handle polynomial nunits.
11116 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11117 * emit-rtl.c (const_vec_series_p_1): Likewise.
11118 (gen_rtx_CONST_VECTOR): Likewise.
11119 * fold-const.c (test_vec_duplicate_folding): Likewise.
11120 * genrecog.c (validate_pattern): Likewise.
11121 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11122 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11123 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11124 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11125 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11126 * rtlanal.c (subreg_get_info): Likewise.
11127 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11128 (vect_grouped_load_supported): Likewise.
11129 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11130 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11131 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11132 (simplify_const_unary_operation, simplify_binary_operation_1)
11133 (simplify_const_binary_operation, simplify_ternary_operation)
11134 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11135 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11136 instead of CONST_VECTOR_NUNITS.
11137 * varasm.c (output_constant_pool_2): Likewise.
11138 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11139 explicit-encoded elements in the XVEC for variable-length vectors.
11141 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11143 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11145 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11146 Alan Hayward <alan.hayward@arm.com>
11147 David Sherwood <david.sherwood@arm.com>
11149 * coretypes.h (fixed_size_mode): Declare.
11150 (fixed_size_mode_pod): New typedef.
11151 * builtins.h (target_builtins::x_apply_args_mode)
11152 (target_builtins::x_apply_result_mode): Change type to
11153 fixed_size_mode_pod.
11154 * builtins.c (apply_args_size, apply_result_size, result_vector)
11155 (expand_builtin_apply_args_1, expand_builtin_apply)
11156 (expand_builtin_return): Update accordingly.
11158 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11160 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11161 * cselib.c (cselib_hash_rtx): Likewise.
11162 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11163 CONST_VECTOR encoding.
11165 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11166 Jeff Law <law@redhat.com>
11169 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11170 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11171 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11172 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11175 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11176 explicitly probe *sp in a noreturn function if there were any callee
11177 register saves or frame pointer is needed.
11179 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11182 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11183 BLKmode for ternary, binary or unary expressions.
11186 * var-tracking.c (delete_vta_debug_insn): New inline function.
11187 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11188 insns from get_insns () to NULL instead of each bb separately.
11189 Use delete_vta_debug_insn. No longer static.
11190 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11191 delete_vta_debug_insns callers.
11192 * rtl.h (delete_vta_debug_insns): Declare.
11193 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11194 instead of variable_tracking_main.
11196 2018-01-03 Martin Sebor <msebor@redhat.com>
11198 PR tree-optimization/83603
11199 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11200 arguments past the endof the argument list in functions declared
11201 without a prototype.
11202 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11203 Avoid checking when arguments are null.
11205 2018-01-03 Martin Sebor <msebor@redhat.com>
11208 * doc/extend.texi (attribute const): Fix a typo.
11209 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11210 issuing -Wsuggest-attribute for void functions.
11212 2018-01-03 Martin Sebor <msebor@redhat.com>
11214 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11215 offset_int::from instead of wide_int::to_shwi.
11216 (maybe_diag_overlap): Remove assertion.
11217 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11218 * gimple-ssa-sprintf.c (format_directive): Same.
11219 (parse_directive): Same.
11220 (sprintf_dom_walker::compute_format_length): Same.
11221 (try_substitute_return_value): Same.
11223 2018-01-03 Jeff Law <law@redhat.com>
11225 PR middle-end/83654
11226 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11227 non-constant residual for zero at runtime and avoid probing in
11228 that case. Reorganize code for trailing problem to mirror handling
11231 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11233 PR tree-optimization/83501
11234 * tree-ssa-strlen.c (get_string_cst): New.
11235 (handle_char_store): Call get_string_cst.
11237 2018-01-03 Martin Liska <mliska@suse.cz>
11239 PR tree-optimization/83593
11240 * tree-ssa-strlen.c: Include tree-cfg.h.
11241 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11242 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11243 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11245 (strlen_dom_walker::before_dom_children): Call
11246 gimple_purge_dead_eh_edges. Dump tranformation with details
11248 (strlen_dom_walker::before_dom_children): Update call by adding
11249 new argument cleanup_eh.
11250 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11252 2018-01-03 Martin Liska <mliska@suse.cz>
11255 * cif-code.def (VARIADIC_THUNK): New enum value.
11256 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11259 2018-01-03 Jan Beulich <jbeulich@suse.com>
11261 * sse.md (mov<mode>_internal): Tighten condition for when to use
11262 vmovdqu<ssescalarsize> for TI and OI modes.
11264 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11266 Update copyright years.
