1 2011-12-02 Richard Guenther <rguenther@suse.de>
4 * ipa.c (varpool_externally_visible_p): Register variables
5 are always externally visible.
7 2011-12-02 Sameera Deshpande <sameera.deshpande@arm.com>
9 * gimple-pretty-print.c (dump_gimple_call): Dereference fn only if
12 2011-12-02 Martin Jambor <mjambor@suse.cz>
14 PR tree-optimization/50622
15 * tree-sra.c (load_assign_lhs_subreplacements): Force gimple operand
16 if both lacc and racc are grp_partial_lhs.
18 2011-12-01 Kaz Kojima <kkojima@gcc.gnu.org>
21 * config/sh/sh.c (expand_ashiftrt): Handle TARGET_SH2A same as
23 (shl_sext_kind): Likewise.
24 * config/sh/sh.h (SH_DYNAMIC_SHIFT_COST): Likewise.
25 * config/sh/sh.md (ashlsi3_sh2a, ashrsi3_sh2a, lshrsi3_sh2a):
27 (ashlsi3_std): Handle TARGET_SH2A same as TARGET_SH3.
29 (ashrsi3_d): Likewise.
30 (lshrsi3_d): Likewise.
33 2011-12-01 Diego Novillo <dnovillo@google.com>
36 * ipa-inline.c (can_inline_edge_p): If the edge E has a
37 statement, use the statement's inline indicator instead
39 Remove consistency check.
41 2011-12-01 Diego Novillo <dnovillo@google.com>
46 2011-11-29 Diego Novillo <dnovillo@google.com>
48 * gimple.c (gimple_call_set_cannot_inline): Move from gimple.h.
49 Update field call_stmt_cannot_inline_p from call
50 graph edge, if needed.
51 * gimple.h (gimple_call_set_cannot_inline): Move to gimple.c.
53 2011-12-01 Nathan Sidwell <nathan@acm.org>
56 * coverage.c (build_var): Keep coverage variables static.
58 2011-12-01 Jakub Jelinek <jakub@redhat.com>
60 PR tree-optimization/51356
61 * tree-vect-patterns.c (vect_recog_bool_pattern): Give up if
62 vectype doesn't have VECTOR_MODE_P.
65 * tree-ssa-dce.c (remove_dead_stmt): Add a debug stmt when removing
66 as unnecessary a store to a variable with gimple reg type.
67 * tree-ssa-live.c (remove_unused_locals): Clear TREE_ADDRESSABLE bit
68 on local unreferenced variables.
69 * cfgexpand.c (expand_gimple_basic_block): Don't emit DEBUG_INSNs
70 for !target_for_debug_bind variables.
72 2011-12-01 Patrick Marlier <patrick.marlier@gmail.com>
75 * cgraph.h (cgraph_call_node_duplication_hooks): Declare.
76 * cgraph.c (cgraph_call_node_duplication_hooks): Make global.
77 * cgraphunit.c (cgraph_copy_node_for_versioning): Call it.
79 2011-12-01 Andrew Pinski <apinski@cavium.com>
82 * tree.c (free_lang_data_in_decl): Clear FIELD_DECL's
85 2011-12-01 Jakub Jelinek <jakub@redhat.com>
87 PR tree-optimization/51246
88 * tree-predcom.c (replace_ref_with): Handle also clobber on the
91 PR rtl-optimization/51014
92 * loop-unroll.c (apply_opt_in_copies): Ignore label DEBUG_INSNs
93 both from bb and orig_bb.
95 2011-12-01 Joern Rennecke <joern.rennecke@embecosm.com>
97 PR tree-optimization/50802
98 * tree-vrp.c (simplify_conversion_using_ranges): Rewrite test
99 considering what happens to ranges during sign changes and/or
100 intermediate narrowing conversions.
102 2011-11-30 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
105 * config/pa/pa.md (in_branch_delay): Disallow frame related insns.
106 (in_nullified_branch_delay): Likewise.
107 (in_call_delay): Likewise.
109 2011-11-30 Richard Henderson <rth@redhat.com>
111 * predict.c (expr_expected_value_1): Assume compare-and-swap builtin
112 boolean return is true.
114 2011-11-30 Jakub Jelinek <jakub@redhat.com>
116 PR rtl-optimization/51044
117 * dse.c (emit_inc_dec_insn_before): Ensure new_insn is an instruction
118 rather than just insn body.
121 * function.c (thread_prologue_and_epilogue_insns): If
122 stack_realign_drap, add drap_reg to SET_UP_BY_PROLOGUE
126 * tree-eh.c (cleanup_empty_eh_merge_phis): Add check to
127 avoid creating duplicate edges here.
128 (cleanup_empty_eh_unsplit): And remove it in the caller.
130 2011-11-30 Andrew Pinski <apinski@cavium.com>
133 * c-parser.c (c_parser_postfix_expression): Check groktypename results
134 before looking at the main variant.
136 2011-11-30 Jakub Jelinek <jakub@redhat.com>
138 PR rtl-optimization/48721
139 * sched-deps.c (sched_analyze_insn): For SIBLING_CALL_P set
140 reg_pending_barrier to TRUE_BARRIER.
142 PR rtl-optimization/48721
143 * sched-ebb.c (begin_move_insn): Insert empty unreachable
144 block after BARRIER if insn is followed by it.
146 2011-11-30 Richard Henderson <rth@redhat.com>
148 * config/sparc/sparc-opts.h (enum sparc_memory_model_type): New.
149 * config/sparc/sparc.opt (mmemory-model=): New option.
150 * doc/invoke.texi (Sparc Options): Document it.
151 * config/sparc/sparc.c (sparc_option_override): Provide default
152 for sparc_memory_model.
153 (sparc_emit_membar_for_model): Omit barrier combinations that are
154 implied by the memory model.
156 2011-11-30 Richard Henderson <rth@redhat.com>
158 * config/sparc/sync.md (I124MODE): New mode iterator.
159 (memory_barrier): Remove.
160 (atomic_exchangesi): New.
161 (swapsi): Rename from *swapsi.
162 (atomic_test_and_set<I124MODE>): New.
163 (ldstubqi): Rename from *ldstubqi, merge with expander.
164 (ldstub<I24MODE>): Rename from *ldstub<I24MODE>, merge with expander.
165 (sync_lock_test_and_setsi): Remove.
166 (sync_lock_test_and_set<I12MODE>): Remove.
168 2011-11-30 Richard Henderson <rth@redhat.com>
170 * config/sparc/constraints.md ("w"): New.
171 * config/sparc/predicates.md (mem_noofs_operand): New.
172 * config/sparc/sparc.c (sparc_expand_compare_and_swap_12): Make static.
173 Generate the boolean result of the atomic_compare_exchange.
174 (sparc_expand_compare_and_swap): New.
175 * config/sparc/sync.md (atomic_compare_and_swap<I>): New.
176 (atomic_compare_and_swap<I48MODE>_1): Rename from
177 *sync_compare_and_swap<I48MODE>, use mem_noofs_operand.
178 (*atomic_compare_and_swapdi_v8plus): Rename from
179 *sync_compare_and_swapdi_v8plus, use mem_noofs_operand.
180 (sync_compare_and_swap<I12MODE>): Remove.
181 (sync_compare_and_swap<I48MODE>): Remove.
183 2011-11-30 Richard Henderson <rth@redhat.com>
185 * config/sparc/predicates.md (register_or_v9_zero_operand): New.
186 * config/sparc/sparc.md (UNSPEC_ATOMIC): New.
187 * config/sparc/sync.md (atomic_load<I>): New.
188 (atomic_loaddi_1, atomic_store<I>, atomic_storedi_1): New.
190 2011-11-30 Richard Henderson <rth@redhat.com>
192 * config/sparc/predicates.md (zero_or_v7_operand): New.
193 * config/sparc/sparc.c (sparc_emit_membar_for_model): New.
194 * config/sparc/sparc-protos.h: Update.
195 * config/sparc/sync.md (mem_thread_fence): New.
196 (memory_barrier): Use sparc_emit_membar_for_model.
197 (membar, *membar_empty, *membar_storestore, *membar_storeload): New.
198 (*membar_v8): Accept and ignore the membar mask.
199 (*membar): Accept and print the membar mask.
201 2011-11-30 Matthew Gretton-Dann <matthew.gretton-dann@arm.com>
203 * config/arm/arm.c (arm_issue_rate): Cortex-A15 can triple issue.
204 * config/arm/arm.md (mul64): New attribute.
205 (generic_sched): Cortex-A15 is not scheduled generically.
206 (cortex-a15.md): Include.
207 * config/arm/cortex-a15.md: New machine description.
208 * config/arm/t-arm (MD_INCLUDES): Add cortex-a15.md.
210 2011-11-30 Matthew Gretton-Dann <matthew.gretton-dann@arm.com>
212 * config/arm/t-arm (MD_INCLUDES): Ensure all md files are listed.
214 2011-11-30 Iain Sandoe <iains@gcc.gnu.org>
216 * doc/tm.texi.in (NEXT_OBJC_RUNTIME): Update macro description.
217 * doc/tm.texi: Regenerate.
218 * flags.h (flag_next_runtime): Remove references.
219 * toplev.c: Likewise.
220 * defaults.h (NEXT_OBJC_RUNTIME): Provide default.
221 * gccspec.c (lang_specific_driver): Adjust use of NEXT_OBJC_RUNTIME.
222 * config/darwin.c (darwin_override_options): Provide default
223 Objective-C abi settings and target conflict checks.
224 * config/darwin.h (SUBTARGET_C_COMMON_OVERRIDE_OPTIONS): Remove unused
225 code. Adjust indenting.
227 2011-11-29 Joseph Myers <joseph@codesourcery.com>
229 * config/sh/sh.h (enum reg_class, REG_CLASS_NAMES)
230 (REG_CLASS_CONTENTS): Add NON_SP_REGS.
231 (REGCLASS_HAS_GENERAL_REG): Handle NON_SP_REGS.
232 * config/sh/constraints.md (u): New constraint.
234 2011-11-29 Oleg Endo <oleg.endo@t-online.de>
237 * config/sh/sh.c (sh_secondary_reload): Add case when FPUL
238 register is being loaded from a pseudo in memory.
240 2011-11-29 DJ Delorie <dj@redhat.com>
242 * config.gcc (rl78-*-elf): New case.
243 * doc/extend.texi: Add RL78 documentation.
244 * doc/invoke.texi: Likewise.
245 * doc/md.texi: Likewise.
246 * doc/contrib.texi: Add RL78.
247 * doc/install.texi: Add rl78-*-elf.
248 * config/rl78: New directory for the Renesas RL78.
250 2011-11-29 Jakub Jelinek <jakub@redhat.com>
252 PR tree-optimization/51247
253 * tree-vrp.c (extract_range_from_assert): For signed 1-bit precision
254 types instead of adding 1 subtract -1 and instead of subtracting 1
255 add -1 to avoid overflows.
257 2011-11-29 Andrew MacLeod <amacleod@redhat.com>
260 * optabs.c (maybe_optimize_fetch_op): New. Look for more optimal
261 instructions for a FECTH_OP or OP_FECTH sequence.
262 (expand_atomic_fetch_op): Call maybe_optimize_fetch_op.
264 2011-11-29 Uros Bizjak <ubizjak@gmail.com>
266 * config/i386/sync.md (UNSPEC_LDA, UNSPEC_STA): New unspecs.
267 (movdi_via_fpu): Remove.
268 (loaddi_via_fpu): New insn pattern.
269 (storedi_via_fpu): Ditto.
270 (atomic_loaddi_fpu): Use loaddi_via_fpu and storedi_via_fpu.
271 (atomic_storedi_fpu): Ditto.
272 * reg-stack.c (get_true_reg): Handle UNSPEC_LDA.
273 (subst_stack_regs_pat): Handle UNSPEC_STA.
275 2011-11-29 Uros Bizjak <ubizjak@gmail.com>
277 * config/i386/i386.md (*floathi<mode>2_i387_with_temp): Do not
278 allocate scratch memory for alternative 0.
280 2011-11-29 Michael Meissner <meissner@linux.vnet.ibm.com>
282 * config/rs6000/rs6000-builtins.def: Completely rewrite builtin
283 handling to centralize more of the builtins in this file. Change
284 some builtin enumerations to be more consistant. Use a new mask
285 to hold the current builtins, including SPE and PAIRED builtins
286 which no longer are set via target_flags masks. Add
287 -mdebug=builtin debug support. For power machines, define all
288 Altivec and VSX buitins when the compiler starts, but don't allow
289 the use of a builtin unless the appropriate switch is used, or
290 #pragma GCC target is used to change the options. If the user
291 uses #pragma GCC target, update the appropriate hardware macros.
292 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Ditto.
293 * config/rs6000/rs6000.opt (rs6000_builtin_mask): Ditto.
294 * config/rs6000/rs6000-c.c (rs6000_macro_to_expand): Ditto.
295 (rs6000_define_or_undefine_macro): Ditto.
296 (rs6000_target_modify_macros): Ditto.
297 (rs6000_cpu_cpp_builtins): Ditto.
298 (altivec_overloaded_builtins): Ditto.
299 (altivec_build_resolved_builtin): Ditto.
300 * config/rs6000/rs6000.c (rs6000_target_modify_macros_ptr): Ditto.
301 (rs6000_builtin_info): Ditto.
302 (def_builtin): Ditto.
303 (enable_mask_for_builtins): Ditto.
304 (DEBUG_FMT_X): Ditto.
305 (rs6000_debug_reg_global): Ditto.
306 (rs6000_builtin_mask_calculate): Ditto.
307 (rs6000_option_override_internal): Ditto.
308 (rs6000_builtin_conversion): Ditto.
309 (rs6000_builtin_vectorized_function): Ditto.
313 (builtin_description_predicates): Ditto.
314 (bdesc_altivec_preds): Ditto.
315 (bdesc_spe_predicates): Ditto.
316 (bdesc_spe_evsel): Ditto.
317 (bdesc_paired_preds): Ditto.
320 (rs6000_overloaded_builtin_p): Ditto.
321 (rs6000_expand_unop_builtin): Ditto.
322 (bdesc_2arg_spe): Ditto.
323 (spe_expand_builtin): Ditto.
324 (rs6000_invalid_builtin): Ditto.
325 (rs6000_expand_builtin): Ditto.
326 (rs6000_init_builtins): Ditto.
327 (spe_init_builtins): Ditto.
328 (paired_init_builtins): Ditto.
329 (altivec_init_builtins): Ditto.
330 (builtin_function_type): Ditto.
331 (rs6000_common_init_builtins): Ditto.
332 (rs6000_builtin_reciprocal): Ditto.
333 (rs6000_builtin_mask_names): Ditto.
334 (rs6000_pragma_target_parse): Ditto.
335 (rs6000_function_specific_print): Ditto.
336 * config/rs6000/rs6000.h (MASK_DEBUG_BUILTIN): Ditto.
337 (MASK_DEBUG_ALL): Ditto.
338 (TARGET_DEBUG_BUILTIN): Ditto.
339 (TARGET_EXTRA_BUILTINS): Ditto.
340 (REGISTER_TARGET_PRAGMAS): Ditto.
341 (enum rs6000_btc): Ditto.
342 (RS6000_BTC_*): Ditto.
343 (RS6000_BTM_*): Ditto.
344 (enum rs6000_builtins): Ditto.
345 * config/rs6000/rs6000-protos.h (rs6000_overloaded_builtin_p): Ditto.
346 (rs6000_target_modify_macros): Ditto.
347 (rs6000_target_modify_macros_ptr): Ditto.
349 * config/rs6000/vector.md (unsfloat<VEC_int><mode>2): Use the
350 standard name for converting vector unsigned values to floating
351 point. Delete the TARGET_VECTORIZE_BUILTIN_CONVERSION hook, which
352 is handled by machine independent code.
353 * config/rs6000/rs6000.c (TARGET_VECTORIZE_BUILTIN_CONVERSION): Ditto.
354 (rs6000_builtin_conversion): Ditto.
356 2011-11-29 Dodji Seketeli <dodji@redhat.com>
358 * tree-diagnostic (struct loc_t): Rename into struct loc_map_pair.
359 (maybe_unwind_expanded_macro_loc): Adjust.
361 2011-11-29 David Edelsohn <dje.gcc@gmail.com>
362 Aldy Hernandez <aldyh@redhat.com>
364 * trans-mem.c (ipa_tm_create_version_alias): Mangle new_decl
366 (ipa_tm_create_version): Same.
368 2011-11-29 Bernd Schmidt <bernds@codesourcery.com>
370 * haifa-sched.c (recompute_todo_spec): Simplify and correct the
371 code checking for a clobber of a condition register when deciding
372 whether to predicate.
374 2011-11-29 Diego Novillo <dnovillo@google.com>
376 * gimple.c (gimple_call_set_cannot_inline): Move from gimple.h.
377 Update field call_stmt_cannot_inline_p from call graph edge, if needed.
378 * gimple.h (gimple_call_set_cannot_inline): Move to gimple.c.
380 2011-11-29 Sameera Deshpande <sameera.deshpande@arm.com>
382 * config/arm/arm.md (arm_movdi): Update NEG_POOL_RANGE.
383 (movdf_soft_insn): Likewise.
384 * config/arm/fpa.md (thumb2_movdf_fpa): Likewise.
385 * config/arm/neon.md (neon_mov<mode>): Likewise.
386 * config/arm/vfp.md (movdi_vfp): Likewise.
387 (movdi_vfp_cortexa8): Likewise.
388 (movdf_vfp): Likewise.
390 2011-11-29 Jakub Jelinek <jakub@redhat.com>
393 * expr.c (expand_expr_addr_expr_1): Don't call force_operand for
396 2011-11-29 Ira Rosen <ira.rosen@linaro.org>
398 PR tree-optimization/51301
399 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Check that
400 the last statement doesn't convert to a bigger type than the original
401 type of the computation.
403 2011-11-28 Richard Henderson <rth@redhat.com>
405 * config/rs6000/rs6000.c (rs6000_expand_atomic_compare_and_swap):
406 Handle overlap between retval and oldval. Always compute boolval
409 2011-11-28 Jakub Jelinek <jakub@redhat.com>
412 * tree-ssa.c (execute_update_addresses_taken): Remove
413 var ={v} {CLOBBER} stmts instead of rewriting them into
414 var_N ={v} {CLOBBER}.
