1 2011-11-01 Andrew Stubbs <ams@codesourcery.com>
3 * config/arm/bpabi.h (BE8_LINK_SPEC): Recognize generic-armv7 tuning.
5 2011-11-01 Uros Bizjak <ubizjak@gmail.com>
7 * config/i386/i386.md (splitters for int-float conversion): Use
8 reg_or_subregno in splitter constraints.
10 2011-11-01 Jakub Jelinek <jakub@redhat.com>
12 * config/i386/i386-protos.h (ix86_expand_adjust_ufix_to_sfix_si): New
14 * config/i386/i386.c (ix86_expand_adjust_ufix_to_sfix_si): New
16 * config/i386/sse.md (fixuns_trunc<mode><sseintvecmodelower>2): Use it.
17 (ssepackfltmode): New mode attr.
18 (vec_pack_ufix_trunc_<mode>): New expander.
20 2011-11-01 Uros Bizjak <ubizjak@gmail.com>
23 * config/i386/i386.md (floatsi<mode>2_vector_sse_with_temp splitter):
24 Compare <ssevecmode>mode with V4SFmode, not V4SImode.
26 2011-11-01 Peter Bergner <bergner@vnet.ibm.com>
28 * config.gcc (powerpc*-*-linux*): Add powerpc*-*-linux*ppc476* variant.
29 * config/rs6000/476.h: New file.
30 * config/rs6000/476.opt: Likewise.
31 * config/rs6000/rs6000.h (TARGET_LINK_STACK): New define.
32 (SET_TARGET_LINK_STACK): Likewise.
33 (TARGET_ASM_CODE_END): Define.
34 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
35 TARGET_LINK_STACK for -mtune=476 and -mtune=476fp.
36 (rs6000_legitimize_tls_address): Emit the link stack preserving GOT
37 code if TARGET_LINK_STACK.
38 (rs6000_emit_load_toc_table): Likewise.
39 (output_function_profiler): Likewise
40 (macho_branch_islands): Likewise
41 (machopic_output_stub): Likewise
42 (get_ppc476_thunk_name): New function.
43 (rs6000_code_end): Likewise.
44 * config/rs6000/rs6000.md (load_toc_v4_PIC_1, load_toc_v4_PIC_1b):
45 Convert to a define_expand.
46 (load_toc_v4_PIC_1_normal): New define_insn.
47 (load_toc_v4_PIC_1_476): Likewise.
48 (load_toc_v4_PIC_1b_normal): Likewise.
49 (load_toc_v4_PIC_1b_476): Likewise.
51 2011-11-01 Georg-Johann Lay <avr@gjlay.de>
54 * config/avr/avr.opt (-mbranch-cost=): New option.
55 * config/avr/avr.h (BRANCH_COST): Define to avr_branch_cost.
56 * config/avr/avr.c (avr_rtx_costs_1): Adjust [U]DIV/[U]MOD costs.
57 * config/avr/avr.md (*addqi3.lt0, *addhi3.lt0, *addsi3.lt0): New insns.
58 (*addhi3_zero_extend1): Remove % in constraint of operand 1.
59 (*addhi3.sign_extend1, *subhi3.sign_extend2): New insns.
61 2011-11-01 Tom de Vries <tom@codesourcery.com>
63 PR tree-optimization/50908
64 * tree-ssa-tail-merge.c (update_vuses): Now that edges are removed
65 before update_vuses, test for 1 predecessor rather than two.
66 (delete_block_update_dominator_info): New function, part of it factored
68 (replace_block_by): Use delete_block_update_dominator_info. Call
69 update_vuses after deleting bb1 and updating dominator info, instead of
72 2011-11-01 David S. Miller <davem@davemloft.net>
74 * config/sparc/sparc.c (sparc_expand_vcond): New function.
75 * config/sparc/sparc-protos.h (sparc_expand_vcond): Declare it.
76 * config/sparc/sparc.md (vcond<mode><mode>): New VIS3 expander.
77 (vconduv8qiv8qi): Likewise.
79 2011-11-01 Alexandre Oliva <aoliva@redhat.com>
82 * cselib.c (cfa_base_preserved_regno): Initialize.
83 (cselib_expand_value_rtx_1): Don't expand it.
84 * var-tracking.c (vt_expand_var_loc_chain): Initialize depth.
85 Check it's only zero if result is NULL.
87 2011-11-01 Jakub Jelinek <jakub@redhat.com>
89 * config/i386/sse.md (fixuns_trunc<mode><sseintvecmodelower>2): New
92 * config/i386/sse.md (sseintvecmode): Remove duplicate modes.
93 (sseintvecmodelower): New mode iterator.
94 (floatv8siv8sf2, floatunsv4siv4sf2): Macroize into...
95 (float<sseintvecmodelower><mode>2): ... this using VF1 iterator.
96 (floatunsv4siv4sf2): Macroize into...
97 (floatuns<sseintvecmodelower><mode>2): ... this using VF1 iterator.
99 2011-10-31 David S. Miller <davem@davemloft.net>
101 * config/sparc/sparc.md (cmask patterns): Allow zero operand.
103 * dwarf2out.c (cached_next_real_insn): New.
104 (dwarf2out_end_epilogue): Set it to NULL_RTX.
105 (dwarf2out_var_location): Remove cached_next_real_insn local static.
107 2011-10-31 Richard Henderson <rth@redhat.com>
109 * config/i386/sse.md (floatv8siv8sf2): Rename from avx_cvtdq2ps256.
110 (floatv4siv4sf2): Rename from sse2_cvtdq2ps.
111 (floatunsv4siv4sf2): Rename from sse2_cvtudq2ps.
112 (fix_truncv8sfv8si2): Rename from avx_cvttps2dq256.
113 (fix_truncv4sfv4si2): Rename from sse2_cvttps2dq.
114 (floatv4siv4df2): Rename from avx_cvtdq2pd256.
115 (fix_truncv4dfv4si2): Rename from avx_cvttpd2dq256.
116 (vec_unpacku_float_hi_v8si): Update for insn pattern name changes.
117 * config/i386/i386.md (splitters for int-float conversion): Likewise.
118 * config/i386/i386.c (ix86_split_convert_uns_si_sse): Likewise.
119 (bdesc_args): Likewise.
120 (enum ix86_builtins) [IX86_BUILTIN_CVTUDQ2PS]: Remove.
121 (ix86_vectorize_builtin_conversion): Remove.
122 (TARGET_VECTORIZE_BUILTIN_CONVERSION): Remove.
124 2011-10-31 Martin Jambor <mjambor@suse.cz>
126 * ipa-prop.c (mark_modified): Moved up in the file.
127 (is_parm_modified_before_call): Renamed to
128 is_parm_modified_before_stmt, moved up in the file.
129 (load_from_unmodified_param): New function.
130 (compute_complex_assign_jump_func): Also attempt to create pass
131 through jump functions for values loaded from (addressable)
134 2011-10-31 Jakub Jelinek <jakub@redhat.com>
136 * tree-vect-stmts.c (vectorizable_shift): If op1 is vect_external_def
137 in a loop and has different type from op0, cast it to op0's type
138 before the loop first. For slp give up. Don't crash if op1_vectype
141 2011-10-31 Paul Brook <paul@codesourcery.com>
143 * cgraphunit.c: Don't mark clones as static constructors.
145 2011-10-31 David Edelsohn <dje.gcc@gmail.com>
147 * gcc-ar: Do not include stdio.h.
149 2011-10-31 Diego Novillo <dnovillo@google.com>
151 * tree-streamer-out.c (pack_ts_base_value_fields): Emit
153 * tree-streamer-in.c (unpack_ts_base_value_fields): Read
156 2011-10-30 David S. Miller <davem@davemloft.net>
158 * config/sparc/sparc.c (vector_init_bshuffle): New function.
159 (vector_init_fpmerge): New function.
160 (sparc_expand_vector_init): Use them to improve non-const cases.
162 * dwarf2out.c (dwarf2out_var_location): When processing several
163 consecutive location notes, cache the result of next_real_insn().
165 2011-10-30 Uros Bizjak <ubizjak@gmail.com>
167 * config/i386/i386.md (avx2_vec_dup<mode>): Macroize insn from
168 avx2_vec_dup{v8sf,v4sf} using VF1 mode iterator.
169 (vec_dupv4sf): Remove expander.
170 (vec_dupv4sf): Merge from *vec_dupv4sf and *vec_dupv4sf_avx.
171 (vec_dupv2df): Remove expander.
172 (vec_dupv2df): Merge from *vec_dupv2df and *vec_dupv2df_sse3.
173 (*vec_concatv2df): Merge *vec_concatv2df_sse3.
174 (*vec_dupv4si): Merge *vec_dupv4si_avx.
175 (*vec_dupv2di): Merge *vec_dupv2di_sse3.
177 2011-10-30 Dmitry Plotnikov <dplotnikov@ispras.ru>
179 * tree-cfg.c (verify_gimple_assign_unary): Allow vector conversions.
180 * optabs.c (supportable_convert_operation): New function.
181 * optabs.h (supportable_convert_operation): New prototype.
182 * tree-vect-stmts.c (vectorizable_conversion): Change condition and
183 behavior for NONE modifier case.
184 * tree.h (VECTOR_INTEGER_TYPE_P): New macro.
186 2011-10-30 Jakub Jelinek <jakub@redhat.com>
188 * tree-vectorizer.h (NUM_PATTERNS): Bump to 9.
189 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern): New
191 (vect_vect_recog_func_ptrs): Add it.
193 2011-10-30 David S. Miller <davem@davemloft.net>
195 * reorg.c (label_before_next_insn): New function.
196 (relax_delay_slots): Use it instead of prev_label.
197 * rtl.h (prev_label): Delete declaration.
198 * emit-rtl.c (prev_label): Remove.
200 2011-10-30 Revital Eres <revital.eres@linaro.org>
202 * modulo-sched.c (generate_prolog_epilog): Mark prolog and epilog
203 as BB_DISABLE_SCHEDULE.
204 (mark_loop_unsched): New function.
205 (sms_schedule): Call it.
207 2011-10-29 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
210 * config/pa/protos.h (attr_length_save_restore_dltp): Delete.
211 (cmpib_comparison_operator): Likewise.
212 (following_cal, output_and, output_ior, output_move_double,
213 output_fp_move_double, output_block_move, output_block_clear,
214 output_cbranch, output_lbranch, output_bb, output_bvb, output_dbra,
215 output_movb, output_parallel_movb, output_parallel_addb, output_call,
216 output_indirect_call, output_millicode_call, output_mul_insn,
217 output_div_insn, output_mod_insn, singlemove_string,
218 output_arg_descriptor, output_global_address, print_operand,
219 legitimize_pic_address, hppa_encode_label, symbolic_expression_p,
220 fmpyaddoperands, fmpysuboperands, emit_bcond_fp, emit_move_sequence,
221 emit_hpdiv_const, is_function_label_plus_const, jump_in_call_delay,
222 hppa_fpstore_bypass_p, attr_length_millicode_call, attr_length_call,
223 attr_length_indirect_call, return_addr_rtx, function_arg_padding,
224 insn_refs_are_delayed, get_deferred_plabel, ldil_cint_p, zdepi_cint_p,
225 output_ascii, compute_frame_size, and_mask_p, cint_ok_for_move,
226 hppa_expand_prologue, hppa_expand_epilogue, ior_mask_p,
227 compute_zdepdi_operands, output_64bit_and, output_64bit_ior,
228 reloc_needed, magic_milli, shadd_constant_p): Consistently prefix
229 exported functions and variables with "pa_".
230 * config/pa/predicates.md: Likewise.
231 * config/pa/pa64-hpux.h: likewise.
232 * config/pa/som.h: Likewise.
233 * config/pa/elf.h: Likewise.
234 * config/pa/pa64-linux.h: Likewise.
235 * config/pa/pa.md: Likewise.
236 * config/pa/pa.c: Likewise.
237 * config/pa/pa-linux.h: Likewise.
238 * config/pa/pa.h: Likewise.
239 * config/pa/constraints.md: Likewise.
241 2011-10-29 Uros Bizjak <ubizjak@gmail.com>
243 * config/i386/i386.md (xop_sha<mode>3): Rename from xop_ashl<mode>3.
245 (xop_shl<mode>3): Rename from xop_lshl<mode>3. Update all uses.
246 * config/i386/i386.c: Update all uses.
248 2011-10-29 Uros Bizjak <ubizjak@gmail.com>
250 * config/i386/i386.md (lshlv16qi3): Remove expander.
251 (lshrv16qi3): New expander.
252 (<shift_insn>v16qi3): Macroize expander from ashrv16qi3 and lshrv16qi3
253 using any_shiftrt code iterator. Cleanup.
254 (ashlv16qi3): Cleanup.
257 2011-10-29 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
260 * config/pa/pa.c (emit_move_sequence): Legitimize TLS symbol
262 (pa_legitimate_constant_p): Return false for TLS_MODEL_GLOBAL_DYNAMIC
263 and TLS_MODEL_LOCAL_DYNAMIC symbol references.
265 2011-10-29 Georg-Johann Lay <avr@gjlay.de>
268 * config/avr/avr.opt (-maccumulate-args): New option.
269 * config/avr/avr.h (STARTING_FRAME_OFFSET): Redefine to
270 avr_starting_frame_offset.
271 (ACCUMULATE_OUTGOING_ARGS): Define to avr_accumulate_outgoing_args.
272 * config/avr/avr.md (UNSPECV_WRITE_SP_IRQ_ON): Remove.
273 (UNSPECV_WRITE_SP_IRQ_OFF): Remove.
274 (UNSPECV_WRITE_SP): New constant.
275 (*addhi3_sp_R): Rewrite to...
276 (*addhi3_sp): ...this new insn.
277 (movhi_sp_r_irq_off, movhi_sp_r_irq_on): Combine to...
278 (movhi_sp_r): ...this new insn.
279 * config/avr/avr-protos.h (avr_accumulate_outgoing_args): New.
280 (avr_starting_frame_offset): New.
281 * config/avr/avr.c (avr_accumulate_outgoing_args): New function.
282 (avr_starting_frame_offset): New function.
283 (avr_outgoing_args_size): New static function.
284 (avr_initial_elimination_offset): Use it.
285 (avr_simple_epilogue): Use it.
286 (avr_asm_function_end_prologue): Use it.
287 (expand_epilogue): Use it.
288 (expand_prologue): Use it. Break out code to...
289 (avr_prologue_setup_frame): ...this new static function.
290 (avr_can_eliminate): Allow eliminating to frame pointer if there
292 (avr_frame_pointer_required_p): Use frame pointer if target has a
294 * config/avr/constraints.md (R): Remove.
295 (Csp): New constraint.
296 * config/avr/predicates.md (avr_sp_immediate_operand): Use it.
298 2011-10-29 Andi Kleen <ak@linux.intel.com>
300 * gcc-ar.c (target_machine): Add.
302 2011-10-29 Anatoly Sokolov <aesok@post.ru>
304 * config/cris/cris.c (reg_ok_for_base_p, reg_ok_for_index_p,
305 cris_constant_index_p, cris_base_p, cris_index_p,
306 cris_base_or_autoincr_p, cris_bdap_index_p, cris_biap_index_p,
307 cris_legitimate_address_p): New functions.
308 (TARGET_LEGITIMATE_ADDRESS_P): Define.
309 (cris_pic_symbol_type, cris_valid_pic_const): Change arguments type
310 from rtx to const_rtx.
311 (cris_print_operand_address, cris_address_cost,
312 cris_side_effect_mode_ok): Use
313 cris_constant_index_p, cris_base_p, cris_base_or_autoincr_p,
314 cris_biap_index_p and cris_bdap_index_p.
315 * config/cris/cris.h (CONSTANT_INDEX_P, BASE_P, BASE_OR_AUTOINCR_P,
316 BDAP_INDEX_P, BIAP_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
317 REG_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P): Remove.
318 (EXTRA_CONSTRAINT_Q, EXTRA_CONSTRAINT_R, EXTRA_CONSTRAINT_T): Use
319 cris_constant_index_p, cris_base_p, cris_base_or_autoincr_p,
320 cris_biap_index_p and cris_bdap_index_p.
321 * config/cris/cris.md (moversideqi movemsideqi peephole2): Use
323 * config/cris/cris-protos.h (cris_constant_index_p, cris_base_p,
324 cris_base_or_autoincr_p, cris_bdap_index_p, cris_biap_index_p): New
326 (cris_pic_symbol_type, cris_valid_pic_const): Update prototype.
328 2011-10-21 Andi Kleen <ak@linux.intel.com>
330 * ggc-page (PAGE_ALIGN): Add.
331 (alloc_page, ggc_pch_total_size, ggc_pch_this_base, ggc_pch_read):
332 Replace ROUND_UP with PAGE_ALIGN.
334 2011-10-20 Andi Kleen <ak@linux.intel.com>
336 * ggc-page (alloc_anon): Add check argument.
337 (alloc_page): Add fallback to 1 page allocation.
338 Adjust alloc_anon calls to new argument.
340 2011-10-18 Andi Kleen <ak@linux.intel.com>
342 * ggc-page (release_pages): First free large continuous
343 chunks in the madvise path.
345 2011-10-18 Andi Kleen <ak@linux.intel.com>
347 * ggc-page.c (alloc_pages): Always round up entry_size.
349 2011-10-19 Andi Kleen <ak@linux.intel.com>
351 * Makefile.in (MOSTLYCLEANFILES): Add gcc-ar/nm/ranlib.
352 (native): Add gcc-ar, gcc-nm, gcc-ranlib.
353 (AR_LIBS, gcc-ar, gcc-ar.o, gcc-ranlib, gcc-ranlib.o,
354 gcc-nm, gcc-nm.o, gcc-ranlib.c, gcc-nm.c): Add.
355 (install): Depend on install-gcc-ar.
356 (install-gcc-ar): Add.
357 (uninstall): Uninstall gcc-ar, gcc-nm, gcc-ranlib.
358 * gcc-ar.c: Add new file.
360 2011-10-28 Pat Haugen <pthaugen@us.ibm.com>
362 * config/rs6000/rs6000.md (define_attr "type"): Add vecdouble.
363 * config/rs6000/vsx.md (VStype_simple, VStype_mul): Use vecdouble
365 (VStype_div): Use vector types for V2DF/V4SF.
366 (VStype_sqrt): Use *sqrt types.
367 (VS_spdp_type): Change type to vecdouble.
368 (*vsx_fmav2df4, *vsx_nfmsv2df4, vsx_xvcvdpsxws, vsx_xvcvdpuxws,
369 vsx_xvcvuxdsp, vsx_xvcvsxwdp, vsx_xvcvuxwdp, vsx_xvcvspsxds,
370 vsx_xvcvspuxds): Likewise.
371 (*vsx_fms<mode>4): Set type via <VStype_mul>.
372 (*vsx_eq_<mode>_p, *vsx_gt_<mode>_p, *vsx_ge_<mode>_p): Set type via
374 * config/rs6000/power7.md (power7-vecstore): Correct VSU pipe.
375 (power7-fpcompare, power7-sdiv, power7-ddiv, power7-sqrt,
376 power7-dsqrt): Correct insn latency.
377 (power7-vecsimple): Add veccmp type and correct dispatch/VSU values.
378 (power7-veccmp): Delete.
379 (power7-vecfloat): Correct latency/dispatch/VSU values.
380 (define_bypass "power7-vecfloat"): Correct latency and types.
381 (power7-veccomplex, power7-vecperm): Correct dispatch/VSU values.
382 (power7-vecdouble, power7-vecfdiv, power7-vecdiv): New.
384 2011-10-28 Uros Bizjak <ubizjak@gmail.com>
386 * config/i386/i386.md (shift_insn): Rename code attribute from
387 shiftrt_insn. Also handle ashift RTX.
388 (shift): Rename code attribute from shiftrt. Also handle ashift RTX.
389 (vshift): New code attribute.
390 (<shift_insn>*): Rename from <shiftrt_insn>*. Update asm templates.
391 (any_lshift): Move and rename code iterator from ...
392 * config/i386/sse.md (lshift): ... here.
393 (lshift_insn): Remove code attribute.
394 (lshift): Remove code attribute.
395 (vlshr<mode>3): Use lshiftrt RTX.
396 (vashr<mode>3, ashrv16qi3, ashrv2di3): Use ashiftrt RTX.
397 (vashl<mode>3, ashlv16qi3): Use ashift RTX.
398 (avx2_<lshift>v<mode>): Rename from avx2_<shift_insn>v<mode>. Use
399 any_lshift code iterator. Update asm template.
400 (<shift_insn><mode>3): Macroize insn from lshr<mode>3 and ashl<mode>3
401 usign any_lshift code iterator.
402 * config/i386/mmx.md (mmx_<shift_insn><mode>3): Macroize insn from
403 mmx_lshr<mode>3 and mmx_ashl<mode>3 usign any_lshift code iterator.
404 * config/i386/i386.c (bdesc_args) <__builtin_ia32_psll>: Update.
406 2011-10-28 Georg-Johann Lay <avr@gjlay.de>
409 * config/avr/avr.md (parityhi2): Expand allowing pseudos.
410 (*parityhi2): New pre-reload insn-and-split to map 16-bit parity
412 (*parityqihi2): Same for 8-bit parity.
414 2011-10-28 Julian Brown <julian@codesourcery.com>
416 PR rtl-optimization/47918
417 * reload1.c (set_initial_label_offsets): Use initial offsets
418 for labels on the nonlocal_goto_handler_labels chain.
420 2011-10-28 Iain Sandoe <iains@gcc.gnu.org>
422 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA):
423 Move darwin-fpsave.asm from here to ... LIB2FUNCS_EXTRA.
424 (LIB2FUNCS_EXTRA): Add darwin-gpsave.asm.
425 (TARGET_LIBGCC2_CFLAGS): Ensure that fPIC and -pipe are inherited from
427 * config/rs6000/darwin.h (FP_SAVE_INLINE): Adjust to enable.
428 (GP_SAVE_INLINE): Likewise.
429 (SAVE_FP_PREFIX, SAVE_FP_SUFFIX, RESTORE_FP_PREFIX,
430 RESTORE_FP_SUFFIX): Set to empty strings.
431 * config/rs6000/rs6000.c (rs6000_savres_strategy): Implement for Darwin.
432 (debug_stack_info): Print savres_strategy.
433 (rs6000_savres_routine_name): Implement for Darwin.
434 (rs6000_make_savres_rtx): Adjust used register for Darwin.
435 (rs6000_emit_prologue): Implement out-of-line saves for Darwin.
436 (rs6000_output_function_prologue): Don't emit .extern for Mach-O.
437 (rs6000_emit_epilogue): Implement out-of-line saves for Darwin.
438 * config/rs6000/darwin-gpsave.asm: New file.
440 2011-10-28 Jakub Jelinek <jakub@redhat.com>
442 * config/i386/sse.md (VI4SD_AVX2): Removed.
443 (VI48_AVX2, VI128_128, VI48_128, VI48_256): New mode iterators.
444 (vashl<mode>3): Use VI12_128 iterator instead of VI124_128.
445 Add another expander using VI48_128 iterator for
446 TARGET_AVX2 || TARGET_XOP and another using VI48_256 iterator
448 (vlshr<mode>3): Likewise. Change register_operand predicate to
449 nonimmediate_operand on last operand in the VI12_128 expander.
450 (vashr<mode>3): Use VI128_128 iterator instead of VI124_128.
451 (vashrv4si3, vashrv8si3): New expanders.
452 (avx2_ashrvv8si, avx2_ashrvv4si, avx2_<lshift>vv8si,
453 avx2_<lshift>vv2di): Removed.
454 (avx2_ashrv<mode>): New insn with VI4_AVX2 iterator.
455 (avx2_<lshift>v<mode>): Macroize using VI48_AVX2
456 iterator. Simplify pattern.
458 2010-10-28 Richard Guenther <rguenther@suse.de>
461 * lto-wrapper.c (get_options_from_collect_gcc_options):
462 Properly count arguments.
463 (run_gcc): Use an obstack to collect argv, properly separate
464 switches and their arguments.
466 2011-10-28 Jakub Jelinek <jakub@redhat.com>
468 * tree-vect-stmts.c (vectorizable_shift): Give up if op1 has different
469 vector mode from vectype's mode.
471 2011-10-28 Chung-Lin Tang <cltang@codesourcery.com>
473 PR rtl-optimization/49720
474 * simplify-rtx.c (simplify_relational_operation_1): Detect
475 infinite recursion condition in "(eq/ne (plus x cst1) cst2)
476 simplifies to (eq/ne x (cst2 - cst1))" case.
478 2011-10-27 David S. Miller <davem@davemloft.net>
480 * config/sparc/sparc.md (snedi_special): Only match when not VIS3.
481 (*snedi_zero): Likewise.
482 (*snedi_zero_trunc): Likewise.
483 (snedi_special_vis3): New expander.
484 (*snedi_zero_vis3): New insn.
485 (*snedi_zero_trunc_vis3): Likewise.
486 (*sltu_insn_vis3): Likewise.
487 (*sltu_insn_vis3_trunc): Likewise.
489 (*addxc_trunc_sp64_vis3): Likewise.
490 * config/sparc/sparc.c (emit_scc_insn): When VIS3 use the
491 gen_snedi_special_vis3 expander, and try GTU/LTU addx based
492 sequences on DImode values.
494 * config/sparc/sparc.md (64-bit vector moves): Use 'e' not 'f'
497 * regcprop.c (copyprop_hardreg_forward_1): Reject the
498 transformation when we narrow the mode on big endian.