11268 2018-01-03 Martin Liska <mliska@suse.cz>
11271 * ipa-visibility.c (function_and_variable_visibility): Skip
11272 functions with noipa attribure.
11274 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11276 * gcc.c (process_command): Update copyright notice dates.
11277 * gcov-dump.c (print_version): Ditto.
11278 * gcov.c (print_version): Ditto.
11279 * gcov-tool.c (print_version): Ditto.
11280 * gengtype.c (create_file): Ditto.
11281 * doc/cpp.texi: Bump @copying's copyright year.
11282 * doc/cppinternals.texi: Ditto.
11283 * doc/gcc.texi: Ditto.
11284 * doc/gccint.texi: Ditto.
11285 * doc/gcov.texi: Ditto.
11286 * doc/install.texi: Ditto.
11287 * doc/invoke.texi: Ditto.
11289 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11291 * vector-builder.h (vector_builder::m_full_nelts): Change from
11292 unsigned int to poly_uint64.
11293 (vector_builder::full_nelts): Update prototype accordingly.
11294 (vector_builder::new_vector): Likewise.
11295 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11296 (vector_builder::operator ==): Likewise.
11297 (vector_builder::finalize): Likewise.
11298 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11299 Take the number of elements as a poly_uint64 rather than an
11301 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11302 from unsigned int to poly_uint64.
11303 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11304 (vec_perm_indices::new_vector): Likewise.
11305 (vec_perm_indices::length): Likewise.
11306 (vec_perm_indices::nelts_per_input): Likewise.
11307 (vec_perm_indices::input_nelts): Likewise.
11308 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11309 number of elements per input as a poly_uint64 rather than an
11310 unsigned int. Use the original encoding for variable-length
11311 vectors, rather than clamping each individual element.
11312 For the second and subsequent elements in each pattern,
11313 clamp the step and base before clamping their sum.
11314 (vec_perm_indices::series_p): Handle polynomial element counts.
11315 (vec_perm_indices::all_in_range_p): Likewise.
11316 (vec_perm_indices_to_tree): Likewise.
11317 (vec_perm_indices_to_rtx): Likewise.
11318 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11319 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11320 (tree_vector_builder::new_binary_operation): Handle polynomial
11321 element counts. Return false if we need to know the number
11322 of elements at compile time.
11323 * fold-const.c (fold_vec_perm): Punt if the number of elements
11324 isn't known at compile time.
11326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11328 * vec-perm-indices.h (vec_perm_builder): Change element type
11329 from HOST_WIDE_INT to poly_int64.
11330 (vec_perm_indices::element_type): Update accordingly.
11331 (vec_perm_indices::clamp): Handle polynomial element_types.
11332 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11333 (vec_perm_indices::all_in_range_p): Likewise.
11334 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11336 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11337 polynomial vec_perm_indices element types.
11338 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11339 * fold-const.c (fold_vec_perm): Likewise.
11340 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11341 * tree-vect-generic.c (lower_vec_perm): Likewise.
11342 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11343 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11344 element type to HOST_WIDE_INT.
11346 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11347 Alan Hayward <alan.hayward@arm.com>
11348 David Sherwood <david.sherwood@arm.com>
11350 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11351 rather than an int. Use plus_constant.
11352 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11353 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11356 Alan Hayward <alan.hayward@arm.com>
11357 David Sherwood <david.sherwood@arm.com>
11359 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11360 a HOST_WIDE_INT to a poly_int64.
11362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11363 Alan Hayward <alan.hayward@arm.com>
11364 David Sherwood <david.sherwood@arm.com>
11366 * calls.c (load_register_parameters): Cope with polynomial
11367 mode sizes. Require a constant size for BLKmode parameters
11368 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11369 forces a parameter to be padded at the lsb end in order to
11370 fill a complete number of words, require the parameter size
11371 to be ordered wrt UNITS_PER_WORD.
11373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11374 Alan Hayward <alan.hayward@arm.com>
11375 David Sherwood <david.sherwood@arm.com>
11377 * reload1.c (spill_stack_slot_width): Change element type
11378 from unsigned int to poly_uint64_pod.
11379 (alter_reg): Treat mode sizes as polynomial.
11381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11382 Alan Hayward <alan.hayward@arm.com>
11383 David Sherwood <david.sherwood@arm.com>
11385 * reload.c (complex_word_subreg_p): New function.
11386 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
11393 operands whose sizes aren't ordered.
11394 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11401 that the mode size is in the set {1, 2, 4, 8, 16}.
11403 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11404 Alan Hayward <alan.hayward@arm.com>
11405 David Sherwood <david.sherwood@arm.com>
11407 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11408 Use plus_constant instead of gen_rtx_PLUS.