417 * function.c (convert_jumps_to_returns): When redirecting an edge
418 succ to EXIT_BLOCK_PTR, clear EDGE_CROSSING flag.
420 PR tree-optimization/50078
421 * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Copy over
422 TREE_THIS_VOLATILE also from the old to new lhs resp. rhs.
424 PR tree-optimization/50682
425 * tree-eh.c (maybe_remove_unreachable_handlers): New function.
426 * tree-flow.h (maybe_remove_unreachable_handlers): New prototype.
427 * tree-optimize.c (execute_cleanup_cfg_post_optimizing): Call it.
429 2011-11-28 Georg-Johann Lay <avr@gjlay.de>
431 * doc/extend.texi (AVR Built-in Functions): Add documentation for
432 __builtin_avr_map8 and __builtin_avr_map16.
434 * config/avr/avr.md: Document new %t and %T asm output codes.
435 (define_c_enum "unspec"): Add UNSPEC_MAP_BITS.
436 (adjust_len): Add map_bits.
437 (map_bitsqi, map_bitshi): New insns.
438 * config/avr/avr-protos.h (avr_out_map_bits): New.
439 * config/avr/avr-protos.c (print_operand): Implement %t and %T.
440 (adjust_insn_length): Handle ADJUST_LEN_MAP_BITS.
441 (avr_double_int_push_digit): New function.
442 (avr_map, avr_revert_map, avr_swap_map, avr_id_map): New functions.
443 (avr_sig_map, avr_map_hamming_byte): New functions.
444 (avr_out_swap_bits, avr_out_revert_bits, avr_move_bits,
445 avr_out_map_bits): New functions.
446 (enum avr_builtin_id): Add AVR_BUILTIN_MAP8, AVR_BUILTIN_MAP16.
447 (avr_init_builtins): Populate __builtin_avr_map8, __builtin_avr_map16.
448 (bdesc_2arg): Add __builtin_avr_map8, __builtin_avr_map16 ...
449 (avr_expand_builtin): ...and expand them.
450 * config/avr/avr-c.c (avr_cpu_cpp_builtins): New built-in defines:
451 __BUILTIN_AVR_MAP8, __BUILTIN_AVR_MAP16.
453 2011-11-27 Richard Sandiford <rdsandiford@googlemail.com>
456 * config/mips/mips.c (mips_rtx_costs): Handle baddu.
458 2011-11-27 Richard Sandiford <rdsandiford@googlemail.com>
460 * config/mips/mips.c (mips16_unextended_reference_p): Test for BLKmode.
462 2011-11-27 Richard Sandiford <rdsandiford@googlemail.com>
464 * config/mips/mips-protos.h (mips_emit_binary): Declare.
465 * config/mips/mips.c (mips_emit_binary): Make global.
466 (mips_set_mips16_mode): Turn off -mfix-r4000 in MIPS16 mode.
467 (mips_conditional_register_usage): Don't treat LO and HI as
468 register operands in MIPS16 mode.
469 (mips_mulsidi3_gen_fn): Use {u,}mulsidi3_{32,64}bit_mips16
471 * config/mips/predicates.md (muldiv_target_operand): New predicate.
472 (move_operand): Allow hilo_operand.
473 * config/mips/mips.md (mul<mode>3): Explicitly specify LO as the
474 target of MIPS16 multiplies, then move it into the target register.
475 (mul<mode>3_internal, *macc2, *msac2): Use muldiv_target_operand.
476 (<u>mulsidi3_32bit_mips16): New expander.
477 (<u>mulsidi3_32bit): Use muldiv_target_operand.
478 (<u>mulsidi3_32bit_r4000): Disable for ISA_HAS_DSP.
479 (<u>mulsidi3_64bit): Require !TARGET_MIPS16. Split into
480 <u>mulsidi3_64bit_split.
481 (<u>mulsidi3_64bit_mips16): New expander.
482 (<u>mulsidi3_64bit_split): Likewise, using expansions from
483 two previous define_splits.
484 (<u>mulsidi3_64bit_hilo, *muls<u>_di, <u>msubsidi4): Use
485 muldiv_target_operand.
486 (<su>mulsi3_highpart): Use <su>mulsi3_highpart_split for MIPS16 code.
487 (<su>mulsi3_highpart_internal): Require !TARGET_MIPS16.
488 Split into <su>mulsi3_highpart_split.
489 (<su>mulsi3_highpart_split): New expander.
490 (<su>muldi3_highpart): Turn into a define_expand.
491 Use <su>muldi3_highpart_split for MIPS16 code.
492 (<su>muldi3_highpart_internal): Renamed from <su>muldi3_highpart.
493 Require !TARGET_MIPS16. Split into <su>muldi3_highpart_split.
494 (<su>muldi3_highpart_split): New expander.
495 (<u>mulditi3): Explicitly specify LO as the target of MIPS16
496 multiplies, then move it into the target register.
497 (<u>mulditi3_internal, <u>maddsidi4): Use muldiv_target_operand.
498 (divmod<mode>4, udivmod<mode>4): Turn into define_expands.
499 Use <u>divmod<mode>4_split for MIPS16 code, then explicitly
500 move LO into operand 0.
501 (divmod<mode>4_internal, udivmod<mode>4_internal): Renamed
502 from <u>divmod<mode>4. Use muldiv_target_operand.
503 Require !TARGET_MIPS16. Split into <u>divmod<mode>4_split.
504 (<u>divmod<mode>4_split): New expander.
505 (<u>divmod<GPR:mode>4_hilo_<HILO:mode>): Use muldiv_target_operand.
506 (mfhi<GPR:mode>_<HILO:mode>): Use hilo_operand.
508 2011-11-27 Richard Sandiford <rdsandiford@googlemail.com>
510 * hard-reg-set.h (target_hard_regs): Add x_accessible_reg_set
511 and x_operand_reg_set.
512 (accessible_reg_set, operand_reg_set): New macros.
513 * reginfo.c (init_reg_sets): Initialize accessible_reg_set and
515 (saved_accessible_reg_set, saved_operand_reg_set): New variables.
516 (save_register_info): Save them.
517 (restore_register_info): Restore them.
518 (init_reg_sets_1): Limit operand_reg_set to accessible_reg_set.
519 Remove NO_REGS registers from operand_reg_set. Treat members
520 of operand_reg_set as fixed.
521 * recog.c (general_operand): Check operand_reg_set rather than NO_REGS.
522 (register_operand, nonmemory_operand): Likewise.
523 * varasm.c (make_decl_rtl): Always use DECL_MODE as the mode of
524 register variables. Check accessible_reg_set and operand_reg_set.
525 * config/mips/mips.c (mips_conditional_register_usage): Remove
526 inaccessible register from accessible_reg_set, rather than just
529 2011-11-27 Gerald Pfeifer <gerald@pfeifer.com>
531 * config/pa/pa-linux.h (TARGET_GAS): Remove comment.
533 2011-11-26 Richard Henderson <rth@redhat.com>
535 * config/arm/arm.c (arm_init_libfuncs): Call init_sync_libfuncs
538 2011-11-26 Richard Henderson <rth@redhat.com>
540 * optabs.c (expand_atomic_fetch_op): Always return result.
542 * optabs.c (init_sync_libfuncs_1): Include max in iteration.
544 * omp-low.c (expand_omp_atomic): Assume anything aligned to
545 BIGGEST_ALIGNMENT is aligned.
547 2011-11-26 Richard Henderson <rth@redhat.com>
549 * config/m68k/m68k.md (UNSPECV_CAS_1, UNSPECV_CAS_2): New.
550 (UNSPECV_TAS_1, UNSPECV_TAS_2): New.
551 (I): New mode iterator.
552 (xz): New mode attribute.
553 * config/m68k/sync.md: New file.
555 * config/m68k/linux.h (TARGET_INIT_LIBFUNCS): New.
556 * config/m68k/m68k.c (m68k_init_sync_libfuncs): New.
558 2011-11-26 Nathan Sidwell <nathan@acm.org>
560 PR gcov-profile/51297
561 * gcov.c (main): Allocate initial names and sources arrays.
562 (find_source): Don't check for null name or source arrays here.
564 2011-11-26 Jakub Jelinek <jakub@redhat.com>
566 PR rtl-optimization/49912
567 * cfgrtl.c (rtl_verify_flow_info_1): Ignore also EDGE_PRESERVE bit
568 when counting n_branch.
570 2011-11-26 Uros Bizjak <ubizjak@gmail.com>
572 * config/i386/sync.md (movdi_via_fpu): Add %Z insn suffixes.
574 2011-11-26 Joern Rennecke <joern.rennecke@embecosm.com>
577 * calls.c (mem_overlaps_already_clobbered_arg_p):
578 Return false if no outgoing arguments have been stored so far.
580 2011-11-25 Georg-Johann Lay <avr@gjlay.de>
582 * config/avr/avr.c (print_operand): Support code = 'i' for CONST_INT.
584 2011-11-24 Andrew MacLeod <amacleod@redhat.com>
587 * tree.h (is_builtin_name): No longer external.
588 * builtins.c (is_builtin_name): Make static.
589 * varasm.c (incorporeal_function_p): __sync and __atomic external calls
590 are not incorporeal and may need asm label processing.
592 2011-11-25 Georg-Johann Lay <avr@gjlay.de>
595 * config/avr/avr-protos.h (avr_log_t): Add field .builtin.
596 * config/avr/avr-log.c (avr_log_set_avr_log): Initialize it.
597 Don't bypass TARGET_ALL_DEBUG. Print self-info with ?.
598 (avr_log_vadump): Support %D and %X to print double_int.
599 (avr_double_int_pop_digit): New static function.
600 (avr_dump_double_int_hex): New static function.
602 2011-11-24 Enkovich Ilya <ilya.enkovich@intel.com>
605 * config/i386/i386.c (distance_non_agu_define_in_bb): Fix insn
608 2011-11-24 Andrew MacLeod <amacleod@redhat.com>
610 * optab.c (maybe_emit_atomic_exchange): New. Try to emit an
611 atomic_exchange pattern.
612 (maybe_emit_sync_lock_test_and_set): New. Try to emit an exchange
613 using __sync_lock_test_and_set.
614 (maybe_emit_compare_and_swap_exchange_loop): New. Try to emit an
615 exchange using a compare_and_swap loop.
616 (expand_sync_lock_test_and_set): New. Expand sync_lock_test_and_set.
617 (expand_atomic_test_and_set): New. Expand test_and_set operation.
618 (expand_atomic_exchange): Use new maybe_emit_* functions.
619 (expand_atomic_store): Use new maybe_emit_* functions.
620 * builtins.c (expand_builtin_sync_lock_test_and_set): Call
621 expand_sync_lock_test_and_set routine.
622 (expand_builtin_atomic_exchange): Remove parameter from call.
623 (expand_builtin_atomic_clear): Use atomic_clear pattern if present.
624 (expand_builtin_atomic_test_and_set): Add target and simply call
625 expand_atomic_test_and_set.
626 (expand_builtin): Add target to expand_builtin_atomic_test_and_set.
627 * expr.h (expand_atomic_exchange): Add parameter.
628 (expand_sync_lock_test_and_set): New prototype.
629 (expand_atomic_test_and_set, expand_atomic_clear): New prototypes.
631 2011-11-24 H.J. Lu <hongjiu.lu@intel.com>
634 * config/i386/i386.h (processor_costs): Revert revision 181357.
635 * config/i386/i386.c (cost models): Likewise.
636 (core_cost): Likewise.
637 (promote_duplicated_reg): Likewise.
638 (promote_duplicated_reg_to_size): Likewise.
639 (processor_target): Likewise.
640 (expand_set_or_movmem_via_loop_with_iter): Likewise.
641 (expand_set_or_movmem_via_loop): Likewise.
642 (emit_strset): Likewise.
643 (expand_movmem_epilogue): Likewise.
644 (expand_setmem_epilogue): Likewise.
645 (expand_movmem_prologue): Likewise.
646 (expand_setmem_prologue): Likewise.
647 (expand_constant_movmem_prologue): Likewise.
648 (expand_constant_setmem_prologue): Likewise.
649 (decide_alg): Likewise.
650 (decide_alignment): Likewise.
651 (ix86_expand_movmem): Likewise.
652 (ix86_expand_setmem): Likewise.
653 (ix86_slow_unaligned_access): Likewise.
654 * config/i386/i386.md (strset): Likewise.
655 * config/i386/sse.md (vec_dupv4si): Likewise.
656 (vec_dupv2di): Likewise.
658 2011-11-24 Georg-Johann Lay <avr@gjlay.de>
660 * config/avr/avr.md (*rotlhi2.15): Set length to 4.
662 2011-11-24 Razya Ladelsky <razya@il.ibm.com>
664 * tree-data-ref.c (initialize_data_dependence_relation): Update
665 comment for the self dependence case.
666 (compute_self_dependence): Remove.
667 * tree-vect-data-refs.c (vect_analyze_data_refs): Remove call to
668 compute_self_dependenc.
670 2011-11-24 Alan Modra <amodra@gmail.com>
672 * c-decl.c (warn_if_shadowing): Don't warn if shadowed
673 identifier is from system header.
675 2011-11-23 Jakub Jelinek <jakub@redhat.com>
678 * config/i386/i386.c (decide_alg): Initialize *dynamic_check
681 2011-11-23 Vladimir Makarov <vmakarov@redhat.com>
683 PR rtl-optimization/48455
684 * doc/invoke.texi (-fira-region): Document default values.
686 * flags-types.h (enum ira_region): Add new value
687 IRA_REGION_AUTODETECT.
689 * common.opt (fira-region): Set up initial value to
690 IRA_REGION_AUTODETECT.
692 * toplev.c (process_options): Set up flag_ira_region depending on
695 * ira.c (ira.c): Remove optimize guard for ira_build.
697 2011-11-23 Chung-Lin Tang <cltang@codesourcery.com>
699 PR rtl-optimization/50496
700 * jump.c (redirect_jump): Assert fail on nlabel == NULL_RTX
701 only after epilogue is created. Add comments.
703 2011-11-22 Richard Henderson <rth@redhat.com>
705 * config/ia64/ia64.c (ia64_expand_atomic_op): Add model parameter.
706 Generate the barrier required for the memory model.
707 (rtx_needs_barrier): Handle UNSPEC_FETCHADD_REL, UNSPEC_CMPXCHG_REL.
708 * config/ia64/ia64-protos.h: Update.
709 * config/ia64/ia64.md (UNSPEC_FETCHADD_REL): New.
710 (UNSPEC_CMPXCHG_REL): New.
711 * config/ia64/sync.md (mem_thread_fence): New.
712 (atomic_load<IMODE>, atomic_store<IMODE>): New.
713 (atomic_compare_and_swap<IMODE>): New.
714 (cmpxchg_acq_<I124MODE>, cmpxchg_acq_di): New.
715 (atomic_exchange<IMODE>): New.
716 (xchg_acq_<IMODE>): Rename from sync_lock_test_and_set<IMODE>.
717 (atomic_<FETCHOP><IMODE>, atomic_nand<IMODE>): New.
718 (atomic_fetch_<FETCHOP><IMODE>, atomic_fetch_nand<IMODE>): New.
719 (atomic_<FETCHOP>_fetch<IMODE>, atomic_nand_fetch<IMODE>): New.
720 (fetchadd_rel_<I48MODE>): New.
721 (sync_<FETCHOP><IMODE>, sync_nand<IMODE>): Remove.
722 (sync_old_<FETCHOP><IMODE>, sync_old_nand<IMODE>): Remove.
723 (sync_new_<FETCHOP><IMODE>, sync_new_nand<IMODE>): Remove.
724 (sync_compare_and_swap<IMODE>): Remove.
725 (sync_lock_release<IMODE>): Remove.
727 * config/ia64/ia64.md: Use define_c_enum for UNSPEC_*
728 and UNSPECV_* constants.
730 2011-11-22 Tom de Vries <tom@codesourcery.com>
732 PR rtl-optimization/50764
733 * sched-deps.c (sched_analyze_insn): Make sure frame-related insns are
734 not moved past preceding jump.
736 2011-11-22 Jeff Law <law@redhat.com>
738 * doc/contrib.texi: Add entry for David Binderman.
740 2011-11-22 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
742 * postreload.c (reload_combine): Mark reg_state as invalid at
743 volatile insns if there has been a use already.
745 2011-11-21 Aldy Hernandez <aldyh@redhat.com>
747 * opts.c (finish_options): Do not fail for -fgnu-tm.
748 * gimple-streamer-out.c (output_gimple_stmt): Handle
750 * gimple-streamer-in.c (input_gimple_stmt): Same.
751 * lto-cgraph.c (input_overwrite_node): Read tm_clone bit.
752 (lto_output_node): Write tm_clone bit.
754 2011-11-22 Ian Lance Taylor <iant@google.com>
756 * doc/install.texi (Configuration): Correct doc of
757 --enable-build-poststage1-with-cxx: it is not experimental.
759 2011-11-22 Jakub Jelinek <jakub@redhat.com>
761 PR tree-optimization/51074
762 * fold-const.c (fold_binary_loc): Fix up VEC_INTERLEAVE_*_EXPR
763 handling for BYTES_BIG_ENDIAN.
764 * optabs.c (can_vec_perm_for_code_p): Likewise.
766 2011-11-22 Michael Matz <matz@suse.de>
769 * trans-mem.c (expand_block_tm): Ignore clobbers.
771 2011-11-22 Iain Sandoe <iains@gcc.gnu.org>
773 * config/darwin.h (ENDFILE_SPEC): New.
774 (DARWIN_CRT1_SPEC): Add crttms.o.
775 (STARTFILE_SPEC): Likewise for dynamic libs and bundles.
776 * config/i386.h (ENDFILE_SPEC): Amend to include crttme.o.
778 2011-11-22 Michael Matz <matz@suse.de>
781 * tree.c (iterative_hash_expr): Remove break after return.
783 2011-11-22 Iain Sandoe <iains@gcc.gnu.org>
785 * target.def (tm_clone_table_section): New hook.
786 * doc/tm.texi.in (TARGET_ASM_TM_CLONE_TABLE_SECTION): Define.
787 * doc/tm.texi: Regenerate.
788 * varasm.c (dump_tm_clone_pairs): Use target tm_clone_table_section.
789 * output.h (default_clone_table_section): New prototype.
790 * config/darwin.h (TARGET_ASM_TM_CLONE_TABLE_SECTION): New.
791 * config/darwin-protos.h (darwin_tm_clone_table_section): New prototype.
