1 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5 remove_member_access_dummy_vars): New functions.
6 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7 lower_omp_1, execute_lower_omp): Use them.
9 PR rtl-optimization/84308
10 * shrink-wrap.c (spread_components): Release todo vector.
12 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
14 PR rtl-optimization/57193
15 * ira-color.c (struct allocno_color_data): Add member
16 conflict_allocno_hard_prefs.
17 (update_conflict_allocno_hard_prefs): New.
18 (bucket_allocno_compare_func): Add a preference based on
19 conflict_allocno_hard_prefs.
20 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
21 (color_allocnos): Remove a dead code. Initiate
22 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
24 2018-02-09 Jakub Jelinek <jakub@redhat.com>
27 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
28 constraint from =wa to wa. Avoid a subreg on the output operand,
29 instead use a pseudo and subreg it in a move.
30 (p9_xxbrd_<mode>): Changed to ...
31 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
32 (p9_xxbrd_v2df): New expander.
33 (p9_xxbrw_<mode>): Changed to ...
34 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
35 (p9_xxbrw_v4sf): New expander.
37 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
39 * config/rx.md: updated "movsicc" expand to be matched by GCC
40 * testsuite/gcc.target/rx/movsicc.c: new test case
42 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
45 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
46 multiply in 32-bit mode.
47 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
48 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
51 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
53 * config/rx/constraints.md: added new constraint CALL_OP_SYMBOL_REF
54 to allow or block "symbol_ref" depending on value of TARGET_JSR
55 * config/rx/rx.md: use CALL_OP_SYMBOL_REF in call_internal and
56 call_value_internal insns
58 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
61 * dwarf2out.c (is_trivial_indirect_ref): New function.
62 (dwarf2out_late_global_decl): Do not generate a location
63 attribute for variables that have a non-trivial DECL_VALUE_EXPR
64 and that are not defined in the current unit.
66 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
68 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
69 instead of a libcall for UNORDERED.
71 2018-02-09 Tamar Christina <tamar.christina@arm.com>
74 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
75 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
77 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
80 * config/s390/s390.c (s390_set_current_function): Invoke
81 s390_indirect_branch_settings also if fndecl didn't change.
83 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
85 * config/rs6000/rs6000.md (blockage): Set length to zero.
87 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
89 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
91 2018-02-09 Jakub Jelinek <jakub@redhat.com>
94 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
95 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
99 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
100 PARALLEL incoming that failed vt_get_decl_and_offset check.
103 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
104 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
106 (get_variable_section): For decls in named .bss* sections pass true as
107 second argument to bss_initializer_p.
109 2018-02-09 Marek Polacek <polacek@redhat.com>
110 Jakub Jelinek <jakub@redhat.com>
113 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
114 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
115 Sync some changes from cxx_fold_indirect_ref.
117 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
119 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
121 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
122 (BLOCK_INLINE_ENTRY_LABEL): New.
123 (dwarf2out_var_location): Disregard inline entry markers.
124 (inline_entry_data): New struct.
125 (inline_entry_data_hasher): New hashtable type.
126 (inline_entry_data_hasher::hash): New.
127 (inline_entry_data_hasher::equal): New.
128 (inline_entry_data_table): New variable.
129 (add_high_low_attributes): Add DW_AT_entry_pc and
130 DW_AT_GNU_entry_view attributes if a pending entry is found
131 in inline_entry_data_table. Add old entry_pc attribute only
132 if debug nonbinding markers are disabled.
133 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
135 (block_within_block_p, dwarf2out_inline_entry): New.
136 (dwarf2out_finish): Check that no entries remained in
137 inline_entry_data_table.
138 * final.c (reemit_insn_block_notes): Handle inline entry notes.
139 (final_scan_insn, notice_source_line): Likewise.
140 (rest_of_clean_state): Skip inline entry markers.
141 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
143 * gimple.c (gimple_build_debug_inline_entry): New.
144 * gimple.h (enum gimple_debug_subcode): Add
145 GIMPLE_DEBUG_INLINE_ENTRY.
146 (gimple_build_debug_inline_entry): Declare.
147 (gimple_debug_inline_entry_p): New.
148 (gimple_debug_nonbind_marker_p): Adjust.
149 * insn-notes.def (INLINE_ENTRY): New.
150 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
151 inline entry marker notes.
152 (print_insn): Likewise.
153 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
154 (INSN_DEBUG_MARKER_KIND): Likewise.
155 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
156 * tree-inline.c (expand_call_inline): Build and insert
157 debug_inline_entry stmt.
158 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
159 inline entry blocks early, if nonbind markers are enabled.
160 (dump_scope_block): Dump fragment info.
161 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
162 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
163 (gimple_build_debug_inline_entry): New.
164 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
165 Enable/disable inline entry points too.
166 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
167 (DEBUG_INSN): Describe inline entry markers.
169 * common.opt (gvariable-location-views): New.
170 (gvariable-location-views=incompat5): New.
171 * config.in: Rebuilt.
172 * configure: Rebuilt.
173 * configure.ac: Test assembler for view support.
174 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
175 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
176 * dwarf2out.c (var_loc_view): New typedef.
177 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
178 (dwarf2out_locviews_in_attribute): New.
179 (dwarf2out_locviews_in_loclist): New.
180 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
181 (enum dw_line_info_opcode): Add LI_adv_address.
182 (struct dw_line_info_table): Add view.
183 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
184 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
185 (zero_view_p): New variable.
186 (ZERO_VIEW_P): New macro.
187 (output_asm_line_debug_info): New.
188 (struct var_loc_node): Add view.
189 (add_AT_view_list, AT_loc_list): New.
190 (add_var_loc_to_decl): Add view param. Test it against last.
191 (new_loc_list): Add view params. Record them.
192 (AT_loc_list_ptr): Handle loc and view lists.
193 (view_list_to_loc_list_val_node): New.
194 (print_dw_val): Handle dw_val_class_view_list.
195 (size_of_die): Likewise.
196 (value_format): Likewise.
197 (loc_list_has_views): New.
198 (gen_llsym): Set vl_symbol too.
199 (maybe_gen_llsym, skip_loc_list_entry): New.
200 (dwarf2out_maybe_output_loclist_view_pair): New.
201 (output_loc_list): Output view list or entries too.
202 (output_view_list_offset): New.
203 (output_die): Handle dw_val_class_view_list.
204 (output_dwarf_version): New.
205 (output_compilation_unit_header): Use it.
206 (output_skeleton_debug_sections): Likewise.
207 (output_rnglists, output_line_info): Likewise.
208 (output_pubnames, output_aranges): Update version comments.
209 (output_one_line_info_table): Output view numbers in asm comments.
210 (dw_loc_list): Determine current endview, pass it to new_loc_list.
211 Call maybe_gen_llsym.
212 (loc_list_from_tree_1): Adjust.
213 (add_AT_location_description): Create view list attribute if
214 needed, check it's absent otherwise.
215 (convert_cfa_to_fb_loc_list): Adjust.
216 (maybe_emit_file): Call output_asm_line_debug_info for test.
217 (dwarf2out_var_location): Reset views as needed. Precompute
218 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
220 (new_line_info_table): Reset next view.
221 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
222 (dwarf2out_source_line): Likewise. Output view resets and labels to
223 the assembler, or select appropriate line info opcodes.
224 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
225 (optimize_string_length): Catch it. Adjust.
226 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
227 dw_val_class_view_list, and remove it if no longer needed.
228 (hash_loc_list): Hash view numbers.
229 (loc_list_hasher::equal): Compare them.
230 (optimize_location_lists): Check whether a view list symbol is
231 needed, and whether the locview attribute is present, and
232 whether they match. Remove the locview attribute if no longer
234 (index_location_lists): Call skip_loc_list_entry for test.
235 (dwarf2out_finish): Call output_asm_line_debug_info for test.
236 Use output_dwarf_version.
237 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
238 (struct dw_val_node): Add val_view_list.
239 * final.c (SEEN_NEXT_VIEW): New.
240 (set_next_view_needed): New.
241 (clear_next_view_needed): New.
242 (maybe_output_next_view): New.
243 (final_start_function): Rename to...
244 (final_start_function_1): ... this. Take pointer to FIRST,
245 add SEEN parameter. Emit param bindings in the initial view.
246 (final_start_function): Reintroduce SEEN-less interface.
247 (final): Rename to...
248 (final_1): ... this. Take SEEN parameter. Output final pending
249 next view at the end.
250 (final): Reintroduce seen-less interface.
251 (final_scan_insn): Output pending next view before switching
252 sections or ending a block. Mark the next view as needed when
253 outputting variable locations. Notify debug backend of section
254 changes, and of location view changes.
255 (rest_of_handle_final): Adjust.
256 * toplev.c (process_options): Autodetect value for debug variable
257 location views option. Warn on incompat5 without -gdwarf-5.
258 * doc/invoke.texi (gvariable-location-views): New.
259 (gvariable-location-views=incompat5): New.
260 (gno-variable-location-views): New.
262 2018-02-08 David Malcolm <dmalcolm@redhat.com>
264 PR tree-optimization/84136
265 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
266 that the result of find_edge is non-NULL.
268 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
271 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
272 storing integer register in SImode. Fix cost of 256 and 512
273 byte aligned SSE register store.
275 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
277 * config/i386/i386.c (ix86_multiplication_cost): Fix
278 multiplication cost for TARGET_AVX512DQ.
280 2018-02-08 Marek Polacek <polacek@redhat.com>
282 PR tree-optimization/84238
283 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
286 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
288 PR tree-optimization/84265
289 * tree-vect-stmts.c (vectorizable_store): Don't treat
290 VMAT_CONTIGUOUS accesses as grouped.
291 (vectorizable_load): Likewise.
293 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
295 PR tree-optimization/81635
296 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
297 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
298 (test_round_for_mask): New functions.
299 (wide_int_cc_tests): Call test_round_for_mask.
300 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
301 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
302 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
303 range returned by get_range_info.
305 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
308 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
309 * symtab.c: Include builtins.h
310 (symtab_node::output_to_lto_symbol_table_p): Move here
311 from lto-streamer-out.c:output_symbol_p.
312 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
313 (output_symbol_p): Move all logic to symtab.c
314 (produce_symtab): Update.
316 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
318 * config/s390/s390-opts.h (enum indirect_branch): Define.
319 * config/s390/s390-protos.h (s390_return_addr_from_memory)
320 (s390_indirect_branch_via_thunk)
321 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
322 (enum s390_indirect_branch_type): Define.
323 * config/s390/s390.c (struct s390_frame_layout, struct
324 machine_function): Remove.
325 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
326 (indirect_branch_table_label_no, indirect_branch_table_name):
328 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
329 (enum s390_indirect_branch_option): Define.
330 (s390_return_addr_from_memory): New function.
331 (s390_handle_string_attribute): New function.
332 (s390_attribute_table): Add new attribute handler.
333 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
334 (s390_indirect_branch_via_thunk): New function.
335 (s390_indirect_branch_via_inline_thunk): New function.
336 (s390_function_ok_for_sibcall): When jumping via thunk disallow
337 sibling call optimization for non z10 compiles.
338 (s390_emit_call): Force indirect branch target to be a single
339 register. Add r1 clobber for non-z10 compiles.
340 (s390_emit_epilogue): Emit return jump via return_use expander.
341 (s390_reorg): Handle JUMP_INSNs as execute targets.
342 (s390_option_override_internal): Perform validity checks for the
343 new command line options.
344 (s390_indirect_branch_attrvalue): New function.
345 (s390_indirect_branch_settings): New function.
346 (s390_set_current_function): Invoke s390_indirect_branch_settings.
347 (s390_output_indirect_thunk_function): New function.
348 (s390_code_end): Implement target hook.
349 (s390_case_values_threshold): Implement target hook.
350 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
352 * config/s390/s390.h (struct s390_frame_layout)
353 (struct machine_function): Move here from s390.c.
354 (TARGET_INDIRECT_BRANCH_NOBP_RET)
355 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
356 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
357 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
358 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
359 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
360 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
361 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
362 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
363 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
364 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
365 (mnemonic attribute): Add values which aren't recognized
367 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
368 pattern for branch conversion. Fix mnemonic attribute.
369 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
370 indirect branch via thunk if requested.
371 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
372 ("*indirect_jump"): Disable for branch conversion using out of
374 ("indirect_jump_via_thunk<mode>_z10")
375 ("indirect_jump_via_thunk<mode>")
376 ("indirect_jump_via_inlinethunk<mode>_z10")
377 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
378 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
379 ("casesi_jump_via_inlinethunk<mode>_z10")
380 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
381 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
382 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
383 ("*indirect2_jump"): Disable for branch conversion.
384 ("casesi_jump"): Turn into expander and expand patterns for branch
386 ("return_use"): New expander.
387 ("*return"): Emit return via thunk and rename it to ...
388 ("*return<mode>"): ... this one.
389 * config/s390/s390.opt: Add new options and and enum for the
392 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
394 * lra-constraints.c (match_reload): Unconditionally use
395 gen_lowpart_SUBREG, rather than selecting between that
396 and equivalent gen_rtx_SUBREG code.
398 2018-02-08 Richard Biener <rguenther@suse.de>
400 PR tree-optimization/84233
401 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
402 changed flag instead of boguously re-using phi_inserted.
404 2018-02-08 Martin Jambor <mjambor@suse.cz>
406 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
407 static local variables.
409 2018-02-08 Richard Biener <rguenther@suse.de>
411 PR tree-optimization/84278
412 * tree-vect-stmts.c (vectorizable_store): When looking for
413 smaller vector types to perform grouped strided loads/stores
414 make sure the mode is supported by the target.
415 (vectorizable_load): Likewise.
417 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
419 * config/aarch64/aarch64.c (aarch64_components_for_bb):
420 Increase LDP/STP opportunities by adding adjacent callee-saves.
422 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
424 PR rtl-optimization/84068
425 PR rtl-optimization/83459
426 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
428 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
430 PR tree-optimization/84224
431 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
432 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
435 2018-02-07 Iain Sandoe <iain@codesourcery.com>
438 * config/rs6000/altivec.md (*restore_world): Remove LR use.
439 * config/rs6000/predicates.md (restore_world_operation): Adjust op
440 count, remove one USE.
442 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
444 * doc/install.texi (Configuration): Document the
445 --with-long-double-format={ibm,ieee} PowerPC configuration
449 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
450 Convert from define_expand to be define_insn_and_split. Rework
451 float/double/_Float128 conversions to QI/HI/SImode to work with
452 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
453 conversions to QI/HImode types did a store and then a load to
454 truncate the value. For conversions to VSX registers, don't split
455 the insn, instead emit the code directly. Use the code iterator
456 any_fix to combine signed and unsigned conversions.
457 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
458 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
459 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
460 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
461 (fix_<mode>di2_hw): Likewise.
462 (fixuns_<mode>di2_hw): Likewise.
463 (fix_<mode>si2_hw): Likewise.
464 (fixuns_<mode>si2_hw): Likewise.
465 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
466 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
467 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
468 fix<uns>_trunc<SFDF:mode>si2_p8.
469 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
471 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
472 (fix<uns>_<mode>_mem): Likewise.
473 (fctiw<u>z_<mode>_mem): Likewise.
474 (fix<uns>_<mode>_mem): Likewise.
475 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
476 the register allocator from doing a direct move to the GPRs to do
477 a store, and instead use the ISA 3.0 store byte/half-word from
478 vector register instruction. For IEEE 128-bit floating point,
479 also optimize stores of 32-bit ints.
480 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
482 2018-02-07 Alan Hayward <alan.hayward@arm.com>
484 * genextract.c (push_pathstr_operand): New function to support
486 (walk_rtx): Call push_pathstr_operand.
487 (print_path): Support [a-zA-Z].
489 2018-02-07 Richard Biener <rguenther@suse.de>
491 PR tree-optimization/84037
492 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
493 (cse_and_gimplify_to_preheader): Declare.
494 (vect_get_place_in_interleaving_chain): Likewise.
495 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
497 (_loop_vec_info::~_loop_vec_info): Delete it.
498 (cse_and_gimplify_to_preheader): New function.
499 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
500 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
501 (vectorizable_load): Likewise. For grouped stores always base
502 the IV on the first element.
503 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
504 condition before gimplifying.
506 2018-02-07 Jakub Jelinek <jakub@redhat.com>
508 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
509 *DIV_EXPR and *MOD_EXPR.
511 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
514 * config/i386/i386.c (ix86_option_override_internal): Mask out
515 the CF_SET bit when checking -fcf-protection.
517 2018-02-07 Tom de Vries <tom@codesourcery.com>
520 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
523 2018-02-07 Richard Biener <rguenther@suse.de>
525 PR tree-optimization/84204
526 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
529 PR tree-optimization/84205
530 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
531 special-case isl_ast_op_zdiv_r.
533 PR tree-optimization/84223
534 * graphite-scop-detection.c (gather_bbs::before_dom_children):
535 Only add conditions from within the region.
536 (gather_bbs::after_dom_children): Adjust.
538 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
541 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
542 * config/avr/avr.md: Only post-reload split REG-REG moves if
543 either register is REGERAL_REG_P.
545 2018-02-07 Jakub Jelinek <jakub@redhat.com>
547 PR tree-optimization/84235
548 * tree-ssa-scopedtables.c
549 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
550 if the subtraction is performed in floating point type where NaNs are
551 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
552 build 1. Formatting fix.
554 2018-02-06 Jakub Jelinek <jakub@redhat.com>
557 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
558 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
559 and skip it regardless of bb boundaries. Use CALL_P macro,
560 don't test INSN_P (insn) together with CALL_P or JUMP_P check
561 unnecessarily, formatting fix.
563 2018-02-06 Michael Collison <michael.collison@arm.com>
565 * config/arm/thumb2.md:
566 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
567 (*thumb_mov_notscc): Ditto.
569 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
572 * config/rs6000/rs6000.md (su code attribute): Use "u" for
573 unsigned_fix, not "s".
575 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
577 * configure.ac (gcc_fn_eh_frame_ro): New function.
578 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
579 correct .eh_frame permissions.
580 * configure: Regenerate.
582 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
584 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
587 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
589 * config/rs6000/rs6000.c (rs6000_option_override_internal):
590 Display warning message for -mno-speculate-indirect-jumps.
592 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
594 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
596 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
598 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
600 PR tree-optimization/84225
601 * tree-eh.c (find_trapping_overflow): Only call
602 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
604 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
607 * config/i386/i386.c: Reimplement the check of possible options
608 -mibt/-mshstk conbination. Change error messages.
609 * doc/invoke.texi: Fix a typo: remove extra '='.
611 2018-02-06 Marek Polacek <polacek@redhat.com>
613 PR tree-optimization/84228
614 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
616 2018-02-06 Tamar Christina <tamar.christina@arm.com>
619 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
620 emitted arch directives.
621 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
622 __ARM_FEATURE_COPROC before changing architectures.
624 2018-02-06 Richard Biener <rguenther@suse.de>
626 * config/i386/i386.c (print_reg): Fix typo.
627 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
629 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
631 * configure: Regenerate.
633 2018-02-05 Martin Sebor <msebor@redhat.com>
635 PR tree-optimization/83369
636 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
639 2018-02-05 Martin Liska <mliska@suse.cz>
641 * doc/invoke.texi: Cherry-pick upstream r323995.
643 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
645 * ira.c (ira_init_register_move_cost): Adjust comment.
647 2018-02-05 Martin Liska <mliska@suse.cz>
649 PR gcov-profile/84137
650 * doc/gcov.texi: Fix typo in documentation.
652 2018-02-05 Martin Liska <mliska@suse.cz>
654 PR gcov-profile/83879
655 * doc/gcov.texi: Document necessity of --dynamic-list-data when
656 using dlopen functionality.
658 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
660 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
661 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
662 _mm_maskz_range_ss, _mm_mask_range_round_ss,
663 _mm_maskz_range_round_ss): New intrinsics.
664 (__builtin_ia32_rangesd128_round)
665 (__builtin_ia32_rangess128_round): Remove.
666 (__builtin_ia32_rangesd128_mask_round,
667 __builtin_ia32_rangess128_mask_round): New builtins.
668 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
669 __builtin_ia32_rangess128_round): Remove.
670 (__builtin_ia32_rangesd128_mask_round,
671 __builtin_ia32_rangess128_mask_round): New builtins.
672 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
673 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
674 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
675 "<round_saeonly_constraint>")): Changed to ...
676 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
677 "<round_saeonly_scalar_constraint>")): ... this.
678 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
679 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
680 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
681 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
682 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
684 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
686 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
688 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
689 Remove all values except native, 8540 and 8548.
691 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
693 * config/i386/i386.c (ix86_output_function_return): Pass
694 INVALID_REGNUM, instead of -1, as invalid register number to
695 indirect_thunk_name and output_indirect_thunk.
697 2018-02-02 Julia Koval <julia.koval@intel.com>
699 * config.gcc: Add -march=icelake.
700 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
701 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
702 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
703 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
704 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
705 (processor_target_table): Add icelake.
706 (ix86_option_override_internal): Handle new PTAs.
707 (get_builtin_code_for_version): Handle icelake.
708 (M_INTEL_COREI7_ICELAKE): New.
709 (fold_builtin_cpu): Handle icelake.
710 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
711 * doc/invoke.texi: Add -march=icelake.
713 2018-02-02 Julia Koval <julia.koval@intel.com>
715 * config/i386/i386.c (ix86_option_override_internal): Change flags type
717 * wide-int-bitmask.h: New.
719 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
722 * config/i386/i386.md: Replace Pmode with word_mode in
723 builtin_setjmp_setup and builtin_longjmp to support x32.
725 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
729 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
731 (rs6000_supported_cpu_names): New static variable.
732 (linux_cpu_translation_table): Likewise.
733 (elf_platform) <cpu>: Define new static variable and use it.
734 Translate kernel AT_PLATFORM name to canonical name if needed.
735 Error if platform name is unknown.
737 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
740 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
742 2018-02-01 Jeff Law <law@redhat.com>
745 * config/i386/i386.c (release_scratch_register_on_entry): Add new
746 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
747 the scratch if RELEASE_VIA_POP is false.
748 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
749 If we have to save a temporary register, decrement SIZE appropriately.
750 Pass new arguments to release_scratch_register_on_entry.
751 (ix86_adjust_stack_and_probe): Likewise.
752 (ix86_emit_probe_stack_range): Pass new arguments to
753 release_scratch_register_on_entry.
755 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
757 PR rtl-optimization/84157
758 * combine.c (change_zero_ext): Use REG_P predicate in
759 front of HARD_REGISTER_P predicate.
761 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
763 * config/avr/avr.c (avr_option_override): Move disabling of
764 -fdelete-null-pointer-checks to...
765 * common/config/avr/avr-common.c (avr_option_optimization_table):
768 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
770 PR tree-optimization/81635
771 * tree-data-ref.c (split_constant_offset_1): For types that
772 wrap on overflow, try to use range info to prove that wrapping
775 2018-02-01 Renlin Li <renlin.li@arm.com>
778 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
780 (aarch64_register_move_cost): Likewise.
781 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
783 (REG_CLASS_NAMES): Likewise.
784 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
785 TAILCALL_ADDR_REGS. Remove IP registers.
786 * config/aarch64/aarch64.md (Ucs): Update register constraint.
788 2018-02-01 Richard Biener <rguenther@suse.de>
790 * domwalk.h (dom_walker::dom_walker): Add additional constructor
791 for specifying RPO order and allow NULL for that.
792 * domwalk.c (dom_walker::dom_walker): Likewise.
793 (dom_walker::walk): Handle NULL RPO order.
794 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
796 (rewrite_update_dom_walker): Likewise.
797 (mark_def_dom_walker): Likewise.
799 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
801 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
802 (aarch64_maybe_expand_sve_subreg_move): Declare.
803 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
804 * config/aarch64/predicates.md (aarch64_any_register_operand): New
806 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
807 that are semantically a reverse operation.
808 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
809 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
810 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
812 (aarch64_can_change_mode_class): For big-endian, forbid changes
813 between two SVE modes if they have different element sizes.
815 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
817 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
818 the TImode handling for big-endian targets.
820 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
822 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
823 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
825 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
826 Remove BSWAP handing for big-endian targets and use the form of
827 LD1RQ appropariate for the mode.
829 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
831 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
832 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
835 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
838 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
839 check for operands that need to go through aarch64_sve_reload_be.
841 2018-02-01 Jakub Jelinek <jakub@redhat.com>
843 PR tree-optimization/81661
844 PR tree-optimization/84117
845 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
846 * tree-eh.c: Include gimplify.h.
847 (find_trapping_overflow, replace_trapping_overflow,
848 rewrite_to_non_trapping_overflow): New functions.
849 * tree-vect-loop.c: Include tree-eh.h.
850 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
851 * tree-data-ref.c: Include tree-eh.h.
852 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
854 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
856 PR rtl-optimization/84123
857 * combine.c (change_zero_ext): Check if hard register satisfies
858 can_change_dest_mode before calling gen_lowpart_SUBREG.
860 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
863 * ira.c (ira_init_register_move_cost): Remove assert.
865 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
867 PR rtl-optimization/84071
868 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
869 * doc/tm.texi: Regenerate.
871 2018-01-31 Richard Biener <rguenther@suse.de>
873 PR tree-optimization/84132
874 * tree-data-ref.c (analyze_miv_subscript): Properly
875 check whether evolution_function_is_affine_multivariate_p
876 before calling gcd_of_steps_may_divide_p.
878 2018-01-31 Julia Koval <julia.koval@intel.com>
881 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
882 * config/i386/i386.md (rdpid_rex64) New.
883 (rdpid): Make 32bit only.
885 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
888 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
889 an IDENTIFIER_NODE for FUNCTION_TYPE's.
891 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
894 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
896 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
898 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
900 PR rtl-optimization/84071
901 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
902 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
904 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
906 * config/arc/arc.c (arc_handle_aux_attribute): New function.
907 (arc_attribute_table): Add 'aux' attribute.
908 (arc_in_small_data_p): Consider aux like variables.
909 (arc_is_aux_reg_p): New function.
