1 2011-11-02 Richard Guenther <rguenther@suse.de>
3 PR tree-optimization/50902
4 * tree-vect-stmts.c (vectorizable_load): Properly convert
5 an invariant initializer element.
7 2010-11-02 Richard Guenther <rguenther@suse.de>
9 PR tree-optimization/50890
10 * gimple.h (gimple_fold_call): Remove.
11 * gimple-fold.c (fold_stmt_1): Move all call related code to ...
12 (gimple_fold_call): ... here. Make static. Update the
13 cannot-inline flag on direct calls.
14 * ipa-inline.c (early_inliner): Copy the cannot-inline flag
15 from the statements to the edges.
17 2011-11-01 Ian Lance Taylor <iant@google.com>
19 * godump.c (struct macro_hash_value): Define.
20 (macro_hash_hashval): New static function.
21 (macro_hash_eq, macro_hash_del): New static functions.
22 (go_define): Use macro_hash_value to store values in macro_hash.
23 Replace an old value on a redefinition. Don't print anything to
25 (go_undef): Delete the entry from the hash table.
26 (go_output_typedef): For an enum, use macro_hash_value, and don't
27 print anything to go_dump_file.
28 (go_print_macro): New static function.
29 (go_finish): Traverse macro_hash with go_print_macro.
30 (dump_go_spec_init): Update macro_hash creation for
33 2011-11-02 Alan Modra <amodra@gmail.com>
35 * config/rs6000/rs6000.c (rs6000_code_end): Declare ATTRIBUTE_UNUSED.
37 2011-11-01 Paolo Carlini <paolo.carlini@oracle.com>
40 * doc/invoke.texi: Document -Wzero-as-null-pointer-constant.
42 2011-11-01 Andrew Stubbs <ams@codesourcery.com>
44 * config/arm/bpabi.h (BE8_LINK_SPEC): Recognize generic-armv7 tuning.
46 2011-11-01 Uros Bizjak <ubizjak@gmail.com>
48 * config/i386/i386.md (splitters for int-float conversion): Use
49 SUBREG_REG on SUBREGs in splitter constraints.
51 2011-11-01 Jakub Jelinek <jakub@redhat.com>
53 * config/i386/i386-protos.h (ix86_expand_adjust_ufix_to_sfix_si): New
55 * config/i386/i386.c (ix86_expand_adjust_ufix_to_sfix_si): New
57 * config/i386/sse.md (fixuns_trunc<mode><sseintvecmodelower>2): Use it.
58 (ssepackfltmode): New mode attr.
59 (vec_pack_ufix_trunc_<mode>): New expander.
61 2011-11-01 Uros Bizjak <ubizjak@gmail.com>
64 * config/i386/i386.md (floatsi<mode>2_vector_sse_with_temp splitter):
65 Compare <ssevecmode>mode with V4SFmode, not V4SImode.
67 2011-11-01 Peter Bergner <bergner@vnet.ibm.com>
69 * config.gcc (powerpc*-*-linux*): Add powerpc*-*-linux*ppc476* variant.
70 * config/rs6000/476.h: New file.
71 * config/rs6000/476.opt: Likewise.
72 * config/rs6000/rs6000.h (TARGET_LINK_STACK): New define.
73 (SET_TARGET_LINK_STACK): Likewise.
74 (TARGET_ASM_CODE_END): Define.
75 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
76 TARGET_LINK_STACK for -mtune=476 and -mtune=476fp.
77 (rs6000_legitimize_tls_address): Emit the link stack preserving GOT
78 code if TARGET_LINK_STACK.
79 (rs6000_emit_load_toc_table): Likewise.
80 (output_function_profiler): Likewise
81 (macho_branch_islands): Likewise
82 (machopic_output_stub): Likewise
83 (get_ppc476_thunk_name): New function.
84 (rs6000_code_end): Likewise.
85 * config/rs6000/rs6000.md (load_toc_v4_PIC_1, load_toc_v4_PIC_1b):
86 Convert to a define_expand.
87 (load_toc_v4_PIC_1_normal): New define_insn.
88 (load_toc_v4_PIC_1_476): Likewise.
89 (load_toc_v4_PIC_1b_normal): Likewise.
90 (load_toc_v4_PIC_1b_476): Likewise.
92 2011-11-01 Georg-Johann Lay <avr@gjlay.de>
95 * config/avr/avr.opt (-mbranch-cost=): New option.
96 * config/avr/avr.h (BRANCH_COST): Define to avr_branch_cost.
97 * config/avr/avr.c (avr_rtx_costs_1): Adjust [U]DIV/[U]MOD costs.
98 * config/avr/avr.md (*addqi3.lt0, *addhi3.lt0, *addsi3.lt0): New insns.
99 (*addhi3_zero_extend1): Remove % in constraint of operand 1.
100 (*addhi3.sign_extend1, *subhi3.sign_extend2): New insns.
102 2011-11-01 Tom de Vries <tom@codesourcery.com>
104 PR tree-optimization/50908
105 * tree-ssa-tail-merge.c (update_vuses): Now that edges are removed
106 before update_vuses, test for 1 predecessor rather than two.
107 (delete_block_update_dominator_info): New function, part of it factored
109 (replace_block_by): Use delete_block_update_dominator_info. Call
110 update_vuses after deleting bb1 and updating dominator info, instead of
113 2011-11-01 David S. Miller <davem@davemloft.net>
115 * config/sparc/sparc.c (vector_init_faligndata): New function.
116 (sparc_expand_vector_init): Use it for V4HImode on VIS1.
118 * config/sparc/sparc.c (sparc_expand_vcond): New function.
119 * config/sparc/sparc-protos.h (sparc_expand_vcond): Declare it.
120 * config/sparc/sparc.md (vcond<mode><mode>): New VIS3 expander.
121 (vconduv8qiv8qi): Likewise.
123 2011-11-01 Alexandre Oliva <aoliva@redhat.com>
126 * cselib.c (cfa_base_preserved_regno): Initialize.
127 (cselib_expand_value_rtx_1): Don't expand it.
128 * var-tracking.c (vt_expand_var_loc_chain): Initialize depth.
129 Check it's only zero if result is NULL.
131 2011-11-01 Jakub Jelinek <jakub@redhat.com>
133 * config/i386/sse.md (fixuns_trunc<mode><sseintvecmodelower>2): New
136 * config/i386/sse.md (sseintvecmode): Remove duplicate modes.
137 (sseintvecmodelower): New mode iterator.
138 (floatv8siv8sf2, floatunsv4siv4sf2): Macroize into...
139 (float<sseintvecmodelower><mode>2): ... this using VF1 iterator.
140 (floatunsv4siv4sf2): Macroize into...
141 (floatuns<sseintvecmodelower><mode>2): ... this using VF1 iterator.
143 2011-10-31 David S. Miller <davem@davemloft.net>
145 * config/sparc/sparc.md (cmask patterns): Allow zero operand.
147 * dwarf2out.c (cached_next_real_insn): New.
148 (dwarf2out_end_epilogue): Set it to NULL_RTX.
149 (dwarf2out_var_location): Remove cached_next_real_insn local static.
151 2011-10-31 Richard Henderson <rth@redhat.com>
153 * config/i386/sse.md (floatv8siv8sf2): Rename from avx_cvtdq2ps256.
154 (floatv4siv4sf2): Rename from sse2_cvtdq2ps.
155 (floatunsv4siv4sf2): Rename from sse2_cvtudq2ps.
156 (fix_truncv8sfv8si2): Rename from avx_cvttps2dq256.
157 (fix_truncv4sfv4si2): Rename from sse2_cvttps2dq.
158 (floatv4siv4df2): Rename from avx_cvtdq2pd256.
159 (fix_truncv4dfv4si2): Rename from avx_cvttpd2dq256.
160 (vec_unpacku_float_hi_v8si): Update for insn pattern name changes.
161 * config/i386/i386.md (splitters for int-float conversion): Likewise.
162 * config/i386/i386.c (ix86_split_convert_uns_si_sse): Likewise.
163 (bdesc_args): Likewise.
164 (enum ix86_builtins) [IX86_BUILTIN_CVTUDQ2PS]: Remove.
165 (ix86_vectorize_builtin_conversion): Remove.
166 (TARGET_VECTORIZE_BUILTIN_CONVERSION): Remove.
168 2011-10-31 Martin Jambor <mjambor@suse.cz>
170 * ipa-prop.c (mark_modified): Moved up in the file.
171 (is_parm_modified_before_call): Renamed to
172 is_parm_modified_before_stmt, moved up in the file.
173 (load_from_unmodified_param): New function.
174 (compute_complex_assign_jump_func): Also attempt to create pass
175 through jump functions for values loaded from (addressable)
178 2011-10-31 Jakub Jelinek <jakub@redhat.com>
180 * tree-vect-stmts.c (vectorizable_shift): If op1 is vect_external_def
181 in a loop and has different type from op0, cast it to op0's type
182 before the loop first. For slp give up. Don't crash if op1_vectype
185 2011-10-31 Paul Brook <paul@codesourcery.com>
187 * cgraphunit.c: Don't mark clones as static constructors.
189 2011-10-31 David Edelsohn <dje.gcc@gmail.com>
191 * gcc-ar: Do not include stdio.h.
193 2011-10-31 Diego Novillo <dnovillo@google.com>
195 * tree-streamer-out.c (pack_ts_base_value_fields): Emit
197 * tree-streamer-in.c (unpack_ts_base_value_fields): Read
200 2011-10-30 David S. Miller <davem@davemloft.net>
202 * config/sparc/sparc.c (vector_init_bshuffle): New function.
203 (vector_init_fpmerge): New function.
204 (sparc_expand_vector_init): Use them to improve non-const cases.
206 * dwarf2out.c (dwarf2out_var_location): When processing several
207 consecutive location notes, cache the result of next_real_insn().
209 2011-10-30 Uros Bizjak <ubizjak@gmail.com>
211 * config/i386/i386.md (avx2_vec_dup<mode>): Macroize insn from
212 avx2_vec_dup{v8sf,v4sf} using VF1 mode iterator.
213 (vec_dupv4sf): Remove expander.
214 (vec_dupv4sf): Merge from *vec_dupv4sf and *vec_dupv4sf_avx.
215 (vec_dupv2df): Remove expander.
216 (vec_dupv2df): Merge from *vec_dupv2df and *vec_dupv2df_sse3.
217 (*vec_concatv2df): Merge *vec_concatv2df_sse3.
218 (*vec_dupv4si): Merge *vec_dupv4si_avx.
219 (*vec_dupv2di): Merge *vec_dupv2di_sse3.
221 2011-10-30 Dmitry Plotnikov <dplotnikov@ispras.ru>
223 * tree-cfg.c (verify_gimple_assign_unary): Allow vector conversions.
224 * optabs.c (supportable_convert_operation): New function.
225 * optabs.h (supportable_convert_operation): New prototype.
226 * tree-vect-stmts.c (vectorizable_conversion): Change condition and
227 behavior for NONE modifier case.
228 * tree.h (VECTOR_INTEGER_TYPE_P): New macro.
230 2011-10-30 Jakub Jelinek <jakub@redhat.com>
232 * tree-vectorizer.h (NUM_PATTERNS): Bump to 9.
233 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern): New
235 (vect_vect_recog_func_ptrs): Add it.
237 2011-10-30 David S. Miller <davem@davemloft.net>
239 * reorg.c (label_before_next_insn): New function.
240 (relax_delay_slots): Use it instead of prev_label.
241 * rtl.h (prev_label): Delete declaration.
242 * emit-rtl.c (prev_label): Remove.
244 2011-10-30 Revital Eres <revital.eres@linaro.org>
246 * modulo-sched.c (generate_prolog_epilog): Mark prolog and epilog
247 as BB_DISABLE_SCHEDULE.
248 (mark_loop_unsched): New function.
249 (sms_schedule): Call it.
251 2011-10-29 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
254 * config/pa/protos.h (attr_length_save_restore_dltp): Delete.
255 (cmpib_comparison_operator): Likewise.
256 (following_cal, output_and, output_ior, output_move_double,
257 output_fp_move_double, output_block_move, output_block_clear,
258 output_cbranch, output_lbranch, output_bb, output_bvb, output_dbra,
259 output_movb, output_parallel_movb, output_parallel_addb, output_call,
260 output_indirect_call, output_millicode_call, output_mul_insn,
261 output_div_insn, output_mod_insn, singlemove_string,
262 output_arg_descriptor, output_global_address, print_operand,
263 legitimize_pic_address, hppa_encode_label, symbolic_expression_p,
264 fmpyaddoperands, fmpysuboperands, emit_bcond_fp, emit_move_sequence,
265 emit_hpdiv_const, is_function_label_plus_const, jump_in_call_delay,
266 hppa_fpstore_bypass_p, attr_length_millicode_call, attr_length_call,
267 attr_length_indirect_call, return_addr_rtx, function_arg_padding,
268 insn_refs_are_delayed, get_deferred_plabel, ldil_cint_p, zdepi_cint_p,
269 output_ascii, compute_frame_size, and_mask_p, cint_ok_for_move,
270 hppa_expand_prologue, hppa_expand_epilogue, ior_mask_p,
271 compute_zdepdi_operands, output_64bit_and, output_64bit_ior,
272 reloc_needed, magic_milli, shadd_constant_p): Consistently prefix
273 exported functions and variables with "pa_".
274 * config/pa/predicates.md: Likewise.
275 * config/pa/pa64-hpux.h: likewise.
276 * config/pa/som.h: Likewise.
277 * config/pa/elf.h: Likewise.
278 * config/pa/pa64-linux.h: Likewise.
279 * config/pa/pa.md: Likewise.
280 * config/pa/pa.c: Likewise.
281 * config/pa/pa-linux.h: Likewise.
282 * config/pa/pa.h: Likewise.
283 * config/pa/constraints.md: Likewise.
285 2011-10-29 Uros Bizjak <ubizjak@gmail.com>
287 * config/i386/i386.md (xop_sha<mode>3): Rename from xop_ashl<mode>3.
289 (xop_shl<mode>3): Rename from xop_lshl<mode>3. Update all uses.
290 * config/i386/i386.c: Update all uses.
292 2011-10-29 Uros Bizjak <ubizjak@gmail.com>
294 * config/i386/i386.md (lshlv16qi3): Remove expander.
295 (lshrv16qi3): New expander.
296 (<shift_insn>v16qi3): Macroize expander from ashrv16qi3 and lshrv16qi3
297 using any_shiftrt code iterator. Cleanup.
298 (ashlv16qi3): Cleanup.
301 2011-10-29 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
304 * config/pa/pa.c (emit_move_sequence): Legitimize TLS symbol
306 (pa_legitimate_constant_p): Return false for TLS_MODEL_GLOBAL_DYNAMIC
307 and TLS_MODEL_LOCAL_DYNAMIC symbol references.
309 2011-10-29 Georg-Johann Lay <avr@gjlay.de>
312 * config/avr/avr.opt (-maccumulate-args): New option.
313 * config/avr/avr.h (STARTING_FRAME_OFFSET): Redefine to
314 avr_starting_frame_offset.
315 (ACCUMULATE_OUTGOING_ARGS): Define to avr_accumulate_outgoing_args.
316 * config/avr/avr.md (UNSPECV_WRITE_SP_IRQ_ON): Remove.
317 (UNSPECV_WRITE_SP_IRQ_OFF): Remove.
318 (UNSPECV_WRITE_SP): New constant.
319 (*addhi3_sp_R): Rewrite to...
320 (*addhi3_sp): ...this new insn.
321 (movhi_sp_r_irq_off, movhi_sp_r_irq_on): Combine to...
322 (movhi_sp_r): ...this new insn.
323 * config/avr/avr-protos.h (avr_accumulate_outgoing_args): New.
324 (avr_starting_frame_offset): New.
325 * config/avr/avr.c (avr_accumulate_outgoing_args): New function.
326 (avr_starting_frame_offset): New function.
327 (avr_outgoing_args_size): New static function.
328 (avr_initial_elimination_offset): Use it.
329 (avr_simple_epilogue): Use it.
330 (avr_asm_function_end_prologue): Use it.
331 (expand_epilogue): Use it.
332 (expand_prologue): Use it. Break out code to...
333 (avr_prologue_setup_frame): ...this new static function.
334 (avr_can_eliminate): Allow eliminating to frame pointer if there
336 (avr_frame_pointer_required_p): Use frame pointer if target has a
338 * config/avr/constraints.md (R): Remove.
339 (Csp): New constraint.
340 * config/avr/predicates.md (avr_sp_immediate_operand): Use it.
342 2011-10-29 Andi Kleen <ak@linux.intel.com>
344 * gcc-ar.c (target_machine): Add.
346 2011-10-29 Anatoly Sokolov <aesok@post.ru>
348 * config/cris/cris.c (reg_ok_for_base_p, reg_ok_for_index_p,
349 cris_constant_index_p, cris_base_p, cris_index_p,
350 cris_base_or_autoincr_p, cris_bdap_index_p, cris_biap_index_p,
351 cris_legitimate_address_p): New functions.
352 (TARGET_LEGITIMATE_ADDRESS_P): Define.
353 (cris_pic_symbol_type, cris_valid_pic_const): Change arguments type
354 from rtx to const_rtx.
355 (cris_print_operand_address, cris_address_cost,
356 cris_side_effect_mode_ok): Use
357 cris_constant_index_p, cris_base_p, cris_base_or_autoincr_p,
358 cris_biap_index_p and cris_bdap_index_p.
359 * config/cris/cris.h (CONSTANT_INDEX_P, BASE_P, BASE_OR_AUTOINCR_P,
360 BDAP_INDEX_P, BIAP_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
361 REG_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P): Remove.
362 (EXTRA_CONSTRAINT_Q, EXTRA_CONSTRAINT_R, EXTRA_CONSTRAINT_T): Use
363 cris_constant_index_p, cris_base_p, cris_base_or_autoincr_p,
364 cris_biap_index_p and cris_bdap_index_p.
365 * config/cris/cris.md (moversideqi movemsideqi peephole2): Use
367 * config/cris/cris-protos.h (cris_constant_index_p, cris_base_p,
368 cris_base_or_autoincr_p, cris_bdap_index_p, cris_biap_index_p): New
370 (cris_pic_symbol_type, cris_valid_pic_const): Update prototype.
372 2011-10-21 Andi Kleen <ak@linux.intel.com>
374 * ggc-page (PAGE_ALIGN): Add.
375 (alloc_page, ggc_pch_total_size, ggc_pch_this_base, ggc_pch_read):
376 Replace ROUND_UP with PAGE_ALIGN.
378 2011-10-20 Andi Kleen <ak@linux.intel.com>
380 * ggc-page (alloc_anon): Add check argument.
381 (alloc_page): Add fallback to 1 page allocation.
382 Adjust alloc_anon calls to new argument.
384 2011-10-18 Andi Kleen <ak@linux.intel.com>
386 * ggc-page (release_pages): First free large continuous
387 chunks in the madvise path.
389 2011-10-18 Andi Kleen <ak@linux.intel.com>
391 * ggc-page.c (alloc_pages): Always round up entry_size.
393 2011-10-19 Andi Kleen <ak@linux.intel.com>
395 * Makefile.in (MOSTLYCLEANFILES): Add gcc-ar/nm/ranlib.
396 (native): Add gcc-ar, gcc-nm, gcc-ranlib.
397 (AR_LIBS, gcc-ar, gcc-ar.o, gcc-ranlib, gcc-ranlib.o,
398 gcc-nm, gcc-nm.o, gcc-ranlib.c, gcc-nm.c): Add.
399 (install): Depend on install-gcc-ar.
400 (install-gcc-ar): Add.
401 (uninstall): Uninstall gcc-ar, gcc-nm, gcc-ranlib.
402 * gcc-ar.c: Add new file.
404 2011-10-28 Pat Haugen <pthaugen@us.ibm.com>
406 * config/rs6000/rs6000.md (define_attr "type"): Add vecdouble.
407 * config/rs6000/vsx.md (VStype_simple, VStype_mul): Use vecdouble
409 (VStype_div): Use vector types for V2DF/V4SF.
410 (VStype_sqrt): Use *sqrt types.
411 (VS_spdp_type): Change type to vecdouble.
412 (*vsx_fmav2df4, *vsx_nfmsv2df4, vsx_xvcvdpsxws, vsx_xvcvdpuxws,
413 vsx_xvcvuxdsp, vsx_xvcvsxwdp, vsx_xvcvuxwdp, vsx_xvcvspsxds,
414 vsx_xvcvspuxds): Likewise.
415 (*vsx_fms<mode>4): Set type via <VStype_mul>.
416 (*vsx_eq_<mode>_p, *vsx_gt_<mode>_p, *vsx_ge_<mode>_p): Set type via
418 * config/rs6000/power7.md (power7-vecstore): Correct VSU pipe.
419 (power7-fpcompare, power7-sdiv, power7-ddiv, power7-sqrt,
420 power7-dsqrt): Correct insn latency.
421 (power7-vecsimple): Add veccmp type and correct dispatch/VSU values.
422 (power7-veccmp): Delete.
423 (power7-vecfloat): Correct latency/dispatch/VSU values.
424 (define_bypass "power7-vecfloat"): Correct latency and types.
425 (power7-veccomplex, power7-vecperm): Correct dispatch/VSU values.
426 (power7-vecdouble, power7-vecfdiv, power7-vecdiv): New.
428 2011-10-28 Uros Bizjak <ubizjak@gmail.com>
430 * config/i386/i386.md (shift_insn): Rename code attribute from
431 shiftrt_insn. Also handle ashift RTX.
432 (shift): Rename code attribute from shiftrt. Also handle ashift RTX.
433 (vshift): New code attribute.
434 (<shift_insn>*): Rename from <shiftrt_insn>*. Update asm templates.
435 (any_lshift): Move and rename code iterator from ...