792 * config/darwin.c (darwin_tm_clone_table_section): New.
794 2011-11-21 Torvald Riegel <triegel@redhat.com>
796 * trans-mem.c (diagnose_tm_1): Print an expression instead of a
797 declaration in error messages for indirect calls.
799 2011-11-21 David S. Miller <davem@davemloft.net>
801 * config/sparc/sparc.c (sparc_regmode_natural_size): New function
802 implementing REGMODE_NATURAL_SIZE taking into consideration vector
804 (sparc_modes_tieable_p): Similarly for MODES_TIEABLE_P.
805 * config/sparc/sparc-protos.h (sparc_regmode_natural_size,
806 sparc_modes_tieable_p): Declare.
807 * gcc/config/sparc/sparc.h (REGMODE_NATURAL_SIZE,
808 MODES_TIEABLE_P): Use new helper functions.
811 2011-11-16 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
814 * expmed.c (store_bit_field_1): Use extract_bit_field on big
815 endian targets if the source cannot be exactly covered by word
818 2011-11-21 Jakub Jelinek <jakub@redhat.com>
821 * var-tracking.c (loc_cmp): For ENTRY_VALUEs use ENTRY_VALUE_EXP
822 macro instead of XEXP.
824 2011-11-21 Andreas Tobler <andreast@fgznet.ch>
826 * configure: Regenerate.
828 2011-11-21 Georg-Johann Lay <avr@gjlay.de>
830 * config/avr/avr.c (output_reload_in_const): Loading a byte with 0
832 * config/avr/avr.md (*movhi, *movpsi, *movsi, *movsf): Zero to any
833 register does not change cc0. Same for any constant to ld-register.
835 2011-11-21 Uros Bizjak <ubizjak@gmail.com>
837 * config/i386/mmx.md (unspec) <UNSPEC_MOVNTQ, UNSPEC_PFRCP,
838 UNSPEC_PFRCPIT1, UNSPEC_PFRCPIT2, UNSPEC_PFRSQRT, UNSPEC_PFRSQIT1>:
839 Move from config/i386/i386.md
840 (unspecv) <UNSPECV_EMMS, UNSPECV_FEMMS>: Ditto.
841 * config/i386/sse.md (unspec) <UNSPEC_MOVNT,UNSPEC_MOVU, UNSPEC_LDDQU,
842 UNSPEC_PSHUFB, UNSPEC_PSIGN, UNSPEC_PALIGNR, UNSPEC_EXTRQI,
843 UNSPEC_EXTRQ, UNSPEC_INSERTQI, UNSPEC_INSERTQ, UNSPEC_BLENDV,
844 UNSPEC_INSERTPS, UNSPEC_DP, UNSPEC_MOVNTDQA, UNSPEC_MPSADBW,
845 UNSPEC_PHMINPOSUW, UNSPEC_PTEST, UNSPEC_PCMPESTR, UNSPEC_PCMPISTR,
846 UNSPEC_FMADDSUB, UNSPEC_XOP_UNSIGNED_CMP, UNSPEC_XOP_TRUEFALSE,
847 UNSPEC_XOP_PERMUTE, UNSPEC_FRCZ, UNSPEC_AESENC, UNSPEC_AESENCLAST,
848 UNSPEC_AESDEC, UNSPEC_AESDECLAST, UNSPEC_AESIMC,
849 UNSPEC_AESKEYGENASSIST, UNSPEC_PCLMUL, UNSPEC_PCMP, UNSPEC_VPERMIL,
850 UNSPEC_VPERMIL2, UNSPEC_VPERMIL2F128, UNSPEC_CAST, UNSPEC_VTESTP,
851 UNSPEC_VCVTPH2PS, UNSPEC_VCVTPS2PH, UNSPEC_VPERMSI, UNSPEC_VPERMDF,
852 UNSPEC_VPERMSF, UNSPEC_VPERMTI, UNSPEC_GATHER, UNSPEC_VSIBADDR>: Ditto.
853 (unspecv) <UNSPECV_LDMXCSR, UNSPECV_STMXCSR, UNSPECV_CLFLUSH,
854 UNSPECV_MONITOR, UNSPECV_MWAIT, UNSPECV_VZEROALL, UNSPECV_VZEROUPPER>:
856 * config/i386/sync.md (unspec) <UNSPEC_LFENCE, UNSPEC_SFENCE,
857 UNSPEC_MFENCE, UNSPEC_MOVA>: Ditto.
858 (unspecv) <UNSPECV_CMPXCHG_1, UNSPECV_CMPXCHG_2, UNSPECV_CMPXCHG_3,
859 UNSPECV_CMPXCHG_4, UNSPECV_XCHG, UNSPECV_LOCK>: Ditto.
860 (sse2_lfence): Move from config/i386/sse.md.
861 (*sse2_lfence): Ditto.
863 (*sse_sfence): Ditto.
864 (sse2_mfence): Ditto.
865 (mfence_sse2): Ditto. Rename from *sse2_mfence. Enable also
867 (mem_thread_fence): Use mfence_sse2.
869 2011-11-21 Georg-Johann Lay <avr@gjlay.de>
871 * config/avr/avr.h (struct base_arch_s): Add field sfr_offset.
872 * config/avr/avr-devices.c: Ditto. And initialize it.
873 * config/avr/avr-c.c (avr_cpu_cpp_builtins): New built-in define
875 * config/avr/avr-protos.h (out_movqi_r_mr, out_movqi_mr_r): Remove.
876 (out_movhi_r_mr, out_movhi_mr_r): Remove.
877 (out_movsi_r_mr, out_movsi_mr_r): Remove.
878 * config/avr/avr.md (*cbi, *sbi): Use %i instead of %m-0x20.
879 (*insv.io, *insv.not.io): Ditto.
880 * config/avr/avr.c (out_movsi_r_mr, out_movsi_mr_r): Make static.
881 (print_operand): Implement "%i" to print address as I/O address.
882 (output_movqi): Clean up call of out_movqi_mr_r.
883 (output_movhi): Clean up call of out_movhi_mr_r.
884 (avr_file_start): Use avr_current_arch->sfr_offset instead of
885 magic -0x20. Use TMP_REGNO, ZERO_REGNO instead of 0, 1.
886 (avr_out_sbxx_branch): Use %i instead of %m-0x20.
887 (out_movqi_r_mr, out_movqi_mr_r): Ditto. And make static.
888 (out_movhi_r_mr, out_movhi_mr_r): Ditto. And use avr_asm_len.
889 (out_shift_with_cnt): Clean up code: Use avr_asm_len.
890 (output_movsisf): Use output_reload_insisf for all CONSTANT_P sources.
891 (avr_out_movpsi): USE avr_out_reload_inpsi for all CONSTANT_P sources.
892 Clean up call of avr_out_store_psi.
893 (output_reload_in_const): Don't cut symbols longer than 2 bytes.
894 (output_reload_insisf): Filter CONST_INT_P or CONST_DOUBLE_P to
895 try if setting pre-cleared register is advantageous.
896 (avr_out_plus_1): Use gen_int_mode instead of GEN_INT.
898 2011-11-20 Joey Ye <joey.ye@arm.com>
900 * expr.c (expand_expr_real_1): Correctly handle strict volatile
901 bitfield loads smaller than mode size.
903 2011-11-20 Richard Henderson <rth@redhat.com>
905 * config/mips/mips.c (mips_init_libfuncs): Call init_sync_libfuncs.
907 2011-11-20 Uros Bizjak <ubizjak@gmail.com>
910 * config/i386/i386.c (ix86_expand_vcond): Generate TARGET_XOP
911 patterns for supported mode only.
914 * config/i386/i386.c (ix86_expand_builtin)
915 <IX86_BUILTIN_GATHERALTSIV4DI>: Use CODE_FOR_avx2_gathersiv4di.
917 2011-11-20 Uros Bizjak <ubizjak@gmail.com>
919 * config/i386/i386.md (UNSPEC_MOVNTI): Remove.
920 (UNSPEC_MOVNTQ): New unspec.
921 * config/i386/mmx.md (sse_movntq): Rename from sse_movntdi.
922 Use UNSPEC_MOVNTQ instead of UNSPEC_MOVNT.
923 * config/i386/sse.md (sse2_movnti<mode>): Use UNSPEC_MOVNT instead of
925 (STORENT_MODE): Add DI and V4DI modes.
927 2011-11-20 Nathan Sidwell <nathan@acm.org>
929 PR gcov-profile/51113
930 * coverage.c (build_var): Propagate visibility for public
931 decls. Use make_decl_one_only for PIC correctness.
934 * gcc.misc-tests/gcov-13.c: Check gcovpart-13b coverage
935 * gcc.misc-tests/gcov-16.c: New.
936 * gcc.misc-tests/gcov-17.c: New.
937 * g++.dg/gcov/gcov-8.C: New.
938 * g++.dg/gcov/gcov-9.C: New.
939 * g++.dg/gcov/gcov-10.C: New.
941 2011-11-19 Eric Botcazou <ebotcazou@adacore.com>
943 PR rtl-optimization/51187
944 * reorg.c (relax_delay_slots): Do not consider a jump useless if there
945 is a barrier between the jump and its target label.
947 2011-11-19 Patrick Marlier <patrick.marlier@gmail.com>
950 * tracer.c (ignore_bb_p): Don't copy GIMPLE_TRANSACTION.
952 2011-11-19 Eric Botcazou <ebotcazou@adacore.com>
954 * expmed.c (store_bit_field_1): Revert bogus formatting change.
956 2011-11-19 Richard Earnshaw <rearnsha@arm.com>
959 * config/arm/arm.c (neon_disambiguate_copy): Correctly handle partial
960 overlap of src and dest operands.
962 2011-11-19 Iain Sandoe <iains@gcc.gnu.org>
964 * config/darwin.h (ASM_DEBUG_SPEC): New.
965 * config/darwin9.h (ASM_DEBUG_SPEC): New.
967 2011-11-18 Jan Hubicka <jh@suse.cz>
969 * config/i386/i386.c (atom_cost): Disable SSE loop until alignment
972 2011-11-18 Anatoly Sokolov <aesok@post.ru>
974 * config/cris/cris.c (cris_function_value_regno_p): Make static.
975 (TARGET_FUNCTION_VALUE_REGNO_P): Define.
976 * config/cris/cris.h (FUNCTION_VALUE_REGNO_P): Remove.
977 * config/cris/cris-protos.h (cris_function_value_regno_p): Remove.
979 2011-11-18 H.J. Lu <hongjiu.lu@intel.com>
982 * doc/extend.texi: Document __builtin_ia32_movnti64.
984 * config/i386/emmintrin.h (_mm_stream_si64): New.
986 * config/i386/i386-builtin-types.def: Add VOID_FTYPE_PLONGLONG_LONGLONG.
988 * config/i386/i386.c (ix86_builtins): Add IX86_BUILTIN_MOVNTI64.
989 (bdesc_special_args): Update __builtin_ia32_movnti. Add
990 __builtin_ia32_movnti64.
991 (ix86_expand_special_args_builtin): Handle
992 VOID_FTYPE_PLONGLONG_LONGLONG.
994 * config/i386/i386.md (UNSPEC_MOVNTI): New.
996 * config/i386/sse.md (sse2_movntsi): Renamed to ...
997 (sse2_movnti<mode>): This.
999 2011-11-18 Georg-Johann Lay <avr@gjlay.de>
1002 * config/avr/avr.h (base_arch_s): Add field n_segments.
1003 (ADDR_SPACE_PGM1, ADDR_SPACE_PGM2, ADDR_SPACE_PGM3,
1004 ADDR_SPACE_PGM4, ADDR_SPACE_PGM5, ADDR_SPACE_PGMX): New address spaces.
1005 (AVR_HAVE_ELPM, AVR_HAVE_ELPMX): New defines.
1006 (INIT_EXPANDERS): New define.
1007 * config/avr/avr-protos.h (avr_mem_pgmx_p): New.
1008 (avr_init_expanders): New.
1009 (avr_emit_movmemhi, avr_out_movmem): New.
1010 (avr_xload_libgcc_p): New.
1011 * config/avr/avr-c.c (avr_register_target_pragmas): Register
1012 address spaces __pgm1, __pgm2, __pgm3, __pgm4 __pgm5, __pgmx.
1013 (avr_cpu_cpp_builtins): Add built-in defines __PGM1,
1014 __PGM2, __PGM3, __PGM4, __PGM5, __PGMX.
1015 * config/avr/avr-devices.c (avr_arch_types): Set field n_segments.
1017 * config/avr/avr.c (AVR_SECTION_PROGMEM): Change define to cover
1018 3 bits instead of just 1.
1019 (xstring_empty, xstring_e, rampz_rtx): New static GTYed variables.
1020 (progmem_section): Change from section to array of sections.
1021 (progmem_section_prefix): New static variable.
1022 (avr_file_start): Print set for __RAMPZ__
1023 (avr_option_override): Move initialization of RTXes from here...
1024 (avr_init_expanders): ...to this new function.
1025 (avr_pgm_segment): New static function.
1026 (avr_decl_pgm_p): Handle error_mark_node.
1027 (avr_mem_pgmx_p, avr_decl_pgmx_p): New static functions.
1028 (avr_out_xload, avr_find_unused_d_reg): New static functions.
1029 (expand_prologue, expand_epilogue): Use rampz_rtx.
1030 (print_operand): Hande CONST_STRING.
1031 (avr_xload_libgcc_p): New static function.
1032 (avr_out_lpm_no_lpmx, avr_out_lpm): Handle ELPM.
1033 (avr_progmem_p): Return 2 for 24-bit flash address space.
1034 (avr_out_sbxx_branch): Clean-up code from ASn macros.
1035 (out_movqi_r_mr, out_movqi_mr_r): Ditto. And recognize RAMPZ's
1036 address and print symbolically.
1037 (avr_asm_named_section, avr_section_type_flags,
1038 avr_encode_section_info, avr_asm_select_section,
1039 avr_addr_space_address_mode, avr_addr_space_pointer_mode,
1040 avr_addr_space_legitimate_address_p, avr_addr_space_convert,
1041 avr_addr_space_legitimize_address): Handle new address spaces.
1042 (avr_output_progmem_section_asm_op): New static function.
1043 (avr_asm_init_sections): Initialize progmem_section[].
1044 (adjust_insn_length): Handle ADJUST_LEN_XLOAD, ADJUST_LEN_MOVMEM.
1045 (avr_const_address_lo16): New static function.
1046 (avr_assemble_integer): Use it to handle 3-byte integers.
1047 (avr_emit_movmemhi, avr_out_movmem): New functions.
1049 * config/avr/predicates.md (nox_general_operand): Handle new
1051 * config/avr/avr.md (unspec): Add UNSPEC_MOVMEM.
1052 (adjust_len): Add xload, movmem.
1053 (SP_ADDR): New define_constants.
1054 (isa): Add "lpm", "lpmx", "elpm", "elpmx".
1055 (enabled): Handle them.
1056 (load<mode>_libgcc): New expander.
1057 (*load.<mode>.libgcc): Rename to load_<mode>_libgcc.
1058 (xload8_A, xload<mode>_A): New insn-and-splits.
1059 (xload_8, xload_<mode>_libgcc, xload_<mode>, loadmem_elpm): New insns.
1060 (mov<mode>): Handle new address spaces.
1061 (movmemhi): Rewrite using avr_emit_movmemhi.
1062 (MOVMEM_r_d): New mode attribute.
1063 (movmem_<mode>, movmem_qi_elpm): New insns.
1064 (setmemhi, *clrmemqi, *clrmemhi, strlenhi, *strlenhi): Unquote
1065 C-code. Use label instead of hard-coded instrunction lengths.
1067 2011-11-18 Martin Jambor <mjambor@suse.cz>
1069 PR tree-optimization/50605
1070 * gimple.c (is_gimple_ip_invariant_address): Also handle MEM_REFs
1071 of IPA invariant decls.
1073 2011-11-18 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1075 * tree-outof-ssa.c (insert_back_edge_copies): Add call to
1076 mark_dfs_back_edges.
1078 2011-11-18 Iain Sandoe <iains@gcc.gnu.org>
1081 * configure.ac: Remove ranlib special-casing for Darwin.
1082 * configure: Regenerate.
1084 2011-11-18 Uros Bizjak <ubizjak@gmail.com>
1086 PR tree-optimization/51118
1087 * fold-const.c (fold_checksum_tree): Check for TS_TYPED structure
1088 before using TREE_TYPE accessor on expr.
1090 2011-11-17 Jan Hubicka <jh@suse.cz>
1093 * i386.c (atom_cost): Fix 32bit memset description.
1094 (expand_set_or_movmem_via_loop_with_iter): Output proper bounds check
1096 (expand_movmem_epilogue): Handle epilogues up to size 15 w/o producing
1098 (decide_alg): sse_loop is not useable wthen SSE2 is disabled;
1099 when not optimizing always use rep movsb or lincall; do not produce
1100 word sized loops when optimizing memset for size (to avoid need
1101 for large constants).
1102 (ix86_expand_movmem): Get into sync with ix86_expand_setmem;
1103 choose unroll factors better; always do 128bit moves when producing
1104 SSE loops; do not produce loopy epilogue when size is too small.
1105 (promote_duplicated_reg_to_size): Do not look into desired alignments
1106 when doing vector expansion.
1107 (ix86_expand_setmem): Track better when promoted value is available;
1108 choose unroll factors more sanely. Output loopy epilogue only
1111 2011-11-17 Steve Ellcey <sje@cup.hp.com>
1114 * output.h (fprint_w): Remove.
1115 * final.c (fprint_w): Remove.
1116 (output_addr_const): Change fprint_w back to fprintf.
1118 2011-11-17 Andrew Pinski <apinski@cavium.com>
1120 * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for MIPS.
1122 2011-11-17 Andrew MacLeod <amacleod@redhat.com>
1124 * builtins.c (expand_builtin): Remove 4th parameter representing
1125 weak/strong mode when __atomic_compare_exchange becomes a library call.
1127 2011-11-17 Richard Henderson <rth@redhat.com>
1129 * config/rs6000/rs6000.c (rs6000_expand_atomic_compare_and_swap): Get
1130 new pseudo for target after convert_modes.
1131 (rs6000_expand_atomic_exchange, rs6000_expand_atomic_op): Likewise.