910 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
911 (arc_get_aux_arg): New function.
912 (prepare_move_operands): Handle aux-register access.
913 (arc_handle_aux_attribute): New function.
914 * doc/extend.texi (ARC Variable attributes): Add subsection.
916 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
918 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
919 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
920 (arc_attribute_table): Add 'uncached' attribute.
921 (arc_print_operand): Print '.di' flag for uncached memory
923 (arc_in_small_data_p): Do not consider for small data the uncached
925 (arc_is_uncached_mem_p): New function.
926 * config/arc/predicates.md (compact_store_memory_operand): Check
927 for uncached memory accesses.
928 (nonvol_nonimm_operand): Likewise.
929 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
931 2018-01-31 Jakub Jelinek <jakub@redhat.com>
934 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
935 falign-loops=): Add Optimization flag.
937 2018-01-30 Jeff Law <law@redhat.com>
940 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
941 INT_REGISTERS_SAVED. Check it prior to calling
942 get_scratch_register_on_entry.
943 (ix86_adjust_stack_and_probe): Similarly.
944 (ix86_emit_probe_stack_range): Similarly.
945 (ix86_expand_prologue): Corresponding changes.
947 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
950 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
951 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
953 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
956 * lra-constraints.c (curr_insn_transform): Process AND in the
959 2018-01-30 Jakub Jelinek <jakub@redhat.com>
961 PR rtl-optimization/83986
962 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
963 dependence against last_pending_memory_flush in addition to
966 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
968 PR tree-optimization/81611
969 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
972 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
974 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
977 2018-01-30 Richard Biener <rguenther@suse.de>
978 Jakub Jelinek <jakub@redhat.com>
980 PR tree-optimization/84111
981 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
982 inner loops added during recursion, as they don't have up-to-date
985 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
988 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
989 (can_inline_edge_by_limits_p): ... here.
990 (can_early_inline_edge_p, check_callers,
991 update_caller_keys, update_callee_keys, recursive_inlining,
992 add_new_edges_to_heap, speculation_useful_p,
993 inline_small_functions,
994 inline_small_functions, flatten_function,
995 inline_to_all_callers_1): Update.
997 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
999 * profile-count.c (profile_count::combine_with_ipa_count): Handle
1002 2018-01-30 Richard Biener <rguenther@suse.de>
1004 PR tree-optimization/83008
1005 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
1006 invariant and constant vector uses in stmts when they need
1009 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1012 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
1013 * configure: Regenerate.
1015 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
1017 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
1019 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
1020 Use gen_rtx_REG rather than gen_lowpart.
1022 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
1024 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
1025 rather than 0 when creating partial subregs.
1027 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
1029 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
1032 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
1035 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
1036 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
1037 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
1038 flags. This restores the settings used before the 2017-07-24.
1039 Turning off pre increment/decrement/modify allows IVOPTS to
1040 optimize DF/SF loops where the index is an int.
1042 2018-01-29 Richard Biener <rguenther@suse.de>
1043 Kelvin Nilsen <kelvin@gcc.gnu.org>
1046 * tree-vect-stmts.c (vectorizable_call): Don't call
1047 targetm.vectorize_builtin_md_vectorized_function if callee is
1050 2018-01-22 Carl Love <cel@us.ibm.com>
1052 * doc/extend.tex: Fix typo in second arg in
1053 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
1055 2018-01-29 Richard Biener <rguenther@suse.de>
1057 PR tree-optimization/84086
1058 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
1059 (flush_ssaname_freelist): When SSA names were released reset
1060 the SCEV hash table.
1062 2018-01-29 Richard Biener <rguenther@suse.de>
1064 PR tree-optimization/84057
1065 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
1066 removed paths when removing edges.
1068 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
1070 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
1071 -mfunction-return=@var{choice}.
1073 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
1076 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
1078 (layout::print_source_line): Likewise.
1079 (test_get_line_width_without_trailing_whitespace): Add test cases.
1081 2018-01-27 Jakub Jelinek <jakub@redhat.com>
1084 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
1087 2018-01-26 Jim Wilson <jimw@sifive.com>
1089 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
1091 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
1094 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1096 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
1097 and CMP + SUB-immediate -> SUBS.
1099 2018-01-26 Martin Sebor <msebor@redhat.com>
1101 PR tree-optimization/83896
1102 * tree-ssa-strlen.c (get_string_len): Rename...
1103 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
1104 Avoid assuming length is constant.
1105 (handle_char_store): Use HOST_WIDE_INT for string length.
1107 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
1110 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
1111 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
1113 2018-01-26 Richard Biener <rguenther@suse.de>
1115 PR rtl-optimization/84003
1116 * dse.c (record_store): Only record redundant stores when
1117 the earlier store aliases at least all accesses the later one does.
1119 2018-01-26 Jakub Jelinek <jakub@redhat.com>
1121 PR rtl-optimization/83985
1122 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
1123 REG_CFA_RESTORE insns.
1124 (delete_unmarked_insns): Don't ignore separate shrink wrapping
1125 REG_CFA_RESTORE insns here.
1128 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
1129 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
1131 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
1133 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
1134 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
1135 (arc_init): Likewise.
1136 (arc_override_options): Likewise.
1137 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
1139 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
1141 * config/arc/arc.h (TARGET_DBNZ): Define.
1142 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
1143 properly set the tune attribute.
1144 (dbnz): Use TARGET_DBNZ guard.
1145 * config/arc/arc.opt (mtune): Add core3 option.
1147 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
1149 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
1150 recognize new pic like addresses.
1151 (arc_delegitimize_address): Clean up.
1153 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
1155 * config/arc/arc-arches.def: Option mrf16 valid for all
1157 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
1158 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
1159 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
1160 * config/arc/arc-tables.opt: Regenerate.
1161 * config/arc/arc.c (arc_conditional_register_usage): Handle
1162 reduced register file case.
1163 (arc_file_start): Set must have build attributes.
1164 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
1166 * config/arc/arc.opt (mrf16): Add new option.
1167 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
1168 * config/arc/genmultilib.awk: Handle new mrf16 option.
1169 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
1170 * config/arc/t-multilib: Regenerate.
1171 * doc/invoke.texi (ARC Options): Document mrf16 option.
1173 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
1175 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
1176 * config/arc/arc.c (arc_handle_secure_attribute): New function.
1177 (arc_attribute_table): Add 'secure_call' attribute.
1178 (arc_print_operand): Print secure call operand.
1179 (arc_function_ok_for_sibcall): Don't optimize tail calls when
1181 (arc_is_secure_call_p): New function. * config/arc/arc.md
1182 (call_i): Add support for sjli instruction.
1183 (call_value_i): Likewise.
1184 * config/arc/constraints.md (Csc): New constraint.
1186 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
1187 John Eric Martin <John.Martin@emmicro-us.com>
1189 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
1190 * config/arc/arc.c (_arc_jli_section): New struct.
1191 (arc_jli_section): New type.
1192 (rc_jli_sections): New static variable.
1193 (arc_handle_jli_attribute): New function.
1194 (arc_attribute_table): Add jli_always and jli_fixed attribute.
1195 (arc_file_end): New function.
1196 (TARGET_ASM_FILE_END): Define.
1197 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
1198 (arc_add_jli_section): New function.
1199 (jli_call_scan): Likewise.
1200 (arc_reorg): Call jli_call_scan.
1201 (arc_output_addsi): Remove 'S' from printing asm operand.
1202 (arc_is_jli_call_p): New function.
1203 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
1205 (movhi_insn): Likewise.
1206 (movsi_insn): Likewise.
1207 (movsi_set_cc_insn): Likewise.
1208 (loadqi_update): Likewise.
1209 (load_zeroextendqisi_update): Likewise.
1210 (load_signextendqisi_update): Likewise.
1211 (loadhi_update): Likewise.
1212 (load_zeroextendhisi_update): Likewise.
1213 (load_signextendhisi_update): Likewise.
1214 (loadsi_update): Likewise.
1215 (loadsf_update): Likewise.
1216 (movsicc_insn): Likewise.
1217 (bset_insn): Likewise.
1218 (bxor_insn): Likewise.
1219 (bclr_insn): Likewise.
1220 (bmsk_insn): Likewise.
1221 (bicsi3_insn): Likewise.
1222 (cmpsi_cc_c_insn): Likewise.
1223 (movsi_ne): Likewise.
1224 (movsi_cond_exec): Likewise.
1225 (clrsbsi2): Likewise.
1236 (call_i): Remove 'S' asm letter, add jli instruction.
1237 (call_value_i): Likewise.
1238 * config/arc/arc.op (mjli-always): New option.
1239 * config/arc/constraints.md (Cji): New constraint.
1240 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
1242 (subsf3_fpx): Likewise.
1243 (mulsf3_fpx): Likewise.
1244 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
1246 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
1248 * doc/invoke.texi (ARC): Document mjli-always option.
1250 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
1252 * config/rl78/rl78.c: if operand 2 is const avoid addition with 0
1253 and use incw and decw where possible
1254 * testsuite/gcc.target/rl78/test_addsi3_internal.c: new file
1256 2018-01-26 Richard Biener <rguenther@suse.de>
1258 PR tree-optimization/81082
1259 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
1260 association if it requires casting to unsigned.
1261 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
1262 from fold_plusminus_mult_expr to catch important cases late when
1263 range info is available.
1265 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1267 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
1268 * configure.ac (hidden_linkonce): New test.
1269 * configure: Regenerate.
1270 * config.in: Regenerate.
1272 2018-01-26 Julia Koval <julia.koval@intel.com>
1274 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
1275 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
1276 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
1277 _mm_mask_bitshuffle_epi64_mask): Fix type.
1278 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
1279 USI_FTYPE_V4DI_V4DI_USI): Remove.
1280 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
1281 __builtin_ia32_vpshufbitqmb256_mask,
1282 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
1283 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
1284 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
1286 2018-01-26 Alan Modra <amodra@gmail.com>
1289 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
1290 UNSPEC_VBPERMQ. Sort other unspecs.
1292 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
1294 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
1296 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
1299 * predict.c (drop_profile): Do not push/pop cfun; update also
1301 (handle_missing_profiles): Fix logic looking for zero profiles.
1303 2018-01-25 Jakub Jelinek <jakub@redhat.com>
1306 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
1307 on functions with #pragma omp declare simd or functions with simd
1309 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
1310 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
1311 Remove trailing \n from warning_at calls.
1313 2018-01-25 Tom de Vries <tom@codesourcery.com>
1316 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
1317 for neutered workers.
1319 2018-01-24 Joseph Myers <joseph@codesourcery.com>
1322 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
1323 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
1325 2018-01-24 Jeff Law <law@redhat.com>
1328 * i386.c (get_probe_interval): Move to earlier point.
1329 (ix86_compute_frame_layout): If -fstack-clash-protection and
1330 the frame is larger than the probe interval, then use pushes
1331 to save registers rather than reg->mem moves.
1332 (ix86_expand_prologue): Remove conditional for int_registers_saved
1335 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
1338 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
1339 min/max for never referenced object.
1341 2018-01-24 Jakub Jelinek <jakub@redhat.com>
1344 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
1346 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
1347 attributes from DECL_ATTRIBUTES (decl) without affecting
1348 DECL_ATTRIBUTES (current_function_decl).
1349 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
1350 functions with non-NULL DECL_ABSTRACT_ORIGIN.
1352 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
1354 PR tree-optimization/83979
1355 * fold-const.c (fold_comparison): Use constant_boolean_node
1356 instead of boolean_{true,false}_node.
1358 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
1360 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
1363 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1365 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
1366 Simplify the clause that sets the length attribute.
1367 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
1368 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
1369 clause that sets the length attribute.
1370 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
1372 2018-01-24 Tom de Vries <tom@codesourcery.com>
1375 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
1376 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
1377 Add strict parameter.
1378 (prevent_branch_around_nothing): Insert dummy insn between branch to
1379 label and label with no ptx insn inbetween.
1380 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
1382 2018-01-24 Tom de Vries <tom@codesourcery.com>
1385 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
1386 for neutered threads in warp.
1387 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
1389 2018-01-24 Richard Biener <rguenther@suse.de>
1391 PR tree-optimization/83176
1392 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
1395 2018-01-24 Richard Biener <rguenther@suse.de>
1397 PR tree-optimization/82819
1398 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
1399 code generating pluses that are no-ops in the target precision.
1401 2018-01-24 Richard Biener <rguenther@suse.de>
1404 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
1406 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
1408 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
1409 to merge probabilities.
1410 * predict.c (probably_never_executed): Also mark as cold functions
1411 with global 0 profile and guessed local profile.
1412 * profile-count.c (profile_probability::combine_with_count): New
1414 * profile-count.h (profile_probability::operator*,
1415 profile_probability::operator*=, profile_probability::operator/,
1416 profile_probability::operator/=): Reduce precision to adjusted
1417 and set value to guessed on contradictory divisions.
1418 (profile_probability::combine_with_freq): Remove.
1419 (profile_probability::combine_wiht_count): Declare.
1420 (profile_count::force_nonzero):: Set to adjusted.
1421 (profile_count::probability_in):: Set quality to adjusted.
1422 * tree-ssa-tail-merge.c (replace_block_by): Use
1425 2018-01-23 Andrew Waterman <andrew@sifive.com>
1426 Jim Wilson <jimw@sifive.com>
1428 * config/riscv/riscv.c (riscv_stack_boundary): New.
1429 (riscv_option_override): Set riscv_stack_boundary. Handle
1430 riscv_preferred_stack_boundary_arg.
1431 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
1432 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
1433 (STACK_BOUNDARY): Set to riscv_stack_boundary.
1434 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
1435 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
1436 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
1438 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
1441 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
1442 of struct ix86_frame.
1443 (ix86_expand_epilogue): Likewise. Add a local variable for
1444 the reg_save_offset field in struct ix86_frame.
1446 2018-01-23 Bin Cheng <bin.cheng@arm.com>
1448 PR tree-optimization/82604
1449 * tree-loop-distribution.c (enum partition_kind): New enum item
1450 PKIND_PARTIAL_MEMSET.
1451 (partition_builtin_p): Support above new enum item.
1452 (generate_code_for_partition): Ditto.
1453 (compute_access_range): Differentiate cases that equality can be
1454 proven at all loops, the innermost loops or no loops.
1455 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
1456 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
1457 (finalize_partitions, distribute_loop): Don't fuse partition of
1458 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
1459 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
1462 2018-01-23 Martin Liska <mliska@suse.cz>
1464 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
1465 order to ignore the predictor.
1466 (PRED_POLYMORPHIC_CALL): Likewise.
1467 (PRED_RECURSIVE_CALL): Likewise.
1469 2018-01-23 Martin Liska <mliska@suse.cz>
1471 * tree-profile.c (tree_profiling): Print function header to
1472 aware reader which function we are working on.
1473 * value-prof.c (gimple_find_values_to_profile): Do not print
1474 not interesting value histograms.
1476 2018-01-23 Martin Liska <mliska@suse.cz>
1478 * profile-count.h (enum profile_quality): Add
1479 profile_uninitialized as the first value. Do not number values
1480 as they are zero based.
1481 (profile_count::verify): Update sanity check.
1482 (profile_probability::verify): Likewise.
1484 2018-01-23 Nathan Sidwell <nathan@acm.org>
1486 * doc/invoke.texi (ffor-scope): Deprecate.
1488 2018-01-23 David Malcolm <dmalcolm@redhat.com>
1490 PR tree-optimization/83510
1491 * domwalk.c (set_all_edges_as_executable): New function.
1492 (dom_walker::dom_walker): Convert bool param
1493 "skip_unreachable_blocks" to enum reachability. Move setup of
1494 edge flags to set_all_edges_as_executable and only do it when
1495 reachability is REACHABLE_BLOCKS.
1496 * domwalk.h (enum dom_walker::reachability): New enum.
1497 (dom_walker::dom_walker): Convert bool param
1498 "skip_unreachable_blocks" to enum reachability.
1499 (set_all_edges_as_executable): New decl.
1500 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
1501 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
1503 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
1504 but converting true to REACHABLE_BLOCKS.
1505 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
1507 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
1508 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
1509 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
1511 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
1512 if check_all_array_refs will be called.
1514 2018-01-23 David Malcolm <dmalcolm@redhat.com>
1516 * tree.c (selftest::test_location_wrappers): Add more test
1519 2018-01-23 David Malcolm <dmalcolm@redhat.com>
1521 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
1522 (selftest::test_bit_in_range): Likewise.
1524 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
1527 * doc/sourcebuild.texi (vect_float): Say that the selector
1528 only describes the situation when -funsafe-math-optimizations is on.
1529 (vect_float_strict): Document.
1531 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
1533 PR tree-optimization/83965
1534 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
1535 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
1536 instead of checking only for a reduction.
1537 (vect_recog_widen_sum_pattern): Likewise.
1539 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
1541 * predict.c (probably_never_executed): Only use precise profile info.
1542 (compute_function_frequency): Skip after inlining hack since we now
1543 have quality checking.
1545 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
1547 * profile-count.h (profile_probability::very_unlikely,
1548 profile_probability::unlikely, profile_probability::even): Set
1549 precision to guessed.
1551 2018-01-23 Richard Biener <rguenther@suse.de>
1553 PR tree-optimization/83963
1554 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
1555 Properly terminate dominator walk when crossing the exit edge not
1556 when visiting its source block.
1558 2018-01-23 Jakub Jelinek <jakub@redhat.com>
1561 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
1562 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
1564 2018-01-22 Jakub Jelinek <jakub@redhat.com>
1566 PR tree-optimization/83957
1567 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
1568 semicolon after for body surrounded by braces.
1570 PR tree-optimization/83081
1571 * profile-count.h (profile_probability::split): New method.
1572 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
1573 Use profile_probability::split.
1574 (do_compare_rtx_and_jump): Fix adjustment of probabilities
1575 when splitting a single conditional jump into 2.
1577 2018-01-22 David Malcolm <dmalcolm@redhat.com>
1579 PR tree-optimization/69452
1580 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
1583 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
1585 * config/rl78/rl78-expand.md: New define_expand "bswaphi2"
1586 * config/rl78/rl78-virt.md: New define_insn "*bswaphi2_virt"
1587 * config/rl78/rl78-real.md: New define_insn "*bswaphi2_real"
1589 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
1591 * config/rl78/rl78-protos.h: New function declaration rl78_split_movdi
1592 * config/rl78/rl78.md: New define_expand "movdi"
1593 * config/rl78/rl78.c: New function definition rl78_split_movdi
1595 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
1598 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
1600 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
1601 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
1602 128-bit to produce an UNSPEC move to get the double word with the
1603 signbit and then a shift directly to do signbit.
1604 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
1605 implementation with a new version that just does either a direct
1606 move or a regular move. Move memory interface to separate insns.
1607 Move insns so they are next to the expander.
1608 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
1609 with signbit move. Split big and little endian case.
1610 (signbit<mode>2_dm_mem_le): Likewise.
1611 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
1612 (signbit<mode>2_dm2): Likewise.
1614 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
1616 * config/rl78/rl78.md: New define_expand "anddi3".
1618 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
1620 * config/rl78/rl78.md: New define_expand "umindi3".
1622 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
1624 * config/rl78/rl78.md: New define_expand "smindi3".
1626 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
1628 * config/rl78/rl78.md: New define_expand "smaxdi3".
1630 2018-01-22 Carl Love <cel@us.ibm.com>
1632 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
1633 LVX_V1TI): Add macro expansion.
1634 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
1635 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
1636 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
1637 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
1638 Change check to determine if the instruction is a byte reversing
1639 entry. Fix typo in comment.
1640 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
1641 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
1642 Add def_builtin calls for new builtins.
1643 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
1644 Add define_insn expansion.
1646 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
1648 * config/rl78/rl78.md: New define_expand "umaxdi3".
1650 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
1652 * config/rl78/rl78.c (rl78_note_reg_set): fixed dead reg check
1653 for non-QImode registers
1655 2018-01-22 Richard Biener <rguenther@suse.de>
1657 PR tree-optimization/83963
1658 * graphite-scop-detection.c (scop_detection::get_sese): Delay
1659 including the loop exit block.
1660 (scop_detection::merge_sese): Likewise.
1661 (scop_detection::add_scop): Do it here instead.
1663 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1665 * doc/sourcebuild.texi (arm_softfloat): Document.
1667 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
1670 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
1671 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
1672 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
1674 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1675 David Edelsohn <dje.gcc@gmail.com>
1678 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
1679 Change "crset eq" to "crset 2".
1680 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
1681 (*call_indirect_aix<mode>_nospec): Likewise.
1682 (*call_value_indirect_aix<mode>_nospec): Likewise.
1683 (*call_indirect_elfv2<mode>_nospec): Likewise.
1684 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
1685 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
1686 change assembly output from . to $.
1687 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
1688 (indirect_jump<mode>_nospec): Change assembly output from . to $.
1689 (*tablejump<mode>_internal1_nospec): Likewise.
1691 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
1694 * config/sh/sh_optimize_sett_clrt.cc:
1695 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
1697 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
1699 PR tree-optimization/83940
1700 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
1701 offset_dt to vect_constant_def rather than vect_unknown_def_type.
1702 (vect_check_load_store_mask): Add a mask_dt_out parameter and
1703 use it to pass back the definition type.
1704 (vect_check_store_rhs): Likewise rhs_dt_out.
1705 (vect_build_gather_load_calls): Add a mask_dt argument and use
1706 it instead of a call to vect_is_simple_use.
1707 (vectorizable_store): Update calls to vect_check_load_store_mask
1708 and vect_check_store_rhs. Use the dt returned by the latter instead
1709 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
1710 instead of calls to vect_is_simple_use. Pass the scalar rather
1711 than the vector operand to vect_is_simple_use when handling
1712 second and subsequent copies of an rhs value.
1713 (vectorizable_load): Update calls to vect_check_load_store_mask
1714 and vect_build_gather_load_calls. Use the cached mask_dt and
1715 gs_info.offset_dt instead of calls to vect_is_simple_use.
1717 2018-01-20 Jakub Jelinek <jakub@redhat.com>
1720 * tree-emutls.c: Include gimplify.h.
1721 (lower_emutls_2): New function.
1722 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
1723 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
1724 it before further processing.
1727 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
1728 UINTVAL (trueop1) instead of INTVAL (op1).
1730 2018-01-19 Jakub Jelinek <jakub@redhat.com>
1734 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
1735 INCOMING_FRAME_SP_OFFSET if not defined.
1736 (scan_trace): Add ENTRY argument. If true and
1737 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
1738 emit a note to adjust the CFA offset.
1739 (create_cfi_notes): Adjust scan_trace callers.
1740 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
1741 INCOMING_FRAME_SP_OFFSET in the CIE.
1742 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
1743 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
1745 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
1746 * doc/tm.texi: Regenerated.
1748 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1750 PR rtl-optimization/83147
1751 * lra-constraints.c (remove_inheritance_pseudos): Use
1752 lra_substitute_pseudo_within_insn.
1754 2018-01-19 Tom de Vries <tom@codesourcery.com>
1755 Cesar Philippidis <cesar@codesourcery.com>
1758 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
1760 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
1763 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
1764 spaces for function labels.
1766 2018-01-19 Martin Liska <mliska@suse.cz>
1768 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
1769 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
1770 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
1771 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
1772 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
1773 (PRED_CONST_RETURN): Change from 69 to 65.
1774 (PRED_NULL_RETURN): Change from 91 to 71.
1775 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
1776 (PRED_LOOP_GUARD): Change from 66 to 73.
1778 2018-01-19 Martin Liska <mliska@suse.cz>
1780 * predict.c (predict_insn_def): Add new assert.
1781 (struct branch_predictor): Change type to signed integer.
1782 (test_prediction_value_range): Amend test to cover
1784 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
1785 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
1786 (PRED_LOOP_ITERATIONS_MAX): Likewise.
1787 (PRED_LOOP_IV_COMPARE): Likewise.
1788 * predict.h (PROB_UNINITIALIZED): Define new constant.
1790 2018-01-19 Martin Liska <mliska@suse.cz>
1792 * predict.c (dump_prediction): Add new format for
1793 analyze_brprob.py script which is enabled with -details
1795 * profile-count.h (precise_p): New function.
1797 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
1799 PR tree-optimization/83922
1800 * tree-vect-loop.c (vect_verify_full_masking): Return false if
1801 there are no statements that need masking.
1802 (vect_active_double_reduction_p): New function.
1803 (vect_analyze_loop_operations): Use it when handling phis that
1804 are not in the loop header.
1806 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
1808 PR tree-optimization/83914
1809 * tree-vect-loop.c (vectorizable_induction): Don't convert
1810 init_expr or apply the peeling adjustment for inductions
1811 that are nested within the vectorized loop.
1813 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1815 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
1818 2018-01-18 Jakub Jelinek <jakub@redhat.com>
1822 * function.h (gimplify_parameters): Add gimple_seq * argument.
1823 * function.c: Include gimple.h and options.h.
1824 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
1825 for the added local temporaries if needed.
1826 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
1827 if there are any parameter cleanups, wrap whole body into a
1828 try/finally with the cleanups.
1830 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
1833 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
1834 Use GET_MODE_CLASS for scalar floating point.
1836 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
1840 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
1841 Fix call of call_cgraph_insertion_hooks.
1843 2018-01-18 Martin Sebor <msebor@redhat.com>
1845 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
1847 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
1850 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
1853 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
1856 * common.opt: (-ffile-prefix-map): New option.
1857 * opts.c (common_handle_option): Defer it.
1858 * opts-global.c (handle_common_deferred_options): Handle it.
1859 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
1860 * file-prefix-map.h: New file.
1861 (remap_debug_filename, add_debug_prefix_map): ...here.
1862 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
1863 * final.c (debug_prefix_map, add_debug_prefix_map
1864 remap_debug_filename): Move to...
1865 * file-prefix-map.c: New file.
1866 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
1867 generalize, get rid of alloca(), use strrchr() instead of strchr().
1868 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
1869 Implement in terms of add_prefix_map().
1870 (remap_macro_filename, remap_debug_filename): Implement in term of
1872 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
1873 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
1874 * dbxout.c: Include file-prefix-map.h.