500 2011-10-27 Jakub Jelinek <jakub@redhat.com>
502 * config/i386/sse.md (avx_cvtpd2dq256_2, avx_cvttpd2dq256_2,
503 vec_pack_sfix_trunc_v4df, vec_pack_sfix_v4df): New expanders.
504 (*avx_cvtpd2dq256_2, *avx_cvttpd2dq256_2): New insns.
506 * config/i386/i386.c (ix86_print_operand): Handle 'q' and 'x'
507 overrides for -masm=intel memory.
508 * config/i386/sse.md (sse2_cvtdq2pd, sse2_cvtps2pd,
509 sse4_1_<code>v8qiv8hi2, avx2_<code>v8qiv8si2,
510 sse4_1_<code>v4hiv4si2, avx2_<code>v4hiv4di2,
511 sse4_1_<code>v2siv2di2): Use %q1 instead of %1 for -masm=intel.
512 (sse4_1_<code>v4qiv4si2, avx2_<code>v4qiv4di2,
513 sse4_1_<code>v2hiv2di2): Use %k1 instead of %1 for -masm=intel.
514 (sse4_1_<code>v2qiv2di2): Use %w1 instead of %1 for -masm=intel.
516 2011-10-27 Martin Jambor <mjambor@suse.cz>
518 * ipa-prop.c (compute_pass_through_member_ptrs): Rename parm_info
520 (ipa_compute_jump_functions_for_edge): Likewise.
521 (ipa_compute_jump_functions): Likewise.
522 (ipa_analyze_indirect_call_uses): Likewise.
523 (ipa_analyze_call_uses): Likewise.
524 (ipa_analyze_params_uses): Likewise.
525 (ipa_analyze_node): Likewise.
527 2011-10-27 Uros Bizjak <ubizjak@gmail.com>
530 * config/i386/sse.md (*avx_unpcklpd256): Remove extra insn
531 constraints. Change alternative 1 to "x,m,1".
533 2011-10-27 Jakub Jelinek <jakub@redhat.com>
535 * Makefile.in (build/gencheck.o): Depend on tree.def and
536 c-family/c-common.def.
538 * tree-ssa-strlen.c: Include expr.h.
539 (get_stridx): Don't use c_strlen, instead use string_constant
540 and compute string length from it.
541 * Makefile.in (tree-ssa-strlen.o): Depend on $(EXPR_H).
543 2011-10-27 Eric Botcazou <ebotcazou@adacore.com>
545 PR rtl-optimization/46603
547 * reload.c (push_reload): In the out case, restore previous behavior
548 for subregs that don't have word mode.
550 2011-10-27 Ian Lance Taylor <iant@google.com>
552 * cppdefault.c: Undef NATIVE_SYSTEM_HEADER_DIR if
553 CROSS_DIRECTORY_STRUCTURE is defined and TARGET_SYSTEM_ROOT is not.
554 (cpp_include_defaults): Only use NATIVE_SYSTEM_HEADER_DIR if it is
557 2011-10-27 Richard Henderson <rth@redhat.com>
559 * optabs.c (expand_vec_perm): Use the correct mode for scaling the
560 selector. Save the qimode constant selector for later use by the
561 qimode vec_perm pattern.
563 2011-10-27 Bernd Schmidt <bernds@codesourcery.com>
565 * config/c6x/c6x.c (unit_req_imbalance, res_mii): Cast the first arg
566 to unit_req_factor to the right enum type.
567 (get_unit_operand_masks, reshuffle_units, try_rename_operands,
568 hwloop_optimize): Remove unused variables.
570 2010-10-27 Richard Guenther <rguenther@suse.de>
573 * tree-vect-generic.c (do_binop): Handle scalar operands.
575 2011-08-27 Uros Bizjak <ubizjak@gmail.com>
578 * config/i386/sse.md (*vec_extract_v4sf_mem): Avoid combining registers
579 from different units in a single alternative.
581 2011-10-26 David S. Miller <davem@davemloft.net>
583 * config/sparc/sparc.c (emit_scc_insn): Force attempt of v9 sequences
584 if we're comparing DImode and comparison is other than EQ or NE.
586 * config/sparc/sparc.c (emit_scc_insn): Do not try v9 sequences until
587 LEU/LTU/GEU/GTU is attempted.
588 * config/sparc/sparc.md (*neg_snesi_sign_extend): New 64-bit insn
590 (*neg_seqsi_sign_extend): Likewise.
591 (*sltu_extend_sp64, *neg_sltu_extend_sp64, *sgeu_extend_sp64,
592 *neg_sgeu_extend_sp64): New insns.
594 * config/sparc/sparc-protos.h (sparc_expand_conditional_move): Declare.
595 * config/sparc/sparc.md (mov<I:mode>cc, mov<F:mode>cc): Call it.
596 (*mov<I:mode>_cc_v9): Normalize to expect operand 0 always in operand 4.
597 (*mov<I:mode>_cc_reg_sp64): Likewise.
598 (*movsf_cc_v9): Likewise.
599 (*movsf_cc_reg_sp64): Likewise.
600 (*movdf_cc_v9): Likewise.
601 (*movdf_cc_reg_sp64): Likewise.
602 (*movtf_cc_hq_v9): Likewise.
603 (*movtf_cc_reg_hq_sp64): Likewise.
604 (*movtf_cc_v9): Likewise.
605 (*movtf_cc_reg_sp64): Likewise.
606 * config/sparc/sparc.c (sparc_expand_conditional_move): New function.
607 (sparc_print_operand): Delete 'c' and 'd' handling, no longer used.
609 2011-10-26 Eric Botcazou <ebotcazou@adacore.com>
611 * reload.c (reload_inner_reg_of_subreg): Change type of return value
612 and type of OUTPUT parameter to bool and adjust. Document MODE and
613 OUTPUT parameters. Use HARD_REGISTER_P. Reorder final condition
614 and improve associated comment.
615 (push_reload): Clarify and update comments about reloading of subregs.
616 Adjust calls to reload_inner_reg_of_subreg. Compute the class upfront
617 for the reloading of subregs in the out case as well.
619 2011-10-26 Alexandre Oliva <aoliva@redhat.com>
622 * var-tracking.c (rtx_debug_expr_p): New.
623 (use_type): Don't use debug exprs to track non-VTA variables.
625 2011-10-26 Jeff Law <law@redhat.com>
627 * doc/invoke.texi (sink-frequency-threshold): Document.
628 * tree-ssa-sink.c: Include params.h.
629 (select_best_block): New function.
630 (statement_sink_location): Use it.
631 * params.def (SINK_FREQUENCY_THRESHOLD): New PARAM.
633 2011-10-26 Iain Sandoe <iains@gcc.gnu.org>
636 * config/darwin.c (top level): Amend comments concerning LTO output.
637 (lto_section_num): New variable. (darwin_lto_section_e): New GTY.
638 (LTO_SECTS_SECTION, LTO_INDEX_SECTION): New.
639 (LTO_NAMES_SECTION): Rename.
640 (darwin_asm_named_section): Record LTO section counts and switches
641 in a vec of darwin_lto_section_e.
642 (darwin_file_start): Remove unused code.
643 (darwin_file_end): Put an LTO section termination label. Handle
644 output of the wrapped LTO sections, index and names table.
646 2011-10-26 Alan Modra <amodra@gmail.com>
648 * config/rs6000/rs6000.c (rs6000_make_savres_rtx): Delete unneeded
650 (rs6000_emit_stack_reset): Only return insn emitted when it adjusts sp.
651 (rs6000_make_savres_rtx): Rename to rs6000_emit_savres_rtx. Use
652 simple_return in pattern, emit instruction, and set jump_label.
653 (rs6000_emit_prologue): Update for rs6000_emit_savres_rtx. Use
654 simple_return rather than return.
655 (emit_cfa_restores): New function.
656 (rs6000_emit_epilogue): Emit cfa_restores when flag_shrink_wrap.
657 Add missing cfa_restores for SAVE_WORLD. Add missing LR cfa_restore
658 when using out-of-line gpr restore. Add missing LR and FP regs
659 cfa_restores for out-of-line fpr restore. Consolidate code setting
660 up cfa_restores. Formatting. Use LR_REGNO define.
661 (rs6000_output_mi_thunk): Use simple_return rather than return.
662 * config/rs6000/rs6000.md (sibcall*, sibcall_value*): Likewise.
663 (return_internal*): Likewise.
664 (any_return, return_pred, return_str): New iterators.
665 (return, conditional return insns): Provide both return and
666 simple_return variants.
667 * config/rs6000/rs6000.h (EARLY_R12, LATE_R12): Define.
668 (REG_ALLOC_ORDER): Move r12 before call-saved regs when FIXED_R13.
669 Move r11 and r0 later to suit shrink-wrapping.
671 2011-10-26 Richard Guenther <rguenther@suse.de>
673 * lto-wrapper.c (run_gcc): Properly init/free obstack.
675 2011-10-26 Jakub Jelinek <jakub@redhat.com>
677 * config/i386/i386.md (UNSPEC_VSIBADDR): New.
678 * config/i386/predicates.md (vsib_address_operand,
679 vsib_mem_operator): New predicates.
680 * config/i386/i386.c (ix86_print_operand_address): Handle
681 UNSPEC_VSIBADDR addresses.
682 * config/i386/sse.md (avx2_gathersi<mode>, avx2_gatherdi<mode>,
683 avx2_gatherdi<mode>256): Adjust expanders to use MEM with
684 UNSPEC_VSIBADDR address.
685 (*avx2_gathersi<mode>, *avx2_gatherdi<mode>, *avx2_gatherdi<mode>256):
686 Adjust insns to use MEM with UNSPEC_VSIBADDR address.
688 2011-10-26 Tom de Vries <tom@codesourcery.com>
690 PR tree-optimization/50763
691 * tree-ssa-tail-merge.c (replace_block_by): Update vops if phi_vuse1 or
692 phi_vuse2 is NULL_TREE only if bb1 dominates or is dominated by bb2.
694 2011-10-26 Richard Guenther <rguenther@suse.de>
697 * Makefile.in (lto-wrapper): Depend on and link against opts-common.o.
698 (lto-wrapper.o): Depend on $(OPTS_H) and $(OPTIONS_H).
699 * lto-wrapper.c (get_options_from_collect_gcc_options): New function.
700 (run_gcc): Use it. Filter out language specific options.
702 2011-10-26 Andreas Tobler <andreast@fgznet.ch>
704 * config/i386/freebsd64.h (LINK_SPEC): Emit the same warning as the
707 2011-10-25 Paolo Carlini <paolo.carlini@oracle.com>
710 * gcc.c (main): Fix fatal_error string for translation.
712 2011-10-25 Ian Lance Taylor <iant@google.com>
714 * tree-eh.c (do_return_redirection): Remove return_value_p
715 parameter. Change all callers.
716 (lower_try_finally_nofallthru): Remove local return_val.
717 (lower_try_finally_onedest): Likewise.
718 (lower_try_finally_copy): Likewise.
719 (lower_try_finally_switch): Likewise.
721 2011-10-25 H.J. Lu <hongjiu.lu@intel.com>
723 * config/i386/mmx.md (*mmx_maskmovq): Replace :SI with :P and
724 remove "&& !TARGET_64BIT"
725 (*mmx_maskmovq_rex): Removed.
727 2011-10-25 Eric Botcazou <ebotcazou@adacore.com>
729 PR rtl-optimization/46603
730 * reload.c (push_reload): In the out case, reload the subreg as well
731 as the reg if it has word mode.
733 2011-10-25 Eric Botcazou <ebotcazou@adacore.com>
735 * dwarf2out.c (add_gnat_descriptive_type_attribute): Temporarily
736 suppress debug info for the parent type.
738 2011-10-25 Eric Botcazou <ebotcazou@adacore.com>
740 * config/ia64/ia64.c (ia64_profile_hook): Fix thinko.
742 2011-10-25 Richard Henderson <rth@redhat.com>
744 * config/i386/sse.md (VEC_EXTRACT_EVENODD_MODE): Remove.
745 (vec_extract_even<mode>, vec_extract_odd<mode>): Remove.
747 * config/rs6000/altivec.md (vec_extract_evenv8hi,
748 vec_extract_evenv16qi, vec_extract_oddv4si,
749 vec_extract_oddv4sf): Remove.
751 * config/spu/spu.md (vec_extract_evenv4si, vec_extract_evenv4sf,
752 vec_extract_evenv8hi, vec_extract_evenv16qi, vec_extract_oddv4si,
753 vec_extract_oddv4sf, vec_extract_oddv8hi, vec_extract_oddv16qi,
754 vec_interleave_highv4sf, vec_interleave_lowv4sf,
755 vec_interleave_highv4si, vec_interleave_lowv4si,
756 vec_interleave_highv8hi, vec_interleave_lowv8hi,
757 vec_interleave_highv16qi, vec_interleave_lowv16qi): Remove.
759 * expr.c (expand_expr_real_2) [VEC_EXTRACT_EVEN_EXPR]: Use binop.
760 [VEC_EXTRACT_ODD_EXPR, VEC_INTERLEAVE_HIGH_EXPR]: Likewise.
761 [VEC_INTERLEAVE_LOW_EXPR]: Likewise.
762 * optabs.c (expand_binop): Implement vec_interleave_high_optab,
763 vec_interleave_low_optab, vec_extract_even_optab,
764 vec_extract_odd_optab with expand_vec_perm.
765 (can_vec_perm_for_code_p): New.
767 * tree-vect-data-refs.c (vect_strided_store_supported): Allow for
768 fallback via can_vec_perm_for_code_p.
769 (vect_strided_load_supported): Likewise.
770 * tree-vect-generic.c (expand_vector_operations_1): Never lower
771 VEC_INTERLEAVE_HIGH_EXPR, VEC_INTERLEAVE_LOW_EXPR,
772 VEC_EXTRACT_EVEN_EXPR, VEC_EXTRACT_ODD_EXPR.
774 * target.def (vec_perm_const_ok): Change parameters to mode and
776 * doc/tm.texi: Rebuild.
777 * config/i386/i386.c (ix86_vectorize_vec_perm_const_ok): Change
778 parameters to mode and array of indicies.
779 * expr.c (expand_expr_real_2) [VEC_PERM_EXPR]: Expand operands here.
780 * optabs.c (can_vec_perm_p): Rename from can_vec_perm_expr_p.
781 Change parameters to mode and array of indicies.
782 (expand_vec_perm_1): Rename from expand_vec_perm_expr_1.
783 (expand_vec_perm): Rename from expand_vec_perm_expr. Change
784 parameters to mode and rtx inputs. Try lowering to QImode
785 vec_perm_const before trying fully variable permutation.
786 * optabs.h: Update decls.
787 * tree-vect-generic.c (lower_vec_perm): Extract array of indices from
788 VECTOR_CST to pass to can_vec_perm_p.
789 * tree-vect-slp.c (vect_get_mask_element): Change mask parameter type
790 from int pointer to unsigned char pointer.
791 (vect_transform_slp_perm_load): Update for change to can_vec_perm_p.
792 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
794 * tree.def (VEC_EXTRACT_EVEN_EXPR): Fix typo in text name.
795 (VEC_EXTRACT_ODD_EXPR, VEC_INTERLEAVE_HIGH_EXPR,
796 VEC_INTERLEAVE_LOW_EXPR): Likewise.
798 2011-10-25 Mike Stump <mikestump@comcast.net>
800 * reload.c (regno_clobbered_p): Fix typo.
802 2011-10-25 Dodji Seketeli <dodji@redhat.com>
804 * input.c (expand_location): Rewrite using linemap_resolve_location
805 and linemap_expand_location. Add a comment.
807 2011-10-25 Jakub Jelinek <jakub@redhat.com>
809 PR tree-optimization/50596
810 * tree-vect-stmts.c (vect_mark_relevant): Only use
811 FOR_EACH_IMM_USE_FAST if lhs is SSA_NAME.
812 (vectorizable_store): If is_pattern_stmt_p look through
813 VIEW_CONVERT_EXPR on lhs.
814 * tree-vect-patterns.c (check_bool_pattern, adjust_bool_pattern):
815 Use unsigned type instead of signed.
816 (vect_recog_bool_pattern): Optimize also stores into bool memory in
817 addition to casts from bool to integral types.
818 (vect_mark_pattern_stmts): If pattern_stmt already has vinfo
819 created, don't create it again.
821 2011-10-25 Kai Tietz <ktietz@redhat.com>
823 * config/i386/i386.c (ix86_frame_pointer_required): Require
824 frame-pointer, if setjmp is used for 32-bit ms-abi.
826 2011-10-24 Michael Meissner <meissner@linux.vnet.ibm.com>
828 * builtins.c (set_builtin_user_assembler_name): Remove extra
829 newline added in October 11th, 2011 change.
831 2011-10-24 David S. Miller <davem@davemloft.net>
833 * config/sparc/little-endian.opt: Delete.
834 * config.gcc: Remove references to config/sparc/little-endian.opt
835 * doc/invoke.texi: Remove documentation of -mlittl-endian on sparc.
836 * config/sparc/linux64.h: Delete references to -mlittle-endian.
837 * config/sparc/netbsd-elf.h: Likewise.
838 * config/sparc/openbsd64.h: Likewise.
839 * config/sparc/sparc.h: Likewise.
840 * config/sparc/sp64-elf.h: Likewise and delete overrides for
841 BYTES_BIG_ENDIAN and WORDS_BIG_ENDIAN.
842 * config/sparc/sparc.c (dump_target_flag_bits): Remove reference
843 to MASK_LITTLE_ENDIAN.
844 * config/sparc/sparc.opt (Mask(LITTLE_ENDIAN)): Delete.
846 * config/sparc/sparc.md: Only use F, G, and C constraints in FP insns.
847 Only use D, Y, and Z constraints in vector insns.
849 * config/sparc/sparc.md (cpu_feature, enabled): New attributes.
850 (*movsi_insn_novis3, *movsi_insn_vis3): Consolidate into one pattern
852 (*movdi_insn_sp32_v9_novis3, *movdi_insn_sp32_v9_vis3): Consolidate
853 into *movdi_insn_sp32.
854 (*movdi_insn_sp64_novis3, *movdi_insn_sp64_vis3): Consolidate into
855 one pattern called *movdi_insn_sp64.
856 (*movsf_insn_novis3, *movsf_insn_vis3, *movsf_insn_no_fpu):
857 Consolidate into one pattern called *movsf_insn.
858 (*movdf_insn_sp32_no_fpu, *movdf_insn_sp32_v9_novis3,
859 *movdf_insn_sp32_v9_vis3, *movdf_insn_sp32_v9_no_fpu): Consolidate
860 into *movdf_insn_sp32.
861 (*movdf_insn_sp64_novis3, *movdf_insn_sp64_vis3,
862 *movdf_insn_sp64_no_fpu): Consolidate into one pattern called
864 (*zero_extendsidi2_insn_sp64_novis3,
865 *zero_extendsidi2_insn_sp64_vis3): Consolidate into one pattern
866 called *zero_extendsidi2_insn_sp64.
867 (*sign_extendsidi2_insn_novis3, *sign_extendsidi2_insn_vis3):
868 Consolidate into one pattern named *sign_extendsidi2_insn.
869 (*mov<VM32:mode>_insn_novis3, *mov<VM32:mode>_insn_vis3):
870 Consolidate into one pattern named *mov<VM32:mode>_insn.
871 (*mov<VM64:mode>_insn_sp64_novis3,
872 *mov<VM64:mode>_insn_sp64_novis3): Consolidate into one pattern
873 named *mov<VM64:mode>_insn_sp64.
874 (*mov<VM64:mode>_insn_sp32_novis3,
875 *mov<VM64:mode>_insn_sp32_vis3): Consolidate into one pattern
876 named *mov<VM64:mode>_insn_sp32.
878 2011-10-24 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
880 * tree-ssa-strlen.c (get_string_length): Change assertion to STPCPY.
881 (zero_length_string): Change assertion to accept strinfo without
882 length but with stmt instead.
883 Set the endptr pointer also if starting a new chain.
884 (adjust_related_strinfos): Ignore strinfos marked for delayed
886 (handle_builtin_strcpy): Mark earlier strinfo elements also for
887 delayed length computation.
889 2011-10-24 Georg-Johann Lay <avr@gjlay.de>
892 Port from 4.6 branch r180379
893 * doc/invoke.texi (AVR Options): New subsubsection to explain EIND
894 handling and indirect jump/calls on devices > 128k.
896 2011-10-24 Anatoly Sokolov <aesok@post.ru>
897 Georg-Johann Lay <avr@gjlay.de>
900 * doc/extend.texi (Declaring Attributes of Functions):
901 Document OS_main and OS_task attributes.
902 (Specifying Attributes of Variables): Move up
903 subsection "AVR Variable Attributes" as of alphabetical order.
905 2011-10-24 Richard Guenther <rguenther@suse.de>
907 * tree-vect-stmts.c (vect_get_vec_def_for_operand): Convert constants
908 to vector element type.
909 (vectorizable_assignment): Bail out for non-mode-precision operations.
910 (vectorizable_shift): Likewise.
911 (vectorizable_operation): Likewise.
912 (vectorizable_type_demotion): Likewise.
913 (vectorizable_type_promotion): Likewise.
914 (vectorizable_store): Handle non-mode-precision stores.
915 (vectorizable_load): Handle non-mode-precision loads.
916 (get_vectype_for_scalar_type_and_size): Return a vector type
917 for non-mode-precision integers.
918 * tree-vect-loop.c (vectorizable_reduction): Bail out for
919 non-mode-precision reductions.
921 2011-10-24 Julian Brown <julian@codesourcery.com>
923 * config/m68k/m68k.c (notice_update_cc): Tighten condition for
924 setting CC_REVERSED for FP comparisons.
926 2011-10-24 Richard Guenther <rguenther@suse.de>
928 PR tree-optimization/50838
929 * tree-data-ref.c (dr_analyze_indices): Properly canonicalize
930 a MEM_REF base if we change it.
932 2011-10-24 Bernd Schmidt <bernds@codesourcery.com>
935 * rtlanal.c: Swap includes of "hard-reg-set.h" and "rtl.h".
937 PR rtl-optimization/50833
938 * function.c (thread_prologue_and_epilogue_insns): Expect the
939 return insn optimization only if optimize.
941 2011-10-24 Georg-Johann Lay <avr@gjlay.de>
943 * config/avr/avr.c: Break long lines.
944 Define target hooks on the fly if applicable.
945 (TARGET_ASM_FUNCTION_RODATA_SECTION): Remove first definition
947 (targetm): Move definition to end of file.
948 (avr_can_eliminate): Make static on the fly.
949 (avr_frame_pointer_required_p): Ditto.
950 (avr_hard_regno_scratch_ok): Ditto.
951 (avr_builtin_setjmp_frame_value): Make static on the fly.
952 Indent according to coding rules.
953 (avr_case_values_threshold): Ditto.
954 (avr_attribute_table): Move down.
956 2011-10-24 Ira Rosen <ira.rosen@linaro.org>
958 PR tree-optimization/50730
959 * tree-vect-data-refs.c (vect_analyze_data_refs): Stop basic block
960 analysis if encountered unsupported data-ref.
962 2011-10-23 David S. Miller <davem@davemloft.net>
964 * config/sparc/sparc.c (sparc_option_override): Remove -mv8plus
966 * config/sparc/linux64.h (CC1_SPEC): When defaulting to 64-bit,
967 append -mcpu=v9 when -mv8plus is given.
969 * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED): We can move
970 between float and non-float regs when VIS3.
971 * config/sparc/sparc.c (eligible_for_restore_insn): We can't
972 use a restore when the source is a float register.
973 (sparc_split_regreg_legitimate): When VIS3 allow moves between
974 float and integer regs.
975 (sparc_register_move_cost): Adjust to account for VIS3 moves.
976 (sparc_preferred_reload_class): On 32-bit with VIS3 when moving an
977 integer reg to a class containing EXTRA_FP_REGS, constrain to FP_REGS.
978 (sparc_secondary_reload): On 32-bit with VIS3 when moving between
979 float and integer regs we sometimes need a FP_REGS class
980 intermediate move to satisfy the reload. When this happens
981 specify an extra cost of 2.
982 (*movsi_insn): Rename to have "_novis3" suffix and add !VIS3 guard.
983 (*movdi_insn_sp32_v9): Likewise.
984 (*movdi_insn_sp64): Likewise.
985 (*movsf_insn): Likewise.
986 (*movdf_insn_sp32_v9): Likewise.
987 (*movdf_insn_sp64): Likewise.
988 (*zero_extendsidi2_insn_sp64): Likewise.
989 (*sign_extendsidi2_insn): Likewise.
990 (*movsi_insn_vis3): New insn.
991 (*movdi_insn_sp32_v9_vis3): New insn.
992 (*movdi_insn_sp64_vis3): New insn.
993 (*movsf_insn_vis3): New insn.
994 (*movdf_insn_sp32_v9_vis3): New insn.
995 (*movdf_insn_sp64_vis3): New insn.
996 (*zero_extendsidi2_insn_sp64_vis3): New insn.
997 (*sign_extendsidi2_insn_vis3): New insn.
998 (TFmode reg/reg split): Make sure both REG operands are float.
999 (*mov<VM32:mode>_insn): Add "_novis3" suffix and !VIS3 guard. Remove
1000 easy constant to integer reg alternatives.
1001 (*mov<VM64:mode>_insn_sp64): Likewise.
1002 (*mov<VM64:mode>_insn_sp32_novis3): Likewise.