1133 2011-11-17 Richard Henderson <rth@redhat.com>
1135 * config/rs6000/rs6000.c (rs6000_expand_atomic_compare_and_swap): Get
1136 new pseudo for target after convert_modes.
1137 (rs6000_expand_atomic_exchange, rs6000_expand_atomic_op): Likewise.
1139 2011-11-17 Richard Henderson <rth@redhat.com>
1141 * builtins.c (expand_builtin_mem_thread_fence): Remove.
1142 (expand_builtin_mem_signal_fence): Remove.
1143 (expand_builtin_atomic_thread_fence): Use expand_mem_thread_fence.
1144 (expand_builtin_sync_synchronize): Likewise.
1145 (expand_builtin_atomic_signal_fence): Use expand_mem_signal_fence.
1146 * optabs.c (expand_asm_memory_barrier): Split out from
1147 expand_builtin_mem_signal_fence.
1148 (expand_mem_thread_fence): New, a combination of code from
1149 expand_builtin_mem_thread_fence and expand_builtin_sync_synchronize.
1150 (expand_mem_signal_fence): Moved and renamed from
1151 expand_builtin_mem_signal_fence.
1152 (expand_atomic_exchange): Use expand_mem_thread_fence.
1153 (expand_atomic_load, expand_atomic_store): Likewise.
1154 * expr.h, optabs.h: Update decls.
1156 2011-11-17 Bin Cheng <bin.cheng@arm.com>
1158 PR rtl-optimization/50663
1159 * cprop.c (implicit_set_indexes): New global variable.
1160 (insert_set_in_table): Add additional parameter and record implicit
1162 (hash_scan_set): Add additional parameter and pass it to above.
1163 (hash_scan_insn): Pass false to hash_scan_set.
1164 (compute_hash_table_work): Pass true to hash_scan_set.
1165 (compute_cprop_data): Add implicit set to AVIN of block which the
1166 implicit set is recorded for.
1167 (one_cprop_pass): Handle implicit_set_indexes array.
1169 2011-11-17 Michael Matz <matz@suse.de>
1173 * tree-ssa-live.c (mark_all_vars_used_1): Recurse only for decls of
1175 (remove_unused_locals): Ditto.
1177 2011-11-16 Richard Henderson <rth@redhat.com>
1179 * config/alpha/alpha.c (alpha_split_compare_and_swap_12): Copy
1180 ALIAS_SET_MEMORY_BARRIER when creating a new memory.
1181 (alpha_split_atomic_exchange_12): Likewise.
1182 * config/rs6000/rs6000.c (rs6000_adjust_atomic_subword): Likewise.
1184 2011-11-16 Joseph Myers <joseph@codesourcery.com>
1186 * config/i386/i386elf.h (PREFERRED_DEBUGGING_TYPE): Remove.
1188 2011-11-16 David Edelsohn <dje.gcc@gmail.com>
1190 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
1191 Default to minimum alignment if decl is NULL.
1193 2011-11-16 Uros Bizjak <ubizjak@gmail.com>
1195 * config/i386/sse.md (round<mode>2_vec_pack_sfix): Optimize V2DFmode
1197 (<sse4_1>_round<ssemodesuffix>_vec_pack_sfix<avxsizesuffix>): Ditto.
1199 2011-11-16 Venkataramanan Kumar <venkataramanan.kumar@amd.com>
1201 * doc/invoke.texi: Document AMD bdver1 and btver1.
1203 2011-11-16 Richard Earnshaw <rearnsha@arm.com>
1204 Bernd Schmidt <bernds@coudesourcery.com>
1205 Sebastian Huber <sebastian.huber@embedded-brains.de>
1208 * config/arm/arm.c (store_multiple_sequence): Avoid cases where
1209 the base reg is stored iff compiling for Thumb1.
1211 2011-11-16 Razya Ladelsky <razya@il.ibm.com>
1213 PR tree-optimization/49960
1214 * tree-data-ref.c (initialize_data_dependence_relation): Add
1216 Remove call to compute_self_dependence.
1217 (compute_affine_dependence): Remove the !DDR_SELF_REFERENCE condition.
1218 (compute_self_dependence): Remove old code. Add call to
1219 compute_affine_dependence.
1220 (compute_all_dependences): Remove call to compute_self_dependence.
1221 Add call to compute_affine_dependence.
1223 2011-11-16 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
1226 * expmed.c (store_bit_field_1): Use extract_bit_field on big
1227 endian targets if the source cannot be exactly covered by word
1230 2011-11-15 Joseph Myers <joseph@codesourcery.com>
1232 * config/i386/i386elf.h (ASM_OUTPUT_ASCII): Change STRING_LIMIT to
1235 2011-11-15 Richard Henderson <rth@redhat.com>
1237 * config/alpha/alpha.c (alpha_pre_atomic_barrier): New.
1238 (alpha_post_atomic_barrier): New.
1239 (alpha_split_atomic_op): New memmodel argument; honor it.
1240 (alpha_split_compare_and_swap): Take array of operands. Honor
1241 memmodel; always set bool output
1242 (alpha_expand_compare_and_swap_12): Similarly.
1243 (alpha_split_compare_and_swap_12): Similarly.
1244 (alpha_split_atomic_exchange): Similarly. Rename from
1245 alpha_split_lock_test_and_set.
1246 (alpha_expand_atomic_exchange_12): Similarly. Rename from
1247 alpha_expand_lock_test_and_set_12.
1248 (alpha_split_atomic_exchange_12): Similarly. Rename from
1249 alpha_split_lock_test_and_set_12.
1250 * config/alpha/alpha-protos.h: Update.
1251 * config/alpha/alpha.md (UNSPECV_CMPXCHG): New.
1252 * config/alpha/constraints.md ("w"): New.
1253 * config/alpha/predicates.md (mem_noofs_operand): New.
1254 * config/alpha/sync.md (atomic_compare_and_swap<mode>): Rename from
1255 sync_compare_and_swap<mode>; add the new parameters.
1256 (atomic_exchange<mode>): Update from sync_test_and_set<mode>.
1257 (atomic_fetch_<op><mode>): Update from sync_old_<op><mode>.
1258 (atomic_<op>_fetch<mode>): Update from sync_new_<op><mode>.
1259 (atomic_<op><mode>): Update from sync_<op><mode>.
1261 2011-11-16 Tom de Vries <tom@codesourcery.com>
1263 * tree-ssa-tail-merge.c (replace_block_by): Add frequency of bb2 to bb1.
1265 2011-11-16 Alan Modra <amodra@gmail.com>
1267 PR rtl-optimization/51051
1269 * function.c (thread_prologue_and_epilogue_insns): Guard
1270 emitting return with single_succ_p test.
1272 2011-11-15 Joern Rennecke <joern.rennecke@embecosm.com>
1274 * config/epiphany/epiphany.md (mov<mode>cc): Fix code to
1275 get mode from CMP_OP1 if CMP_OP0 is VOIDmode.
1277 2011-11-15 Uros Bizjak <ubizjak@gmail.com>
1279 * config/i386/sse.md (vec_pack_trunc_v2df): Optimize sequence for AVX.
1280 (vec_pack_sfix_trunc_v2df): Ditto.
1281 (vec_pack_sfix_v2df): Ditto.
1282 (vec_pack_sfix_trunc_v4df): Generate fix_truncv4dfv4si2 and
1283 avx_vec_concatv8si patterns.
1284 (vec_pack_sfix_v4df): Generate avx_cvtpd2dq256 and
1285 avx_vec_concatv8si patterns.
1287 2011-11-15 Torvald Riegel <triegel@redhat.com>
1289 * c-parser.c (c_parser_transaction_expression): Require parentheses
1290 when parsing transaction expressions.
1292 2011-11-15 Tristan Gingold <gingold@adacore.com>
1294 * incpath.c (get_added_cpp_dirs): New function.
1295 * incpath.h (get_added_cpp_dirs): Declare.
1296 * config/vms/vms-c.c (vms_c_register_includes): New function.
1297 (vms_std_modules): New variable.
1298 * config/vms/vms.h (TARGET_EXTRA_INCLUDES): Define.
1299 (vms_c_register_includes): Declare.
1301 2011-11-15 Tristan Gingold <gingold@adacore.com>
1303 * c-family/c-pragma.h (pragma_extern_prefix): Declare.
1304 * c-family/c-pragma.c (pragma_extern_prefix): Make it public.
1305 * config/vms/vms-c.c (saved_extern_prefix): New variable.
1306 (vms_pragma_extern_prefix): New function.
1307 (vms_c_register_pragma): Register vms_pragma_extern_prefix.
1309 2011-11-15 Georg-Johann Lay <avr@gjlay.de>
1312 * config/avr/avr.h (ADDR_SPACE_PGM): New address spaces.
1313 (REGISTER_TARGET_PRAGMAS): New define.
1314 * config/avr/avr-protos.h (avr_mem_pgm_p): New.
1315 (avr_load_libgcc_p): New.
1316 (avr_register_target_pragmas): New.
1317 (asm_output_external_libcall): Remove.
1318 (avr_log_t): Add field "progmem". Order alphabetically.
1319 * config/avr/avr-log.c (avr_log_set_avr_log): Set avr_log.progmem.
1320 * config/avr/avr-c.c (langhooks.h): New include.
1321 (avr_register_target_pragmas): New function. Register address
1323 (avr_cpu_cpp_builtins): Add built-in define __PGM.
1325 * config/avr/avr.c: Include "c-family/c-common.h".
1326 (TARGET_LEGITIMATE_ADDRESS_P): Remove define.
1327 (TARGET_LEGITIMIZE_ADDRESS): Remove define.
1328 (TARGET_ADDR_SPACE_SUBSET_P): Define to...
1329 (avr_addr_space_subset_p): ...this new static function.
1330 (TARGET_ADDR_SPACE_CONVERT): Define to...
1331 (avr_addr_space_convert): ...this new static function.
1332 (TARGET_ADDR_SPACE_ADDRESS_MODE): Define to...
1333 (avr_addr_space_address_mode): ...this new static function.
1334 (TARGET_ADDR_SPACE_POINTER_MODE): Define to...
1335 (avr_addr_space_pointer_mode): ...this new static function.
1336 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): Define to...
1337 (avr_addr_space_legitimate_address_p): ...this new static function.
1338 (TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS): Define to...
1339 (avr_addr_space_legitimize_address): ...this new static function.
1340 (avr_mode_code_base_reg_class): Handle address spaces.
1341 (avr_regno_mode_code_ok_for_base_p): Ditto.
1342 (lpm_addr_reg_rtx, lpm_reg_rtx, all_regs_rtx): New static variables.
1343 (avr_option_override): Initialize them.
1344 (output_reload_in_const): Use all_regs_rtx. Fix signedness of loop
1346 (avr_pgm_segment): New static function.
1347 (avr_decl_pgm_p, avr_mem_pgm_p): New static functions.
1348 (avr_out_lpm, avr_out_lpm_no_lpmx): New static functions.
1349 (output_movqi, output_movhi, output_movsisf, avr_out_movpsi): Call
1350 avr_out_lpm to handle loads from progmem.
1351 (avr_load_libgcc_p): New static function.
1352 (avr_progmem_p): Test if decl is in flash.
1353 (avr_pgm_pointer_const_p): New static function.
1354 (avr_nonconst_pointer_addrspace): New static function.
1355 (avr_pgm_check_var_decl): New static function.
1356 (avr_insert_attributes): Use it. Change error message to report
1357 cause (progmem or address space) when code wants to write to flash.
1358 (avr_section_type_flags): Unset section flag SECTION_BSS for
1361 * config/avr/predicates.md (nop_general_operand): New predicate.
1362 (nox_general_operand): New predicate.
1363 * config/avr/avr.md (LPM_REGNO): New define_constant.
1364 (load<mode>_libgcc): New expander.
1365 (*load.<mode>.libgcc): New insn.
1366 (mov<mode>): Handle loads from non-generic AS.
1367 (movmemhi): Ditto. Propagate address space information to newly
1369 (movqi_insn, *movhi, *movpsi, *movsi, *movsf): Change predicate #1
1370 to nox_general_operand.
1371 (ashrqi3, ashrhi3, ashrsi3): Change predicate #1 to
1372 nop_general_operand.
1373 (ashlqi3, *ashlqi3, ashlhi3, ashlsi3): Ditto.
1374 (lshrqi3, *lshrqi3, lshrhi3, lshrsi3): Ditto.
1375 (split-lpmx): New split.
1376 (*ashlhi3_const, *ashlsi3_const, *ashrhi3_const, *ashrsi3_const,
1377 *lshrhi3_const, *lshrsi3_const): Indent, unquote C.
1379 2011-11-15 Maxim Kuvyrkov <maxim@codesourcery.com>
1381 * ipa-cp.c (ipa_value_from_jfunc): Make global.
1382 (ipa_cst_from_jfunc): Remove, use ipa_value_from_jfunc instead.
1383 (get_indirect_edge_target): Rename, make global.
1384 (devirtualization_time_bonus, estimate_local_effects,)
1385 (ipcp_discover_new_direct_edges): Update.
1386 * ipa-inline-analysis.c (evaluate_conditions_for_edge):
1387 Generalize to also handle types. Rename to ...
1388 (evaluate_properties_for_edge): Use instead of
1389 evaluate_conditions_for_edge.
1390 (estimate_edge_devirt_benefit): New function.
1391 (estimate_calls_size_and_time): Use it.
1392 (estimate_node_size_and_time, estimate_ipcp_clone_size_and_time,)
1393 (inline_merge_summary): Update.
1394 (do_estimate_edge_time, do_estimate_edge_growth): Update. Calculate
1395 parameter information at the call site and pass it on to subroutines.
1396 * tree-inline.c (estimate_num_insns): Distinguish between direct and
1398 (init_inline_once): Set size and time costs or indirect calls.
1399 * tree-inline.h (eni_weights): Add indirect_call_cost.
1401 2011-11-15 Tom de Vries <tom@codesourcery.com>
1403 PR tree-optimization/51005
1404 * tree-ssa-tail-merge.c (delete_basic_block_same_succ): Rename to
1405 mark_basic_block_deleted.
1406 (update_worklist): Inline purge_bbs.
1407 (purge_bbs, unlink_virtual_phi, update_vuses, vop_at_entry)
1408 (delete_block_update_dominator_info): Remove.
1409 (replace_block_by): Remove update_vops parameter. Partially evaluate
1410 for update_vops == false.
1411 (apply_clusters): Remove update_vops parameter. Remove update_vops
1412 argument in replace_block_by call.
1413 (update_debug_stmts): Remove MAY_HAVE_DEBUG_STMTS test.
1414 (tail_merge_optimize): Remove update_vops argument to apply_clusters.
1415 Remove call to purge_bbs. Add calls to calculate_dominance_info and
1416 free_dominance_info. Add MAY_HAVE_DEBUG_STMTS before calling
1417 update_debug_stmts. Mark vop var for renaming, if necessary.
1419 2011-11-15 Bernd Schmidt <bernds@codesourcery.com>
1421 PR rtl-optimization/51051
1422 * cfgrtl.c (cfg_layout_can_merge_blocks_p): Return FALSE if the
1423 move would cause fallthrough into the exit block.
1425 2011-11-14 Richard Henderson <rth@redhat.com>
1427 * config/rs6000/rs6000.c (emit_load_locked): Assert the mode is
1429 (emit_store_conditional): Likewise.
1430 (rs6000_pre_atomic_barrier, rs6000_post_atomic_barrier): New.
1431 (rs6000_adjust_atomic_subword): New.
1432 (rs6000_mask_atomic_subword, rs6000_finish_atomic_subword): New.
1433 (rs6000_expand_atomic_op): Rename from rs6000_emit_sync; rewrite
1434 for pre-reload expansion.
1435 (rs6000_split_compare_and_swap, rs6000_split_compare_and_swapqhi,
1436 rs6000_expand_compare_and_swapqhi): Merge into ...
1437 (rs6000_expand_atomic_compare_and_swap): ... here. New function.
1438 rs6000_split_lock_test_and_set; expand immediately. Handle
1440 * config/rs6000/rs6000.md (UNSPEC_LWSYNC): Move and rename
1441 from UNSPECV_LWSYNC.
1442 * config/rs6000/sync.md (fetchopsi_constr, fetchopdi_constr): Remove.
1443 (mem_thread_fence): New.
1444 (hwsync): Rename from memory_barrier.
1445 (*hwsync): Rename from *sync_internal.
1446 (lwsync, *lwsync): Mirror hwsync implementation.
1447 (isync): Don't reference memory.
1449 (atomic_load<INT>, atomic_store<INT>): New.
1450 (ATOMIC): New mode iterator.
1451 (load_locked<ATOMIC>): Rename from load_locked_<GPR>.
1452 (store_conditional<ATOMIC>): Rename from store_conditional_<GPR>.
1453 (sync_compare_and_swap<GPR>): Remove.
1454 (sync_compare_and_swaphi, sync_compare_and_swapqi): Remove.
1455 (sync_compare_and_swapqhi_internal): Remove.
1456 (sync_lock_test_and_set<GPR>): Remove.
1457 (sync_<FETCHOP><INT1>): Remove.
1458 (*sync_<FETCHOP>si_internal, *sync_<FETCHOP>di_internal): Remove.
1459 (sync_nand<INT1>, *sync_nand<GPR>_internal): Remove.
1460 (sync_old_<FETCHOP><GPR>): Remove.
1461 (*sync_old_<FETCHOP>si_internal): Remove.
1462 (*sync_old_<FETCHOP>di_internal): Remove.
1463 (sync_old_nand<INT1>): Remove.
1464 (*sync_old_nand<GPR>_internal): Remove.
1465 (sync_new_<FETCHOP><GPR>): Remove.
1466 (*sync_new_<FETCHOP>si_internal): Remove.
1467 (*sync_new_<FETCHOP>di_internal): Remove.
1468 (sync_new_nand<INT1>): Remove.
1469 (*sync_new_nand<GPR>_internal): Remove.
1470 (*atomic_andsi, *atomic_anddi): Remove.
1471 (*sync_addshort_internal, *sync_subshort_internal): Remove.
1472 (*sync_andsi_internal, *sync_boolsi_internal): Remove.
1473 (*sync_boolcshort_internal): Remove.
1474 (sync_lock_release<INT1>): Remove.
1475 (atomic_compare_and_swap<INT1>): New.
1476 (atomic_exchange<INT1>): New.
1477 (atomic_<FETCHOP><INT1>, atomic_nand<INT1>): New.