1875 * varasm.c: Likewise.
1876 * vmsdbgout.c: Likewise.
1877 * xcoffout.c: Likewise.
1878 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
1879 * doc/cppopts.texi (-fmacro-prefix-map): Document.
1880 * doc/invoke.texi (-ffile-prefix-map): Document.
1881 (-fdebug-prefix-map): Update description.
1883 2018-01-18 Martin Liska <mliska@suse.cz>
1885 * config/i386/i386.c (indirect_thunk_name): Document that also
1887 (output_indirect_thunk): Document why both instructions
1888 (pause and lfence) are generated.
1890 2018-01-18 Richard Biener <rguenther@suse.de>
1892 PR tree-optimization/83887
1893 * graphite-scop-detection.c
1894 (scop_detection::get_nearest_dom_with_single_entry): Remove.
1895 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
1896 (scop_detection::merge_sese): Re-implement with a flood-fill
1897 algorithm that properly finds a SESE region if it exists.
1899 2018-01-18 Jakub Jelinek <jakub@redhat.com>
1902 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
1903 pointer_diff optimizations use view_convert instead of convert.
1905 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1907 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
1908 Generate different code for -mno-speculate-indirect-jumps.
1909 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
1910 (*call_indirect_aix<mode>): Disable for
1911 -mno-speculate-indirect-jumps.
1912 (*call_indirect_aix<mode>_nospec): New define_insn.
1913 (*call_value_indirect_aix<mode>): Disable for
1914 -mno-speculate-indirect-jumps.
1915 (*call_value_indirect_aix<mode>_nospec): New define_insn.
1916 (*sibcall_nonlocal_sysv<mode>): Generate different code for
1917 -mno-speculate-indirect-jumps.
1918 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
1920 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1922 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
1923 long double type, set the flags for noting the default long double
1924 type, even if we don't pass or return a long double type.
1926 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
1929 * ipa-inline.c (flatten_function): Do not overwrite final inlining
1932 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
1934 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
1935 support for merge[hl].
1936 (fold_mergehl_helper): New helper function.
1937 (tree-vector-builder.h): New #include for tree_vector_builder usage.
1938 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
1939 (altivec_vmrglw_direct): Add xxmrglw insn.
1941 2018-01-17 Andrew Waterman <andrew@sifive.com>
1943 * config/riscv/riscv.c (riscv_conditional_register_usage): If
1944 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
1946 2018-01-17 David Malcolm <dmalcolm@redhat.com>
1949 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
1950 call the lto_location_cache before reading the
1951 DECL_SOURCE_LOCATION of the types.
1953 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
1954 Richard Sandiford <richard.sandiford@linaro.org>
1956 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
1957 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
1958 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
1959 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
1960 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
1962 * config/aarch64/constraints.md (aarch64_movti_operand):
1964 * config/aarch64/predicates.md (Uti): Add new constraint.
1966 2018-01-17 Carl Love <cel@us.ibm.com>
1967 * config/rs6000/vsx.md (define_expand xl_len_r,
1968 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
1969 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
1971 (define_expand, define_insn): Move the shift left from the
1972 define_insn to the define_expand for lxvl and stxvl instructions.
1973 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
1974 and XL_LEN_R definitions to PURE.
1976 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
1978 * config/i386/i386.c (indirect_thunk_name): Declare regno
1979 as unsigned int. Compare regno with INVALID_REGNUM.
1980 (output_indirect_thunk): Ditto.
1981 (output_indirect_thunk_function): Ditto.
1982 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
1983 in the call to output_indirect_thunk_function.
1985 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
1988 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
1989 rather than the size of inner_type to determine the stack slot size
1990 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
1992 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
1995 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
1998 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
2000 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
2001 endian Linux systems to optionally enable multilibs for selecting
2002 the long double type if the user configured an explicit type.
2003 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
2004 have no long double multilibs if not defined.
2005 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
2006 warn if the user used -mabi={ieee,ibm}longdouble and we built
2007 multilibs for long double.
2008 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
2009 appropriate multilib option.
2010 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
2012 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
2013 for building long double multilibs.
2014 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
2016 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
2018 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
2021 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
2023 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
2026 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
2029 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
2032 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
2034 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
2035 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
2037 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
2039 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
2040 different rtl trees depending on TARGET_64BIT.
2041 (rs6000_gen_lvx): Likewise.
2043 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
2045 * config/visium/visium.md (nop): Tweak comment.
2046 (hazard_nop): Likewise.
2048 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2050 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
2051 -mspeculate-indirect-jumps.
2052 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
2053 for -mno-speculate-indirect-jumps.
2054 (*call_indirect_elfv2<mode>_nospec): New define_insn.
2055 (*call_value_indirect_elfv2<mode>): Disable for
2056 -mno-speculate-indirect-jumps.
2057 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
2058 (indirect_jump): Emit different RTL for
2059 -mno-speculate-indirect-jumps.
2060 (*indirect_jump<mode>): Disable for
2061 -mno-speculate-indirect-jumps.
2062 (*indirect_jump<mode>_nospec): New define_insn.
2063 (tablejump): Emit different RTL for
2064 -mno-speculate-indirect-jumps.
2065 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
2066 (tablejumpsi_nospec): New define_expand.
2067 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
2068 (tablejumpdi_nospec): New define_expand.
2069 (*tablejump<mode>_internal1): Disable for
2070 -mno-speculate-indirect-jumps.
2071 (*tablejump<mode>_internal1_nospec): New define_insn.
2072 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
2075 2018-01-16 Artyom Skrobov tyomitch@gmail.com
2077 * caller-save.c (insert_save): Drop unnecessary parameter. All
2080 2018-01-16 Jakub Jelinek <jakub@redhat.com>
2081 Richard Biener <rguenth@suse.de>
2084 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
2085 return early, inline manually is_gimple_sizepos. Make sure if we
2086 call gimplify_expr we don't end up with a gimple constant.
2087 * tree.c (variably_modified_type_p): Don't return true for
2088 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
2089 * gimplify.h (is_gimple_sizepos): Remove.
2091 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
2093 PR tree-optimization/83857
2094 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
2095 vectorizable_live_operation for pure SLP statements.
2096 (vectorizable_live_operation): Handle PHIs.
2098 2018-01-16 Richard Biener <rguenther@suse.de>
2100 PR tree-optimization/83867
2101 * tree-vect-stmts.c (vect_transform_stmt): Precompute
2102 nested_in_vect_loop_p since the scalar stmt may get invalidated.
2104 2018-01-16 Jakub Jelinek <jakub@redhat.com>
2107 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
2108 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
2109 If off is not INTEGER_CST, issue a may not be aligned warning
2110 rather than isn't aligned. Use isn%'t rather than isn't.
2111 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
2113 <case MULT_EXPR>: Improve the case when bottom and one of the
2114 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
2115 operand, in that case check if the other operand is multiple of
2116 bottom divided by the INTEGER_CST operand.
2118 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
2121 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
2122 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
2123 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
2124 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
2125 * config/pa/pa.c (pa_function_arg_advance): Likewise.
2126 (pa_function_arg, pa_arg_partial_bytes): Likewise.
2127 (pa_function_arg_size): New function.
2129 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
2131 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
2132 in a separate statement.
2134 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
2136 PR tree-optimization/83847
2137 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
2138 group gathers and scatters.
2140 2018-01-16 Jakub Jelinek <jakub@redhat.com>
2142 PR rtl-optimization/86620
2143 * params.def (max-sched-ready-insns): Bump minimum value to 1.
2145 PR rtl-optimization/83213
2146 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
2147 to last if both are JUMP_INSNs.
2149 PR tree-optimization/83843
2150 * gimple-ssa-store-merging.c
2151 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
2152 store_immediate_info for bswap/nop orig_stores.
2154 2018-01-15 Andrew Waterman <andrew@sifive.com>
2156 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
2158 <UDIV>: Increase cost if !TARGET_DIV.
2160 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
2162 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
2163 (define_attr "cr_logical_3op"): New.
2164 (cceq_ior_compare): Adjust.
2165 (cceq_ior_compare_complement): Adjust.
2166 (*cceq_rev_compare): Adjust.
2167 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
2168 (is_cracked_insn): Adjust.
2169 (insn_must_be_first_in_group): Adjust.
2170 * config/rs6000/40x.md: Adjust.
2171 * config/rs6000/440.md: Adjust.
2172 * config/rs6000/476.md: Adjust.
2173 * config/rs6000/601.md: Adjust.
2174 * config/rs6000/603.md: Adjust.
2175 * config/rs6000/6xx.md: Adjust.
2176 * config/rs6000/7450.md: Adjust.
2177 * config/rs6000/7xx.md: Adjust.
2178 * config/rs6000/8540.md: Adjust.
2179 * config/rs6000/cell.md: Adjust.
2180 * config/rs6000/e300c2c3.md: Adjust.
2181 * config/rs6000/e500mc.md: Adjust.
2182 * config/rs6000/e500mc64.md: Adjust.
2183 * config/rs6000/e5500.md: Adjust.
2184 * config/rs6000/e6500.md: Adjust.
2185 * config/rs6000/mpc.md: Adjust.
2186 * config/rs6000/power4.md: Adjust.
2187 * config/rs6000/power5.md: Adjust.
2188 * config/rs6000/power6.md: Adjust.
2189 * config/rs6000/power7.md: Adjust.
2190 * config/rs6000/power8.md: Adjust.
2191 * config/rs6000/power9.md: Adjust.
2192 * config/rs6000/rs64.md: Adjust.
2193 * config/rs6000/titan.md: Adjust.
2195 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
2197 * config/i386/predicates.md (indirect_branch_operand): Rewrite
2198 ix86_indirect_branch_register logic.
2200 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
2202 * config/i386/constraints.md (Bs): Update
2203 ix86_indirect_branch_register check. Don't check
2204 ix86_indirect_branch_register with GOT_memory_operand.
2206 * config/i386/predicates.md (GOT_memory_operand): Don't check
2207 ix86_indirect_branch_register here.
2208 (GOT32_symbol_operand): Likewise.
2210 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
2212 * config/i386/predicates.md (constant_call_address_operand):
2213 Rewrite ix86_indirect_branch_register logic.
2214 (sibcall_insn_operand): Likewise.
2216 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
2218 * config/i386/constraints.md (Bs): Replace
2219 ix86_indirect_branch_thunk_register with
2220 ix86_indirect_branch_register.
2222 * config/i386/i386.md (indirect_jump): Likewise.
2223 (tablejump): Likewise.
2224 (*sibcall_memory): Likewise.
2225 (*sibcall_value_memory): Likewise.
2226 Peepholes of indirect call and jump via memory: Likewise.
2227 * config/i386/i386.opt: Likewise.
2228 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2229 (GOT_memory_operand): Likewise.
2230 (call_insn_operand): Likewise.
2231 (sibcall_insn_operand): Likewise.
2232 (GOT32_symbol_operand): Likewise.
2234 2018-01-15 Jakub Jelinek <jakub@redhat.com>
2237 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
2238 type rather than type addr's type points to.
2239 (expand_omp_atomic_mutex): Likewise.
2240 (expand_omp_atomic): Likewise.
2242 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
2245 * config/i386/i386.c (output_indirect_thunk_function): Use
2246 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
2247 for __x86_return_thunk.
2249 2018-01-15 Richard Biener <rguenther@suse.de>
2252 * expmed.c (extract_bit_field_1): Fix typo.
2254 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2257 * config/arm/iterators.md (VF): New mode iterator.
2258 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
2259 Remove integer-related logic from pattern.
2260 (neon_vabd<mode>_3): Likewise.
2262 2018-01-15 Jakub Jelinek <jakub@redhat.com>
2265 * common.opt (fstrict-overflow): No longer an alias.
2266 (fwrapv-pointer): New option.
2267 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
2268 also for pointer types based on flag_wrapv_pointer.
2269 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
2270 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
2271 opts->x_flag_wrapv got set.
2272 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
2273 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
2274 POINTER_TYPE_OVERFLOW_UNDEFINED.
2275 * match.pd: Likewise in address comparison pattern.
2276 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
2278 2018-01-15 Richard Biener <rguenther@suse.de>
2281 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
2282 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
2283 Reset type names to their identifier if their TYPE_DECL doesn't
2284 have linkage (and thus is used for ODR and devirt).
2285 (save_debug_info_for_decl): Remove.
2286 (save_debug_info_for_type): Likewise.
2287 (add_tree_to_fld_list): Adjust.
2288 * tree-pretty-print.c (dump_generic_node): Make dumping of
2289 type names more robust.
2291 2018-01-15 Richard Biener <rguenther@suse.de>
2293 * BASE-VER: Bump to 8.0.1.
2295 2018-01-14 Martin Sebor <msebor@redhat.com>
2298 * builtins.c (check_access): Avoid warning when the no-warning bit
2301 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
2303 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
2304 * ira-color (allocno_hard_regs_compare): Likewise.
2306 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
2309 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
2310 Use .pushsection/.popsection.
2312 2018-01-14 Martin Sebor <msebor@redhat.com>
2315 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
2317 2018-01-14 Jakub Jelinek <jakub@redhat.com>
2319 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
2320 entry from extra_headers.
2321 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
2322 extra_headers, make the list bitwise identical to the i?86-*-* one.
2324 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
2326 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
2327 -mcmodel=large with -mindirect-branch=thunk,
2328 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
2329 -mfunction-return=thunk-extern.
2330 * doc/invoke.texi: Document -mcmodel=large is incompatible with
2331 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
2332 -mfunction-return=thunk and -mfunction-return=thunk-extern.
2334 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
2336 * config/i386/i386.c (print_reg): Print the name of the full
2337 integer register without '%'.
2338 (ix86_print_operand): Handle 'V'.
2339 * doc/extend.texi: Document 'V' modifier.
2341 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
2343 * config/i386/constraints.md (Bs): Disallow memory operand for
2344 -mindirect-branch-register.
2346 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2347 (GOT_memory_operand): Likewise.
2348 (call_insn_operand): Likewise.
2349 (sibcall_insn_operand): Likewise.
2350 (GOT32_symbol_operand): Likewise.
2351 * config/i386/i386.md (indirect_jump): Call convert_memory_address
2352 for -mindirect-branch-register.
2353 (tablejump): Likewise.
2354 (*sibcall_memory): Likewise.
2355 (*sibcall_value_memory): Likewise.
2356 Disallow peepholes of indirect call and jump via memory for
2357 -mindirect-branch-register.
2358 (*call_pop): Replace m with Bw.
2359 (*call_value_pop): Likewise.
2360 (*sibcall_pop_memory): Replace m with Bs.
2361 * config/i386/i386.opt (mindirect-branch-register): New option.
2362 * doc/invoke.texi: Document -mindirect-branch-register option.
2364 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
2366 * config/i386/i386-protos.h (ix86_output_function_return): New.
2367 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
2368 set function_return_type.
2369 (indirect_thunk_name): Add ret_p to indicate thunk for function
2371 (output_indirect_thunk_function): Pass false to
2372 indirect_thunk_name.
2373 (ix86_output_indirect_branch_via_reg): Likewise.
2374 (ix86_output_indirect_branch_via_push): Likewise.
2375 (output_indirect_thunk_function): Create alias for function
2376 return thunk if regno < 0.
2377 (ix86_output_function_return): New function.
2378 (ix86_handle_fndecl_attribute): Handle function_return.
2379 (ix86_attribute_table): Add function_return.
2380 * config/i386/i386.h (machine_function): Add
2381 function_return_type.
2382 * config/i386/i386.md (simple_return_internal): Use
2383 ix86_output_function_return.
2384 (simple_return_internal_long): Likewise.
2385 * config/i386/i386.opt (mfunction-return=): New option.
2386 (indirect_branch): Mention -mfunction-return=.
2387 * doc/extend.texi: Document function_return function attribute.
2388 * doc/invoke.texi: Document -mfunction-return= option.
2390 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
2392 * config/i386/i386-opts.h (indirect_branch): New.
2393 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
2394 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
2395 with local indirect jump when converting indirect call and jump.
2396 (ix86_set_indirect_branch_type): New.
2397 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
2398 (indirectlabelno): New.
2399 (indirect_thunk_needed): Likewise.
2400 (indirect_thunk_bnd_needed): Likewise.
2401 (indirect_thunks_used): Likewise.
2402 (indirect_thunks_bnd_used): Likewise.
2403 (INDIRECT_LABEL): Likewise.
2404 (indirect_thunk_name): Likewise.
2405 (output_indirect_thunk): Likewise.
2406 (output_indirect_thunk_function): Likewise.
2407 (ix86_output_indirect_branch_via_reg): Likewise.
2408 (ix86_output_indirect_branch_via_push): Likewise.
2409 (ix86_output_indirect_branch): Likewise.
2410 (ix86_output_indirect_jmp): Likewise.
2411 (ix86_code_end): Call output_indirect_thunk_function if needed.
2412 (ix86_output_call_insn): Call ix86_output_indirect_branch if
2414 (ix86_handle_fndecl_attribute): Handle indirect_branch.
2415 (ix86_attribute_table): Add indirect_branch.
2416 * config/i386/i386.h (machine_function): Add indirect_branch_type
2417 and has_local_indirect_jump.
2418 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
2420 (tablejump): Likewise.
2421 (*indirect_jump): Use ix86_output_indirect_jmp.
2422 (*tablejump_1): Likewise.
2423 (simple_return_indirect_internal): Likewise.
2424 * config/i386/i386.opt (mindirect-branch=): New option.
2425 (indirect_branch): New.
2428 (thunk-inline): Likewise.
2429 (thunk-extern): Likewise.
2430 * doc/extend.texi: Document indirect_branch function attribute.
2431 * doc/invoke.texi: Document -mindirect-branch= option.
2433 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
2436 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
2438 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
2440 * ipa-inline.c (want_inline_small_function_p): Return false if
2441 inlining has already failed with CIF_FINAL_ERROR.
2442 (update_caller_keys): Call want_inline_small_function_p before
2444 (update_callee_keys): Likewise.
2446 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2448 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
2450 (rs6000_quadword_masked_address_p): Likewise.
2451 (quad_aligned_load_p): Likewise.
2452 (quad_aligned_store_p): Likewise.
2453 (const_load_sequence_p): Add comment to describe the outer-most loop.
2454 (mimic_memory_attributes_and_flags): New function.
2455 (rs6000_gen_stvx): Likewise.
2456 (replace_swapped_aligned_store): Likewise.
2457 (rs6000_gen_lvx): Likewise.
2458 (replace_swapped_aligned_load): Likewise.
2459 (replace_swapped_load_constant): Capitalize argument name in
2460 comment describing this function.
2461 (rs6000_analyze_swaps): Add a third pass to search for vector loads
2462 and stores that access quad-word aligned addresses and replace
2463 with stvx or lvx instructions when appropriate.
2464 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
2465 New function prototype.
2466 (rs6000_quadword_masked_address_p): Likewise.
2467 (rs6000_gen_lvx): Likewise.
2468 (rs6000_gen_stvx): Likewise.
2469 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
2470 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
2471 when memory address is aligned.
2472 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
2473 this split to select lvx instruction when memory address is aligned.
2474 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
2475 instruction when memory address is aligned.
2476 (*vsx_le_perm_load_v16qi): Likewise.
2477 (four unnamed splitters): Modify to select the stvx instruction
2478 when memory is aligned.
2480 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
2482 * predict.c (determine_unlikely_bbs): Handle correctly BBs
2483 which appears in the queue multiple times.
2485 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2486 Alan Hayward <alan.hayward@arm.com>
2487 David Sherwood <david.sherwood@arm.com>
2489 * tree-vectorizer.h (vec_lower_bound): New structure.
2490 (_loop_vec_info): Add check_nonzero and lower_bounds.
2491 (LOOP_VINFO_CHECK_NONZERO): New macro.
2492 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
2493 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
2494 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
2495 fields. Make seg_len the distance travelled, not including the
2497 (dr_direction_indicator): Declare.
2498 (dr_zero_step_indicator): Likewise.
2499 (dr_known_forward_stride_p): Likewise.
2500 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
2502 (runtime_alias_check_p): Allow runtime alias checks with
2504 (operator ==): Compare access_size and align.
2505 (prune_runtime_alias_test_list): Rework for new distinction between
2506 the access_size and seg_len.
2507 (create_intersect_range_checks_index): Likewise. Cope with polynomial
2509 (get_segment_min_max): New function.
2510 (create_intersect_range_checks): Use it.
2511 (dr_step_indicator): New function.
2512 (dr_direction_indicator): Likewise.
2513 (dr_zero_step_indicator): Likewise.
2514 (dr_known_forward_stride_p): Likewise.
2515 * tree-loop-distribution.c (data_ref_segment_size): Return
2516 DR_STEP * (niters - 1).
2517 (compute_alias_check_pairs): Update call to the dr_with_seg_len
2519 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
2520 (vect_preserves_scalar_order_p): New function, split out from...
2521 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
2522 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
2523 (vect_vfa_access_size): New function.
2524 (vect_vfa_align): Likewise.
2525 (vect_compile_time_alias): Take access_size_a and access_b arguments.
2526 (dump_lower_bound): New function.
2527 (vect_check_lower_bound): Likewise.
2528 (vect_small_gap_p): Likewise.
2529 (vectorizable_with_step_bound_p): Likewise.
2530 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
2531 depencies if the vectorization factor is 1. Convert the checks
2532 for nonzero steps into checks on the bounds of DR_STEP. Try using
2533 a bunds check for variable steps if the minimum required step is
2534 relatively small. Update calls to the dr_with_seg_len
2535 constructor and to vect_compile_time_alias.
2536 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
2538 (vect_loop_versioning): Call it.
2539 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
2541 (vect_estimate_min_profitable_iters): Account for any bounds checks.
2543 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2544 Alan Hayward <alan.hayward@arm.com>
2545 David Sherwood <david.sherwood@arm.com>
2547 * doc/sourcebuild.texi (vect_scatter_store): Document.
2548 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
2550 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
2552 * genopinit.c (main): Add supports_vec_scatter_store and
2553 supports_vec_scatter_store_cached to target_optabs.
2554 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
2555 IFN_MASK_SCATTER_STORE.
2556 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
2558 * internal-fn.h (internal_store_fn_p): Declare.
2559 (internal_fn_stored_value_index): Likewise.
2560 * internal-fn.c (scatter_store_direct): New macro.
2561 (expand_scatter_store_optab_fn): New function.
2562 (direct_scatter_store_optab_supported_p): New macro.
2563 (internal_store_fn_p): New function.
2564 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
2565 IFN_MASK_SCATTER_STORE.
2566 (internal_fn_mask_index): Likewise.
2567 (internal_fn_stored_value_index): New function.
2568 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
2570 * optabs-query.h (supports_vec_scatter_store_p): Declare.
2571 * optabs-query.c (supports_vec_scatter_store_p): New function.
2572 * tree-vectorizer.h (vect_get_store_rhs): Declare.
2573 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
2574 true for scatter stores.
2575 (vect_gather_scatter_fn_p): Handle scatter stores too.
2576 (vect_check_gather_scatter): Consider using scatter stores if
2577 supports_vec_scatter_store_p.
2578 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
2580 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
2581 internal_fn_stored_value_index.
2582 (check_load_store_masking): Handle scatter stores too.
2583 (vect_get_store_rhs): Make public.
2584 (vectorizable_call): Use internal_store_fn_p.
2585 (vectorizable_store): Handle scatter store internal functions.
2586 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
2587 when deciding whether the end of the group has been reached.
2588 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
2589 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
2590 (mask_scatter_store<mode>): New insns.
2592 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2593 Alan Hayward <alan.hayward@arm.com>
2594 David Sherwood <david.sherwood@arm.com>
2596 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
2597 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
2598 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
2600 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
2601 Use vect_truncate_gather_scatter_offset if we can't treat the
2602 operation as a normal gather load or scatter store.
2603 (get_group_load_store_type): Take the gather_scatter_info
2604 as argument. Try using a gather load or scatter store for
2605 single-element groups.
2606 (get_load_store_type): Update calls to get_group_load_store_type
2607 and vect_use_strided_gather_scatters_p.
2609 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2610 Alan Hayward <alan.hayward@arm.com>
2611 David Sherwood <david.sherwood@arm.com>
2613 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
2614 optional tree argument.
2615 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
2617 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
2618 but continue to use the current value as a fallback.
2619 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
2620 to compare the updates.
2621 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
2622 (get_load_store_type): Use it when handling a strided access.
2623 (vect_get_strided_load_store_ops): New function.
2624 (vect_get_data_ptr_increment): Likewise.
2625 (vectorizable_load): Handle strided gather loads. Always pass
2626 a step to vect_create_data_ref_ptr and bump_vector_ptr.
2628 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2629 Alan Hayward <alan.hayward@arm.com>
2630 David Sherwood <david.sherwood@arm.com>
2632 * doc/md.texi (gather_load@var{m}): Document.
2633 (mask_gather_load@var{m}): Likewise.
2634 * genopinit.c (main): Add supports_vec_gather_load and
2635 supports_vec_gather_load_cached to target_optabs.
2636 * optabs-tree.c (init_tree_optimization_optabs): Use
2637 ggc_cleared_alloc to allocate target_optabs.
2638 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
2639 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
2641 * internal-fn.h (internal_load_fn_p): Declare.
2642 (internal_gather_scatter_fn_p): Likewise.
2643 (internal_fn_mask_index): Likewise.
2644 (internal_gather_scatter_fn_supported_p): Likewise.
2645 * internal-fn.c (gather_load_direct): New macro.
2646 (expand_gather_load_optab_fn): New function.
2647 (direct_gather_load_optab_supported_p): New macro.
2648 (direct_internal_fn_optab): New function.
2649 (internal_load_fn_p): Likewise.
2650 (internal_gather_scatter_fn_p): Likewise.
2651 (internal_fn_mask_index): Likewise.
2652 (internal_gather_scatter_fn_supported_p): Likewise.
2653 * optabs-query.c (supports_at_least_one_mode_p): New function.
2654 (supports_vec_gather_load_p): Likewise.