436 * config/i386/sse.md (lshift): ... here.
437 (lshift_insn): Remove code attribute.
438 (lshift): Remove code attribute.
439 (vlshr<mode>3): Use lshiftrt RTX.
440 (vashr<mode>3, ashrv16qi3, ashrv2di3): Use ashiftrt RTX.
441 (vashl<mode>3, ashlv16qi3): Use ashift RTX.
442 (avx2_<lshift>v<mode>): Rename from avx2_<shift_insn>v<mode>. Use
443 any_lshift code iterator. Update asm template.
444 (<shift_insn><mode>3): Macroize insn from lshr<mode>3 and ashl<mode>3
445 usign any_lshift code iterator.
446 * config/i386/mmx.md (mmx_<shift_insn><mode>3): Macroize insn from
447 mmx_lshr<mode>3 and mmx_ashl<mode>3 usign any_lshift code iterator.
448 * config/i386/i386.c (bdesc_args) <__builtin_ia32_psll>: Update.
450 2011-10-28 Georg-Johann Lay <avr@gjlay.de>
453 * config/avr/avr.md (parityhi2): Expand allowing pseudos.
454 (*parityhi2): New pre-reload insn-and-split to map 16-bit parity
456 (*parityqihi2): Same for 8-bit parity.
458 2011-10-28 Julian Brown <julian@codesourcery.com>
460 PR rtl-optimization/47918
461 * reload1.c (set_initial_label_offsets): Use initial offsets
462 for labels on the nonlocal_goto_handler_labels chain.
464 2011-10-28 Iain Sandoe <iains@gcc.gnu.org>
466 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA):
467 Move darwin-fpsave.asm from here to ... LIB2FUNCS_EXTRA.
468 (LIB2FUNCS_EXTRA): Add darwin-gpsave.asm.
469 (TARGET_LIBGCC2_CFLAGS): Ensure that fPIC and -pipe are inherited from
471 * config/rs6000/darwin.h (FP_SAVE_INLINE): Adjust to enable.
472 (GP_SAVE_INLINE): Likewise.
473 (SAVE_FP_PREFIX, SAVE_FP_SUFFIX, RESTORE_FP_PREFIX,
474 RESTORE_FP_SUFFIX): Set to empty strings.
475 * config/rs6000/rs6000.c (rs6000_savres_strategy): Implement for Darwin.
476 (debug_stack_info): Print savres_strategy.
477 (rs6000_savres_routine_name): Implement for Darwin.
478 (rs6000_make_savres_rtx): Adjust used register for Darwin.
479 (rs6000_emit_prologue): Implement out-of-line saves for Darwin.
480 (rs6000_output_function_prologue): Don't emit .extern for Mach-O.
481 (rs6000_emit_epilogue): Implement out-of-line saves for Darwin.
482 * config/rs6000/darwin-gpsave.asm: New file.
484 2011-10-28 Jakub Jelinek <jakub@redhat.com>
486 * config/i386/sse.md (VI4SD_AVX2): Removed.
487 (VI48_AVX2, VI128_128, VI48_128, VI48_256): New mode iterators.
488 (vashl<mode>3): Use VI12_128 iterator instead of VI124_128.
489 Add another expander using VI48_128 iterator for
490 TARGET_AVX2 || TARGET_XOP and another using VI48_256 iterator
492 (vlshr<mode>3): Likewise. Change register_operand predicate to
493 nonimmediate_operand on last operand in the VI12_128 expander.
494 (vashr<mode>3): Use VI128_128 iterator instead of VI124_128.
495 (vashrv4si3, vashrv8si3): New expanders.
496 (avx2_ashrvv8si, avx2_ashrvv4si, avx2_<lshift>vv8si,
497 avx2_<lshift>vv2di): Removed.
498 (avx2_ashrv<mode>): New insn with VI4_AVX2 iterator.
499 (avx2_<lshift>v<mode>): Macroize using VI48_AVX2
500 iterator. Simplify pattern.
502 2010-10-28 Richard Guenther <rguenther@suse.de>
505 * lto-wrapper.c (get_options_from_collect_gcc_options):
506 Properly count arguments.
507 (run_gcc): Use an obstack to collect argv, properly separate
508 switches and their arguments.
510 2011-10-28 Jakub Jelinek <jakub@redhat.com>
512 * tree-vect-stmts.c (vectorizable_shift): Give up if op1 has different
513 vector mode from vectype's mode.
515 2011-10-28 Chung-Lin Tang <cltang@codesourcery.com>
517 PR rtl-optimization/49720
518 * simplify-rtx.c (simplify_relational_operation_1): Detect
519 infinite recursion condition in "(eq/ne (plus x cst1) cst2)
520 simplifies to (eq/ne x (cst2 - cst1))" case.
522 2011-10-27 David S. Miller <davem@davemloft.net>
524 * config/sparc/sparc.md (snedi_special): Only match when not VIS3.
525 (*snedi_zero): Likewise.
526 (*snedi_zero_trunc): Likewise.
527 (snedi_special_vis3): New expander.
528 (*snedi_zero_vis3): New insn.
529 (*snedi_zero_trunc_vis3): Likewise.
530 (*sltu_insn_vis3): Likewise.
531 (*sltu_insn_vis3_trunc): Likewise.
533 (*addxc_trunc_sp64_vis3): Likewise.
534 * config/sparc/sparc.c (emit_scc_insn): When VIS3 use the
535 gen_snedi_special_vis3 expander, and try GTU/LTU addx based
536 sequences on DImode values.
538 * config/sparc/sparc.md (64-bit vector moves): Use 'e' not 'f'
541 * regcprop.c (copyprop_hardreg_forward_1): Reject the
542 transformation when we narrow the mode on big endian.
544 2011-10-27 Jakub Jelinek <jakub@redhat.com>
546 * config/i386/sse.md (avx_cvtpd2dq256_2, avx_cvttpd2dq256_2,
547 vec_pack_sfix_trunc_v4df, vec_pack_sfix_v4df): New expanders.
548 (*avx_cvtpd2dq256_2, *avx_cvttpd2dq256_2): New insns.
550 * config/i386/i386.c (ix86_print_operand): Handle 'q' and 'x'
551 overrides for -masm=intel memory.
552 * config/i386/sse.md (sse2_cvtdq2pd, sse2_cvtps2pd,
553 sse4_1_<code>v8qiv8hi2, avx2_<code>v8qiv8si2,
554 sse4_1_<code>v4hiv4si2, avx2_<code>v4hiv4di2,
555 sse4_1_<code>v2siv2di2): Use %q1 instead of %1 for -masm=intel.
556 (sse4_1_<code>v4qiv4si2, avx2_<code>v4qiv4di2,
557 sse4_1_<code>v2hiv2di2): Use %k1 instead of %1 for -masm=intel.
558 (sse4_1_<code>v2qiv2di2): Use %w1 instead of %1 for -masm=intel.
560 2011-10-27 Martin Jambor <mjambor@suse.cz>
562 * ipa-prop.c (compute_pass_through_member_ptrs): Rename parm_info
564 (ipa_compute_jump_functions_for_edge): Likewise.
565 (ipa_compute_jump_functions): Likewise.
566 (ipa_analyze_indirect_call_uses): Likewise.
567 (ipa_analyze_call_uses): Likewise.
568 (ipa_analyze_params_uses): Likewise.
569 (ipa_analyze_node): Likewise.
571 2011-10-27 Uros Bizjak <ubizjak@gmail.com>
574 * config/i386/sse.md (*avx_unpcklpd256): Remove extra insn
575 constraints. Change alternative 1 to "x,m,1".
577 2011-10-27 Jakub Jelinek <jakub@redhat.com>
579 * Makefile.in (build/gencheck.o): Depend on tree.def and
580 c-family/c-common.def.
582 * tree-ssa-strlen.c: Include expr.h.
583 (get_stridx): Don't use c_strlen, instead use string_constant
584 and compute string length from it.
585 * Makefile.in (tree-ssa-strlen.o): Depend on $(EXPR_H).
587 2011-10-27 Eric Botcazou <ebotcazou@adacore.com>
589 PR rtl-optimization/46603
591 * reload.c (push_reload): In the out case, restore previous behavior
592 for subregs that don't have word mode.
594 2011-10-27 Ian Lance Taylor <iant@google.com>
596 * cppdefault.c: Undef NATIVE_SYSTEM_HEADER_DIR if
597 CROSS_DIRECTORY_STRUCTURE is defined and TARGET_SYSTEM_ROOT is not.
598 (cpp_include_defaults): Only use NATIVE_SYSTEM_HEADER_DIR if it is
601 2011-10-27 Richard Henderson <rth@redhat.com>
603 * optabs.c (expand_vec_perm): Use the correct mode for scaling the
604 selector. Save the qimode constant selector for later use by the
605 qimode vec_perm pattern.
607 2011-10-27 Bernd Schmidt <bernds@codesourcery.com>
609 * config/c6x/c6x.c (unit_req_imbalance, res_mii): Cast the first arg
610 to unit_req_factor to the right enum type.
611 (get_unit_operand_masks, reshuffle_units, try_rename_operands,
612 hwloop_optimize): Remove unused variables.
614 2010-10-27 Richard Guenther <rguenther@suse.de>
617 * tree-vect-generic.c (do_binop): Handle scalar operands.
619 2011-08-27 Uros Bizjak <ubizjak@gmail.com>
622 * config/i386/sse.md (*vec_extract_v4sf_mem): Avoid combining registers
623 from different units in a single alternative.
625 2011-10-26 David S. Miller <davem@davemloft.net>
627 * config/sparc/sparc.c (emit_scc_insn): Force attempt of v9 sequences
628 if we're comparing DImode and comparison is other than EQ or NE.
630 * config/sparc/sparc.c (emit_scc_insn): Do not try v9 sequences until
631 LEU/LTU/GEU/GTU is attempted.
632 * config/sparc/sparc.md (*neg_snesi_sign_extend): New 64-bit insn
634 (*neg_seqsi_sign_extend): Likewise.
635 (*sltu_extend_sp64, *neg_sltu_extend_sp64, *sgeu_extend_sp64,
636 *neg_sgeu_extend_sp64): New insns.
638 * config/sparc/sparc-protos.h (sparc_expand_conditional_move): Declare.
639 * config/sparc/sparc.md (mov<I:mode>cc, mov<F:mode>cc): Call it.
640 (*mov<I:mode>_cc_v9): Normalize to expect operand 0 always in operand 4.
641 (*mov<I:mode>_cc_reg_sp64): Likewise.
642 (*movsf_cc_v9): Likewise.
643 (*movsf_cc_reg_sp64): Likewise.
644 (*movdf_cc_v9): Likewise.
645 (*movdf_cc_reg_sp64): Likewise.
646 (*movtf_cc_hq_v9): Likewise.
647 (*movtf_cc_reg_hq_sp64): Likewise.
648 (*movtf_cc_v9): Likewise.
649 (*movtf_cc_reg_sp64): Likewise.
650 * config/sparc/sparc.c (sparc_expand_conditional_move): New function.
651 (sparc_print_operand): Delete 'c' and 'd' handling, no longer used.
653 2011-10-26 Eric Botcazou <ebotcazou@adacore.com>
655 * reload.c (reload_inner_reg_of_subreg): Change type of return value
656 and type of OUTPUT parameter to bool and adjust. Document MODE and
657 OUTPUT parameters. Use HARD_REGISTER_P. Reorder final condition
658 and improve associated comment.
659 (push_reload): Clarify and update comments about reloading of subregs.
660 Adjust calls to reload_inner_reg_of_subreg. Compute the class upfront
661 for the reloading of subregs in the out case as well.
663 2011-10-26 Alexandre Oliva <aoliva@redhat.com>
666 * var-tracking.c (rtx_debug_expr_p): New.
667 (use_type): Don't use debug exprs to track non-VTA variables.
669 2011-10-26 Jeff Law <law@redhat.com>
671 * doc/invoke.texi (sink-frequency-threshold): Document.
672 * tree-ssa-sink.c: Include params.h.
673 (select_best_block): New function.
674 (statement_sink_location): Use it.
675 * params.def (SINK_FREQUENCY_THRESHOLD): New PARAM.
677 2011-10-26 Iain Sandoe <iains@gcc.gnu.org>
680 * config/darwin.c (top level): Amend comments concerning LTO output.
681 (lto_section_num): New variable. (darwin_lto_section_e): New GTY.
682 (LTO_SECTS_SECTION, LTO_INDEX_SECTION): New.
683 (LTO_NAMES_SECTION): Rename.
684 (darwin_asm_named_section): Record LTO section counts and switches
685 in a vec of darwin_lto_section_e.
686 (darwin_file_start): Remove unused code.
687 (darwin_file_end): Put an LTO section termination label. Handle
688 output of the wrapped LTO sections, index and names table.
690 2011-10-26 Alan Modra <amodra@gmail.com>
692 * config/rs6000/rs6000.c (rs6000_make_savres_rtx): Delete unneeded
694 (rs6000_emit_stack_reset): Only return insn emitted when it adjusts sp.
695 (rs6000_make_savres_rtx): Rename to rs6000_emit_savres_rtx. Use
696 simple_return in pattern, emit instruction, and set jump_label.
697 (rs6000_emit_prologue): Update for rs6000_emit_savres_rtx. Use
698 simple_return rather than return.
699 (emit_cfa_restores): New function.
700 (rs6000_emit_epilogue): Emit cfa_restores when flag_shrink_wrap.
701 Add missing cfa_restores for SAVE_WORLD. Add missing LR cfa_restore
702 when using out-of-line gpr restore. Add missing LR and FP regs
703 cfa_restores for out-of-line fpr restore. Consolidate code setting
704 up cfa_restores. Formatting. Use LR_REGNO define.
705 (rs6000_output_mi_thunk): Use simple_return rather than return.
706 * config/rs6000/rs6000.md (sibcall*, sibcall_value*): Likewise.
707 (return_internal*): Likewise.
708 (any_return, return_pred, return_str): New iterators.
709 (return, conditional return insns): Provide both return and
710 simple_return variants.
711 * config/rs6000/rs6000.h (EARLY_R12, LATE_R12): Define.
712 (REG_ALLOC_ORDER): Move r12 before call-saved regs when FIXED_R13.
713 Move r11 and r0 later to suit shrink-wrapping.
715 2011-10-26 Richard Guenther <rguenther@suse.de>
717 * lto-wrapper.c (run_gcc): Properly init/free obstack.
719 2011-10-26 Jakub Jelinek <jakub@redhat.com>
721 * config/i386/i386.md (UNSPEC_VSIBADDR): New.
722 * config/i386/predicates.md (vsib_address_operand,
723 vsib_mem_operator): New predicates.
724 * config/i386/i386.c (ix86_print_operand_address): Handle
725 UNSPEC_VSIBADDR addresses.
726 * config/i386/sse.md (avx2_gathersi<mode>, avx2_gatherdi<mode>,
727 avx2_gatherdi<mode>256): Adjust expanders to use MEM with
728 UNSPEC_VSIBADDR address.
729 (*avx2_gathersi<mode>, *avx2_gatherdi<mode>, *avx2_gatherdi<mode>256):
730 Adjust insns to use MEM with UNSPEC_VSIBADDR address.
732 2011-10-26 Tom de Vries <tom@codesourcery.com>
734 PR tree-optimization/50763
735 * tree-ssa-tail-merge.c (replace_block_by): Update vops if phi_vuse1 or
736 phi_vuse2 is NULL_TREE only if bb1 dominates or is dominated by bb2.
738 2011-10-26 Richard Guenther <rguenther@suse.de>
741 * Makefile.in (lto-wrapper): Depend on and link against opts-common.o.
742 (lto-wrapper.o): Depend on $(OPTS_H) and $(OPTIONS_H).
743 * lto-wrapper.c (get_options_from_collect_gcc_options): New function.
744 (run_gcc): Use it. Filter out language specific options.
746 2011-10-26 Andreas Tobler <andreast@fgznet.ch>
748 * config/i386/freebsd64.h (LINK_SPEC): Emit the same warning as the
751 2011-10-25 Paolo Carlini <paolo.carlini@oracle.com>
754 * gcc.c (main): Fix fatal_error string for translation.
756 2011-10-25 Ian Lance Taylor <iant@google.com>
758 * tree-eh.c (do_return_redirection): Remove return_value_p
759 parameter. Change all callers.
760 (lower_try_finally_nofallthru): Remove local return_val.
761 (lower_try_finally_onedest): Likewise.
762 (lower_try_finally_copy): Likewise.
763 (lower_try_finally_switch): Likewise.
765 2011-10-25 H.J. Lu <hongjiu.lu@intel.com>
767 * config/i386/mmx.md (*mmx_maskmovq): Replace :SI with :P and
768 remove "&& !TARGET_64BIT"
769 (*mmx_maskmovq_rex): Removed.
771 2011-10-25 Eric Botcazou <ebotcazou@adacore.com>
773 PR rtl-optimization/46603
774 * reload.c (push_reload): In the out case, reload the subreg as well
775 as the reg if it has word mode.
777 2011-10-25 Eric Botcazou <ebotcazou@adacore.com>
779 * dwarf2out.c (add_gnat_descriptive_type_attribute): Temporarily
780 suppress debug info for the parent type.
782 2011-10-25 Eric Botcazou <ebotcazou@adacore.com>
784 * config/ia64/ia64.c (ia64_profile_hook): Fix thinko.
786 2011-10-25 Richard Henderson <rth@redhat.com>
788 * config/i386/sse.md (VEC_EXTRACT_EVENODD_MODE): Remove.
789 (vec_extract_even<mode>, vec_extract_odd<mode>): Remove.
791 * config/rs6000/altivec.md (vec_extract_evenv8hi,
792 vec_extract_evenv16qi, vec_extract_oddv4si,
793 vec_extract_oddv4sf): Remove.
795 * config/spu/spu.md (vec_extract_evenv4si, vec_extract_evenv4sf,
796 vec_extract_evenv8hi, vec_extract_evenv16qi, vec_extract_oddv4si,
797 vec_extract_oddv4sf, vec_extract_oddv8hi, vec_extract_oddv16qi,
798 vec_interleave_highv4sf, vec_interleave_lowv4sf,
799 vec_interleave_highv4si, vec_interleave_lowv4si,
800 vec_interleave_highv8hi, vec_interleave_lowv8hi,
801 vec_interleave_highv16qi, vec_interleave_lowv16qi): Remove.
803 * expr.c (expand_expr_real_2) [VEC_EXTRACT_EVEN_EXPR]: Use binop.
804 [VEC_EXTRACT_ODD_EXPR, VEC_INTERLEAVE_HIGH_EXPR]: Likewise.
805 [VEC_INTERLEAVE_LOW_EXPR]: Likewise.
806 * optabs.c (expand_binop): Implement vec_interleave_high_optab,
807 vec_interleave_low_optab, vec_extract_even_optab,
808 vec_extract_odd_optab with expand_vec_perm.
809 (can_vec_perm_for_code_p): New.
811 * tree-vect-data-refs.c (vect_strided_store_supported): Allow for
812 fallback via can_vec_perm_for_code_p.
813 (vect_strided_load_supported): Likewise.
814 * tree-vect-generic.c (expand_vector_operations_1): Never lower
815 VEC_INTERLEAVE_HIGH_EXPR, VEC_INTERLEAVE_LOW_EXPR,
816 VEC_EXTRACT_EVEN_EXPR, VEC_EXTRACT_ODD_EXPR.
818 * target.def (vec_perm_const_ok): Change parameters to mode and
820 * doc/tm.texi: Rebuild.
821 * config/i386/i386.c (ix86_vectorize_vec_perm_const_ok): Change
822 parameters to mode and array of indicies.
823 * expr.c (expand_expr_real_2) [VEC_PERM_EXPR]: Expand operands here.
824 * optabs.c (can_vec_perm_p): Rename from can_vec_perm_expr_p.
825 Change parameters to mode and array of indicies.
826 (expand_vec_perm_1): Rename from expand_vec_perm_expr_1.
827 (expand_vec_perm): Rename from expand_vec_perm_expr. Change
828 parameters to mode and rtx inputs. Try lowering to QImode
829 vec_perm_const before trying fully variable permutation.
830 * optabs.h: Update decls.
831 * tree-vect-generic.c (lower_vec_perm): Extract array of indices from
832 VECTOR_CST to pass to can_vec_perm_p.
833 * tree-vect-slp.c (vect_get_mask_element): Change mask parameter type
834 from int pointer to unsigned char pointer.
835 (vect_transform_slp_perm_load): Update for change to can_vec_perm_p.
836 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
838 * tree.def (VEC_EXTRACT_EVEN_EXPR): Fix typo in text name.
839 (VEC_EXTRACT_ODD_EXPR, VEC_INTERLEAVE_HIGH_EXPR,
840 VEC_INTERLEAVE_LOW_EXPR): Likewise.
842 2011-10-25 Mike Stump <mikestump@comcast.net>
844 * reload.c (regno_clobbered_p): Fix typo.
846 2011-10-25 Dodji Seketeli <dodji@redhat.com>
848 * input.c (expand_location): Rewrite using linemap_resolve_location
849 and linemap_expand_location. Add a comment.
851 2011-10-25 Jakub Jelinek <jakub@redhat.com>
853 PR tree-optimization/50596
854 * tree-vect-stmts.c (vect_mark_relevant): Only use
855 FOR_EACH_IMM_USE_FAST if lhs is SSA_NAME.
856 (vectorizable_store): If is_pattern_stmt_p look through
857 VIEW_CONVERT_EXPR on lhs.
858 * tree-vect-patterns.c (check_bool_pattern, adjust_bool_pattern):
859 Use unsigned type instead of signed.
860 (vect_recog_bool_pattern): Optimize also stores into bool memory in
861 addition to casts from bool to integral types.