1478 (atomic_fetch_<FETCHOP><INT1>, atomic_fetch_nand<INT1>): New.
1479 (atomic_<FETCHOP>_fetch<INT1>, atomic_nand_fetch<INT1>): New.
1481 2011-11-14 Uros Bizjak <ubizjak@gmail.com>
1483 * config/i386/sse.md (round<mode>2): Use register_operand for
1484 operand 1 predicate.
1485 (round<mode>2_sfix): Ditto.
1486 (round<mode>2_vec_pack_sfix): Ditto for operand 1 and operand 2.
1488 2011-11-14 Nathan Sidwell <nathan@acm.org>
1490 * gcov.c (canonicalize_name): Protect use of S_ISLNK.
1492 2011-11-14 Jan Hubicka <jh@suse.cz>
1495 * config/i386/i386.c (core cost model): Correct pasto.
1497 2011-11-14 Uros Bizjak <ubizjak@gmail.com>
1499 * config/i386/sse.md (round<mode>2_sfix): New expander.
1500 (round<mode>2_vec_pack_sfix): Ditto.
1501 (<sse4_1>_round<ssemodesuffix>_sfix<avxsizesuffix>): Ditto.
1502 (<sse4_1>_round<ssemodesuffix>_vec_pack_sfix<avxsizesuffix>): Ditto.
1503 * config/i386/builtin-types.def (V4SI_FTYPE_V4SF_ROUND,
1504 V8SI_FTYPE_V8SF_ROUND, V4SI_FTYPE_V2DF_V2DF_ROUND,
1505 V8SI_FTYPE_V4DF_V4DF_ROUND): New builtin types.
1506 * config/i386/i386.c (ix86_builtins): Add
1507 IX86_BUILTIN_{FLOORPD,CEILPD,ROUNDPD_AZ}_VEC_PACK_SFIX{,256} and
1508 IX86_BUILTIN_{FLOORPS,CEILPS,ROUNDPS_AZ}_SFIX{,256} defines.
1509 (bdesc_args): Add __builtin_ia32_{floorpd,ceilpd}_vec_pack_sfix{,256},
1510 __builtin_ia32_roundpd_az_vec_pack_sfix{,256},
1511 __builtin_ia32_{floorps,ceilps}_sfix{,256}and
1512 __builtin_ia32_roundps_az_sfix{,256} descriptions.
1513 (ix86_expand_sse_round_vec_pack_sfix): New.
1514 (ix86_expand_args_builtin): Handle V4SI_FTYPE_V4SF_ROUND,
1515 V8SI_FTYPE_V8SF_ROUND, V4SI_FTYPE_V2DF_V2DF_ROUND and
1516 V8SI_FTYPE_V4DF_V4DF_ROUND types. Check last argument of
1517 CODE_FOR_sse4_1_roundpd_vec_pack_sfix, CODE_FOR_sse4_1_roundps_sfix,
1518 CODE_FOR_avx_roundpd_vec_pack_sfix256 and CODE_FOR_avx_roundps_sfix256.
1519 (ix86_builtin_vectorized_function): Handle
1520 BUILT_IN_{I,L,LL}FLOOR{,F}, BUILT_IN_{I,L,LL}CEIL{,F} and
1521 BUILT_IN_{I,L,LL}ROUND{,F}
1523 2011-11-14 Jan Hubicka <jh@suse.cz>
1526 * cgraph.h (varpool_externally_visible_p): Declare.
1527 (varpool_add_new_variable): Declare.
1528 * tree-emultls.c (get_emutls_init_templ_addr,
1529 new_emutls_decl): Use varpool_add_new_variable.
1530 * ipa.c (varpool_externally_visible_p): Export.
1531 * varpool.c (varpool_add_new_variable): New function.
1533 2011-11-14 Michael Zolotukhin <michael.v.zolotukhin@gmail.com>
1534 Jan Hubicka <jh@suse.cz>
1536 * config/i386/i386.h (processor_costs): Add second dimension to
1537 stringop_algs array.
1538 * config/i386/i386.c (cost models): Initialize second dimension of
1539 stringop_algs arrays.
1540 (core_cost): New costs based on generic64 costs with updated stringop
1542 (promote_duplicated_reg): Add support for vector modes, add
1544 (promote_duplicated_reg_to_size): Likewise.
1545 (processor_target): Set core costs for core variants.
1546 (expand_set_or_movmem_via_loop_with_iter): New function.
1547 (expand_set_or_movmem_via_loop): Enable reuse of the same iters in
1548 different loops, produced by this function.
1549 (emit_strset): New function.
1550 (expand_movmem_epilogue): Add epilogue generation for bigger sizes,
1551 use SSE-moves where possible.
1552 (expand_setmem_epilogue): Likewise.
1553 (expand_movmem_prologue): Likewise for prologue.
1554 (expand_setmem_prologue): Likewise.
1555 (expand_constant_movmem_prologue): Likewise.
1556 (expand_constant_setmem_prologue): Likewise.
1557 (decide_alg): Add new argument align_unknown. Fix algorithm of
1558 strategy selection if TARGET_INLINE_ALL_STRINGOPS is set; Skip sse_loop
1559 (decide_alignment): Update desired alignment according to chosen move
1561 (ix86_expand_movmem): Change unrolled_loop strategy to use SSE-moves.
1562 (ix86_expand_setmem): Likewise.
1563 (ix86_slow_unaligned_access): Implementation of new hook
1564 slow_unaligned_access.
1565 * config/i386/i386.md (strset): Enable half-SSE moves.
1566 * config/i386/sse.md (vec_dupv4si): Add expand for vec_dupv4si.
1567 (vec_dupv2di): Add expand for vec_dupv2di.
1569 2011-11-14 Dimitrios Apostolou <jimis@gmx.net>
1573 * config/elfos.h (TARGET_ASM_INTERNAL_LABEL): Remove.
1574 * varasm.c (default_elf_internal_label): Remove.
1576 2011-11-14 Kai Tietz <ktietz@redhat.com>
1578 * gcov.c (generate_results): Add missing semicolon and correct indent.
1580 2011-11-14 Ira Rosen <ira.rosen@linaro.org>
1583 * tree-vect-stmts.c (vectorizable_condition): Initialize comp_vectype.
1585 2011-11-14 Mingjie Xing <mingjie.xing@gmail.com>
1587 * doc/invoke.texi (Wunused-result): Change @pxref{Variable Attributes}
1588 to @pxref{Function Attributes}.
1590 2011-11-13 Oleg Endo <oleg.endo@t-online.de>
1593 * config/sh/sh.h (IS_LITTLE_ENDIAN_OPTION, UNSUPPORTED_SH2A):
1595 (DRIVER_SELF_SPECS): Use new macros to filter out unsupported options
1596 taking the default configuration into account.
1598 2011-11-13 Jonathan Wakely <jwakely.gcc@gmail.com>
1600 * doc/install.texi (Specific): Fix anchor for
1601 x86_64-*-solaris2.1[0-9]*.
1603 2011-11-13 Nathan Sidwell <nathan@acm.org>
1605 * gcov.c (source_prefix, source_length): New globals.
1606 (flag_relative_only): Likewise.
1607 (print_usage, options, process_args): Update.
1608 (generate_results): Use coverage.name, check
1609 flag_relative_only. Adjust messages.
1610 (find_source): Check source_prefix.
1611 (output_lines): Use coverage.name, adjust messages.
1612 * doc/gcov.texi (Invoking Gcov): Document new options.
1614 2011-11-12 Jason Merrill <jason@redhat.com>
1617 * gimplify.c (gimplify_target_expr): Add a clobber to the cleanup.
1618 (gimplify_modify_expr): Don't try to simplify it.
1620 2011-11-12 Dimitrios Apostolou <jimis@gmx.net>
1623 * config/elfos.h (STRING_ASM_OP): Define.
1624 * varasm.c (default_elf_asm_output_limited_string): Use it.
1627 * gcc/configure.ac: Check for stpcpy declaration.
1628 * system.h: Declare it if not.
1629 * config.in, configure: Regenerate.
1631 2011-11-12 Richard Henderson <rth@redhat.com>
1633 * config/rs6000/rs6000.md (fix_trunc<SFDF>si2_stfiwx): Use
1634 nonimmediate_operand for the destination.
1635 (fixuns_trunc<SFDF>si2_stfiwx): Likewise.
1637 * builtins.c (expand_builtin_compare_and_swap): If target is const0,
1638 don't pass the target to expand_atomic_compare_and_swap.
1639 (expand_builtin_atomic_compare_exchange): Likewise.
1641 * config/sh/linux.h (TARGET_INIT_LIBFUNCS): New.
1642 * config/sh/sh.c (sh_init_sync_libfuncs): New.
1644 * config/pa/pa-linux.h (TARGET_SYNC_LIBCALL): New.
1645 * config/pa/pa.h (TARGET_SYNC_LIBCALL): New default.
1646 * config/pa/pa.c (pa_init_libfuncs): Use init_sync_libfuncs.
1648 2011-11-12 Iain Sandoe <iains@gcc.gnu.org>
1651 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
1652 Only expand a symbol ref. into an access when the entity is defined
1655 2011-11-12 Jan Hubicka <jh@suse.cz>
1658 * cgraphunit.c (output_weakrefs): Output really only weakrefs.
1660 2011-11-12 Nathan Sidwell <nathan@acm.org>
1662 * gcov.c (struct name_map): New.
1663 (names, n_names, a_names): New global vars.
1664 (print_usage): Adjust usage.
1665 (generate_results): Canonicalize main file name.
1666 (release_structures): Adjust.
1667 (name_search, name_sort): New callbacks.
1668 (find_source): Look for and create a canonical name.
1669 (canonicalize_name): New.
1670 (make_gcov_file_name): Reimplement and fix mangling.
1672 * doc/gcov.texi: Update documentation about path preservation.
1674 2011-11-11 David S. Miller <davem@davemloft.net>
1676 * config/sparc/linux.h (ASM_GENERATE_INTERNAL_LABEL): Delete.
1677 * config/sparc/linux64.h (ASM_GENERATE_INTERNAL_LABEL): Delete.
1679 2011-11-11 Jakub Jelinek <jakub@redhat.com>
1681 * config/i386/i386-protos.h (ix86_maybe_emit_epilogue_vzeroupper):
1683 * config/i386/i386.c (ix86_maybe_emit_epilogue_vzeroupper): New
1685 (ix86_expand_epilogue): Use it.
1686 * config/i386/i386.md (return, simple_return): Call it in the
1689 PR tree-optimization/51091
1690 * tree-stdarg.c (execute_optimize_stdarg): Ignore TREE_CLOBBER_P
1691 rhs also in the va_list_simple_ptr case.
1693 PR tree-optimization/51058
1694 * tree-vect-slp.c (vect_remove_slp_scalar_calls): New function.
1695 (vect_schedule_slp): Call it.
1696 * tree-vect-stmts.c (vectorizable_call): If slp_node != NULL,
1697 don't replace scalar calls with setting of their lhs to zero here.
1699 PR tree-optimization/51074
1700 * fold-const.c (vec_cst_ctor_to_array, fold_vec_perm): New functions.
1701 (fold_binary_loc): Handle VEC_EXTRACT_EVEN_EXPR,
1702 VEC_EXTRACT_ODD_EXPR, VEC_INTERLEAVE_HIGH_EXPR and
1703 VEC_INTERLEAVE_LOW_EXPR with VECTOR_CST or CONSTRUCTOR operands.
1704 (fold_ternary_loc): Handle VEC_PERM_EXPR with VECTOR_CST or
1705 CONSTRUCTOR operands.
1706 * tree-ssa-propagate.c (valid_gimple_rhs_p): Handle ternary
1708 * tree-vect-generic.c (lower_vec_perm): Mask sel_int elements
1709 to 0 .. 2 * elements - 1.
1711 2011-11-11 Richard Henderson <rth@redhat.com>
1713 * gimple-pretty-print.c (dump_gimple_omp_atomic_load): Dump needed.
1714 (dump_gimple_omp_atomic_store): Likewise.
1715 * optabs.c (can_atomic_exchange_p): New.
1716 * optabs.h (can_atomic_exchange_p): Declare.
1717 * omp-low.c (expand_omp_atomic_load): Implement.
1718 (expand_omp_atomic_store): Likewise.
1719 (expand_omp_atomic): Update for new arguments to load/store.
1721 2011-11-11 Eric Botcazou <ebotcazou@adacore.com>
1723 * cprop.c: Adjust outdated comments throughout.
1724 (hash_scan_set): Rename PAT parameter into SET.
1725 (cprop_pavloc): Rename into...
1726 (cprop_avloc): ...this.
1727 (cprop_absaltered): Rename into...
1728 (cprop_kill): ...this.
1729 (alloc_cprop_mem): Adjust for above renaming.
1730 (free_cprop_mem): Likewise.
1731 (compute_cprop_data): Likewise.
1732 (compute_local_properties): Rename TRANSP parameter into KILL and
1733 adjust throughout. Rework comments.
1734 (try_replace_reg): Fix long line.
1735 (cprop_jump): Likewise.
1736 (constprop_register): Fix prototype and take INSN last.
1737 (cprop_insn): Adjust calls to above function. Fix long lines.
1738 (bypass_block): Likewise.
1739 (one_cprop_pass): Likewise.
1741 2011-11-11 Uros Bizjak <ubizjak@gmail.com>
1743 * config/alpha/elf.h (ELF_ASCII_ESCAPES): Rename from ESCAPES.
1744 (ELF_STRING_LIMIT): Rename from STRING_LIMIT.
1746 2011-11-11 Jason Merrill <jason@redhat.com>
1748 * varasm.c (default_elf_asm_output_limited_string): #ifdef
1750 (default_elf_asm_output_ascii): Likewise.
1752 2011-11-11 Uros Bizjak <ubizjak@gmail.com>
1755 * config/i386/constraints.md (j): New address constraint.
1756 * config/i386/predicates.md (lea_address_operand): Redefine as
1758 * config/i386/i386.md (*lea_3_zext): Use "j" constraint for operand 1.
1759 (*lea_4_zext): Ditto.
1761 2011-11-11 David S. Miller <davem@davemloft.net>
1764 2011-11-05 David S. Miller <davem@davemloft.net>
1766 * config/sparc/sparc.md (UNSPEC_SHORT_LOAD): New unspec.
1767 (zero-extend_v8qi_vis, zero_extend_v4hi_vis): New expanders.
1768 (*zero_extend_v8qi_<P:mode>_insn,
1769 *zero_extend_v4hi_<P:mode>_insn): New insns.
1770 * config/sparc/sparc.c (vector_init_move_words,
1771 vector_init_prepare_elts, sparc_expand_vector_init_vis2,
1772 sparc_expand_vector_init_vis1): New functions.
1773 (vector_init_bshuffle): Rewrite to handle more cases and make use
1774 of locs[] array prepared by vector_init_prepare_elts.
1775 (vector_init_fpmerge, vector_init_faligndata): Delete.
1776 (sparc_expand_vector_init): Rewrite using new infrastructure.
1778 2011-11-11 Jakub Jelinek <jakub@redhat.com>
1780 * opts-common.c (generate_canonical_option): Free opt_text
1781 it it has been allocated here and not stored anywhere.
1783 * tree-vect-slp.c (vect_free_slp_tree): Also free SLP_TREE_CHILDREN
1785 (vect_create_new_slp_node): Don't allocate node before checking stmt
1787 (vect_free_oprnd_info): Remove FREE_DEF_STMTS argument, always
1788 free def_stmts vectors and additionally free oprnd_info.
1789 (vect_build_slp_tree): Adjust callers. Call it even if
1790 stop_recursion. If vect_create_new_slp_node or
1791 vect_build_slp_tree fails, properly handle freeing memory.
1792 If it succeeded, clear def_stmts in oprnd_info.
1794 2011-11-11 Dimitrios Apostolou <jimis@gmx.net>
1796 * defaults.h (ASM_OUTPUT_LABEL, ASM_OUTPUT_INTERNAL_LABEL): Expand on
1797 multiple lines for readability.
1798 * output.h (fprint_whex, fprint_w, fprint_ul, sprint_ul): Declare.
1799 (ASM_OUTPUT_LABELREF): Call fputs instead of asm_fprintf.
1800 * dwarf2asm.c (dw2_assemble_integer, dw2_asm_output_data,
1801 dw2_asm_data_uleb128, dw2_asm_delta_uleb128,
1802 dw2_asm_delta_sleb128): Convert fprintf calls to the new functions.
1803 * dwarf2out.c (dwarf2out_source_line): Call fprint_ul instead of
1805 * final.c (fprint_whex, fprint_w, fprint_ul, sprint_ul): New functions
1806 serving as fast replacements for fprintf integer to string conversions.
1807 (sprint_ul_rev): Internal helper for the above.
1808 (output_addr_const) <CONST_INT>: Don't use fprintf.
1809 * varasm.c: Fix some whitespace inconsistencies.
1810 (default_elf_asm_output_limited_string): New.
1811 (default_elf_asm_output_ascii): Likewise.
1812 (default_elf_internal_label): Likewise.
1813 * config/elfos.h (ASM_GENERATE_INTERNAL_LABEL): Use sprint_ul and
1814 stpcpy instead of sprintf.
1815 (TARGET_ASM_INTERNAL_LABEL): Define as default_elf_internal_label.
1816 (ELF_ASCII_ESCAPES, ELF_STRING_LIMIT): Rename from old ESCAPES and
1817 STRING_LIMIT macros.
1818 (ASM_OUTPUT_LIMITED_STRING, ASM_OUTPUT_ASCII): Call functions that
1819 provide the same functionality.
1820 * config/i386/i386.c (print_reg): Call fprint_ul instead of fprintf.
1822 2011-11-10 Andrew MacLeod <amacleod@redhat.com>
1824 * doc/extend.texi: Document __atomic_test_and_set and __atomic_clear.
1826 2011-11-10 Andrew MacLeod <amacleod@redhat.com>
1829 * builtins.c (expand_builtin_atomic_clear): New. Expand atomic_clear.
1830 (expand_builtin_atomic_test_and_set): New. Expand atomic test_and_set.
1831 (expand_builtin): Add cases for test_and_set and clear.
1832 * sync-builtins.def (BUILT_IN_ATOMIC_TEST_AND_SET): New.
1833 (BUILT_IN_ATOMIC_CLEAR): New.