2655 * optabs-query.h (supports_vec_gather_load_p): Declare.
2656 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
2657 and memory_type field.
2658 (NUM_PATTERNS): Bump to 15.
2659 * tree-vect-data-refs.c: Include internal-fn.h.
2660 (vect_gather_scatter_fn_p): New function.
2661 (vect_describe_gather_scatter_call): Likewise.
2662 (vect_check_gather_scatter): Try using internal functions for
2663 gather loads. Recognize existing calls to a gather load function.
2664 (vect_analyze_data_refs): Consider using gather loads if
2665 supports_vec_gather_load_p.
2666 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
2667 (vect_get_gather_scatter_offset_type): Likewise.
2668 (vect_convert_mask_for_vectype): Likewise.
2669 (vect_add_conversion_to_patterm): Likewise.
2670 (vect_try_gather_scatter_pattern): Likewise.
2671 (vect_recog_gather_scatter_pattern): New pattern recognizer.
2672 (vect_vect_recog_func_ptrs): Add it.
2673 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
2674 internal_fn_mask_index and internal_gather_scatter_fn_p.
2675 (check_load_store_masking): Take the gather_scatter_info as an
2676 argument and handle gather loads.
2677 (vect_get_gather_scatter_ops): New function.
2678 (vectorizable_call): Check internal_load_fn_p.
2679 (vectorizable_load): Likewise. Handle gather load internal
2681 (vectorizable_store): Update call to check_load_store_masking.
2682 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
2683 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
2684 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
2685 (aarch64_gather_scale_operand_d): New predicates.
2686 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
2687 (mask_gather_load<mode>): New insns.
2689 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2690 Alan Hayward <alan.hayward@arm.com>
2691 David Sherwood <david.sherwood@arm.com>
2693 * optabs.def (fold_left_plus_optab): New optab.
2694 * doc/md.texi (fold_left_plus_@var{m}): Document.
2695 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
2696 * internal-fn.c (fold_left_direct): Define.
2697 (expand_fold_left_optab_fn): Likewise.
2698 (direct_fold_left_optab_supported_p): Likewise.
2699 * fold-const-call.c (fold_const_fold_left): New function.
2700 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
2701 * tree-parloops.c (valid_reduction_p): New function.
2702 (gather_scalar_reductions): Use it.
2703 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
2704 (vect_finish_replace_stmt): Declare.
2705 * tree-vect-loop.c (fold_left_reduction_fn): New function.
2706 (needs_fold_left_reduction_p): New function, split out from...
2707 (vect_is_simple_reduction): ...here. Accept reductions that
2708 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
2709 (vect_force_simple_reduction): Also store the reduction type in
2710 the assignment's STMT_VINFO_REDUC_TYPE.
2711 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
2712 (merge_with_identity): New function.
2713 (vect_expand_fold_left): Likewise.
2714 (vectorize_fold_left_reduction): Likewise.
2715 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
2716 scalar phi in place for it. Check for target support and reject
2717 cases that would reassociate the operation. Defer the transform
2718 phase to vectorize_fold_left_reduction.
2719 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
2720 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
2721 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
2723 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2725 * tree-if-conv.c (predicate_mem_writes): Remove redundant
2726 call to ifc_temp_var.
2728 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2729 Alan Hayward <alan.hayward@arm.com>
2730 David Sherwood <david.sherwood@arm.com>
2732 * target.def (legitimize_address_displacement): Take the original
2733 offset as a poly_int.
2734 * targhooks.h (default_legitimize_address_displacement): Update
2736 * targhooks.c (default_legitimize_address_displacement): Likewise.
2737 * doc/tm.texi: Regenerate.
2738 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
2739 as an argument, moving assert of ad->disp == ad->disp_term to...
2740 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
2741 Try calling targetm.legitimize_address_displacement before expanding
2742 the address rather than afterwards, and adjust for the new interface.
2743 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
2744 Match the new hook interface. Handle SVE addresses.
2745 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
2748 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2750 * Makefile.in (OBJS): Add early-remat.o.
2751 * target.def (select_early_remat_modes): New hook.
2752 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
2753 * doc/tm.texi: Regenerate.
2754 * targhooks.h (default_select_early_remat_modes): Declare.
2755 * targhooks.c (default_select_early_remat_modes): New function.
2756 * timevar.def (TV_EARLY_REMAT): New timevar.
2757 * passes.def (pass_early_remat): New pass.
2758 * tree-pass.h (make_pass_early_remat): Declare.
2759 * early-remat.c: New file.
2760 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
2762 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
2764 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2765 Alan Hayward <alan.hayward@arm.com>
2766 David Sherwood <david.sherwood@arm.com>
2768 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
2769 vfm1 with a bound_epilog parameter.
2770 (vect_do_peeling): Update calls accordingly, and move the prologue
2771 call earlier in the function. Treat the base bound_epilog as 0 for
2772 fully-masked loops and retain vf - 1 for other loops. Add 1 to
2773 this base when peeling for gaps.
2774 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
2775 with fully-masked loops.
2776 (vect_estimate_min_profitable_iters): Handle the single peeled
2777 iteration in that case.
2779 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2780 Alan Hayward <alan.hayward@arm.com>
2781 David Sherwood <david.sherwood@arm.com>
2783 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
2784 single-element interleaving even if the size is not a power of 2.
2785 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
2786 accesses for single-element interleaving if the group size is
2789 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2790 Alan Hayward <alan.hayward@arm.com>
2791 David Sherwood <david.sherwood@arm.com>
2793 * doc/md.texi (fold_extract_last_@var{m}): Document.
2794 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
2795 * optabs.def (fold_extract_last_optab): New optab.
2796 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
2797 * internal-fn.c (fold_extract_direct): New macro.
2798 (expand_fold_extract_optab_fn): Likewise.
2799 (direct_fold_extract_optab_supported_p): Likewise.
2800 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
2801 * tree-vect-loop.c (vect_model_reduction_cost): Handle
2802 EXTRACT_LAST_REDUCTION.
2803 (get_initial_def_for_reduction): Do not create an initial vector
2804 for EXTRACT_LAST_REDUCTION reductions.
2805 (vectorizable_reduction): Leave the scalar phi in place for
2806 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
2807 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
2808 epilogue code for EXTRACT_LAST_REDUCTION and defer the
2809 transform phase to vectorizable_condition.
2810 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
2812 (vect_finish_stmt_generation): ...here.
2813 (vect_finish_replace_stmt): New function.
2814 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
2815 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
2817 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
2819 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2820 Alan Hayward <alan.hayward@arm.com>
2821 David Sherwood <david.sherwood@arm.com>
2823 * doc/md.texi (extract_last_@var{m}): Document.
2824 * optabs.def (extract_last_optab): New optab.
2825 * internal-fn.def (EXTRACT_LAST): New internal function.
2826 * internal-fn.c (cond_unary_direct): New macro.
2827 (expand_cond_unary_optab_fn): Likewise.
2828 (direct_cond_unary_optab_supported_p): Likewise.
2829 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
2830 loops using EXTRACT_LAST.
2831 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
2832 (extract_last_<mode>): ...this optab.
2833 (vec_extract<mode><Vel>): Update accordingly.
2835 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2836 Alan Hayward <alan.hayward@arm.com>
2837 David Sherwood <david.sherwood@arm.com>
2839 * target.def (empty_mask_is_expensive): New hook.
2840 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
2841 * doc/tm.texi: Regenerate.
2842 * targhooks.h (default_empty_mask_is_expensive): Declare.
2843 * targhooks.c (default_empty_mask_is_expensive): New function.
2844 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
2845 if the target says that empty masks are expensive.
2846 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
2848 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
2850 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2851 Alan Hayward <alan.hayward@arm.com>
2852 David Sherwood <david.sherwood@arm.com>
2854 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
2855 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
2856 (vect_use_loop_mask_for_alignment_p): New function.
2857 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
2858 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
2859 niters_skip argument. Make sure that the first niters_skip elements
2860 of the first iteration are inactive.
2861 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
2862 Update call to vect_set_loop_masks_directly.
2863 (get_misalign_in_elems): New function, split out from...
2864 (vect_gen_prolog_loop_niters): ...here.
2865 (vect_update_init_of_dr): Take a code argument that specifies whether
2866 the adjustment should be added or subtracted.
2867 (vect_update_init_of_drs): Likewise.
2868 (vect_prepare_for_masked_peels): New function.
2869 (vect_do_peeling): Skip prologue peeling if we're using a mask
2870 instead. Update call to vect_update_inits_of_drs.
2871 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2873 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
2874 alignment. Do not include the number of peeled iterations in
2875 the minimum threshold in that case.
2876 (vectorizable_induction): Adjust the start value down by
2877 LOOP_VINFO_MASK_SKIP_NITERS iterations.
2878 (vect_transform_loop): Call vect_prepare_for_masked_peels.
2879 Take the number of skipped iterations into account when calculating
2881 * tree-vect-stmts.c (vect_gen_while_not): New function.
2883 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2884 Alan Hayward <alan.hayward@arm.com>
2885 David Sherwood <david.sherwood@arm.com>
2887 * doc/sourcebuild.texi (vect_fully_masked): Document.
2888 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
2890 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
2892 (vect_analyze_loop_2): ...here. Don't check the vectorization
2893 factor against the number of loop iterations if the loop is
2896 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2897 Alan Hayward <alan.hayward@arm.com>
2898 David Sherwood <david.sherwood@arm.com>
2900 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
2901 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
2902 (dump_groups): Update accordingly.
2903 (iv_use::mem_type): New member variable.
2904 (address_p): New function.
2905 (record_use): Add a mem_type argument and initialize the new
2907 (record_group_use): Add a mem_type argument. Use address_p.
2908 Remove obsolete null checks of base_object. Update call to record_use.
2909 (find_interesting_uses_op): Update call to record_group_use.
2910 (find_interesting_uses_cond): Likewise.
2911 (find_interesting_uses_address): Likewise.
2912 (get_mem_type_for_internal_fn): New function.
2913 (find_address_like_use): Likewise.
2914 (find_interesting_uses_stmt): Try find_address_like_use before
2915 calling find_interesting_uses_op.
2916 (addr_offset_valid_p): Use the iv mem_type field as the type
2917 of the addressed memory.
2918 (add_autoinc_candidates): Likewise.
2919 (get_address_cost): Likewise.
2920 (split_small_address_groups_p): Use address_p.
2921 (split_address_groups): Likewise.
2922 (add_iv_candidate_for_use): Likewise.
2923 (autoinc_possible_for_pair): Likewise.
2924 (rewrite_groups): Likewise.
2925 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
2926 (determine_group_iv_cost): Update after split of USE_ADDRESS.
2927 (get_alias_ptr_type_for_ptr_address): New function.
2928 (rewrite_use_address): Rewrite address uses in calls that were
2929 identified by find_address_like_use.
2931 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2932 Alan Hayward <alan.hayward@arm.com>
2933 David Sherwood <david.sherwood@arm.com>
2935 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
2937 * gimple-expr.h (is_gimple_addressable: Likewise.
2938 * gimple-expr.c (is_gimple_address): Likewise.
2939 * internal-fn.c (expand_call_mem_ref): New function.
2940 (expand_mask_load_optab_fn): Use it.
2941 (expand_mask_store_optab_fn): Likewise.
2943 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2944 Alan Hayward <alan.hayward@arm.com>
2945 David Sherwood <david.sherwood@arm.com>
2947 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
2948 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
2949 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
2950 (cond_umax@var{mode}): Document.
2951 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
2952 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
2953 (cond_umin_optab, cond_umax_optab): New optabs.
2954 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
2955 (COND_IOR, COND_XOR): New internal functions.
2956 * internal-fn.h (get_conditional_internal_fn): Declare.
2957 * internal-fn.c (cond_binary_direct): New macro.
2958 (expand_cond_binary_optab_fn): Likewise.
2959 (direct_cond_binary_optab_supported_p): Likewise.
2960 (get_conditional_internal_fn): New function.
2961 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
2962 Cope with reduction statements that are vectorized as calls rather
2964 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
2965 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
2966 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
2967 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
2968 (UNSPEC_COND_EOR): New unspecs.
2969 (optab): Add mappings for them.
2970 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
2971 (sve_int_op, sve_fp_op): New int attributes.
2973 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2974 Alan Hayward <alan.hayward@arm.com>
2975 David Sherwood <david.sherwood@arm.com>
2977 * optabs.def (while_ult_optab): New optab.
2978 * doc/md.texi (while_ult@var{m}@var{n}): Document.
2979 * internal-fn.def (WHILE_ULT): New internal function.
2980 * internal-fn.h (direct_internal_fn_supported_p): New override
2981 that takes two types as argument.
2982 * internal-fn.c (while_direct): New macro.
2983 (expand_while_optab_fn): New function.
2984 (convert_optab_supported_p): Likewise.
2985 (direct_while_optab_supported_p): New macro.
2986 * wide-int.h (wi::udiv_ceil): New function.
2987 * tree-vectorizer.h (rgroup_masks): New structure.
2988 (vec_loop_masks): New typedef.
2989 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
2991 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
2992 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
2993 (vect_max_vf): New function.
2994 (slpeel_make_loop_iterate_ntimes): Delete.
2995 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
2996 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
2997 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
2998 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
2999 internal-fn.h, stor-layout.h and optabs-query.h.
3000 (vect_set_loop_mask): New function.
3001 (add_preheader_seq): Likewise.
3002 (add_header_seq): Likewise.
3003 (interleave_supported_p): Likewise.
3004 (vect_maybe_permute_loop_masks): Likewise.
3005 (vect_set_loop_masks_directly): Likewise.
3006 (vect_set_loop_condition_masked): Likewise.
3007 (vect_set_loop_condition_unmasked): New function, split out from
3008 slpeel_make_loop_iterate_ntimes.
3009 (slpeel_make_loop_iterate_ntimes): Rename to..
3010 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
3011 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
3012 (vect_do_peeling): Update call accordingly.
3013 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
3015 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3016 mask_compare_type, can_fully_mask_p and fully_masked_p.
3017 (release_vec_loop_masks): New function.
3018 (_loop_vec_info): Use it to free the loop masks.
3019 (can_produce_all_loop_masks_p): New function.
3020 (vect_get_max_nscalars_per_iter): Likewise.
3021 (vect_verify_full_masking): Likewise.
3022 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
3023 retries, and free the mask rgroups before retrying. Check loop-wide
3024 reasons for disallowing fully-masked loops. Make the final decision
3025 about whether use a fully-masked loop or not.
3026 (vect_estimate_min_profitable_iters): Do not assume that peeling
3027 for the number of iterations will be needed for fully-masked loops.
3028 (vectorizable_reduction): Disable fully-masked loops.
3029 (vectorizable_live_operation): Likewise.
3030 (vect_halve_mask_nunits): New function.
3031 (vect_double_mask_nunits): Likewise.
3032 (vect_record_loop_mask): Likewise.
3033 (vect_get_loop_mask): Likewise.
3034 (vect_transform_loop): Handle the case in which the final loop
3035 iteration might handle a partial vector. Call vect_set_loop_condition
3036 instead of slpeel_make_loop_iterate_ntimes.
3037 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
3038 (check_load_store_masking): New function.
3039 (prepare_load_store_mask): Likewise.
3040 (vectorizable_store): Handle fully-masked loops.
3041 (vectorizable_load): Likewise.
3042 (supportable_widening_operation): Use vect_halve_mask_nunits for
3044 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
3045 (vect_gen_while): New function.
3046 * config/aarch64/aarch64.md (umax<mode>3): New expander.
3047 (aarch64_uqdec<mode>): New insn.
3049 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3050 Alan Hayward <alan.hayward@arm.com>
3051 David Sherwood <david.sherwood@arm.com>
3053 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
3054 (reduc_xor_scal_optab): New optabs.
3055 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
3056 (reduc_xor_scal_@var{m}): Document.
3057 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
3058 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
3060 * fold-const-call.c (fold_const_call): Handle them.
3061 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
3062 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
3063 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
3064 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
3065 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
3066 (UNSPEC_XORV): New unspecs.
3067 (optab): Add entries for them.
3068 (BITWISEV): New int iterator.
3069 (bit_reduc_op): New int attributes.
3071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3072 Alan Hayward <alan.hayward@arm.com>
3073 David Sherwood <david.sherwood@arm.com>
3075 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
3076 * internal-fn.def (VEC_SHL_INSERT): New internal function.
3077 * optabs.def (vec_shl_insert_optab): New optab.
3078 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
3079 (duplicate_and_interleave): Likewise.
3080 * tree-vect-loop.c: Include internal-fn.h.
3081 (neutral_op_for_slp_reduction): New function, split out from
3082 get_initial_defs_for_reduction.
3083 (get_initial_def_for_reduction): Handle option 2 for variable-length
3084 vectors by loading the neutral value into a vector and then shifting
3085 the initial value into element 0.
3086 (get_initial_defs_for_reduction): Replace the code argument with
3087 the neutral value calculated by neutral_op_for_slp_reduction.
3088 Use gimple_build_vector for constant-length vectors.
3089 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
3090 but the first group_size elements have a neutral value.
3091 Use duplicate_and_interleave otherwise.
3092 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
3093 Update call to get_initial_defs_for_reduction. Handle SLP
3094 reductions for variable-length vectors by creating one vector
3095 result for each scalar result, with the elements associated
3096 with other scalar results stubbed out with the neutral value.
3097 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
3098 Require IFN_VEC_SHL_INSERT for double reductions on
3099 variable-length vectors, or SLP reductions that have
3100 a neutral value. Require can_duplicate_and_interleave_p
3101 support for variable-length unchained SLP reductions if there
3102 is no neutral value, such as for MIN/MAX reductions. Also require
3103 the number of vector elements to be a multiple of the number of
3104 SLP statements when doing variable-length unchained SLP reductions.
3105 Update call to vect_create_epilog_for_reduction.
3106 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
3107 and remove initial values.
3108 (duplicate_and_interleave): Make public.
3109 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
3110 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
3112 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3113 Alan Hayward <alan.hayward@arm.com>
3114 David Sherwood <david.sherwood@arm.com>
3116 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
3117 (can_duplicate_and_interleave_p): New function.
3118 (vect_get_and_check_slp_defs): Take the vector of statements
3119 rather than just the current one. Remove excess parentheses.
3120 Restriction rejectinon of vect_constant_def and vect_external_def
3121 for variable-length vectors to boolean types, or types for which
3122 can_duplicate_and_interleave_p is false.
3123 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
3124 (duplicate_and_interleave): New function.
3125 (vect_get_constant_vectors): Use gimple_build_vector for
3126 constant-length vectors and suitable variable-length constant
3127 vectors. Use duplicate_and_interleave for other variable-length
3128 vectors. Don't defer the update when inserting new statements.
3130 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3131 Alan Hayward <alan.hayward@arm.com>
3132 David Sherwood <david.sherwood@arm.com>
3134 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
3135 min_profitable_iters doesn't go negative.
3137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3138 Alan Hayward <alan.hayward@arm.com>
3139 David Sherwood <david.sherwood@arm.com>
3141 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
3142 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
3143 * optabs.def (vec_mask_load_lanes_optab): New optab.
3144 (vec_mask_store_lanes_optab): Likewise.
3145 * internal-fn.def (MASK_LOAD_LANES): New internal function.
3146 (MASK_STORE_LANES): Likewise.
3147 * internal-fn.c (mask_load_lanes_direct): New macro.
3148 (mask_store_lanes_direct): Likewise.
3149 (expand_mask_load_optab_fn): Handle masked operations.
3150 (expand_mask_load_lanes_optab_fn): New macro.
3151 (expand_mask_store_optab_fn): Handle masked operations.
3152 (expand_mask_store_lanes_optab_fn): New macro.
3153 (direct_mask_load_lanes_optab_supported_p): Likewise.
3154 (direct_mask_store_lanes_optab_supported_p): Likewise.
3155 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
3157 (vect_load_lanes_supported): Likewise.
3158 * tree-vect-data-refs.c (strip_conversion): New function.
3159 (can_group_stmts_p): Likewise.
3160 (vect_analyze_data_ref_accesses): Use it instead of checking
3161 for a pair of assignments.
3162 (vect_store_lanes_supported): Take a masked_p parameter.
3163 (vect_load_lanes_supported): Likewise.
3164 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
3165 vect_store_lanes_supported and vect_load_lanes_supported.
3166 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3167 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
3168 parameter. Don't allow gaps for masked accesses.
3169 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
3170 and vect_load_lanes_supported.
3171 (get_load_store_type): Take a masked_p parameter and update
3172 call to get_group_load_store_type.
3173 (vectorizable_store): Update call to get_load_store_type.
3174 Handle IFN_MASK_STORE_LANES.
3175 (vectorizable_load): Update call to get_load_store_type.
3176 Handle IFN_MASK_LOAD_LANES.
3178 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3179 Alan Hayward <alan.hayward@arm.com>
3180 David Sherwood <david.sherwood@arm.com>
3182 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
3184 * config/aarch64/aarch64-protos.h
3185 (aarch64_sve_struct_memory_operand_p): Declare.
3186 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
3187 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
3188 (VPRED, vpred): Handle SVE structure modes.
3189 * config/aarch64/constraints.md (Utx): New constraint.
3190 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
3191 (aarch64_sve_struct_nonimmediate_operand): New predicates.
3192 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
3193 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
3194 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
3195 structure modes. Split into pieces after RA.
3196 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
3197 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
3199 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
3200 SVE structure modes.
3201 (aarch64_classify_address): Likewise.
3202 (sizetochar): Move earlier in file.
3203 (aarch64_print_operand): Handle SVE register lists.
3204 (aarch64_array_mode): New function.
3205 (aarch64_sve_struct_memory_operand_p): Likewise.
3206 (TARGET_ARRAY_MODE): Redefine.
3208 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3209 Alan Hayward <alan.hayward@arm.com>
3210 David Sherwood <david.sherwood@arm.com>
3212 * target.def (array_mode): New target hook.
3213 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
3214 * doc/tm.texi: Regenerate.
3215 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
3216 * hooks.c (hook_optmode_mode_uhwi_none): New function.
3217 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
3219 * stor-layout.c (mode_for_array): Likewise. Support polynomial
3222 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3223 Alan Hayward <alan.hayward@arm.com>
3224 David Sherwood <david.sherwood@arm.com>
3226 * fold-const.c (fold_binary_loc): Check the argument types
3227 rather than the result type when testing for a vector operation.
3229 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3231 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
3232 * doc/tm.texi: Regenerate.
3234 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3235 Alan Hayward <alan.hayward@arm.com>
3236 David Sherwood <david.sherwood@arm.com>
3238 * doc/invoke.texi (-msve-vector-bits=): Document new option.
3239 (sve): Document new AArch64 extension.
3240 * doc/md.texi (w): Extend the description of the AArch64
3241 constraint to include SVE vectors.
3242 (Upl, Upa): Document new AArch64 predicate constraints.
3243 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
3245 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
3246 (msve-vector-bits=): New option.
3247 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
3248 SVE when these are disabled.
3249 (sve): New extension.
3250 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
3251 modes. Adjust their number of units based on aarch64_sve_vg.
3252 (MAX_BITSIZE_MODE_ANY_MODE): Define.
3253 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
3254 aarch64_addr_query_type.
3255 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
3256 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
3257 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
3258 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
3259 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
3260 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
3261 (aarch64_simd_imm_zero_p): Delete.
3262 (aarch64_check_zero_based_sve_index_immediate): Declare.
3263 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
3264 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
3265 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
3266 (aarch64_sve_float_mul_immediate_p): Likewise.
3267 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
3269 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
3270 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
3271 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
3272 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
3273 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
3274 (aarch64_regmode_natural_size): Likewise.
3275 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
3276 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
3278 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
3279 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
3280 for VG and the SVE predicate registers.
3281 (V_ALIASES): Add a "z"-prefixed alias.
3282 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
3283 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
3284 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
3285 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
3286 (REG_CLASS_NAMES): Add entries for them.
3287 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
3288 and the predicate registers.
3289 (aarch64_sve_vg): Declare.
3290 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
3291 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
3292 (REGMODE_NATURAL_SIZE): Define.
3293 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3295 * config/aarch64/aarch64.c: Include cfgrtl.h.
3296 (simd_immediate_info): Add a constructor for series vectors,
3297 and an associated step field.
3298 (aarch64_sve_vg): New variable.
3299 (aarch64_dbx_register_number): Handle VG and the predicate registers.
3300 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
3301 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
3302 (VEC_ANY_DATA, VEC_STRUCT): New constants.
3303 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
3304 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
3305 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
3306 (aarch64_get_mask_mode): New functions.
3307 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
3308 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
3309 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
3310 predicate modes and predicate registers. Explicitly restrict
3311 GPRs to modes of 16 bytes or smaller. Only allow FP registers
3312 to store a vector mode if it is recognized by
3313 aarch64_classify_vector_mode.
3314 (aarch64_regmode_natural_size): New function.
3315 (aarch64_hard_regno_caller_save_mode): Return the original mode
3317 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
3318 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
3319 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
3320 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
3322 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
3323 does not overlap dest if the function is frame-related. Handle
3325 (aarch64_split_add_offset): New function.
3326 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
3327 them aarch64_add_offset.
3328 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
3329 and update call to aarch64_sub_sp.
3330 (aarch64_add_cfa_expression): New function.
3331 (aarch64_expand_prologue): Pass extra temporary registers to the
3332 functions above. Handle the case in which we need to emit new
3333 DW_CFA_expressions for registers that were originally saved
3334 relative to the stack pointer, but now have to be expressed
3335 relative to the frame pointer.
3336 (aarch64_output_mi_thunk): Pass extra temporary registers to the
3338 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
3339 IP0 and IP1 values for SVE frames.
3340 (aarch64_expand_vec_series): New function.
3341 (aarch64_expand_sve_widened_duplicate): Likewise.
3342 (aarch64_expand_sve_const_vector): Likewise.
3343 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
3344 Handle SVE constants. Use emit_move_insn to move a force_const_mem
3345 into the register, rather than emitting a SET directly.
3346 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
3347 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
3348 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
3349 (offset_9bit_signed_scaled_p): New functions.