11410 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11411 Alan Hayward <alan.hayward@arm.com>
11412 David Sherwood <david.sherwood@arm.com>
11414 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11415 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11416 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11417 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11418 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11419 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11420 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11421 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11422 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11423 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11425 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11426 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11427 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11428 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11429 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11430 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11431 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11432 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11433 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11434 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11436 * expr.c (emit_move_resolve_push): Treat the input and result
11437 of PUSH_ROUNDING as a poly_int64.
11438 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11439 (emit_push_insn): Likewise.
11440 * lra-eliminations.c (mark_not_eliminable): Likewise.
11441 * recog.c (push_operand): Likewise.
11442 * reload1.c (elimination_effects): Likewise.
11443 * rtlanal.c (nonzero_bits1): Likewise.
11444 * calls.c (store_one_arg): Likewise. Require the padding to be
11445 known at compile time.
11447 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11448 Alan Hayward <alan.hayward@arm.com>
11449 David Sherwood <david.sherwood@arm.com>
11451 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11452 Use plus_constant instead of gen_rtx_PLUS.
11454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11455 Alan Hayward <alan.hayward@arm.com>
11456 David Sherwood <david.sherwood@arm.com>
11458 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11459 rather than an int.
11461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11462 Alan Hayward <alan.hayward@arm.com>
11463 David Sherwood <david.sherwood@arm.com>
11465 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11466 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11467 via stack temporaries. Treat the mode size as polynomial too.
11469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11470 Alan Hayward <alan.hayward@arm.com>
11471 David Sherwood <david.sherwood@arm.com>
11473 * expr.c (expand_expr_real_2): When handling conversions involving
11474 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11475 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11476 as a poly_uint64 too.
11478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11479 Alan Hayward <alan.hayward@arm.com>
11480 David Sherwood <david.sherwood@arm.com>
11482 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
11484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11485 Alan Hayward <alan.hayward@arm.com>
11486 David Sherwood <david.sherwood@arm.com>
11488 * combine.c (can_change_dest_mode): Handle polynomial
11489 REGMODE_NATURAL_SIZE.
11490 * expmed.c (store_bit_field_1): Likewise.
11491 * expr.c (store_constructor): Likewise.
11492 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11493 and polynomial REGMODE_NATURAL_SIZE.
11494 (gen_lowpart_common): Likewise.
11495 * reginfo.c (record_subregs_of_mode): Likewise.
11496 * rtlanal.c (read_modify_subreg_p): Likewise.
11498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11499 Alan Hayward <alan.hayward@arm.com>
11500 David Sherwood <david.sherwood@arm.com>
11502 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11503 numbers of elements.
11505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11506 Alan Hayward <alan.hayward@arm.com>
11507 David Sherwood <david.sherwood@arm.com>
11509 * match.pd: Cope with polynomial numbers of vector elements.
11511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11512 Alan Hayward <alan.hayward@arm.com>
11513 David Sherwood <david.sherwood@arm.com>
11515 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11516 in a POINTER_PLUS_EXPR.
11518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11519 Alan Hayward <alan.hayward@arm.com>
11520 David Sherwood <david.sherwood@arm.com>
11522 * omp-simd-clone.c (simd_clone_subparts): New function.
11523 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11524 (ipa_simd_modify_function_body): Likewise.
11526 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11527 Alan Hayward <alan.hayward@arm.com>
11528 David Sherwood <david.sherwood@arm.com>
11530 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11531 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11532 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11533 (expand_vector_condition, vector_element): Likewise.
11534 (subparts_gt): New function.
11535 (get_compute_type): Use subparts_gt.
11536 (count_type_subparts): Delete.
11537 (expand_vector_operations_1): Use subparts_gt instead of
11538 count_type_subparts.
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-vect-data-refs.c (vect_no_alias_p): Replace with...
11545 (vect_compile_time_alias): ...this new function. Do the calculation
11546 on poly_ints rather than trees.
11547 (vect_prune_runtime_alias_test_list): Update call accordingly.
11549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11550 Alan Hayward <alan.hayward@arm.com>
11551 David Sherwood <david.sherwood@arm.com>
11553 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11555 (vect_schedule_slp_instance): Likewise.
11557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11558 Alan Hayward <alan.hayward@arm.com>
11559 David Sherwood <david.sherwood@arm.com>
11561 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11562 constant and extern definitions for variable-length vectors.
11563 (vect_get_constant_vectors): Note that the number of units
11564 is known to be constant.