1003 (*mov<VM32:mode>_insn_vis3): New insn.
1004 (*mov<VM64:mode>_insn_sp64_vis3): New insn.
1005 (*mov<VM64:mode>_insn_sp32_vis3): New insn.
1006 (VM64 reg<-->reg split): New splitter for 32-bit.
1008 * config/sparc/sparc.c (sparc_split_regreg_legitimate): New function.
1009 * config/sparc/sparc-protos.h (sparc_split_regreg_legitimate):
1011 * config/sparc/sparc.md (DImode reg/reg split): Use it.
1012 (DFmode reg/reg split): Likewise.
1014 * config/sparc/sparc.md (*movdi_insn_sp32_v9): Add alternatives for
1015 generating fzero and fone instructions.
1016 (DImode const_int --> reg splitter): Only trigger for integer regs.
1018 * config/sparc/predicates.md (input_operand): Disallow vector
1019 constants other than 0 and -1.
1020 * config/sparc/sparc.c (sparc_preferred_reload_class): Return
1021 NO_REGS for vector constants other than 0 and -1.
1023 * config/sparc/sparc.h (SPARC_FIRST_INT_REG, SPARC_LAST_INT_REG,
1024 SPARC_INT_REG_P): Define.
1025 (HARD_REGNO_NREGS): Use SPARC_INT_REG_P.
1026 (REGNO_OK_FOR_INDEX_P): Likewise.
1027 * config/sparc/sparc.c (gen_df_reg): Likewise.
1028 (eligible_for_return_delay): Likewise.
1029 (eligible_for_sibcall_delay): Likewise.
1030 (sparc_legitimate_address_p): Likewise.
1031 (emit_save_or_restore_regs): Likewise.
1032 (registers_ok_for_ldd_peep): Likewise.
1033 * config/spac/sparc.md (DI mode splitters): Likewise.
1034 (SF mode const splitters): Likewise.
1035 (DF mode splitters): Likewise.
1036 (32-bit DI mode logical op splitters): Likewise.
1038 2011-10-23 Paolo Carlini <paolo.carlini@oracle.com>
1042 2011-10-23 Paolo Carlini <paolo.carlini@oracle.com>
1045 * doc/invoke.texi ([-Wnarrowing], [-Wc++0x-compat]): Update.
1047 2011-10-23 Uros Bizjak <ubizjak@gmail.com>
1049 * config/i386/sse.md (avx2_lshl<mode>3): Remove insn pattern.
1050 (VI248_256): Remove mode iterator.
1051 * config/i386/i386.c (ix86_expand_vec_perm): Use gen_ashlv4di3
1052 instead of gen_avx2_lshlv4di3.
1053 (bdesc_args): Use CODE_FOR_ashl{v16hi,v8si,v4di}3 instead of
1054 CODE_FOR_avx2_lshl{v16hi,v8si,v4di}3.
1056 2011-10-23 Uros Bizjak <ubizjak@gmail.com>
1058 * config/i386/sse.md (sseintprefix): Rename from gthrfirstp.
1059 (<avx_avx2>_maskload<ssemodesuffix><avxsizesuffix>): Delete expander.
1060 (<avx_avx2>_maskload<ssemodesuffix><avxsizesuffix>) Merge insn
1061 pattern from *avx2_maskload<ssemodesuffix><avxsizesuffix> and
1062 *avx_maskload<ssemodesuffix><avxsizesuffix> using V48_AVX mode
1063 iterator. Use sseintprefix mode attribute.
1064 (<avx_avx2>_maskstore<ssemodesuffix><avxsizesuffix>): Delete expander.
1065 (<avx_avx2>_maskstore<ssemodesuffix><avxsizesuffix>) Merge insn
1066 pattern from *avx2_maskstore<ssemodesuffix><avxsizesuffix> and
1067 *avx_maskstore<ssemodesuffix><avxsizesuffix> using V48_AVX mode
1068 iterator. Use sseintprefix mode attribute.
1069 (*avx2_gathersi<mode>) Use sseintprefix and ssemodesuffix mode
1071 (*avx2_gatherdi<mode>): Ditto.
1072 (*avx2_gatherdi<mode>256): Ditto.
1073 (VI48_AVX2): Remove mode iterator.
1074 (gthrlastfp): Remove mode attribute.
1076 2011-10-23 Paolo Carlini <paolo.carlini@oracle.com>
1079 * doc/invoke.texi ([-Wnarrowing], [-Wc++0x-compat]): Update.
1081 2011-10-23 Tom de Vries <tom@codesourcery.com>
1083 PR tree-optimization/50763
1084 * tree-ssa-tail-merge.c (same_succ_flush_bb): New function, factored
1086 (same_succ_flush_bbs): Use same_succ_flush_bb.
1087 (purge_bbs): Remove argument. Remove calls to same_succ_flush_bbs,
1088 release_last_vdef and delete_basic_block.
1089 (unlink_virtual_phi): New function.
1090 (update_vuses): Add and use vuse1_phi_args argument. Set var to
1091 SSA_NAME_VAR of vuse1 or vuse2, and use var. Handle case that
1092 def_stmt2 is NULL. Use phi result as phi arg in case vuse1 or vuse2
1093 is NULL_TREE. Replace uses of vuse1 if vuse2 is NULL_TREE. Fix code
1094 to limit replacement of uses. Propagate phi argument for phis with a
1096 (replace_block_by): Update vops if phi_vuse1 or phi_vuse2 is NULL_TREE.
1097 Set vuse1_phi_args if vuse1 is a phi defined in bb1. Add
1098 vuse1_phi_args as argument to call to update_vuses. Call
1099 release_last_vdef, same_succ_flush_bb, delete_basic_block. Update
1100 CDI_DOMINATORS info.
1101 (tail_merge_optimize): Remove argument in call to purge_bbs. Remove
1102 call to free_dominance_info. Only call calculate_dominance_info once.
1104 2011-10-23 Eric Botcazou <ebotcazou@adacore.com>
1106 * fold-const.c (invert_tree_comparison): Always invert EQ_EXPR/NE_EXPR.
1108 PR tree-optimization/44683
1109 * tree-ssa-dom.c (record_edge_info): Record simple equivalences only if
1110 we can be sure that there are no signed zeros involved.
1112 2011-10-23 Jan Hubicka <jh@suse.cz>
1114 * ipa-inline.c (estimate_badness): Scale up and handle overflows.
1116 2011-10-23 Uros Bizjak <ubizjak@gmail.com>
1119 * config/i386/sse.md (<avx_avx2>_maskload<ssemodesuffix><avxsizesuffix>):
1120 Remove (match_dup 0).
1121 (*avx2_maskload<ssemodesuffix><avxsizesuffix>): New insn pattern.
1122 (*avx_maskload<ssemodesuffix><avxsizesuffix>): Ditto.
1123 (*avx2_maskstore<ssemodesuffix><avxsizesuffix>): Ditto.
1124 (*avx_maskstore<ssemodesuffix><avxsizesuffix>): Ditto.
1125 (*avx2_maskmov<ssemodesuffix><avxsizesuffix>): Remove insn pattern.
1126 (*avx_maskmov<ssemodesuffix><avxsizesuffix>): Ditto.
1128 2011-10-23 Ira Rosen <ira.rosen@linaro.org>
1130 PR tree-optimization/50819
1131 * tree-vectorizer.h (vect_analyze_data_ref_dependences): Remove
1133 * tree-vect-loop.c (vect_analyze_loop_2): Update call to
1134 vect_analyze_data_ref_dependences.
1135 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Remove
1136 the last argument. Check load-after-store dependence for unknown
1137 dependencies in basic blocks.
1138 (vect_analyze_data_ref_dependences): Update call to
1139 vect_analyze_data_ref_dependences.
1140 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo.
1141 * tree-vect-slp.c (vect_bb_vectorizable_with_dependencies): Remove.
1142 (vect_slp_analyze_bb_1): Update call to
1143 vect_analyze_data_ref_dependences. Don't call
1144 vect_bb_vectorizable_with_dependencies.
1146 2011-10-22 David S. Miller <davem@davemloft.net>
1148 * config/sparc/sparc.h (SECONDARY_INPUT_RELOAD_CLASS,
1149 SECONDARY_OUTPUT_RELOAD_CLASS): Delete.
1150 * config/sparc/sparc.c (TARGET_SECONDARY_RELOAD): Redefine.
1151 (sparc_secondary_reload): New function.
1153 * config/sparc/sparc.h (sparc_costs): Remove extern decl.
1154 (struct processor_costs): Move from here..
1155 * config/sparc/sparc.c (struct processor_costs): To here.
1156 (sparc_costs): Mark static.
1158 * config/sparc/sparc.c (short_branch, reg_unused_after): Delete.
1159 * config/sparc/sparc-protos.h (short_branch, reg_unused_after):
1160 Get rid of declarations.
1162 2011-10-21 Paul Brook <paul@codesourcery.com>
1164 * config/c6x/c6x.c (c6x_asm_emit_except_personality,
1165 c6x_asm_init_sections): New functions.
1166 (TARGET_ASM_EMIT_EXCEPT_PERSONALITY, TARGET_ASM_INIT_SECTIONS): Define.
1168 2011-10-21 Jakub Jelinek <jakub@redhat.com>
1171 * config/i386/i386.c (expand_vec_perm_even_odd_1): Handle
1172 V4DImode and V8SImode for !TARGET_AVX2.
1174 2011-10-21 Bernd Schmidt <bernds@codesourcery.com>
1177 * sched-deps.c (add_dependence): If not doing predication, promote
1178 REG_DEP_CONTROL to REG_DEP_ANTI.
1180 2011-10-21 Georg-Johann Lay <avr@gjlay.de>
1182 * config/avr/avr.h (LEGITIMIZE_RELOAD_ADDRESS): Pass address of X
1183 instead of X to avr_legitimize_reload_address.
1184 * config/avr/avr-protos.h (avr_legitimize_reload_address): Change
1185 first argument's type from rtx to rtx*.
1186 * config/avr/avr.c (avr_legitimize_reload_address): Ditto.
1187 Pass PX to push_reload instead of &X. Change log messages for
1190 2011-10-21 Roland Stigge <stigge@antcom.de>
1192 PR translation/47064
1193 * params.def: Fix typo "compilatoin" -> "compilation".
1195 2011-10-21 H.J. Lu <hongjiu.lu@intel.com>
1196 Kirill Yukhin <kirill.yukhin@intel.com>
1199 * config/i386/driver-i386.c (host_detect_local_cpu): Do cpuid 7 only
1200 if max_level allows that.
1202 2011-10-21 Bernd Schmidt <bernds@codesourcery.com>
1204 * reg-notes.def (DEP_CONTROL): New.
1205 * sched-ebb.c (add_deps_for_risky_insns): Add a REG_DEP_CONTROL when
1206 not doing speculation.
1207 * rtlanal.c (record_hard_reg_sets, find_all_hard_reg_sets,
1208 record_hard_reg_uses_1, record_hard_reg_uses): New functions.
1209 * function.c (record_hard_reg_sets, record_hard_reg_uses,
1210 record_hard_reg_uses_1): Remove; move to rtlanal.c.
1211 * lists.c (copy_INSN_LIST, concat_INSN_LIST): New functions.
1212 * haifa-sched.c: Swap includes of "rtl.h" and "hard-reg-set.h".
1213 (MUST_RECOMPUTE_SPEC_P): New macro.
1214 (real_insn_for_shadow): New function.
1215 (cond_clobbered_p, recompute_todo_spec, check_clobbered_conditions,
1216 toggle_cancelled_flags): New static functions.
1217 (schedule_insn): Relax an assert to only check for empty hard back
1218 dependencies. Skip cancelled dependencies. Call
1219 check_clobbered_conditions.
1220 (copy_insn_list): Remove function, renamed moved to lists.c.
1221 (save_backtrack_point): Use new spelling copy_INSN_LIST.
1222 (unschedule_insns_until): Ensure TODO_SPEC is reset properly.
1223 (restore_last_backtrack_point): Likewise. Call toggle_cancelled_flags.
1224 (estimate_insn_tick): Ignore cancelled dependencies.
1225 (haifa_speculate_insn): Move declaration.
1226 (try_ready): Move code into recompute_todo_spec and call it. Tweak
1227 some asserts. Ensure predicated patterns are restored if necessary.
1228 Dump DEP_CONTROL flag.
1229 (haifa_change_pattern): Merge with sched_change_pattern.
1230 (sched_change_pattern): Remove function.
1231 * sched-deps.c (NON_FLUSH_JUMP_KIND, NON_FLUSH_JUMP): Remove. All
1232 uses changed to simply not test NON_FLUSH_JUMP_P.
1233 (ds_to_dk, dk_to_ds, dump_dep, ds_to_dt, dump_ds, check_dep): Handle
1235 (dep_spec_p): If DO_PREDICATION, REG_DEP_CONTROL is speculative.
1236 (reg_pending_control_uses, control_dependency_cache): New static
1238 (sched_get_reverse_condition_uncached): New function.
1239 (sd_find_dep_between): Remove pointless assert. Look in
1240 control_dependency_cache.
1241 (ask_dependency_caches, set_dependency_caches, sd_delete_dep,
1242 extend_dependency_caches, sched_deps_finish): Handle REG_DEP_CONTROL
1243 and control_dependency_cache.
1244 (sd_unresolve_dep): Use dep_spec_p.
1245 (add_dependence): Now a wrapper around add_dependence_1, handling
1246 REG_DEP_CONTROL specially.
1247 (flush_pending_lists): Clear pending_jump_insns.
1248 (sched_analyze_1): Handle pending_jump_insns like a memory flush.
1249 (sched_analyze_2): Unconditionally add to pending memory flushes,
1250 keep previous behaviour but apply it to pending_jump_insns instead.
1251 (sched_analyze_insn): Defer adding jump reg dependencies using
1252 reg_pending_control_uses; add them to the control_uses list. Handle
1253 pending_jump_insns and control_uses when adding dependence lists.
1254 (deps_analyze_insn): Update INSN_COND_DEPS.
1255 (deps_analyze_insn): Add jumps to pending_jump_insns rather than
1256 last_pending_memory_flush.
1257 (init_deps): Initialize pending_jump_insns.
1258 (free_deps): Free control_uses.
1259 (remove_from_deps): Remove from pending_jump_insns.
1260 (init_deps_global): Allocate reg_pending_control_uses).
1261 (finish_deps_global): Free it.
1262 (add_dependence_1): Renamed from add_dependence. Handle
1264 * rtl.h (record_hard_reg_uses, find_all_hard_reg_sets): Declare.
1265 (copy_INSN_LIST, concat_INSN_LIST): Declare.
1266 * sched-int.h (struct deps_reg): Add control_uses.
1267 (struct deps_desc): Add pending_jump_insns.
1268 (struct _haifa_deps_insn_data): Add cond_deps.
1269 (struct _haifa_insn_data): Add must_recompute_spec and predicated_pat.
1270 (INSN_COND_DEPS, PREDICATED_PAT): New macros.
1271 (BITS_PER_DEP_WEAK): Adjust for two extra bits in the word.
1272 (DEP_CONTROL): New macro.
1273 (DEP_TYPES): Include it.
1274 (HARD_DEP): Adjust definition.
1275 (DEP_CANCELLED): New macro.
1276 (enum SCHED_FLAGS): Add DO_PREDICATION.
1277 (sched_get_reverse_condition_uncached, real_insn_for_shadow): Declare.
1278 * sched-rgn.c (concat_INSN_LIST): Remove function.
1279 (deps_join): Handle pending_jump_insns.
1280 (free_pending_lists): Likewise.
1281 * config/c6x/c6x.c (c6x_set_sched_flags): Set DO_PREDICATION for final
1284 2011-10-21 Georg-Johann Lay <avr@gjlay.de>
1287 * config/avr/libgcc.S (__EIND__): New define to 0x3C.
1288 (__tablejump__): Consistently use EIND for indirect jump/call.
1289 (__tablejump_elpm__): Ditto.
1291 2011-10-21 Bernd Schmidt <bernds@codesourcery.com>
1293 * config/c6x/c6x.md (attr "op_pattern"): New.
1294 (load_sdata_pic, mov<mode>_insn for QIHIM and SISFVM): Set it.
1295 * config/c6x/c6x-mult.md.in (mulhi3_VARIANT_, mulhisi3_insn_VARIANT_):
1297 * config/c6x/c6x-mult.md: Regenerate.
1298 * config/c6x/c6x.c: Include "regrename.h".
1299 (unit_req_table): New typedef.
1300 (unit_reqs): Use it for the declaration.
1301 (unit_req_factor, get_unit_reqs, merge_unit_reqs, unit_req_imbalance,
1302 get_unit_operand_masks, try_rename_operands, reshuffle_units): New
1304 (count_unit_reqs): New arg reqs. All callers changed. Use
1305 get_unit_reqs, and don't merge here.
1306 (res_mii): New arg reqs. All callers changed. Rewrite to use a loop
1307 using unit_req_factor.
1308 (hwloop_optimize): Call reshuffle_units. Call merge_unit_reqs after
1310 (c6x_reorg): Add reg notes problem, and call df_analyze.
1311 * Makefile.in ($(out_object_file)): Depend on regrename.h.
1313 2011-10-21 Kai Tietz <ktietz@redhat.com>
1315 * fold-const.c (simple_operand_p_2): Handle integral
1316 casts from boolean-operands.
1318 2011-10-21 Jan Hubicka <jh@suse.cz>
1320 * cgraph.c (dump_cgraph_node): Dump alias flag.
1321 * cgraphunit.c (handle_alias_pairs): Handle weakrefs with
1323 (get_alias_symbol): New function.
1324 (output_weakrefs): Output also weakrefs with no destinatoin.
1325 (lto_output_node): Output weakref alias flag when at function boundary.
1327 2011-10-21 Andrew Stubbs <ams@codesourcery.com>
1330 * config/arm/driver-arm.c (vendors): Make static.
1332 2011-10-21 Uros Bizjak <ubizjak@gmail.com>
1334 * config/i386/i386.c (ix86_emit_swdivsf): Force b into register.
1335 (ix86_emit_swsqrtsf): Force a into register.
1337 2011-10-20 Mike Stump <mikestump@comcast.net>
1339 * regcprop.c (copyprop_hardreg_forward_1): Update recog_data
1340 after validate_change wipes it out.
1342 2011-10-20 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1344 * config/spu/spu.md ("vec_permv16qi"): Reduce selector modulo 32
1345 before using the shufb instruction.
1347 2011-10-20 Kirill Yukhin <kirill.yukhin@intel.com>
1350 * config/i386/i386.md (bmi_bextr_<mode>): Update register/
1351 memory operand order.
1352 (bmi2_bzhi_<mode>3): Ditto.
1353 (bmi2_pdep_<mode>3): Ditto.
1354 (bmi2_pext_<mode>3): Ditto.
1356 2011-10-20 Richard Henderson <rth@redhat.com>
1358 * target.def (vec_perm_const_ok): Rename from builtin_vec_perm_ok.
1359 * optabs.c (can_vec_perm_expr_p): Update to match.
1360 (expand_vec_perm_expr): Likewise.
1361 * config/i386/i386.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Rename
1362 from TARGET_VECTORIZE_BUILTIN_VEC_PERM_OK.
1363 * doc/tm.texi.in: Likewise.
1365 2011-10-20 Sergey Ostanevich <sergos.gnu@gmail.com>
1368 * config/i386/i386.c (processor_target_table): Change Atom
1369 align_loops_max_skip to 15.
1371 2011-10-20 Richard Henderson <rth@redhat.com>
1373 * target.def (builtin_vec_perm): Remove.
1374 * doc/tm.texi.in (TARGET_VECTORIZE_BUILTIN_VEC_PERM): Remove.
1376 * config/i386/i386.c (ix86_expand_vec_perm_builtin): Remove.
1377 (IX86_BUILTIN_VEC_PERM_*): Remove.
1378 (bdesc_args): Remove vec_perm builtins
1379 (ix86_expand_builtin): Likewise.
1380 (ix86_expand_vec_perm_const_1): Rename from
1381 ix86_expand_vec_perm_builtin_1.
1382 (extract_vec_perm_cst): Merge into...
1383 (ix86_vectorize_vec_perm_const_ok): ... here. Rename from
1384 ix86_vectorize_builtin_vec_perm_ok.
1385 (TARGET_VECTORIZE_BUILTIN_VEC_PERM): Remove.
1387 * config/rs6000/rs6000.c (rs6000_builtin_vec_perm): Remove.
1388 (TARGET_VECTORIZE_BUILTIN_VEC_PERM): Remove.
1390 * config/spu/spu.c (spu_builtin_vec_perm): Remove.
1391 (TARGET_VECTORIZE_BUILTIN_VEC_PERM): Remove.
1393 2011-10-20 Uros Bizjak <ubizjak@gmail.com>
1396 * config/i386/i386.h (RECIP_MASK_DEFAULT): New define.
1397 * config/i386/i386.op (recip_mask): Initialize with RECIP_MASK_DEFAULT.
1398 * doc/invoke.texi (ix86 Options, -mrecip): Document that GCC
1399 implements vectorized single float division and vectorized sqrtf(x)
1400 with reciprocal sequence with additional Newton-Raphson step with
1403 2011-10-20 Dodji Seketeli <dodji@redhat.com>
1405 * ggc-zone.c (ggc_internal_alloc_zone_stat): Rename
1406 ggc_alloced_size_order_for_request into ggc_round_alloc_size like
1407 it was done in ggc-page.c.
1410 * doc/cppopts.texi: Use @smallexample/@end smallexample in
1411 documentation for -fdebug-cpp instead of @quotation/@end quotation
1412 that is not supported by contrib/texi2pod.pl.
1414 2011-10-19 Jan Hubicka <jh@suse.cz>
1416 * ipa-inline.c (inline_small_functions): Always update all calles after
1419 2011-10-19 Jan Hubicka <jh@suse.cz>
1422 * ipa-inline.c (inline_small_functions): Fix checking code to not make
1423 effect on fibheap stability.
1425 2011-10-20 Maxim Kuvyrkov <maxim@codesourcery.com>
1427 * config/m68k/t-linux (M68K_MLIB_CPU): Add ColdFire CPUs.
1429 2011-10-20 Ramana Radhakrishnan <ramana.radhakrishnan@linaro.org>
1432 * config/arm/arm.c (thumb_unexpanded_epilogue): Handle return
1435 2011-10-20 Richard Guenther <rguenther@suse.de>
1437 * tree-ssa-loop-im.c (stmt_cost): Add WIDEN_*, FMA_EXPR
1438 and rotates to the set of expensive operations.
1440 2011-10-19 David S. Miller <davem@davemloft.net>
1442 * config/sparc/sparc.c (sparc_expand_move): Use can_create_pseudo_p.
1443 (sparc_emit_set_const32): Likewise.
1444 (sparc_emit_set_const64_longway): Likewise.
1445 (sparc_emit_set_const64): Likewise.
1446 (sparc_legitimize_pic_address): Likewise.
1447 (memory_ok_for_ldd): Likewise.
1449 2011-10-20 Dehao Chen <dehao@google.com>
1451 * profile.c (compute_branch_probabilities): Compute and dump the
1452 overlap between the static estimation and the instrumentation profile.
1453 (OVERLAP_BASE): New macro.
1454 (compute_frequency_overlap): New function
1456 2011-10-19 Jakub Jelinek <jakub@redhat.com>
1458 * config/i386/i386.c (expand_vec_perm_vpshufb2_vpermq_even_odd): Use
1459 d->op1 instead of d->op0 for the second vpshufb.
1460 (expand_vec_perm_even_odd_1): For V8SImode fix vpshufd immediates.
1461 (ix86_expand_vec_perm_const): If mask indicates two operands are
1462 needed, but both are the same and expanding them as d.op0 == d.op1
1463 failed, retry with d.op0 != d.op1.
1464 (ix86_expand_vec_perm_builtin): Likewise. Handle sorry printing
1465 also for d.nelt == 32.
1468 * cfgexpand.c (expand_debug_expr): Handle WIDEN_LSHIFT_EXPR, ignore
1471 2011-10-19 Bernd Schmidt <bernds@codesourcery.com>
1473 * regrename.h: New file.
1474 * regrename.c: Include it. Also include "emit-rtl.h".
1475 (struct du_head, struct du_chain, du_head_p DEF_VEC and
1476 DEF_VEC_ALLOC_P): Move to regrename.h.
1477 (do_replace): Remove declaration.
1478 (insn_rr): New variable.
1479 (cur_operand): New static variable.
1480 (regrename_chain_from_id): Renamed from chain_from_id and no longer
1481 static. All callers changed.
1482 (record_operand_use): New static function.
1483 (scan_rtx_reg): Use it.
1484 (find_best_rename_reg): New function, broken out of rename_chains.
1485 (rename_chains): Use it. Don't update chain regno and nregs here, ...
1486 (regrename_do_replace): ... do it here instead. Renamed from
1487 do_replace, and no longer static. All callers changed.
1488 (regrename_analyze): No longer static. New arg bb_mask.
1489 All callers changed. If bb_mask is nonzero, use it to limit the
1490 number of basic blocks we analyze. If we failed to analyze a block,
1491 clear insn operand data.
1492 (record_out_operands): New arg insn_info. Update cur_operand if it is
1494 (build_def_use): If insn_rr is nonnull, pass an insn_info to
1495 record_out_operands, and update cur_operand here as well.
1496 (regrename_init, regrename_finish): New functions.
1497 (regrename_optimize): Use them.
1498 * Makefile.in (regrename.o): Adjust dependencies.