862 (vect_mark_pattern_stmts): If pattern_stmt already has vinfo
863 created, don't create it again.
865 2011-10-25 Kai Tietz <ktietz@redhat.com>
867 * config/i386/i386.c (ix86_frame_pointer_required): Require
868 frame-pointer, if setjmp is used for 32-bit ms-abi.
870 2011-10-24 Michael Meissner <meissner@linux.vnet.ibm.com>
872 * builtins.c (set_builtin_user_assembler_name): Remove extra
873 newline added in October 11th, 2011 change.
875 2011-10-24 David S. Miller <davem@davemloft.net>
877 * config/sparc/little-endian.opt: Delete.
878 * config.gcc: Remove references to config/sparc/little-endian.opt
879 * doc/invoke.texi: Remove documentation of -mlittl-endian on sparc.
880 * config/sparc/linux64.h: Delete references to -mlittle-endian.
881 * config/sparc/netbsd-elf.h: Likewise.
882 * config/sparc/openbsd64.h: Likewise.
883 * config/sparc/sparc.h: Likewise.
884 * config/sparc/sp64-elf.h: Likewise and delete overrides for
885 BYTES_BIG_ENDIAN and WORDS_BIG_ENDIAN.
886 * config/sparc/sparc.c (dump_target_flag_bits): Remove reference
887 to MASK_LITTLE_ENDIAN.
888 * config/sparc/sparc.opt (Mask(LITTLE_ENDIAN)): Delete.
890 * config/sparc/sparc.md: Only use F, G, and C constraints in FP insns.
891 Only use D, Y, and Z constraints in vector insns.
893 * config/sparc/sparc.md (cpu_feature, enabled): New attributes.
894 (*movsi_insn_novis3, *movsi_insn_vis3): Consolidate into one pattern
896 (*movdi_insn_sp32_v9_novis3, *movdi_insn_sp32_v9_vis3): Consolidate
897 into *movdi_insn_sp32.
898 (*movdi_insn_sp64_novis3, *movdi_insn_sp64_vis3): Consolidate into
899 one pattern called *movdi_insn_sp64.
900 (*movsf_insn_novis3, *movsf_insn_vis3, *movsf_insn_no_fpu):
901 Consolidate into one pattern called *movsf_insn.
902 (*movdf_insn_sp32_no_fpu, *movdf_insn_sp32_v9_novis3,
903 *movdf_insn_sp32_v9_vis3, *movdf_insn_sp32_v9_no_fpu): Consolidate
904 into *movdf_insn_sp32.
905 (*movdf_insn_sp64_novis3, *movdf_insn_sp64_vis3,
906 *movdf_insn_sp64_no_fpu): Consolidate into one pattern called
908 (*zero_extendsidi2_insn_sp64_novis3,
909 *zero_extendsidi2_insn_sp64_vis3): Consolidate into one pattern
910 called *zero_extendsidi2_insn_sp64.
911 (*sign_extendsidi2_insn_novis3, *sign_extendsidi2_insn_vis3):
912 Consolidate into one pattern named *sign_extendsidi2_insn.
913 (*mov<VM32:mode>_insn_novis3, *mov<VM32:mode>_insn_vis3):
914 Consolidate into one pattern named *mov<VM32:mode>_insn.
915 (*mov<VM64:mode>_insn_sp64_novis3,
916 *mov<VM64:mode>_insn_sp64_novis3): Consolidate into one pattern
917 named *mov<VM64:mode>_insn_sp64.
918 (*mov<VM64:mode>_insn_sp32_novis3,
919 *mov<VM64:mode>_insn_sp32_vis3): Consolidate into one pattern
920 named *mov<VM64:mode>_insn_sp32.
922 2011-10-24 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
924 * tree-ssa-strlen.c (get_string_length): Change assertion to STPCPY.
925 (zero_length_string): Change assertion to accept strinfo without
926 length but with stmt instead.
927 Set the endptr pointer also if starting a new chain.
928 (adjust_related_strinfos): Ignore strinfos marked for delayed
930 (handle_builtin_strcpy): Mark earlier strinfo elements also for
931 delayed length computation.
933 2011-10-24 Georg-Johann Lay <avr@gjlay.de>
936 Port from 4.6 branch r180379
937 * doc/invoke.texi (AVR Options): New subsubsection to explain EIND
938 handling and indirect jump/calls on devices > 128k.
940 2011-10-24 Anatoly Sokolov <aesok@post.ru>
941 Georg-Johann Lay <avr@gjlay.de>
944 * doc/extend.texi (Declaring Attributes of Functions):
945 Document OS_main and OS_task attributes.
946 (Specifying Attributes of Variables): Move up
947 subsection "AVR Variable Attributes" as of alphabetical order.
949 2011-10-24 Richard Guenther <rguenther@suse.de>
951 * tree-vect-stmts.c (vect_get_vec_def_for_operand): Convert constants
952 to vector element type.
953 (vectorizable_assignment): Bail out for non-mode-precision operations.
954 (vectorizable_shift): Likewise.
955 (vectorizable_operation): Likewise.
956 (vectorizable_type_demotion): Likewise.
957 (vectorizable_type_promotion): Likewise.
958 (vectorizable_store): Handle non-mode-precision stores.
959 (vectorizable_load): Handle non-mode-precision loads.
960 (get_vectype_for_scalar_type_and_size): Return a vector type
961 for non-mode-precision integers.
962 * tree-vect-loop.c (vectorizable_reduction): Bail out for
963 non-mode-precision reductions.
965 2011-10-24 Julian Brown <julian@codesourcery.com>
967 * config/m68k/m68k.c (notice_update_cc): Tighten condition for
968 setting CC_REVERSED for FP comparisons.
970 2011-10-24 Richard Guenther <rguenther@suse.de>
972 PR tree-optimization/50838
973 * tree-data-ref.c (dr_analyze_indices): Properly canonicalize
974 a MEM_REF base if we change it.
976 2011-10-24 Bernd Schmidt <bernds@codesourcery.com>
979 * rtlanal.c: Swap includes of "hard-reg-set.h" and "rtl.h".
981 PR rtl-optimization/50833
982 * function.c (thread_prologue_and_epilogue_insns): Expect the
983 return insn optimization only if optimize.
985 2011-10-24 Georg-Johann Lay <avr@gjlay.de>
987 * config/avr/avr.c: Break long lines.
988 Define target hooks on the fly if applicable.
989 (TARGET_ASM_FUNCTION_RODATA_SECTION): Remove first definition
991 (targetm): Move definition to end of file.
992 (avr_can_eliminate): Make static on the fly.
993 (avr_frame_pointer_required_p): Ditto.
994 (avr_hard_regno_scratch_ok): Ditto.
995 (avr_builtin_setjmp_frame_value): Make static on the fly.
996 Indent according to coding rules.
997 (avr_case_values_threshold): Ditto.
998 (avr_attribute_table): Move down.
1000 2011-10-24 Ira Rosen <ira.rosen@linaro.org>
1002 PR tree-optimization/50730
1003 * tree-vect-data-refs.c (vect_analyze_data_refs): Stop basic block
1004 analysis if encountered unsupported data-ref.
1006 2011-10-23 David S. Miller <davem@davemloft.net>
1008 * config/sparc/sparc.c (sparc_option_override): Remove -mv8plus
1010 * config/sparc/linux64.h (CC1_SPEC): When defaulting to 64-bit,
1011 append -mcpu=v9 when -mv8plus is given.
1013 * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED): We can move
1014 between float and non-float regs when VIS3.
1015 * config/sparc/sparc.c (eligible_for_restore_insn): We can't
1016 use a restore when the source is a float register.
1017 (sparc_split_regreg_legitimate): When VIS3 allow moves between
1018 float and integer regs.
1019 (sparc_register_move_cost): Adjust to account for VIS3 moves.
1020 (sparc_preferred_reload_class): On 32-bit with VIS3 when moving an
1021 integer reg to a class containing EXTRA_FP_REGS, constrain to FP_REGS.
1022 (sparc_secondary_reload): On 32-bit with VIS3 when moving between
1023 float and integer regs we sometimes need a FP_REGS class
1024 intermediate move to satisfy the reload. When this happens
1025 specify an extra cost of 2.
1026 (*movsi_insn): Rename to have "_novis3" suffix and add !VIS3 guard.
1027 (*movdi_insn_sp32_v9): Likewise.
1028 (*movdi_insn_sp64): Likewise.
1029 (*movsf_insn): Likewise.
1030 (*movdf_insn_sp32_v9): Likewise.
1031 (*movdf_insn_sp64): Likewise.
1032 (*zero_extendsidi2_insn_sp64): Likewise.
1033 (*sign_extendsidi2_insn): Likewise.
1034 (*movsi_insn_vis3): New insn.
1035 (*movdi_insn_sp32_v9_vis3): New insn.
1036 (*movdi_insn_sp64_vis3): New insn.
1037 (*movsf_insn_vis3): New insn.
1038 (*movdf_insn_sp32_v9_vis3): New insn.
1039 (*movdf_insn_sp64_vis3): New insn.
1040 (*zero_extendsidi2_insn_sp64_vis3): New insn.
1041 (*sign_extendsidi2_insn_vis3): New insn.
1042 (TFmode reg/reg split): Make sure both REG operands are float.
1043 (*mov<VM32:mode>_insn): Add "_novis3" suffix and !VIS3 guard. Remove
1044 easy constant to integer reg alternatives.
1045 (*mov<VM64:mode>_insn_sp64): Likewise.
1046 (*mov<VM64:mode>_insn_sp32_novis3): Likewise.
1047 (*mov<VM32:mode>_insn_vis3): New insn.
1048 (*mov<VM64:mode>_insn_sp64_vis3): New insn.
1049 (*mov<VM64:mode>_insn_sp32_vis3): New insn.
1050 (VM64 reg<-->reg split): New splitter for 32-bit.
1052 * config/sparc/sparc.c (sparc_split_regreg_legitimate): New function.
1053 * config/sparc/sparc-protos.h (sparc_split_regreg_legitimate):
1055 * config/sparc/sparc.md (DImode reg/reg split): Use it.
1056 (DFmode reg/reg split): Likewise.
1058 * config/sparc/sparc.md (*movdi_insn_sp32_v9): Add alternatives for
1059 generating fzero and fone instructions.
1060 (DImode const_int --> reg splitter): Only trigger for integer regs.
1062 * config/sparc/predicates.md (input_operand): Disallow vector
1063 constants other than 0 and -1.
1064 * config/sparc/sparc.c (sparc_preferred_reload_class): Return
1065 NO_REGS for vector constants other than 0 and -1.
1067 * config/sparc/sparc.h (SPARC_FIRST_INT_REG, SPARC_LAST_INT_REG,
1068 SPARC_INT_REG_P): Define.
1069 (HARD_REGNO_NREGS): Use SPARC_INT_REG_P.
1070 (REGNO_OK_FOR_INDEX_P): Likewise.
1071 * config/sparc/sparc.c (gen_df_reg): Likewise.
1072 (eligible_for_return_delay): Likewise.
1073 (eligible_for_sibcall_delay): Likewise.
1074 (sparc_legitimate_address_p): Likewise.
1075 (emit_save_or_restore_regs): Likewise.
1076 (registers_ok_for_ldd_peep): Likewise.
1077 * config/spac/sparc.md (DI mode splitters): Likewise.
1078 (SF mode const splitters): Likewise.
1079 (DF mode splitters): Likewise.
1080 (32-bit DI mode logical op splitters): Likewise.
1082 2011-10-23 Paolo Carlini <paolo.carlini@oracle.com>
1086 2011-10-23 Paolo Carlini <paolo.carlini@oracle.com>
1089 * doc/invoke.texi ([-Wnarrowing], [-Wc++0x-compat]): Update.
1091 2011-10-23 Uros Bizjak <ubizjak@gmail.com>
1093 * config/i386/sse.md (avx2_lshl<mode>3): Remove insn pattern.
1094 (VI248_256): Remove mode iterator.
1095 * config/i386/i386.c (ix86_expand_vec_perm): Use gen_ashlv4di3
1096 instead of gen_avx2_lshlv4di3.
1097 (bdesc_args): Use CODE_FOR_ashl{v16hi,v8si,v4di}3 instead of
1098 CODE_FOR_avx2_lshl{v16hi,v8si,v4di}3.
1100 2011-10-23 Uros Bizjak <ubizjak@gmail.com>
1102 * config/i386/sse.md (sseintprefix): Rename from gthrfirstp.
1103 (<avx_avx2>_maskload<ssemodesuffix><avxsizesuffix>): Delete expander.
1104 (<avx_avx2>_maskload<ssemodesuffix><avxsizesuffix>) Merge insn
1105 pattern from *avx2_maskload<ssemodesuffix><avxsizesuffix> and
1106 *avx_maskload<ssemodesuffix><avxsizesuffix> using V48_AVX mode
1107 iterator. Use sseintprefix mode attribute.
1108 (<avx_avx2>_maskstore<ssemodesuffix><avxsizesuffix>): Delete expander.
1109 (<avx_avx2>_maskstore<ssemodesuffix><avxsizesuffix>) Merge insn
1110 pattern from *avx2_maskstore<ssemodesuffix><avxsizesuffix> and
1111 *avx_maskstore<ssemodesuffix><avxsizesuffix> using V48_AVX mode
1112 iterator. Use sseintprefix mode attribute.
1113 (*avx2_gathersi<mode>) Use sseintprefix and ssemodesuffix mode
1115 (*avx2_gatherdi<mode>): Ditto.
1116 (*avx2_gatherdi<mode>256): Ditto.
1117 (VI48_AVX2): Remove mode iterator.
1118 (gthrlastfp): Remove mode attribute.
1120 2011-10-23 Paolo Carlini <paolo.carlini@oracle.com>
1123 * doc/invoke.texi ([-Wnarrowing], [-Wc++0x-compat]): Update.
1125 2011-10-23 Tom de Vries <tom@codesourcery.com>
1127 PR tree-optimization/50763
1128 * tree-ssa-tail-merge.c (same_succ_flush_bb): New function, factored
1130 (same_succ_flush_bbs): Use same_succ_flush_bb.
1131 (purge_bbs): Remove argument. Remove calls to same_succ_flush_bbs,
1132 release_last_vdef and delete_basic_block.
1133 (unlink_virtual_phi): New function.
1134 (update_vuses): Add and use vuse1_phi_args argument. Set var to
1135 SSA_NAME_VAR of vuse1 or vuse2, and use var. Handle case that
1136 def_stmt2 is NULL. Use phi result as phi arg in case vuse1 or vuse2
1137 is NULL_TREE. Replace uses of vuse1 if vuse2 is NULL_TREE. Fix code
1138 to limit replacement of uses. Propagate phi argument for phis with a
1140 (replace_block_by): Update vops if phi_vuse1 or phi_vuse2 is NULL_TREE.
1141 Set vuse1_phi_args if vuse1 is a phi defined in bb1. Add
1142 vuse1_phi_args as argument to call to update_vuses. Call
1143 release_last_vdef, same_succ_flush_bb, delete_basic_block. Update
1144 CDI_DOMINATORS info.
1145 (tail_merge_optimize): Remove argument in call to purge_bbs. Remove
1146 call to free_dominance_info. Only call calculate_dominance_info once.
1148 2011-10-23 Eric Botcazou <ebotcazou@adacore.com>
1150 * fold-const.c (invert_tree_comparison): Always invert EQ_EXPR/NE_EXPR.
1152 PR tree-optimization/44683
1153 * tree-ssa-dom.c (record_edge_info): Record simple equivalences only if
1154 we can be sure that there are no signed zeros involved.
1156 2011-10-23 Jan Hubicka <jh@suse.cz>
1158 * ipa-inline.c (estimate_badness): Scale up and handle overflows.
1160 2011-10-23 Uros Bizjak <ubizjak@gmail.com>
1163 * config/i386/sse.md (<avx_avx2>_maskload<ssemodesuffix><avxsizesuffix>):
1164 Remove (match_dup 0).
1165 (*avx2_maskload<ssemodesuffix><avxsizesuffix>): New insn pattern.
1166 (*avx_maskload<ssemodesuffix><avxsizesuffix>): Ditto.
1167 (*avx2_maskstore<ssemodesuffix><avxsizesuffix>): Ditto.
1168 (*avx_maskstore<ssemodesuffix><avxsizesuffix>): Ditto.
1169 (*avx2_maskmov<ssemodesuffix><avxsizesuffix>): Remove insn pattern.
1170 (*avx_maskmov<ssemodesuffix><avxsizesuffix>): Ditto.
1172 2011-10-23 Ira Rosen <ira.rosen@linaro.org>
1174 PR tree-optimization/50819
1175 * tree-vectorizer.h (vect_analyze_data_ref_dependences): Remove
1177 * tree-vect-loop.c (vect_analyze_loop_2): Update call to
1178 vect_analyze_data_ref_dependences.
1179 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Remove
1180 the last argument. Check load-after-store dependence for unknown
1181 dependencies in basic blocks.
1182 (vect_analyze_data_ref_dependences): Update call to
1183 vect_analyze_data_ref_dependences.
1184 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo.
1185 * tree-vect-slp.c (vect_bb_vectorizable_with_dependencies): Remove.
1186 (vect_slp_analyze_bb_1): Update call to
1187 vect_analyze_data_ref_dependences. Don't call
1188 vect_bb_vectorizable_with_dependencies.
1190 2011-10-22 David S. Miller <davem@davemloft.net>
1192 * config/sparc/sparc.h (SECONDARY_INPUT_RELOAD_CLASS,
1193 SECONDARY_OUTPUT_RELOAD_CLASS): Delete.
1194 * config/sparc/sparc.c (TARGET_SECONDARY_RELOAD): Redefine.
1195 (sparc_secondary_reload): New function.
1197 * config/sparc/sparc.h (sparc_costs): Remove extern decl.
1198 (struct processor_costs): Move from here..
1199 * config/sparc/sparc.c (struct processor_costs): To here.
1200 (sparc_costs): Mark static.
1202 * config/sparc/sparc.c (short_branch, reg_unused_after): Delete.
1203 * config/sparc/sparc-protos.h (short_branch, reg_unused_after):
1204 Get rid of declarations.
1206 2011-10-21 Paul Brook <paul@codesourcery.com>
1208 * config/c6x/c6x.c (c6x_asm_emit_except_personality,
1209 c6x_asm_init_sections): New functions.
1210 (TARGET_ASM_EMIT_EXCEPT_PERSONALITY, TARGET_ASM_INIT_SECTIONS): Define.
1212 2011-10-21 Jakub Jelinek <jakub@redhat.com>
1215 * config/i386/i386.c (expand_vec_perm_even_odd_1): Handle
1216 V4DImode and V8SImode for !TARGET_AVX2.
1218 2011-10-21 Bernd Schmidt <bernds@codesourcery.com>
1221 * sched-deps.c (add_dependence): If not doing predication, promote
1222 REG_DEP_CONTROL to REG_DEP_ANTI.
1224 2011-10-21 Georg-Johann Lay <avr@gjlay.de>
1226 * config/avr/avr.h (LEGITIMIZE_RELOAD_ADDRESS): Pass address of X
1227 instead of X to avr_legitimize_reload_address.
1228 * config/avr/avr-protos.h (avr_legitimize_reload_address): Change
1229 first argument's type from rtx to rtx*.
1230 * config/avr/avr.c (avr_legitimize_reload_address): Ditto.
1231 Pass PX to push_reload instead of &X. Change log messages for
1234 2011-10-21 Roland Stigge <stigge@antcom.de>
1236 PR translation/47064
1237 * params.def: Fix typo "compilatoin" -> "compilation".
1239 2011-10-21 H.J. Lu <hongjiu.lu@intel.com>
1240 Kirill Yukhin <kirill.yukhin@intel.com>
1243 * config/i386/driver-i386.c (host_detect_local_cpu): Do cpuid 7 only
1244 if max_level allows that.
1246 2011-10-21 Bernd Schmidt <bernds@codesourcery.com>
1248 * reg-notes.def (DEP_CONTROL): New.
1249 * sched-ebb.c (add_deps_for_risky_insns): Add a REG_DEP_CONTROL when
1250 not doing speculation.
1251 * rtlanal.c (record_hard_reg_sets, find_all_hard_reg_sets,
1252 record_hard_reg_uses_1, record_hard_reg_uses): New functions.
1253 * function.c (record_hard_reg_sets, record_hard_reg_uses,
1254 record_hard_reg_uses_1): Remove; move to rtlanal.c.
1255 * lists.c (copy_INSN_LIST, concat_INSN_LIST): New functions.
1256 * haifa-sched.c: Swap includes of "rtl.h" and "hard-reg-set.h".
1257 (MUST_RECOMPUTE_SPEC_P): New macro.
1258 (real_insn_for_shadow): New function.
1259 (cond_clobbered_p, recompute_todo_spec, check_clobbered_conditions,
1260 toggle_cancelled_flags): New static functions.
1261 (schedule_insn): Relax an assert to only check for empty hard back
1262 dependencies. Skip cancelled dependencies. Call
1263 check_clobbered_conditions.
1264 (copy_insn_list): Remove function, renamed moved to lists.c.
1265 (save_backtrack_point): Use new spelling copy_INSN_LIST.
1266 (unschedule_insns_until): Ensure TODO_SPEC is reset properly.
1267 (restore_last_backtrack_point): Likewise. Call toggle_cancelled_flags.
1268 (estimate_insn_tick): Ignore cancelled dependencies.
1269 (haifa_speculate_insn): Move declaration.
1270 (try_ready): Move code into recompute_todo_spec and call it. Tweak
1271 some asserts. Ensure predicated patterns are restored if necessary.
1272 Dump DEP_CONTROL flag.
1273 (haifa_change_pattern): Merge with sched_change_pattern.
1274 (sched_change_pattern): Remove function.