1835 2011-11-10 Roberto Agostino Vitillo <ravitillo@lbl.gov>
1838 * dwarf2out.c (set_cur_line_info_table): Restore the last is_stmt
1839 value in the current line table.
1841 2011-11-10 Nathan Sidwell <nathan@acm.org>
1843 * gcov.c (struct function_info): Make src an index, not a pointer.
1844 (struct source_info): Remove index and next source fields.
1845 (fn_end): New static var.
1846 (sources_index): Remove.
1847 (sources): Now a pointer to an array, not a list.
1848 (n_sources, a_sources): New.
1849 (process_file): Adjust for changes to read_graph_file. Insert
1850 functions into source lists and check line numbers here.
1851 (generate_results): Only allocate lines for sources with
1852 contents. Adjust for source array.
1853 (release_structures): Likewise.
1854 (find_source): Return source index, adjust for source array.
1855 (read_graph_file): Return function list. Don't insert into source
1857 (read_count_file): Take list of functions.
1858 (solve_flow_graph): Reverse the arc lists here.
1859 (add_line_counts): Adjust for source array.
1861 2011-11-10 Jakub Jelinek <jakub@redhat.com>
1864 * tree-object-size.c (addr_object_size): Check TREE_CODE of
1865 MEM_REF's operand rather than code of the MEM_REF itself.
1867 2011-11-10 Andrew MacLeod <amacleod@redhat.com>
1869 PR rtl-optimization/51040
1870 * optabs.c (expand_atomic_fetch_op): Patchup code for NAND should
1871 be AND followed by NOT.
1872 * builtins.c (expand_builtin_atomic_fetch_op): Patchup code for NAND
1873 should be AND followed by NOT.
1875 2011-11-10 Jakub Jelinek <jakub@redhat.com>
1877 * vec.h (VEC_BASE): If base is at offset 0 in the structure,
1878 use &(P)->base even if P is NULL.
1880 PR rtl-optimization/51023
1881 * combine.c (simplify_comparison) <case SIGN_EXTEND>: Don't use
1882 val_signbit_known_clear_p for signed comparison narrowing
1883 optimization. Don't check for non-VOIDmode, use
1884 HWI_COMPUTABLE_MODE_P macro.
1885 <case ZERO_EXTEND>: Don't check for non-VOIDmode.
1886 Optimize even when const_op is equal to GET_MODE_MASK (mode),
1887 don't optimize if const_op is negative.
1889 2011-11-10 Richard Guenther <rguenther@suse.de>
1891 PR tree-optimization/51042
1892 * tree-ssa-pre.c (phi_translate_1): Avoid recursing on
1893 self-referential expressions. Refactor code to avoid duplication.
1895 2011-11-10 Richard Guenther <rguenther@suse.de>
1897 PR tree-optimization/51070
1898 * tree-loop-distribution.c (generate_builtin): Do not replace
1899 the loop with a builtin if the partition contains statements which
1900 results are used outside of the loop.
1901 (pass_loop_distribution): Verify and collect.
1903 2011-11-10 Richard Guenther <rguenther@suse.de>
1905 PR tree-optimization/51030
1906 * tree-ssa-phiopt.c (jump_function_from_stmt): New function.
1907 (value_replacement): Use it to handle trivial non-empty
1908 intermediate blocks.
1910 2011-11-10 Richard Guenther <rguenther@suse.de>
1913 * gimple.c (gimple_has_side_effects): Remove checking code
1914 that doesn't belong here.
1916 2011-11-10 Ira Rosen <ira.rosen@linaro.org>
1918 PR tree-optimization/51058
1919 * tree-vect-slp.c (vect_get_constant_vectors): Handle CALL_EXPR.
1921 2011-11-10 Jakub Jelinek <jakub@redhat.com>
1923 PR tree-optimization/51000
1924 * tree-vect-patterns.c (vect_recog_bool_pattern): If adding
1925 a pattern stmt for a bool store, adjust DR_STMT too.
1926 Don't handle bool conversions to single bit precision lhs.
1927 * tree-vect-stmts.c (vect_remove_stores): If next is a pattern
1928 stmt, remove its related stmt and free its stmt_vinfo.
1929 (free_stmt_vec_info): Free also pattern stmt's vinfo and
1930 pattern def stmt's vinfo.
1931 * tree-vect-loop.c (destroy_loop_vec_info): Don't try to
1932 free pattern stmt's vinfo here.
1933 (vect_transform_loop): When calling vect_remove_stores,
1934 do gsi_next first and don't call gsi_remove. If not strided
1935 store, free stmt vinfo for gsi_stmt (si) rather than stmt.
1937 2011-11-09 Jakub Jelinek <jakub@redhat.com>
1940 * config/i386/i386.c (expand_vec_perm_interleave2): If d->vmode is
1941 V4SImode, !TARGET_SSE2 and punpck[lh]* is needed, change dremap.vmode
1944 * function.h (requires_stack_frame_p): New prototype.
1945 * function.c (requires_stack_frame_p): No longer static.
1946 * config/i386/i386.c (ix86_finalize_stack_realign_flags): If
1947 stack_realign_fp was just a conservative guess for a function
1948 which doesn't use sp/fp/argp at all, clear frame_pointer_needed
1949 and stack realignment.
1951 2011-11-09 Paolo Carlini <paolo.carlini@oracle.com>
1953 PR preprocessor/51061
1954 * incpath.c (merge_include_chains): Make sure to not pass null
1955 pointers to free_path.
1957 2011-11-09 Paolo Carlini <paolo.carlini@oracle.com>
1959 * doc/invoke.texi ([Option Summary, C++ Language Options]):
1960 Add -Wzero-as-null-pointer-constant.
1962 2011-11-07 Matthew Gretton-Dann <matthew.gretton-dann@arm.com>
1964 * config/arm/arm-cores.def: Add -mcpu=cortex-a7.
1965 * config/arm/arm-tables.opt: Regenerate.
1966 * config/arm/arm-tune.md: Likewise.
1967 * config/arm/bpabi.h (BE8_LINK_SPEC): Add Cortex A-7.
1968 * doc/invoke.texi: Document -mcpu=cortex-a7.
1970 2011-11-09 Richard Guenther <rguenther@suse.de>
1973 2011-11-09 Richard Guenther <rguenther@suse.de>
1975 * tree-cfg.c (verify_gimple_call): Verify that
1976 gimple_call_cannot_inline_p is returning a conservative
1977 correct result according to gimple_check_call_matching_types.
1979 2011-11-09 Richard Guenther <rguenther@suse.de>
1981 PR tree-optimization/51039
1982 * tree-cfg.c (verify_gimple_call): Verify that
1983 gimple_call_cannot_inline_p is returning a conservative
1984 correct result according to gimple_check_call_matching_types.
1985 * ipa-inline-analysis.c (estimate_function_body_sizes): Remove
1986 code dealing with un-inlinablility.
1987 * gimple-streamer-in.c (input_gimple_stmt): Update the
1990 2011-11-09 Richard Guenther <rguenther@suse.de>
1992 PR tree-optimization/51039
1993 * tree-inline.c (setup_one_parameter): Always perform a
1994 valid gimple type change.
1995 (declare_return_variable): Likewise.
1997 2011-11-09 Jakub Jelinek <jakub@redhat.com>
1999 * config/rs6000/vector.md (vcondv4sfv4si, vcondv4siv4sf,
2000 vconduv4sfv4si): New patterns.
2001 * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Handle
2002 different dest_mode from comparison mode.
2004 2011-11-09 Richard Guenther <rguenther@suse.de>
2006 * gimple-fold.c (canonicalize_constructor_val): Mark
2007 address-taken variables addressable.
2009 2011-11-09 Ira Rosen <ira.rosen@linaro.org>
2011 PR tree-optimization/51015
2012 * tree-vect-loop.c (vect_determine_vectorization_factor): Expect
2013 vectype to be set for pattern def stmts.
2015 2011-11-09 Alan Modra <amodra@gmail.com>
2017 * function.c (bb_active_p): Delete.
2018 (dup_block_and_redirect, active_insn_between): New functions.
2019 (convert_jumps_to_returns, emit_return_for_exit): New functions,
2021 (thread_prologue_and_epilogue_insns): ..here. Delete
2022 shadowing variables. Don't do prologue register clobber tests
2023 when shrink wrapping already failed. Delete all last_bb_active
2024 code. Instead compute tail block candidates for duplicating
2025 exit path. Remove these from antic set. Duplicate tails when
2026 reached from both blocks needing a prologue/epilogue and
2027 blocks not needing such.
2028 * ifcvt.c (dead_or_predicable): Test both flag_shrink_wrap and
2030 * bb-reorder.c (get_uncond_jump_length): Make global.
2031 * bb-reorder.h (get_uncond_jump_length): Declare.
2032 * cfgrtl.c (rtl_create_basic_block): Comment typo fix.
2033 (rtl_split_edge): Likewise. Warning fix.
2034 (rtl_duplicate_bb): New function.
2035 (rtl_cfg_hooks): Enable can_duplicate_block_p and duplicate_block.
2036 * Makefile.in (function.o): Update dependencies.
2038 2011-11-08 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
2039 Georg-Johann Lay <avr@gjlay.de>
2041 * doc/tm.texi.in (MODE_CODE_BASE_REG_CLASS): Add address space
2043 (REGNO_MODE_CODE_OK_FOR_BASE_P): Likewise.
2044 * doc/tm.texi: Regenerate.
2046 * config/cris/cris.h (MODE_CODE_BASE_REG_CLASS): Add address
2048 (REGNO_MODE_CODE_OK_FOR_BASE_P): Likewise.
2049 * config/bfin/bfin.h (MODE_CODE_BASE_REG_CLASS): Likewise.
2050 (REGNO_MODE_CODE_OK_FOR_BASE_P): Likewise.
2051 * config/avr/avr.h (MODE_CODE_BASE_REG_CLASS): Add address space
2053 (REGNO_MODE_CODE_OK_FOR_BASE_P): Ditto.
2054 * config/avr/avr-protos.h (avr_mode_code_base_reg_class): Ditto.
2055 (avr_regno_mode_code_ok_for_base_p): Ditto.
2056 * config/avr/avr.c (avr_mode_code_base_reg_class): Ditto.
2057 (avr_regno_mode_code_ok_for_base_p): Ditto.
2058 (avr_reg_ok_for_addr_p): Pass AS down to
2059 avr_regno_mode_code_ok_for_base_p.
2061 * addresses.h (base_reg_class): Add address space argument.
2062 Pass to MODE_CODE_BASE_REG_CLASS.
2063 (ok_for_base_p_1): Add address space argument. Pass to
2064 REGNO_MODE_CODE_OK_FOR_BASE_P.
2065 (regno_ok_for_base_p): Add address space argument. Pass to
2068 * regrename.c (scan_rtx_address): Add address space argument.
2069 Pass address space to regno_ok_for_base_p and base_reg_class.
2070 Update recursive calls.
2071 (scan_rtx): Pass address space to scan_rtx_address.
2072 (build_def_use): Likewise.
2073 * regcprop.c (replace_oldest_value_addr): Add address space
2074 argument. Pass to regno_ok_for_base_p and base_reg_class.
2075 Update recursive calls.
2076 (replace_oldest_value_mem): Pass address space to
2077 replace_oldest_value_addr.
2078 (copyprop_hardreg_forward_1): Likewise.
2080 * reload.c (find_reloads_address_1): Add address space argument.
2081 Pass address space to base_reg_class and regno_ok_for_base_p.
2082 Update recursive calls.
2083 (find_reloads_address): Pass address space to base_reg_class,
2084 regno_ok_for_base_p, and find_reloads_address_1.
2085 (find_reloads): Pass address space to base_reg_class.
2086 (find_reloads_subreg_address): Likewise.
2088 * ira-costs.c (record_reg_classes): Update calls to base_reg_class.
2089 (ok_for_base_p_nonstrict): Add address space argument. Pass to
2091 (record_address_regs): Add address space argument. Pass to
2092 base_reg_class and ok_for_base_p_nonstrict. Update recursive calls.
2093 (record_operand_costs): Pass address space to record_address_regs.
2094 (scan_one_insn): Likewise.
2096 * caller-save.c (init_caller_save): Update call to base_reg_class.
2097 * ira-conflicts.c (ira_build_conflicts): Likewise.
2098 * reload1.c (maybe_fix_stack_asms): Likewise.
2100 2011-11-08 Michael Matz <matz@suse.de>
2102 * gengtype.c (write_field_root): Avoid out-of-scope access of newv.
2104 * tree-stdarg.c (execute_optimize_stdarg): Accept clobbers.
2106 * tree.h (TREE_CLOBBER_P): New macro.
2107 * gimple.h (gimple_clobber_p): New inline function.
2108 * gimplify.c (gimplify_bind_expr): Add clobbers for all variables
2109 that go out of scope and live in memory.
2110 * tree-ssa-operands.c (get_expr_operands): Transfer volatility also
2112 * cfgexpand.c (decl_to_stack_part): New static variable.
2113 (add_stack_var): Allocate it, and remember mapping.
2114 (fini_vars_expansion): Deallocate it.
2115 (stack_var_conflict_p): Add early outs.
2116 (visit_op, visit_conflict, add_scope_conflicts_1,
2117 add_scope_conflicts): New static functions.
2118 (expand_used_vars_for_block): Don't call add_stack_var_conflict, tidy.
2119 (expand_used_vars): Add scope conflicts.
2120 (expand_gimple_stmt_1): Expand clobbers to nothing.
2121 (expand_debug_expr): Ditto.
2123 * tree-pretty-print.c (dump_generic_node): Dump clobbers nicely.
2124 * tree-ssa-live.c (remove_unused_locals): Remove clobbers that
2125 refer to otherwise unused locals.
2126 * tree-sra.c (build_accesses_from_assign): Ignore clobbers.
2127 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Clobbers of
2128 SSA names aren't necessary.
2129 (propagate_necessity): Accept and ignore constructors on the rhs, tidy.
2130 * gimple.c (walk_gimple_op): Accept constructors like mem_rhs.
2131 * tree-ssa-structalias.c (find_func_aliases): Clobbers don't store
2133 * tree-ssa-sccvn.c (vn_reference_lookup_3): Ditto, in particular they
2134 don't zero-initialize something.
2135 * tree-ssa-phiopt.c (cond_if_else_store_replacement_1): Ignore
2136 clobber RHS, we don't want PHI nodes with those.
2138 2011-11-08 Jakub Jelinek <jakub@redhat.com>
2140 * config/i386/i386.c (ix86_preferred_simd_mode): Even for TARGET_AVX
2141 if not TARGET_PREFER_AVX128 return 32-byte integer vectors.
2143 2011-11-08 Richard Guenther <rguenther@suse.de>
2145 PR tree-optimization/51012
2146 * ipa-prop.c (update_indirect_edges_after_inlining): Fixup
2147 non-inlinable state.
2148 * cgraph.c (cgraph_make_edge_direct): Likewise.
2150 2011-11-08 Eric Botcazou <ebotcazou@adacore.com>
2152 PR rtl-optimization/47698
2153 * ifcvt.c (noce_operand_ok): Move around comment.
2155 2011-11-08 Richard Guenther <rguenther@suse.de>
2158 * lto-opts.c (append_to_collect_gcc_options): Split out from...
2159 (lto_write_options): ... here. Prepend frontend specific flags.
2161 2011-11-08 Jakub Jelinek <jakub@redhat.com>
2163 * config/i386/sse.md (*avx_vperm2f128<mode>3_nozero): Emit mask
2164 0x12 and 0x20 as vinsert[fi]128 instead of vperm2[fi]128.
2166 * tree-vect-stmts.c (vectorizable_call): Add SLP_NODE argument.
2167 Handle vectorization of SLP calls.
2168 (vect_analyze_stmt): Adjust caller, add call to it for SLP too.
2169 (vect_transform_stmt): Adjust vectorizable_call caller, remove
2171 * tree-vect-slp.c (vect_get_and_check_slp_defs): For calls start
2173 (vect_build_slp_tree): Allow CALL_EXPR.
2175 2011-11-08 Richard Guenther <rguenther@suse.de>
2177 * gimple-fold.c (canonicalize_constructor_val): Make sure
2178 we have referenced vars setup before adding to them.
2180 2011-11-07 Richard Henderson <rth@redhat.com>
2181 Aldy Hernandez <aldyh@redhat.com>
2182 Andrew MacLeod <amacleod@redhat.com>
2183 Torvald Riegel <triegel@redhat.com>
2185 Merged from transactional-memory.
2187 * gtm-builtins.def: New file.
2188 * trans-mem.c: New file.
2189 * trans-mem.h: New file.
2191 * opts.c (finish_options): Error out when using -flto and -fgnu-tm.
2193 * config/i386/i386.c: Define TARGET_VECTORIZE* transactional variants.
2194 (ix86_handle_tm_regparm_attribute, struct bdesc_tm,
2195 ix86_builtin_tm_load, ix86_builtin_tm_store,
2196 ix86_init_tm_builtins): New.
2197 (ix86_init_builtins): Initialize TM builtins.
2198 (struct ix86_attribute_table): Add "*tm regparm".
2199 * config/i386/i386-builtin-types.def (PV2SI): Define.
2201 Define V2SI_FTYPE_PCV2SI.
2202 Define V4SF_FTYPE_PCV4SF.
2203 Define V8SF_FTYPE_PCV8SF.
2204 Define VOID_PV2SI_V2SI.
2206 * doc/invoke.texi (C Dialect Options): Document -fgnu-tm and
2207 tm-max-aggregate-size.
2208 * doc/tm.texi.in: Add TARGET_VECTORIZE_BUILTIN_TM_LOAD and
2209 TARGET_VECTORIZE_BUILTIN_TM_STORE hooks.
2210 * doc/tm.texi: Regenerate.
2212 * attribs.c (apply_tm_attr): New.
2213 (init_attributes): Allow '*' prefix for overrides.
2214 (register_attribute): Likewise.
2215 * builtin-attrs.def (ATTR_TM_TMPURE, ATTR_TM_REGPARM): New.