1500 2011-10-19 Tom de Vries <tom@codesourcery.com>
1502 PR tree-optimization/50769
1503 * tree-ssa-tail-merge.c (replace_block_by): Calculate phi_vuse2
1504 unconditionally. Handle case that phi_vuse2 is not an SSA_NAME. Add
1505 dummy argument .MEM to phi when increasing number of arguments of phi by
1506 redirecting edges to the block with phi.
1508 2011-10-19 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1511 * configure.ac: Save and restore CXXFLAGS around
1512 gcc_AC_CHECK_DECLS uses.
1513 Check for madvise() declaration with g++ if --enable-build-with-cxx.
1514 * configure: Regenerate.
1515 * config.in: Regenerate.
1516 * ggc-page.c (USING_MADVISE): Also check HAVE_DECL_MADVISE.
1518 2011-10-19 Alexandre Oliva <aoliva@redhat.com>
1521 * var-tracking.c (loc_exp_dep, onepart_aux): New structs.
1522 (variable_part): Replace offset with union.
1523 (enum onepart_enum, onepart_enum_t): New.
1524 (variable_def): Drop cur_loc_changed, add onepart.
1525 (value_chain_def, const_value_chain): Remove.
1526 (VAR_PART_OFFSET, VAR_LOC_1PAUX): New macros, with checking.
1527 (VAR_LOC_DEP_LST, VAR_LOC_DEP_LSTP): New macros.
1528 (VAR_LOC_FROM, VAR_LOC_DEPTH, VAR_LOC_DEP_VEC): Likewise.
1529 (value_chain_pool, value_chains): Remove.
1530 (dropped_values): New.
1531 (struct parm_reg): Only if HAVE_window_save.
1532 (vt_stack_adjustments): Don't record register arguments.
1534 (dv_onepart_p): Return a onepart_enum_t.
1535 (onepart_pool): New.
1538 (variable_htab_free): Release onepart aux data. Reset flags.
1539 (value_chain_htab_hash, value_chain_htab_eq): Remove.
1540 (unshare_variable): Use onepart field. Propagate onepart aux
1541 data or offset. Drop cur_loc_changed.
1542 (val_store): Cope with NULL insn. Rephrase dump output. Check
1543 for unsuitable locs. Add FIXME on using cselib locs.
1544 (val_reset): Remove FIXME of unfounded concerns.
1545 (val_resolve): Check for unsuitable locs. Add FIXME on using
1547 (variable_union): Use onepart field, adjust access to offset.
1549 (VALUE_CHANGED, DECL_CHANGED): Update doc.
1550 (set_dv_changed): Clear NO_LOC_P when changed.
1551 (find_loc_in_1pdv): Use onepart field.
1552 (intersect_loc_chains): Likewise.
1553 (unsuitable_loc): New.
1554 (loc_cmp): Keep ENTRY_VALUEs at the end of the loc list.
1555 (add_value_chain, add_value_chains): Remove.
1556 (add_cselib_value_chains, remove_value_chain): Likewise.
1557 (remove_value_chains, remove_cselib_value_chains): Likewise.
1558 (canonicalize_loc_order_check): Use onepart. Drop cur_loc_changed.
1559 (canonicalize_values_star, canonicalize_vars_star): Use onepart.
1560 (variable_merge_over_cur): Likewise. Adjust access to offset.
1561 Drop cur_loc_changed.
1562 (variable_merge_over_src): Use onepart field.
1563 (remove_duplicate_values): Likewise.
1564 (variable_post_merge_new_vals): Likewise.
1565 (find_mem_expr_in_1pdv): Likewise.
1566 (dataflow_set_preserve_mem_locs): Likewise. Drop cur_loc_changed
1568 (dataflow_set_remove_mem_locs): Likewise. Use VAR_LOC_FROM.
1569 (variable_different_p): Use onepart field. Move onepart test out
1571 (argument_reg_set): Drop.
1572 (add_uses, add_stores): Preserve but do not record in dynamic
1573 tables equivalences for ENTRY_VALUEs and CFA_based addresses.
1574 Avoid unsuitable address expressions.
1575 (EXPR_DEPTH): Unlimit.
1576 (EXPR_USE_DEPTH): Repurpose PARAM_MAX_VARTRACK_EXPR_DEPTH.
1577 (prepare_call_arguments): Use DECL_RTL_IF_SET.
1578 (dump_var): Adjust access to offset.
1579 (variable_from_dropped, recover_dropped_1paux): New.
1580 (variable_was_changed): Drop cur_loc_changed. Use onepart.
1581 Preserve onepart aux in empty_var. Recover empty_var and onepart
1582 aux from dropped_values.
1583 (find_variable_location_part): Special-case onepart. Adjust
1585 (set_slot_part): Use onepart. Drop cur_loc_changed. Adjust
1586 access to offset. Initialize onepaux. Drop value chains.
1587 (delete_slot_part): Drop value chains. Use VAR_LOC_FROM.
1588 (VEC (variable, heap), VEC (rtx, stack)): Define.
1589 (expand_loc_callback_data): Drop dummy, cur_loc_changed,
1590 ignore_cur_loc. Add expanding, pending, depth.
1591 (loc_exp_dep_alloc, loc_exp_dep_clear): New.
1592 (loc_exp_dep_insert, loc_exp_dep_set): New.
1593 (notify_dependents_of_resolved_value): New.
1594 (update_depth, vt_expand_var_loc_chain): New.
1595 (vt_expand_loc_callback): Revamped.
1596 (resolve_expansions_pending_recursion): New.
1597 (INIT_ELCD, FINI_ELCD): New.
1598 (vt_expand_loc): Use the new macros above. Drop ignore_cur_loc
1599 parameter, adjust all callers.
1600 (vt_expand_loc_dummy): Drop.
1601 (vt_expand_1pvar): New.
1602 (emit_note_insn_var_location): Operate on non-debug decls only.
1603 Revamp multi-part cur_loc recomputation and one-part expansion.
1604 Drop cur_loc_changed. Adjust access to offset.
1605 (VEC (variable, heap)): Drop.
1606 (changed_variables_stack, changed_values_stack): Drop.
1607 (check_changed_vars_0, check_changed_vars_1): Remove.
1608 (check_changed_vars_2, check_changed_vars_3): Remove.
1609 (values_to_stack, remove_value_from_changed_variables): New.
1610 (notify_dependents_of_changed_value, process_changed_values): New.
1611 (emit_notes_for_changes): Revamp onepart updates.
1612 (emit_notes_for_differences_1): Use onepart. Drop cur_loc_changed
1613 and value chains. Propagate onepaux. Recover empty_var and onepaux
1614 from dropped_values.
1615 (emit_notes_for_differences_2): Drop value chains.
1616 (emit_notes_in_bb): Adjust.
1617 (vt_emit_notes): Drop value chains, changed_variables_stack.
1618 Initialize and release dropped_values.
1619 (create_entry_value): Revamp.
1620 (vt_add_function_parameter): Use new interface.
1621 (note_register_arguments): Remove.
1622 (vt_initialize): Drop value chains and register arguments.
1623 (vt_finalize): Drop value chains. Release windowed_parm_regs only
1624 if HAVE_window_save.
1625 * rtl.h: Document various pass-local uses of RTL flags.
1626 * tree.h (DECL_RTL_KNOWN_SET): New.
1627 * doc/invoke.texi (param max-vartrack-expr-depth): Update
1628 description and default.
1630 2011-10-19 Georg-Johann Lay <avr@gjlay.de>
1633 * config/avr/avr.md (cc): New alternative out_plus_noclobber.
1634 (adjust_len): Ditto.
1635 (addhi3): Don't pipe through short; use gen_int_mode instead.
1636 Prior to reload, expand to gen_addhi3_clobber.
1637 (*addhi3): Use avr_out_plus_noclobber if applicable, use
1638 out_plus_noclobber in cc and adjust_len attribute.
1639 (addhi3_clobber): 2 new RTL peepholes.
1640 (addhi3_clobber): New insn.
1641 * config/avr/avr-protos.h: (avr_out_plus_noclobber): New prototype.
1642 * config/avr/avr.c (avr_out_plus_noclobber): New function.
1643 (notice_update_cc): Handle CC_OUT_PLUS_NOCLOBBER.
1644 (avr_out_plus_1): Tweak if only MSB is +/-1 and other bytes are 0.
1645 Set cc0 to set_zn for adiw on 16-bit values.
1646 (adjust_insn_length): Handle ADJUST_LEN_OUT_PLUS_NOCLOBBER.
1647 (expand_epilogue): No need to add 0 to frame_pointer_rtx.
1649 2011-10-19 Richard Guenther <rguenther@suse.de>
1652 * tree-ssa-forwprop.c (forward_propagate_into_cond): Verify
1653 the condition is properly gimple before using it.
1654 * tree-eh (stmt_could_throw_1_p): Properly extract the
1655 operation type from comparisons.
1657 2011-10-19 Roland Stigge <stigge@antcom.de>
1659 PR translation/48638
1660 * plugin.c (add_new_plugin): Fix typo in fatal_error message.
1662 2011-10-19 Roland Stigge <stigge@antcom.de>
1664 PR translation/49517
1665 * config/rx/rx.c (rx_print_operand): Fix typo in warning message.
1667 2011-10-19 Richard Guenther <rguenther@suse.de>
1670 * gimple-fold.c (gimplify_and_update_call_from_tree): Rewrite.
1672 2011-10-19 Andrey Belevantsev <abel@ispras.ru>
1674 PR rtl-optimization/50340
1675 * sel-sched-ir.c (update_target_availability): LHS register
1676 availability is not known if the unavailable LHS of the other
1677 expression is a different register.
1679 2011-10-19 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1682 * config/spu/spu.c (spu_emit_vector_compare): Support unordered
1683 floating-point comparisons.
1685 2011-10-19 Jan Hubicka <jh@suse.cz>
1687 * cgraphunit.c (handle_alias_pairs): Also handle wekref with
1688 destination declared.
1689 (output_weakrefs): New function.
1690 * varpool.c (varpool_create_variable_alias): Handle external aliases.
1692 2011-10-19 Jakub Jelinek <jakub@redhat.com>
1694 * dwarf2out.c (loc_descriptor): For SUBREG pass SUBREG_REG's mode as
1695 second argument instead of mode.
1697 2011-10-18 Jakub Jelinek <jakub@redhat.com>
1699 * config/i386/i386.c (ix86_expand_vec_perm): In merge_two use
1700 mode SUBREG of operands[0] as target.
1701 (valid_perm_using_mode_p): Don't ignore higher bits of d->perm.
1702 (expand_vec_pshufb): For V8SImode vmode emit avx2_permvarv8si.
1703 (expand_vec_perm_1): Handle identity and some broadcast
1705 (expand_vec_perm_interleave2): Handle also 32-byte modes, using
1706 vperm2[fi]128 or vpunpck[lh]* followed by single insn permutation.
1707 For d->testing_p return true earlier to avoid creating more GC
1709 (expand_vec_perm_vpermq_perm_1): New function.
1710 (expand_vec_perm_vpshufb2_vpermq): For d->testing_p return true
1711 earlier to avoid creating more GC garbage. Fix handling of
1712 V16HImode. Avoid some SUBREGs in SET_DEST.
1713 (expand_vec_perm_broadcast_1): Return false for 32-byte integer
1715 (expand_vec_perm_vpshufb4_vpermq2): New function.
1716 (ix86_expand_vec_perm_builtin_1): Call expand_vec_perm_vpermq_perm_1
1717 and expand_vec_perm_vpshufb4_vpermq2.
1719 2011-10-18 Andrew Stubbs <ams@codesourcery.com>
1721 * config/arm/driver-arm.c (host_detect_local_cpu): Close the file
1724 2011-10-18 Andrew Stubbs <ams@codesourcery.com>
1726 PR tree-optimization/50717
1727 * tree-ssa-math-opts.c (is_widening_mult_p): Remove the 'type'
1728 parameter. Calculate 'type' from stmt.
1729 (convert_mult_to_widen): Update call the is_widening_mult_p.
1730 (convert_plusminus_to_widen): Likewise.
1732 2011-10-18 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1734 * config/spu/spu.c (struct machine_function): New data structure.
1735 (spu_init_machine_status): New function.
1736 (spu_option_override): Install it.
1737 (get_pic_reg): Set and use cfun->machine->pic_reg.
1738 (spu_split_immediate): Do not set crtl->uses_pic_offset_table.
1739 (need_to_save_reg): Use cfun->machine->pic_reg instead of
1740 checking crtl->uses_pic_offset_table.
1741 (spu_expand_prologue): Likewise.
1743 2011-10-18 Jakub Jelinek <jakub@redhat.com>
1745 PR tree-optimization/50735
1746 * function.c (gimplify_parameters): Use create_tmp_var instead of
1747 create_tmp_reg. If parm is not TREE_ADDRESSABLE and type is complex
1748 or vector type, set DECL_GIMPLE_REG_P.
1750 2011-10-18 Andrew Stubbs <ams@codesourcery.com>
1752 * config.host (arm*-*-linux*): Add driver-arm.o and x-arm.
1753 * config/arm/arm.opt: Add 'native' processor_type and
1754 arm_arch enum values.
1755 * config/arm/arm.h (host_detect_local_cpu): New prototype.
1756 (EXTRA_SPEC_FUNCTIONS): New define.
1757 (MCPU_MTUNE_NATIVE_SPECS): New define.
1758 (DRIVER_SELF_SPECS): New define.
1759 * config/arm/driver-arm.c: New file.
1760 * config/arm/x-arm: New file.
1761 * doc/invoke.texi (ARM Options): Document -mcpu=native,
1762 -mtune=native and -march=native.
1764 2011-10-18 Alexander Monakov <amonakov@ispras.ru>
1766 PR rtl-optimization/50205
1767 * sel-sched.c (count_occurrences_1): Simplify on the assumption that
1768 p->x is a register. Forbid substitution when the same register is
1769 found in a different mode.
1770 (count_occurrences_equiv): Assert that 'what' is a register.
1772 2011-10-18 Richard Guenther <rguenther@suse.de>
1774 PR tree-optimization/50767
1775 * tree-ssa-pre.c (create_expression_by_pieces): Update the
1778 2011-10-18 Julian Brown <julian@codesourcery.com>
1780 * config/arm/arm.c (arm_block_move_unaligned_straight)
1781 (arm_adjust_block_mem, arm_block_move_unaligned_loop)
1782 (arm_movmemqi_unaligned): New.
1783 (arm_gen_movmemqi): Support unaligned block copies.
1785 2011-10-18 Ira Rosen <ira.rosen@linaro.org>
1787 * doc/md.texi (vec_widen_ushiftl_hi, vec_widen_ushiftl_lo,
1788 vec_widen_sshiftl_hi, vec_widen_sshiftl_lo): Document.
1789 * tree-pretty-print.c (dump_generic_node): Handle WIDEN_LSHIFT_EXPR,
1790 VEC_WIDEN_LSHIFT_HI_EXPR and VEC_WIDEN_LSHIFT_LO_EXPR.
1791 (op_code_prio): Likewise.
1792 (op_symbol_code): Handle WIDEN_LSHIFT_EXPR.
1793 * optabs.c (optab_for_tree_code): Handle
1794 VEC_WIDEN_LSHIFT_HI_EXPR and VEC_WIDEN_LSHIFT_LO_EXPR.
1795 (init-optabs): Initialize optab codes for vec_widen_u/sshiftl_hi/lo.
1796 * optabs.h (enum optab_index): Add OTI_vec_widen_u/sshiftl_hi/lo.
1797 * genopinit.c (optabs): Initialize the new optabs.
1798 * expr.c (expand_expr_real_2): Handle
1799 VEC_WIDEN_LSHIFT_HI_EXPR and VEC_WIDEN_LSHIFT_LO_EXPR.
1800 * gimple-pretty-print.c (dump_binary_rhs): Likewise.
1801 * tree-vectorizer.h (NUM_PATTERNS): Increase to 8.
1802 * tree.def (WIDEN_LSHIFT_EXPR, VEC_WIDEN_LSHIFT_HI_EXPR,
1803 VEC_WIDEN_LSHIFT_LO_EXPR): New.
1804 * cfgexpand.c (expand_debug_expr): Handle new tree codes.
1805 * tree-vect-patterns.c (vect_vect_recog_func_ptrs): Add
1806 vect_recog_widen_shift_pattern.
1807 (vect_handle_widen_mult_by_const): Rename...
1808 (vect_handle_widen_op_by_const): ...to this. Handle shifts.
1809 Add a new argument, update documentation.
1810 (vect_recog_widen_mult_pattern): Assume that only second
1811 operand can be constant. Update call to
1812 vect_handle_widen_op_by_const.
1813 (vect_recog_over_widening_pattern): Fix typo.
1814 (vect_recog_widen_shift_pattern): New.
1815 * tree-vect-stmts.c (vectorizable_type_promotion): Handle
1817 (supportable_widening_operation): Likewise.
1818 * tree-inline.c (estimate_operator_cost): Handle new tree codes.
1819 * tree-vect-generic.c (expand_vector_operations_1): Likewise.
1820 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
1821 * config/arm/neon.md (neon_vec_<US>shiftl_<mode>): New.
1822 (vec_widen_<US>shiftl_lo_<mode>, neon_vec_<US>shiftl_hi_<mode>,
1823 vec_widen_<US>shiftl_hi_<mode>, neon_vec_<US>shift_left_<mode>):
1825 * config/arm/predicates.md (const_neon_scalar_shift_amount_operand):
1827 * config/arm/iterators.md (V_innermode): New.
1828 * tree-vect-slp.c (vect_build_slp_tree): Require same shift operand
1831 2011-10-18 Richard Guenther <rguenther@suse.de>
1833 * tree-ssa-alias.h (struct pt_solution): Remove
1834 vars_contains_restrict member.
1835 (pt_solutions_same_restrict_base): Remove.
1836 (pt_solution_set): Adjust.
1837 * tree-ssa-alias.c (ptr_deref_may_alias_decl_p): Remove
1838 vars_contains_restrict handling.
1839 (dump_points_to_solution): Likewise.
1840 (ptr_derefs_may_alias_p): Do not call pt_solutions_same_restrict_base.
1841 * tree-ssa-structalias.c (struct variable_info): Remove is_restrict_var
1843 (new_var_info): Do not initialize it.
1844 (ipa_escaped_pt): Adjust.
1845 (make_constraint_from_restrict): Make the tag global.
1846 (make_constraint_from_global_restrict): New function.
1847 (make_constraint_from_heapvar): Remove.
1848 (create_variable_info_for): Do not make restrict vars point
1850 (intra_create_variable_infos): Likewise.
1851 (find_what_var_points_to): Remove vars_contains_restrict handling.
1852 (pt_solution_set): Adjust.
1853 (pt_solution_ior_into): Likewise.
1854 (pt_solutions_same_restrict_base): Remove.
1855 (compute_points_to_sets): Do not test is_restrict_var.
1856 * cfgexpand.c (update_alias_info_with_stack_vars): Adjust.
1857 * gimple-pretty-print.c (pp_points_to_solution): Likewise.
1859 2011-10-18 Tom de Vries <tom@codesourcery.com>
1861 PR tree-optimization/50672
1862 * tree-ssa-dce.c (mark_virtual_operand_for_renaming): New function,
1864 (mark_virtual_phi_result_for_renaming): Use
1865 mark_virtual_operand_for_renaming.
1866 * tree-flow.h (mark_virtual_operand_for_renaming): Declare.
1867 * tree-ssa-tail-merge.c (release_last_vdef): New function.
1868 (purge_bbs): Add update_vops parameter. Call release_last_vdef for each
1869 deleted basic block.
1870 (tail_merge_optimize): Add argument to call to purge_bbs.
1872 2011-10-18 Richard Guenther <rguenther@suse.de>
1875 * expr.c (get_object_or_type_alignment): New function.
1876 (expand_assignment): Use it.
1877 (expand_expr_real_1): Likewise.
1879 2011-10-18 Dodji Seketeli <dodji@redhat.com>
1882 * input.c (dump_line_table_statistics): Use long, not size_t.
1884 2011-10-17 Eric Botcazou <ebotcazou@adacore.com>
1886 * config/sparc/sparc.md (in_call_delay): Fix formatting issues.
1888 2011-10-17 Simon Baldwin <simonb@google.com>
1889 Ian Lance Taylor <iant@google.com>
1891 * configure.ac: Add --with-native-system-header-dir. Set and
1892 substitute NATIVE_SYSTEM_HEADER_DIR. Use native_system_header
1893 when setting target_header_dir.
1894 * config.gcc: Always set native_system_header_dir.
1895 (*-*-gnu*): Set native_system_header_dir. Don't use t-gnu.
1896 (i[34567]86-pc-msdosdjgpp*): Set native_system_header_dir. Don't
1898 (i[34567]86-*-mingw* | x86_64-*-mingw*): Set
1899 native_system_header_dir.
1900 (spu-*-elf*): Set native_system_header_dir.
1901 * Makefile.in (NATIVE_SYSTEM_HEADER_DIR): Set to
1902 @NATIVE_SYSTEM_HEADER_DIR@.
1903 (PREPROCESSOR_DEFINES): Define NATIVE_SYSTEM_HEADER_DIR.
1904 * cppdefault.c (STANDARD_INCLUDE_DIR): Don't define.
1905 (NATIVE_SYSTEM_HEADER_COMPONENT): Rename from
1906 STANDARD_INCLUDE_COMPONENT.
1907 (cpp_include_defaults): Don't use SYSTEM_INCLUDE_DIR. Rename
1908 STANDARD_INCLUDE_DIR to NATIVE_SYSTEM_HEADER_DIR.
1909 * system.h: Poison SYSTEM_INCLUDE_DIR, STANDARD_INCLUDE_DIR, and
1910 STANDARD_INCLUDE_COMPONENT.
1911 * config/i386/t-mingw32 (NATIVE_SYSTEM_HEADER_DIR): Remove.
1912 * config/i386/t-mingw-w32: Likewise.
1913 * config/i386/t-mingw-w64: Likewise.
1914 * config/spu/t-spu-elf: Likewise.
1915 * config/i386/t-djgpp: Remove.
1916 * config/t-gnu: Remove.
1917 * config/i386/mingw32.h (STANDARD_INCLUDE_DIR): Don't define.
1918 (NATIVE_SYSTEM_HEADER_COMPONENT): Rename from
1919 STANDARD_INCLUDE_COMPONENT.
1920 * config/i386/djgpp.h (STANDARD_INCLUDE_DIR): Don't define.
1921 * config/spu/spu-elf.h: Likewise.
1922 * config/vms/xm-vms.h: Likewise.
1923 * config/gnu.h: Likewise.
1924 * config/openbsd.h (INCLUDE_DEFAULTS): Change STANDARD_INCLUDE_DIR
1925 and STANDARD_INCLUDE_COMPONENT to NATIVE_SYSTEM_HEADER_DIR and
1926 NATIVE_SYSTME_HEADER_COMPONENT.
1927 * doc/install.texi (Configuration): Document
1928 --with-native-system-header-dir. Mention it in the documentation
1929 for --with-sysroot and --with-build-sysroot.
1930 * doc/tm.texi.in (Driver): Don't document SYSTEM_INCLUDE_DIR or
1931 STANDARD_INCLUDE_DIR. Rename STANDARD_INCLUDE_COMPONENT to
1932 NATIVE_SYSTEM_HEADER_COMPONENT. Rename uses of
1933 STANDARD_INCLUDE_DIR to NATIVE_SYSTEM_HEADER_DIR.
1934 * doc/fragments.texi (Target Fragment): Don't document
1935 NATIVE_SYSTEM_HEADER_DIR.
1936 * configure, doc/tm.texi: Rebuild.
1938 2011-10-17 Richard Henderson <rth@redhat.com>
1940 * config/sparc/sparc.md: Use register_or_zero_operand where rJ
1943 * config/sparc/sparc.md (vec_perm_constv8qi, vec_perm<mode>): New
1945 * config/sparc/sparc.c (sparc_expand_vec_perm_bmask): New function.
1946 * config/sparc/sparc-protos.h (sparc_expand_vec_perm_bmask): Declare.
1948 2011-10-17 David S. Miller <davem@davemloft.net>
1950 * config/sparc/sparc-modes.def: Add single entry vector modes for
1952 * config/sparc/sparc.md (V32, V32I, V64, V64I, V64N8): Delete
1954 (mov<V32:mode>): Revert back to plain SFmode pattern.
1955 (*movsf_insn): Likewise.
1956 (mov<V64:mode>): Revert back to plain DFmode pattern.
1957 (*movdf_insn_sp32): Likewise.
1958 (*movdf_insn_sp32_v9): Likewise.
1959 (*movdf_insn_sp64): Likewise.
1960 (V64 mode splitters) Likewise.
1961 (addsi3): Remove VIS alternatives.
1963 (and<V64I:mode>3): Revert to DImode only pattern.
1964 (and<V64I:mode>3_sp32): Likewise.
1965 (*and<V64I:mode>3_sp64): Likewise.
1966 (and<V32I:mode>3): Likewise.
1967 (*and_not_<V64I:mode>_sp32): Likewise.
1968 (*and_not_<V64I:mode>_sp64): Likewise.
1969 (*and_not_<V32I:mode>): Likewise.
1970 (ior<V64I:mode>3): Likewise.
1971 (*ior<V64I:mode>3_sp32): Likewise.
1972 (*ior<V64I:mode>3_sp64): Likewise.
1973 (ior<V32I:mode>3): Likewise.