1275 * sched-deps.c (NON_FLUSH_JUMP_KIND, NON_FLUSH_JUMP): Remove. All
1276 uses changed to simply not test NON_FLUSH_JUMP_P.
1277 (ds_to_dk, dk_to_ds, dump_dep, ds_to_dt, dump_ds, check_dep): Handle
1279 (dep_spec_p): If DO_PREDICATION, REG_DEP_CONTROL is speculative.
1280 (reg_pending_control_uses, control_dependency_cache): New static
1282 (sched_get_reverse_condition_uncached): New function.
1283 (sd_find_dep_between): Remove pointless assert. Look in
1284 control_dependency_cache.
1285 (ask_dependency_caches, set_dependency_caches, sd_delete_dep,
1286 extend_dependency_caches, sched_deps_finish): Handle REG_DEP_CONTROL
1287 and control_dependency_cache.
1288 (sd_unresolve_dep): Use dep_spec_p.
1289 (add_dependence): Now a wrapper around add_dependence_1, handling
1290 REG_DEP_CONTROL specially.
1291 (flush_pending_lists): Clear pending_jump_insns.
1292 (sched_analyze_1): Handle pending_jump_insns like a memory flush.
1293 (sched_analyze_2): Unconditionally add to pending memory flushes,
1294 keep previous behaviour but apply it to pending_jump_insns instead.
1295 (sched_analyze_insn): Defer adding jump reg dependencies using
1296 reg_pending_control_uses; add them to the control_uses list. Handle
1297 pending_jump_insns and control_uses when adding dependence lists.
1298 (deps_analyze_insn): Update INSN_COND_DEPS.
1299 (deps_analyze_insn): Add jumps to pending_jump_insns rather than
1300 last_pending_memory_flush.
1301 (init_deps): Initialize pending_jump_insns.
1302 (free_deps): Free control_uses.
1303 (remove_from_deps): Remove from pending_jump_insns.
1304 (init_deps_global): Allocate reg_pending_control_uses).
1305 (finish_deps_global): Free it.
1306 (add_dependence_1): Renamed from add_dependence. Handle
1308 * rtl.h (record_hard_reg_uses, find_all_hard_reg_sets): Declare.
1309 (copy_INSN_LIST, concat_INSN_LIST): Declare.
1310 * sched-int.h (struct deps_reg): Add control_uses.
1311 (struct deps_desc): Add pending_jump_insns.
1312 (struct _haifa_deps_insn_data): Add cond_deps.
1313 (struct _haifa_insn_data): Add must_recompute_spec and predicated_pat.
1314 (INSN_COND_DEPS, PREDICATED_PAT): New macros.
1315 (BITS_PER_DEP_WEAK): Adjust for two extra bits in the word.
1316 (DEP_CONTROL): New macro.
1317 (DEP_TYPES): Include it.
1318 (HARD_DEP): Adjust definition.
1319 (DEP_CANCELLED): New macro.
1320 (enum SCHED_FLAGS): Add DO_PREDICATION.
1321 (sched_get_reverse_condition_uncached, real_insn_for_shadow): Declare.
1322 * sched-rgn.c (concat_INSN_LIST): Remove function.
1323 (deps_join): Handle pending_jump_insns.
1324 (free_pending_lists): Likewise.
1325 * config/c6x/c6x.c (c6x_set_sched_flags): Set DO_PREDICATION for final
1328 2011-10-21 Georg-Johann Lay <avr@gjlay.de>
1331 * config/avr/libgcc.S (__EIND__): New define to 0x3C.
1332 (__tablejump__): Consistently use EIND for indirect jump/call.
1333 (__tablejump_elpm__): Ditto.
1335 2011-10-21 Bernd Schmidt <bernds@codesourcery.com>
1337 * config/c6x/c6x.md (attr "op_pattern"): New.
1338 (load_sdata_pic, mov<mode>_insn for QIHIM and SISFVM): Set it.
1339 * config/c6x/c6x-mult.md.in (mulhi3_VARIANT_, mulhisi3_insn_VARIANT_):
1341 * config/c6x/c6x-mult.md: Regenerate.
1342 * config/c6x/c6x.c: Include "regrename.h".
1343 (unit_req_table): New typedef.
1344 (unit_reqs): Use it for the declaration.
1345 (unit_req_factor, get_unit_reqs, merge_unit_reqs, unit_req_imbalance,
1346 get_unit_operand_masks, try_rename_operands, reshuffle_units): New
1348 (count_unit_reqs): New arg reqs. All callers changed. Use
1349 get_unit_reqs, and don't merge here.
1350 (res_mii): New arg reqs. All callers changed. Rewrite to use a loop
1351 using unit_req_factor.
1352 (hwloop_optimize): Call reshuffle_units. Call merge_unit_reqs after
1354 (c6x_reorg): Add reg notes problem, and call df_analyze.
1355 * Makefile.in ($(out_object_file)): Depend on regrename.h.
1357 2011-10-21 Kai Tietz <ktietz@redhat.com>
1359 * fold-const.c (simple_operand_p_2): Handle integral
1360 casts from boolean-operands.
1362 2011-10-21 Jan Hubicka <jh@suse.cz>
1364 * cgraph.c (dump_cgraph_node): Dump alias flag.
1365 * cgraphunit.c (handle_alias_pairs): Handle weakrefs with
1367 (get_alias_symbol): New function.
1368 (output_weakrefs): Output also weakrefs with no destinatoin.
1369 (lto_output_node): Output weakref alias flag when at function boundary.
1371 2011-10-21 Andrew Stubbs <ams@codesourcery.com>
1374 * config/arm/driver-arm.c (vendors): Make static.
1376 2011-10-21 Uros Bizjak <ubizjak@gmail.com>
1378 * config/i386/i386.c (ix86_emit_swdivsf): Force b into register.
1379 (ix86_emit_swsqrtsf): Force a into register.
1381 2011-10-20 Mike Stump <mikestump@comcast.net>
1383 * regcprop.c (copyprop_hardreg_forward_1): Update recog_data
1384 after validate_change wipes it out.
1386 2011-10-20 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1388 * config/spu/spu.md ("vec_permv16qi"): Reduce selector modulo 32
1389 before using the shufb instruction.
1391 2011-10-20 Kirill Yukhin <kirill.yukhin@intel.com>
1394 * config/i386/i386.md (bmi_bextr_<mode>): Update register/
1395 memory operand order.
1396 (bmi2_bzhi_<mode>3): Ditto.
1397 (bmi2_pdep_<mode>3): Ditto.
1398 (bmi2_pext_<mode>3): Ditto.
1400 2011-10-20 Richard Henderson <rth@redhat.com>
1402 * target.def (vec_perm_const_ok): Rename from builtin_vec_perm_ok.
1403 * optabs.c (can_vec_perm_expr_p): Update to match.
1404 (expand_vec_perm_expr): Likewise.
1405 * config/i386/i386.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Rename
1406 from TARGET_VECTORIZE_BUILTIN_VEC_PERM_OK.
1407 * doc/tm.texi.in: Likewise.
1409 2011-10-20 Sergey Ostanevich <sergos.gnu@gmail.com>
1412 * config/i386/i386.c (processor_target_table): Change Atom
1413 align_loops_max_skip to 15.
1415 2011-10-20 Richard Henderson <rth@redhat.com>
1417 * target.def (builtin_vec_perm): Remove.
1418 * doc/tm.texi.in (TARGET_VECTORIZE_BUILTIN_VEC_PERM): Remove.
1420 * config/i386/i386.c (ix86_expand_vec_perm_builtin): Remove.
1421 (IX86_BUILTIN_VEC_PERM_*): Remove.
1422 (bdesc_args): Remove vec_perm builtins
1423 (ix86_expand_builtin): Likewise.
1424 (ix86_expand_vec_perm_const_1): Rename from
1425 ix86_expand_vec_perm_builtin_1.
1426 (extract_vec_perm_cst): Merge into...
1427 (ix86_vectorize_vec_perm_const_ok): ... here. Rename from
1428 ix86_vectorize_builtin_vec_perm_ok.
1429 (TARGET_VECTORIZE_BUILTIN_VEC_PERM): Remove.
1431 * config/rs6000/rs6000.c (rs6000_builtin_vec_perm): Remove.
1432 (TARGET_VECTORIZE_BUILTIN_VEC_PERM): Remove.
1434 * config/spu/spu.c (spu_builtin_vec_perm): Remove.
1435 (TARGET_VECTORIZE_BUILTIN_VEC_PERM): Remove.
1437 2011-10-20 Uros Bizjak <ubizjak@gmail.com>
1440 * config/i386/i386.h (RECIP_MASK_DEFAULT): New define.
1441 * config/i386/i386.op (recip_mask): Initialize with RECIP_MASK_DEFAULT.
1442 * doc/invoke.texi (ix86 Options, -mrecip): Document that GCC
1443 implements vectorized single float division and vectorized sqrtf(x)
1444 with reciprocal sequence with additional Newton-Raphson step with
1447 2011-10-20 Dodji Seketeli <dodji@redhat.com>
1449 * ggc-zone.c (ggc_internal_alloc_zone_stat): Rename
1450 ggc_alloced_size_order_for_request into ggc_round_alloc_size like
1451 it was done in ggc-page.c.
1454 * doc/cppopts.texi: Use @smallexample/@end smallexample in
1455 documentation for -fdebug-cpp instead of @quotation/@end quotation
1456 that is not supported by contrib/texi2pod.pl.
1458 2011-10-19 Jan Hubicka <jh@suse.cz>
1460 * ipa-inline.c (inline_small_functions): Always update all calles after
1463 2011-10-19 Jan Hubicka <jh@suse.cz>
1466 * ipa-inline.c (inline_small_functions): Fix checking code to not make
1467 effect on fibheap stability.
1469 2011-10-20 Maxim Kuvyrkov <maxim@codesourcery.com>
1471 * config/m68k/t-linux (M68K_MLIB_CPU): Add ColdFire CPUs.
1473 2011-10-20 Ramana Radhakrishnan <ramana.radhakrishnan@linaro.org>
1476 * config/arm/arm.c (thumb_unexpanded_epilogue): Handle return
1479 2011-10-20 Richard Guenther <rguenther@suse.de>
1481 * tree-ssa-loop-im.c (stmt_cost): Add WIDEN_*, FMA_EXPR
1482 and rotates to the set of expensive operations.
1484 2011-10-19 David S. Miller <davem@davemloft.net>
1486 * config/sparc/sparc.c (sparc_expand_move): Use can_create_pseudo_p.
1487 (sparc_emit_set_const32): Likewise.
1488 (sparc_emit_set_const64_longway): Likewise.
1489 (sparc_emit_set_const64): Likewise.
1490 (sparc_legitimize_pic_address): Likewise.
1491 (memory_ok_for_ldd): Likewise.
1493 2011-10-20 Dehao Chen <dehao@google.com>
1495 * profile.c (compute_branch_probabilities): Compute and dump the
1496 overlap between the static estimation and the instrumentation profile.
1497 (OVERLAP_BASE): New macro.
1498 (compute_frequency_overlap): New function
1500 2011-10-19 Jakub Jelinek <jakub@redhat.com>
1502 * config/i386/i386.c (expand_vec_perm_vpshufb2_vpermq_even_odd): Use
1503 d->op1 instead of d->op0 for the second vpshufb.
1504 (expand_vec_perm_even_odd_1): For V8SImode fix vpshufd immediates.
1505 (ix86_expand_vec_perm_const): If mask indicates two operands are
1506 needed, but both are the same and expanding them as d.op0 == d.op1
1507 failed, retry with d.op0 != d.op1.
1508 (ix86_expand_vec_perm_builtin): Likewise. Handle sorry printing
1509 also for d.nelt == 32.
1512 * cfgexpand.c (expand_debug_expr): Handle WIDEN_LSHIFT_EXPR, ignore
1515 2011-10-19 Bernd Schmidt <bernds@codesourcery.com>
1517 * regrename.h: New file.
1518 * regrename.c: Include it. Also include "emit-rtl.h".
1519 (struct du_head, struct du_chain, du_head_p DEF_VEC and
1520 DEF_VEC_ALLOC_P): Move to regrename.h.
1521 (do_replace): Remove declaration.
1522 (insn_rr): New variable.
1523 (cur_operand): New static variable.
1524 (regrename_chain_from_id): Renamed from chain_from_id and no longer
1525 static. All callers changed.
1526 (record_operand_use): New static function.
1527 (scan_rtx_reg): Use it.
1528 (find_best_rename_reg): New function, broken out of rename_chains.
1529 (rename_chains): Use it. Don't update chain regno and nregs here, ...
1530 (regrename_do_replace): ... do it here instead. Renamed from
1531 do_replace, and no longer static. All callers changed.
1532 (regrename_analyze): No longer static. New arg bb_mask.
1533 All callers changed. If bb_mask is nonzero, use it to limit the
1534 number of basic blocks we analyze. If we failed to analyze a block,
1535 clear insn operand data.
1536 (record_out_operands): New arg insn_info. Update cur_operand if it is
1538 (build_def_use): If insn_rr is nonnull, pass an insn_info to
1539 record_out_operands, and update cur_operand here as well.
1540 (regrename_init, regrename_finish): New functions.
1541 (regrename_optimize): Use them.
1542 * Makefile.in (regrename.o): Adjust dependencies.
1544 2011-10-19 Tom de Vries <tom@codesourcery.com>
1546 PR tree-optimization/50769
1547 * tree-ssa-tail-merge.c (replace_block_by): Calculate phi_vuse2
1548 unconditionally. Handle case that phi_vuse2 is not an SSA_NAME. Add
1549 dummy argument .MEM to phi when increasing number of arguments of phi by
1550 redirecting edges to the block with phi.
1552 2011-10-19 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1555 * configure.ac: Save and restore CXXFLAGS around
1556 gcc_AC_CHECK_DECLS uses.
1557 Check for madvise() declaration with g++ if --enable-build-with-cxx.
1558 * configure: Regenerate.
1559 * config.in: Regenerate.
1560 * ggc-page.c (USING_MADVISE): Also check HAVE_DECL_MADVISE.
1562 2011-10-19 Alexandre Oliva <aoliva@redhat.com>
1565 * var-tracking.c (loc_exp_dep, onepart_aux): New structs.
1566 (variable_part): Replace offset with union.
1567 (enum onepart_enum, onepart_enum_t): New.
1568 (variable_def): Drop cur_loc_changed, add onepart.
1569 (value_chain_def, const_value_chain): Remove.
1570 (VAR_PART_OFFSET, VAR_LOC_1PAUX): New macros, with checking.
1571 (VAR_LOC_DEP_LST, VAR_LOC_DEP_LSTP): New macros.
1572 (VAR_LOC_FROM, VAR_LOC_DEPTH, VAR_LOC_DEP_VEC): Likewise.
1573 (value_chain_pool, value_chains): Remove.
1574 (dropped_values): New.
1575 (struct parm_reg): Only if HAVE_window_save.
1576 (vt_stack_adjustments): Don't record register arguments.
1578 (dv_onepart_p): Return a onepart_enum_t.
1579 (onepart_pool): New.
1582 (variable_htab_free): Release onepart aux data. Reset flags.
1583 (value_chain_htab_hash, value_chain_htab_eq): Remove.
1584 (unshare_variable): Use onepart field. Propagate onepart aux
1585 data or offset. Drop cur_loc_changed.
1586 (val_store): Cope with NULL insn. Rephrase dump output. Check
1587 for unsuitable locs. Add FIXME on using cselib locs.
1588 (val_reset): Remove FIXME of unfounded concerns.
1589 (val_resolve): Check for unsuitable locs. Add FIXME on using
1591 (variable_union): Use onepart field, adjust access to offset.
1593 (VALUE_CHANGED, DECL_CHANGED): Update doc.
1594 (set_dv_changed): Clear NO_LOC_P when changed.
1595 (find_loc_in_1pdv): Use onepart field.
1596 (intersect_loc_chains): Likewise.
1597 (unsuitable_loc): New.
1598 (loc_cmp): Keep ENTRY_VALUEs at the end of the loc list.
1599 (add_value_chain, add_value_chains): Remove.
1600 (add_cselib_value_chains, remove_value_chain): Likewise.
1601 (remove_value_chains, remove_cselib_value_chains): Likewise.
1602 (canonicalize_loc_order_check): Use onepart. Drop cur_loc_changed.
1603 (canonicalize_values_star, canonicalize_vars_star): Use onepart.
1604 (variable_merge_over_cur): Likewise. Adjust access to offset.
1605 Drop cur_loc_changed.
1606 (variable_merge_over_src): Use onepart field.
1607 (remove_duplicate_values): Likewise.
1608 (variable_post_merge_new_vals): Likewise.
1609 (find_mem_expr_in_1pdv): Likewise.
1610 (dataflow_set_preserve_mem_locs): Likewise. Drop cur_loc_changed
1612 (dataflow_set_remove_mem_locs): Likewise. Use VAR_LOC_FROM.
1613 (variable_different_p): Use onepart field. Move onepart test out
1615 (argument_reg_set): Drop.
1616 (add_uses, add_stores): Preserve but do not record in dynamic
1617 tables equivalences for ENTRY_VALUEs and CFA_based addresses.
1618 Avoid unsuitable address expressions.
1619 (EXPR_DEPTH): Unlimit.
1620 (EXPR_USE_DEPTH): Repurpose PARAM_MAX_VARTRACK_EXPR_DEPTH.
1621 (prepare_call_arguments): Use DECL_RTL_IF_SET.
1622 (dump_var): Adjust access to offset.
1623 (variable_from_dropped, recover_dropped_1paux): New.
1624 (variable_was_changed): Drop cur_loc_changed. Use onepart.
1625 Preserve onepart aux in empty_var. Recover empty_var and onepart
1626 aux from dropped_values.
1627 (find_variable_location_part): Special-case onepart. Adjust
1629 (set_slot_part): Use onepart. Drop cur_loc_changed. Adjust
1630 access to offset. Initialize onepaux. Drop value chains.
1631 (delete_slot_part): Drop value chains. Use VAR_LOC_FROM.
1632 (VEC (variable, heap), VEC (rtx, stack)): Define.
1633 (expand_loc_callback_data): Drop dummy, cur_loc_changed,
1634 ignore_cur_loc. Add expanding, pending, depth.
1635 (loc_exp_dep_alloc, loc_exp_dep_clear): New.
1636 (loc_exp_dep_insert, loc_exp_dep_set): New.
1637 (notify_dependents_of_resolved_value): New.
1638 (update_depth, vt_expand_var_loc_chain): New.
1639 (vt_expand_loc_callback): Revamped.
1640 (resolve_expansions_pending_recursion): New.
1641 (INIT_ELCD, FINI_ELCD): New.
1642 (vt_expand_loc): Use the new macros above. Drop ignore_cur_loc
1643 parameter, adjust all callers.
1644 (vt_expand_loc_dummy): Drop.
1645 (vt_expand_1pvar): New.
1646 (emit_note_insn_var_location): Operate on non-debug decls only.
1647 Revamp multi-part cur_loc recomputation and one-part expansion.
1648 Drop cur_loc_changed. Adjust access to offset.
1649 (VEC (variable, heap)): Drop.
1650 (changed_variables_stack, changed_values_stack): Drop.
1651 (check_changed_vars_0, check_changed_vars_1): Remove.
1652 (check_changed_vars_2, check_changed_vars_3): Remove.
1653 (values_to_stack, remove_value_from_changed_variables): New.
1654 (notify_dependents_of_changed_value, process_changed_values): New.
1655 (emit_notes_for_changes): Revamp onepart updates.
1656 (emit_notes_for_differences_1): Use onepart. Drop cur_loc_changed
1657 and value chains. Propagate onepaux. Recover empty_var and onepaux
1658 from dropped_values.
1659 (emit_notes_for_differences_2): Drop value chains.
1660 (emit_notes_in_bb): Adjust.
1661 (vt_emit_notes): Drop value chains, changed_variables_stack.
1662 Initialize and release dropped_values.
1663 (create_entry_value): Revamp.
1664 (vt_add_function_parameter): Use new interface.
1665 (note_register_arguments): Remove.
1666 (vt_initialize): Drop value chains and register arguments.
1667 (vt_finalize): Drop value chains. Release windowed_parm_regs only
1668 if HAVE_window_save.
1669 * rtl.h: Document various pass-local uses of RTL flags.
1670 * tree.h (DECL_RTL_KNOWN_SET): New.
1671 * doc/invoke.texi (param max-vartrack-expr-depth): Update
1672 description and default.
1674 2011-10-19 Georg-Johann Lay <avr@gjlay.de>
1677 * config/avr/avr.md (cc): New alternative out_plus_noclobber.
1678 (adjust_len): Ditto.
1679 (addhi3): Don't pipe through short; use gen_int_mode instead.
1680 Prior to reload, expand to gen_addhi3_clobber.
1681 (*addhi3): Use avr_out_plus_noclobber if applicable, use
1682 out_plus_noclobber in cc and adjust_len attribute.
1683 (addhi3_clobber): 2 new RTL peepholes.
1684 (addhi3_clobber): New insn.
1685 * config/avr/avr-protos.h: (avr_out_plus_noclobber): New prototype.
1686 * config/avr/avr.c (avr_out_plus_noclobber): New function.
1687 (notice_update_cc): Handle CC_OUT_PLUS_NOCLOBBER.
1688 (avr_out_plus_1): Tweak if only MSB is +/-1 and other bytes are 0.
1689 Set cc0 to set_zn for adiw on 16-bit values.
1690 (adjust_insn_length): Handle ADJUST_LEN_OUT_PLUS_NOCLOBBER.
1691 (expand_epilogue): No need to add 0 to frame_pointer_rtx.
1693 2011-10-19 Richard Guenther <rguenther@suse.de>
1696 * tree-ssa-forwprop.c (forward_propagate_into_cond): Verify
1697 the condition is properly gimple before using it.