3350 (aarch64_replicate_bitmask_imm): New function.
3351 (aarch64_bitmask_imm): Use it.
3352 (aarch64_cannot_force_const_mem): Reject expressions involving
3353 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
3354 (aarch64_classify_index): Handle SVE indices, by requiring
3355 a plain register index with a scale that matches the element size.
3356 (aarch64_classify_address): Handle SVE addresses. Assert that
3357 the mode of the address is VOIDmode or an integer mode.
3358 Update call to aarch64_classify_symbol.
3359 (aarch64_classify_symbolic_expression): Update call to
3360 aarch64_classify_symbol.
3361 (aarch64_const_vec_all_in_range_p): New function.
3362 (aarch64_print_vector_float_operand): Likewise.
3363 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
3364 "vN" for FP registers with SVE modes. Handle (const ...) vectors
3365 and the FP immediates 1.0 and 0.5.
3366 (aarch64_print_address_internal): Handle SVE addresses.
3367 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
3368 (aarch64_regno_regclass): Handle predicate registers.
3369 (aarch64_secondary_reload): Handle big-endian reloads of SVE
3371 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
3372 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
3373 (aarch64_convert_sve_vector_bits): New function.
3374 (aarch64_override_options): Use it to handle -msve-vector-bits=.
3375 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
3377 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
3378 Handle SVE vector and predicate modes. Accept VL-based constants
3379 that need only one temporary register, and VL offsets that require
3380 no temporary registers.
3381 (aarch64_conditional_register_usage): Mark the predicate registers
3382 as fixed if SVE isn't available.
3383 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
3384 Return true for SVE vector and predicate modes.
3385 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
3386 rather than an unsigned int. Handle SVE modes.
3387 (aarch64_preferred_simd_mode): Update call accordingly. Handle
3389 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
3391 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
3392 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
3393 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
3394 (aarch64_sve_float_mul_immediate_p): New functions.
3395 (aarch64_sve_valid_immediate): New function.
3396 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
3397 Explicitly reject structure modes. Check for INDEX constants.
3398 Handle PTRUE and PFALSE constants.
3399 (aarch64_check_zero_based_sve_index_immediate): New function.
3400 (aarch64_simd_imm_zero_p): Delete.
3401 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
3402 vector modes. Accept constants in the range of CNT[BHWD].
3403 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
3404 ask for an Advanced SIMD mode.
3405 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
3406 (aarch64_simd_vector_alignment): Handle SVE predicates.
3407 (aarch64_vectorize_preferred_vector_alignment): New function.
3408 (aarch64_simd_vector_alignment_reachable): Use it instead of
3410 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
3411 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
3413 (MAX_VECT_LEN): Delete.
3414 (expand_vec_perm_d): Add a vec_flags field.
3415 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
3416 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
3417 (aarch64_evpc_ext): Don't apply a big-endian lane correction
3419 (aarch64_evpc_rev): Rename to...
3420 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
3421 (aarch64_evpc_rev_global): New function.
3422 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
3423 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
3425 (aarch64_evpc_sve_tbl): New function.
3426 (aarch64_expand_vec_perm_const_1): Update after rename of
3427 aarch64_evpc_rev. Handle SVE permutes too, trying
3428 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
3429 than aarch64_evpc_tbl.
3430 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
3431 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
3432 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
3433 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
3434 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
3435 (aarch64_expand_sve_vcond): New functions.
3436 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
3437 of aarch64_vector_mode_p.
3438 (aarch64_dwarf_poly_indeterminate_value): New function.
3439 (aarch64_compute_pressure_classes): Likewise.
3440 (aarch64_can_change_mode_class): Likewise.
3441 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
3442 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
3443 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
3444 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
3445 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
3446 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
3447 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
3448 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
3450 (Dn, Dl, Dr): Accept const as well as const_vector.
3451 (Dz): Likewise. Compare against CONST0_RTX.
3452 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
3453 of "vector" where appropriate.
3454 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
3455 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
3456 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
3457 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
3458 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
3459 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
3460 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
3461 (v_int_equiv): Extend to SVE modes.
3462 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
3464 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
3465 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
3466 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
3467 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
3468 (SVE_COND_FP_CMP): New int iterators.
3469 (perm_hilo): Handle the new unpack unspecs.
3470 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
3472 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
3473 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
3474 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
3475 (aarch64_equality_operator, aarch64_constant_vector_operand)
3476 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
3477 (aarch64_sve_nonimmediate_operand): Likewise.
3478 (aarch64_sve_general_operand): Likewise.
3479 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
3480 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
3481 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
3482 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
3483 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
3484 (aarch64_sve_float_arith_immediate): Likewise.
3485 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
3486 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
3487 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
3488 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
3489 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
3490 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
3491 (aarch64_sve_float_arith_operand): Likewise.
3492 (aarch64_sve_float_arith_with_sub_operand): Likewise.
3493 (aarch64_sve_float_mul_operand): Likewise.
3494 (aarch64_sve_vec_perm_operand): Likewise.
3495 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
3496 (aarch64_mov_operand): Accept const_poly_int and const_vector.
3497 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
3498 as well as const_vector.
3499 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
3500 in file. Use CONST0_RTX and CONSTM1_RTX.
3501 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
3502 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
3503 Use aarch64_simd_imm_zero.
3504 * config/aarch64/aarch64-sve.md: New file.
3505 * config/aarch64/aarch64.md: Include it.
3506 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
3507 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
3508 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
3509 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
3510 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
3511 (sve): New attribute.
3512 (enabled): Disable instructions with the sve attribute unless
3514 (movqi, movhi): Pass CONST_POLY_INT operaneds through
3515 aarch64_expand_mov_immediate.
3516 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
3517 CNT[BHSD] immediates.
3518 (movti): Split CONST_POLY_INT moves into two halves.
3519 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
3520 Split additions that need a temporary here if the destination
3521 is the stack pointer.
3522 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
3523 (*add<mode>3_poly_1): New instruction.
3524 (set_clobber_cc): New expander.
3526 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3528 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
3529 parameter and use it instead of GET_MODE_SIZE (innermode). Use
3530 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
3531 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
3532 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
3533 Change innermode from fixed_mode_size to machine_mode.
3534 (simplify_subreg): Update call accordingly. Handle a constant-sized
3535 subreg of a variable-length CONST_VECTOR.
3537 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3538 Alan Hayward <alan.hayward@arm.com>
3539 David Sherwood <david.sherwood@arm.com>
3541 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
3542 (add_offset_to_base): New function, split out from...
3543 (create_mem_ref): ...here. When handling a scale other than 1,
3544 check first whether the address is valid without the offset.
3545 Add it into the base if so, leaving the index and scale as-is.
3547 2018-01-12 Jakub Jelinek <jakub@redhat.com>
3550 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
3551 fold_for_warn before checking if arg2 is INTEGER_CST.
3553 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
3555 * config/rs6000/predicates.md (load_multiple_operation): Delete.
3556 (store_multiple_operation): Delete.
3557 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
3558 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
3559 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
3560 guarded by TARGET_STRING.
3561 (rs6000_output_load_multiple): Delete.
3562 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
3563 OPTION_MASK_STRING / TARGET_STRING handling.
3564 (print_operand) <'N', 'O'>: Add comment that these are unused now.
3565 (const rs6000_opt_masks) <"string">: Change mask to 0.
3566 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
3567 (MASK_STRING): Delete.
3568 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
3570 (load_multiple): Delete.
3577 (store_multiple): Delete.
3584 (movmemsi_8reg): Delete.
3585 (corresponding unnamed define_insn): Delete.
3586 (movmemsi_6reg): Delete.
3587 (corresponding unnamed define_insn): Delete.
3588 (movmemsi_4reg): Delete.
3589 (corresponding unnamed define_insn): Delete.
3590 (movmemsi_2reg): Delete.
3591 (corresponding unnamed define_insn): Delete.
3592 (movmemsi_1reg): Delete.
3593 (corresponding unnamed define_insn): Delete.
3594 * config/rs6000/rs6000.opt (mno-string): New.
3595 (mstring): Replace by deprecation warning stub.
3596 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
3598 2018-01-12 Jakub Jelinek <jakub@redhat.com>
3600 * regrename.c (regrename_do_replace): If replacing the same
3601 reg multiple times, try to reuse last created gen_raw_REG.
3604 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
3605 main to workaround a bug in GDB.
3607 2018-01-12 Tom de Vries <tom@codesourcery.com>
3610 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
3612 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
3614 PR rtl-optimization/80481
3615 * ira-color.c (get_cap_member): New function.
3616 (allocnos_conflict_by_live_ranges_p): Use it.
3617 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
3618 (setup_slot_coalesced_allocno_live_ranges): Ditto.
3620 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
3623 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
3624 (*saddl_se_1): Ditto.
3626 (*ssubl_se_1): Ditto.
3628 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
3630 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
3631 rather than wi::to_widest for DR_INITs.
3632 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
3633 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
3634 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
3636 (vect_analyze_group_access_1): Note that here.
3638 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
3640 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
3641 polynomial type sizes.
3643 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
3645 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
3646 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
3647 (gimple_add_tmp_var): Likewise.
3649 2018-01-12 Martin Liska <mliska@suse.cz>
3651 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
3652 (gimple_alloc_sizes): Likewise.
3653 (dump_gimple_statistics): Use PRIu64 in printf format.
3654 * gimple.h: Change uint64_t to int.
3656 2018-01-12 Martin Liska <mliska@suse.cz>
3658 * tree-core.h: Use uint64_t instead of int.
3659 * tree.c (tree_node_counts): Likewise.
3660 (tree_node_sizes): Likewise.
3661 (dump_tree_statistics): Use PRIu64 in printf format.
3663 2018-01-12 Martin Liska <mliska@suse.cz>
3665 * Makefile.in: As qsort_chk is implemented in vec.c, add
3666 vec.o to linkage of gencfn-macros.
3667 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
3668 passing the info to record_node_allocation_statistics.
3669 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
3671 * ggc-common.c (struct ggc_usage): Add operator== and use
3672 it in operator< and compare function.
3673 * mem-stats.h (struct mem_usage): Likewise.
3674 * vec.c (struct vec_usage): Remove operator< and compare
3675 function. Can be simply inherited.
3677 2018-01-12 Martin Jambor <mjambor@suse.cz>
3680 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
3681 * tree-ssa-math-opts.c: Include domwalk.h.
3682 (convert_mult_to_fma_1): New function.
3683 (fma_transformation_info): New type.
3684 (fma_deferring_state): Likewise.
3685 (cancel_fma_deferring): New function.
3686 (result_of_phi): Likewise.
3687 (last_fma_candidate_feeds_initial_phi): Likewise.
3688 (convert_mult_to_fma): Added deferring logic, split actual
3689 transformation to convert_mult_to_fma_1.
3690 (math_opts_dom_walker): New type.
3691 (math_opts_dom_walker::after_dom_children): New method, body moved
3692 here from pass_optimize_widening_mul::execute, added deferring logic
3694 (pass_optimize_widening_mul::execute): Moved most of code to
3695 math_opts_dom_walker::after_dom_children.
3696 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
3697 * config/i386/i386.c (ix86_option_override_internal): Added
3698 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
3700 2018-01-12 Richard Biener <rguenther@suse.de>
3703 * dwarf2out.c (gen_variable_die): Do not reset old_die for
3704 inline instance vars.
3706 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
3709 * config/rx/rx.c (rx_is_restricted_memory_address):
3712 2018-01-12 Richard Biener <rguenther@suse.de>
3714 PR tree-optimization/80846
3715 * target.def (split_reduction): New target hook.
3716 * targhooks.c (default_split_reduction): New function.
3717 * targhooks.h (default_split_reduction): Declare.
3718 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
3719 target requests first reduce vectors by combining low and high
3721 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
3722 (get_vectype_for_scalar_type_and_size): Export.
3723 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
3724 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
3725 * doc/tm.texi: Regenerate.
3726 * config/i386/i386.c (ix86_split_reduction): Implement
3727 TARGET_VECTORIZE_SPLIT_REDUCTION.
3729 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3732 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
3733 in PIC mode except for TARGET_VXWORKS_RTP.
3734 * config/sparc/sparc.c: Include cfgrtl.h.
3735 (TARGET_INIT_PIC_REG): Define.
3736 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
3737 (sparc_pic_register_p): New predicate.
3738 (sparc_legitimate_address_p): Use it.
3739 (sparc_legitimize_pic_address): Likewise.
3740 (sparc_delegitimize_address): Likewise.
3741 (sparc_mode_dependent_address_p): Likewise.
3742 (gen_load_pcrel_sym): Remove 4th parameter.
3743 (load_got_register): Adjust call to above. Remove obsolete stuff.
3744 (sparc_expand_prologue): Do not call load_got_register here.
3745 (sparc_flat_expand_prologue): Likewise.
3746 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
3747 (sparc_use_pseudo_pic_reg): New function.
3748 (sparc_init_pic_reg): Likewise.
3749 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3750 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
3752 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
3754 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
3755 Add item for branch_cost.
3757 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3759 PR rtl-optimization/83565
3760 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
3761 not extend the result to a larger mode for rotate operations.
3762 (num_sign_bit_copies1): Likewise.
3764 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3767 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
3769 Use values-Xc.o for -pedantic.
3770 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
3772 2018-01-12 Martin Liska <mliska@suse.cz>
3775 * ipa-devirt.c (final_warning_record::grow_type_warnings):
3777 (possible_polymorphic_call_targets): Use it.
3778 (ipa_devirt): Likewise.
3780 2018-01-12 Martin Liska <mliska@suse.cz>
3782 * profile-count.h (enum profile_quality): Use 0 as invalid
3783 enum value of profile_quality.
3785 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
3787 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
3788 -mext-string options.
3790 2018-01-12 Richard Biener <rguenther@suse.de>
3792 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
3793 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
3794 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
3796 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
3798 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
3800 * configure.ac (--with-long-double-format): Add support for the
3801 configuration option to change the default long double format on
3803 * config.gcc (powerpc*-linux*-*): Likewise.
3804 * configure: Regenerate.
3805 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
3806 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
3807 used without modification.
3809 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3811 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
3812 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
3813 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
3814 MISC_BUILTIN_SPEC_BARRIER.
3815 (rs6000_init_builtins): Likewise.
3816 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
3818 (speculation_barrier): New define_insn.
3819 * doc/extend.texi: Document __builtin_speculation_barrier.
3821 2018-01-11 Jakub Jelinek <jakub@redhat.com>
3824 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
3825 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
3826 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
3828 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
3829 integral modes instead of "ss" and "sd".
3830 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
3831 vectors with 32-bit and 64-bit elements.
3832 (vecdupssescalarmodesuffix): New mode attribute.
3833 (vec_dup<mode>): Use it.
3835 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
3838 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
3839 frame if argument is passed on stack.
3841 2018-01-11 Jakub Jelinek <jakub@redhat.com>
3844 * ree.c (combine_reaching_defs): Optimize also
3845 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
3846 reg2=any_extend(exp); reg1=reg2;, formatting fix.
3848 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
3851 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
3853 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
3856 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
3857 after they are computed.
3859 2018-01-11 Bin Cheng <bin.cheng@arm.com>
3861 PR tree-optimization/83695
3862 * gimple-loop-linterchange.cc
3863 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
3864 reset cached scev information after interchange.
3865 (pass_linterchange::execute): Remove call to scev_reset_htab.
3867 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3869 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
3870 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
3871 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
3872 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
3873 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
3874 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
3875 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
3876 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
3877 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
3878 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
3879 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
3880 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
3881 (V_lane_reg): Likewise.
3882 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
3884 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
3885 (vfmal_lane_low<mode>_intrinsic,
3886 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
3887 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
3888 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
3889 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
3890 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
3891 vfmsl_lane_high<mode>_intrinsic): New define_insns.
3893 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3895 * config/arm/arm-cpus.in (fp16fml): New feature.
3896 (ALL_SIMD): Add fp16fml.
3897 (armv8.2-a): Add fp16fml as an option.
3898 (armv8.3-a): Likewise.
3899 (armv8.4-a): Add fp16fml as part of fp16.
3900 * config/arm/arm.h (TARGET_FP16FML): Define.
3901 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
3903 * config/arm/arm-modes.def (V2HF): Define.
3904 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
3905 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
3906 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
3907 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
3908 vfmsl_low, vfmsl_high): New set of builtins.
3909 * config/arm/iterators.md (PLUSMINUS): New code iterator.
3910 (vfml_op): New code attribute.
3911 (VFMLHALVES): New int iterator.
3912 (VFML, VFMLSEL): New mode attributes.
3913 (V_reg): Define mapping for V2HF.
3914 (V_hi, V_lo): New mode attributes.
3915 (VF_constraint): Likewise.
3916 (vfml_half, vfml_half_selector): New int attributes.
3917 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
3919 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
3920 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
3922 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
3923 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
3924 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
3925 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
3927 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
3928 Document new effective target and option set.
3930 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3932 * config/arm/arm-cpus.in (armv8_4): New feature.
3933 (ARMv8_4a): New fgroup.
3934 (armv8.4-a): New arch.
3935 * config/arm/arm-tables.opt: Regenerate.
3936 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
3937 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
3938 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
3939 Add matching rules for -march=armv8.4-a and extensions.
3940 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
3942 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
3945 * config/rx/rx.md (BW): New mode attribute.
3946 (sync_lock_test_and_setsi): Add mode suffix to insn output.
3948 2018-01-11 Richard Biener <rguenther@suse.de>
3950 PR tree-optimization/83435
3951 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
3952 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
3953 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
3955 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
3956 Alan Hayward <alan.hayward@arm.com>
3957 David Sherwood <david.sherwood@arm.com>
3959 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
3961 (aarch64_classify_address): Initialize it. Track polynomial offsets.
3962 (aarch64_print_address_internal): Use it to check for a zero offset.
3964 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
3965 Alan Hayward <alan.hayward@arm.com>
3966 David Sherwood <david.sherwood@arm.com>
3968 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
3969 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
3970 Return a poly_int64 rather than a HOST_WIDE_INT.
3971 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
3972 rather than a HOST_WIDE_INT.
3973 * config/aarch64/aarch64.h (aarch64_frame): Protect with
3974 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
3975 hard_fp_offset, frame_size, initial_adjust, callee_offset and
3976 final_offset from HOST_WIDE_INT to poly_int64.
3977 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
3978 to_constant when getting the number of units in an Advanced SIMD
3980 (aarch64_builtin_vectorized_function): Check for a constant number
3982 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
3984 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
3985 attribute instead of GET_MODE_NUNITS.
3986 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
3987 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
3988 GET_MODE_SIZE for fixed-size registers.
3989 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
3990 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
3991 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
3992 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
3993 (aarch64_print_operand, aarch64_print_address_internal)
3994 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
3995 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
3996 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
3997 Handle polynomial GET_MODE_SIZE.
3998 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
3999 wider than SImode without modification.
4000 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
4001 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
4002 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
4003 passing and returning SVE modes.
4004 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
4005 rather than GEN_INT.
4006 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
4007 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
4008 (aarch64_allocate_and_probe_stack_space): Likewise.
4009 (aarch64_layout_frame): Cope with polynomial offsets.
4010 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
4011 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
4013 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
4014 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
4015 poly_int64 rather than a HOST_WIDE_INT.
4016 (aarch64_get_separate_components, aarch64_process_components)
4017 (aarch64_expand_prologue, aarch64_expand_epilogue)
4018 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
4019 (aarch64_anchor_offset): New function, split out from...
4020 (aarch64_legitimize_address): ...here.
4021 (aarch64_builtin_vectorization_cost): Handle polynomial
4022 TYPE_VECTOR_SUBPARTS.
4023 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
4025 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
4026 number of elements from the PARALLEL rather than the mode.
4027 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
4028 rather than GET_MODE_BITSIZE.
4029 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
4030 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
4031 (aarch64_expand_vec_perm_const_1): Handle polynomial
4032 d->perm.length () and d->perm elements.
4033 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
4034 Apply to_constant to d->perm elements.
4035 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
4036 polynomial CONST_VECTOR_NUNITS.
4037 (aarch64_move_pointer): Take amount as a poly_int64 rather
4039 (aarch64_progress_pointer): Avoid temporary variable.
4040 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
4041 the mode attribute instead of GET_MODE.
4043 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4044 Alan Hayward <alan.hayward@arm.com>
4045 David Sherwood <david.sherwood@arm.com>
4047 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
4048 x exists before using it.
4049 (aarch64_add_constant_internal): Rename to...
4050 (aarch64_add_offset_1): ...this. Replace regnum with separate
4051 src and dest rtxes. Handle the case in which they're different,
4052 including when the offset is zero. Replace scratchreg with an rtx.
4053 Use 2 additions if there is no spare register into which we can
4054 move a 16-bit constant.
4055 (aarch64_add_constant): Delete.
4056 (aarch64_add_offset): Replace reg with separate src and dest
4057 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
4058 Use aarch64_add_offset_1.
4059 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
4060 an rtx rather than an int. Take the delta as a poly_int64
4061 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
4062 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
4063 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
4064 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
4065 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
4067 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
4068 aarch64_add_constant.
4070 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4072 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
4073 Use scalar_float_mode.
4075 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4077 * config/aarch64/aarch64-simd.md
4078 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
4079 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
4080 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
4081 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
4082 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
4083 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
4084 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
4085 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
4086 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
4087 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
4089 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4092 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
4093 targ_options->x_arm_arch_string is non NULL.
4095 2018-01-11 Tamar Christina <tamar.christina@arm.com>
4097 * config/aarch64/aarch64.h
4098 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
4100 2018-01-11 Sudakshina Das <sudi.das@arm.com>
4103 * expmed.c (emit_store_flag_force): Swap if const op0
4104 and change VOIDmode to mode of op0.
4106 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4108 PR rtl-optimization/83761
4109 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
4110 than bytes to mode_for_size.
4112 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
4115 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
4116 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
4119 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
4122 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
4123 when in layout mode.
4124 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
4125 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
4128 2018-01-10 Michael Collison <michael.collison@arm.com>
4130 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
4131 * config/aarch64/aarch64-option-extension.def: Add
4132 AARCH64_OPT_EXTENSION of 'fp16fml'.
4133 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
4134 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
4135 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
4136 * config/aarch64/constraints.md (Ui7): New constraint.
4137 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
4138 (VFMLA_SEL_W): Ditto.
4141 (VFMLA16_LOW): New int iterator.
4142 (VFMLA16_HIGH): Ditto.
4143 (UNSPEC_FMLAL): New unspec.
4144 (UNSPEC_FMLSL): Ditto.
4145 (UNSPEC_FMLAL2): Ditto.
4146 (UNSPEC_FMLSL2): Ditto.
4147 (f16mac): New code attribute.
4148 * config/aarch64/aarch64-simd-builtins.def
4149 (aarch64_fmlal_lowv2sf): Ditto.
4150 (aarch64_fmlsl_lowv2sf): Ditto.
4151 (aarch64_fmlalq_lowv4sf): Ditto.
4152 (aarch64_fmlslq_lowv4sf): Ditto.
4153 (aarch64_fmlal_highv2sf): Ditto.
4154 (aarch64_fmlsl_highv2sf): Ditto.
4155 (aarch64_fmlalq_highv4sf): Ditto.
4156 (aarch64_fmlslq_highv4sf): Ditto.
4157 (aarch64_fmlal_lane_lowv2sf): Ditto.
4158 (aarch64_fmlsl_lane_lowv2sf): Ditto.
4159 (aarch64_fmlal_laneq_lowv2sf): Ditto.
4160 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
4161 (aarch64_fmlalq_lane_lowv4sf): Ditto.
4162 (aarch64_fmlsl_lane_lowv4sf): Ditto.
4163 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
4164 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
4165 (aarch64_fmlal_lane_highv2sf): Ditto.
4166 (aarch64_fmlsl_lane_highv2sf): Ditto.
4167 (aarch64_fmlal_laneq_highv2sf): Ditto.
4168 (aarch64_fmlsl_laneq_highv2sf): Ditto.
4169 (aarch64_fmlalq_lane_highv4sf): Ditto.
4170 (aarch64_fmlsl_lane_highv4sf): Ditto.
4171 (aarch64_fmlalq_laneq_highv4sf): Ditto.
4172 (aarch64_fmlsl_laneq_highv4sf): Ditto.
4173 * config/aarch64/aarch64-simd.md:
4174 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
4175 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
4176 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
4177 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
4178 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
4179 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
4180 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
4181 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
4182 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
4183 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
4184 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
4185 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
4186 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
4187 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
4188 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
4189 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
4190 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
4191 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
4192 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
4193 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
4194 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
4195 (vfmlsl_low_u32): Ditto.
4196 (vfmlalq_low_u32): Ditto.
4197 (vfmlslq_low_u32): Ditto.
4198 (vfmlal_high_u32): Ditto.
4199 (vfmlsl_high_u32): Ditto.
4200 (vfmlalq_high_u32): Ditto.
4201 (vfmlslq_high_u32): Ditto.
4202 (vfmlal_lane_low_u32): Ditto.
4203 (vfmlsl_lane_low_u32): Ditto.
4204 (vfmlal_laneq_low_u32): Ditto.
4205 (vfmlsl_laneq_low_u32): Ditto.
4206 (vfmlalq_lane_low_u32): Ditto.
4207 (vfmlslq_lane_low_u32): Ditto.
4208 (vfmlalq_laneq_low_u32): Ditto.
4209 (vfmlslq_laneq_low_u32): Ditto.
4210 (vfmlal_lane_high_u32): Ditto.
4211 (vfmlsl_lane_high_u32): Ditto.
4212 (vfmlal_laneq_high_u32): Ditto.
4213 (vfmlsl_laneq_high_u32): Ditto.
4214 (vfmlalq_lane_high_u32): Ditto.
4215 (vfmlslq_lane_high_u32): Ditto.
4216 (vfmlalq_laneq_high_u32): Ditto.
4217 (vfmlslq_laneq_high_u32): Ditto.
4218 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
4219 (AARCH64_FL_FOR_ARCH8_4): New.