11566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11567 Alan Hayward <alan.hayward@arm.com>
11568 David Sherwood <david.sherwood@arm.com>
11570 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11571 of units as polynomial. Choose between WIDE and NARROW based
11574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11575 Alan Hayward <alan.hayward@arm.com>
11576 David Sherwood <david.sherwood@arm.com>
11578 * tree-vect-stmts.c (simd_clone_subparts): New function.
11579 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11582 Alan Hayward <alan.hayward@arm.com>
11583 David Sherwood <david.sherwood@arm.com>
11585 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11586 vectors as polynomial. Use build_index_vector for
11587 IFN_GOMP_SIMD_LANE.
11589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11590 Alan Hayward <alan.hayward@arm.com>
11591 David Sherwood <david.sherwood@arm.com>
11593 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11594 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11595 for variable-length vectors.
11596 (vectorizable_mask_load_store): Treat the number of units as
11597 polynomial, asserting that it is constant if the condition has
11598 already been enforced.
11599 (vectorizable_store, vectorizable_load): Likewise.
11601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11602 Alan Hayward <alan.hayward@arm.com>
11603 David Sherwood <david.sherwood@arm.com>
11605 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11606 of units as polynomial. Punt if we can't tell at compile time
11607 which vector contains the final result.
11609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11610 Alan Hayward <alan.hayward@arm.com>
11611 David Sherwood <david.sherwood@arm.com>
11613 * tree-vect-loop.c (vectorizable_induction): Treat the number
11614 of units as polynomial. Punt on SLP inductions. Use an integer
11615 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11616 cast of such a series for variable-length floating-point
11619 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11620 Alan Hayward <alan.hayward@arm.com>
11621 David Sherwood <david.sherwood@arm.com>
11623 * tree.h (build_index_vector): Declare.
11624 * tree.c (build_index_vector): New function.
11625 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11626 of units as polynomial, forcibly converting it to a constant if
11627 vectorizable_reduction has already enforced the condition.
11628 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11629 to create a {1,2,3,...} vector.
11630 (vectorizable_reduction): Treat the number of units as polynomial.
11631 Choose vectype_in based on the largest scalar element size rather
11632 than the smallest number of units. Enforce the restrictions
11635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11636 Alan Hayward <alan.hayward@arm.com>
11637 David Sherwood <david.sherwood@arm.com>
11639 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11640 number of units as polynomial.
11642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11643 Alan Hayward <alan.hayward@arm.com>
11644 David Sherwood <david.sherwood@arm.com>
11646 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11647 * target.def (autovectorize_vector_sizes): Return the vector sizes
11648 by pointer, using vector_sizes rather than a bitmask.
11649 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11650 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11651 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11653 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11654 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11655 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11656 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11657 * omp-general.c (omp_max_vf): Likewise.
11658 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11659 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11660 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11661 * tree-vect-slp.c (vect_slp_bb): Likewise.
11662 * doc/tm.texi: Regenerate.
11663 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11665 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11666 the vector size as a poly_uint64 rather than an unsigned int.
11667 (current_vector_size): Change from an unsigned int to a poly_uint64.
11668 (get_vectype_for_scalar_type): Update accordingly.
11669 * tree.h (build_truth_vector_type): Take the size and number of
11670 units as a poly_uint64 rather than an unsigned int.
11671 (build_vector_type): Add a temporary overload that takes
11672 the number of units as a poly_uint64 rather than an unsigned int.
11673 * tree.c (make_vector_type): Likewise.
11674 (build_truth_vector_type): Take the number of units as a poly_uint64
11675 rather than an unsigned int.
11677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11678 Alan Hayward <alan.hayward@arm.com>
11679 David Sherwood <david.sherwood@arm.com>
11681 * target.def (get_mask_mode): Take the number of units and length
11682 as poly_uint64s rather than unsigned ints.
11683 * targhooks.h (default_get_mask_mode): Update accordingly.
11684 * targhooks.c (default_get_mask_mode): Likewise.
11685 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11686 * doc/tm.texi: Regenerate.
11688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11689 Alan Hayward <alan.hayward@arm.com>
11690 David Sherwood <david.sherwood@arm.com>
11692 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11693 * omp-general.c (omp_max_vf): Likewise.
11694 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11695 (expand_omp_simd): Handle polynomial safelen.
11696 * omp-low.c (omplow_simd_context): Add a default constructor.
11697 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11698 (lower_rec_simd_input_clauses): Update accordingly.
11699 (lower_rec_input_clauses): Likewise.
11701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11702 Alan Hayward <alan.hayward@arm.com>
11703 David Sherwood <david.sherwood@arm.com>
11705 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11706 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11707 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11708 (vect_analyze_slp_cost): Likewise.