1698 * tree-eh (stmt_could_throw_1_p): Properly extract the
1699 operation type from comparisons.
1701 2011-10-19 Roland Stigge <stigge@antcom.de>
1703 PR translation/48638
1704 * plugin.c (add_new_plugin): Fix typo in fatal_error message.
1706 2011-10-19 Roland Stigge <stigge@antcom.de>
1708 PR translation/49517
1709 * config/rx/rx.c (rx_print_operand): Fix typo in warning message.
1711 2011-10-19 Richard Guenther <rguenther@suse.de>
1714 * gimple-fold.c (gimplify_and_update_call_from_tree): Rewrite.
1716 2011-10-19 Andrey Belevantsev <abel@ispras.ru>
1718 PR rtl-optimization/50340
1719 * sel-sched-ir.c (update_target_availability): LHS register
1720 availability is not known if the unavailable LHS of the other
1721 expression is a different register.
1723 2011-10-19 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1726 * config/spu/spu.c (spu_emit_vector_compare): Support unordered
1727 floating-point comparisons.
1729 2011-10-19 Jan Hubicka <jh@suse.cz>
1731 * cgraphunit.c (handle_alias_pairs): Also handle wekref with
1732 destination declared.
1733 (output_weakrefs): New function.
1734 * varpool.c (varpool_create_variable_alias): Handle external aliases.
1736 2011-10-19 Jakub Jelinek <jakub@redhat.com>
1738 * dwarf2out.c (loc_descriptor): For SUBREG pass SUBREG_REG's mode as
1739 second argument instead of mode.
1741 2011-10-18 Jakub Jelinek <jakub@redhat.com>
1743 * config/i386/i386.c (ix86_expand_vec_perm): In merge_two use
1744 mode SUBREG of operands[0] as target.
1745 (valid_perm_using_mode_p): Don't ignore higher bits of d->perm.
1746 (expand_vec_pshufb): For V8SImode vmode emit avx2_permvarv8si.
1747 (expand_vec_perm_1): Handle identity and some broadcast
1749 (expand_vec_perm_interleave2): Handle also 32-byte modes, using
1750 vperm2[fi]128 or vpunpck[lh]* followed by single insn permutation.
1751 For d->testing_p return true earlier to avoid creating more GC
1753 (expand_vec_perm_vpermq_perm_1): New function.
1754 (expand_vec_perm_vpshufb2_vpermq): For d->testing_p return true
1755 earlier to avoid creating more GC garbage. Fix handling of
1756 V16HImode. Avoid some SUBREGs in SET_DEST.
1757 (expand_vec_perm_broadcast_1): Return false for 32-byte integer
1759 (expand_vec_perm_vpshufb4_vpermq2): New function.
1760 (ix86_expand_vec_perm_builtin_1): Call expand_vec_perm_vpermq_perm_1
1761 and expand_vec_perm_vpshufb4_vpermq2.
1763 2011-10-18 Andrew Stubbs <ams@codesourcery.com>
1765 * config/arm/driver-arm.c (host_detect_local_cpu): Close the file
1768 2011-10-18 Andrew Stubbs <ams@codesourcery.com>
1770 PR tree-optimization/50717
1771 * tree-ssa-math-opts.c (is_widening_mult_p): Remove the 'type'
1772 parameter. Calculate 'type' from stmt.
1773 (convert_mult_to_widen): Update call the is_widening_mult_p.
1774 (convert_plusminus_to_widen): Likewise.
1776 2011-10-18 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1778 * config/spu/spu.c (struct machine_function): New data structure.
1779 (spu_init_machine_status): New function.
1780 (spu_option_override): Install it.
1781 (get_pic_reg): Set and use cfun->machine->pic_reg.
1782 (spu_split_immediate): Do not set crtl->uses_pic_offset_table.
1783 (need_to_save_reg): Use cfun->machine->pic_reg instead of
1784 checking crtl->uses_pic_offset_table.
1785 (spu_expand_prologue): Likewise.
1787 2011-10-18 Jakub Jelinek <jakub@redhat.com>
1789 PR tree-optimization/50735
1790 * function.c (gimplify_parameters): Use create_tmp_var instead of
1791 create_tmp_reg. If parm is not TREE_ADDRESSABLE and type is complex
1792 or vector type, set DECL_GIMPLE_REG_P.
1794 2011-10-18 Andrew Stubbs <ams@codesourcery.com>
1796 * config.host (arm*-*-linux*): Add driver-arm.o and x-arm.
1797 * config/arm/arm.opt: Add 'native' processor_type and
1798 arm_arch enum values.
1799 * config/arm/arm.h (host_detect_local_cpu): New prototype.
1800 (EXTRA_SPEC_FUNCTIONS): New define.
1801 (MCPU_MTUNE_NATIVE_SPECS): New define.
1802 (DRIVER_SELF_SPECS): New define.
1803 * config/arm/driver-arm.c: New file.
1804 * config/arm/x-arm: New file.
1805 * doc/invoke.texi (ARM Options): Document -mcpu=native,
1806 -mtune=native and -march=native.
1808 2011-10-18 Alexander Monakov <amonakov@ispras.ru>
1810 PR rtl-optimization/50205
1811 * sel-sched.c (count_occurrences_1): Simplify on the assumption that
1812 p->x is a register. Forbid substitution when the same register is
1813 found in a different mode.
1814 (count_occurrences_equiv): Assert that 'what' is a register.
1816 2011-10-18 Richard Guenther <rguenther@suse.de>
1818 PR tree-optimization/50767
1819 * tree-ssa-pre.c (create_expression_by_pieces): Update the
1822 2011-10-18 Julian Brown <julian@codesourcery.com>
1824 * config/arm/arm.c (arm_block_move_unaligned_straight)
1825 (arm_adjust_block_mem, arm_block_move_unaligned_loop)
1826 (arm_movmemqi_unaligned): New.
1827 (arm_gen_movmemqi): Support unaligned block copies.
1829 2011-10-18 Ira Rosen <ira.rosen@linaro.org>
1831 * doc/md.texi (vec_widen_ushiftl_hi, vec_widen_ushiftl_lo,
1832 vec_widen_sshiftl_hi, vec_widen_sshiftl_lo): Document.
1833 * tree-pretty-print.c (dump_generic_node): Handle WIDEN_LSHIFT_EXPR,
1834 VEC_WIDEN_LSHIFT_HI_EXPR and VEC_WIDEN_LSHIFT_LO_EXPR.
1835 (op_code_prio): Likewise.
1836 (op_symbol_code): Handle WIDEN_LSHIFT_EXPR.
1837 * optabs.c (optab_for_tree_code): Handle
1838 VEC_WIDEN_LSHIFT_HI_EXPR and VEC_WIDEN_LSHIFT_LO_EXPR.
1839 (init-optabs): Initialize optab codes for vec_widen_u/sshiftl_hi/lo.
1840 * optabs.h (enum optab_index): Add OTI_vec_widen_u/sshiftl_hi/lo.
1841 * genopinit.c (optabs): Initialize the new optabs.
1842 * expr.c (expand_expr_real_2): Handle
1843 VEC_WIDEN_LSHIFT_HI_EXPR and VEC_WIDEN_LSHIFT_LO_EXPR.
1844 * gimple-pretty-print.c (dump_binary_rhs): Likewise.
1845 * tree-vectorizer.h (NUM_PATTERNS): Increase to 8.
1846 * tree.def (WIDEN_LSHIFT_EXPR, VEC_WIDEN_LSHIFT_HI_EXPR,
1847 VEC_WIDEN_LSHIFT_LO_EXPR): New.
1848 * cfgexpand.c (expand_debug_expr): Handle new tree codes.
1849 * tree-vect-patterns.c (vect_vect_recog_func_ptrs): Add
1850 vect_recog_widen_shift_pattern.
1851 (vect_handle_widen_mult_by_const): Rename...
1852 (vect_handle_widen_op_by_const): ...to this. Handle shifts.
1853 Add a new argument, update documentation.
1854 (vect_recog_widen_mult_pattern): Assume that only second
1855 operand can be constant. Update call to
1856 vect_handle_widen_op_by_const.
1857 (vect_recog_over_widening_pattern): Fix typo.
1858 (vect_recog_widen_shift_pattern): New.
1859 * tree-vect-stmts.c (vectorizable_type_promotion): Handle
1861 (supportable_widening_operation): Likewise.
1862 * tree-inline.c (estimate_operator_cost): Handle new tree codes.
1863 * tree-vect-generic.c (expand_vector_operations_1): Likewise.
1864 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
1865 * config/arm/neon.md (neon_vec_<US>shiftl_<mode>): New.
1866 (vec_widen_<US>shiftl_lo_<mode>, neon_vec_<US>shiftl_hi_<mode>,
1867 vec_widen_<US>shiftl_hi_<mode>, neon_vec_<US>shift_left_<mode>):
1869 * config/arm/predicates.md (const_neon_scalar_shift_amount_operand):
1871 * config/arm/iterators.md (V_innermode): New.
1872 * tree-vect-slp.c (vect_build_slp_tree): Require same shift operand
1875 2011-10-18 Richard Guenther <rguenther@suse.de>
1877 * tree-ssa-alias.h (struct pt_solution): Remove
1878 vars_contains_restrict member.
1879 (pt_solutions_same_restrict_base): Remove.
1880 (pt_solution_set): Adjust.
1881 * tree-ssa-alias.c (ptr_deref_may_alias_decl_p): Remove
1882 vars_contains_restrict handling.
1883 (dump_points_to_solution): Likewise.
1884 (ptr_derefs_may_alias_p): Do not call pt_solutions_same_restrict_base.
1885 * tree-ssa-structalias.c (struct variable_info): Remove is_restrict_var
1887 (new_var_info): Do not initialize it.
1888 (ipa_escaped_pt): Adjust.
1889 (make_constraint_from_restrict): Make the tag global.
1890 (make_constraint_from_global_restrict): New function.
1891 (make_constraint_from_heapvar): Remove.
1892 (create_variable_info_for): Do not make restrict vars point
1894 (intra_create_variable_infos): Likewise.
1895 (find_what_var_points_to): Remove vars_contains_restrict handling.
1896 (pt_solution_set): Adjust.
1897 (pt_solution_ior_into): Likewise.
1898 (pt_solutions_same_restrict_base): Remove.
1899 (compute_points_to_sets): Do not test is_restrict_var.
1900 * cfgexpand.c (update_alias_info_with_stack_vars): Adjust.
1901 * gimple-pretty-print.c (pp_points_to_solution): Likewise.
1903 2011-10-18 Tom de Vries <tom@codesourcery.com>
1905 PR tree-optimization/50672
1906 * tree-ssa-dce.c (mark_virtual_operand_for_renaming): New function,
1908 (mark_virtual_phi_result_for_renaming): Use
1909 mark_virtual_operand_for_renaming.
1910 * tree-flow.h (mark_virtual_operand_for_renaming): Declare.
1911 * tree-ssa-tail-merge.c (release_last_vdef): New function.
1912 (purge_bbs): Add update_vops parameter. Call release_last_vdef for each
1913 deleted basic block.
1914 (tail_merge_optimize): Add argument to call to purge_bbs.
1916 2011-10-18 Richard Guenther <rguenther@suse.de>
1919 * expr.c (get_object_or_type_alignment): New function.
1920 (expand_assignment): Use it.
1921 (expand_expr_real_1): Likewise.
1923 2011-10-18 Dodji Seketeli <dodji@redhat.com>
1926 * input.c (dump_line_table_statistics): Use long, not size_t.
1928 2011-10-17 Eric Botcazou <ebotcazou@adacore.com>
1930 * config/sparc/sparc.md (in_call_delay): Fix formatting issues.
1932 2011-10-17 Simon Baldwin <simonb@google.com>
1933 Ian Lance Taylor <iant@google.com>
1935 * configure.ac: Add --with-native-system-header-dir. Set and
1936 substitute NATIVE_SYSTEM_HEADER_DIR. Use native_system_header
1937 when setting target_header_dir.
1938 * config.gcc: Always set native_system_header_dir.
1939 (*-*-gnu*): Set native_system_header_dir. Don't use t-gnu.
1940 (i[34567]86-pc-msdosdjgpp*): Set native_system_header_dir. Don't
1942 (i[34567]86-*-mingw* | x86_64-*-mingw*): Set
1943 native_system_header_dir.
1944 (spu-*-elf*): Set native_system_header_dir.
1945 * Makefile.in (NATIVE_SYSTEM_HEADER_DIR): Set to
1946 @NATIVE_SYSTEM_HEADER_DIR@.
1947 (PREPROCESSOR_DEFINES): Define NATIVE_SYSTEM_HEADER_DIR.
1948 * cppdefault.c (STANDARD_INCLUDE_DIR): Don't define.
1949 (NATIVE_SYSTEM_HEADER_COMPONENT): Rename from
1950 STANDARD_INCLUDE_COMPONENT.
1951 (cpp_include_defaults): Don't use SYSTEM_INCLUDE_DIR. Rename
1952 STANDARD_INCLUDE_DIR to NATIVE_SYSTEM_HEADER_DIR.
1953 * system.h: Poison SYSTEM_INCLUDE_DIR, STANDARD_INCLUDE_DIR, and
1954 STANDARD_INCLUDE_COMPONENT.
1955 * config/i386/t-mingw32 (NATIVE_SYSTEM_HEADER_DIR): Remove.
1956 * config/i386/t-mingw-w32: Likewise.
1957 * config/i386/t-mingw-w64: Likewise.
1958 * config/spu/t-spu-elf: Likewise.
1959 * config/i386/t-djgpp: Remove.
1960 * config/t-gnu: Remove.
1961 * config/i386/mingw32.h (STANDARD_INCLUDE_DIR): Don't define.
1962 (NATIVE_SYSTEM_HEADER_COMPONENT): Rename from
1963 STANDARD_INCLUDE_COMPONENT.
1964 * config/i386/djgpp.h (STANDARD_INCLUDE_DIR): Don't define.
1965 * config/spu/spu-elf.h: Likewise.
1966 * config/vms/xm-vms.h: Likewise.
1967 * config/gnu.h: Likewise.
1968 * config/openbsd.h (INCLUDE_DEFAULTS): Change STANDARD_INCLUDE_DIR
1969 and STANDARD_INCLUDE_COMPONENT to NATIVE_SYSTEM_HEADER_DIR and
1970 NATIVE_SYSTME_HEADER_COMPONENT.
1971 * doc/install.texi (Configuration): Document
1972 --with-native-system-header-dir. Mention it in the documentation
1973 for --with-sysroot and --with-build-sysroot.
1974 * doc/tm.texi.in (Driver): Don't document SYSTEM_INCLUDE_DIR or
1975 STANDARD_INCLUDE_DIR. Rename STANDARD_INCLUDE_COMPONENT to
1976 NATIVE_SYSTEM_HEADER_COMPONENT. Rename uses of
1977 STANDARD_INCLUDE_DIR to NATIVE_SYSTEM_HEADER_DIR.
1978 * doc/fragments.texi (Target Fragment): Don't document
1979 NATIVE_SYSTEM_HEADER_DIR.
1980 * configure, doc/tm.texi: Rebuild.
1982 2011-10-17 Richard Henderson <rth@redhat.com>
1984 * config/sparc/sparc.md: Use register_or_zero_operand where rJ
1987 * config/sparc/sparc.md (vec_perm_constv8qi, vec_perm<mode>): New
1989 * config/sparc/sparc.c (sparc_expand_vec_perm_bmask): New function.
1990 * config/sparc/sparc-protos.h (sparc_expand_vec_perm_bmask): Declare.
1992 2011-10-17 David S. Miller <davem@davemloft.net>
1994 * config/sparc/sparc-modes.def: Add single entry vector modes for
1996 * config/sparc/sparc.md (V32, V32I, V64, V64I, V64N8): Delete
1998 (mov<V32:mode>): Revert back to plain SFmode pattern.
1999 (*movsf_insn): Likewise.
2000 (mov<V64:mode>): Revert back to plain DFmode pattern.
2001 (*movdf_insn_sp32): Likewise.
2002 (*movdf_insn_sp32_v9): Likewise.
2003 (*movdf_insn_sp64): Likewise.
2004 (V64 mode splitters) Likewise.
2005 (addsi3): Remove VIS alternatives.
2007 (and<V64I:mode>3): Revert to DImode only pattern.
2008 (and<V64I:mode>3_sp32): Likewise.
2009 (*and<V64I:mode>3_sp64): Likewise.
2010 (and<V32I:mode>3): Likewise.
2011 (*and_not_<V64I:mode>_sp32): Likewise.
2012 (*and_not_<V64I:mode>_sp64): Likewise.
2013 (*and_not_<V32I:mode>): Likewise.
2014 (ior<V64I:mode>3): Likewise.
2015 (*ior<V64I:mode>3_sp32): Likewise.
2016 (*ior<V64I:mode>3_sp64): Likewise.
2017 (ior<V32I:mode>3): Likewise.
2018 (*or_not_<V64I:mode>_sp32): Likewise.
2019 (*or_not_<V64I:mode>_sp64): Likewise.
2020 (*or_not_<V32I:mode>): Likewise.
2021 (xor<V64I:mode>3): Likewise.
2022 (*xor<V64I:mode>3_sp32): Likewise.
2023 (*xor<V64I:mode>3_sp64): Likewise.
2024 (xor<V32I:mode>3): Likewise.
2025 (V64I mode splitters): Likewise.
2026 (*xor_not_<V64I:mode>_sp32): Likewise.
2027 (*xor_not_<V64I:mode>_sp64): Likewise.
2028 (*xor_not_<V32I:mode>): Likewise.
2029 (one_cmpl<V64I:mode>2): Likewise.
2030 (*one_cmpl<V64I:mode>2_sp32): Likewise.
2031 (*one_cmpl<V64I:mode>2_sp64): Likewise.
2032 (one_cmpl<V32I:mode>2): Likewise.
2033 (VM32, VM64, VMALL): New mode iterators.
2034 (vbits, vconstr, vfptype): New mode attributes.
2035 (mov<VMALL:mode>): New expander.
2036 (*mov<VM32:mode>_insn): New insn.
2037 (*mov<VM64:mode>_insn_sp64): New insn.
2038 (*mov<VM64:mode>_insn_sp32): New insn, and associated splitter
2039 specifically for the register to memory case.
2040 (vec_init<mode>): New expander.
2041 (VADDSUB): New mode iterator.
2042 (<plusminus_insn>v2si3, <plusminus_insn>v2hi3): Remove and replace
2044 (<plusminus_insn><mode>3): New consolidated pattern.
2045 (VL): New mode iterator for logical operations.
2046 (vlsuf): New more attribute.
2047 (vlop): New code iterator.
2048 (vlinsn, vlninsn): New code attributes.
2049 (<code><mode>3): New insn to non-negated vector logical ops.
2050 (*not_<code><mode>3): Likewise for negated variants.
2051 (*nand<mode>_vis): New insn.
2052 (vlnotop): New code iterator.
2053 (*<code>_not1<mode>_vis, *<code>_not2<mode>_vis): New insns.
2054 (one_cmpl<mode>2): New insn.
2055 (faligndata<V64I:mode>_vis): Rewrite to use VM64 iterator.
2056 (bshuffle<VM64:mode>_vis): Likewise.
2057 (v<vis3_shift_patname><mode>3): Use GCM mode iterator.
2058 (fp<plusminus_insn>64_vis): Use V1DI mode.
2059 (VASS mode iterator): Use V1SI not SI mode.
2060 * config/sparc/sparc.c (sparc_vis_init_builtins): Account for
2061 single-entry vector mode changes.
2062 (sparc_expand_builtin): Likewise.
2063 (sparc_expand_vector_init): New function.
2064 * config/sparc/sparc-protos.h (sparc_expand_vector_init): Declare.
2066 2011-10-17 Kai Tietz <ktietz@redhat.com>
2068 * fold-const.c (simple_operand_p_2): New function.
2069 (fold_truthop): Rename to
2070 (fold_truth_andor_1): function name.
2071 Additionally remove branching creation for logical and/or.
2072 (fold_truth_andor): Handle branching creation for logical and/or here.
2074 2011-10-17 Andi Kleen <ak@linux.intel.com>
2076 * ggc-page.c (USING_MADVISE): Adjust ifdef to check for USING_MMAP.
2078 2011-10-17 Georg-Johann Lay <avr@gjlay.de>
2080 * config/avr/avr.h (ASSEMBLER_DIALECT): Remove.
2081 * config/avr/avr.md (mcu_have_movw, mcu_mega): Remove attributes.
2082 (adjust_len): Add alternative "call".
2083 (isa, enabled): New insn attributes.
2084 (length): Use match_test with AVR_HAVE_JMP_CALL instead of
2086 (*sbrx_branch<mode>): Ditto.
2087 (*sbrx_and_branch<mode>): Ditto.
2088 (*sbix_branch): Ditto.
2089 (*sbix_branch_bit7): Ditto.
2090 (*sbix_branch_tmp): Ditto.
2091 (*sbix_branch_tmp_bit7): Ditto.
2093 (negsi2): Use attribute "isa" instead of assembler dialect.
2094 (extendhisi2): Ditto.
2095 (call_insn, call_value_insn): Set adjust_len attribute.
2096 (indirect_jump): Indent to coding rules.
2097 (call_prologue_saves): Use isa attribute instead of mcu_mega.
2098 (epilogue_restores): Ditto. Fix setting of SP as described in the
2100 (*indirect_jump): Fusion of *jcindirect_jump, *njcindirect_jump
2101 and *indirect_jump_avr6.
2102 (*tablejump): Fusion of *tablejump_rjmp and *tablejump_lib.
2103 (*jcindirect_jump, *njcindirect_jump, *indirect_jump_avr6): Remove.
2104 (*tablejump_rjmp, *tablejump_lib): Remove.