1974 (*or_not_<V64I:mode>_sp32): Likewise.
1975 (*or_not_<V64I:mode>_sp64): Likewise.
1976 (*or_not_<V32I:mode>): Likewise.
1977 (xor<V64I:mode>3): Likewise.
1978 (*xor<V64I:mode>3_sp32): Likewise.
1979 (*xor<V64I:mode>3_sp64): Likewise.
1980 (xor<V32I:mode>3): Likewise.
1981 (V64I mode splitters): Likewise.
1982 (*xor_not_<V64I:mode>_sp32): Likewise.
1983 (*xor_not_<V64I:mode>_sp64): Likewise.
1984 (*xor_not_<V32I:mode>): Likewise.
1985 (one_cmpl<V64I:mode>2): Likewise.
1986 (*one_cmpl<V64I:mode>2_sp32): Likewise.
1987 (*one_cmpl<V64I:mode>2_sp64): Likewise.
1988 (one_cmpl<V32I:mode>2): Likewise.
1989 (VM32, VM64, VMALL): New mode iterators.
1990 (vbits, vconstr, vfptype): New mode attributes.
1991 (mov<VMALL:mode>): New expander.
1992 (*mov<VM32:mode>_insn): New insn.
1993 (*mov<VM64:mode>_insn_sp64): New insn.
1994 (*mov<VM64:mode>_insn_sp32): New insn, and associated splitter
1995 specifically for the register to memory case.
1996 (vec_init<mode>): New expander.
1997 (VADDSUB): New mode iterator.
1998 (<plusminus_insn>v2si3, <plusminus_insn>v2hi3): Remove and replace
2000 (<plusminus_insn><mode>3): New consolidated pattern.
2001 (VL): New mode iterator for logical operations.
2002 (vlsuf): New more attribute.
2003 (vlop): New code iterator.
2004 (vlinsn, vlninsn): New code attributes.
2005 (<code><mode>3): New insn to non-negated vector logical ops.
2006 (*not_<code><mode>3): Likewise for negated variants.
2007 (*nand<mode>_vis): New insn.
2008 (vlnotop): New code iterator.
2009 (*<code>_not1<mode>_vis, *<code>_not2<mode>_vis): New insns.
2010 (one_cmpl<mode>2): New insn.
2011 (faligndata<V64I:mode>_vis): Rewrite to use VM64 iterator.
2012 (bshuffle<VM64:mode>_vis): Likewise.
2013 (v<vis3_shift_patname><mode>3): Use GCM mode iterator.
2014 (fp<plusminus_insn>64_vis): Use V1DI mode.
2015 (VASS mode iterator): Use V1SI not SI mode.
2016 * config/sparc/sparc.c (sparc_vis_init_builtins): Account for
2017 single-entry vector mode changes.
2018 (sparc_expand_builtin): Likewise.
2019 (sparc_expand_vector_init): New function.
2020 * config/sparc/sparc-protos.h (sparc_expand_vector_init): Declare.
2022 2011-10-17 Kai Tietz <ktietz@redhat.com>
2024 * fold-const.c (simple_operand_p_2): New function.
2025 (fold_truthop): Rename to
2026 (fold_truth_andor_1): function name.
2027 Additionally remove branching creation for logical and/or.
2028 (fold_truth_andor): Handle branching creation for logical and/or here.
2030 2011-10-17 Andi Kleen <ak@linux.intel.com>
2032 * ggc-page.c (USING_MADVISE): Adjust ifdef to check for USING_MMAP.
2034 2011-10-17 Georg-Johann Lay <avr@gjlay.de>
2036 * config/avr/avr.h (ASSEMBLER_DIALECT): Remove.
2037 * config/avr/avr.md (mcu_have_movw, mcu_mega): Remove attributes.
2038 (adjust_len): Add alternative "call".
2039 (isa, enabled): New insn attributes.
2040 (length): Use match_test with AVR_HAVE_JMP_CALL instead of
2042 (*sbrx_branch<mode>): Ditto.
2043 (*sbrx_and_branch<mode>): Ditto.
2044 (*sbix_branch): Ditto.
2045 (*sbix_branch_bit7): Ditto.
2046 (*sbix_branch_tmp): Ditto.
2047 (*sbix_branch_tmp_bit7): Ditto.
2049 (negsi2): Use attribute "isa" instead of assembler dialect.
2050 (extendhisi2): Ditto.
2051 (call_insn, call_value_insn): Set adjust_len attribute.
2052 (indirect_jump): Indent to coding rules.
2053 (call_prologue_saves): Use isa attribute instead of mcu_mega.
2054 (epilogue_restores): Ditto. Fix setting of SP as described in the
2056 (*indirect_jump): Fusion of *jcindirect_jump, *njcindirect_jump
2057 and *indirect_jump_avr6.
2058 (*tablejump): Fusion of *tablejump_rjmp and *tablejump_lib.
2059 (*jcindirect_jump, *njcindirect_jump, *indirect_jump_avr6): Remove.
2060 (*tablejump_rjmp, *tablejump_lib): Remove.
2061 * config/avr/avr.c (adjust_insn_length): Handle ADJUST_LEN_CALL.
2063 2011-10-17 Paolo Carlini <paolo.carlini@oracle.com>
2066 * doc/invoke.texi ([Wnonnull]): Update.
2068 2011-10-17 Richard Henderson <rth@redhat.com>
2071 * optabs.c (expand_vec_perm_expr): Fix indexing error.
2073 2011-10-17 Sergio Durigan Junior <sergiodj@redhat.com>
2075 * configure.ac: Display `yes' if the SystemTap header has been found.
2076 * configure: Regenerate.
2078 2011-10-08 Andi Kleen <ak@linux.intel.com>
2081 * config.in, configure: Regenerate.
2082 * configure.ac (madvise): Add to AC_CHECK_FUNCS.
2083 * ggc-page.c (USING_MADVISE): Add.
2084 (page_entry): Add discarded field.
2085 (alloc_page): Check for discarded pages.
2086 (release_pages): Add USING_MADVISE branch.
2088 2011-10-17 Richard Guenther <rguenther@suse.de>
2090 PR tree-optimization/50729
2091 * tree-vrp.c (extract_range_from_unary_expr_1): Remove redundant test.
2092 (simplify_conversion_using_ranges): Properly test the
2093 intermediate result.
2095 2011-10-15 Tom Tromey <tromey@redhat.com>
2096 Dodji Seketeli <dodji@redhat.com>
2098 * ggc.h (ggc_round_alloc_size): Declare new public entry point.
2099 * ggc-none.c (ggc_round_alloc_size): New public stub function.
2100 * ggc-page.c (ggc_alloced_size_order_for_request): New static
2101 function. Factorized from ggc_internal_alloc_stat.
2102 (ggc_round_alloc_size): New public function. Uses
2103 ggc_alloced_size_order_for_request.
2104 (ggc_internal_alloc_stat): Use ggc_alloced_size_order_for_request.
2105 * ggc-zone.c (ggc_round_alloc_size): New public function extracted
2106 from ggc_internal_alloc_zone_stat.
2107 (ggc_internal_alloc_zone_stat): Use ggc_round_alloc_size.
2108 * toplev.c (general_init): Initialize
2109 line_table->alloced_size_for_request.
2111 2011-10-15 Tom Tromey <tromey@redhat.com>
2112 Dodji Seketeli <dodji@redhat.com>
2114 * input.c (ONE_K, ONE_M, SCALE, STAT_LABEL, FORMAT_AMOUNT): New macros.
2115 (num_expanded_macros_counter, num_macro_tokens_counter): Declare
2117 (dump_line_table_statistics): Define new function.
2118 * input.h (dump_line_table_statistics): Declare new function.
2119 * toplev.c (dump_memory_report): Call dump_line_table_statistics.
2121 2011-10-15 Tom Tromey <tromey@redhat.com>
2122 Dodji Seketeli <dodji@redhat.com>
2124 * doc/cppopts.texi: Document -fdebug-cpp.
2125 * doc/invoke.texi: Add -fdebug-cpp to the list of preprocessor options.
2127 2011-10-15 Tom Tromey <tromey@redhat.com>
2128 Dodji Seketeli <dodji@redhat.com>
2130 * gcc/diagnostic.h (diagnostic_report_current_module): Add a
2132 * diagnostic.c (diagnostic_report_current_module): Add a location
2133 parameter to the function definition. Use it instead of
2134 input_location. Resolve the virtual location rather than just
2135 looking up its map and risking to touch a resulting macro map.
2136 (default_diagnostic_starter): Pass the relevant diagnostic
2137 location to diagnostic_report_current_module.
2138 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): New.
2139 (virt_loc_aware_diagnostic_finalizer): Likewise.
2140 (diagnostic_report_current_function): Pass the
2141 relevant location to diagnostic_report_current_module.
2142 * tree-diagnostic.h (virt_loc_aware_diagnostic_finalizer): Declare
2144 * toplev.c (general_init): By default, use the new
2145 virt_loc_aware_diagnostic_finalizer as diagnostic finalizer.
2146 * Makefile.in: Add vec.h dependency to tree-diagnostic.c.
2148 2011-10-15 Tom Tromey <tromey@redhat.com>
2149 Dodji Seketeli <dodji@redhat.com>
2151 * doc/cppopts.texi (-ftrack-macro-expansion): Document new option.
2152 * doc/invoke.texi (-ftrack-macro-expansion): Add this to the list of
2153 preprocessor related options.
2155 2011-10-15 Tom Tromey <tromey@redhat>
2156 Dodji Seketeli <dodji@redhat.com>
2158 * input.h (struct expanded_location): Move to libcpp/line-map.h.
2159 (LOCATION_COLUMN): New accessor
2160 (in_system_header_at): Use linemap_location_in_system_header_p.
2161 * diagnostic.c (diagnostic_report_current_module): Adjust to avoid
2162 touching the internals of struct line_map. Use the public API instead.
2163 (diagnostic_report_diagnostic): Don't use relational operator '<'
2164 on virtual locations. Use linemap_location_before_p instead.
2165 * input.c (expand_location): Adjust to expand to the tokens'
2166 spelling location when macro location tracking is on.
2169 2011-10-08 Andi Kleen <ak@linux.intel.com>
2171 * ggc-page.c (GGC_QUIRE_SIZE): Increase to 512
2173 2011-10-13 Andi Kleen <ak@linux.intel.com>
2175 * toplev.c (compile_file): Rename __gnu_slim_lto to __gnu_lto_slim.
2177 2011-10-16 Ira Rosen <ira.rosen@linaro.org>
2179 PR tree-optimization/50727
2180 * tree-vect-patterns.c (vect_operation_fits_smaller_type): Add
2181 DEF_STMT to the list of statements to be replaced by the
2184 2011-10-16 Eric Botcazou <ebotcazou@adacore.com>
2186 PR rtl-optimization/50615
2187 * combine.c (distribute_notes) <REG_ARGS_SIZE>: Skip if I3 is a no-op.
2189 2011-10-16 Jakub Jelinek <jakub@redhat.com>
2191 PR tree-optimization/50596
2192 * tree-vectorizer.h (NUM_PATTERNS): Increase to 7.
2193 * tree-vect-patterns.c (vect_vect_recog_func_ptrs): Add
2194 vect_recog_bool_pattern.
2195 (check_bool_pattern, adjust_bool_pattern_cast,
2196 adjust_bool_pattern, vect_recog_bool_pattern): New functions.
2198 2011-10-16 Ira Rosen <ira.rosen@linaro.org>
2200 * tree-vect-stmts.c (vectorizable_load): For SLP without permutation
2201 treat the first load of the node as the first element in its
2203 * tree-vect-slp.c (vect_get_and_check_slp_defs): Swap the operands if
2204 necessary and possible.
2205 (vect_build_slp_tree): Add new argument. Allow load groups of any size
2206 in basic blocks. Keep all the loads for further permutation check.
2207 Use the new argument to determine if there is a permutation. Update
2208 the recursive calls.
2209 (vect_supported_load_permutation_p): Allow subchains of interleaving
2210 chains in basic block vectorization.
2211 (vect_analyze_slp_instance): Update the call to vect_build_slp_tree.
2212 Check load permutation based on the new parameter.
2213 (vect_schedule_slp_instance): Don't start from the first element in
2214 interleaving chain unless the loads are permuted.
2216 2011-10-15 Jan Hubicka <jh@suse.cz>
2220 * cgraphunit.c (cgraph_expand_function): Expand thunks and alises
2221 after function body.
2223 2011-10-15 Richard Henderson <rth@redhat.com>
2225 * tree-vect-slp.c: Include langhooks.h.
2226 (vect_create_mask_and_perm): Emit VEC_PERM_EXPR, not a builtin.
2227 (vect_transform_slp_perm_load): Use can_vec_perm_expr_p. Simplify
2228 mask creation for VEC_PERM_EXPR.
2229 * tree-vect-stmts.c (perm_mask_for_reverse): Return the mask,
2231 (reverse_vec_elements): Emit VEC_PERM_EXPR not a builtin.
2232 * Makefile.in (tree-vect-slp.o): Update dependency.
2233 * optabs.c (can_vec_perm_expr_p): Allow NULL as unknown constant.
2235 2011-10-15 Alan Modra <amodra@gmail.com>
2238 * ifcvt.c (dead_or_predicable): Revert accidental commit with
2239 HAVE_simple_return test.
2241 2011-10-15 Alan Modra <amodra@gmail.com>
2243 * ifcvt.c (dead_or_predicable): Disable if-conversion when
2244 doing so is likely to kill a shrink-wrapping opportunity.
2246 PR rtl-optimization/49941
2247 * jump.c (mark_jump_label_1): Set JUMP_LABEL for simple_return jumps.
2249 * rtl.h (set_return_jump_label): Declare.
2250 * function.c (set_return_jump_label): New function, extracted..
2251 (thread_prologue_and_epilogue_insns): ..from here. Use it in
2252 another instance to set return jump_label.
2253 * cfgrtl.c (force_nonfallthru_and_redirect): Use set_return_jump_label.
2254 * reorg.c (find_end_label): Likewise.
2256 2011-10-14 David S. Miller <davem@davemloft.net>
2258 * config/sparc/sol2.h: Protect -m{cpu,tune}=native handling
2259 with a more complete cpp test.
2260 * config/sparc/linux64.h: Likewise.
2261 * config/sparc/linux.h: Likewise.
2262 * config/sparc/sparc.opt (sparc_debug): New target variable.
2263 (mdebug): New target option.
2264 * config/sparc/sparc.h (MASK_DEBUG_OPTIONS, MASK_DEBUG_ALL,
2265 TARGET_DEBUG_OPTIONS): New defines.
2266 * config/sparc/sparc.c (debug_target_flag_bits,
2267 debug_target_flags): New functions.
2268 (sparc_option_override): Add name strings back to cpu_table[].
2269 Parse -mdebug string. When TARGET_DEBUG_OPTIONS is true, print
2270 out the target flags before and after override processing as well
2271 as the selected cpu. If MASK_V8PLUS, make sure that the selected
2274 2011-10-15 Oleg Endo <oleg.endo@t-online.de>
2277 * config/sh/sh.h (ZERO_EXTRACT_ANDMASK): New macro.
2278 * config/sh/sh.c (sh_rtx_costs): Add test instruction case.
2279 * config/sh/sh.md (tstsi_t): Name existing insn. Make inner
2280 and instruction commutative.
2281 (tsthi_t, tstqi_t, tstqi_t_zero, tstsi_t_and_not,
2282 tstsi_t_zero_extract_eq, tstsi_t_zero_extract_xor,
2283 tstsi_t_zero_extract_subreg_xor_little,
2284 tstsi_t_zero_extract_subreg_xor_big): New insns.
2285 (*movsicc_t_false, *movsicc_t_true): Replace space with tab in
2287 (*andsi_compact): Reorder alternatives so that K08 is considered first.
2289 2011-10-14 Eric Botcazou <ebotcazou@adacore.com>
2292 * config/sparc/linux64.h (TARGET_DEFAULT): Only override if the default
2293 processor is at least V9 and TARGET_64BIT_DEFAULT is defined.
2295 2011-10-14 Gerald Pfeifer <gerald@pfeifer.com>
2297 * invoke.texi (AVR Options): Avoid \leq{}.
2299 2011-10-14 Kai Tietz <ktietz@redhat.com>
2301 * gimplify.c (gimplify_expr): Take care that for bitwise-binary
2302 transformation the operands have compatible types.
2304 2011-10-14 Jakub Jelinek <jakub@redhat.com>
2306 * config/i386/sse.md (vec_widen_smult_hi_v8hi,
2307 vec_widen_smult_lo_v8hi, vec_widen_umult_hi_v8hi,
2308 vec_widen_umult_lo_v8hi): Macroize using VI2_AVX2
2309 mode iterator and any_extend code iterator.
2310 (vec_widen_<s>mult_hi_v8si, vec_widen_<s>mult_lo_v8si): New expanders.
2311 (vec_widen_smult_hi_v4si, vec_widen_smult_lo_v4si): Enable
2312 also for TARGET_SSE4_1 using pmuldq insn.
2313 (sdot_prodv8hi): Macroize using VI2_AVX2 iterator.
2314 (sse2_sse4_1): New code attr.
2315 (udot_prodv4si): Macroize using any_extend code iterator.
2316 (<s>dot_prodv8si): New expander.
2318 2011-10-14 Yakovlev Vladimir <vladimir.b.yakovlev@intel.com>
2320 * config/i386/i386.c (atom_cost): Changed cost for loading
2321 QImode using movzbl.
2323 2011-10-14 Michael Meissner <meissner@linux.vnet.ibm.com>
2325 * config/rs6000/rs6000.c (rs6000_init_builtins): Fix typo in my
2326 change on October 11th, 2011.
2328 2011-10-14 Jakub Jelinek <jakub@redhat.com>
2330 * config/i386/sse.md (vec_interleave_high<mode>,
2331 vec_interleave_low<mode>): Add AVX2 expanders for VI_256 modes.
2332 * config/i386/i386.c (expand_vec_perm_interleave3): New function.
2333 (ix86_expand_vec_perm_builtin_1): Call it.
2335 2011-10-14 Georg-Johann Lay <avr@gjlay.de>
2337 Fix thinko from r179765
2338 * config/avr/avr.c (avr_option_override): Don't override
2339 flag_omit_frame_pointer if not actually needed.
2341 2011-10-14 Georg-Johann Lay <avr@gjlay.de>
2344 * doc/invoke.texi (AVR Options): Document -mstrict-X.
2346 * config/avr/avr.opt (-mstrict-X): New option.
2347 (avr_strict_X): New variable reflecting -mstrict-X.
2348 * config/avr/avr.c (avr_reg_ok_for_addr_p): Add parameter
2349 outer_code and pass it down to avr_regno_mode_code_ok_for_base_p.
2350 (avr_legitimate_address_p): Pass outer_code to
2351 avr_reg_ok_for_addr_p and use that function in case PLUS.
2352 (avr_mode_code_base_reg_class): Depend on avr_strict_X.
2353 (avr_regno_mode_code_ok_for_base_p): Ditto, and depend on outer_code.
2354 (avr_option_override): Disable -fcaller-saves if -mstrict-X is on.
2356 2011-10-14 Jakub Jelinek <jakub@redhat.com>
2358 * config/i386/sse.md (neg<mode>2): Use VI_AVX2 iterator instead
2361 * config/i386/sse.md (mulv2di3): Macroize using VI8_AVX2 iterator.
2362 (ashl<mode>3): Use VI248_AVX2 iterator instead of VI248_128.
2363 Use <sseinsnmode> instead of TI in mode attr.
2365 2011-10-14 David Alan Gilbert <david.gilbert@linaro.org>
2367 * config/arm/linux-atomic-64bit.c: New (based on linux-atomic.c).
2368 * config/arm/linux-atomic.c: Change comment to point to 64bit version.
2369 (SYNC_LOCK_RELEASE): Instantiate 64bit version.
2370 * config/arm/t-linux-eabi: Pull in linux-atomic-64bit.c.
2372 2011-10-14 David Alan Gilbert <david.gilbert@linaro.org>
2374 * config/arm/arm.c (arm_output_ldrex): Support ldrexd.
2375 (arm_output_strex): Support strexd.
2376 (arm_output_it): New helper to output it in Thumb2 mode only.
2377 (arm_output_sync_loop): Support DI mode. Change comment to
2378 not support const_int.
2379 (arm_expand_sync): Support DI mode.
2380 * config/arm/arm.h (TARGET_HAVE_LDREXBHD): Split into LDREXBH
2382 * config/arm/iterators.md (NARROW): move from sync.md.
2383 (QHSD): New iterator for all current ARM integer modes.
2384 (SIDI): New iterator for SI and DI modes only.
2385 * config/arm/sync.md (sync_predtab): New mode_attr.
2386 (sync_compare_and_swapsi): Fold into sync_compare_and_swap<mode>.
2387 (sync_lock_test_and_setsi): Fold into sync_lock_test_and_setsi<mode>.
2388 (sync_<sync_optab>si): Fold into sync_<sync_optab><mode>.
2389 (sync_nandsi): Fold into sync_nand<mode>.
2390 (sync_new_<sync_optab>si): Fold into sync_new_<sync_optab><mode>.
2391 (sync_new_nandsi): Fold into sync_new_nand<mode>.
2392 (sync_old_<sync_optab>si): Fold into sync_old_<sync_optab><mode>.
2393 (sync_old_nandsi): Fold into sync_old_nand<mode>.
2394 (sync_compare_and_swap<mode>): Support SI & DI.
2395 (sync_lock_test_and_set<mode>): Likewise.
2396 (sync_<sync_optab><mode>): Likewise.
2397 (sync_nand<mode>): Likewise.
2398 (sync_new_<sync_optab><mode>): Likewise.
2399 (sync_new_nand<mode>): Likewise.
2400 (sync_old_<sync_optab><mode>): Likewise.
2401 (sync_old_nand<mode>): Likewise.
2402 (arm_sync_compare_and_swapsi): Turn into iterator on SI & DI.
2403 (arm_sync_lock_test_and_setsi): Likewise.
2404 (arm_sync_new_<sync_optab>si): Likewise.
2405 (arm_sync_new_nandsi): Likewise.
2406 (arm_sync_old_<sync_optab>si): Likewise.
2407 (arm_sync_old_nandsi): Likewise.
2408 (arm_sync_compare_and_swap<mode> NARROW): use sync_predtab, fix indent.
2409 (arm_sync_lock_test_and_setsi<mode> NARROW): Likewise.
2410 (arm_sync_new_<sync_optab><mode> NARROW): Likewise.
2411 (arm_sync_new_nand<mode> NARROW): Likewise.
2412 (arm_sync_old_<sync_optab><mode> NARROW): Likewise.
2413 (arm_sync_old_nand<mode> NARROW): Likewise.
2415 2011-10-14 David Alan Gilbert <david.gilbert@linaro.org>
2418 * config/arm/arm.c (arm_output_sync_loop): Move label before barrier.
2420 2011-10-14 David Alan Gilbert <david.gilbert@linaro.org>
2422 * config/arm/arm.h (TARGET_HAVE_DMB_MCR): MCR Not available in Thumb1.
2424 2011-10-14 Paolo Carlini <paolo.carlini@oracle.com>
2426 * doc/invoke.texi ([Wformat-zero-length]): Tidy.
2428 2011-10-14 Jakub Jelinek <jakub@redhat.com>
2430 * gimple.c (walk_stmt_load_store_addr_ops): Call visit_addr also
2431 on COND_EXPR/VEC_COND_EXPR comparison operands if they are ADDR_EXPRs.
2433 2011-10-14 Richard Guenther <rguenther@suse.de>
2435 PR tree-optimization/50723
2436 * ipa-split.c (split_function): Use GSI_CONTINUE_LINKING.
2438 2011-10-14 Nicola Pero <nicola.pero@meta-innovation.com>
2440 * gengtype.c (files_rules): Added rules for objc/objc-map.h and
2443 2011-10-14 Paolo Carlini <paolo.carlini@oracle.com>
2446 * doc/invoke.texi ([Wformat-zero-length]): Update.
2448 2011-10-14 Iain Sandoe <iains@gcc.gnu.org>
2451 * config/darwin.c (darwin_patch_builtin): Adjust argument type. Only
2452 build for powerpc targets. (darwin_patch_builtins): Only build for
2455 2011-10-14 Jakub Jelinek <jakub@redhat.com>
2457 * config/i386/sse.md (*avx_cvtdq2pd256_2): Rename to...
2458 (avx_cvtdq2pd256_2): ... this.
2459 (sseunpackfltmode): New mode attr.
2460 (vec_unpacks_float_hi_v8hi, vec_unpacks_float_lo_v8hi,
2461 vec_unpacku_float_hi_v8hi, vec_unpacku_float_lo_v8hi): Macroize
2462 using VI2_AVX2 iterator.
2463 (vec_unpacku_float_hi_v8si, vec_unpacku_float_lo_v8si): New expanders.
2465 2011-10-13 David S. Miller <davem@davemloft.net>
2467 * config/sparc/sparc.md (plusminus): New code iterator.
2468 (plusminus_insn): New code attr.
2469 (addv2si3, subv2si3, addv4hi3, subv4hi3, addv2hi3, subv2hi3): Merge
2470 using plusminus and plusminus_insn.
2471 (fpadd64_vis, fpsub64_vis): Likewise.
2473 2011-10-13 Richard Henderson <rth@redhat.com>
2475 * doc/md.texi (vec_perm): Document fallback to byte permutation.
2476 * genopinit.c (optabs): Add vec_perm_const.