4220 (AARCH64_ISA_F16FML): New ISA flag.
4221 (TARGET_F16FML): New feature flag for fp16fml.
4222 (doc/invoke.texi): Document new fp16fml option.
4224 2018-01-10 Michael Collison <michael.collison@arm.com>
4226 * config/aarch64/aarch64-builtins.c:
4227 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
4228 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
4229 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
4230 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
4231 (AARCH64_ISA_SHA3): New ISA flag.
4232 (TARGET_SHA3): New feature flag for sha3.
4233 * config/aarch64/iterators.md (sha512_op): New int attribute.
4234 (CRYPTO_SHA512): New int iterator.
4235 (UNSPEC_SHA512H): New unspec.
4236 (UNSPEC_SHA512H2): Ditto.
4237 (UNSPEC_SHA512SU0): Ditto.
4238 (UNSPEC_SHA512SU1): Ditto.
4239 * config/aarch64/aarch64-simd-builtins.def
4240 (aarch64_crypto_sha512hqv2di): New builtin.
4241 (aarch64_crypto_sha512h2qv2di): Ditto.
4242 (aarch64_crypto_sha512su0qv2di): Ditto.
4243 (aarch64_crypto_sha512su1qv2di): Ditto.
4244 (aarch64_eor3qv8hi): Ditto.
4245 (aarch64_rax1qv2di): Ditto.
4246 (aarch64_xarqv2di): Ditto.
4247 (aarch64_bcaxqv8hi): Ditto.
4248 * config/aarch64/aarch64-simd.md:
4249 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
4250 (aarch64_crypto_sha512su0qv2di): Ditto.
4251 (aarch64_crypto_sha512su1qv2di): Ditto.
4252 (aarch64_eor3qv8hi): Ditto.
4253 (aarch64_rax1qv2di): Ditto.
4254 (aarch64_xarqv2di): Ditto.
4255 (aarch64_bcaxqv8hi): Ditto.
4256 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
4257 (vsha512h2q_u64): Ditto.
4258 (vsha512su0q_u64): Ditto.
4259 (vsha512su1q_u64): Ditto.
4260 (veor3q_u16): Ditto.
4261 (vrax1q_u64): Ditto.
4263 (vbcaxq_u16): Ditto.
4264 * config/arm/types.md (crypto_sha512): New type attribute.
4265 (crypto_sha3): Ditto.
4266 (doc/invoke.texi): Document new sha3 option.
4268 2018-01-10 Michael Collison <michael.collison@arm.com>
4270 * config/aarch64/aarch64-builtins.c:
4271 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
4272 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
4273 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
4274 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
4275 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
4276 (AARCH64_ISA_SM4): New ISA flag.
4277 (TARGET_SM4): New feature flag for sm4.
4278 * config/aarch64/aarch64-simd-builtins.def
4279 (aarch64_sm3ss1qv4si): Ditto.
4280 (aarch64_sm3tt1aq4si): Ditto.
4281 (aarch64_sm3tt1bq4si): Ditto.
4282 (aarch64_sm3tt2aq4si): Ditto.
4283 (aarch64_sm3tt2bq4si): Ditto.
4284 (aarch64_sm3partw1qv4si): Ditto.
4285 (aarch64_sm3partw2qv4si): Ditto.
4286 (aarch64_sm4eqv4si): Ditto.
4287 (aarch64_sm4ekeyqv4si): Ditto.
4288 * config/aarch64/aarch64-simd.md:
4289 (aarch64_sm3ss1qv4si): Ditto.
4290 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
4291 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
4292 (aarch64_sm4eqv4si): Ditto.
4293 (aarch64_sm4ekeyqv4si): Ditto.
4294 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
4295 (sm3part_op): Ditto.
4296 (CRYPTO_SM3TT): Ditto.
4297 (CRYPTO_SM3PART): Ditto.
4298 (UNSPEC_SM3SS1): New unspec.
4299 (UNSPEC_SM3TT1A): Ditto.
4300 (UNSPEC_SM3TT1B): Ditto.
4301 (UNSPEC_SM3TT2A): Ditto.
4302 (UNSPEC_SM3TT2B): Ditto.
4303 (UNSPEC_SM3PARTW1): Ditto.
4304 (UNSPEC_SM3PARTW2): Ditto.
4305 (UNSPEC_SM4E): Ditto.
4306 (UNSPEC_SM4EKEY): Ditto.
4307 * config/aarch64/constraints.md (Ui2): New constraint.
4308 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
4309 * config/arm/types.md (crypto_sm3): New type attribute.
4310 (crypto_sm4): Ditto.
4311 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
4312 (vsm3tt1aq_u32): Ditto.
4313 (vsm3tt1bq_u32): Ditto.
4314 (vsm3tt2aq_u32): Ditto.
4315 (vsm3tt2bq_u32): Ditto.
4316 (vsm3partw1q_u32): Ditto.
4317 (vsm3partw2q_u32): Ditto.
4318 (vsm4eq_u32): Ditto.
4319 (vsm4ekeyq_u32): Ditto.
4320 (doc/invoke.texi): Document new sm4 option.
4322 2018-01-10 Michael Collison <michael.collison@arm.com>
4324 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
4325 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
4326 (AARCH64_FL_FOR_ARCH8_4): New.
4327 (AARCH64_FL_V8_4): New flag.
4328 (doc/invoke.texi): Document new armv8.4-a option.
4330 2018-01-10 Michael Collison <michael.collison@arm.com>
4332 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
4333 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
4334 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
4335 * config/aarch64/aarch64-option-extension.def: Add
4336 AARCH64_OPT_EXTENSION of 'sha2'.
4337 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
4338 (crypto): Disable sha2 and aes if crypto disabled.
4339 (crypto): Enable aes and sha2 if enabled.
4340 (simd): Disable sha2 and aes if simd disabled.
4341 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
4343 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
4344 (TARGET_SHA2): New feature flag for sha2.
4345 (TARGET_AES): New feature flag for aes.
4346 * config/aarch64/aarch64-simd.md:
4347 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
4348 conditional on TARGET_AES.
4349 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
4350 (aarch64_crypto_sha1hsi): Make pattern conditional
4352 (aarch64_crypto_sha1hv4si): Ditto.
4353 (aarch64_be_crypto_sha1hv4si): Ditto.
4354 (aarch64_crypto_sha1su1v4si): Ditto.
4355 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
4356 (aarch64_crypto_sha1su0v4si): Ditto.
4357 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
4358 (aarch64_crypto_sha256su0v4si): Ditto.
4359 (aarch64_crypto_sha256su1v4si): Ditto.
4360 (doc/invoke.texi): Document new aes and sha2 options.
4362 2018-01-10 Martin Sebor <msebor@redhat.com>
4364 PR tree-optimization/83781
4365 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
4368 2018-01-11 Martin Sebor <msebor@gmail.com>
4369 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4371 PR tree-optimization/83501
4372 PR tree-optimization/81703
4374 * tree-ssa-strlen.c (get_string_cst): Rename...
4375 (get_string_len): ...to this. Handle global constants.
4376 (handle_char_store): Adjust.
4378 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
4379 Jim Wilson <jimw@sifive.com>
4381 * config/riscv/riscv-protos.h (riscv_output_return): New.
4382 * config/riscv/riscv.c (struct machine_function): New naked_p field.
4383 (riscv_attribute_table, riscv_output_return),
4384 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
4385 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
4386 (riscv_compute_frame_info): Only compute frame->mask if not a naked
4388 (riscv_expand_prologue): Add early return for naked function.
4389 (riscv_expand_epilogue): Likewise.
4390 (riscv_function_ok_for_sibcall): Return false for naked function.
4391 (riscv_set_current_function): New.
4392 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
4393 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
4394 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
4395 * doc/extend.texi (RISC-V Function Attributes): New.
4397 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4399 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
4400 check for 128-bit long double before checking TCmode.
4401 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
4402 128-bit long doubles before checking TFmode or TCmode.
4403 (FLOAT128_IBM_P): Likewise.
4405 2018-01-10 Martin Sebor <msebor@redhat.com>
4407 PR tree-optimization/83671
4408 * builtins.c (c_strlen): Unconditionally return zero for the empty
4410 Use -Warray-bounds for warnings.
4411 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
4412 for non-constant array indices with COMPONENT_REF, arrays of
4413 arrays, and pointers to arrays.
4414 (gimple_fold_builtin_strlen): Determine and set length range for
4415 non-constant character arrays.
4417 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
4420 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
4423 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
4425 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
4427 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
4430 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
4431 VECTOR_MEM_ALTIVEC_OR_VSX_P.
4432 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
4433 indexed_or_indirect_operand predicate.
4434 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
4435 (*vsx_le_perm_load_v8hi): Likewise.
4436 (*vsx_le_perm_load_v16qi): Likewise.
4437 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
4438 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
4439 (*vsx_le_perm_store_v8hi): Likewise.
4440 (*vsx_le_perm_store_v16qi): Likewise.
4441 (eight unnamed splitters): Likewise.
4443 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
4445 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
4446 * config/rs6000/emmintrin.h: Likewise.
4447 * config/rs6000/mmintrin.h: Likewise.
4448 * config/rs6000/xmmintrin.h: Likewise.
4450 2018-01-10 David Malcolm <dmalcolm@redhat.com>
4453 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
4455 * tree.c (tree_nop_conversion): Return true for location wrapper
4457 (maybe_wrap_with_location): New function.
4458 (selftest::check_strip_nops): New function.
4459 (selftest::test_location_wrappers): New function.
4460 (selftest::tree_c_tests): Call it.
4461 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
4462 (maybe_wrap_with_location): New decl.
4463 (EXPR_LOCATION_WRAPPER_P): New macro.
4464 (location_wrapper_p): New inline function.
4465 (tree_strip_any_location_wrapper): New inline function.
4467 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
4470 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
4471 stack_realign_offset for the largest alignment of stack slot
4473 (ix86_find_max_used_stack_alignment): New function.
4474 (ix86_finalize_stack_frame_flags): Use it. Set
4475 max_used_stack_alignment if we don't realign stack.
4476 * config/i386/i386.h (machine_function): Add
4477 max_used_stack_alignment.
4479 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
4481 * config/arm/arm.opt (-mbranch-cost): New option.
4482 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
4485 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
4488 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
4489 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
4491 2018-01-10 Richard Biener <rguenther@suse.de>
4494 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
4495 early out so it also covers the case where we have a non-NULL
4498 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
4500 PR tree-optimization/83753
4501 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
4502 for non-strided grouped accesses if the number of elements is 1.
4504 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
4507 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
4508 * i386.h (TARGET_USE_GATHER): Define.
4509 * x86-tune.def (X86_TUNE_USE_GATHER): New.
4511 2018-01-10 Martin Liska <mliska@suse.cz>
4514 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
4515 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
4517 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
4518 CLEANUP_NO_PARTITIONING is not set.
4520 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
4522 * doc/rtl.texi: Remove documentation of (const ...) wrappers
4523 for vectors, as a partial revert of r254296.
4524 * rtl.h (const_vec_p): Delete.
4525 (const_vec_duplicate_p): Don't test for vector CONSTs.
4526 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
4527 * expmed.c (make_tree): Likewise.
4530 * common.md (E, F): Use CONSTANT_P instead of checking for
4532 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
4533 checking for CONST_VECTOR.
4535 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
4538 * predict.c (force_edge_cold): Handle in more sane way edges
4541 2018-01-09 Carl Love <cel@us.ibm.com>
4543 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
4545 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
4546 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
4547 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
4548 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
4549 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
4550 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
4551 * config/rs6000/rs6000-protos.h: Add extern defition for
4552 rs6000_generate_float2_double_code.
4553 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
4555 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
4556 (float2_v2df): Add define_expand.
4558 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
4561 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
4562 op_mode in the force_to_mode call.
4564 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
4566 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
4567 instead of checking each element individually.
4568 (aarch64_evpc_uzp): Likewise.
4569 (aarch64_evpc_zip): Likewise.
4570 (aarch64_evpc_ext): Likewise.
4571 (aarch64_evpc_rev): Likewise.
4572 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
4573 instead of checking each element individually. Return true without
4575 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
4576 whether all selected elements come from the same input, instead of
4577 checking each element individually. Remove calls to gen_rtx_REG,
4578 start_sequence and end_sequence and instead assert that no rtl is
4581 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
4583 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
4584 order of HIGH and CONST checks.
4586 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
4588 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
4589 if the destination isn't an SSA_NAME.
4591 2018-01-09 Richard Biener <rguenther@suse.de>
4593 PR tree-optimization/83668
4594 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
4596 (canonicalize_loop_form): ... here, renamed from ...
4597 (canonicalize_loop_closed_ssa_form): ... this and amended to
4598 swap successor edges for loop exit blocks to make us use
4599 the RPO order we need for initial schedule generation.
4601 2018-01-09 Joseph Myers <joseph@codesourcery.com>
4603 PR tree-optimization/64811
4604 * match.pd: When optimizing comparisons with Inf, avoid
4605 introducing or losing exceptions from comparisons with NaN.
4607 2018-01-09 Martin Liska <mliska@suse.cz>
4610 * asan.c (shadow_mem_size): Add gcc_assert.
4612 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
4614 Don't save registers in main().
4617 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
4618 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
4619 * config/avr/avr.c (avr_set_current_function): Don't error if
4620 naked, OS_task or OS_main are specified at the same time.
4621 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
4623 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
4625 * common/config/avr/avr-common.c (avr_option_optimization_table):
4626 Switch on -mmain-is-OS_task for optimizing compilations.
4628 2018-01-09 Richard Biener <rguenther@suse.de>
4630 PR tree-optimization/83572
4631 * graphite.c: Include cfganal.h.
4632 (graphite_transform_loops): Connect infinite loops to exit
4633 and remove fake edges at the end.
4635 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
4637 * ipa-inline.c (edge_badness): Revert accidental checkin.
4639 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
4642 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
4643 symbols; not inline clones.
4645 2018-01-09 Jakub Jelinek <jakub@redhat.com>
4648 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
4649 hard registers. Formatting fixes.
4651 PR preprocessor/83722
4652 * gcc.c (try_generate_repro): Pass
4653 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
4654 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
4657 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
4658 Kito Cheng <kito.cheng@gmail.com>
4660 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
4661 (riscv_leaf_function_p): Delete.
4662 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
4664 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4666 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
4668 (do_ifelse): New function.
4669 (do_isel): New function.
4670 (do_sub3): New function.
4671 (do_add3): New function.
4672 (do_load_mask_compare): New function.
4673 (do_overlap_load_compare): New function.
4674 (expand_compare_loop): New function.
4675 (expand_block_compare): Call expand_compare_loop() when appropriate.
4676 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
4678 (-mblock-compare-inline-loop-limit): New option.
4680 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4683 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
4684 Reverse order of second and third operands in first alternative.
4685 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
4686 of first and second elements in UNSPEC_VPERMR vector.
4687 (altivec_expand_vec_perm_le): Likewise.
4689 2018-01-08 Jeff Law <law@redhat.com>
4691 PR rtl-optimizatin/81308
4692 * tree-switch-conversion.c (cfg_altered): New file scoped static.
4693 (process_switch): If group_case_labels makes a change, then set
4695 (pass_convert_switch::execute): If a switch is converted, then
4696 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
4698 PR rtl-optimization/81308
4699 * recog.c (split_all_insns): Conditionally cleanup the CFG after
4702 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
4704 PR target/83663 - Revert r255946
4705 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
4706 generation for cases where splatting a value is not useful.
4707 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
4708 across a vec_duplicate and a paradoxical subreg forming a vector
4709 mode to a vec_concat.
4711 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4713 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
4714 -march=armv8.3-a variants.
4715 * config/arm/t-multilib: Likewise.
4716 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
4718 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4720 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
4722 (cceq_ior_compare_complement): Give it a name so I can use it, and
4723 change boolean_or_operator predicate to boolean_operator so it can
4724 be used to generate a crand.
4725 (eqne): New code iterator.
4726 (bd/bd_neg): New code_attrs.
4727 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
4728 a single define_insn.
4729 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
4730 decrement (bdnzt/bdnzf/bdzt/bdzf).
4731 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
4732 with the new names of the branch decrement patterns, and added the
4733 names of the branch decrement conditional patterns.
4735 2018-01-08 Richard Biener <rguenther@suse.de>
4737 PR tree-optimization/83563
4738 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
4741 2018-01-08 Richard Biener <rguenther@suse.de>
4744 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
4746 2018-01-08 Richard Biener <rguenther@suse.de>
4748 PR tree-optimization/83685
4749 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
4750 references to abnormals.
4752 2018-01-08 Richard Biener <rguenther@suse.de>
4755 * dwarf2out.c (output_indirect_strings): Handle empty
4756 skeleton_debug_str_hash.
4757 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
4759 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
4761 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
4762 (emit_store_direct): Likewise.
4763 (arc_trampoline_adjust_address): Likewise.
4764 (arc_asm_trampoline_template): New function.
4765 (arc_initialize_trampoline): Use asm_trampoline_template.
4766 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
4767 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
4768 * config/arc/arc.md (flush_icache): Delete pattern.
4770 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
4772 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
4773 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
4776 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
4779 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
4780 by not USED_FOR_TARGET.
4781 (make_pass_resolve_sw_modes): Likewise.
4783 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
4785 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
4788 2018-01-08 Richard Biener <rguenther@suse.de>
4791 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
4793 2018-01-08 Richard Biener <rguenther@suse.de>
4796 * match.pd ((t * 2) / 2) -> t): Add missing :c.
4798 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
4801 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
4802 basic blocks with a small number of successors.
4803 (convert_control_dep_chain_into_preds): Improve handling of
4805 (dump_predicates): Split apart into...
4806 (dump_pred_chain): ...here...
4807 (dump_pred_info): ...and here.
4808 (can_one_predicate_be_invalidated_p): Add debugging printfs.
4809 (can_chain_union_be_invalidated_p): Improve check for invalidation
4811 (uninit_uses_cannot_happen): Avoid unnecessary if
4812 convert_control_dep_chain_into_preds yielded nothing.
4814 2018-01-06 Martin Sebor <msebor@redhat.com>
4816 PR tree-optimization/83640
4817 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
4818 subtracting negative offset from size.
4819 (builtin_access::overlap): Adjust offset bounds of the access to fall
4820 within the size of the object if possible.
4822 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
4824 PR rtl-optimization/83699
4825 * expmed.c (extract_bit_field_1): Restrict the vector usage of
4826 extract_bit_field_as_subreg to cases in which the extracted
4827 value is also a vector.
4829 * lra-constraints.c (process_alt_operands): Test for the equivalence
4830 substitutions when detecting a possible reload cycle.
4832 2018-01-06 Jakub Jelinek <jakub@redhat.com>
4835 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
4836 by default if flag_selective_schedling{,2}. Formatting fixes.
4838 PR rtl-optimization/83682
4839 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
4840 if it has non-VECTOR_MODE element mode.
4841 (vec_duplicate_p): Likewise.
4844 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
4845 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
4847 2018-01-05 Jakub Jelinek <jakub@redhat.com>
4850 * config/i386/i386-builtin.def
4851 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
4852 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
4853 Require also OPTION_MASK_ISA_AVX512F in addition to
4854 OPTION_MASK_ISA_GFNI.
4855 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
4856 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
4857 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
4858 to OPTION_MASK_ISA_GFNI.
4859 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
4860 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
4861 OPTION_MASK_ISA_AVX512BW.
4862 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
4863 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
4864 addition to OPTION_MASK_ISA_GFNI.
4865 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
4866 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
4867 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
4868 to OPTION_MASK_ISA_GFNI.
4869 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
4870 a requirement for all ISAs rather than any of them with a few
4872 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
4874 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
4875 bitmasks to be enabled with 3 exceptions, instead of requiring any
4876 enabled ISA with lots of exceptions.
4877 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
4878 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
4879 Change avx512bw in isa attribute to avx512f.
4880 * config/i386/sgxintrin.h: Add license boilerplate.
4881 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
4882 to __AVX512F__ and __AVX512VL to __AVX512VL__.
4883 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
4884 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
4886 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
4887 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
4888 temporarily sse2 rather than sse if not enabled already.
4891 * config/i386/sse.md (VI248_VLBW): Rename to ...
4892 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
4893 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
4894 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
4895 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
4896 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
4897 mode iterator instead of VI248_VLBW.
4899 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
4901 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
4902 (record_modified): Skip clobbers; add debug output.
4903 (param_change_prob): Use sreal frequencies.
4905 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
4907 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
4908 punt for user-aligned variables.
4910 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
4912 * tree-chrec.c (chrec_contains_symbols): Return true for
4915 2018-01-05 Sudakshina Das <sudi.das@arm.com>
4918 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
4919 of (x|y) == x for BICS pattern.
4921 2018-01-05 Jakub Jelinek <jakub@redhat.com>
4923 PR tree-optimization/83605
4924 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
4925 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
4928 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
4930 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
4931 * config/epiphany/rtems.h: New file.
4933 2018-01-04 Jakub Jelinek <jakub@redhat.com>
4934 Uros Bizjak <ubizjak@gmail.com>
4937 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
4938 QIreg_operand instead of register_operand predicate.
4939 * config/i386/i386.c (ix86_rop_should_change_byte_p,
4940 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
4941 comments instead of -fmitigate[-_]rop.
4943 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4946 * cgraphunit.c (symbol_table::compile): Switch to text_section
4947 before calling assembly_start debug hook.
4948 * run-rtl-passes.c (run_rtl_passes): Likewise.
4951 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
4953 * tree-vrp.c (extract_range_from_binary_expr_1): Check
4954 range_int_cst_p rather than !symbolic_range_p before calling
4955 extract_range_from_multiplicative_op_1.
4957 2018-01-04 Jeff Law <law@redhat.com>
4959 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
4960 redundant test in assertion.
4962 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
4964 * doc/rtl.texi: Document machine_mode wrapper classes.
4966 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
4968 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
4971 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
4973 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
4974 the VEC_PERM_EXPR fold to fail.
4976 2018-01-04 Jakub Jelinek <jakub@redhat.com>
4979 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
4980 to switched_sections.
4982 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
4985 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
4988 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
4991 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
4992 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
4994 2018-01-04 Jakub Jelinek <jakub@redhat.com>
4997 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
4998 is BLKmode and bitpos not zero or mode change is needed.
5000 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5003 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
5006 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
5009 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
5010 instead of MULT rtx. Update all corresponding splitters.
5012 (*ssub<modesuffix>): Ditto.
5014 (*cmp_sadd_di): Update split patterns.
5015 (*cmp_sadd_si): Ditto.
5016 (*cmp_sadd_sidi): Ditto.
5017 (*cmp_ssub_di): Ditto.
5018 (*cmp_ssub_si): Ditto.
5019 (*cmp_ssub_sidi): Ditto.
5020 * config/alpha/predicates.md (const23_operand): New predicate.
5021 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
5022 Look for ASHIFT, not MULT inner operand.
5023 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
5025 2018-01-04 Martin Liska <mliska@suse.cz>
5027 PR gcov-profile/83669
5028 * gcov.c (output_intermediate_file): Add version to intermediate
5030 * doc/gcov.texi: Document new field 'version' in intermediate
5031 file format. Fix location of '-k' option of gcov command.
5033 2018-01-04 Martin Liska <mliska@suse.cz>
5036 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
5038 2018-01-04 Jakub Jelinek <jakub@redhat.com>
5040 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
5042 2018-01-03 Martin Sebor <msebor@redhat.com>
5044 PR tree-optimization/83655
5045 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
5046 checking calls with invalid arguments.
5048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5050 * tree-vect-stmts.c (vect_get_store_rhs): New function.
5051 (vectorizable_mask_load_store): Delete.
5052 (vectorizable_call): Return false for masked loads and stores.
5053 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
5054 instead of gimple_assign_rhs1.
5055 (vectorizable_load): Handle IFN_MASK_LOAD.
5056 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
5058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5060 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
5062 (vectorizable_mask_load_store): ...here.
5063 (vectorizable_load): ...and here.
5065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5067 * tree-vect-stmts.c (vect_build_all_ones_mask)
5068 (vect_build_zero_merge_argument): New functions, split out from...
5069 (vectorizable_load): ...here.
5071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5073 * tree-vect-stmts.c (vect_check_store_rhs): New function,
5075 (vectorizable_mask_load_store): ...here.
5076 (vectorizable_store): ...and here.
5078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5080 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
5082 (vectorizable_mask_load_store): ...here.
5084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5086 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
5087 (vect_model_store_cost): Take a vec_load_store_type instead of a
5089 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
5090 (vect_model_store_cost): Take a vec_load_store_type instead of a
5092 (vectorizable_mask_load_store): Update accordingly.
5093 (vectorizable_store): Likewise.
5094 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
5096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5098 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
5099 IFN_MASK_LOAD calls here rather than...
5100 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
5102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5103 Alan Hayward <alan.hayward@arm.com>
5104 David Sherwood <david.sherwood@arm.com>
5106 * expmed.c (extract_bit_field_1): For vector extracts,
5107 fall back to extract_bit_field_as_subreg if vec_extract
5110 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5111 Alan Hayward <alan.hayward@arm.com>
5112 David Sherwood <david.sherwood@arm.com>
5114 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
5115 they are variable or constant sized.
5116 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
5117 slots for constant-sized data.
5119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5120 Alan Hayward <alan.hayward@arm.com>
5121 David Sherwood <david.sherwood@arm.com>
5123 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
5124 handling COND_EXPRs with boolean comparisons, try to find a better
5125 basis for the mask type than the boolean itself.
5127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5129 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
5130 is calculated and how it can be overridden.
5131 * genmodes.c (max_bitsize_mode_any_mode): New variable.
5132 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
5134 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
5137 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5138 Alan Hayward <alan.hayward@arm.com>
5139 David Sherwood <david.sherwood@arm.com>
5141 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
5142 Remove the mode argument.
5143 (aarch64_simd_valid_immediate): Remove the mode and inverse
5145 * config/aarch64/iterators.md (bitsize): New iterator.
5146 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
5147 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
5148 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
5149 aarch64_simd_valid_immediate.