2105 * config/avr/avr.c (adjust_insn_length): Handle ADJUST_LEN_CALL.
2107 2011-10-17 Paolo Carlini <paolo.carlini@oracle.com>
2110 * doc/invoke.texi ([Wnonnull]): Update.
2112 2011-10-17 Richard Henderson <rth@redhat.com>
2115 * optabs.c (expand_vec_perm_expr): Fix indexing error.
2117 2011-10-17 Sergio Durigan Junior <sergiodj@redhat.com>
2119 * configure.ac: Display `yes' if the SystemTap header has been found.
2120 * configure: Regenerate.
2122 2011-10-08 Andi Kleen <ak@linux.intel.com>
2125 * config.in, configure: Regenerate.
2126 * configure.ac (madvise): Add to AC_CHECK_FUNCS.
2127 * ggc-page.c (USING_MADVISE): Add.
2128 (page_entry): Add discarded field.
2129 (alloc_page): Check for discarded pages.
2130 (release_pages): Add USING_MADVISE branch.
2132 2011-10-17 Richard Guenther <rguenther@suse.de>
2134 PR tree-optimization/50729
2135 * tree-vrp.c (extract_range_from_unary_expr_1): Remove redundant test.
2136 (simplify_conversion_using_ranges): Properly test the
2137 intermediate result.
2139 2011-10-15 Tom Tromey <tromey@redhat.com>
2140 Dodji Seketeli <dodji@redhat.com>
2142 * ggc.h (ggc_round_alloc_size): Declare new public entry point.
2143 * ggc-none.c (ggc_round_alloc_size): New public stub function.
2144 * ggc-page.c (ggc_alloced_size_order_for_request): New static
2145 function. Factorized from ggc_internal_alloc_stat.
2146 (ggc_round_alloc_size): New public function. Uses
2147 ggc_alloced_size_order_for_request.
2148 (ggc_internal_alloc_stat): Use ggc_alloced_size_order_for_request.
2149 * ggc-zone.c (ggc_round_alloc_size): New public function extracted
2150 from ggc_internal_alloc_zone_stat.
2151 (ggc_internal_alloc_zone_stat): Use ggc_round_alloc_size.
2152 * toplev.c (general_init): Initialize
2153 line_table->alloced_size_for_request.
2155 2011-10-15 Tom Tromey <tromey@redhat.com>
2156 Dodji Seketeli <dodji@redhat.com>
2158 * input.c (ONE_K, ONE_M, SCALE, STAT_LABEL, FORMAT_AMOUNT): New macros.
2159 (num_expanded_macros_counter, num_macro_tokens_counter): Declare
2161 (dump_line_table_statistics): Define new function.
2162 * input.h (dump_line_table_statistics): Declare new function.
2163 * toplev.c (dump_memory_report): Call dump_line_table_statistics.
2165 2011-10-15 Tom Tromey <tromey@redhat.com>
2166 Dodji Seketeli <dodji@redhat.com>
2168 * doc/cppopts.texi: Document -fdebug-cpp.
2169 * doc/invoke.texi: Add -fdebug-cpp to the list of preprocessor options.
2171 2011-10-15 Tom Tromey <tromey@redhat.com>
2172 Dodji Seketeli <dodji@redhat.com>
2174 * gcc/diagnostic.h (diagnostic_report_current_module): Add a
2176 * diagnostic.c (diagnostic_report_current_module): Add a location
2177 parameter to the function definition. Use it instead of
2178 input_location. Resolve the virtual location rather than just
2179 looking up its map and risking to touch a resulting macro map.
2180 (default_diagnostic_starter): Pass the relevant diagnostic
2181 location to diagnostic_report_current_module.
2182 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): New.
2183 (virt_loc_aware_diagnostic_finalizer): Likewise.
2184 (diagnostic_report_current_function): Pass the
2185 relevant location to diagnostic_report_current_module.
2186 * tree-diagnostic.h (virt_loc_aware_diagnostic_finalizer): Declare
2188 * toplev.c (general_init): By default, use the new
2189 virt_loc_aware_diagnostic_finalizer as diagnostic finalizer.
2190 * Makefile.in: Add vec.h dependency to tree-diagnostic.c.
2192 2011-10-15 Tom Tromey <tromey@redhat.com>
2193 Dodji Seketeli <dodji@redhat.com>
2195 * doc/cppopts.texi (-ftrack-macro-expansion): Document new option.
2196 * doc/invoke.texi (-ftrack-macro-expansion): Add this to the list of
2197 preprocessor related options.
2199 2011-10-15 Tom Tromey <tromey@redhat>
2200 Dodji Seketeli <dodji@redhat.com>
2202 * input.h (struct expanded_location): Move to libcpp/line-map.h.
2203 (LOCATION_COLUMN): New accessor
2204 (in_system_header_at): Use linemap_location_in_system_header_p.
2205 * diagnostic.c (diagnostic_report_current_module): Adjust to avoid
2206 touching the internals of struct line_map. Use the public API instead.
2207 (diagnostic_report_diagnostic): Don't use relational operator '<'
2208 on virtual locations. Use linemap_location_before_p instead.
2209 * input.c (expand_location): Adjust to expand to the tokens'
2210 spelling location when macro location tracking is on.
2213 2011-10-08 Andi Kleen <ak@linux.intel.com>
2215 * ggc-page.c (GGC_QUIRE_SIZE): Increase to 512
2217 2011-10-13 Andi Kleen <ak@linux.intel.com>
2219 * toplev.c (compile_file): Rename __gnu_slim_lto to __gnu_lto_slim.
2221 2011-10-16 Ira Rosen <ira.rosen@linaro.org>
2223 PR tree-optimization/50727
2224 * tree-vect-patterns.c (vect_operation_fits_smaller_type): Add
2225 DEF_STMT to the list of statements to be replaced by the
2228 2011-10-16 Eric Botcazou <ebotcazou@adacore.com>
2230 PR rtl-optimization/50615
2231 * combine.c (distribute_notes) <REG_ARGS_SIZE>: Skip if I3 is a no-op.
2233 2011-10-16 Jakub Jelinek <jakub@redhat.com>
2235 PR tree-optimization/50596
2236 * tree-vectorizer.h (NUM_PATTERNS): Increase to 7.
2237 * tree-vect-patterns.c (vect_vect_recog_func_ptrs): Add
2238 vect_recog_bool_pattern.
2239 (check_bool_pattern, adjust_bool_pattern_cast,
2240 adjust_bool_pattern, vect_recog_bool_pattern): New functions.
2242 2011-10-16 Ira Rosen <ira.rosen@linaro.org>
2244 * tree-vect-stmts.c (vectorizable_load): For SLP without permutation
2245 treat the first load of the node as the first element in its
2247 * tree-vect-slp.c (vect_get_and_check_slp_defs): Swap the operands if
2248 necessary and possible.
2249 (vect_build_slp_tree): Add new argument. Allow load groups of any size
2250 in basic blocks. Keep all the loads for further permutation check.
2251 Use the new argument to determine if there is a permutation. Update
2252 the recursive calls.
2253 (vect_supported_load_permutation_p): Allow subchains of interleaving
2254 chains in basic block vectorization.
2255 (vect_analyze_slp_instance): Update the call to vect_build_slp_tree.
2256 Check load permutation based on the new parameter.
2257 (vect_schedule_slp_instance): Don't start from the first element in
2258 interleaving chain unless the loads are permuted.
2260 2011-10-15 Jan Hubicka <jh@suse.cz>
2264 * cgraphunit.c (cgraph_expand_function): Expand thunks and alises
2265 after function body.
2267 2011-10-15 Richard Henderson <rth@redhat.com>
2269 * tree-vect-slp.c: Include langhooks.h.
2270 (vect_create_mask_and_perm): Emit VEC_PERM_EXPR, not a builtin.
2271 (vect_transform_slp_perm_load): Use can_vec_perm_expr_p. Simplify
2272 mask creation for VEC_PERM_EXPR.
2273 * tree-vect-stmts.c (perm_mask_for_reverse): Return the mask,
2275 (reverse_vec_elements): Emit VEC_PERM_EXPR not a builtin.
2276 * Makefile.in (tree-vect-slp.o): Update dependency.
2277 * optabs.c (can_vec_perm_expr_p): Allow NULL as unknown constant.
2279 2011-10-15 Alan Modra <amodra@gmail.com>
2282 * ifcvt.c (dead_or_predicable): Revert accidental commit with
2283 HAVE_simple_return test.
2285 2011-10-15 Alan Modra <amodra@gmail.com>
2287 * ifcvt.c (dead_or_predicable): Disable if-conversion when
2288 doing so is likely to kill a shrink-wrapping opportunity.
2290 PR rtl-optimization/49941
2291 * jump.c (mark_jump_label_1): Set JUMP_LABEL for simple_return jumps.
2293 * rtl.h (set_return_jump_label): Declare.
2294 * function.c (set_return_jump_label): New function, extracted..
2295 (thread_prologue_and_epilogue_insns): ..from here. Use it in
2296 another instance to set return jump_label.
2297 * cfgrtl.c (force_nonfallthru_and_redirect): Use set_return_jump_label.
2298 * reorg.c (find_end_label): Likewise.
2300 2011-10-14 David S. Miller <davem@davemloft.net>
2302 * config/sparc/sol2.h: Protect -m{cpu,tune}=native handling
2303 with a more complete cpp test.
2304 * config/sparc/linux64.h: Likewise.
2305 * config/sparc/linux.h: Likewise.
2306 * config/sparc/sparc.opt (sparc_debug): New target variable.
2307 (mdebug): New target option.
2308 * config/sparc/sparc.h (MASK_DEBUG_OPTIONS, MASK_DEBUG_ALL,
2309 TARGET_DEBUG_OPTIONS): New defines.
2310 * config/sparc/sparc.c (debug_target_flag_bits,
2311 debug_target_flags): New functions.
2312 (sparc_option_override): Add name strings back to cpu_table[].
2313 Parse -mdebug string. When TARGET_DEBUG_OPTIONS is true, print
2314 out the target flags before and after override processing as well
2315 as the selected cpu. If MASK_V8PLUS, make sure that the selected
2318 2011-10-15 Oleg Endo <oleg.endo@t-online.de>
2321 * config/sh/sh.h (ZERO_EXTRACT_ANDMASK): New macro.
2322 * config/sh/sh.c (sh_rtx_costs): Add test instruction case.
2323 * config/sh/sh.md (tstsi_t): Name existing insn. Make inner
2324 and instruction commutative.
2325 (tsthi_t, tstqi_t, tstqi_t_zero, tstsi_t_and_not,
2326 tstsi_t_zero_extract_eq, tstsi_t_zero_extract_xor,
2327 tstsi_t_zero_extract_subreg_xor_little,
2328 tstsi_t_zero_extract_subreg_xor_big): New insns.
2329 (*movsicc_t_false, *movsicc_t_true): Replace space with tab in
2331 (*andsi_compact): Reorder alternatives so that K08 is considered first.
2333 2011-10-14 Eric Botcazou <ebotcazou@adacore.com>
2336 * config/sparc/linux64.h (TARGET_DEFAULT): Only override if the default
2337 processor is at least V9 and TARGET_64BIT_DEFAULT is defined.
2339 2011-10-14 Gerald Pfeifer <gerald@pfeifer.com>
2341 * invoke.texi (AVR Options): Avoid \leq{}.
2343 2011-10-14 Kai Tietz <ktietz@redhat.com>
2345 * gimplify.c (gimplify_expr): Take care that for bitwise-binary
2346 transformation the operands have compatible types.
2348 2011-10-14 Jakub Jelinek <jakub@redhat.com>
2350 * config/i386/sse.md (vec_widen_smult_hi_v8hi,
2351 vec_widen_smult_lo_v8hi, vec_widen_umult_hi_v8hi,
2352 vec_widen_umult_lo_v8hi): Macroize using VI2_AVX2
2353 mode iterator and any_extend code iterator.
2354 (vec_widen_<s>mult_hi_v8si, vec_widen_<s>mult_lo_v8si): New expanders.
2355 (vec_widen_smult_hi_v4si, vec_widen_smult_lo_v4si): Enable
2356 also for TARGET_SSE4_1 using pmuldq insn.
2357 (sdot_prodv8hi): Macroize using VI2_AVX2 iterator.
2358 (sse2_sse4_1): New code attr.
2359 (udot_prodv4si): Macroize using any_extend code iterator.
2360 (<s>dot_prodv8si): New expander.
2362 2011-10-14 Yakovlev Vladimir <vladimir.b.yakovlev@intel.com>
2364 * config/i386/i386.c (atom_cost): Changed cost for loading
2365 QImode using movzbl.
2367 2011-10-14 Michael Meissner <meissner@linux.vnet.ibm.com>
2369 * config/rs6000/rs6000.c (rs6000_init_builtins): Fix typo in my
2370 change on October 11th, 2011.
2372 2011-10-14 Jakub Jelinek <jakub@redhat.com>
2374 * config/i386/sse.md (vec_interleave_high<mode>,
2375 vec_interleave_low<mode>): Add AVX2 expanders for VI_256 modes.
2376 * config/i386/i386.c (expand_vec_perm_interleave3): New function.
2377 (ix86_expand_vec_perm_builtin_1): Call it.
2379 2011-10-14 Georg-Johann Lay <avr@gjlay.de>
2381 Fix thinko from r179765
2382 * config/avr/avr.c (avr_option_override): Don't override
2383 flag_omit_frame_pointer if not actually needed.
2385 2011-10-14 Georg-Johann Lay <avr@gjlay.de>
2388 * doc/invoke.texi (AVR Options): Document -mstrict-X.
2390 * config/avr/avr.opt (-mstrict-X): New option.
2391 (avr_strict_X): New variable reflecting -mstrict-X.
2392 * config/avr/avr.c (avr_reg_ok_for_addr_p): Add parameter
2393 outer_code and pass it down to avr_regno_mode_code_ok_for_base_p.
2394 (avr_legitimate_address_p): Pass outer_code to
2395 avr_reg_ok_for_addr_p and use that function in case PLUS.
2396 (avr_mode_code_base_reg_class): Depend on avr_strict_X.
2397 (avr_regno_mode_code_ok_for_base_p): Ditto, and depend on outer_code.
2398 (avr_option_override): Disable -fcaller-saves if -mstrict-X is on.
2400 2011-10-14 Jakub Jelinek <jakub@redhat.com>
2402 * config/i386/sse.md (neg<mode>2): Use VI_AVX2 iterator instead
2405 * config/i386/sse.md (mulv2di3): Macroize using VI8_AVX2 iterator.
2406 (ashl<mode>3): Use VI248_AVX2 iterator instead of VI248_128.
2407 Use <sseinsnmode> instead of TI in mode attr.
2409 2011-10-14 David Alan Gilbert <david.gilbert@linaro.org>
2411 * config/arm/linux-atomic-64bit.c: New (based on linux-atomic.c).
2412 * config/arm/linux-atomic.c: Change comment to point to 64bit version.
2413 (SYNC_LOCK_RELEASE): Instantiate 64bit version.
2414 * config/arm/t-linux-eabi: Pull in linux-atomic-64bit.c.
2416 2011-10-14 David Alan Gilbert <david.gilbert@linaro.org>
2418 * config/arm/arm.c (arm_output_ldrex): Support ldrexd.
2419 (arm_output_strex): Support strexd.
2420 (arm_output_it): New helper to output it in Thumb2 mode only.
2421 (arm_output_sync_loop): Support DI mode. Change comment to
2422 not support const_int.
2423 (arm_expand_sync): Support DI mode.
2424 * config/arm/arm.h (TARGET_HAVE_LDREXBHD): Split into LDREXBH
2426 * config/arm/iterators.md (NARROW): move from sync.md.
2427 (QHSD): New iterator for all current ARM integer modes.
2428 (SIDI): New iterator for SI and DI modes only.
2429 * config/arm/sync.md (sync_predtab): New mode_attr.
2430 (sync_compare_and_swapsi): Fold into sync_compare_and_swap<mode>.
2431 (sync_lock_test_and_setsi): Fold into sync_lock_test_and_setsi<mode>.
2432 (sync_<sync_optab>si): Fold into sync_<sync_optab><mode>.
2433 (sync_nandsi): Fold into sync_nand<mode>.
2434 (sync_new_<sync_optab>si): Fold into sync_new_<sync_optab><mode>.
2435 (sync_new_nandsi): Fold into sync_new_nand<mode>.
2436 (sync_old_<sync_optab>si): Fold into sync_old_<sync_optab><mode>.
2437 (sync_old_nandsi): Fold into sync_old_nand<mode>.
2438 (sync_compare_and_swap<mode>): Support SI & DI.
2439 (sync_lock_test_and_set<mode>): Likewise.
2440 (sync_<sync_optab><mode>): Likewise.
2441 (sync_nand<mode>): Likewise.
2442 (sync_new_<sync_optab><mode>): Likewise.
2443 (sync_new_nand<mode>): Likewise.
2444 (sync_old_<sync_optab><mode>): Likewise.
2445 (sync_old_nand<mode>): Likewise.
2446 (arm_sync_compare_and_swapsi): Turn into iterator on SI & DI.
2447 (arm_sync_lock_test_and_setsi): Likewise.
2448 (arm_sync_new_<sync_optab>si): Likewise.
2449 (arm_sync_new_nandsi): Likewise.
2450 (arm_sync_old_<sync_optab>si): Likewise.
2451 (arm_sync_old_nandsi): Likewise.
2452 (arm_sync_compare_and_swap<mode> NARROW): use sync_predtab, fix indent.
2453 (arm_sync_lock_test_and_setsi<mode> NARROW): Likewise.
2454 (arm_sync_new_<sync_optab><mode> NARROW): Likewise.
2455 (arm_sync_new_nand<mode> NARROW): Likewise.
2456 (arm_sync_old_<sync_optab><mode> NARROW): Likewise.
2457 (arm_sync_old_nand<mode> NARROW): Likewise.
2459 2011-10-14 David Alan Gilbert <david.gilbert@linaro.org>
2462 * config/arm/arm.c (arm_output_sync_loop): Move label before barrier.
2464 2011-10-14 David Alan Gilbert <david.gilbert@linaro.org>
2466 * config/arm/arm.h (TARGET_HAVE_DMB_MCR): MCR Not available in Thumb1.
2468 2011-10-14 Paolo Carlini <paolo.carlini@oracle.com>
2470 * doc/invoke.texi ([Wformat-zero-length]): Tidy.
2472 2011-10-14 Jakub Jelinek <jakub@redhat.com>
2474 * gimple.c (walk_stmt_load_store_addr_ops): Call visit_addr also
2475 on COND_EXPR/VEC_COND_EXPR comparison operands if they are ADDR_EXPRs.
2477 2011-10-14 Richard Guenther <rguenther@suse.de>
2479 PR tree-optimization/50723
2480 * ipa-split.c (split_function): Use GSI_CONTINUE_LINKING.
2482 2011-10-14 Nicola Pero <nicola.pero@meta-innovation.com>
2484 * gengtype.c (files_rules): Added rules for objc/objc-map.h and
2487 2011-10-14 Paolo Carlini <paolo.carlini@oracle.com>
2490 * doc/invoke.texi ([Wformat-zero-length]): Update.
2492 2011-10-14 Iain Sandoe <iains@gcc.gnu.org>
2495 * config/darwin.c (darwin_patch_builtin): Adjust argument type. Only
2496 build for powerpc targets. (darwin_patch_builtins): Only build for
2499 2011-10-14 Jakub Jelinek <jakub@redhat.com>
2501 * config/i386/sse.md (*avx_cvtdq2pd256_2): Rename to...
2502 (avx_cvtdq2pd256_2): ... this.
2503 (sseunpackfltmode): New mode attr.
2504 (vec_unpacks_float_hi_v8hi, vec_unpacks_float_lo_v8hi,
2505 vec_unpacku_float_hi_v8hi, vec_unpacku_float_lo_v8hi): Macroize
2506 using VI2_AVX2 iterator.
2507 (vec_unpacku_float_hi_v8si, vec_unpacku_float_lo_v8si): New expanders.
2509 2011-10-13 David S. Miller <davem@davemloft.net>
2511 * config/sparc/sparc.md (plusminus): New code iterator.
2512 (plusminus_insn): New code attr.
2513 (addv2si3, subv2si3, addv4hi3, subv4hi3, addv2hi3, subv2hi3): Merge
2514 using plusminus and plusminus_insn.
2515 (fpadd64_vis, fpsub64_vis): Likewise.
2517 2011-10-13 Richard Henderson <rth@redhat.com>
2519 * doc/md.texi (vec_perm): Document fallback to byte permutation.
2520 * genopinit.c (optabs): Add vec_perm_const.
2521 * optabs.c (can_vec_perm_expr_p): Rename from expand_vec_perm_expr_p.
2522 Reject non-vector modes. Allow fallback to byte permutation.
2523 (expand_vec_perm_expr_1): Split out from ...
2524 (expand_vec_perm_expr): ... here. Allow fallback to byte permutation.
2525 * optabs.h (DOI_vec_perm_const, vec_perm_const_optab): New.
2526 * tree-vect-generic.c (lower_vec_perm): Update for name change.
2528 2011-10-13 Richard Henderson <rth@redhat.com>
2530 * config/rs6000/altivec.md (vec_permv16qi): New pattern.
2532 * config/rs6000/spu.md (vec_permv16qi): New pattern.
2534 * config/i386/i386.c (ix86_expand_vec_perm_const): New.
2535 * config/i386/i386-protos.h: Update.
2536 * config/i386/sse.md (VEC_PERM_CONST): New mode iterator.
2537 (vec_perm_const<VEC_PERM_CONST>): New expander.
2539 * optabs.c (expand_vector_broadcast): New.
2540 (expand_binop): Expand scalar shifts of vectors to vector shifts
2541 of vectors, if the former isn't supported.