11709 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11710 (vect_model_load_cost): Likewise.
11712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11713 Alan Hayward <alan.hayward@arm.com>
11714 David Sherwood <david.sherwood@arm.com>
11716 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11717 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11718 from an unsigned int * to a poly_uint64_pod *.
11719 (calculate_unrolling_factor): New function.
11720 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11722 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11723 Alan Hayward <alan.hayward@arm.com>
11724 David Sherwood <david.sherwood@arm.com>
11726 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11727 from an unsigned int to a poly_uint64.
11728 (_loop_vec_info::slp_unrolling_factor): Likewise.
11729 (_loop_vec_info::vectorization_factor): Change from an int
11731 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11732 (vect_get_num_vectors): New function.
11733 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11734 (vect_get_num_copies): Use vect_get_num_vectors.
11735 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11736 to an unsigned int *.
11737 (vect_analyze_data_refs): Change min_vf from an int * to a
11739 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11740 than an unsigned HOST_WIDE_INT.
11741 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11742 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11743 to an unsigned int *.
11744 (vect_analyze_data_ref_dependences): Likewise.
11745 (vect_compute_data_ref_alignment): Handle polynomial vf.
11746 (vect_enhance_data_refs_alignment): Likewise.
11747 (vect_prune_runtime_alias_test_list): Likewise.
11748 (vect_shift_permute_load_chain): Likewise.
11749 (vect_supportable_dr_alignment): Likewise.
11750 (dependence_distance_ge_vf): Take the vectorization factor as a
11751 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11752 (vect_analyze_data_refs): Change min_vf from an int * to a
11754 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11755 vfm1 as a poly_uint64 rather than an int. Make the same change
11756 for the returned bound_scalar.
11757 (vect_gen_vector_loop_niters): Handle polynomial vf.
11758 (vect_do_peeling): Likewise. Update call to
11759 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11760 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11762 * tree-vect-loop.c (vect_determine_vectorization_factor)
11763 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11764 (vect_get_known_peeling_cost): Likewise.
11765 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11766 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11767 (vect_transform_loop): Likewise. Use the lowest possible VF when
11768 updating the upper bounds of the loop.
11769 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11770 rather than an int.
11771 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11772 polynomial unroll factors.
11773 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11774 (vect_make_slp_decision): Likewise.
11775 (vect_supported_load_permutation_p): Likewise, and polynomial
11777 (vect_analyze_slp_cost): Handle polynomial vf.
11778 (vect_slp_analyze_node_operations): Likewise.
11779 (vect_slp_analyze_bb_1): Likewise.
11780 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11781 than an unsigned HOST_WIDE_INT.
11782 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11783 (vectorizable_load): Handle polynomial vf.
11784 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11786 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11789 Alan Hayward <alan.hayward@arm.com>
11790 David Sherwood <david.sherwood@arm.com>
11792 * match.pd: Handle bit operations involving three constants
11793 and try to fold one pair.
11795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11797 * tree-vect-loop-manip.c: Include gimple-fold.h.
11798 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11799 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11800 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11801 Add a path that uses a step of VF instead of 1, but disable it
11803 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11804 and niters_no_overflow parameters. Update calls to
11805 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11806 Create a new SSA name if the latter choses to use a ste other
11807 than zero, and return it via niters_vector_mult_vf_var.
11808 * tree-vect-loop.c (vect_transform_loop): Update calls to
11809 vect_do_peeling, vect_gen_vector_loop_niters and
11810 slpeel_make_loop_iterate_ntimes.
11811 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11812 (vect_gen_vector_loop_niters): Update declarations after above changes.
11814 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11816 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11817 128-bit round to integer instructions.
11818 (ceil<mode>2): Likewise.
11819 (btrunc<mode>2): Likewise.
11820 (round<mode>2): Likewise.
11822 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11824 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11825 unaligned VSX load/store on P8/P9.
11826 (expand_block_clear): Allow the use of unaligned VSX
11827 load/store on P8/P9.
11829 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11831 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11833 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11834 swap associated with both a load and a store.
11836 2018-01-02 Andrew Waterman <andrew@sifive.com>
11838 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11839 * config/riscv/riscv.md (clear_cache): Use it.
11841 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11843 * web.c: Remove out-of-date comment.
11845 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11847 * expr.c (fixup_args_size_notes): Check that any existing
11848 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11849 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11850 (emit_single_push_insn): ...here.
11852 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11854 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11855 (const_vector_encoded_nelts): New function.
11856 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11857 (const_vector_int_elt, const_vector_elt): Declare.