2216 (ATTR_TM_NOTHROW_LIST, ATTR_TM_TMPURE_NOTHROW_LIST,
2217 ATTR_TM_PURE_TMPURE_NOTHROW_LIST, ATTR_TM_NORETURN_NOTHROW_LIST,
2218 ATTR_TM_CONST_NOTHROW_LIST, ATTR_TMPURE_MALLOC_NOTHROW_LIST,
2219 ATTR_TMPURE_NOTHROW_LIST): New.
2220 * builtin-types.def (BT_FN_I[1248]_VPTR, BT_FN_FLOAT_VPTR,
2221 BT_FN_DOUBLE_VPTR, BT_FN_LDOUBLE_VPTR, BT_FN_VOID_VPTR_I[1248],
2222 BT_FN_VOID_VPTR_FLOAT, BT_FN_VOID_VPTR_DOUBLE,
2223 BT_FN_VOID_VPTR_LDOUBLE, BT_FN_VOID_VPTR_SIZE): New.
2224 * builtins.def: Include gtm-builtins.def. Add comments regarding
2225 transactional memory synchronization.
2226 (DEF_TM_BUILTIN): New.
2227 * c-parser.c (struct c_parser): Add in_transaction.
2228 (c_parser_transaction, c_parser_transaction_expression,
2229 c_parser_transaction_cancel, c_parser_transaction_attributes): New.
2230 (c_parser_attribute_any_word): Split out from c_parser_attributes.
2231 (c_parser_statement_after_labels): Handle RID_TRANSACTION*.
2232 (c_parser_unary_expression): Same.
2233 * c-tree.h (c_finish_transaction): Declare.
2234 * c-typeck.c (c_finish_transaction): New.
2235 (build_function_call_vec): Call tm_malloc_replacement.
2236 * calls.c (is_tm_builtin): New.
2237 (flags_from_decl_or_type): Add ECF_TM_BUILTIN and ECF_TM_PURE.
2238 * cfgbuild.c (make_edges): Add edges for REG_TM notes.
2239 * cfgexpand.c (expand_call_stmt): Call
2240 mark_transaction_restart_calls.
2241 (gimple_expand_cfg): Free the tm_restart map.
2242 (mark_transaction_restart_calls): New.
2243 * cfgrtl.c (purge_dead_edges): Look for REG_TM notes.
2244 * cgraph.c (dump_cgraph_node): Handle tm_clone.
2245 * cgraph.h (struct cgraph_node): Add tm_clone field.
2246 (decl_is_tm_clone): New.
2247 (struct cgraph_local_info): Add tm_may_enter_irr.
2248 (cgraph_copy_node_for_versioning): Declare.
2249 * cgraphunit.c (cgraph_copy_node_for_versioning): Export;
2250 copy analyzed from old version.
2251 * combine.c (distribute_notes): Handle REG_TM notes.
2252 * common.opt: Add -fgnu-tm.
2253 * crtstuff.c (__TMC_LIST__, __TMC_END__): New.
2254 (__do_global_dtors_aux): Deregister clone table.
2255 (frame_dummy): Register clone table.
2256 * emit-rtl.c (try_split): Handle REG_TM. Early return if no function
2258 * gimple-low.c (lower_stmt): Handle GIMPLE_EH_ELSE and
2260 (gimple_stmt_may_fallthru): Handle GIMPLE_EH_ELSE.
2261 * gimple-pretty-print.c: Include trans-mem.h.
2262 (dump_gimple_fmt): Add %x.
2263 (dump_gimple_call): Dump arguments for calls to _ITM_beginTransaction.
2264 (dump_gimple_eh_else, dump_gimple_transaction): New.
2265 (dump_gimple_stmt): Handle GIMPLE_EH_ELSE and GIMPLE_TRANSACTION.
2266 * gimple.c (gimple_build_eh_else, gimple_build_transaction): New.
2267 (walk_gimple_seq): Honor removed_stmt. Document usage of removed_stmt
2269 (walk_gimple_op): Handle GIMPLE_TRANSACTION.
2270 (walk_gimple_stmt): Initialize and honor removed_stmt.
2271 Handle GIMPLE_EH_ELSE and GIMPLE_TRANSACTION.
2272 (gimple_copy): Handle GIMPLE_EH_ELSE and GIMPLE_TRANSACTION.
2273 * gimple.def (GIMPLE_TRANSACTION, GIMPLE_EH_ELSE): New.
2274 * gimple.h (struct gimple_statement_eh_else,
2275 gimple_statement_transaction, GTMA_*): New.
2276 (gimple_statement_d): Add gimple_statement_eh_else and
2278 (gimple_build_eh_else, gimple_build_transaction,
2279 gimple_fold_call, diagnose_tm_safe_errors): Declare.
2280 (get_call_expr_in): Remove prototype.
2281 (gimple_has_substatements): Add GIMPLE_EH_ELSE and GIMPLE_TRANSACTION.
2282 (gimple_eh_else_n_body, gimple_eh_else_e_body,
2283 gimple_eh_else_set_n_body, gimple_eh_else_set_e_body,
2284 gimple_transaction_body, gimple_transaction_label,
2285 gimple_transaction_label_ptr, gimple_transaction_subcode,
2286 gimple_transaction_set_body, gimple_transaction_set_label,
2287 gimple_transaction_set_subcode): New.
2288 (struct walk_stmt_info): Use BOOL_BITFIELD; add removed_stmt.
2289 * gimplify.c (create_tmp_var_name): Use clean_symbol_name.
2290 (voidify_wrapper_expr): Handle TRANSACTION_EXPR.
2291 (gimplify_transaction): New.
2292 (gimplify_expr): Handle TRANSACTION_EXPR.
2293 * gsstruct.def (GSS_EH_ELSE, GSS_TRANSACTION): New.
2294 * ipa-inline.c (can_inline_edge_p): Do not inline TM safe calling
2296 * Makefile.in: Add trans-mem.o and dependencies.
2297 (BUILTINS_DEF): Add gtm-builtins.def.
2298 (gimple-pretty-print.o): Depend on TRANS_MEM_H.
2299 (GTFILES): Add trans-mem.c.
2300 * omp-low.c (WALK_SUBSTMTS): Add GIMPLE_TRANSACTION.
2301 * output.h (record_tm_clone_pair, finish_tm_clone_pairs,
2302 get_tm_clone_pair): Declare.
2303 * params.def (PARAM_TM_MAX_AGGREGATE_SIZE): New.
2304 * passes.c (init_optimization_passes): Place transactional memory
2306 * print-tree.c (print_node): Dump tm-clone.
2307 * recog.c (peep2_attempt): Handle REG_TM.
2308 * reg-notes.def (TM): New.
2309 * rtlanal.c (alloc_reg_note): Handle REG_TM.
2310 * target.def (builtin_tm_load, builtin_tm_store): New.
2311 * targhooks.c (default_builtin_tm_load_store): New.
2312 * targhooks.h (default_builtin_tm_load_store): Declare.
2313 * timevar.def (TV_TRANS_MEM): New.
2314 * toplev.c (compile_file): Call finish_tm_clone_pairs.
2315 * tree-cfg.c (make_edges): Handle GIMPLE_TRANSACTION.
2316 (cleanup_dead_labels): Handle GIMPLE_TRANSACTION. Avoid unnecessary
2317 writes into the statements to update labels.
2318 (is_ctrl_altering_stmt): Add TM ending statements. Handle
2320 (verify_gimple_transaction): New.
2321 (verify_gimple_stmt): Handle GIMPLE_TRANSACTION.
2322 (verify_gimple_in_seq_2): Handle GIMPLE_EH_ELSE and GIMPLE_TRANSACTION.
2323 (gimple_redirect_edge_and_branch): Handle TM_TRANSACTION.
2324 (dump_function_to_file): Display [tm-clone] if applicable.
2325 * tree-eh.c (struct_ptr_eq): Make inline and move to tree.h.
2326 (struct_ptr_hash): Same.
2327 (collect_finally_tree): Handle GIMPLE_EH_ELSE.
2328 (replace_goto_queue_1): Likewise.
2330 (honor_protect_cleanup_actions): Handle GIMPLE_EH_ELSE.
2331 (lower_try_finally_nofallthru): Likewise.
2332 (lower_try_finally_onedest): Likewise.
2333 (lower_try_finally_copy): Likewise.
2334 (lower_try_finally_switch): Likewise.
2335 (lower_try_finally): Likewise.
2336 (decide_copy_try_finally): Likewise.
2337 (lower_eh_constructs_2): Likewise.
2338 (refactor_eh_r): Likewise.
2339 * tree-flow.h (struct gimple_df): Add tm_restart field.
2340 Define tm_restart_node.
2341 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_TRANSACTION.
2342 (estimate_num_insns): Likewise.
2343 (init_inline_once): Init tm_cost.
2344 * tree-inline.h (struct eni_weights_d): Add tm_cost.
2345 * tree-pass.h (pass_diagnose_tm_blocks, pass_lower_tm, pass_tm_init,
2346 pass_tm_mark, pass_tm_memopt, pass_tm_edges, pass_ipa_tm): Declare.
2347 * tree-pretty-print.c (dump_generic_node): Handle TRANSACTION_EXPR.
2348 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Handle
2349 BUILT_IN_TM_MEMSET, BUILT_IN_TM_MEMCPY, BUILT_IN_TM_MEMMOVE.
2350 Add support for TM vector loads. Add support for TM logging builtins.
2351 (call_may_clobber_ref_p_1): Add support for vector stores.
2352 * tree-ssa-structalias.c (find_func_aliases): Add support for TM
2353 vector stores and loads. Handle BUILT_IN_TM_MEMSET,
2354 BUILT_IN_TM_MEMCPY, BUILT_IN_TM_MEMMOVE.
2355 * tree.c (strip_invariant_refs): Moved from gimple.c to here.
2356 (local_define_builtin): Handle ECF_TM_PURE.
2357 (build_common_builtin_nodes): Set __builtin_eh_pointer to ECF_TM_PURE.
2358 * tree.def (TRANSACTION_EXPR): New.
2359 * tree.h (strip_invariant_refs): Moved from gimple.h to here.
2360 (TRANSACTION_EXPR_BODY, TRANSACTION_EXPR_CHECK,
2361 TRANSACTION_EXPR_OUTER, TRANSACTION_EXPR_RELAXED,
2362 BUILTIN_TM_LOAD_STORE_P, BUILTIN_TM_LOAD_P, BUILTIN_TM_STORE_P,
2363 CASE_BUILT_IN_TM_LOAD, CASE_BUILT_IN_TM_STORE): New.
2364 (ECF_TM_PURE, ECF_TM_BUILTIN): New.
2365 (struct tree_function_decl): Add tm_clone_flag.
2366 (struct_ptr_eq, struct_ptr_hash): New.
2367 (apply_tm_attr): Declare.
2368 (is_tm_safe_or_pure): New.
2369 (build_tm_abort_call, is_tm_safe, is_tm_pure,
2370 is_tm_may_cancel_outer, is_tm_ending_fndecl, record_tm_replacement,
2371 tm_malloc_replacement): Declare.
2372 * varasm.c (tm_clone_hash): New.
2373 (record_tm_clone_pair, finish_tm_clone_pairs, get_tm_clone_pair,
2374 dump_tm_clone_to_vec, dump_tm_clone_pairs, tm_alias_pair_cmp): New.
2375 (struct tm_alias_pair): New. Declare VEC types for object.
2377 2011-11-07 Richard Henderson <rth@redhat.com>
2379 * optabs.h (OTI_sync_compare_and_swap, OTI_sync_lock_test_and_set,
2380 OTI_sync_old_add, OTI_sync_old_sub, OTI_sync_old_ior,
2381 OTI_sync_old_and, OTI_sync_old_xor, OTI_sync_old_nand,
2382 OTI_sync_new_add, OTI_sync_new_sub, OTI_sync_new_ior,
2383 OTI_sync_new_and, OTI_sync_new_xor, OTI_sync_new_nand): Move and
2384 rename from the direct_optab_index enum.
2385 (sync_compare_and_swap_optab, sync_lock_test_and_set_optab,
2386 sync_old_add_optab, sync_old_sub_optab, sync_old_ior_optab,
2387 sync_old_and_optab, sync_old_xor_optab, sync_old_nand_optab,
2388 sync_new_add_optab, sync_new_sub_optab, sync_new_ior_optab,
2389 sync_new_and_optab, sync_new_xor_optab, sync_new_nand_optab): Read
2390 from the optab_table, not the direct_optab_table.
2391 (init_sync_libfuncs): Declare.
2392 (can_compare_and_swap_p): Update parameters.
2393 * optabs.c (init_sync_libfuncs_1, init_sync_libfuncs): New.
2394 (can_compare_and_swap_p): Add allow_libcall parameter; if true,
2395 test for the legacy compare-and-swap libcall.
2396 (expand_atomic_exchange): Use the legacy test-and-set libcall.
2397 (expand_atomic_compare_and_swap): Use the legacy CAS libcall.
2398 (struct atomic_op_functions): Update for optab type changes.
2399 (maybe_emit_op): Likewise.
2400 (expand_atomic_fetch_op): Use the legacy fetch-op libcalls.
2401 * builtins.c (fold_builtin_atomic_always_lock_free): Update call
2402 to can_compare_and_swap_p.
2403 * omp-low.c (expand_omp_atomic_fetch_op): Likewise.
2404 (expand_omp_atomic_pipeline): Likewise.
2405 * genopinit.c (optabs): Make sync_old_*_optab, sync_new_*_optab,
2406 sync_compare_and_swap_optab, sync_lock_test_and_set_optab regular
2408 * doc/md.texi (sync_compare_and_swap): Update docs for libcalls.
2410 2011-11-07 Jakub Jelinek <jakub@redhat.com>
2412 * config/i386/i386-bultin-types.def (V8SI_FTYPE_V4DF_V4DF): Add.
2413 * config/i386/i386.c (enum ix86_builtins): Add
2414 IX86_BUILTIN_VEC_PACK_SFIX256.
2415 (bdesc_args): Add __builtin_ia32_vec_pack_sfix256.
2416 (ix86_expand_args_builtin): Handle V8SI_FTYPE_V4DF_V4DF.
2417 (ix86_builtin_vectorized_function): Also vectorize lrint using
2418 256-bit vectors for -mavx.
2420 2011-11-07 Anatoly Sokolov <aesok@post.ru>
2422 * config/cris/constraints.md: New file.
2423 * config/cris/cris.h (REG_CLASS_FROM_LETTER, CONSTRAINT_LEN,
2424 CRIS_CONST_OK_FOR_LETTER_P, CONST_OK_FOR_CONSTRAINT_P,
2425 CONST_DOUBLE_OK_FOR_LETTER_P, EXTRA_MEMORY_CONSTRAINT,
2426 EXTRA_CONSTRAINT, EXTRA_CONSTRAINT_Q, EXTRA_CONSTRAINT_R,
2427 EXTRA_CONSTRAINT_T, EXTRA_CONSTRAINT_S, EXTRA_CONSTRAINT_U): Remove.
2428 * config/cris/cris.c: Incule "tm-constrs.h".
2429 (cris_print_operand): Use satisfies_constraint_O.
2430 (cris_normal_notice_update_cc, cris_rtx_costs): Use
2431 satisfies_constraint_I.
2432 (cris_address_cost): Use satisfies_constraint_L.
2433 * config/cris/cris.md: Include "constraints.md".
2434 (*mov_side<mode>, *mov_sidesisf, *mov_side<mode>_mem,
2435 *mov_sidesisf_mem, *clear_side<mode>, *ext_sideqihi,
2436 *ext_side<mode>si, *op_side<mode>, *op_swap_side<mode>,
2437 *extopqihi_side, *extop<mode>si_side, *extopqihi_swap_side,
2438 *extop<mode>si_swap_side): Use satisfies_constraint_N and
2439 satisfies_constraint_J.
2440 (moversideqi movemsideqi mover2side peephole2): Use
2441 satisfies_constraint_N and satisfies_constraint_J.
2442 (andu peephole2): Use satisfies_constraint_I and
2443 satisfies_constraint_O.
2445 2011-11-07 Uros Bizjak <ubizjak@gmail.com>
2447 * config/i386/i386.c (ix86_builtin_vectorized_function): Handle
2448 BUILT_IN_IRINT, BUILT_IN_IRINTF, BUILT_IN_LLRINT and BUILT_IN_LLRINTF.
2450 2011-11-07 Andrew MacLeod <amacleod@redhat.com>
2452 * optabs.c (get_atomic_op_for_code): Fill in optab table at runtime so
2453 SWITCHABLE_TARGET can change the values during compilation.
2454 (expand_atomic_fetch_op): Handle parameter change ripples for
2455 get_atomic_op_for_code call.
2457 2011-11-07 Andrew MacLeod <amacleod@redhat.com>
2459 * doc/extend.texi: Docuemnt behaviour change for __atomic_exchange and
2461 * optabs.c (expand_atomic_exchange): Expand to __sync_lock_test_and_set
2462 only when originated from that builtin.
2463 (expand_atomic_store): Expand to __sync_lock_release when originated
2465 * builtins.c (expand_builtin_sync_lock_test_and_set): Add flag that
2466 expand_atomic_exchange call originated from here.
2467 (expand_builtin_sync_lock_release): Add flag that expand_atomic_store
2468 call originated from here.
2469 (expand_builtin_atomic_exchange): Add origination flag.
2470 (expand_builtin_atomic_store): Add origination flag.
2471 * expr.h (expand_atomic_exchange, expand_atomic_store): Add boolean
2472 parameters to indicate implementation fall back options.
2474 2011-11-07 Georg-Johann Lay <avr@gjlay.de>
2476 * config/avr/avr.c (output_reload_in_const): Can handle CONSTANT_P
2477 now, not only CONST_INT and CONST_DOUBLE.
2478 (output_movqi): Use output_reload_in_const.
2479 (output_reload_inhi): Ditto.
2480 (output_reload_insisf): Move assertion to output_reload_in_const.
2481 (avr_out_reload_inpsi): Ditto.
2483 2011-11-07 Nathan Sidwell <nathan@acm.org>
2485 * gcov.c (object_summary): Replace with ...
2486 (object_runs): ... this.
2487 (process_file): Remove functions with no data.
2488 (generate_results): Ignore files with no lines.
2489 (release_function): New helper, broken out of ...
2490 (release_structures): ... here. Use it.
2491 (read_count_file): Adjust for new data file format.
2492 (output_lines): Use object_runs.
2493 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Obsolete.
2494 (struct gcov_ctr_info): Move definition.
2495 (struct gcov_fn_info): Add key field, use gcov_ctr_info for
2497 (struct gcov_info): Add merge function array, remove mask and
2498 counts. Trailing array is array of pointers to function info.
2499 * coverage.c (struct function_list): Replace counter numbers with
2500 counter arrays. Add fndecl field. GTYify.