2477 * optabs.c (can_vec_perm_expr_p): Rename from expand_vec_perm_expr_p.
2478 Reject non-vector modes. Allow fallback to byte permutation.
2479 (expand_vec_perm_expr_1): Split out from ...
2480 (expand_vec_perm_expr): ... here. Allow fallback to byte permutation.
2481 * optabs.h (DOI_vec_perm_const, vec_perm_const_optab): New.
2482 * tree-vect-generic.c (lower_vec_perm): Update for name change.
2484 2011-10-13 Richard Henderson <rth@redhat.com>
2486 * config/rs6000/altivec.md (vec_permv16qi): New pattern.
2488 * config/rs6000/spu.md (vec_permv16qi): New pattern.
2490 * config/i386/i386.c (ix86_expand_vec_perm_const): New.
2491 * config/i386/i386-protos.h: Update.
2492 * config/i386/sse.md (VEC_PERM_CONST): New mode iterator.
2493 (vec_perm_const<VEC_PERM_CONST>): New expander.
2495 * optabs.c (expand_vector_broadcast): New.
2496 (expand_binop): Expand scalar shifts of vectors to vector shifts
2497 of vectors, if the former isn't supported.
2498 * tree-vect-generic.c (expand_vector_operations_1): Don't do that
2499 here; always simplify to scalar shift of vector if possible.
2501 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Fix mode
2502 test for vector splat.
2504 2011-10-13 Jakub Jelinek <jakub@redhat.com>
2506 * config/i386/sse.md (vec_set<mode>): Change V_128 iterator mode to V.
2508 2011-10-13 Jakub Jelinek <jakub@redhat.com>
2509 Richard Guenther <rguenther@suse.de>
2511 * tree-ssa.c (maybe_optimize_var): Drop TREE_ADDRESSABLE from vector
2512 or complex vars even if their DECL_UID is in not_reg_needs bitmap.
2514 2011-10-13 Jakub Jelinek <jakub@redhat.com>
2516 * config/i386/sse.md (reduc_umin_v8hi): New pattern.
2517 * config/i386/i386.c (ix86_build_const_vector): Handle
2518 also V32QI, V16QI, V16HI and V8HI modes.
2519 (emit_reduc_half): New function.
2520 (ix86_expand_reduc): Use phminposuw insn for V8HImode UMIN.
2521 Use emit_reduc_half helper function.
2523 2011-10-13 Lawrence Crowl <crowl@google.com>
2524 Diego Novillo <dnovillo@google.com>
2526 * lto-streamer-in.c (input_struct_function_base): Factor out of ...
2527 (input_function): ... here.
2528 * lto-streamer-out.c (output_struct_function_base): Factor out of ...
2529 (output_function): ... here.
2531 2011-10-13 Gabriel Charette <gchare@google.com>
2532 Diego Novillo <dnovillo@google.com>
2534 * streamer-hooks.h (struct streamer_hooks): Add hooks
2535 input_location and output_location.
2536 * lto-streamer-in.c (lto_input_location): Use
2537 streamer_hooks.input_location, if set.
2538 * lto-streamer-out.c (lto_output_location): Use
2539 streamer_hooks.output_location, if set.
2541 2011-10-13 Eric Botcazou <ebotcazou@adacore.com>
2543 * doc/invoke.texi (SPARC options): Document -mfix-at697f.
2544 * config/sparc/sparc.opt (mfix-at697f): New option.
2545 * config/sparc/sparc.c (TARGET_MACHINE_DEPENDENT_REORG): Define.
2546 (sparc_reorg): New function.
2548 2011-10-13 Richard Guenther <rguenther@suse.de>
2550 PR tree-optimization/50712
2551 * ipa-split.c (split_function): Always re-gimplify parameters
2552 when they are not gimple vals before passing them. Properly
2553 check for type compatibility.
2555 2011-10-13 Tom de Vries <tom@codesourcery.com>
2557 * function.c (gimplify_parameters): Set number of arguments of call to
2558 BUILT_IN_ALLOCA_WITH_ALIGN to 2.
2560 2011-10-13 Tom de Vries <tom@codesourcery.com>
2562 * emit-rtl.c (set_mem_attributes_minus_bitpos): Set MEM_READONLY_P
2563 for static const strings.
2564 * varasm.c (build_constant_desc): Generate the memory location of the
2565 constant using gen_const_mem.
2567 2011-10-13 Richard Guenther <rguenther@suse.de>
2569 PR tree-optimization/50698
2570 * tree-data-ref.c (split_constant_offset_1): Also process
2573 2011-10-12 David S. Miller <davem@davemloft.net>
2575 * config/sparc/sparc.md (UNSPEC_FPMERGE): Delete.
2576 (UNSPEC_MUL16AU, UNSPEC_MUL8, UNSPEC_MUL8SU, UNSPEC_MULDSU): New
2578 (fpmerge_vis): Remove inaccurate comment, represent using vec_select
2580 (vec_interleave_lowv8qi, vec_interleave_highv8qi): New insns.
2581 (fmul8x16_vis, fmul8x16au_vis, fmul8sux16_vis, fmuld8sux16_vis):
2582 Reimplement as unspecs and remove inaccurate comments.
2583 (vis3_shift_patname): New code attr.
2584 (<vis3_shift_insn><vbits>_vis): Rename to
2585 "v<vis3_shift_patname><mode>3".
2586 (vis3_addsub_ss_patname): New code attr.
2587 (<vis3_addsub_ss_insn><vbits>_vis): Rename to
2588 "<vis3_addsub_ss_patname><mode>3".
2589 * config/sparc/sparc.c (sparc_vis_init_builtins): Update to
2590 accommodate pattern name changes.
2592 * config/sparc/sparc.h: Do not force TARGET_VIS3 and TARGET_FMAF
2593 to zero when assembler lacks support for such instructions.
2594 * config/sparc/sparc.c (sparc_option_override): Clear MASK_VIS3
2595 and MASK_FMAF in defaults when assembler lacks necessary support.
2597 2011-10-12 Jakub Jelinek <jakub@redhat.com>
2599 * config/i386/sse.md (vec_unpacks_lo_<mode>,
2600 vec_unpacks_hi_<mode>, vec_unpacku_lo_<mode>,
2601 vec_unpacku_hi_<mode>): Change VI124_128 mode to VI124_AVX2.
2602 * config/i386/i386.c (ix86_expand_sse_unpack): Handle
2603 V32QImode, V16HImode and V8SImode for TARGET_AVX2.
2605 * config/i386/sse.md (vec_avx2): New mode_attr.
2606 (mulv16qi3): Macroize to cover also mulv32qi3 for TARGET_AVX2 into ...
2607 (mul<mode>3): ... this.
2609 * config/i386/i386.md (UNSPEC_VPERMDI): Remove.
2610 * config/i386/i386.c (ix86_expand_vec_perm): Handle
2611 V16QImode and V32QImode for TARGET_AVX2.
2612 (MAX_VECT_LEN): Increase to 32.
2613 (expand_vec_perm_blend): Add support for 32-byte integer
2614 vectors with TARGET_AVX2.
2615 (valid_perm_using_mode_p): New function.
2616 (expand_vec_perm_pshufb): Add support for 32-byte integer
2617 vectors with TARGET_AVX2.
2618 (expand_vec_perm_vpshufb2_vpermq): New function.
2619 (expand_vec_perm_vpshufb2_vpermq_even_odd): New function.
2620 (expand_vec_perm_even_odd_1): Handle 32-byte integer vectors
2622 (ix86_expand_vec_perm_builtin_1): Try expand_vec_perm_vpshufb2_vpermq
2623 and expand_vec_perm_vpshufb2_vpermq_even_odd.
2624 * config/i386/sse.md (VEC_EXTRACT_EVENODD_MODE): Add for TARGET_AVX2
2625 32-byte integer vector modes.
2626 (vec_pack_trunc_<mode>): Use VI248_AVX2 instead of VI248_128.
2627 (avx2_interleave_highv32qi, avx2_interleave_lowv32qi): Remove pasto.
2628 (avx2_pshufdv3, avx2_pshuflwv3, avx2_pshufhwv3): Generate
2630 (avx2_pshufd_1, avx2_pshuflw_1, avx2_pshufhw_1): Don't use
2631 match_dup, instead add 4 new operands and require they have
2632 right cross-lane values.
2633 (avx2_permv4di): Change into define_expand.
2634 (avx2_permv4di_1): New instruction.
2635 (avx2_permv2ti): Use nonimmediate_operand instead of register_operand
2636 for "xm" constrained operand.
2637 (VEC_PERM_AVX2): Add V32QI and V16QI for TARGET_AVX2.
2639 * config/i386/sse.md (avx2_gathersi<mode>,
2640 avx2_gatherdi<mode>, avx2_gatherdi<mode>256): Add clobber of
2641 match_scratch, change memory_operand to register_operand,
2642 add (mem:BLK (scratch)) use.
2643 (*avx2_gathersi<mode>, *avx2_gatherdi<mode>,
2644 *avx2_gatherdi<mode>256): Add clobber of match_scratch,
2645 add earlyclobber to the output operand and match_scratch,
2646 add (mem:BLK (scratch)) use, change the other mem to match_operand.
2647 Use %p6 instead of %c6 in the pattern.
2648 * config/i386/i386.c (ix86_expand_builtin): Adjust for
2649 operand 2 being a Pmode register_operand instead of memory_operand.
2651 2011-10-12 Kai Tietz <ktietz@redhat.com>
2653 * config/i386/i386.md (simple_return): Disable if TARGET_SEH is active.
2655 2011-10-12 Steve Ellcey <sje@cup.hp.com>
2657 * config/ia64/ia64.c (ia64_init_builtins): Fix unbalanced parentheses.
2659 2011-10-12 Richard Guenther <rguenther@suse.de>
2661 * tree-ssa-alias.c (maybe_skip_until): Cache also at the point
2662 of the first store we visit in a basic-block.
2663 (get_continuation_for_phi): Search for a candidate VUSE that
2664 might dominates all others. Do pairwise disambiguation against
2667 2011-10-12 Paul Koning <pkoning@gcc.gnu.org>
2669 PR tree-optimization/50189
2670 * tree-vrp.c (extract_range_from_assert): Use the type of
2671 the variable, not the limit.
2673 2011-10-12 Richard Guenther <rguenther@suse.de>
2675 PR tree-optimization/50700
2676 * tree-object-size.c (addr_object_size): Simplify and treat
2677 MEM_REF bases consistently.
2679 2011-10-12 Bernd Schmidt <bernds@codesourcery.com>
2681 * function.c (prepare_shrink_wrap, bb_active_p): New function.
2682 (thread_prologue_and_epilogue_insns): Use bb_active_p. Call
2683 prepare_shrink_wrap, then recompute bb_active_p for the last block.
2685 2011-10-12 Joseph Myers <joseph@codesourcery.com>
2688 * convert.c (convert_to_integer): Do not narrow operands of
2689 pointer subtraction.
2691 2011-10-12 Nick Clifton <nickc@redhat.com>
2693 * config/arm/arm.h (EMIT_EABI_ATTRIBUTE): New macro. Used to
2694 emit a .eabi_attribute assembler directive, possibly with a
2696 * config/arm/arm.c (arm_file_start): Use the new macro.
2697 * config/arm/arm-c.c (arm_output_c_attributes): Likewise.
2699 2011-10-12 Georg-Johann Lay <avr@gjlay.de>
2702 * config/avr/avr.md (*movqi): Rename to movqi_insn.
2703 (*call_insn): Rename to call_insn.
2704 (*call_value_insn): Rename to call_value_insn.
2705 * config/avr/avr.c (avr_2word_insn_p): New static function.
2706 (jump_over_one_insn_p): Use it.
2708 2011-10-12 Richard Sandiford <richard.sandiford@linaro.org>
2710 * expr.h (copy_blkmode_to_reg): Declare.
2711 * expr.c (copy_blkmode_to_reg): New function.
2712 (expand_assignment): Don't expand register RESULT_DECLs before
2713 the lhs. Use copy_blkmode_to_reg to copy BLKmode values into a
2714 RESULT_DECL register.
2715 (expand_expr_real_1): Handle BLKmode decls when looking for promotion.
2716 * stmt.c (expand_return): Move BLKmode-to-register code into
2717 copy_blkmode_to_reg.
2719 2011-10-11 Eric Botcazou <ebotcazou@adacore.com>
2722 * config/sparc/sparc.md (mov<I:mode>cc): Do not save comparison code.
2723 (mov<F:mode>cc): Likewise.
2725 2011-10-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2727 * tree-ssa-address.c (copy_ref_info): Remove copy of TREE_THIS_NOTRAP.
2729 2011-10-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2731 * tree.h (built_in_decls): Delete old interface with two parallel
2732 arrays to hold standard builtin declarations, and replace it with
2733 a function based interface that can support creating builtins on
2734 the fly in the future. Change all uses, and poison the old
2735 names. Make sure 0 is not a legitimate builtin index.
2736 (implicit_built_in_decls): Ditto.
2737 (built_in_info): Ditto.
2738 (BUILTIN_VALID_P): Ditto.
2739 (builtin_decl_explicit): Ditto.
2740 (builtin_decl_implicit): Ditto.
2741 (set_builtin_decl): Ditto.
2742 (set_builtin_decl_implicit_p): Ditto.
2743 (builtin_decl_explicit_p): Ditto.
2744 (builtin_decl_implicit_p): Ditto.
2745 * tree-complex.c (expand_complex_libcall): Ditto.
2746 * tree-loop-distribution.c (generate_memset_zero): Ditto.
2747 * tree-ssa-strlen.c (get_string_length): Ditto.
2748 (handle_builtin_strcpy): Ditto.
2749 (handle_builtin_strcat): Ditto.
2750 * tree.c (iterative_hash_expr): Ditto.
2751 (local_define_builtin): Ditto.
2752 (build_common_builtin_nodes): Ditto.
2753 * builtins.c (built_in_decls): Ditto.
2754 (implicit_built_in_decls): Ditto.
2755 (built_in_info): Ditto
2756 (expand_builtin_classify_type): Ditto.
2757 (mathfn_built_in_1): Ditto.
2758 (expand_builtin_cexpi): Ditto.
2759 (expand_builtin_mempcpy_args): Ditto.
2760 (expand_builtin_stpcpy): Ditto.
2761 (gimplify_va_arg_expr): Ditto.
2762 (expand_builtin_sync_operation): Ditto.
2763 (build_builtin_expect_predicate): Ditto.
2764 (fold_builtin_memory_op): Ditto.
2765 (fold_builtin_strcpy): Ditto.
2766 (fold_builtin_stpcpy): Ditto.
2767 (fold_builtin_strncpy): Ditto.
2768 (fold_builtin_interclass_mathfn): Ditto.
2769 (fold_builtin_classify): Ditto.
2770 (fold_builtin_2): Ditto.
2771 (fold_builtin_strstr): Ditto.
2772 (fold_builtin_strrchr): Ditto.
2773 (fold_builtin_strpbrk): Ditto.
2774 (fold_builtin_strcat): Ditto.
2775 (fold_builtin_strncat): Ditto.
2776 (fold_builtin_strcspn): Ditto.
2777 (fold_builtin_fputs): Ditto.
2778 (fold_builtin_sprintf): Ditto.
2779 (fold_builtin_snprintf): Ditto.
2780 (expand_builtin_memory_chk): Ditto.
2781 (fold_builtin_memory_chk): Ditto.
2782 (fold_builtin_stxcpy_chk): Ditto.
2783 (fold_builtin_strncpy_chk): Ditto.
2784 (fold_builtin_strcat_chk): Ditto.
2785 (fold_builtin_strncat_chk): Ditto.
2786 (fold_builtin_sprintf_chk_1): Ditto.
2787 (fold_builtin_snprintf_chk_1): Ditto.
2788 (fold_builtin_printf): Ditto.
2789 (fold_builtin_fprintf): Ditto.
2790 (fold_call_stmt): Ditto.
2791 (set_builtin_user_assembler_name): Ditto.
2792 * tree-emutls.c (emutls_common_1): Ditto.
2793 * omp-low.c (scan_omp): Ditto.
2794 (lower_rec_input_clauses): Ditto.
2795 (lower_reduction_clauses): Ditto.
2796 (expand_parallel_call): Ditto.
2797 (expand_task_call): Ditto.
2798 (maybe_catch_exception): Ditto.
2799 (optimize_omp_library_calls): Ditto.
2800 (expand_omp_for_generic): Ditto.
2801 (expand_omp_for_static_nochunk): Ditto.
2802 (expand_omp_for_static_chunk): Ditto.
2803 (expand_omp_sections): Ditto.
2804 (expand_omp_atomic_fetch_op): Ditto.
2805 (expand_omp_atomic_pipeline): Ditto.
2806 (expand_omp_atomic_mutex): Ditto.
2807 (lower_omp_single_simple): Ditto.
2808 (lower_omp_single_copy): Ditto.
2809 (lower_omp_master): Ditto.
2810 (lower_omp_ordered): Ditto.
2811 (lower_omp_critical): Ditto.
2812 * tree-ssa-ccp.c (optimize_stdarg_builtin): Ditto.
2813 * builtins.c (DEF_BUILTIN_STUB): Ditto.
2814 (BUILT_IN_NONE): Ditto.
2815 * tree-ssa-math-opts.c (execute_optimize_bswap): Ditto.
2816 * gimple-low.c (lower_function_body): Ditto.
2817 (lower_builtin_setjmp): Ditto.
2818 * c-decl.c (merge_decls): Ditto.
2819 * tree-eh.c (lower_resx): Ditto.
2820 (lower_resx): Ditto.
2821 (lower_eh_dispatch): Ditto.
2822 * function (gimplify_parameters): Ditto.
2823 * c-typeck.c (build_function_call_vec): Ditto.
2824 * gimplify.c (build_stack_save_restore): Ditto.
2825 (gimplify_vla_decl): Ditto.
2826 (gimplify_modify_expr_to_memcpy): Ditto.
2827 (gimplify_modify_expr_to_memset): Ditto.
2828 (gimplify_variable_sized_compare): Ditto.
2829 (gimplify_function_tree): Ditto.
2830 * calls.c (emit_call_1): Ditto.
2831 * tree-ssa-forprop.c (simplify_builtin_call): Ditto.
2832 * tree-nested.c (convert_nl_goto_reference): Ditto.
2833 (convert_tramp_reference_op): Ditto.
2834 (finalize_nesting_tree_1): Ditto.
2835 * tree-ssa-loop-prefetch.c (issue_prefetch_ref): Ditto.
2836 (tree_ssa_prefetch_arrays): Ditto.
2837 * tree-streamer-in.c (streamer_get_builtin_tree): Ditto.
2838 * system.h (built_in_decls): Ditto.
2839 (implicit_built_in_decls): Ditto.
2840 * tree-vect-generic.c (expand_vector_operations_1): Ditto.
2841 * config/sparc/sparc.c (sparc_gimplify_va_arg): Ditto.
2842 * config/i386/i386.c (ix86_gimplify_va_arg): Ditto.
2843 (ix86_veclibabi_svml): Ditto.
2844 (ix86_veclibabi_acml): Ditto.
2845 * config/vms/vms.c (vms_patch_builtins): Ditto.
2846 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
2847 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Ditto.
2848 (rs6000_builtin_vectorized_libmass): Ditto.
2849 (rs6000_init_builtins): Ditto.
2850 * config/darwin.c (darwin_override_options): Ditto.
2851 (darwin_patch_builtin): Ditto.
2852 (darwin_rename_builtins): Ditto.
2853 * config/pa/pa.c (pa_init_builtins): Ditto.
2855 2011-10-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2857 * tree.h (copy_ref_info): Expose existing function.
2858 * tree-ssa-loop-ivopts.c (copy_ref_info): Move function to...
2859 * tree-ssa-address.c (copy_ref_info): ...here, and remove static token.
2861 2011-10-11 Georg-Johann Lay <avr@gjlay.de>
2863 * config/avr/avr-protos.h (avr_mode_code_base_reg_class): New
2865 (avr_regno_mode_code_ok_for_base_p): New prototype.
2866 * config/avr/avr.h (BASE_REG_CLASS): Remove.
2867 (REGNO_OK_FOR_BASE_P): Remove.
2868 (REG_OK_FOR_BASE_NOSTRICT_P): Remove.
2869 (REG_OK_FOR_BASE_STRICT_P): Remove.
2870 (MODE_CODE_BASE_REG_CLASS): New define.
2871 (REGNO_MODE_CODE_OK_FOR_BASE_P): New define.
2872 * config/avr/avr.c (avr_mode_code_base_reg_class): New function.
2873 (avr_regno_mode_code_ok_for_base_p): New function.
2874 (avr_reg_ok_for_addr_p): New static function.
2875 (avr_legitimate_address_p): Use it. Beautify.
2877 2011-10-11 Georg-Johann Lay <avr@gjlay.de>
2880 * config/avr/avr.md (cc): Add out_plus attribute alternative.
2881 (addsi3): Use it. Adapt avr_out_plus to new prototype. Use
2882 avr_out_plus for all CONST_INT addends.
2883 * config/avr/avr-protos.h (avr_out_plus): Change prototype.
2884 * config/avr/avr.c (notice_update_cc): Call avr_out_plus on
2886 (avr_out_plus_1): Change prototype and report effect on cc0.
2887 (avr_out_plus): Ditto.
2888 (adjust_insn_length): Adapt call to avr_out_plus to new prototype.
2890 2011-10-11 H.J. Lu <hongjiu.lu@intel.com>
2892 * config/i386/i386.c (ix86_expand_special_args_builtin): Remove
2895 2011-10-11 Artjoms Sinkarovs <artyom.shinkaroff@gmail.com>
2897 * doc/invoke.texi: Document new warning.
2898 * common.opt (Wvector-operation-performance): Define new warning.
2899 * tree-vect-generic.c (expand_vector_piecewise): Warn about expanded
2901 (exapnd_vector_parallel): Warn about expanded vector operation.
2902 (lower_vec_shuffle): Warn about expanded vector operation.
2903 * c-typeck.c (c_build_vec_perm_expr): Store correct location
2904 when creating VEC_PERM_EXPR.
2906 2011-10-11 Richard Guenther <rguenther@suse.de>
2908 PR tree-optimization/50204
2909 * tree-ssa-alias.c (get_continuation_for_phi_1): Split out
2910 two argument handling from ...
2911 (get_continuation_for_phi): ... here. Handle arbitrary number
2914 2011-10-11 Richard Sandiford <richard.sandiford@linaro.org>
2916 * modulo-sched.c: Fix comment typo. Mention the possibility
2917 of using scheduling windows of II+1 cycles.
2919 2011-10-11 Tristan Gingold <gingold@adacore.com>
2921 * doc/invoke.texi (C Dialect Options): Document
2922 -fallow-parameterless-variadic-functions.
2923 * c-parser.c (c_parser_parms_list_declarator): Handle it.
2925 2011-10-10 Georg-Johann Lay <avr@gjlay.de>
2927 * config/avr/avr.c (avr_option_override): Set flag_omit_frame_pointer
2928 to 0 if frame pointer is needed for unwinding.
2930 2011-10-10 Uros Bizjak <ubizjak@gmail.com>
2933 * optabs.h (DOI_vec_perm): Rename from OTI_vec_perm. Move from enum
2934 optab_index to enum direct_optab_index.
2935 (vec_perm_optab): Update.
2937 2011-10-10 Anatoly Sokolov <aesok@post.ru>
2939 * config/cris/cris.c (cris_preferred_reload_class): New function.
2940 (TARGET_PREFERRED_RELOAD_CLASS): Define.
2941 * config/cris/cris.h (PREFERRED_RELOAD_CLASS): Remove.
2943 2011-10-10 Georg-Johann Lay <avr@gjlay.de>
2945 * config/avr/avr.md (*tablejump_rjmp): Change insn condition to
2947 (*tablejump_lib): Change insn condition to AVR_HAVE_JMP_CALL.
2948 (*tablejump_enh, *tablejump): Remove insns.
2949 * config/avr/libgcc.S (__tablejump__): Use RET instead of EIND +
2950 EIJMP for indirect jump. Use LPM Z+ where available.
2952 2011-10-10 Richard Henderson <rth@redhat.com>
2954 * doc/md.texi (vec_perm_const): Fix typo in cindex.
2956 2011-10-10 Kirill Yukhin <kirill.yukhin@intel.com>
2957 Yakovlev Vladimir <vladimir.b.yakovlev@intel.com>
2959 * config/i386/sse.md (fma_fnmsub_<mode>): Fix a typo.
2961 2011-10-10 Richard Guenther <rguenther@suse.de>
2964 * gimple-fold.c (gimplify_and_update_call_from_tree): Do not
2965 mark symbols for renaming. Append the VUSE to all statements
2966 that possibly can have one.
2968 2011-10-10 Richard Guenther <rguenther@suse.de>
2970 * ipa-split.c (pass_split_functions): Add verification TODOs.
2971 (pass_feedback_split_functions): Likewise.
2973 2011-10-10 Richard Guenther <rguenther@suse.de>
2976 * fold-const.c (fold_binary_loc): Canonicalize x*x to pow (x, 2)
2977 only when optimizing.
2979 2011-10-10 Nick Clifton <nickc@redhat.com>
2982 * compare-elim.c (find_comparisons_in_bb): Use df_get_live_in and
2983 df_get_live_out instead of accessing the bitmaps directly.
2984 (execute_compare_elim_after_reload): Remove calls to df_set_flags,
2985 df_live_add_problem and df_analyze.