11858 * emit-rtl.c (const_vector_int_elt_1): New function.
11859 (const_vector_elt): Likewise.
11860 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11861 of CONST_VECTOR_ELT.
11863 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11865 * expr.c: Include rtx-vector-builder.h.
11866 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11867 directly on the tree encoding.
11868 (const_vector_from_tree): Likewise.
11869 * optabs.c: Include rtx-vector-builder.h.
11870 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11871 sequence of "u" values.
11872 * vec-perm-indices.c: Include rtx-vector-builder.h.
11873 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11874 directly on the vec_perm_indices encoding.
11876 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11878 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11879 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11880 * rtx-vector-builder.h: New file.
11881 * rtx-vector-builder.c: Likewise.
11882 * rtl.h (rtx_def::u2): Add a const_vector field.
11883 (CONST_VECTOR_NPATTERNS): New macro.
11884 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11885 (CONST_VECTOR_DUPLICATE_P): Likewise.
11886 (CONST_VECTOR_STEPPED_P): Likewise.
11887 (CONST_VECTOR_ENCODED_ELT): Likewise.
11888 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11889 (unwrap_const_vec_duplicate): Likewise.
11890 (const_vec_series_p): Check for a non-duplicated vector encoding.
11891 Say that the function only returns true for integer vectors.
11892 * emit-rtl.c: Include rtx-vector-builder.h.
11893 (gen_const_vec_duplicate_1): Delete.
11894 (gen_const_vector): Call gen_const_vec_duplicate instead of
11895 gen_const_vec_duplicate_1.
11896 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11897 (gen_const_vec_duplicate): Use rtx_vector_builder.
11898 (gen_const_vec_series): Likewise.
11899 (gen_rtx_CONST_VECTOR): Likewise.
11900 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11901 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11902 Build a new vector rather than modifying a CONST_VECTOR in-place.
11903 (handle_special_swappables): Update call accordingly.
11904 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11905 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11906 Build a new vector rather than modifying a CONST_VECTOR in-place.
11907 (handle_special_swappables): Update call accordingly.
11909 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11911 * simplify-rtx.c (simplify_const_binary_operation): Use
11912 CONST_VECTOR_ELT instead of XVECEXP.
11914 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11916 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11917 the selector elements to be different from the data elements
11918 if the selector is a VECTOR_CST.
11919 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11920 ssizetype for the selector.
11922 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11924 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11925 before testing each element individually.
11926 * tree-vect-generic.c (lower_vec_perm): Likewise.
11928 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11930 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11931 * selftest-run-tests.c (selftest::run_tests): Call it.
11932 * vector-builder.h (vector_builder::operator ==): New function.
11933 (vector_builder::operator !=): Likewise.
11934 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11935 (vec_perm_indices::all_from_input_p): New function.
11936 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11937 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11938 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11939 instead of reading the VECTOR_CST directly. Detect whether both
11940 vector inputs are the same before constructing the vec_perm_indices,
11941 and update the number of inputs argument accordingly. Use the
11942 utility functions added above. Only construct sel2 if we need to.
11944 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11946 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11947 the broadcast of the low byte.
11948 (expand_mult_highpart): Use an explicit encoding for the permutes.
11949 * optabs-query.c (can_mult_highpart_p): Likewise.
11950 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11951 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11952 (vectorizable_bswap): Likewise.
11953 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11954 explicit encoding for the power-of-2 permutes.
11955 (vect_permute_store_chain): Likewise.
11956 (vect_grouped_load_supported): Likewise.
11957 (vect_permute_load_chain): Likewise.
11959 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11961 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11962 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11963 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11964 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11965 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11966 (vect_gen_perm_mask_any): Likewise.
11968 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11970 * int-vector-builder.h: New file.
11971 * vec-perm-indices.h: Include int-vector-builder.h.
11972 (vec_perm_indices): Redefine as an int_vector_builder.
11973 (auto_vec_perm_indices): Delete.
11974 (vec_perm_builder): Redefine as a stand-alone class.
11975 (vec_perm_indices::vec_perm_indices): New function.
11976 (vec_perm_indices::clamp): Likewise.
11977 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11978 (vec_perm_indices::new_vector): New function.
11979 (vec_perm_indices::new_expanded_vector): Update for new
11980 vec_perm_indices class.
11981 (vec_perm_indices::rotate_inputs): New function.
11982 (vec_perm_indices::all_in_range_p): Operate directly on the
11983 encoded form, without computing elided elements.
11984 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11985 encoding. Update for new vec_perm_indices class.