5150 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
5151 (aarch64_reg_or_bic_imm): Likewise.
5152 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
5153 with an insn_type enum and msl with a modifier_type enum.
5154 Replace element_width with a scalar_mode. Change the shift
5155 to unsigned int. Add constructors for scalar_float_mode and
5156 scalar_int_mode elements.
5157 (aarch64_vect_float_const_representable_p): Delete.
5158 (aarch64_can_const_movi_rtx_p)
5159 (aarch64_simd_scalar_immediate_valid_for_move)
5160 (aarch64_simd_make_constant): Update call to
5161 aarch64_simd_valid_immediate.
5162 (aarch64_advsimd_valid_immediate_hs): New function.
5163 (aarch64_advsimd_valid_immediate): Likewise.
5164 (aarch64_simd_valid_immediate): Remove mode and inverse
5165 arguments. Rewrite to use the above. Use const_vec_duplicate_p
5166 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
5167 and aarch64_float_const_representable_p on the result.
5168 (aarch64_output_simd_mov_immediate): Remove mode argument.
5169 Update call to aarch64_simd_valid_immediate and use of
5170 simd_immediate_info.
5171 (aarch64_output_scalar_simd_mov_immediate): Update call
5174 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5175 Alan Hayward <alan.hayward@arm.com>
5176 David Sherwood <david.sherwood@arm.com>
5178 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
5179 (mode_nunits): Likewise CONST_MODE_NUNITS.
5180 * machmode.def (ADJUST_NUNITS): Document.
5181 * genmodes.c (mode_data::need_nunits_adj): New field.
5182 (blank_mode): Update accordingly.
5183 (adj_nunits): New variable.
5184 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
5186 (emit_mode_size_inline): Set need_bytesize_adj for all modes
5187 listed in adj_nunits.
5188 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
5189 listed in adj_nunits. Don't emit case statements for such modes.
5190 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
5191 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
5192 nothing if adj_nunits is nonnull.
5193 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
5194 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
5195 (emit_mode_fbit): Update use of print_maybe_const_decl.
5196 (emit_move_size): Likewise. Treat the array as non-const
5198 (emit_mode_adjustments): Handle adj_nunits.
5200 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5202 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
5203 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
5204 (VECTOR_MODES): Use it.
5205 (make_vector_modes): Take the prefix as an argument.
5207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5208 Alan Hayward <alan.hayward@arm.com>
5209 David Sherwood <david.sherwood@arm.com>
5211 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
5212 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
5213 for MODE_VECTOR_BOOL.
5214 * machmode.def (VECTOR_BOOL_MODE): Document.
5215 * genmodes.c (VECTOR_BOOL_MODE): New macro.
5216 (make_vector_bool_mode): New function.
5217 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
5219 * lto-streamer-in.c (lto_input_mode_table): Likewise.
5220 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
5222 * stor-layout.c (int_mode_for_mode): Likewise.
5223 * tree.c (build_vector_type_for_mode): Likewise.
5224 * varasm.c (output_constant_pool_2): Likewise.
5225 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
5226 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
5227 for MODE_VECTOR_BOOL.
5228 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
5229 of mode class checks.
5230 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
5231 instead of a list of mode class checks.
5232 (expand_vector_scalar_condition): Likewise.
5233 (type_for_widest_vector_mode): Handle BImode as an inner mode.
5235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5236 Alan Hayward <alan.hayward@arm.com>
5237 David Sherwood <david.sherwood@arm.com>
5239 * machmode.h (mode_size): Change from unsigned short to
5241 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
5242 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
5243 or if measurement_type is not polynomial.
5244 (fixed_size_mode::includes_p): Check for constant-sized modes.
5245 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
5246 return a poly_uint16 rather than an unsigned short.
5247 (emit_mode_size): Change the type of mode_size from unsigned short
5248 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
5249 (emit_mode_adjustments): Cope with polynomial vector sizes.
5250 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
5252 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
5254 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
5255 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
5256 * caller-save.c (setup_save_areas): Likewise.
5257 (replace_reg_with_saved_mem): Likewise.
5258 * calls.c (emit_library_call_value_1): Likewise.
5259 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
5260 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
5261 (gen_lowpart_for_combine): Likewise.
5262 * convert.c (convert_to_integer_1): Likewise.
5263 * cse.c (equiv_constant, cse_insn): Likewise.
5264 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
5265 (cselib_subst_to_values): Likewise.
5266 * dce.c (word_dce_process_block): Likewise.
5267 * df-problems.c (df_word_lr_mark_ref): Likewise.
5268 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
5269 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
5270 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
5271 (rtl_for_decl_location): Likewise.
5272 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
5273 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
5274 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
5275 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
5276 (expand_expr_real_1): Likewise.
5277 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
5278 (pad_below): Likewise.
5279 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
5280 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
5281 * ira.c (get_subreg_tracking_sizes): Likewise.
5282 * ira-build.c (ira_create_allocno_objects): Likewise.
5283 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
5284 (ira_sort_regnos_for_alter_reg): Likewise.
5285 * ira-costs.c (record_operand_costs): Likewise.
5286 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
5287 (resolve_simple_move): Likewise.
5288 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
5289 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
5290 (lra_constraints): Likewise.
5291 (CONST_POOL_OK_P): Reject variable-sized modes.
5292 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
5293 (add_pseudo_to_slot, lra_spill): Likewise.
5294 * omp-low.c (omp_clause_aligned_alignment): Likewise.
5295 * optabs-query.c (get_best_extraction_insn): Likewise.
5296 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
5297 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
5298 (expand_mult_highpart, valid_multiword_target_p): Likewise.
5299 * recog.c (offsettable_address_addr_space_p): Likewise.
5300 * regcprop.c (maybe_mode_change): Likewise.
5301 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
5302 * regrename.c (build_def_use): Likewise.
5303 * regstat.c (dump_reg_info): Likewise.
5304 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
5305 (find_reloads, find_reloads_subreg_address): Likewise.
5306 * reload1.c (eliminate_regs_1): Likewise.
5307 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
5308 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
5309 (simplify_binary_operation_1, simplify_subreg): Likewise.
5310 * targhooks.c (default_function_arg_padding): Likewise.
5311 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
5312 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
5313 (verify_gimple_assign_ternary): Likewise.
5314 * tree-inline.c (estimate_move_cost): Likewise.
5315 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
5316 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
5317 (get_address_cost_ainc): Likewise.
5318 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
5319 (vect_supportable_dr_alignment): Likewise.
5320 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
5321 (vectorizable_reduction): Likewise.
5322 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
5323 (vectorizable_operation, vectorizable_load): Likewise.
5324 * tree.c (build_same_sized_truth_vector_type): Likewise.
5325 * valtrack.c (cleanup_auto_inc_dec): Likewise.
5326 * var-tracking.c (emit_note_insn_var_location): Likewise.
5327 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
5328 (ADDR_VEC_ALIGN): Likewise.
5330 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5331 Alan Hayward <alan.hayward@arm.com>
5332 David Sherwood <david.sherwood@arm.com>
5334 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
5336 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
5337 or if measurement_type is polynomial.
5338 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
5339 * combine.c (make_extraction): Likewise.
5340 * dse.c (find_shift_sequence): Likewise.
5341 * dwarf2out.c (mem_loc_descriptor): Likewise.
5342 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
5343 (extract_bit_field, extract_low_bits): Likewise.
5344 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
5345 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
5346 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
5347 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
5348 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
5349 * reload.c (find_reloads): Likewise.
5350 * reload1.c (alter_reg): Likewise.
5351 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
5352 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
5353 * tree-if-conv.c (predicate_mem_writes): Likewise.
5354 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
5355 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
5356 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
5357 * valtrack.c (dead_debug_insert_temp): Likewise.
5358 * varasm.c (mergeable_constant_section): Likewise.
5359 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
5361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5362 Alan Hayward <alan.hayward@arm.com>
5363 David Sherwood <david.sherwood@arm.com>
5365 * expr.c (expand_assignment): Cope with polynomial mode sizes
5366 when assigning to a CONCAT.
5368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5369 Alan Hayward <alan.hayward@arm.com>
5370 David Sherwood <david.sherwood@arm.com>
5372 * machmode.h (mode_precision): Change from unsigned short to
5374 (mode_to_precision): Return a poly_uint16 rather than an unsigned
5376 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
5377 or if measurement_type is not polynomial.
5378 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
5379 in which the mode is already known to be a scalar_int_mode.
5380 * genmodes.c (emit_mode_precision): Change the type of mode_precision
5381 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
5383 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
5384 for GET_MODE_PRECISION.
5385 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
5386 for GET_MODE_PRECISION.
5387 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
5389 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
5390 (expand_field_assignment, make_extraction): Likewise.
5391 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
5392 (get_last_value): Likewise.
5393 * convert.c (convert_to_integer_1): Likewise.
5394 * cse.c (cse_insn): Likewise.
5395 * expr.c (expand_expr_real_1): Likewise.
5396 * lra-constraints.c (simplify_operand_subreg): Likewise.
5397 * optabs-query.c (can_atomic_load_p): Likewise.
5398 * optabs.c (expand_atomic_load): Likewise.
5399 (expand_atomic_store): Likewise.
5400 * ree.c (combine_reaching_defs): Likewise.
5401 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
5402 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
5403 * tree.h (type_has_mode_precision_p): Likewise.
5404 * ubsan.c (instrument_si_overflow): Likewise.
5406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5407 Alan Hayward <alan.hayward@arm.com>
5408 David Sherwood <david.sherwood@arm.com>
5410 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
5411 polynomial numbers of units.
5412 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
5413 (valid_vector_subparts_p): New function.
5414 (build_vector_type): Remove temporary shim and take the number
5415 of units as a poly_uint64 rather than an int.
5416 (build_opaque_vector_type): Take the number of units as a
5417 poly_uint64 rather than an int.
5418 * tree.c (build_vector_from_ctor): Handle polynomial
5419 TYPE_VECTOR_SUBPARTS.
5420 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
5421 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
5422 (build_vector_from_val): If the number of units is variable,
5423 use build_vec_duplicate_cst for constant operands and
5424 VEC_DUPLICATE_EXPR otherwise.
5425 (make_vector_type): Remove temporary is_constant ().
5426 (build_vector_type, build_opaque_vector_type): Take the number of
5427 units as a poly_uint64 rather than an int.
5428 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
5430 * cfgexpand.c (expand_debug_expr): Likewise.
5431 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
5432 (store_constructor, expand_expr_real_1): Likewise.
5433 (const_scalar_mask_from_tree): Likewise.
5434 * fold-const-call.c (fold_const_reduction): Likewise.
5435 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
5436 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
5437 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
5438 (fold_relational_const): Likewise.
5439 (native_interpret_vector): Likewise. Change the size from an
5440 int to an unsigned int.
5441 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
5442 TYPE_VECTOR_SUBPARTS.
5443 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
5444 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
5445 duplicating a non-constant operand into a variable-length vector.
5446 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
5447 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
5448 * ipa-icf.c (sem_variable::equals): Likewise.
5449 * match.pd: Likewise.
5450 * omp-simd-clone.c (simd_clone_subparts): Likewise.
5451 * print-tree.c (print_node): Likewise.
5452 * stor-layout.c (layout_type): Likewise.
5453 * targhooks.c (default_builtin_vectorization_cost): Likewise.
5454 * tree-cfg.c (verify_gimple_comparison): Likewise.
5455 (verify_gimple_assign_binary): Likewise.
5456 (verify_gimple_assign_ternary): Likewise.
5457 (verify_gimple_assign_single): Likewise.
5458 * tree-pretty-print.c (dump_generic_node): Likewise.
5459 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
5460 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
5461 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
5462 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
5463 (vect_shift_permute_load_chain): Likewise.
5464 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
5465 (expand_vector_condition, optimize_vector_constructor): Likewise.
5466 (lower_vec_perm, get_compute_type): Likewise.
5467 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
5468 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
5469 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
5470 (vect_recog_mask_conversion_pattern): Likewise.
5471 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
5472 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
5473 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
5474 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
5475 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
5476 (vectorizable_shift, vectorizable_operation, vectorizable_store)
5477 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
5478 (supportable_widening_operation): Likewise.
5479 (supportable_narrowing_operation): Likewise.
5480 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
5482 * varasm.c (output_constant): Likewise.
5484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5485 Alan Hayward <alan.hayward@arm.com>
5486 David Sherwood <david.sherwood@arm.com>
5488 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
5489 so that both the length == 3 and length != 3 cases set up their
5490 own permute vectors. Add comments explaining why we know the
5491 number of elements is constant.
5492 (vect_permute_load_chain): Likewise.
5494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5495 Alan Hayward <alan.hayward@arm.com>
5496 David Sherwood <david.sherwood@arm.com>
5498 * machmode.h (mode_nunits): Change from unsigned char to
5500 (ONLY_FIXED_SIZE_MODES): New macro.
5501 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
5502 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
5503 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
5505 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
5506 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
5507 or if measurement_type is not polynomial.
5508 * genmodes.c (ZERO_COEFFS): New macro.
5509 (emit_mode_nunits_inline): Make mode_nunits_inline return a
5511 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
5512 Use ZERO_COEFFS when emitting initializers.
5513 * data-streamer.h (bp_pack_poly_value): New function.
5514 (bp_unpack_poly_value): Likewise.
5515 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
5516 for GET_MODE_NUNITS.
5517 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
5518 for GET_MODE_NUNITS.
5519 * tree.c (make_vector_type): Remove temporary shim and make
5520 the real function take the number of units as a poly_uint64
5522 (build_vector_type_for_mode): Handle polynomial nunits.
5523 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
5524 * emit-rtl.c (const_vec_series_p_1): Likewise.
5525 (gen_rtx_CONST_VECTOR): Likewise.
5526 * fold-const.c (test_vec_duplicate_folding): Likewise.
5527 * genrecog.c (validate_pattern): Likewise.
5528 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
5529 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
5530 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
5531 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
5532 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
5533 * rtlanal.c (subreg_get_info): Likewise.
5534 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
5535 (vect_grouped_load_supported): Likewise.
5536 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
5537 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
5538 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
5539 (simplify_const_unary_operation, simplify_binary_operation_1)
5540 (simplify_const_binary_operation, simplify_ternary_operation)
5541 (test_vector_ops_duplicate, test_vector_ops): Likewise.
5542 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
5543 instead of CONST_VECTOR_NUNITS.
5544 * varasm.c (output_constant_pool_2): Likewise.
5545 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
5546 explicit-encoded elements in the XVEC for variable-length vectors.
5548 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5550 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
5552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5553 Alan Hayward <alan.hayward@arm.com>
5554 David Sherwood <david.sherwood@arm.com>
5556 * coretypes.h (fixed_size_mode): Declare.
5557 (fixed_size_mode_pod): New typedef.
5558 * builtins.h (target_builtins::x_apply_args_mode)
5559 (target_builtins::x_apply_result_mode): Change type to
5560 fixed_size_mode_pod.
5561 * builtins.c (apply_args_size, apply_result_size, result_vector)
5562 (expand_builtin_apply_args_1, expand_builtin_apply)
5563 (expand_builtin_return): Update accordingly.
5565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5567 * cse.c (hash_rtx_cb): Hash only the encoded elements.
5568 * cselib.c (cselib_hash_rtx): Likewise.
5569 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
5570 CONST_VECTOR encoding.
5572 2018-01-03 Jakub Jelinek <jakub@redhat.com>
5573 Jeff Law <law@redhat.com>
5576 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
5577 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
5578 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
5579 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
5582 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
5583 explicitly probe *sp in a noreturn function if there were any callee
5584 register saves or frame pointer is needed.
5586 2018-01-03 Jakub Jelinek <jakub@redhat.com>
5589 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
5590 BLKmode for ternary, binary or unary expressions.
5593 * var-tracking.c (delete_vta_debug_insn): New inline function.
5594 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
5595 insns from get_insns () to NULL instead of each bb separately.
5596 Use delete_vta_debug_insn. No longer static.
5597 (vt_debug_insns_local, variable_tracking_main_1): Adjust
5598 delete_vta_debug_insns callers.
5599 * rtl.h (delete_vta_debug_insns): Declare.
5600 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
5601 instead of variable_tracking_main.
5603 2018-01-03 Martin Sebor <msebor@redhat.com>
5605 PR tree-optimization/83603
5606 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
5607 arguments past the endof the argument list in functions declared
5608 without a prototype.
5609 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
5610 Avoid checking when arguments are null.
5612 2018-01-03 Martin Sebor <msebor@redhat.com>
5615 * doc/extend.texi (attribute const): Fix a typo.
5616 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
5617 issuing -Wsuggest-attribute for void functions.
5619 2018-01-03 Martin Sebor <msebor@redhat.com>
5621 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
5622 offset_int::from instead of wide_int::to_shwi.
5623 (maybe_diag_overlap): Remove assertion.
5624 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
5625 * gimple-ssa-sprintf.c (format_directive): Same.
5626 (parse_directive): Same.
5627 (sprintf_dom_walker::compute_format_length): Same.
5628 (try_substitute_return_value): Same.
5630 2018-01-03 Jeff Law <law@redhat.com>
5633 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
5634 non-constant residual for zero at runtime and avoid probing in
5635 that case. Reorganize code for trailing problem to mirror handling
5638 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5640 PR tree-optimization/83501
5641 * tree-ssa-strlen.c (get_string_cst): New.
5642 (handle_char_store): Call get_string_cst.
5644 2018-01-03 Martin Liska <mliska@suse.cz>
5646 PR tree-optimization/83593
5647 * tree-ssa-strlen.c: Include tree-cfg.h.
5648 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
5649 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
5650 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
5652 (strlen_dom_walker::before_dom_children): Call
5653 gimple_purge_dead_eh_edges. Dump tranformation with details
5655 (strlen_dom_walker::before_dom_children): Update call by adding
5656 new argument cleanup_eh.
5657 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
5659 2018-01-03 Martin Liska <mliska@suse.cz>
5662 * cif-code.def (VARIADIC_THUNK): New enum value.
5663 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
5666 2018-01-03 Jan Beulich <jbeulich@suse.com>
5668 * sse.md (mov<mode>_internal): Tighten condition for when to use
5669 vmovdqu<ssescalarsize> for TI and OI modes.
5671 2018-01-03 Jakub Jelinek <jakub@redhat.com>
5673 Update copyright years.
5675 2018-01-03 Martin Liska <mliska@suse.cz>
5678 * ipa-visibility.c (function_and_variable_visibility): Skip
5679 functions with noipa attribure.
5681 2018-01-03 Jakub Jelinek <jakub@redhat.com>
5683 * gcc.c (process_command): Update copyright notice dates.
5684 * gcov-dump.c (print_version): Ditto.
5685 * gcov.c (print_version): Ditto.
5686 * gcov-tool.c (print_version): Ditto.
5687 * gengtype.c (create_file): Ditto.
5688 * doc/cpp.texi: Bump @copying's copyright year.
5689 * doc/cppinternals.texi: Ditto.
5690 * doc/gcc.texi: Ditto.
5691 * doc/gccint.texi: Ditto.
5692 * doc/gcov.texi: Ditto.
5693 * doc/install.texi: Ditto.
5694 * doc/invoke.texi: Ditto.
5696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5698 * vector-builder.h (vector_builder::m_full_nelts): Change from
5699 unsigned int to poly_uint64.
5700 (vector_builder::full_nelts): Update prototype accordingly.
5701 (vector_builder::new_vector): Likewise.
5702 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
5703 (vector_builder::operator ==): Likewise.
5704 (vector_builder::finalize): Likewise.
5705 * int-vector-builder.h (int_vector_builder::int_vector_builder):
5706 Take the number of elements as a poly_uint64 rather than an
5708 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
5709 from unsigned int to poly_uint64.
5710 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
5711 (vec_perm_indices::new_vector): Likewise.
5712 (vec_perm_indices::length): Likewise.
5713 (vec_perm_indices::nelts_per_input): Likewise.
5714 (vec_perm_indices::input_nelts): Likewise.
5715 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
5716 number of elements per input as a poly_uint64 rather than an
5717 unsigned int. Use the original encoding for variable-length
5718 vectors, rather than clamping each individual element.
5719 For the second and subsequent elements in each pattern,
5720 clamp the step and base before clamping their sum.
5721 (vec_perm_indices::series_p): Handle polynomial element counts.
5722 (vec_perm_indices::all_in_range_p): Likewise.
5723 (vec_perm_indices_to_tree): Likewise.
5724 (vec_perm_indices_to_rtx): Likewise.
5725 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
5726 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
5727 (tree_vector_builder::new_binary_operation): Handle polynomial
5728 element counts. Return false if we need to know the number
5729 of elements at compile time.
5730 * fold-const.c (fold_vec_perm): Punt if the number of elements
5731 isn't known at compile time.
5733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5735 * vec-perm-indices.h (vec_perm_builder): Change element type
5736 from HOST_WIDE_INT to poly_int64.
5737 (vec_perm_indices::element_type): Update accordingly.
5738 (vec_perm_indices::clamp): Handle polynomial element_types.
5739 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
5740 (vec_perm_indices::all_in_range_p): Likewise.
5741 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
5743 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
5744 polynomial vec_perm_indices element types.
5745 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
5746 * fold-const.c (fold_vec_perm): Likewise.
5747 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
5748 * tree-vect-generic.c (lower_vec_perm): Likewise.
5749 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
5750 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
5751 element type to HOST_WIDE_INT.
5753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5754 Alan Hayward <alan.hayward@arm.com>
5755 David Sherwood <david.sherwood@arm.com>
5757 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
5758 rather than an int. Use plus_constant.
5759 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
5760 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
5762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5763 Alan Hayward <alan.hayward@arm.com>
5764 David Sherwood <david.sherwood@arm.com>
5766 * calls.c (emit_call_1, expand_call): Change struct_value_size from
5767 a HOST_WIDE_INT to a poly_int64.
5769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5770 Alan Hayward <alan.hayward@arm.com>
5771 David Sherwood <david.sherwood@arm.com>
5773 * calls.c (load_register_parameters): Cope with polynomial
5774 mode sizes. Require a constant size for BLKmode parameters
5775 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
5776 forces a parameter to be padded at the lsb end in order to
5777 fill a complete number of words, require the parameter size
5778 to be ordered wrt UNITS_PER_WORD.
5780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5781 Alan Hayward <alan.hayward@arm.com>
5782 David Sherwood <david.sherwood@arm.com>
5784 * reload1.c (spill_stack_slot_width): Change element type
5785 from unsigned int to poly_uint64_pod.
5786 (alter_reg): Treat mode sizes as polynomial.
5788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5789 Alan Hayward <alan.hayward@arm.com>
5790 David Sherwood <david.sherwood@arm.com>
5792 * reload.c (complex_word_subreg_p): New function.
5793 (reload_inner_reg_of_subreg, push_reload): Use it.
5795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5796 Alan Hayward <alan.hayward@arm.com>
5797 David Sherwood <david.sherwood@arm.com>
5799 * lra-constraints.c (process_alt_operands): Reject matched
5800 operands whose sizes aren't ordered.
5801 (match_reload): Refer to this check here.
5803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5804 Alan Hayward <alan.hayward@arm.com>
5805 David Sherwood <david.sherwood@arm.com>
5807 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
5808 that the mode size is in the set {1, 2, 4, 8, 16}.
5810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5811 Alan Hayward <alan.hayward@arm.com>
5812 David Sherwood <david.sherwood@arm.com>
5814 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
5815 Use plus_constant instead of gen_rtx_PLUS.
5817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5818 Alan Hayward <alan.hayward@arm.com>
5819 David Sherwood <david.sherwood@arm.com>
5821 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
5822 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
5823 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
5824 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
5825 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
5826 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
5827 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
5828 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
5829 * config/i386/i386.c (ix86_push_rounding): ...this new function.
5830 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
5832 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
5833 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
5834 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
5835 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
5836 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
5837 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
5838 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
5839 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
5840 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
5841 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
5843 * expr.c (emit_move_resolve_push): Treat the input and result
5844 of PUSH_ROUNDING as a poly_int64.
5845 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
5846 (emit_push_insn): Likewise.
5847 * lra-eliminations.c (mark_not_eliminable): Likewise.
5848 * recog.c (push_operand): Likewise.
5849 * reload1.c (elimination_effects): Likewise.
5850 * rtlanal.c (nonzero_bits1): Likewise.
5851 * calls.c (store_one_arg): Likewise. Require the padding to be
5852 known at compile time.
5854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5855 Alan Hayward <alan.hayward@arm.com>
5856 David Sherwood <david.sherwood@arm.com>
5858 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
5859 Use plus_constant instead of gen_rtx_PLUS.
5861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5862 Alan Hayward <alan.hayward@arm.com>
5863 David Sherwood <david.sherwood@arm.com>
5865 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
5868 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5869 Alan Hayward <alan.hayward@arm.com>
5870 David Sherwood <david.sherwood@arm.com>
5872 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
5873 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
5874 via stack temporaries. Treat the mode size as polynomial too.
5876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5877 Alan Hayward <alan.hayward@arm.com>
5878 David Sherwood <david.sherwood@arm.com>
5880 * expr.c (expand_expr_real_2): When handling conversions involving
5881 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
5882 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
5883 as a poly_uint64 too.
5885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5886 Alan Hayward <alan.hayward@arm.com>
5887 David Sherwood <david.sherwood@arm.com>
5889 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
5891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5892 Alan Hayward <alan.hayward@arm.com>
5893 David Sherwood <david.sherwood@arm.com>
5895 * combine.c (can_change_dest_mode): Handle polynomial
5896 REGMODE_NATURAL_SIZE.
5897 * expmed.c (store_bit_field_1): Likewise.
5898 * expr.c (store_constructor): Likewise.
5899 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
5900 and polynomial REGMODE_NATURAL_SIZE.
5901 (gen_lowpart_common): Likewise.
5902 * reginfo.c (record_subregs_of_mode): Likewise.
5903 * rtlanal.c (read_modify_subreg_p): Likewise.
5905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5906 Alan Hayward <alan.hayward@arm.com>
5907 David Sherwood <david.sherwood@arm.com>
5909 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
5910 numbers of elements.