2542 * tree-vect-generic.c (expand_vector_operations_1): Don't do that
2543 here; always simplify to scalar shift of vector if possible.
2545 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Fix mode
2546 test for vector splat.
2548 2011-10-13 Jakub Jelinek <jakub@redhat.com>
2550 * config/i386/sse.md (vec_set<mode>): Change V_128 iterator mode to V.
2552 2011-10-13 Jakub Jelinek <jakub@redhat.com>
2553 Richard Guenther <rguenther@suse.de>
2555 * tree-ssa.c (maybe_optimize_var): Drop TREE_ADDRESSABLE from vector
2556 or complex vars even if their DECL_UID is in not_reg_needs bitmap.
2558 2011-10-13 Jakub Jelinek <jakub@redhat.com>
2560 * config/i386/sse.md (reduc_umin_v8hi): New pattern.
2561 * config/i386/i386.c (ix86_build_const_vector): Handle
2562 also V32QI, V16QI, V16HI and V8HI modes.
2563 (emit_reduc_half): New function.
2564 (ix86_expand_reduc): Use phminposuw insn for V8HImode UMIN.
2565 Use emit_reduc_half helper function.
2567 2011-10-13 Lawrence Crowl <crowl@google.com>
2568 Diego Novillo <dnovillo@google.com>
2570 * lto-streamer-in.c (input_struct_function_base): Factor out of ...
2571 (input_function): ... here.
2572 * lto-streamer-out.c (output_struct_function_base): Factor out of ...
2573 (output_function): ... here.
2575 2011-10-13 Gabriel Charette <gchare@google.com>
2576 Diego Novillo <dnovillo@google.com>
2578 * streamer-hooks.h (struct streamer_hooks): Add hooks
2579 input_location and output_location.
2580 * lto-streamer-in.c (lto_input_location): Use
2581 streamer_hooks.input_location, if set.
2582 * lto-streamer-out.c (lto_output_location): Use
2583 streamer_hooks.output_location, if set.
2585 2011-10-13 Eric Botcazou <ebotcazou@adacore.com>
2587 * doc/invoke.texi (SPARC options): Document -mfix-at697f.
2588 * config/sparc/sparc.opt (mfix-at697f): New option.
2589 * config/sparc/sparc.c (TARGET_MACHINE_DEPENDENT_REORG): Define.
2590 (sparc_reorg): New function.
2592 2011-10-13 Richard Guenther <rguenther@suse.de>
2594 PR tree-optimization/50712
2595 * ipa-split.c (split_function): Always re-gimplify parameters
2596 when they are not gimple vals before passing them. Properly
2597 check for type compatibility.
2599 2011-10-13 Tom de Vries <tom@codesourcery.com>
2601 * function.c (gimplify_parameters): Set number of arguments of call to
2602 BUILT_IN_ALLOCA_WITH_ALIGN to 2.
2604 2011-10-13 Tom de Vries <tom@codesourcery.com>
2606 * emit-rtl.c (set_mem_attributes_minus_bitpos): Set MEM_READONLY_P
2607 for static const strings.
2608 * varasm.c (build_constant_desc): Generate the memory location of the
2609 constant using gen_const_mem.
2611 2011-10-13 Richard Guenther <rguenther@suse.de>
2613 PR tree-optimization/50698
2614 * tree-data-ref.c (split_constant_offset_1): Also process
2617 2011-10-12 David S. Miller <davem@davemloft.net>
2619 * config/sparc/sparc.md (UNSPEC_FPMERGE): Delete.
2620 (UNSPEC_MUL16AU, UNSPEC_MUL8, UNSPEC_MUL8SU, UNSPEC_MULDSU): New
2622 (fpmerge_vis): Remove inaccurate comment, represent using vec_select
2624 (vec_interleave_lowv8qi, vec_interleave_highv8qi): New insns.
2625 (fmul8x16_vis, fmul8x16au_vis, fmul8sux16_vis, fmuld8sux16_vis):
2626 Reimplement as unspecs and remove inaccurate comments.
2627 (vis3_shift_patname): New code attr.
2628 (<vis3_shift_insn><vbits>_vis): Rename to
2629 "v<vis3_shift_patname><mode>3".
2630 (vis3_addsub_ss_patname): New code attr.
2631 (<vis3_addsub_ss_insn><vbits>_vis): Rename to
2632 "<vis3_addsub_ss_patname><mode>3".
2633 * config/sparc/sparc.c (sparc_vis_init_builtins): Update to
2634 accommodate pattern name changes.
2636 * config/sparc/sparc.h: Do not force TARGET_VIS3 and TARGET_FMAF
2637 to zero when assembler lacks support for such instructions.
2638 * config/sparc/sparc.c (sparc_option_override): Clear MASK_VIS3
2639 and MASK_FMAF in defaults when assembler lacks necessary support.
2641 2011-10-12 Jakub Jelinek <jakub@redhat.com>
2643 * config/i386/sse.md (vec_unpacks_lo_<mode>,
2644 vec_unpacks_hi_<mode>, vec_unpacku_lo_<mode>,
2645 vec_unpacku_hi_<mode>): Change VI124_128 mode to VI124_AVX2.
2646 * config/i386/i386.c (ix86_expand_sse_unpack): Handle
2647 V32QImode, V16HImode and V8SImode for TARGET_AVX2.
2649 * config/i386/sse.md (vec_avx2): New mode_attr.
2650 (mulv16qi3): Macroize to cover also mulv32qi3 for TARGET_AVX2 into ...
2651 (mul<mode>3): ... this.
2653 * config/i386/i386.md (UNSPEC_VPERMDI): Remove.
2654 * config/i386/i386.c (ix86_expand_vec_perm): Handle
2655 V16QImode and V32QImode for TARGET_AVX2.
2656 (MAX_VECT_LEN): Increase to 32.
2657 (expand_vec_perm_blend): Add support for 32-byte integer
2658 vectors with TARGET_AVX2.
2659 (valid_perm_using_mode_p): New function.
2660 (expand_vec_perm_pshufb): Add support for 32-byte integer
2661 vectors with TARGET_AVX2.
2662 (expand_vec_perm_vpshufb2_vpermq): New function.
2663 (expand_vec_perm_vpshufb2_vpermq_even_odd): New function.
2664 (expand_vec_perm_even_odd_1): Handle 32-byte integer vectors
2666 (ix86_expand_vec_perm_builtin_1): Try expand_vec_perm_vpshufb2_vpermq
2667 and expand_vec_perm_vpshufb2_vpermq_even_odd.
2668 * config/i386/sse.md (VEC_EXTRACT_EVENODD_MODE): Add for TARGET_AVX2
2669 32-byte integer vector modes.
2670 (vec_pack_trunc_<mode>): Use VI248_AVX2 instead of VI248_128.
2671 (avx2_interleave_highv32qi, avx2_interleave_lowv32qi): Remove pasto.
2672 (avx2_pshufdv3, avx2_pshuflwv3, avx2_pshufhwv3): Generate
2674 (avx2_pshufd_1, avx2_pshuflw_1, avx2_pshufhw_1): Don't use
2675 match_dup, instead add 4 new operands and require they have
2676 right cross-lane values.
2677 (avx2_permv4di): Change into define_expand.
2678 (avx2_permv4di_1): New instruction.
2679 (avx2_permv2ti): Use nonimmediate_operand instead of register_operand
2680 for "xm" constrained operand.
2681 (VEC_PERM_AVX2): Add V32QI and V16QI for TARGET_AVX2.
2683 * config/i386/sse.md (avx2_gathersi<mode>,
2684 avx2_gatherdi<mode>, avx2_gatherdi<mode>256): Add clobber of
2685 match_scratch, change memory_operand to register_operand,
2686 add (mem:BLK (scratch)) use.
2687 (*avx2_gathersi<mode>, *avx2_gatherdi<mode>,
2688 *avx2_gatherdi<mode>256): Add clobber of match_scratch,
2689 add earlyclobber to the output operand and match_scratch,
2690 add (mem:BLK (scratch)) use, change the other mem to match_operand.
2691 Use %p6 instead of %c6 in the pattern.
2692 * config/i386/i386.c (ix86_expand_builtin): Adjust for
2693 operand 2 being a Pmode register_operand instead of memory_operand.
2695 2011-10-12 Kai Tietz <ktietz@redhat.com>
2697 * config/i386/i386.md (simple_return): Disable if TARGET_SEH is active.
2699 2011-10-12 Steve Ellcey <sje@cup.hp.com>
2701 * config/ia64/ia64.c (ia64_init_builtins): Fix unbalanced parentheses.
2703 2011-10-12 Richard Guenther <rguenther@suse.de>
2705 * tree-ssa-alias.c (maybe_skip_until): Cache also at the point
2706 of the first store we visit in a basic-block.
2707 (get_continuation_for_phi): Search for a candidate VUSE that
2708 might dominates all others. Do pairwise disambiguation against
2711 2011-10-12 Paul Koning <pkoning@gcc.gnu.org>
2713 PR tree-optimization/50189
2714 * tree-vrp.c (extract_range_from_assert): Use the type of
2715 the variable, not the limit.
2717 2011-10-12 Richard Guenther <rguenther@suse.de>
2719 PR tree-optimization/50700
2720 * tree-object-size.c (addr_object_size): Simplify and treat
2721 MEM_REF bases consistently.
2723 2011-10-12 Bernd Schmidt <bernds@codesourcery.com>
2725 * function.c (prepare_shrink_wrap, bb_active_p): New function.
2726 (thread_prologue_and_epilogue_insns): Use bb_active_p. Call
2727 prepare_shrink_wrap, then recompute bb_active_p for the last block.
2729 2011-10-12 Joseph Myers <joseph@codesourcery.com>
2732 * convert.c (convert_to_integer): Do not narrow operands of
2733 pointer subtraction.
2735 2011-10-12 Nick Clifton <nickc@redhat.com>
2737 * config/arm/arm.h (EMIT_EABI_ATTRIBUTE): New macro. Used to
2738 emit a .eabi_attribute assembler directive, possibly with a
2740 * config/arm/arm.c (arm_file_start): Use the new macro.
2741 * config/arm/arm-c.c (arm_output_c_attributes): Likewise.
2743 2011-10-12 Georg-Johann Lay <avr@gjlay.de>
2746 * config/avr/avr.md (*movqi): Rename to movqi_insn.
2747 (*call_insn): Rename to call_insn.
2748 (*call_value_insn): Rename to call_value_insn.
2749 * config/avr/avr.c (avr_2word_insn_p): New static function.
2750 (jump_over_one_insn_p): Use it.
2752 2011-10-12 Richard Sandiford <richard.sandiford@linaro.org>
2754 * expr.h (copy_blkmode_to_reg): Declare.
2755 * expr.c (copy_blkmode_to_reg): New function.
2756 (expand_assignment): Don't expand register RESULT_DECLs before
2757 the lhs. Use copy_blkmode_to_reg to copy BLKmode values into a
2758 RESULT_DECL register.
2759 (expand_expr_real_1): Handle BLKmode decls when looking for promotion.
2760 * stmt.c (expand_return): Move BLKmode-to-register code into
2761 copy_blkmode_to_reg.
2763 2011-10-11 Eric Botcazou <ebotcazou@adacore.com>
2766 * config/sparc/sparc.md (mov<I:mode>cc): Do not save comparison code.
2767 (mov<F:mode>cc): Likewise.
2769 2011-10-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2771 * tree-ssa-address.c (copy_ref_info): Remove copy of TREE_THIS_NOTRAP.
2773 2011-10-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2775 * tree.h (built_in_decls): Delete old interface with two parallel
2776 arrays to hold standard builtin declarations, and replace it with
2777 a function based interface that can support creating builtins on
2778 the fly in the future. Change all uses, and poison the old
2779 names. Make sure 0 is not a legitimate builtin index.
2780 (implicit_built_in_decls): Ditto.
2781 (built_in_info): Ditto.
2782 (BUILTIN_VALID_P): Ditto.
2783 (builtin_decl_explicit): Ditto.
2784 (builtin_decl_implicit): Ditto.
2785 (set_builtin_decl): Ditto.
2786 (set_builtin_decl_implicit_p): Ditto.
2787 (builtin_decl_explicit_p): Ditto.
2788 (builtin_decl_implicit_p): Ditto.
2789 * tree-complex.c (expand_complex_libcall): Ditto.
2790 * tree-loop-distribution.c (generate_memset_zero): Ditto.
2791 * tree-ssa-strlen.c (get_string_length): Ditto.
2792 (handle_builtin_strcpy): Ditto.
2793 (handle_builtin_strcat): Ditto.
2794 * tree.c (iterative_hash_expr): Ditto.
2795 (local_define_builtin): Ditto.
2796 (build_common_builtin_nodes): Ditto.
2797 * builtins.c (built_in_decls): Ditto.
2798 (implicit_built_in_decls): Ditto.
2799 (built_in_info): Ditto
2800 (expand_builtin_classify_type): Ditto.
2801 (mathfn_built_in_1): Ditto.
2802 (expand_builtin_cexpi): Ditto.
2803 (expand_builtin_mempcpy_args): Ditto.
2804 (expand_builtin_stpcpy): Ditto.
2805 (gimplify_va_arg_expr): Ditto.
2806 (expand_builtin_sync_operation): Ditto.
2807 (build_builtin_expect_predicate): Ditto.
2808 (fold_builtin_memory_op): Ditto.
2809 (fold_builtin_strcpy): Ditto.
2810 (fold_builtin_stpcpy): Ditto.
2811 (fold_builtin_strncpy): Ditto.
2812 (fold_builtin_interclass_mathfn): Ditto.
2813 (fold_builtin_classify): Ditto.
2814 (fold_builtin_2): Ditto.
2815 (fold_builtin_strstr): Ditto.
2816 (fold_builtin_strrchr): Ditto.
2817 (fold_builtin_strpbrk): Ditto.
2818 (fold_builtin_strcat): Ditto.
2819 (fold_builtin_strncat): Ditto.
2820 (fold_builtin_strcspn): Ditto.
2821 (fold_builtin_fputs): Ditto.
2822 (fold_builtin_sprintf): Ditto.
2823 (fold_builtin_snprintf): Ditto.
2824 (expand_builtin_memory_chk): Ditto.
2825 (fold_builtin_memory_chk): Ditto.
2826 (fold_builtin_stxcpy_chk): Ditto.
2827 (fold_builtin_strncpy_chk): Ditto.
2828 (fold_builtin_strcat_chk): Ditto.
2829 (fold_builtin_strncat_chk): Ditto.
2830 (fold_builtin_sprintf_chk_1): Ditto.
2831 (fold_builtin_snprintf_chk_1): Ditto.
2832 (fold_builtin_printf): Ditto.
2833 (fold_builtin_fprintf): Ditto.
2834 (fold_call_stmt): Ditto.
2835 (set_builtin_user_assembler_name): Ditto.
2836 * tree-emutls.c (emutls_common_1): Ditto.
2837 * omp-low.c (scan_omp): Ditto.
2838 (lower_rec_input_clauses): Ditto.
2839 (lower_reduction_clauses): Ditto.
2840 (expand_parallel_call): Ditto.
2841 (expand_task_call): Ditto.
2842 (maybe_catch_exception): Ditto.
2843 (optimize_omp_library_calls): Ditto.
2844 (expand_omp_for_generic): Ditto.
2845 (expand_omp_for_static_nochunk): Ditto.
2846 (expand_omp_for_static_chunk): Ditto.
2847 (expand_omp_sections): Ditto.
2848 (expand_omp_atomic_fetch_op): Ditto.
2849 (expand_omp_atomic_pipeline): Ditto.
2850 (expand_omp_atomic_mutex): Ditto.
2851 (lower_omp_single_simple): Ditto.
2852 (lower_omp_single_copy): Ditto.
2853 (lower_omp_master): Ditto.
2854 (lower_omp_ordered): Ditto.
2855 (lower_omp_critical): Ditto.
2856 * tree-ssa-ccp.c (optimize_stdarg_builtin): Ditto.
2857 * builtins.c (DEF_BUILTIN_STUB): Ditto.
2858 (BUILT_IN_NONE): Ditto.
2859 * tree-ssa-math-opts.c (execute_optimize_bswap): Ditto.
2860 * gimple-low.c (lower_function_body): Ditto.
2861 (lower_builtin_setjmp): Ditto.
2862 * c-decl.c (merge_decls): Ditto.
2863 * tree-eh.c (lower_resx): Ditto.
2864 (lower_resx): Ditto.
2865 (lower_eh_dispatch): Ditto.
2866 * function (gimplify_parameters): Ditto.
2867 * c-typeck.c (build_function_call_vec): Ditto.
2868 * gimplify.c (build_stack_save_restore): Ditto.
2869 (gimplify_vla_decl): Ditto.
2870 (gimplify_modify_expr_to_memcpy): Ditto.
2871 (gimplify_modify_expr_to_memset): Ditto.
2872 (gimplify_variable_sized_compare): Ditto.
2873 (gimplify_function_tree): Ditto.
2874 * calls.c (emit_call_1): Ditto.
2875 * tree-ssa-forprop.c (simplify_builtin_call): Ditto.
2876 * tree-nested.c (convert_nl_goto_reference): Ditto.
2877 (convert_tramp_reference_op): Ditto.
2878 (finalize_nesting_tree_1): Ditto.
2879 * tree-ssa-loop-prefetch.c (issue_prefetch_ref): Ditto.
2880 (tree_ssa_prefetch_arrays): Ditto.
2881 * tree-streamer-in.c (streamer_get_builtin_tree): Ditto.
2882 * system.h (built_in_decls): Ditto.
2883 (implicit_built_in_decls): Ditto.
2884 * tree-vect-generic.c (expand_vector_operations_1): Ditto.
2885 * config/sparc/sparc.c (sparc_gimplify_va_arg): Ditto.
2886 * config/i386/i386.c (ix86_gimplify_va_arg): Ditto.
2887 (ix86_veclibabi_svml): Ditto.
2888 (ix86_veclibabi_acml): Ditto.
2889 * config/vms/vms.c (vms_patch_builtins): Ditto.
2890 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
2891 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Ditto.
2892 (rs6000_builtin_vectorized_libmass): Ditto.
2893 (rs6000_init_builtins): Ditto.
2894 * config/darwin.c (darwin_override_options): Ditto.
2895 (darwin_patch_builtin): Ditto.
2896 (darwin_rename_builtins): Ditto.
2897 * config/pa/pa.c (pa_init_builtins): Ditto.
2899 2011-10-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2901 * tree.h (copy_ref_info): Expose existing function.
2902 * tree-ssa-loop-ivopts.c (copy_ref_info): Move function to...
2903 * tree-ssa-address.c (copy_ref_info): ...here, and remove static token.
2905 2011-10-11 Georg-Johann Lay <avr@gjlay.de>
2907 * config/avr/avr-protos.h (avr_mode_code_base_reg_class): New
2909 (avr_regno_mode_code_ok_for_base_p): New prototype.
2910 * config/avr/avr.h (BASE_REG_CLASS): Remove.
2911 (REGNO_OK_FOR_BASE_P): Remove.
2912 (REG_OK_FOR_BASE_NOSTRICT_P): Remove.
2913 (REG_OK_FOR_BASE_STRICT_P): Remove.
2914 (MODE_CODE_BASE_REG_CLASS): New define.
2915 (REGNO_MODE_CODE_OK_FOR_BASE_P): New define.
2916 * config/avr/avr.c (avr_mode_code_base_reg_class): New function.
2917 (avr_regno_mode_code_ok_for_base_p): New function.
2918 (avr_reg_ok_for_addr_p): New static function.
2919 (avr_legitimate_address_p): Use it. Beautify.
2921 2011-10-11 Georg-Johann Lay <avr@gjlay.de>
2924 * config/avr/avr.md (cc): Add out_plus attribute alternative.
2925 (addsi3): Use it. Adapt avr_out_plus to new prototype. Use
2926 avr_out_plus for all CONST_INT addends.
2927 * config/avr/avr-protos.h (avr_out_plus): Change prototype.
2928 * config/avr/avr.c (notice_update_cc): Call avr_out_plus on
2930 (avr_out_plus_1): Change prototype and report effect on cc0.
2931 (avr_out_plus): Ditto.
2932 (adjust_insn_length): Adapt call to avr_out_plus to new prototype.
2934 2011-10-11 H.J. Lu <hongjiu.lu@intel.com>
2936 * config/i386/i386.c (ix86_expand_special_args_builtin): Remove
2939 2011-10-11 Artjoms Sinkarovs <artyom.shinkaroff@gmail.com>
2941 * doc/invoke.texi: Document new warning.
2942 * common.opt (Wvector-operation-performance): Define new warning.
2943 * tree-vect-generic.c (expand_vector_piecewise): Warn about expanded
2945 (exapnd_vector_parallel): Warn about expanded vector operation.
2946 (lower_vec_shuffle): Warn about expanded vector operation.
2947 * c-typeck.c (c_build_vec_perm_expr): Store correct location
2948 when creating VEC_PERM_EXPR.
2950 2011-10-11 Richard Guenther <rguenther@suse.de>
2952 PR tree-optimization/50204
2953 * tree-ssa-alias.c (get_continuation_for_phi_1): Split out
2954 two argument handling from ...
2955 (get_continuation_for_phi): ... here. Handle arbitrary number
2958 2011-10-11 Richard Sandiford <richard.sandiford@linaro.org>
2960 * modulo-sched.c: Fix comment typo. Mention the possibility
2961 of using scheduling windows of II+1 cycles.
2963 2011-10-11 Tristan Gingold <gingold@adacore.com>
2965 * doc/invoke.texi (C Dialect Options): Document
2966 -fallow-parameterless-variadic-functions.
2967 * c-parser.c (c_parser_parms_list_declarator): Handle it.