11986 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11987 the given vec_perm_builder.
11988 (expand_vec_perm_var): Update vec_perm_builder constructor.
11989 (expand_mult_highpart): Use vec_perm_builder instead of
11990 auto_vec_perm_indices.
11991 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11992 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11993 or double series encoding as appropriate.
11994 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11995 vec_perm_indices instead of auto_vec_perm_indices.
11996 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11997 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11998 (vect_permute_store_chain): Likewise.
11999 (vect_grouped_load_supported): Likewise.
12000 (vect_permute_load_chain): Likewise.
12001 (vect_shift_permute_load_chain): Likewise.
12002 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12003 (vect_transform_slp_perm_load): Likewise.
12004 (vect_schedule_slp_instance): Likewise.
12005 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12006 (vectorizable_mask_load_store): Likewise.
12007 (vectorizable_bswap): Likewise.
12008 (vectorizable_store): Likewise.
12009 (vectorizable_load): Likewise.
12010 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12011 vec_perm_indices instead of auto_vec_perm_indices. Use
12012 tree_to_vec_perm_builder to read the vector from a tree.
12013 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12014 vec_perm_builder instead of a vec_perm_indices.
12015 (have_whole_vector_shift): Use vec_perm_builder and
12016 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12017 truncation to calc_vec_perm_mask_for_shift.
12018 (vect_create_epilog_for_reduction): Likewise.
12019 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12020 from auto_vec_perm_indices to vec_perm_indices.
12021 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12022 instead of changing individual elements.
12023 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12024 the vector in d.perm.
12025 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12026 from auto_vec_perm_indices to vec_perm_indices.
12027 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12028 instead of changing individual elements.
12029 (arm_vectorize_vec_perm_const): Use new_vector to install
12030 the vector in d.perm.
12031 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12032 Update vec_perm_builder constructor.
12033 (rs6000_expand_interleave): Likewise.
12034 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12035 (rs6000_expand_interleave): Likewise.
12037 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12039 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12040 to qimode could truncate the indices.
12041 * optabs.c (expand_vec_perm_var): Likewise.
12043 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12045 * Makefile.in (OBJS): Add vec-perm-indices.o.
12046 * vec-perm-indices.h: New file.
12047 * vec-perm-indices.c: Likewise.
12048 * target.h (vec_perm_indices): Replace with a forward class
12050 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12051 * optabs.h: Include vec-perm-indices.h.
12052 (expand_vec_perm): Delete.
12053 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12054 (expand_vec_perm_const): Declare.
12055 * target.def (vec_perm_const_ok): Replace with...
12056 (vec_perm_const): ...this new hook.
12057 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12058 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12059 * doc/tm.texi: Regenerate.
12060 * optabs.def (vec_perm_const): Delete.
12061 * doc/md.texi (vec_perm_const): Likewise.
12062 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12063 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12064 expand_vec_perm for constant permutation vectors. Assert that
12065 the mode of variable permutation vectors is the integer equivalent
12066 of the mode that is being permuted.
12067 * optabs-query.h (selector_fits_mode_p): Declare.
12068 * optabs-query.c: Include vec-perm-indices.h.
12069 (selector_fits_mode_p): New function.
12070 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12071 is defined, instead of checking whether the vec_perm_const_optab
12072 exists. Use targetm.vectorize.vec_perm_const instead of
12073 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12074 fit in the vector mode before using a variable permute.
12075 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12076 vec_perm_indices instead of an rtx.
12077 (expand_vec_perm): Replace with...
12078 (expand_vec_perm_const): ...this new function. Take the selector
12079 as a vec_perm_indices rather than an rtx. Also take the mode of
12080 the selector. Update call to shift_amt_for_vec_perm_mask.
12081 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12082 Use vec_perm_indices::new_expanded_vector to expand the original
12083 selector into bytes. Check whether the indices fit in the vector
12084 mode before using a variable permute.
12085 (expand_vec_perm_var): Make global.
12086 (expand_mult_highpart): Use expand_vec_perm_const.
12087 * fold-const.c: Includes vec-perm-indices.h.
12088 * tree-ssa-forwprop.c: Likewise.
12089 * tree-vect-data-refs.c: Likewise.
12090 * tree-vect-generic.c: Likewise.
12091 * tree-vect-loop.c: Likewise.
12092 * tree-vect-slp.c: Likewise.
12093 * tree-vect-stmts.c: Likewise.
12094 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12096 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12097 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12098 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12099 (aarch64_vectorize_vec_perm_const): ...this new function.
12100 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12101 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12102 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12103 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12104 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12105 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12106 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12108 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12109 check for NEON modes.