2987 2011-10-10 Michael Matz <matz@suse.de>
2990 * tree-emutls.c (gen_emutls_addr): Call add_referenced_var.
2992 2011-10-10 Richard Sandiford <richard.sandiford@linaro.org>
2994 * modulo-sched.c (ps_reg_move_info): Add num_consecutive_stages.
2995 (SCHED_FIRST_REG_MOVE, SCHED_NREG_MOVES): Delete.
2996 (node_sched_params): Remove first_reg_move and nreg_moves.
2997 (ps_num_consecutive_stages, extend_node_sched_params): New functions.
2998 (update_node_sched_params): Move up file.
2999 (print_node_sched_params): Print the stage. Don't dump info related
3000 to first_reg_move and nreg_moves.
3001 (set_columns_for_row): New function.
3002 (set_columns_for_ps): Move up file and use set_columns_for_row.
3003 (schedule_reg_move): New function.
3004 (schedule_reg_moves): Call extend_node_sched_params and
3005 schedule_reg_move. Extend size of uses bitmap. Initialize
3006 num_consecutive_stages. Return false if a move could not be
3008 (apply_reg_moves): Don't emit moves here.
3009 (permute_partial_schedule): Handle register moves.
3010 (duplicate_insns_of_cycles): Remove for_prolog. Emit moves according
3011 to the same stage-count test as ddg nodes.
3012 (generate_prolog_epilog): Update calls accordingly.
3013 (sms_schedule): Allow move-scheduling to add a new first stage.
3015 2011-10-10 Richard Sandiford <richard.sandiford@linaro.org>
3017 * modulo-sched.c (ps_insn): Adjust comment.
3018 (ps_reg_move_info): New structure.
3019 (partial_schedule): Add reg_moves field.
3020 (SCHED_PARAMS): Use node_sched_param_vec instead of node_sched_params.
3021 (node_sched_params): Turn first_reg_move into an identifier.
3022 (ps_reg_move): New function.
3023 (ps_rtl_insn): Cope with register moves.
3024 (ps_first_note): Adjust comment and assert that the instruction
3025 isn't a register move.
3026 (node_sched_params): Replace with...
3027 (node_sched_param_vec): ...this vector.
3028 (set_node_sched_params): Adjust accordingly.
3029 (print_node_sched_params): Take a partial schedule instead of a ddg.
3030 Use ps_rtl_insn and ps_reg_move.
3031 (generate_reg_moves): Rename to...
3032 (schedule_reg_moves): ...this. Remove rescan parameter. Record each
3033 move in the partial schedule, but don't emit it here. Don't perform
3034 register substitutions here either.
3035 (apply_reg_moves): New function.
3036 (duplicate_insns_of_cycles): Use register indices directly,
3037 rather than finding instructions using PREV_INSN. Use ps_reg_move.
3038 (sms_schedule): Call schedule_reg_moves before committing to
3039 a partial schedule. Try the next ii if the schedule fails.
3040 Use apply_reg_moves instead of generate_reg_moves. Adjust
3041 call to print_node_sched_params. Free node_sched_param_vec
3042 instead of node_sched_params.
3043 (create_partial_schedule): Initialize reg_moves.
3044 (free_partial_schedule): Free reg_moves.
3046 2011-10-10 Richard Sandiford <richard.sandiford@linaro.org>
3048 * modulo-sched.c (ps_insn): Replace node field with an identifier.
3049 (SCHED_ASAP): Replace with..
3050 (NODE_ASAP): ...this macro.
3051 (SCHED_PARAMS): New macro.
3052 (SCHED_TIME, SCHED_FIRST_REG_MOVE, SCHED_NREG_MOVES, SCHED_ROW)
3053 (SCHED_STAGE, SCHED_COLUMN): Redefine using SCHED_PARAMS.
3054 (node_sched_params): Remove asap.
3055 (ps_rtl_insn, ps_first_note): New functions.
3056 (set_node_sched_params): Use XCNEWVEC. Don't copy across the
3058 (print_node_sched_params): Use SCHED_PARAMS and NODE_ASAP.
3059 (generate_reg_moves): Pass ids to the SCHED_* macros.
3060 (update_node_sched_params): Take a ps insn identifier rather than
3061 a node as parameter. Use ps_rtl_insn.
3062 (set_columns_for_ps): Update for above field and SCHED_* macro changes.
3063 (permute_partial_schedule): Use ps_rtl_insn and ps_first_note.
3064 (optimize_sc): Update for above field and SCHED_* macro changes.
3065 Update calls to try_scheduling_node_in_cycle and
3066 update_node_sched_params.
3067 (duplicate_insns_of_cycles): Adjust for above field and SCHED_*
3068 macro changes. Use ps_rtl_insn and ps_first_note.
3069 (sms_schedule): Pass ids to the SCHED_* macros.
3070 (get_sched_window): Adjust for above field and SCHED_* macro changes.
3071 Use NODE_ASAP instead of SCHED_ASAP.
3072 (try_scheduling_node_in_cycle): Remove node parameter. Update
3073 call to ps_add_node_check_conflicts. Pass ids to the SCHED_* macros.
3074 (sms_schedule_by_order): Update call to try_scheduling_node_in_cycle.
3075 (ps_insert_empty_row): Adjust for above field changes.
3076 (compute_split_row): Use ids rather than nodes.
3077 (verify_partial_schedule): Adjust for above field changes.
3078 (print_partial_schedule): Use ps_rtl_insn.
3079 (create_ps_insn): Take an id rather than a node.
3080 (ps_insn_find_column): Adjust for above field changes.
3082 (ps_insn_advance_column): Adjust for above field changes.
3083 (add_node_to_ps): Remove node parameter. Update call to
3085 (ps_has_conflicts): Use ps_rtl_insn.
3086 (ps_add_node_check_conflicts): Replace node parameter than an id.
3088 2011-10-10 Richard Sandiford <richard.sandiford@linaro.org>
3090 * modulo-sched.c (undo_replace_buff_elem): Delete.
3091 (generate_reg_moves): Don't build and return an undo list.
3092 (free_undo_replace_buff): Delete.
3093 (sms_schedule): Adjust call to generate_reg_moves.
3094 Don't call free_undo_replace_buff.
3096 2011-10-10 Matthias Klose <doko@ubuntu.com>
3098 * common/config/m32c: Remove empty directory.
3100 2011-10-10 Georg-Johann Lay <avr@gjlay.de>
3102 * config/avr/avr.md (*mulhi3_enh): Treat squaring smarter.
3104 2011-10-10 Georg-Johann Lay <avr@gjlay.de>
3107 * config/avr/avr-mcus.def (AVR_MCU): Set .data_section_start of
3108 atmega164a to 0x100.
3110 2011-10-09 Eric Botcazou <ebotcazou@adacore.com>
3112 * tree-ssa-forwprop.c (forward_propagate_into_comparison): Do not make
3113 the replacement if the conversion to the LHS type is not useless.
3115 2011-10-09 Ira Rosen <ira.rosen@linaro.org>
3117 PR tree-optimization/50635
3118 * tree-vect-patterns.c (vect_handle_widen_mult_by_const): Add
3119 DEF_STMT to the list of statements to be replaced by the
3121 (vect_handle_widen_mult_by_const): Don't check TYPE_OUT.
3123 2011-10-09 Anatoly Sokolov <aesok@post.ru>
3125 * system.h: Commit forgotten hunk in previous patch.
3126 (OUTPUT_ADDR_CONST_EXTRA): Poison.
3128 2011-10-08 Nicola Pero <nicola.pero@meta-innovation.com>
3131 * doc/objc.texi (Garbage Collection): Updated example to protect
3132 +initialize against execution in subclasses.
3134 2011-10-07 Richard Henderson <rth@redhat.com>
3136 * doc/extend.texi (__builtin_shuffle): Improve the description to
3137 include the modulus of the selector. Mention OpenCL.
3138 * doc/md.texi (vec_perm, vec_perm_const): Document named patterns.
3140 * tree.def (VEC_PERM_EXPR): Rename from VEC_SHUFFLE_EXPR.
3141 * genopinit.c (optabs): Rename vshuffle to vec_perm.
3142 * c-typeck.c (c_build_vec_perm_expr): Rename from
3143 c_build_vec_shuffle_expr. Update for name changes.
3144 * optabs.c (expand_vec_perm_expr_p): Rename from
3145 expand_vec_shuffle_expr_p.
3146 (expand_vec_perm_expr): Rename from expand_vec_shuffle_expr.
3147 * optabs.h (OTI_vec_perm): Rename from DOI_vshuffle.
3148 (vec_perm_optab): Rename from vshuffle_optab.
3149 * expr.c, gimple-pretty-print.c, gimple.c, gimplify.c,
3150 c-tree.h, c-parser.c, tree-cfg.c, tree-inline.c, tree-pretty-print.c,
3151 tree-ssa-operands.c, tree-vect-generic.c: Update for name changes.
3153 * config/i386/i386.c (ix86_expand_vec_perm): Rename from
3154 ix86_expand_vshuffle.
3155 * config/i386/i386-protos.h: Update.
3156 * config/i386/sse.md (VEC_PERM_AVX2): Rename from VSHUFFLE_AVX2.
3157 (vec_perm<VEC_PERM_AVX2>): Rename from vshuffle<VSHUFFLE_AVX2>.
3159 2011-10-07 Richard Henderson <rth@redhat.com>
3161 * config/i386/predicates.md (avx2_pblendw_operand): New.
3162 * config/i386/sse.md (sse4_1_pblendw): Un-macroize.
3163 (avx2_pblendw, *avx2_pblendw): New expander and insn.
3165 2011-10-07 Richard Henderson <rth@redhat.com>
3167 * config/i386/i386.c (bdesc_args): Update code for
3168 __builtin_ia32_palignr256. Change type of __builtin_ia32_pslldqi256,
3169 and __builtin_ia32_psrldqi256 to V4DI_FTYPE_V4DI_INT_CONVERT.
3170 (ix86_expand_args_builtin): Handle V4DI_FTYPE_V4DI_INT_CONVERT.
3171 * config/i386/sse.md (mode iterator V16): Add V2TI.
3172 (mode iterator SSESCALARMODE): Use V2TI not V4DI.
3173 (mode attr ssse3_avx2): Add V2TI.
3174 (avx2_lshrqv4di3, avx2_lshlqv4di3): Remove.
3176 2011-10-07 David S. Miller <davem@davemloft.net>
3179 * configure.ac: Add .register directives to VIS3 test.
3180 * configure: Regenerate.
3182 2011-10-07 Richard Henderson <rth@redhat.com>
3184 * config.gcc (x86_64-*): Add core-avx-i, core-avx2 for with_cpu.
3186 2011-10-07 Richard Henderson <rth@redhat.com>
3189 * fold-const.c (fold_checksum_tree): Remove out-of-date assert.
3191 2011-10-07 Andrew Stubbs <ams@codesourcery.com>
3193 * config/arm/predicates.md (shift_amount_operand): Remove constant
3195 (shift_operator): Check range of constants for all shift operators.
3197 2011-10-07 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
3199 * config/s390/s390.c (s390_emit_tls_call_insn): Remove assertion.
3200 Load GOT pointer for non-pic builds.
3201 (s390_load_got): Replace pic_offset_table_rtx with hardcoded r12.
3202 (s390_emit_call): Likewise.
3204 2011-10-07 Tom de Vries <tom@codesourcery.com>
3207 * tree.c (build_common_builtin_nodes): Add local_define_builtin for
3208 BUILT_IN_ALLOCA_WITH_ALIGN. Mark that BUILT_IN_ALLOCA_WITH_ALIGN can
3210 * builtins.c (expand_builtin_alloca): Handle BUILT_IN_ALLOCA_WITH_ALIGN
3211 arglist. Set align for BUILT_IN_ALLOCA_WITH_ALIGN.
3212 (expand_builtin): Handle BUILT_IN_ALLOCA_WITH_ALIGN.
3213 (is_inexpensive_builtin): Handle BUILT_IN_ALLOCA_WITH_ALIGN.
3214 * tree-ssa-ccp.c (evaluate_stmt): Set align for
3215 BUILT_IN_ALLOCA_WITH_ALIGN.
3216 (fold_builtin_alloca_for_var): Rename to ...
3217 (fold_builtin_alloca_with_align): Set DECL_ALIGN from 2nd
3218 BUILT_IN_ALLOCA_WITH_ALIGN argument.
3219 (ccp_fold_stmt): Try folding BUILT_IN_ALLOCA_WITH_ALIGN using
3220 fold_builtin_alloca_with_align.
3221 (optimize_stack_restore): Handle BUILT_IN_ALLOCA_WITH_ALIGN.
3222 * builtins.def (BUILT_IN_ALLOCA_WITH_ALIGN): Declare using
3224 * ipa-pure-const.c (special_builtin_state): Handle
3225 BUILT_IN_ALLOCA_WITH_ALIGN.
3226 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1)
3227 (call_may_clobber_ref_p_1): Same.
3228 * function.c (gimplify_parameters): Lower vla to
3229 BUILT_IN_ALLOCA_WITH_ALIGN.
3230 * gimplify.c (gimplify_vla_decl): Same.
3231 * cfgexpand.c (expand_call_stmt): Handle BUILT_IN_ALLOCA_WITH_ALIGN.
3232 * tree-mudflap.c (mf_xform_statements): Same.
3233 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary)
3234 (mark_all_reaching_defs_necessary_1, propagate_necessity): Same.
3235 * varasm.c (incorporeal_function_p): Same.
3236 * tree-object-size.c (alloc_object_size): Same.
3237 * gimple.c (gimple_build_call_from_tree): Same.
3239 2011-10-07 Bernd Schmidt <bernds@codesourcery.com>
3241 * function.c (frame_required_for_rtx): Remove function.
3242 (requires_stack_frame_p): New arg set_up_by_prologue. All callers
3243 changed. Compute a set of mentioned registers and compare against
3244 the new arg rather than calling frame_required_for_rtx.
3245 (thread_prologue_and_epilogue_insns): Compute the set_up_by_prologue
3246 reg set. Convert the unconverted_simple_returns mechanism to store
3247 jump insns rather than their basic blocks. Also check the
3248 orig_entry_edge destination for new blocks.
3250 2011-10-07 Jakub Jelinek <jakub@redhat.com>
3252 PR tree-optimization/50650
3253 * tree-vect-patterns.c (vect_recog_mixed_size_cond_pattern): Don't
3254 call vect_is_simple_cond here, instead fail if cond_expr isn't
3255 COMPARISON_CLASS_P or if get_vectype_for_scalar_type returns NULL
3256 for cond_expr's first operand.
3257 * tree-vect-stmts.c (vect_is_simple_cond): Static again.
3258 * tree-vectorizer.h (vect_is_simple_cond): Remove prototype.
3260 2011-10-07 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
3262 * config/s390/s390.md (DWH, dwh): New mode macros.
3263 ("umulsidi3"): Extend to support "umulditi3" as well.
3265 2011-10-07 Uros Bizjak <ubizjak@gmail.com>
3266 H.J. Lu <hongjiu.lu@intel.com>
3269 * config/i386/i386.c (ix86_fixup_binary_operands): Force src2 of
3270 integer PLUS RTX to a register to improve address combine.
3272 2011-10-06 Richard Henderson <rth@redhat.com>
3274 * combine-stack-adjust.c (maybe_move_args_size_note): Add after
3275 parameter; use it to decide whether to merge two notes.
3276 (combine_stack_adjustments_for_block): Use maybe_move_args_size_note
3277 for the deallocation case as well.
3279 2011-10-06 Anatoly Sokolov <aesok@post.ru>
3281 * system.h (OUTPUT_ADDR_CONST_EXTRA): Poison.
3282 * doc/tm.texi.in (OUTPUT_ADDR_CONST_EXTRA): Remove documentation.
3283 * doc/tm.texi: Regenerate.
3284 * target.def (output_addr_const_extra): Use
3285 hook_bool_FILEptr_rtx_false.
3286 * targhooks.c (default_asm_output_addr_const_extra): Remove.
3287 * targhooks.h (default_asm_output_addr_const_extra): Remove.
3288 * hooks.c (hook_bool_FILEptr_rtx_false): New functions.
3289 * hooks.h (hook_bool_FILEptr_rtx_false): Declare.
3291 2011-10-06 David S. Miller <davem@davemloft.net>
3293 * config/sparc/sparc.md (popcount<mode>2, clz<mode>2): Split up into...
3294 (popcountdi2, popcountsi2, clzdi2, clzsi2): Explicit expanders, in the
3295 SI mode 64-bit code gen case explicitly zero-extend and truncate.
3296 (*popcount<mode>_sp64): Split up into...
3297 (*popcountdi_sp64, *popcountsi_64): Explicit instantiations, and in the
3298 SImode case use truncate.
3299 (*clzsi_sp64): Rewrite to use truncate, and let the expander emit the
3300 subtract so the compiler can optimize it.
3301 (SIDI): Remove unused mode iterator.
3303 2011-10-06 Bernd Schmidt <bernds@codesourcery.com>
3305 * function.c (thread_prologue_and_epilogue_insns): Emit split prologue
3306 on the orig_entry_edge. Don't account for it in prologue_clobbered.
3308 2011-10-06 Jakub Jelinek <jakub@redhat.com>
3310 PR tree-optimization/50596
3311 * tree-vectorizer.h (vect_is_simple_cond): New prototype.
3312 (NUM_PATTERNS): Change to 6.
3313 * tree-vect-patterns.c (vect_recog_mixed_size_cond_pattern): New
3315 (vect_vect_recog_func_ptrs): Add vect_recog_mixed_size_cond_pattern.
3316 (vect_mark_pattern_stmts): Don't create stmt_vinfo for def_stmt
3317 if it already has one, and don't set STMT_VINFO_VECTYPE in it
3318 if it is already set.
3319 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Handle
3320 COND_EXPR in pattern stmts.
3321 (vect_is_simple_cond): No longer static.
3323 2001-10-06 Richard Henderson <rth@redhat.com>
3325 * config/i386/i386.c (ix86_expand_vshuffle): Add AVX2 support.
3326 * config/i386/sse.md (sseshuffint): Remove.
3327 (sseintvecmode): Support V16HI, V8HI, V32QI, V16QI.
3328 (VSHUFFLE_AVX2): New mode iterator.
3329 (vshuffle<mode>): Use it.
3330 (avx_vec_concat<V_256>): Rename from *vec_concat<V_256>_avx.
3332 * config/i386/i386.c (ix86_expand_sse_movcc): Use correct mode
3333 for vector_all_ones_operand.
3334 (ix86_expand_int_vcond): Distinguish between comparison mode
3335 and data mode. Allow them to differ.
3336 (ix86_expand_vshuffle): Don't force data mode to match maskmode.
3338 2001-10-06 Richard Henderson <rth@redhat.com>
3340 * optabs.c (expand_vec_shuffle_expr): Use the proper mode for the
3341 mask operand. Tidy the code.
3343 2011-10-06 Jakub Jelinek <jakub@redhat.com>
3345 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3346 vect_recog_func_ptr typedef for the first argument.
3347 (vect_pattern_recog): Rename vect_recog_func_ptr variable
3348 to vect_recog_func, use vect_recog_func_ptr typedef for it.
3350 PR tree-optimization/49279
3351 * tree-ssa-structalias.c (find_func_aliases): Don't handle
3353 * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Allow
3354 restrict propagation.
3355 * tree-ssa.c (useless_type_conversion_p): Don't return false
3356 if TYPE_RESTRICT differs.
3358 2011-10-06 Bernd Schmidt <bernds@codesourcery.com>
3360 * function.c (thread_prologue_and_epilogue_insns): Build a vector
3361 of unconverted simple return blocks rather than trying to
3362 recompute them later based on bb_flags bitmap tests.
3364 2011-10-06 Michael Matz <matz@suse.de>
3366 * tree-flow.h (get_var_ann): Don't declare.
3367 * tree-flow-inline.h (get_var_ann): Remove.
3368 (set_is_used): Use var_ann, not get_var_ann.
3369 * tree-dfa.c (add_referenced_var): Inline body of get_var_ann.
3370 * tree-profile.c (gimple_gen_edge_profiler): Call
3371 find_referenced_var_in.
3372 (gimple_gen_interval_profiler): Ditto.
3373 (gimple_gen_pow2_profiler): Ditto.
3374 (gimple_gen_one_value_profiler): Ditto.
3375 (gimple_gen_average_profiler): Ditto.
3376 (gimple_gen_ior_profiler): Ditto.
3377 (gimple_gen_ic_profiler): Ditto plus call add_referenced_var.
3378 (gimple_gen_ic_func_profiler): Call add_referenced_var.
3379 * tree-mudflap.c (execute_mudflap_function_ops): Call
3382 2011-10-06 Jakub Jelinek <jakub@redhat.com>
3384 * tree-vect-patterns.c (vect_handle_widen_mult_by_const): For lhs
3385 don't set SSA_NAME_DEF_STMT that has been already set by
3386 gimple_build_assign_with_ops.
3387 (vect_recog_pow_pattern, vect_recog_widen_sum_pattern,
3388 vect_operation_fits_smaller_type, vect_recog_over_widening_pattern):
3391 * tree.h (avoid_folding_inline_builtin): New prototype.
3392 * builtins.c (avoid_folding_inline_builtin): No longer static.
3393 * gimple-fold.c (gimple_fold_builtin): Give up if
3394 avoid_folding_inline_builtin returns true.
3396 2011-10-06 Richard Guenther <rguenther@suse.de>
3398 * tree-vect-generic.c (vector_element): Look at previous
3401 2011-10-06 David Edelsohn <dje.gcc@gmail.com>
3404 * config/rs6000/aix.h (TARGET_OS_AIX_CPP_BUILTINS): Define
3405 __powerpc__, __PPC__, __unix__.
3407 2011-10-06 Michael Matz <matz@suse.de>
3409 * i386/i386.opt (recip_mask, recip_mask_explicit,
3410 x_recip_mask_explicit): New variables and cl_target member.
3411 (mrecip=): New option.
3412 * i386/i386.h (RECIP_MASK_DIV, RECIP_MASK_SQRT, RECIP_MASK_VEC_DIV,
3413 RECIP_MASK_VEC_SQRT, RECIP_MASK_ALL, RECIP_MASK_NONE): New bitmasks.
3414 (TARGET_RECIP_DIV, TARGET_RECIP_SQRT, TARGET_RECIP_VEC_DIV,
3415 TARGET_RECIP_VEC_SQRT): New tests.
3416 * i386/i386.md (divsf3): Check TARGET_RECIP_DIV.
3417 (sqrt<mode>2): Check TARGET_RECIP_SQRT.
3418 * i386/sse.md (div<mode>3): Check TARGET_RECIP_VEC_DIV.
3419 (sqrt<mode>2): Check TARGET_RECIP_VEC_SQRT.
3420 * i386/i386.c (ix86_option_override_internal): Set recip_mask
3421 for -mrecip and -mrecip=options.
3422 (ix86_function_specific_save): Save recip_mask_explicit.
3423 (ix86_function_specific_restore): Restore recip_mask_explicit.
3425 * doc/invoke.texi (ix86 Options): Document the new option.
3427 2011-10-06 Bernd Schmidt <bernds@codesourcery.com>
3430 * config/arm/arm.md (arm_subsi3_insn): Lose the last alternative.
3432 2011-10-06 Ulrich Weigand <ulrich.weigand@linaro.org>
3435 * config/arm/arm.c (arm_legitimize_reload_address): Recognize
3436 output of a previous pass through legitimize_reload_address.
3437 Do not attempt to optimize addresses if the base register is
3438 equivalent to a constant.
3440 2011-10-06 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
3442 * function.c (thread_prologue_and_epilogue_insns): Mark
3443 last_bb_active as possibly unused. It is unused for targets which
3444 do neither have "return" nor "simple_return" expanders.
3446 2011-10-06 Richard Guenther <rguenther@suse.de>
3448 * fold-const.c (fold_ternary_loc): Also fold non-constant
3449 vector CONSTRUCTORs. Make more efficient.
3450 * tree-ssa-dom.c (cprop_operand): Don't handle virtual operands.
3451 (cprop_into_stmt): Don't propagate into virtual operands.
3452 (optimize_stmt): Really dump original statement.
3454 2011-10-06 Nick Clifton <nickc@redhat.com>
3456 * config/rx/rx.md (smin3): Revert previous delta.
3458 2011-10-06 Richard Guenther <rguenther@suse.de>
3460 PR tree-optimization/38884
3461 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle partial
3462 reads from aggregate SSA names.
3464 2011-10-05 Jakub Jelinek <jakub@redhat.com>
3466 * tree-vect-patterns.c (vect_pattern_recog_1): Add stmts_to_replace
3467 argument, truncate it at the beginning instead of allocating there
3468 and freeing at the end.
3469 (vect_pattern_recog): Allocate stmts_to_replace here and free at end,
3470 pass its address to vect_pattern_recog_1.
3472 2011-10-05 David S. Miller <davem@davemloft.net>
3474 * config/sparc/sparc.opt (POPC): New option.
3475 * doc/invoke.texi: Document it.
3476 * config/sparc/sparc.c (sparc_option_override): Enable MASK_POPC by
3477 default on Niagara-2 and later.
3478 * config/sparc/sparc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3479 * config/sparc/sparc.md (SIDI): New mode iterator.
3480 (ffsdi2): Delete commented out pattern and comments.
3481 (popcount<mode>2, clz<mode>2): New expanders.