5912 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5913 Alan Hayward <alan.hayward@arm.com>
5914 David Sherwood <david.sherwood@arm.com>
5916 * match.pd: Cope with polynomial numbers of vector elements.
5918 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5919 Alan Hayward <alan.hayward@arm.com>
5920 David Sherwood <david.sherwood@arm.com>
5922 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
5923 in a POINTER_PLUS_EXPR.
5925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5926 Alan Hayward <alan.hayward@arm.com>
5927 David Sherwood <david.sherwood@arm.com>
5929 * omp-simd-clone.c (simd_clone_subparts): New function.
5930 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
5931 (ipa_simd_modify_function_body): Likewise.
5933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5934 Alan Hayward <alan.hayward@arm.com>
5935 David Sherwood <david.sherwood@arm.com>
5937 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
5938 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
5939 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
5940 (expand_vector_condition, vector_element): Likewise.
5941 (subparts_gt): New function.
5942 (get_compute_type): Use subparts_gt.
5943 (count_type_subparts): Delete.
5944 (expand_vector_operations_1): Use subparts_gt instead of
5945 count_type_subparts.
5947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5948 Alan Hayward <alan.hayward@arm.com>
5949 David Sherwood <david.sherwood@arm.com>
5951 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
5952 (vect_compile_time_alias): ...this new function. Do the calculation
5953 on poly_ints rather than trees.
5954 (vect_prune_runtime_alias_test_list): Update call accordingly.
5956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5957 Alan Hayward <alan.hayward@arm.com>
5958 David Sherwood <david.sherwood@arm.com>
5960 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
5962 (vect_schedule_slp_instance): Likewise.
5964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5965 Alan Hayward <alan.hayward@arm.com>
5966 David Sherwood <david.sherwood@arm.com>
5968 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
5969 constant and extern definitions for variable-length vectors.
5970 (vect_get_constant_vectors): Note that the number of units
5971 is known to be constant.
5973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5974 Alan Hayward <alan.hayward@arm.com>
5975 David Sherwood <david.sherwood@arm.com>
5977 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
5978 of units as polynomial. Choose between WIDE and NARROW based
5981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5982 Alan Hayward <alan.hayward@arm.com>
5983 David Sherwood <david.sherwood@arm.com>
5985 * tree-vect-stmts.c (simd_clone_subparts): New function.
5986 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
5988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5989 Alan Hayward <alan.hayward@arm.com>
5990 David Sherwood <david.sherwood@arm.com>
5992 * tree-vect-stmts.c (vectorizable_call): Treat the number of
5993 vectors as polynomial. Use build_index_vector for
5996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5997 Alan Hayward <alan.hayward@arm.com>
5998 David Sherwood <david.sherwood@arm.com>
6000 * tree-vect-stmts.c (get_load_store_type): Treat the number of
6001 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
6002 for variable-length vectors.
6003 (vectorizable_mask_load_store): Treat the number of units as
6004 polynomial, asserting that it is constant if the condition has
6005 already been enforced.
6006 (vectorizable_store, vectorizable_load): Likewise.
6008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6009 Alan Hayward <alan.hayward@arm.com>
6010 David Sherwood <david.sherwood@arm.com>
6012 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
6013 of units as polynomial. Punt if we can't tell at compile time
6014 which vector contains the final result.
6016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6017 Alan Hayward <alan.hayward@arm.com>
6018 David Sherwood <david.sherwood@arm.com>
6020 * tree-vect-loop.c (vectorizable_induction): Treat the number
6021 of units as polynomial. Punt on SLP inductions. Use an integer
6022 VEC_SERIES_EXPR for variable-length integer reductions. Use a
6023 cast of such a series for variable-length floating-point
6026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6027 Alan Hayward <alan.hayward@arm.com>
6028 David Sherwood <david.sherwood@arm.com>
6030 * tree.h (build_index_vector): Declare.
6031 * tree.c (build_index_vector): New function.
6032 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
6033 of units as polynomial, forcibly converting it to a constant if
6034 vectorizable_reduction has already enforced the condition.
6035 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
6036 to create a {1,2,3,...} vector.
6037 (vectorizable_reduction): Treat the number of units as polynomial.
6038 Choose vectype_in based on the largest scalar element size rather
6039 than the smallest number of units. Enforce the restrictions
6042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6043 Alan Hayward <alan.hayward@arm.com>
6044 David Sherwood <david.sherwood@arm.com>
6046 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
6047 number of units as polynomial.
6049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6050 Alan Hayward <alan.hayward@arm.com>
6051 David Sherwood <david.sherwood@arm.com>
6053 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
6054 * target.def (autovectorize_vector_sizes): Return the vector sizes
6055 by pointer, using vector_sizes rather than a bitmask.
6056 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
6057 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
6058 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
6060 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
6061 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
6062 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
6063 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
6064 * omp-general.c (omp_max_vf): Likewise.
6065 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6066 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
6067 * tree-vect-loop.c (vect_analyze_loop): Likewise.
6068 * tree-vect-slp.c (vect_slp_bb): Likewise.
6069 * doc/tm.texi: Regenerate.
6070 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
6072 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
6073 the vector size as a poly_uint64 rather than an unsigned int.
6074 (current_vector_size): Change from an unsigned int to a poly_uint64.
6075 (get_vectype_for_scalar_type): Update accordingly.
6076 * tree.h (build_truth_vector_type): Take the size and number of
6077 units as a poly_uint64 rather than an unsigned int.
6078 (build_vector_type): Add a temporary overload that takes
6079 the number of units as a poly_uint64 rather than an unsigned int.
6080 * tree.c (make_vector_type): Likewise.
6081 (build_truth_vector_type): Take the number of units as a poly_uint64
6082 rather than an unsigned int.
6084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6085 Alan Hayward <alan.hayward@arm.com>
6086 David Sherwood <david.sherwood@arm.com>
6088 * target.def (get_mask_mode): Take the number of units and length
6089 as poly_uint64s rather than unsigned ints.
6090 * targhooks.h (default_get_mask_mode): Update accordingly.
6091 * targhooks.c (default_get_mask_mode): Likewise.
6092 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
6093 * doc/tm.texi: Regenerate.
6095 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6096 Alan Hayward <alan.hayward@arm.com>
6097 David Sherwood <david.sherwood@arm.com>
6099 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
6100 * omp-general.c (omp_max_vf): Likewise.
6101 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
6102 (expand_omp_simd): Handle polynomial safelen.
6103 * omp-low.c (omplow_simd_context): Add a default constructor.
6104 (omplow_simd_context::max_vf): Change from int to poly_uint64.
6105 (lower_rec_simd_input_clauses): Update accordingly.
6106 (lower_rec_input_clauses): Likewise.
6108 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6109 Alan Hayward <alan.hayward@arm.com>
6110 David Sherwood <david.sherwood@arm.com>
6112 * tree-vectorizer.h (vect_nunits_for_cost): New function.
6113 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
6114 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
6115 (vect_analyze_slp_cost): Likewise.
6116 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
6117 (vect_model_load_cost): Likewise.
6119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6120 Alan Hayward <alan.hayward@arm.com>
6121 David Sherwood <david.sherwood@arm.com>
6123 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
6124 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
6125 from an unsigned int * to a poly_uint64_pod *.
6126 (calculate_unrolling_factor): New function.
6127 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
6129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6130 Alan Hayward <alan.hayward@arm.com>
6131 David Sherwood <david.sherwood@arm.com>
6133 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
6134 from an unsigned int to a poly_uint64.
6135 (_loop_vec_info::slp_unrolling_factor): Likewise.
6136 (_loop_vec_info::vectorization_factor): Change from an int
6138 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
6139 (vect_get_num_vectors): New function.
6140 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
6141 (vect_get_num_copies): Use vect_get_num_vectors.
6142 (vect_analyze_data_ref_dependences): Change max_vf from an int *
6143 to an unsigned int *.
6144 (vect_analyze_data_refs): Change min_vf from an int * to a
6146 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
6147 than an unsigned HOST_WIDE_INT.
6148 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
6149 (vect_analyze_data_ref_dependence): Change max_vf from an int *
6150 to an unsigned int *.
6151 (vect_analyze_data_ref_dependences): Likewise.
6152 (vect_compute_data_ref_alignment): Handle polynomial vf.
6153 (vect_enhance_data_refs_alignment): Likewise.
6154 (vect_prune_runtime_alias_test_list): Likewise.
6155 (vect_shift_permute_load_chain): Likewise.
6156 (vect_supportable_dr_alignment): Likewise.
6157 (dependence_distance_ge_vf): Take the vectorization factor as a
6158 poly_uint64 rather than an unsigned HOST_WIDE_INT.
6159 (vect_analyze_data_refs): Change min_vf from an int * to a
6161 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
6162 vfm1 as a poly_uint64 rather than an int. Make the same change
6163 for the returned bound_scalar.
6164 (vect_gen_vector_loop_niters): Handle polynomial vf.
6165 (vect_do_peeling): Likewise. Update call to
6166 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
6167 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
6169 * tree-vect-loop.c (vect_determine_vectorization_factor)
6170 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
6171 (vect_get_known_peeling_cost): Likewise.
6172 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
6173 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
6174 (vect_transform_loop): Likewise. Use the lowest possible VF when
6175 updating the upper bounds of the loop.
6176 (vect_min_worthwhile_factor): Make static. Return an unsigned int
6178 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
6179 polynomial unroll factors.
6180 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
6181 (vect_make_slp_decision): Likewise.
6182 (vect_supported_load_permutation_p): Likewise, and polynomial
6184 (vect_analyze_slp_cost): Handle polynomial vf.
6185 (vect_slp_analyze_node_operations): Likewise.
6186 (vect_slp_analyze_bb_1): Likewise.
6187 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
6188 than an unsigned HOST_WIDE_INT.
6189 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
6190 (vectorizable_load): Handle polynomial vf.
6191 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
6193 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
6195 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6196 Alan Hayward <alan.hayward@arm.com>
6197 David Sherwood <david.sherwood@arm.com>
6199 * match.pd: Handle bit operations involving three constants
6200 and try to fold one pair.
6202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6204 * tree-vect-loop-manip.c: Include gimple-fold.h.
6205 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
6206 niters_maybe_zero parameters. Handle other cases besides a step of 1.
6207 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
6208 Add a path that uses a step of VF instead of 1, but disable it
6210 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
6211 and niters_no_overflow parameters. Update calls to
6212 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
6213 Create a new SSA name if the latter choses to use a ste other
6214 than zero, and return it via niters_vector_mult_vf_var.
6215 * tree-vect-loop.c (vect_transform_loop): Update calls to
6216 vect_do_peeling, vect_gen_vector_loop_niters and
6217 slpeel_make_loop_iterate_ntimes.
6218 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
6219 (vect_gen_vector_loop_niters): Update declarations after above changes.
6221 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
6223 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
6224 128-bit round to integer instructions.
6225 (ceil<mode>2): Likewise.
6226 (btrunc<mode>2): Likewise.
6227 (round<mode>2): Likewise.
6229 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6231 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
6232 unaligned VSX load/store on P8/P9.
6233 (expand_block_clear): Allow the use of unaligned VSX
6234 load/store on P8/P9.
6236 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6238 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
6240 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
6241 swap associated with both a load and a store.
6243 2018-01-02 Andrew Waterman <andrew@sifive.com>
6245 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
6246 * config/riscv/riscv.md (clear_cache): Use it.
6248 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
6250 * web.c: Remove out-of-date comment.
6252 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6254 * expr.c (fixup_args_size_notes): Check that any existing
6255 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
6256 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
6257 (emit_single_push_insn): ...here.
6259 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6261 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
6262 (const_vector_encoded_nelts): New function.
6263 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
6264 (const_vector_int_elt, const_vector_elt): Declare.
6265 * emit-rtl.c (const_vector_int_elt_1): New function.
6266 (const_vector_elt): Likewise.
6267 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
6268 of CONST_VECTOR_ELT.
6270 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6272 * expr.c: Include rtx-vector-builder.h.
6273 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
6274 directly on the tree encoding.
6275 (const_vector_from_tree): Likewise.
6276 * optabs.c: Include rtx-vector-builder.h.
6277 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
6278 sequence of "u" values.
6279 * vec-perm-indices.c: Include rtx-vector-builder.h.
6280 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
6281 directly on the vec_perm_indices encoding.
6283 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6285 * doc/rtl.texi (const_vector): Describe new encoding scheme.
6286 * Makefile.in (OBJS): Add rtx-vector-builder.o.
6287 * rtx-vector-builder.h: New file.
6288 * rtx-vector-builder.c: Likewise.
6289 * rtl.h (rtx_def::u2): Add a const_vector field.
6290 (CONST_VECTOR_NPATTERNS): New macro.
6291 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
6292 (CONST_VECTOR_DUPLICATE_P): Likewise.
6293 (CONST_VECTOR_STEPPED_P): Likewise.
6294 (CONST_VECTOR_ENCODED_ELT): Likewise.
6295 (const_vec_duplicate_p): Check for a duplicated vector encoding.
6296 (unwrap_const_vec_duplicate): Likewise.
6297 (const_vec_series_p): Check for a non-duplicated vector encoding.
6298 Say that the function only returns true for integer vectors.
6299 * emit-rtl.c: Include rtx-vector-builder.h.
6300 (gen_const_vec_duplicate_1): Delete.
6301 (gen_const_vector): Call gen_const_vec_duplicate instead of
6302 gen_const_vec_duplicate_1.
6303 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
6304 (gen_const_vec_duplicate): Use rtx_vector_builder.
6305 (gen_const_vec_series): Likewise.
6306 (gen_rtx_CONST_VECTOR): Likewise.
6307 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
6308 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
6309 Build a new vector rather than modifying a CONST_VECTOR in-place.
6310 (handle_special_swappables): Update call accordingly.
6311 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
6312 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
6313 Build a new vector rather than modifying a CONST_VECTOR in-place.
6314 (handle_special_swappables): Update call accordingly.
6316 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6318 * simplify-rtx.c (simplify_const_binary_operation): Use
6319 CONST_VECTOR_ELT instead of XVECEXP.
6321 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6323 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
6324 the selector elements to be different from the data elements
6325 if the selector is a VECTOR_CST.
6326 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
6327 ssizetype for the selector.
6329 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6331 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
6332 before testing each element individually.
6333 * tree-vect-generic.c (lower_vec_perm): Likewise.
6335 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6337 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
6338 * selftest-run-tests.c (selftest::run_tests): Call it.
6339 * vector-builder.h (vector_builder::operator ==): New function.
6340 (vector_builder::operator !=): Likewise.
6341 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
6342 (vec_perm_indices::all_from_input_p): New function.
6343 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
6344 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
6345 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
6346 instead of reading the VECTOR_CST directly. Detect whether both
6347 vector inputs are the same before constructing the vec_perm_indices,
6348 and update the number of inputs argument accordingly. Use the
6349 utility functions added above. Only construct sel2 if we need to.
6351 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6353 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
6354 the broadcast of the low byte.
6355 (expand_mult_highpart): Use an explicit encoding for the permutes.
6356 * optabs-query.c (can_mult_highpart_p): Likewise.
6357 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
6358 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6359 (vectorizable_bswap): Likewise.
6360 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
6361 explicit encoding for the power-of-2 permutes.
6362 (vect_permute_store_chain): Likewise.
6363 (vect_grouped_load_supported): Likewise.
6364 (vect_permute_load_chain): Likewise.
6366 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6368 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
6369 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
6370 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
6371 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
6372 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
6373 (vect_gen_perm_mask_any): Likewise.
6375 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6377 * int-vector-builder.h: New file.
6378 * vec-perm-indices.h: Include int-vector-builder.h.
6379 (vec_perm_indices): Redefine as an int_vector_builder.
6380 (auto_vec_perm_indices): Delete.
6381 (vec_perm_builder): Redefine as a stand-alone class.
6382 (vec_perm_indices::vec_perm_indices): New function.
6383 (vec_perm_indices::clamp): Likewise.
6384 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
6385 (vec_perm_indices::new_vector): New function.
6386 (vec_perm_indices::new_expanded_vector): Update for new
6387 vec_perm_indices class.
6388 (vec_perm_indices::rotate_inputs): New function.
6389 (vec_perm_indices::all_in_range_p): Operate directly on the
6390 encoded form, without computing elided elements.
6391 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
6392 encoding. Update for new vec_perm_indices class.
6393 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
6394 the given vec_perm_builder.
6395 (expand_vec_perm_var): Update vec_perm_builder constructor.
6396 (expand_mult_highpart): Use vec_perm_builder instead of
6397 auto_vec_perm_indices.
6398 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
6399 vec_perm_indices instead of auto_vec_perm_indices. Use a single
6400 or double series encoding as appropriate.
6401 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
6402 vec_perm_indices instead of auto_vec_perm_indices.
6403 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6404 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
6405 (vect_permute_store_chain): Likewise.
6406 (vect_grouped_load_supported): Likewise.
6407 (vect_permute_load_chain): Likewise.
6408 (vect_shift_permute_load_chain): Likewise.
6409 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
6410 (vect_transform_slp_perm_load): Likewise.
6411 (vect_schedule_slp_instance): Likewise.
6412 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6413 (vectorizable_mask_load_store): Likewise.
6414 (vectorizable_bswap): Likewise.
6415 (vectorizable_store): Likewise.
6416 (vectorizable_load): Likewise.
6417 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
6418 vec_perm_indices instead of auto_vec_perm_indices. Use
6419 tree_to_vec_perm_builder to read the vector from a tree.
6420 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
6421 vec_perm_builder instead of a vec_perm_indices.
6422 (have_whole_vector_shift): Use vec_perm_builder and
6423 vec_perm_indices instead of auto_vec_perm_indices. Leave the
6424 truncation to calc_vec_perm_mask_for_shift.
6425 (vect_create_epilog_for_reduction): Likewise.
6426 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
6427 from auto_vec_perm_indices to vec_perm_indices.
6428 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
6429 instead of changing individual elements.
6430 (aarch64_vectorize_vec_perm_const): Use new_vector to install
6431 the vector in d.perm.
6432 * config/arm/arm.c (expand_vec_perm_d::perm): Change
6433 from auto_vec_perm_indices to vec_perm_indices.
6434 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
6435 instead of changing individual elements.
6436 (arm_vectorize_vec_perm_const): Use new_vector to install
6437 the vector in d.perm.
6438 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
6439 Update vec_perm_builder constructor.
6440 (rs6000_expand_interleave): Likewise.
6441 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
6442 (rs6000_expand_interleave): Likewise.
6444 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6446 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
6447 to qimode could truncate the indices.
6448 * optabs.c (expand_vec_perm_var): Likewise.
6450 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6452 * Makefile.in (OBJS): Add vec-perm-indices.o.
6453 * vec-perm-indices.h: New file.
6454 * vec-perm-indices.c: Likewise.
6455 * target.h (vec_perm_indices): Replace with a forward class
6457 (auto_vec_perm_indices): Move to vec-perm-indices.h.
6458 * optabs.h: Include vec-perm-indices.h.
6459 (expand_vec_perm): Delete.
6460 (selector_fits_mode_p, expand_vec_perm_var): Declare.
6461 (expand_vec_perm_const): Declare.
6462 * target.def (vec_perm_const_ok): Replace with...
6463 (vec_perm_const): ...this new hook.
6464 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
6465 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
6466 * doc/tm.texi: Regenerate.
6467 * optabs.def (vec_perm_const): Delete.
6468 * doc/md.texi (vec_perm_const): Likewise.
6469 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
6470 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
6471 expand_vec_perm for constant permutation vectors. Assert that
6472 the mode of variable permutation vectors is the integer equivalent
6473 of the mode that is being permuted.
6474 * optabs-query.h (selector_fits_mode_p): Declare.
6475 * optabs-query.c: Include vec-perm-indices.h.
6476 (selector_fits_mode_p): New function.
6477 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
6478 is defined, instead of checking whether the vec_perm_const_optab
6479 exists. Use targetm.vectorize.vec_perm_const instead of
6480 targetm.vectorize.vec_perm_const_ok. Check whether the indices
6481 fit in the vector mode before using a variable permute.
6482 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
6483 vec_perm_indices instead of an rtx.
6484 (expand_vec_perm): Replace with...
6485 (expand_vec_perm_const): ...this new function. Take the selector
6486 as a vec_perm_indices rather than an rtx. Also take the mode of
6487 the selector. Update call to shift_amt_for_vec_perm_mask.
6488 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
6489 Use vec_perm_indices::new_expanded_vector to expand the original
6490 selector into bytes. Check whether the indices fit in the vector
6491 mode before using a variable permute.
6492 (expand_vec_perm_var): Make global.
6493 (expand_mult_highpart): Use expand_vec_perm_const.
6494 * fold-const.c: Includes vec-perm-indices.h.
6495 * tree-ssa-forwprop.c: Likewise.
6496 * tree-vect-data-refs.c: Likewise.
6497 * tree-vect-generic.c: Likewise.
6498 * tree-vect-loop.c: Likewise.
6499 * tree-vect-slp.c: Likewise.
6500 * tree-vect-stmts.c: Likewise.
6501 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
6503 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
6504 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
6505 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
6506 (aarch64_vectorize_vec_perm_const): ...this new function.
6507 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
6508 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
6509 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
6510 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
6511 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
6512 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
6513 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
6515 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
6516 check for NEON modes.
6517 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
6518 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
6519 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
6520 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
6522 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
6523 the old VEC_PERM_CONST conditions.
6524 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
6525 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
6526 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
6527 (ia64_vectorize_vec_perm_const_ok): Merge into...
6528 (ia64_vectorize_vec_perm_const): ...this new function.
6529 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
6530 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
6531 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
6532 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
6533 * config/mips/mips.c (mips_expand_vec_perm_const)
6534 (mips_vectorize_vec_perm_const_ok): Merge into...
6535 (mips_vectorize_vec_perm_const): ...this new function.
6536 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
6537 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
6538 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
6539 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
6540 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
6541 (rs6000_expand_vec_perm_const): Delete.
6542 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
6544 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
6545 (altivec_expand_vec_perm_const_le): Take each operand individually.
6546 Operate on constant selectors rather than rtxes.
6547 (altivec_expand_vec_perm_const): Likewise. Update call to
6548 altivec_expand_vec_perm_const_le.
6549 (rs6000_expand_vec_perm_const): Delete.
6550 (rs6000_vectorize_vec_perm_const_ok): Delete.
6551 (rs6000_vectorize_vec_perm_const): New function.
6552 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
6553 an element count and rtx array.
6554 (rs6000_expand_extract_even): Update call accordingly.
6555 (rs6000_expand_interleave): Likewise.
6556 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
6557 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
6558 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
6559 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
6560 (rs6000_expand_vec_perm_const): Delete.
6561 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
6562 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
6563 (altivec_expand_vec_perm_const_le): Take each operand individually.
6564 Operate on constant selectors rather than rtxes.
6565 (altivec_expand_vec_perm_const): Likewise. Update call to
6566 altivec_expand_vec_perm_const_le.
6567 (rs6000_expand_vec_perm_const): Delete.
6568 (rs6000_vectorize_vec_perm_const_ok): Delete.
6569 (rs6000_vectorize_vec_perm_const): New function. Remove stray
6570 reference to the SPE evmerge intructions.
6571 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
6572 an element count and rtx array.
6573 (rs6000_expand_extract_even): Update call accordingly.
6574 (rs6000_expand_interleave): Likewise.
6575 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
6576 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
6578 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
6580 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6582 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
6583 vector mode and that that mode matches the mode of the data
6585 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
6586 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
6587 directly using expand_vec_perm_1 when forcing selectors into
6589 (expand_vec_perm_var): New function, split out from expand_vec_perm.
6591 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6593 * optabs-query.h (can_vec_perm_p): Delete.
6594 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
6595 * optabs-query.c (can_vec_perm_p): Split into...
6596 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
6597 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
6598 particular selector is valid.
6599 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6600 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
6601 (vect_grouped_load_supported): Likewise.
6602 (vect_shift_permute_load_chain): Likewise.
6603 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
6604 (vect_transform_slp_perm_load): Likewise.
6605 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6606 (vectorizable_bswap): Likewise.
6607 (vect_gen_perm_mask_checked): Likewise.
6608 * fold-const.c (fold_ternary_loc): Likewise. Don't take
6609 implementations of variable permutation vectors into account
6610 when deciding which selector to use.
6611 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
6612 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
6613 with a false third argument.
6614 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
6615 to test whether the constant selector is valid and can_vec_perm_var_p
6616 to test whether a variable selector is valid.
6618 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6620 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
6621 * optabs-query.c (can_vec_perm_p): Likewise.
6622 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
6623 instead of vec_perm_indices.
6624 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
6625 (vect_gen_perm_mask_checked): Likewise,
6626 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
6627 (vect_gen_perm_mask_checked): Likewise,
6629 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6631 * optabs-query.h (qimode_for_vec_perm): Declare.
6632 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
6633 (qimode_for_vec_perm): ...this new function.
6634 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
6636 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6638 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
6639 does not have a conditional at the top.
6641 2018-01-02 Richard Biener <rguenther@suse.de>
6643 * ipa-inline.c (big_speedup_p): Fix expression.
6645 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
6648 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
6651 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
6655 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
6656 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
6657 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
6658 cond_taken_branch_cost 3->4.
6660 2018-01-01 Jakub Jelinek <jakub@redhat.com>
6662 PR tree-optimization/83581
6663 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
6664 TODO_cleanup_cfg if any changes have been made.
6667 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
6668 convert_modes if target mode has the right side, but different mode
6672 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
6673 last argument when extracting from CONCAT. If either from_real or
6674 from_imag is NULL, use expansion through memory. If result is not
6675 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
6676 the parts directly to inner mode, if even that fails, use expansion
6680 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
6681 check for bswap in mode rather than HImode and use that in expand_unop
6684 Copyright (C) 2018 Free Software Foundation, Inc.
6686 Copying and distribution of this file, with or without modification,
6687 are permitted in any medium without royalty provided the copyright
6688 notice and this notice are preserved.