12110 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12111 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12112 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12113 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12115 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12116 the old VEC_PERM_CONST conditions.
12117 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12118 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12119 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12120 (ia64_vectorize_vec_perm_const_ok): Merge into...
12121 (ia64_vectorize_vec_perm_const): ...this new function.
12122 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12123 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12124 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12125 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12126 * config/mips/mips.c (mips_expand_vec_perm_const)
12127 (mips_vectorize_vec_perm_const_ok): Merge into...
12128 (mips_vectorize_vec_perm_const): ...this new function.
12129 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12130 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12131 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12132 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12133 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12134 (rs6000_expand_vec_perm_const): Delete.
12135 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12137 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12138 (altivec_expand_vec_perm_const_le): Take each operand individually.
12139 Operate on constant selectors rather than rtxes.
12140 (altivec_expand_vec_perm_const): Likewise. Update call to
12141 altivec_expand_vec_perm_const_le.
12142 (rs6000_expand_vec_perm_const): Delete.
12143 (rs6000_vectorize_vec_perm_const_ok): Delete.
12144 (rs6000_vectorize_vec_perm_const): New function.
12145 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12146 an element count and rtx array.
12147 (rs6000_expand_extract_even): Update call accordingly.
12148 (rs6000_expand_interleave): Likewise.
12149 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12150 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12151 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12152 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12153 (rs6000_expand_vec_perm_const): Delete.
12154 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12155 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12156 (altivec_expand_vec_perm_const_le): Take each operand individually.
12157 Operate on constant selectors rather than rtxes.
12158 (altivec_expand_vec_perm_const): Likewise. Update call to
12159 altivec_expand_vec_perm_const_le.
12160 (rs6000_expand_vec_perm_const): Delete.
12161 (rs6000_vectorize_vec_perm_const_ok): Delete.
12162 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12163 reference to the SPE evmerge intructions.
12164 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12165 an element count and rtx array.
12166 (rs6000_expand_extract_even): Update call accordingly.
12167 (rs6000_expand_interleave): Likewise.
12168 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12169 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12171 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12173 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12175 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12176 vector mode and that that mode matches the mode of the data
12178 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12179 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12180 directly using expand_vec_perm_1 when forcing selectors into
12182 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12184 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12186 * optabs-query.h (can_vec_perm_p): Delete.
12187 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12188 * optabs-query.c (can_vec_perm_p): Split into...
12189 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12190 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12191 particular selector is valid.
12192 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12193 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12194 (vect_grouped_load_supported): Likewise.
12195 (vect_shift_permute_load_chain): Likewise.
12196 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12197 (vect_transform_slp_perm_load): Likewise.
12198 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12199 (vectorizable_bswap): Likewise.
12200 (vect_gen_perm_mask_checked): Likewise.
12201 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12202 implementations of variable permutation vectors into account
12203 when deciding which selector to use.
12204 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12205 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12206 with a false third argument.
12207 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12208 to test whether the constant selector is valid and can_vec_perm_var_p
12209 to test whether a variable selector is valid.
12211 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12213 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12214 * optabs-query.c (can_vec_perm_p): Likewise.
12215 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12216 instead of vec_perm_indices.
12217 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12218 (vect_gen_perm_mask_checked): Likewise,
12219 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12220 (vect_gen_perm_mask_checked): Likewise,
12222 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12224 * optabs-query.h (qimode_for_vec_perm): Declare.
12225 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12226 (qimode_for_vec_perm): ...this new function.
12227 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12229 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12231 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12232 does not have a conditional at the top.
12234 2018-01-02 Richard Biener <rguenther@suse.de>
12236 * ipa-inline.c (big_speedup_p): Fix expression.
12238 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12241 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12244 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12248 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12249 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12250 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12251 cond_taken_branch_cost 3->4.
12253 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12255 PR tree-optimization/83581
12256 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12257 TODO_cleanup_cfg if any changes have been made.
12259 PR middle-end/83608
12260 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12261 convert_modes if target mode has the right side, but different mode
12264 PR middle-end/83609
12265 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12266 last argument when extracting from CONCAT. If either from_real or
12267 from_imag is NULL, use expansion through memory. If result is not
12268 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12269 the parts directly to inner mode, if even that fails, use expansion
12272 PR middle-end/83623
12273 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12274 check for bswap in mode rather than HImode and use that in expand_unop
12277 Copyright (C) 2018 Free Software Foundation, Inc.
12279 Copying and distribution of this file, with or without modification,
12280 are permitted in any medium without royalty provided the copyright
12281 notice and this notice are preserved.