3482 (*popcount<mode>_sp64, popcountsi_v8plus, popcountdi_v8plus,
3483 *clzdi_sp64, clzdi_v8plus, *clzsi_sp64, clzsi_v8plus): New insns.
3485 2011-10-06 Artjoms Sinkarovs <artyom.shinkaroff@gmail.com>
3488 * c-tree.h (c_expr_t): New typedef for struct c_expr.
3489 (C_EXPR_APPEND): New macro.
3490 * c-parser.c (c_parser_get_builtin_args): Preserve
3491 original_tree_code of c_expr structure.
3492 (c_parser_postfix_expression): Adjust to the new function.
3494 2011-10-05 Bernd Schmidt <bernds@codesourcery.com>
3496 * function.c (thread_prologue_and_epilogue_insns): Don't shrink-wrap
3497 if profiling after the prologue.
3499 2011-10-05 Jakub Jelinek <jakub@redhat.com>
3501 PR tree-optimization/50613
3502 * tree-ssa-strlen.c (find_equal_ptrs): If CASE_CONVERT
3503 operand is ADDR_EXPR, fallthru into ADDR_EXPR handling,
3504 and if it is neither that not SSA_NAME, give up.
3506 2011-10-05 Richard Henderson <rth@redhat.com>
3508 * tree-vect-generic.c (vector_element): Never fail. Use
3509 build_zero_cst. Tidy up type references.
3510 (lower_vec_shuffle): Never fail. Mask shuffle indicies. Reduce
3511 code duplication. Do update_stmt here ...
3512 (expand_vector_operations_1): ... not here.
3514 * config/i386/i386.c (ix86_expand_vshuffle): Never fail. Handle
3515 TARGET_XOP. Fix pshufb constant vector creation. Reduce code
3516 duplication. Handle V2DI without SSE4.1.
3517 * config/i386/i386-protos.h (ix86_expand_vshuffle): Update decl.
3518 * config/i386/i386.md (vshuffle<V_128>): Remove assert for ok.
3520 2011-10-05 Uros Bizjak <ubizjak@gmail.com>
3522 * config/i386/i386.c (distance_non_agu_define): Simplify calculation
3523 of "found". Simplify return value calculation.
3524 (distance_agu_use): Ditto.
3526 2011-10-05 Bernd Schmidt <bernds@codesourcery.com>
3529 * config/i386/i386.c (ix86_add_cfa_restore_note): Omit notes only
3530 if the function was not shrink-wrapped.
3531 (ix86_expand_epilogue): Ensure queued cfa_adjust notes are attached
3533 * function.c (thread_prologue_and_epilogue_insns): Make sure the
3534 shrink_wrapped flag is set even if there is no dump file.
3536 2011-10-05 DJ Delorie <dj@redhat.com>
3537 Nick Clifton <nickc@redhat.com>
3539 * config/rx/rx.opt (mpid): Define.
3540 * config/rx/t-rx (MULTILIB_OPTIONS): Add -mpid
3541 (MULTILIB_DIRNAMES): Add pid.
3542 * config/rx/rx.c (rx_gp_base_regnum_val, rx_pid_base_regnum_val)
3543 (rx_num_interrupt_regs): New variable.
3544 (rx_gp_base_regnum): New function. Returns the number of the
3545 small data area register.
3546 (rx_pid_base_regnum): New function. Returns the number of the pid
3548 (rx_decl_for_addr): New function. Returns the symbolic part of a MEM.
3549 (rx_pid_data_operand): New function. Returns whether an object is
3550 in the position independent data area.
3551 (rx_legitimize_address): New function. Puts undecided PID
3552 objects in the PID data area.
3553 (rx_is_legitimate_address): Add support for PID operands.
3554 (rx_print_operand_address): Likewise.
3555 (rx_print_operand): Likewise.
3556 (rx_maybe_pidify_operand): New function. Determine if an operand
3557 is suitable for PID addressing.
3558 (rx_gen_move_template): Add PID support.
3559 (rx_conditional_register_usage): Likewise.
3560 (rx_option_override): Initialise rx_num_interrupt_regs.
3561 (rx_is_legitimate_constant): Add support for PID constants.
3562 (TARGET_LEGITIMIZE_ADDRESS): Define.
3563 * config/rx/constraints.md (Rpid): Define.
3565 * config/rx/rx.md (UNSPEC_PID_ADDR): Define.
3566 (tablejump): Add PID support.
3568 (mov<>_internal): Likewise.
3569 (addsi3): Convert to an expander. Add PID support.
3570 (pid_addr): New pattern.
3571 * config/rx/rx.h (CPP_SPEC): Define.
3572 (ASM_SPEC): Pass -mpid and -mint-register on to assembler.
3573 (CASE_VECTOR_PC_RELATIVE): Define.
3574 (JUMP_TABLES_IN_TEXT_SECTION): Enable for PID mode.
3575 * config/rx/rx-protos.h (rx_maybe_pidify_operand): Prototype.
3576 * doc/invoke.texi (RX Options): Document -mpid command line option.
3578 2011-10-05 Richard Guenther <rguenther@suse.de>
3580 PR tree-optimization/38885
3581 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle partial reads
3584 2011-10-05 Bernd Schmidt <bernds@codesourcery.com>
3586 * doc/invoke.texi (-fshrink-wrap): Document.
3587 * opts.c (default_options_table): Add it.
3588 * common.opt (fshrink-wrap): Add.
3589 * function.c (emit_return_into_block): Remove useless declaration.
3590 (record_hard_reg_uses_1, record_hard_reg_uses, frame_required_for_rtx,
3591 requires_stack_frame_p, gen_return_pattern): New static functions.
3592 (emit_return_into_block): New arg simple_p. All callers changed.
3593 Use gen_return_pattern.
3594 (thread_prologue_and_epilogue_insns): Implement shrink-wrapping.
3595 * config/i386/i386.md (return): Expand into a simple_return.
3596 (simple_return): New expander):
3597 (simple_return_internal, simple_return_internal_long,
3598 simple_return_pop_internal_long, simple_return_indirect_internal):
3599 Renamed from return_internal, return_internal_long,
3600 return_pop_internal_long and return_indirect_internal; changed to use
3602 * config/i386/i386.c (ix86_expand_epilogue): Adjust to expand
3604 (ix86_pad_returns): Likewise.
3605 * function.h (struct rtl_data): Add member shrink_wrapped.
3606 * cfgcleanup.c (outgoing_edges_match): If shrink-wrapped, edges that
3607 are not jumps or sibcalls can't be compared.
3609 2011-10-05 Richard Guenther <rguenther@suse.de>
3611 * tree-ssa-sccvn.c (vn_get_expr_for): Handle CONSTRUCTOR of
3613 (simplify_unary_expression): Handle BIT_FIELD_REFs.
3614 (try_to_simplify): Handle BIT_FIELD_REFs.
3616 2011-10-05 Georg-Johann Lay <avr@gjlay.de>
3618 * config/avr/avr-protos.h (avr_out_addto_sp): New prototype.
3619 * config/avr/avr.c (avr_out_addto_sp): New function.
3620 (adjust_insn_length): Handle ADJUST_LEN_ADDTO_SP.
3621 * config/avr/avr.md (adjust_len): Add "addto_sp".
3622 (*movhi_sp): Remove insn.
3623 (*addhi3_sp_R_pc2, *addhi3_sp_R_pc3): Merge to *addhi3_sp_R.
3625 2011-10-05 Richard Guenther <rguenther@suse.de>
3627 * gimple-fold.c (gimple_fold_stmt_to_constant_1): For ternary ops
3628 with an embedded expression valueize and fold that as well.
3629 * tree-ssa-sccvn.c (try_to_simplify): Also allow SSA name
3630 results from gimple_fold_stmt_to_constant_1.
3632 2011-10-05 Nick Clifton <nickc@redhat.com>
3634 * config/rx/rx.md (tablejump): Add missing label.
3635 (adddi3_internal): Mark operand 0 as early-clobbered.
3636 (smaxsi3): Revert previous delta.
3637 (adc_internal): Fix whitespace in generated asm.
3638 (adc_flags): Likewise.
3640 2011-10-05 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
3642 * expmed.c (expand_mult_highpart_optab): Replace optab_handler
3643 with the new widening_optab_handler.
3645 2011-10-05 Richard Guenther <rguenther@suse.de>
3647 PR tree-optimization/50609
3648 * gimple-fold.c (fold_array_ctor_reference): Also handle
3649 vector typed constructors.
3650 (fold_ctor_reference): Dispatch to fold_array_ctor_reference
3651 for vector typed constructors.
3653 2011-10-05 Uros Bizjak <ubizjak@gmail.com>
3655 * config/i386/i386.c (ix86_emit_binop): New static function.
3656 (ix86_split_lea_for_addr): Use ix86_emit_binop to emit add and shl
3658 (x86_output_mi_thunk): Use ix86_emit_binop to emit add instructions.
3660 2011-10-04 David S. Miller <davem@davemloft.net>
3662 * config/sparc/sparc.md (UNSPEC_FHADD, UNSPEC_FHSUB,
3663 UNSPEC_XMUL): New unspecs.
3664 (muldi3_v8plus): Use output_v8plus_mult.
3665 (*naddsf3, *nadddf3, *nmulsf3, *nmuldf3, *nmuldf3_extend):
3666 New VIS 3.0 combiner patterns.
3667 (fhaddsf_vis, fhadddf_vis, fhsubsf_vis, fhsubdf_vis,
3668 fnhaddsf_vis, fnhaddf_vis, umulxhi_vis, *umulxhi_sp64,
3669 umulxhi_v8plus, xmulx_vis, *xmulx_sp64, xmulx_v8plus,
3670 xmulxhi_vis, *xmulxhi_sp64, xmulxhi_v8plus): New VIS 3.0
3672 * config/sparc/sparc.c (sparc_vis_init_builtins): Emit new builtins.
3673 (output_v8plus_mult): New function.
3674 * config/sparc/sparc-protos.h: Declare it.
3675 * config/sparc/visintrin.h (__vis_fhadds, __vis_fhaddd,
3676 __vis_fhsubs, __vis_fhsubd, __vis_fnhadds, __vis_fnhaddd,
3677 __vis_umulxhi, __vis_xmulx, __vis_xmulxhi): New intrinsics.
3678 * doc/extend.texi: Document new builtins.
3680 2011-10-04 Richard Henderson <rth@redhat.com>
3682 * c-typeck.c (c_build_vec_shuffle_expr): Fix uninitialized variable.
3683 Avoid save_expr unless two_arguments.
3685 2011-10-04 Ozkan Sezer <sezeroz@gmail.com>
3687 * config/i386/mingw-w64.h (CPP_SPEC): Rename _REENTRANCE to _REENTRANT.
3688 * config/i386/mingw32.h (CPP_SPEC): Likewise.
3690 2011-10-04 David S. Miller <davem@davemloft.net>
3692 * config/sparc/driver-sparc.c (cpu_names): Fix string for supersparc
3695 2011-10-04 Jakub Jelinek <jakub@redhat.com>
3697 PR tree-optimization/50604
3698 * builtins.c (fold_builtin_strcpy, fold_builtin_stpcpy,
3699 fold_builtin_strncpy, fold_builtin_stxcpy_chk): Ensure
3700 last argument to memcpy has size_type_node type instead of ssizetype.
3701 * tree-ssa-strlen.c (handle_builtin_memcpy): Use size_type_node
3702 instead of TREE_TYPE (len) as type for newlen.
3704 PR tree-optimization/50522
3705 * tree-ssa-alias.c (ptr_deref_may_alias_decl_p): Don't test
3707 (ptr_derefs_may_alias_p): Call pt_solutions_same_restrict_base
3710 * fold-const.c (fold_unary_loc): Don't optimize
3711 POINTER_PLUS_EXPR casted to TYPE_RESTRICT pointer by
3712 casting the inner pointer if it isn't TYPE_RESTRICT.
3713 * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Don't through
3714 casts from non-TYPE_RESTRICT pointer to TYPE_RESTRICT pointer.
3716 2011-10-04 Joseph Myers <joseph@codesourcery.com>
3718 * config.gcc (i[34567]86-*-elf*, x86_64-*-elf*): Use i386/t-crtstuff.
3720 2011-10-04 Jan Hubicka <jh@suse.cz>
3722 * lto-streamer.h (lto_input_toplevel_asms): Add order_base parameter.
3723 * lto-streamer-in.c (lto_input_toplevel_asms): Stream in order.
3724 * lto-streamer-out.c (lto_output_toplevel_asms): Stream out order.
3725 * lto-cgraph.c (order_base): New static var.
3726 (lto_output_node): Stream in order.
3727 (lto_output_varpool_node): Stream out order.
3728 (input_node): Stream in order.
3729 (input_varpool_node): Stream out order.
3730 (input_cgraph_1): Initialize order base; update call of
3731 lto_input_toplevel_asms.
3733 2011-10-04 Georg-Johann Lay <avr@gjlay.de>
3736 * config/avr/avr-protos.h (avr_legitimize_reload_address): New
3738 * config/avr/avr.h (LEGITIMIZE_RELOAD_ADDRESS): Copy worker code
3740 * config/avr/avr.c (avr_legitimize_reload_address) ...to this new
3741 function. Log if avr_log.legitimize_reload_address.
3743 2011-10-04 Eric Botcazou <ebotcazou@adacore.com>
3745 * config/sparc/sparc.c (sparc_fold_builtin): Use a sequence of tests.
3747 2011-10-04 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
3749 * longlong.h (smul_ppmm, sdiv_qrnnd): Add underscores to the local
3750 variables. Fix second operand of DR. Swap inputs for sdiv_qrnnd.
3752 2011-10-03 David S. Miller <davem@davemloft.net>
3754 * config/sparc/sparc.md (fpack16_vis, fpackfix_vis, fpack32_vis): Make
3755 GSR_REG an input operand to UNSPEC instead of a parallel USE.
3756 (faligndata<V64I:mode>_vis): Likewise and use DI mode.
3757 (alignaddrsi_vis, alignaddrdi_vis, alignaddrlsi_vis, alignaddrldi_vis):
3758 Reference GSR_REG in DI mode, simplify convoluted expressions by using
3760 (bshuffle<V64I:mode>_vis): Reference GSR_REG in DI mode.
3762 2011-10-03 Maxim Kuvyrkov <maxim@codesourcery.com>
3764 * tree-eh.c (remove_unreachable_handlers): Obvious cleanup.
3766 2011-10-03 Jakub Jelinek <jakub@redhat.com>
3767 Ian Lance Taylor <iant@google.com>
3769 * godump.c (go_output_typedef): Support printing enum values that
3770 don't fit in a signed HOST_WIDE_INT.
3772 2011-10-03 Anatoly Sokolov <aesok@post.ru>
3774 * config/cris/cris.c (cris_output_addr_const_extra): Make static.
3775 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
3776 * config/cris/cris.h (OUTPUT_ADDR_CONST_EXTRA): Remove.
3777 * config/cris/cris-protos.h (cris_output_addr_const_extra): Remove.
3779 2011-10-03 Anatoly Sokolov <aesok@post.ru>
3781 * config/m68k/m68k.c (m68k_output_addr_const_extra): Make static.
3782 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
3783 * config/m68k/m68k.h (OUTPUT_ADDR_CONST_EXTRA): Remove.
3784 * config/m68k/m68k-protos.h (m68k_output_addr_const_extra): Remove.
3786 2011-10-03 Steve Ellcey <sje@cup.hp.com>
3789 * configure.ac (gcc_cv_ld_static_dynamic): Define for *-*-hpux*.
3790 (gcc_cv_ld_static_option): Ditto.
3791 (gcc_cv_ld_dynamic_option): Ditto.
3792 * configure: Regenerate.
3794 2011-10-03 David S. Miller <davem@davemloft.net>
3796 * config/sparc/sparc.md (bmask<P:mode>_vis): Split into explicit 'di'
3797 and 'si' patterns which describe the GSR changes explicitly in the
3798 RTL using zero_extract.
3799 (bshuffle<V64I:mode>_vis): Put the GSR use inside of the unspec.
3801 * config/sparc/sparc.md (bshuffle<V64I:mode>_vis): Don't wrap
3802 GSR_REG in a USE, since it's now a true arg to the UNSPEC.
3804 2011-10-03 Artjoms Sinkarovs <artyom.shinkaroff@gmail.com>
3806 * optabs.c (expand_vec_shuffle_expr_p): New function. Checks
3807 if given expression can be expanded by the target.
3808 (expand_vec_shuffle_expr): New function. Expand VEC_SHUFFLE_EXPR
3809 using target vector instructions.
3810 * optabs.h: New optab vshuffle.
3811 (expand_vec_shuffle_expr_p): New prototype.
3812 (expand_vec_shuffle_expr): New prototype.
3813 (vshuffle_optab): New optab.
3814 * genopinit.c: Adjust to support vecshuffle.
3815 * c-tree.h (c_build_vec_shuffle_expr): New prototype.
3816 * expr.c (expand_expr_real_2): Adjust.
3817 * c-typeck.c: (c_build_vec_shuffle_expr): Build a VEC_SHUFFLE_EXPR
3818 recognizing the cases of two and three arguments.
3819 (convert_arguments) (build_binary_op)
3820 (scalar_to_vector) (build_array_ref): Spurious whitespace.
3821 * gimplify.c (gimplify_exp): Adjusted to support VEC_SHUFFLE_EXPR.
3822 * tree.def: New tree code VEC_SHUFFLE_EXPR.
3823 * tree-inline.c (estimate_operator_cost): Recognize VEC_SHUFFLE_EXPR.
3824 * tree-vect-generic.c (vector_element): New function. Returns an
3825 element of the vector at the given position.
3826 (lower_vec_shuffle): Checks if VEC_SHUFLLE_EXPR is supported
3827 by the backend or expand an expression piecewise.
3828 (expand_vector_operations_1): Adjusted.
3829 (gate_expand_vector_operations_noop): New gate function.
3830 * Makefile.in (tree-vect-generic.o): New include.
3831 * gimple.c (get_gimple_rhs_num_ops): Adjust.
3832 * tree-cfg.c (verify_gimple_assign_trenary): Verify VEC_SHUFFLE_EXPR.
3833 * passes.c: Move veclower down.
3834 * tree-pretty-print.c (dump_generic_node): Recognize
3835 VEC_SHUFFLE_EXPR as valid expression.
3836 * c-parser.c (c_parser_get_builtin_args): Helper function for the
3837 builtins with variable number of arguments.
3838 (c_parser_postfix_expression): Use a new helper function for
3839 RID_CHOOSE_EXPR, RID_BUILTIN_COMPLEX and RID_BUILTIN_SHUFFLE.
3840 * tree-ssa-operands: Adjust.
3841 * c-family/c-common.c: New __builtin_shuffle keyword.
3842 * c-family/c-common.h: New __builtin_shuffle keyword.
3843 * gcc/doc/extend.texi: Adjust.
3845 * gcc/config/i386/sse.md: (sseshuffint) New mode_attr. Correspondence
3846 between the vector and the type of the mask when shuffling.
3847 (vecshuffle<mode>): New expansion.
3848 * gcc/config/i386/i386-protos.h (ix86_expand_vshuffle): New prototype.
3849 * gcc/config/i386/i386.c (ix86_expand_vshuffle): New function.
3850 (ix86_vectorize_builtin_vec_perm_ok): Adjust.
3852 2011-10-03 Jakub Jelinek <jakub@redhat.com>
3854 PR tree-optimization/50587
3855 * tree-ssa-reassoc.c (init_range_entry): Stop iterating when
3856 arg0 is not a SSA_NAME.
3858 2011-10-03 Richard Sandiford <rdsandiford@googlemail.com>
3860 * ipa-inline-analysis.c (MAX_TIME): Update comment.
3862 2011-10-02 Richard Henderson <rth@redhat.com>
3863 David S. Miller <davem@davemloft.net>
3865 * config/sparc/sparc.md (ashlsi3, *ashldi3_sp64): Remove
3866 conditional insn type setting, we always emit a shift.
3867 (*ashlsi3_extend, *lshrsi3_extend0): New patterns.
3868 (*lshrsi3_extend): Rename to *lshrsi3_extend1.
3869 * config/sparc/predicates.md (const_one_operand): Delete.
3871 2011-10-02 Gerald Pfeifer <gerald@pfeifer.com>
3873 * invoke.texi (SPARC Options): Refer to GNU/Linux.
3875 2011-10-02 Richard Sandiford <rdsandiford@googlemail.com>
3877 * config/mips/mips.c (mips_frame_barrier): New function.
3878 (mips_expand_prologue): Call it after allocating stack space.
3879 (mips_deallocate_stack): New function.
3880 (mips_expand_epilogue): Call mips_frame_barrier and
3881 mips_deallocate_stack.
3883 2011-10-02 Richard Sandiford <rdsandiford@googlemail.com>
3886 * config/mips/sync.md (sync_<optab>_12): Allow zero operands.
3887 (sync_old_<optab>_12, sync_new_<optab>_12, sync_nand_12): Likewise.
3888 (sync_old_nand_12, sync_new_nand_12, test_and_set_12): Likewise.
3890 2011-10-02 Jan Hubicka <jh@suse.cz>
3892 * cgraphunit.c (verify_edge_count_and_frequency): Bounds check.
3894 * cgraphunit.c (ipa_passes): Remove unrechable nodes.
3895 * lto-streamer-out.c (produce_symtab): Skip unused extern declarations.
3896 * ipa.c (cgraph_remove_unreachable_nodes): Do not assume that external
3897 functions are reachable when address is taken.
3898 * tree-sra.c (modify_function): Free dominance info.
3900 2011-10-02 Jan Hubicka <jh@suse.cz>
3902 * ipa-inline-analysis.c (inline_summary_alloc): Bounds check.
3904 * ipa-inline-analysis.c (reset_inline_edge_summary): New function.
3905 (reset_inline_summary): New function.
3906 (compute_inline_parameters, inline_node_removal_hook,
3907 inline_edge_removal_hook): Use it.
3908 (inline_free_summary): Reset holders correctly.
3909 (inline_generate_summary): Free summary before computing it.
3911 2011-10-02 Paolo Carlini <paolo.carlini@oracle.com>
3913 PR preprocessor/36819
3914 * incpath.c (merge_include_chains): Call free_path on
3915 heads[QUOTE] and tails[QUOTE].
3917 2011-10-02 Jan Hubicka <jh@suse.cz>
3920 * lto-symtab.c (lto_symtab_resolve_symbols): Do not resolve
3921 when resolution is already availbale from plugin.
3922 (lto_symtab_merge_decls_1): Handle LDPR_PREVAILING_DEF_IRONLY_EXP.
3923 * cgraph.c (ld_plugin_symbol_resolution): Add
3924 prevailing_def_ironly_exp.
3925 * lto-cgraph.c (LDPR_NUM_KNOWN): Update.
3926 * ipa.c (varpool_externally_visible_p): IRONLY variables are never
3928 * varasm.c (resolution_to_local_definition_p): Add
3929 LDPR_PREVAILING_DEF_IRONLY_EXP.
3930 (resolution_local_p): Likewise.
3932 2011-10-01 David S. Miller <davem@davemloft.net>
3934 * config/sparc/sparc.opt (VIS3): New option.
3935 * doc/invoke.texi: Document it.
3936 * config/sparc/sparc.h: Force TARGET_VIS3 to zero if assembler is
3937 not capable of such instructions.
3938 * config/sparc/sparc-c.c (sparc_target_macros): Define __VIS__
3939 to 0x300 when TARGET_VIS3.
3940 * config/sparc/sparc-modes.def: Create 16-byte vector modes.
3941 * config/sparc/sparc.md (UNSPEC_CMASK8, UNSPEC_CMASK16, UNSPEC_CMASK32,
3942 UNSPEC_FCHKSM16, UNSPEC_PDISTN, UNSPC_FUCMP): New unspecs.
3943 (V64N8, VASS): New mode iterators.
3944 (vis3_shift, vis3_addsub_ss): New code iterators.
3945 (vbits, vconstr): New mode attributes.
3946 (vis3_shift_insn, vis3_addsub_ss_insn): New code attributes.
3947 (cmask8<P:mode>_vis, cmask16<P:mode>_vis, cmask32<P:mode>_vis,
3948 fchksm16_vis, <vis3_shift_insn><vbits>_vis, pdistn<mode>_vis,
3949 fmean16_vis, fpadd64_vis, fpsub64_vis,
3950 <vis3_addsub_ss_insn><vbits>_vis, fucmp<code>8<P:mode>_vis): New
3951 VIS 3.0 instruction patterns.
3952 * config/sparc/sparc.c (sparc_option_override): Set MASK_VIS3 by
3953 default when targetting capable cpus. TARGET_VIS3 implies
3954 TARGET_VIS2 and TARGET_VIS, and clear them when TARGET_FPU is disabled.
3955 (sparc_vis_init_builtins): Emit new VIS 3.0 builtins.
3956 (sparc_fold_builtin): Do not eliminate cmask{8,16,32} when result
3958 * config/sparc/visintrin.h (__vis_cmask8, __vis_cmask16,
3959 __vis_cmask32, __vis_fchksm16, __vis_fsll16, __vis_fslas16,
3960 __vis_fsrl16, __vis_fsra16, __vis_fsll32, __vis_fslas32,
3961 __vis_fsrl32, __vis_fsra32, __vis_pdistn, __vis_fmean16,
3962 __vis_fpadd64, __vis_fpsub64, __vis_fpadds16, __vis_fpadds16s,
3963 __vis_fpsubs16, __vis_fpsubs16s, __vis_fpadds32, __