2969 2011-10-10 Georg-Johann Lay <avr@gjlay.de>
2971 * config/avr/avr.c (avr_option_override): Set flag_omit_frame_pointer
2972 to 0 if frame pointer is needed for unwinding.
2974 2011-10-10 Uros Bizjak <ubizjak@gmail.com>
2977 * optabs.h (DOI_vec_perm): Rename from OTI_vec_perm. Move from enum
2978 optab_index to enum direct_optab_index.
2979 (vec_perm_optab): Update.
2981 2011-10-10 Anatoly Sokolov <aesok@post.ru>
2983 * config/cris/cris.c (cris_preferred_reload_class): New function.
2984 (TARGET_PREFERRED_RELOAD_CLASS): Define.
2985 * config/cris/cris.h (PREFERRED_RELOAD_CLASS): Remove.
2987 2011-10-10 Georg-Johann Lay <avr@gjlay.de>
2989 * config/avr/avr.md (*tablejump_rjmp): Change insn condition to
2991 (*tablejump_lib): Change insn condition to AVR_HAVE_JMP_CALL.
2992 (*tablejump_enh, *tablejump): Remove insns.
2993 * config/avr/libgcc.S (__tablejump__): Use RET instead of EIND +
2994 EIJMP for indirect jump. Use LPM Z+ where available.
2996 2011-10-10 Richard Henderson <rth@redhat.com>
2998 * doc/md.texi (vec_perm_const): Fix typo in cindex.
3000 2011-10-10 Kirill Yukhin <kirill.yukhin@intel.com>
3001 Yakovlev Vladimir <vladimir.b.yakovlev@intel.com>
3003 * config/i386/sse.md (fma_fnmsub_<mode>): Fix a typo.
3005 2011-10-10 Richard Guenther <rguenther@suse.de>
3008 * gimple-fold.c (gimplify_and_update_call_from_tree): Do not
3009 mark symbols for renaming. Append the VUSE to all statements
3010 that possibly can have one.
3012 2011-10-10 Richard Guenther <rguenther@suse.de>
3014 * ipa-split.c (pass_split_functions): Add verification TODOs.
3015 (pass_feedback_split_functions): Likewise.
3017 2011-10-10 Richard Guenther <rguenther@suse.de>
3020 * fold-const.c (fold_binary_loc): Canonicalize x*x to pow (x, 2)
3021 only when optimizing.
3023 2011-10-10 Nick Clifton <nickc@redhat.com>
3026 * compare-elim.c (find_comparisons_in_bb): Use df_get_live_in and
3027 df_get_live_out instead of accessing the bitmaps directly.
3028 (execute_compare_elim_after_reload): Remove calls to df_set_flags,
3029 df_live_add_problem and df_analyze.
3031 2011-10-10 Michael Matz <matz@suse.de>
3034 * tree-emutls.c (gen_emutls_addr): Call add_referenced_var.
3036 2011-10-10 Richard Sandiford <richard.sandiford@linaro.org>
3038 * modulo-sched.c (ps_reg_move_info): Add num_consecutive_stages.
3039 (SCHED_FIRST_REG_MOVE, SCHED_NREG_MOVES): Delete.
3040 (node_sched_params): Remove first_reg_move and nreg_moves.
3041 (ps_num_consecutive_stages, extend_node_sched_params): New functions.
3042 (update_node_sched_params): Move up file.
3043 (print_node_sched_params): Print the stage. Don't dump info related
3044 to first_reg_move and nreg_moves.
3045 (set_columns_for_row): New function.
3046 (set_columns_for_ps): Move up file and use set_columns_for_row.
3047 (schedule_reg_move): New function.
3048 (schedule_reg_moves): Call extend_node_sched_params and
3049 schedule_reg_move. Extend size of uses bitmap. Initialize
3050 num_consecutive_stages. Return false if a move could not be
3052 (apply_reg_moves): Don't emit moves here.
3053 (permute_partial_schedule): Handle register moves.
3054 (duplicate_insns_of_cycles): Remove for_prolog. Emit moves according
3055 to the same stage-count test as ddg nodes.
3056 (generate_prolog_epilog): Update calls accordingly.
3057 (sms_schedule): Allow move-scheduling to add a new first stage.
3059 2011-10-10 Richard Sandiford <richard.sandiford@linaro.org>
3061 * modulo-sched.c (ps_insn): Adjust comment.
3062 (ps_reg_move_info): New structure.
3063 (partial_schedule): Add reg_moves field.
3064 (SCHED_PARAMS): Use node_sched_param_vec instead of node_sched_params.
3065 (node_sched_params): Turn first_reg_move into an identifier.
3066 (ps_reg_move): New function.
3067 (ps_rtl_insn): Cope with register moves.
3068 (ps_first_note): Adjust comment and assert that the instruction
3069 isn't a register move.
3070 (node_sched_params): Replace with...
3071 (node_sched_param_vec): ...this vector.
3072 (set_node_sched_params): Adjust accordingly.
3073 (print_node_sched_params): Take a partial schedule instead of a ddg.
3074 Use ps_rtl_insn and ps_reg_move.
3075 (generate_reg_moves): Rename to...
3076 (schedule_reg_moves): ...this. Remove rescan parameter. Record each
3077 move in the partial schedule, but don't emit it here. Don't perform
3078 register substitutions here either.
3079 (apply_reg_moves): New function.
3080 (duplicate_insns_of_cycles): Use register indices directly,
3081 rather than finding instructions using PREV_INSN. Use ps_reg_move.
3082 (sms_schedule): Call schedule_reg_moves before committing to
3083 a partial schedule. Try the next ii if the schedule fails.
3084 Use apply_reg_moves instead of generate_reg_moves. Adjust
3085 call to print_node_sched_params. Free node_sched_param_vec
3086 instead of node_sched_params.
3087 (create_partial_schedule): Initialize reg_moves.
3088 (free_partial_schedule): Free reg_moves.
3090 2011-10-10 Richard Sandiford <richard.sandiford@linaro.org>
3092 * modulo-sched.c (ps_insn): Replace node field with an identifier.
3093 (SCHED_ASAP): Replace with..
3094 (NODE_ASAP): ...this macro.
3095 (SCHED_PARAMS): New macro.
3096 (SCHED_TIME, SCHED_FIRST_REG_MOVE, SCHED_NREG_MOVES, SCHED_ROW)
3097 (SCHED_STAGE, SCHED_COLUMN): Redefine using SCHED_PARAMS.
3098 (node_sched_params): Remove asap.
3099 (ps_rtl_insn, ps_first_note): New functions.
3100 (set_node_sched_params): Use XCNEWVEC. Don't copy across the
3102 (print_node_sched_params): Use SCHED_PARAMS and NODE_ASAP.
3103 (generate_reg_moves): Pass ids to the SCHED_* macros.
3104 (update_node_sched_params): Take a ps insn identifier rather than
3105 a node as parameter. Use ps_rtl_insn.
3106 (set_columns_for_ps): Update for above field and SCHED_* macro changes.
3107 (permute_partial_schedule): Use ps_rtl_insn and ps_first_note.
3108 (optimize_sc): Update for above field and SCHED_* macro changes.
3109 Update calls to try_scheduling_node_in_cycle and
3110 update_node_sched_params.
3111 (duplicate_insns_of_cycles): Adjust for above field and SCHED_*
3112 macro changes. Use ps_rtl_insn and ps_first_note.
3113 (sms_schedule): Pass ids to the SCHED_* macros.
3114 (get_sched_window): Adjust for above field and SCHED_* macro changes.
3115 Use NODE_ASAP instead of SCHED_ASAP.
3116 (try_scheduling_node_in_cycle): Remove node parameter. Update
3117 call to ps_add_node_check_conflicts. Pass ids to the SCHED_* macros.
3118 (sms_schedule_by_order): Update call to try_scheduling_node_in_cycle.
3119 (ps_insert_empty_row): Adjust for above field changes.
3120 (compute_split_row): Use ids rather than nodes.
3121 (verify_partial_schedule): Adjust for above field changes.
3122 (print_partial_schedule): Use ps_rtl_insn.
3123 (create_ps_insn): Take an id rather than a node.
3124 (ps_insn_find_column): Adjust for above field changes.
3126 (ps_insn_advance_column): Adjust for above field changes.
3127 (add_node_to_ps): Remove node parameter. Update call to
3129 (ps_has_conflicts): Use ps_rtl_insn.
3130 (ps_add_node_check_conflicts): Replace node parameter than an id.
3132 2011-10-10 Richard Sandiford <richard.sandiford@linaro.org>
3134 * modulo-sched.c (undo_replace_buff_elem): Delete.
3135 (generate_reg_moves): Don't build and return an undo list.
3136 (free_undo_replace_buff): Delete.
3137 (sms_schedule): Adjust call to generate_reg_moves.
3138 Don't call free_undo_replace_buff.
3140 2011-10-10 Matthias Klose <doko@ubuntu.com>
3142 * common/config/m32c: Remove empty directory.
3144 2011-10-10 Georg-Johann Lay <avr@gjlay.de>
3146 * config/avr/avr.md (*mulhi3_enh): Treat squaring smarter.
3148 2011-10-10 Georg-Johann Lay <avr@gjlay.de>
3151 * config/avr/avr-mcus.def (AVR_MCU): Set .data_section_start of
3152 atmega164a to 0x100.
3154 2011-10-09 Eric Botcazou <ebotcazou@adacore.com>
3156 * tree-ssa-forwprop.c (forward_propagate_into_comparison): Do not make
3157 the replacement if the conversion to the LHS type is not useless.
3159 2011-10-09 Ira Rosen <ira.rosen@linaro.org>
3161 PR tree-optimization/50635
3162 * tree-vect-patterns.c (vect_handle_widen_mult_by_const): Add
3163 DEF_STMT to the list of statements to be replaced by the
3165 (vect_handle_widen_mult_by_const): Don't check TYPE_OUT.
3167 2011-10-09 Anatoly Sokolov <aesok@post.ru>
3169 * system.h: Commit forgotten hunk in previous patch.
3170 (OUTPUT_ADDR_CONST_EXTRA): Poison.
3172 2011-10-08 Nicola Pero <nicola.pero@meta-innovation.com>
3175 * doc/objc.texi (Garbage Collection): Updated example to protect
3176 +initialize against execution in subclasses.
3178 2011-10-07 Richard Henderson <rth@redhat.com>
3180 * doc/extend.texi (__builtin_shuffle): Improve the description to
3181 include the modulus of the selector. Mention OpenCL.
3182 * doc/md.texi (vec_perm, vec_perm_const): Document named patterns.
3184 * tree.def (VEC_PERM_EXPR): Rename from VEC_SHUFFLE_EXPR.
3185 * genopinit.c (optabs): Rename vshuffle to vec_perm.
3186 * c-typeck.c (c_build_vec_perm_expr): Rename from
3187 c_build_vec_shuffle_expr. Update for name changes.
3188 * optabs.c (expand_vec_perm_expr_p): Rename from
3189 expand_vec_shuffle_expr_p.
3190 (expand_vec_perm_expr): Rename from expand_vec_shuffle_expr.
3191 * optabs.h (OTI_vec_perm): Rename from DOI_vshuffle.
3192 (vec_perm_optab): Rename from vshuffle_optab.
3193 * expr.c, gimple-pretty-print.c, gimple.c, gimplify.c,
3194 c-tree.h, c-parser.c, tree-cfg.c, tree-inline.c, tree-pretty-print.c,
3195 tree-ssa-operands.c, tree-vect-generic.c: Update for name changes.
3197 * config/i386/i386.c (ix86_expand_vec_perm): Rename from
3198 ix86_expand_vshuffle.
3199 * config/i386/i386-protos.h: Update.
3200 * config/i386/sse.md (VEC_PERM_AVX2): Rename from VSHUFFLE_AVX2.
3201 (vec_perm<VEC_PERM_AVX2>): Rename from vshuffle<VSHUFFLE_AVX2>.
3203 2011-10-07 Richard Henderson <rth@redhat.com>
3205 * config/i386/predicates.md (avx2_pblendw_operand): New.
3206 * config/i386/sse.md (sse4_1_pblendw): Un-macroize.
3207 (avx2_pblendw, *avx2_pblendw): New expander and insn.
3209 2011-10-07 Richard Henderson <rth@redhat.com>
3211 * config/i386/i386.c (bdesc_args): Update code for
3212 __builtin_ia32_palignr256. Change type of __builtin_ia32_pslldqi256,
3213 and __builtin_ia32_psrldqi256 to V4DI_FTYPE_V4DI_INT_CONVERT.
3214 (ix86_expand_args_builtin): Handle V4DI_FTYPE_V4DI_INT_CONVERT.
3215 * config/i386/sse.md (mode iterator V16): Add V2TI.
3216 (mode iterator SSESCALARMODE): Use V2TI not V4DI.
3217 (mode attr ssse3_avx2): Add V2TI.
3218 (avx2_lshrqv4di3, avx2_lshlqv4di3): Remove.
3220 2011-10-07 David S. Miller <davem@davemloft.net>
3223 * configure.ac: Add .register directives to VIS3 test.
3224 * configure: Regenerate.
3226 2011-10-07 Richard Henderson <rth@redhat.com>
3228 * config.gcc (x86_64-*): Add core-avx-i, core-avx2 for with_cpu.
3230 2011-10-07 Richard Henderson <rth@redhat.com>
3233 * fold-const.c (fold_checksum_tree): Remove out-of-date assert.
3235 2011-10-07 Andrew Stubbs <ams@codesourcery.com>
3237 * config/arm/predicates.md (shift_amount_operand): Remove constant
3239 (shift_operator): Check range of constants for all shift operators.
3241 2011-10-07 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
3243 * config/s390/s390.c (s390_emit_tls_call_insn): Remove assertion.
3244 Load GOT pointer for non-pic builds.
3245 (s390_load_got): Replace pic_offset_table_rtx with hardcoded r12.
3246 (s390_emit_call): Likewise.
3248 2011-10-07 Tom de Vries <tom@codesourcery.com>
3251 * tree.c (build_common_builtin_nodes): Add local_define_builtin for
3252 BUILT_IN_ALLOCA_WITH_ALIGN. Mark that BUILT_IN_ALLOCA_WITH_ALIGN can
3254 * builtins.c (expand_builtin_alloca): Handle BUILT_IN_ALLOCA_WITH_ALIGN
3255 arglist. Set align for BUILT_IN_ALLOCA_WITH_ALIGN.
3256 (expand_builtin): Handle BUILT_IN_ALLOCA_WITH_ALIGN.
3257 (is_inexpensive_builtin): Handle BUILT_IN_ALLOCA_WITH_ALIGN.
3258 * tree-ssa-ccp.c (evaluate_stmt): Set align for
3259 BUILT_IN_ALLOCA_WITH_ALIGN.
3260 (fold_builtin_alloca_for_var): Rename to ...
3261 (fold_builtin_alloca_with_align): Set DECL_ALIGN from 2nd
3262 BUILT_IN_ALLOCA_WITH_ALIGN argument.
3263 (ccp_fold_stmt): Try folding BUILT_IN_ALLOCA_WITH_ALIGN using
3264 fold_builtin_alloca_with_align.
3265 (optimize_stack_restore): Handle BUILT_IN_ALLOCA_WITH_ALIGN.
3266 * builtins.def (BUILT_IN_ALLOCA_WITH_ALIGN): Declare using
3268 * ipa-pure-const.c (special_builtin_state): Handle
3269 BUILT_IN_ALLOCA_WITH_ALIGN.
3270 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1)
3271 (call_may_clobber_ref_p_1): Same.
3272 * function.c (gimplify_parameters): Lower vla to
3273 BUILT_IN_ALLOCA_WITH_ALIGN.
3274 * gimplify.c (gimplify_vla_decl): Same.
3275 * cfgexpand.c (expand_call_stmt): Handle BUILT_IN_ALLOCA_WITH_ALIGN.
3276 * tree-mudflap.c (mf_xform_statements): Same.
3277 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary)
3278 (mark_all_reaching_defs_necessary_1, propagate_necessity): Same.
3279 * varasm.c (incorporeal_function_p): Same.
3280 * tree-object-size.c (alloc_object_size): Same.
3281 * gimple.c (gimple_build_call_from_tree): Same.
3283 2011-10-07 Bernd Schmidt <bernds@codesourcery.com>
3285 * function.c (frame_required_for_rtx): Remove function.
3286 (requires_stack_frame_p): New arg set_up_by_prologue. All callers
3287 changed. Compute a set of mentioned registers and compare against
3288 the new arg rather than calling frame_required_for_rtx.
3289 (thread_prologue_and_epilogue_insns): Compute the set_up_by_prologue
3290 reg set. Convert the unconverted_simple_returns mechanism to store
3291 jump insns rather than their basic blocks. Also check the
3292 orig_entry_edge destination for new blocks.
3294 2011-10-07 Jakub Jelinek <jakub@redhat.com>
3296 PR tree-optimization/50650
3297 * tree-vect-patterns.c (vect_recog_mixed_size_cond_pattern): Don't
3298 call vect_is_simple_cond here, instead fail if cond_expr isn't
3299 COMPARISON_CLASS_P or if get_vectype_for_scalar_type returns NULL
3300 for cond_expr's first operand.
3301 * tree-vect-stmts.c (vect_is_simple_cond): Static again.
3302 * tree-vectorizer.h (vect_is_simple_cond): Remove prototype.
3304 2011-10-07 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
3306 * config/s390/s390.md (DWH, dwh): New mode macros.
3307 ("umulsidi3"): Extend to support "umulditi3" as well.
3309 2011-10-07 Uros Bizjak <ubizjak@gmail.com>
3310 H.J. Lu <hongjiu.lu@intel.com>
3313 * config/i386/i386.c (ix86_fixup_binary_operands): Force src2 of
3314 integer PLUS RTX to a register to improve address combine.
3316 2011-10-06 Richard Henderson <rth@redhat.com>
3318 * combine-stack-adjust.c (maybe_move_args_size_note): Add after
3319 parameter; use it to decide whether to merge two notes.
3320 (combine_stack_adjustments_for_block): Use maybe_move_args_size_note
3321 for the deallocation case as well.
3323 2011-10-06 Anatoly Sokolov <aesok@post.ru>
3325 * system.h (OUTPUT_ADDR_CONST_EXTRA): Poison.
3326 * doc/tm.texi.in (OUTPUT_ADDR_CONST_EXTRA): Remove documentation.
3327 * doc/tm.texi: Regenerate.
3328 * target.def (output_addr_const_extra): Use
3329 hook_bool_FILEptr_rtx_false.
3330 * targhooks.c (default_asm_output_addr_const_extra): Remove.
3331 * targhooks.h (default_asm_output_addr_const_extra): Remove.
3332 * hooks.c (hook_bool_FILEptr_rtx_false): New functions.
3333 * hooks.h (hook_bool_FILEptr_rtx_false): Declare.
3335 2011-10-06 David S. Miller <davem@davemloft.net>
3337 * config/sparc/sparc.md (popcount<mode>2, clz<mode>2): Split up into...
3338 (popcountdi2, popcountsi2, clzdi2, clzsi2): Explicit expanders, in the
3339 SI mode 64-bit code gen case explicitly zero-extend and truncate.
3340 (*popcount<mode>_sp64): Split up into...
3341 (*popcountdi_sp64, *popcountsi_64): Explicit instantiations, and in the
3342 SImode case use truncate.
3343 (*clzsi_sp64): Rewrite to use truncate, and let the expander emit the
3344 subtract so the compiler can optimize it.
3345 (SIDI): Remove unused mode iterator.
3347 2011-10-06 Bernd Schmidt <bernds@codesourcery.com>
3349 * function.c (thread_prologue_and_epilogue_insns): Emit split prologue
3350 on the orig_entry_edge. Don't account for it in prologue_clobbered.
3352 2011-10-06 Jakub Jelinek <jakub@redhat.com>
3354 PR tree-optimization/50596
3355 * tree-vectorizer.h (vect_is_simple_cond): New prototype.
3356 (NUM_PATTERNS): Change to 6.
3357 * tree-vect-patterns.c (vect_recog_mixed_size_cond_pattern): New
3359 (vect_vect_recog_func_ptrs): Add vect_recog_mixed_size_cond_pattern.
3360 (vect_mark_pattern_stmts): Don't create stmt_vinfo for def_stmt
3361 if it already has one, and don't set STMT_VINFO_VECTYPE in it
3362 if it is already set.
3363 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Handle
3364 COND_EXPR in pattern stmts.
3365 (vect_is_simple_cond): No longer static.
3367 2001-10-06 Richard Henderson <rth@redhat.com>
3369 * config/i386/i386.c (ix86_expand_vshuffle): Add AVX2 support.
3370 * config/i386/sse.md (sseshuffint): Remove.
3371 (sseintvecmode): Support V16HI, V8HI, V32QI, V16QI.
3372 (VSHUFFLE_AVX2): New mode iterator.
3373 (vshuffle<mode>): Use it.
3374 (avx_vec_concat<V_256>): Rename from *vec_concat<V_256>_avx.
3376 * config/i386/i386.c (ix86_expand_sse_movcc): Use correct mode
3377 for vector_all_ones_operand.
3378 (ix86_expand_int_vcond): Distinguish between comparison mode
3379 and data mode. Allow them to differ.
3380 (ix86_expand_vshuffle): Don't force data mode to match maskmode.
3382 2001-10-06 Richard Henderson <rth@redhat.com>
3384 * optabs.c (expand_vec_shuffle_expr): Use the proper mode for the
3385 mask operand. Tidy the code.
3387 2011-10-06 Jakub Jelinek <jakub@redhat.com>
3389 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3390 vect_recog_func_ptr typedef for the first argument.
3391 (vect_pattern_recog): Rename vect_recog_func_ptr variable
3392 to vect_recog_func, use vect_recog_func_ptr typedef for it.