2501 (counts_entry): Remove chain workspace.
2502 (functions_head): GTYify.
2503 (prg_n_ctrs): Remove.
2505 (tree_ctr_tables): Remove.
2506 (read_counts_file): Cope with blank entries and expect program
2507 summaries before functions. Don't warn on missing entries.
2508 (coverage_counter_alloc): Allocate individual function arrays.
2509 (tree_coverage_counter_ref, tree_coverage_counter_addr): Adjust
2510 for individual function arrays.
2511 (coverage_end_function): GTYify function list object. Finalize
2512 function's counter arrays.
2513 (build_var): New. Create a counter-related variable with
2514 appropriate linkage.
2515 (build_fn_info_type): Adjust for new runtime structure.
2516 (build_fn_info_value): Rename to ...
2517 (build_fn_info): ... here. Build new format data.
2518 (build_ctr_info_type, build_ctr_info_value): Remove.
2519 (build_info_type): New. Build new format data structure.
2520 (build_info): Adjust for new format data.
2521 (create_coverage): Likewise.
2522 * gcov-dump.c (tag_function): Recognize placeholders.
2524 2011-11-07 Georg-Johann Lay <avr@gjlay.de>
2526 * config/avr/constraints.md (Cm2): New constraint for int -2.
2527 * config/avr/avr.md (addqi3): Use it. New alternatives for +/-2.
2528 (*negqihi2): New insn.
2530 2011-11-07 H.J. Lu <hongjiu.lu@intel.com>
2532 * dwarf2cfi.c (dwarf2out_frame_debug_expr): Check
2533 HARD_FRAME_POINTER_REGNUM instead of hard_frame_pointer_rtx in Rule 18.
2535 2011-11-07 Jakub Jelinek <jakub@redhat.com>
2537 * config/i386/i386.c (ix86_expand_builtin): If gather mask
2538 argument is known to have all high bits set, pass pc_rtx as
2539 second argument to the expander instead of op0.
2540 * config/i386/sse.md (*avx2_gathersi<mode>_2,
2541 *avx2_gatherdi<mode>_2): New patterns.
2542 * config/i386/avx2intrin.h (_mm256_i32gather_pd,
2543 _mm256_i64gather_pd, _mm256_i32gather_ps): Set mask using
2544 _mm256_cmp_pd with zero vector arguments and _CMP_EQ_OQ instead of
2547 PR tree-optimization/50789
2548 * tree-vect-stmts.c (process_use): Add force argument, avoid
2549 exist_non_indexing_operands_for_use_p check if true.
2550 (vect_mark_stmts_to_be_vectorized): Adjust callers. Handle
2551 STMT_VINFO_GATHER_P.
2552 (gen_perm_mask): New function.
2553 (perm_mask_for_reverse): Use it.
2554 (reverse_vec_element): Rename to...
2555 (permute_vec_elements): ... this. Add Y and MASK_VEC arguments,
2556 generalize for any permutations.
2557 (vectorizable_load): Adjust caller. Handle STMT_VINFO_GATHER_P.
2558 * target.def (TARGET_VECTORIZE_BUILTIN_GATHER): New hook.
2559 * doc/tm.texi.in (TARGET_VECTORIZE_BUILTIN_GATHER): Document it.
2560 * doc/tm.texi: Regenerate.
2561 * tree-data-ref.c (initialize_data_dependence_relation,
2562 compute_self_dependence): No longer static.
2563 * tree-data-ref.h (initialize_data_dependence_relation,
2564 compute_self_dependence): New prototypes.
2565 * tree-vect-data-refs.c (vect_check_gather): New function.
2566 (vect_analyze_data_refs): Detect possible gather load data refs.
2567 * tree-vectorizer.h (struct _stmt_vec_info): Add gather_p field.
2568 (STMT_VINFO_GATHER_P): Define.
2569 (vect_check_gather): New prototype.
2570 * config/i386/i386-builtin-types.def: Add types for alternate
2572 * config/i386/sse.md (AVXMODE48P_DI): Remove.
2573 (VEC_GATHER_MODE): Rename mode_attr to...
2574 (VEC_GATHER_IDXSI): ... this.
2575 (VEC_GATHER_IDXDI, VEC_GATHER_SRCDI): New mode_attrs.
2576 (avx2_gathersi<mode>, *avx2_gathersi<mode>): Use <VEC_GATHER_IDXSI>
2577 instead of <VEC_GATHER_MODE>.
2578 (avx2_gatherdi<mode>): Use <VEC_GATHER_IDXDI> instead of
2579 <AVXMODE48P_DI> and <VEC_GATHER_SRCDI> instead of VEC_GATHER_MODE
2580 on src and mask operands.
2581 (*avx2_gatherdi<mode>): Likewise. Use VEC_GATHER_MODE iterator
2582 instead of AVXMODE48P_DI.
2583 (avx2_gatherdi<mode>256, *avx2_gatherdi<mode>256): Removed.
2584 * config/i386/i386.c (enum ix86_builtins): Add
2585 IX86_BUILTIN_GATHERALTSIV4DF, IX86_BUILTIN_GATHERALTDIV8SF,
2586 IX86_BUILTIN_GATHERALTSIV4DI and IX86_BUILTIN_GATHERALTDIV8SI.
2587 (ix86_init_mmx_sse_builtins): Create those builtins.
2588 (ix86_expand_builtin): Handle those builtins and adjust expansions
2589 of other gather builtins.
2590 (ix86_vectorize_builtin_gather): New function.
2591 (TARGET_VECTORIZE_BUILTIN_GATHER): Define.
2593 2011-11-07 Uros Bizjak <ubizjak@gmail.com>
2595 * config/i386/f16cintrin.h: Remove extra _X86INTRIN_H_INCLUDED check.
2597 2011-11-07 Tristan Gingold <gingold@adacore.com>
2599 * config.gcc (*-*-*vms*): Set c_target_objs, cxx_target_objs.
2600 * config/vms/t-vms: Add vms-c.o rule.
2601 * config/vms/vms-c.c: New file.
2602 * config/vms/vms-protos.h (vms_c_register_pragma): New prototype.
2603 * config/vms/vms.h (REGISTER_TARGET_PRAGMAS): Define.
2605 2011-11-07 Tristan Gingold <gingold@adacore.com>
2607 * config/alpha/vms.h (TARGET_OBJECT_SUFFIX,
2608 TARGET_EXECUTABLE_SUFFIX, TARGET_OS_CPP_BUILTINS,
2609 TARGET_ABI_OPEN_VMS, LONG_TYPE_SIZE, ADA_LONG_TYPE_SIZE,
2610 POINTER_SIZE, POINTERS_EXTEND_UNSIGNED): Move to config/vms/vms.h
2611 (SUBTARGET_OS_CPP_BUILTINS): Define.
2612 (TARGET_DEFAULT): Tune according to POINTER_SIZE.
2613 (MASK_RETURN_ADDR): Define in 64 bit mode.
2614 * config/ia64/vms.h: Likewise.
2615 * config/vms/vms.h: New file.
2616 * config/vms/vms64.h: New file.
2617 * config/alpha/vms64.h: Removed.
2618 * config/ia64/vms64.h: Removed.
2619 * config.gcc (alpha64-dec-*vms*, alpha*-dec-*vms*, ia64-hp-*vms*):
2620 Adjust for above change.
2622 2011-11-07 Enkovich Ilya <ilya.enkovich@intel.com>
2625 * config/i386/i386-protos.h (ix86_use_lea_for_mov): New.
2626 * config/i386/i386.c (ix86_use_lea_for_mov): Likewise.
2627 * config/i386/i386.md (movsi_internal): Emit lea if profitable.
2628 (movdi_internal_rex64): Likewise.
2630 2011-11-07 Sergey Ostanevich <sergos.gnu@gmail.com>
2632 PR rtl-optimization/47698
2633 * ifcvt.c (noce_operand_ok): Return false for mems with side effects.
2635 2011-11-07 Tristan Gingold <gingold@adacore.com>
2637 * common/config/alpha/alpha-common.c (alpha_option_init_struct):
2639 (TARGET_OPTION_INIT_STRUCT): Define.
2640 * config/alpha/alpha.c (TARGET_MIN_ANCHOR_OFFSET)
2641 (TARGET_MAX_ANCHOR_OFFSET)
2642 (TARGET_USE_BLOCKS_FOR_CONSTANT_P): Redefine.
2644 2011-11-06 Quentin Neill <quentin.neill@amd.com>
2646 Fix r180999, update ChangeLog
2647 * config.gcc: Add f16cintrin.h.
2648 * config/i386/f16cintrin.h: Add missing endif.
2650 2011-11-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
2652 * config.gcc (arm*-*-rtemseabi*): New target.
2653 * config/arm/rtems-eabi.h: New.
2654 * config/arm/t-rtems-eabi: New.
2656 2011-11-06 David S. Miller <davem@davemloft.net>
2658 * config/sparc/sparc.md (UNSPEC_SHORT_LOAD): Delete.
2659 (zero_extend_v8qi_vis, zero_extend_v4hi_vis,
2660 *zero_extend_v8qi_<P:mode>_insn,
2661 *zero_extend_v4hi_<P:mode>_insn): Express using vec_merge
2662 and vec_duplicate instead of using an UNSPEC.
2664 2011-11-07 Alan Modra <amodra@gmail.com>
2667 * config/rs6000/rs6000.c (rs6000_emit_stack_reset): Always emit
2668 blockage for ABI_V4.
2670 2011-11-06 Dave Korn <dave.korn.cygwin@gmail.com>
2672 * config/i386/cygwin.h (LIBGCJ_SONAME): Updated to match recent
2673 libgcj version bump.
2674 * config/i386/mingw32.h (LIBGCJ_SONAME): Likewise.
2676 2011-11-06 Eric Botcazou <ebotcazou@adacore.com>
2678 * gcse.c: Adjust outdated comments throughout.
2679 (struct mem_conflict_info): New structure.
2680 (mems_conflict_for_gcse_p): Use it to communicate with caller.
2681 (load_killed_in_block_p): Pass it to note_stores.
2682 (hash_expr): Remove superfluous line break.
2683 (hash_scan_set): Rename PAT parameter into SET.
2684 (hash_scan_insn): Reorder cases.
2685 (canon_list_insert): Fix long line.
2686 (edge_list): Delete.
2687 (prune_expressions): Rename E local variable into EXPR.
2688 (compute_pre_data): Return struct edge_list * object.
2689 (pre_expr_reaches_here_p_work): Fix formatting.
2690 (process_insert_insn): Move around comment.
2691 (pre_edge_insert): Fix long line.
2692 (pre_insert_copies): Likewise.
2693 (gcse_emit_move_after): Swap SRC and DEST parameters.
2694 (pre_delete): Adjust call to gcse_emit_move_after.
2695 (pre_gcse): Take struct edge_list * parameter. Fix long line.
2696 (one_pre_gcse_pass): Use flag_gcse_lm condition for all routines.
2697 Use a local list of edges.
2698 (hoist_code): Fix long line. Adjust call to gcse_emit_move_after.
2699 (pre_ldst_expr_hash): Fix long line.
2700 (free_ldst_mems): Rename into...
2701 (free_ld_motion_mems): ...this.
2702 (first_ls_expr): Delete.
2703 (next_ls_expr): Likewise.
2704 (print_ldst_list): Do not use above two functions.
2705 (simple_mem): Adjust interface.
2706 (compute_ld_motion_mems): Fix formatting.
2707 (update_ld_motion_stores): Reuse local variable.
2709 2011-11-06 Joseph Myers <joseph@codesourcery.com>
2711 * c-decl.c (shadow_tag_warned, grokdeclarator): Handle _Alignas
2713 (build_null_declspecs): Initialize align_log and alignas_p fields.
2714 (declspecs_add_alignas): New.
2715 * c-parser.c (c_token_starts_declspecs): Handle RID_ALIGNAS.
2716 (c_parser_declspecs): Handle _Alignas specifiers.
2717 (c_parser_alignas_specifier): New.
2718 (c_parser_alignof_expression): Diagnose alignof use for non-C1X.
2719 Diagnose _Alignof (expression).
2720 * c-tree.h (struct c_declspecs): Add align_log and alignas_p fields.
2721 (declspecs_add_alignas): Declare.
2722 * ginclude/stddef.h (max_align_t): Define for C1X and C++11.
2723 * ginclude/stdalign.h: New.
2724 * Makefile.in (USER_H): Add stdalign.h.
2726 2011-11-06 Joern Rennecke <joern.rennecke@embecosm.com>
2728 * regset.h (fixed_reg_set_regset): Declare.
2729 * dse.c: Include regset.h .
2730 (struct insn_info): Add member fixed_regs_live.
2731 (note_add_store_info): New typedef.
2732 (note_add_store): New function.
2733 (emit_inc_dec_insn_before): Expect arg to be of type insn_info_t.
2734 Use gen_add3_insn / gen_move_insn.
2735 Check new insn for unwanted clobbers before emitting it.
2736 (check_for_inc_dec): Rename to...
2737 (check_for_inc_dec_1:) ... this. Return bool. Take insn_info
2738 parameter. Changed all callers in file.
2739 (check_for_inc_dec, copy_fixed_regs): New functions.
2740 (scan_insn): Set fixed_regs_live field of insn_info.
2741 * rtl.h (check_for_inc_dec): Update prototype.
2742 * postreload.c (reload_cse_simplify): Take new signature of
2743 check_ind_dec into account.
2744 * reginfo.c (fixed_reg_set_regset): New variable.
2745 (init_reg_sets_1): Initialize it.
2747 2011-11-06 Jakub Jelinek <jakub@redhat.com>
2749 * tree-cfg.c (gimple_can_merge_blocks_p): For -O0 don't remove
2752 2011-11-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2754 * config/pa/pa.c (pa_hpux_init_libfuncs): Rename to pa_init_libfuncs.
2755 Remove dependence of declaration and target define on definition of
2756 HPUX_LONG_DOUBLE_LIBRARY. Update implementation.
2758 2011-11-06 Andrew Macleod <amacleod@redhat.com>
2759 Richard Henderson <rth@redhat.com>
2760 Aldy Hernandez <aldyh@redhat.com>
2762 Merged from cxx-mem-model.
2764 * cppbuiltin.c (define__GNUC__): Define __ATOMIC memory models
2765 * coretypes.h (enum memmodel): New. enumerated memory model type.
2766 * Makefile.in (cppbuiltin.o) Add missing dependency on $(TREE_H)
2767 * genopinit,c (optabs): Add atomic direct optab handlers.
2768 * sync-builtins.def (BUILT_IN_ATOMIC_*): New builtins.
2769 * builtin-types.def (BT_CONST_VOLATILE_PTR,
2770 BT_FN_I{1,2,4,8,16}_CONST_VPTR_INT, BT_FN_VOID_VPTR_INT,
2771 BT_FN_BOOL_VPTR_INT, BT_FN_BOOL_SIZE_CONST_VPTR,
2772 BT_FN_I{1,2,4,8,16}_VPTR_I{1,2,4,8,16}_INT,
2773 BT_FN_VOID_VPTR_I{1,2,4,8,16}_INT, BT_FN_VOID_SIZE_VPTR_PTR_INT,
2774 BT_FN_VOID_SIZE_CONST_VPTR_PTR_INT, BT_FN_VOID_SIZE_VPTR_PTR_PTR_INT,
2775 BT_FN_BOOL_VPTR_PTR_I{1,2,4,8,16}_BOOL_INT_INT): New builtin types.
2776 * expr.h (expand_atomic_*): Add prototypes.
2777 (expand_{bool,val}_compare_and_swap): Remove prototypes.
2778 * c-typeck.c (build_function_call_vec): Don't reprocess __atomic
2780 * common.opt (Winvalid-memory-model): New warning flag.
2781 (finline-atomics): New. Flag to disable atomic inlining.
2782 * params.h (ALLOW_LOAD_DATA_RACES): New.
2783 (ALLOW_PACKED_LOAD_DATA_RACES): New.
2784 (ALLOW_PACKED_STORE_DATA_RACES): New.
2785 * params.def (PARAM_ALLOW_LOAD_DATA_RACES): New.
2786 (PARAM_ALLOW_PACKED_LOAD_DATA_RACES): New.
2787 (PARAM_ALLOW_PACKED_STORE_DATA_RACES): New.
2788 * builtins.c (is_builtin_name): Handle __atomic.
2789 (get_memmodel): New. Extract memory model.
2790 (expand_expr_force_mode): New. Factor out common code for ensuring an
2791 integer argument is in the proper mode.
2792 (expand_builtin_sync_operation): Remove ignore param. Always call
2793 expand_atomic_fetch_op instead of the old expanders.
2794 (expand_builtin_compare_and_swap,
2795 expand_builtin_sync_lock_test_and_set): Use expand_expr_force_mode,
2796 call atomic expanders instead of sync expanders.
2797 (expand_builtin_sync_lock_release): Call atomic_store expander.
2798 (expand_builtin_atomic_compare_exchange, expand_builtin_atomic_load,
2799 expand_builtin_atomic_store, expand_builtin_atomic_fetch_op): New.
2800 (expand_builtin_atomic_exchange): New.
2801 (fold_builtin_atomic_always_lock_free,
2802 expand_builtin_atomic_always_lock_free,
2803 fold_builtin_atomic_is_lock_free, expand_builtin_atomic_is_lock_free):
2805 (expand_builtin_mem_thread_fence, expand_builtin_atomic_thread_fence,
2806 expand_builtin_atomic_signal_fence): New.
2807 (expand_builtin_mem_signal_fence): New.
2808 (expand_builtin): Add cases for BUILT_IN_ATOMIC_*.
2809 (fold_builtin_2): Add cases for BUILT_IN_ATOMIC_{IS,ALWAYS}_LOCK_FREE.
2810 * optabs.h (DOI_atomic_*): Define new atomics.
2811 (atomic_*_optab): Define.
2812 (can_compare_and_swap_p, expand_atomic_compare_and_swap): New
2814 * optabs.c (expand_sync_operation, expand_sync_fetch_operation): Remove.
2815 (expand_sync_lock_test_and_set): Remove.
2816 (expand_atomic_load, expand_atomic_store): New.
2817 (expand_atomic_exchange): New.
2818 (expand_atomic_compare_and_swap): New. Implements
2819 atomic_compare_exchange via compare and swap.
2820 (struct atomic_op_functions): Opcode table struct for fetch ops.
2821 (get_atomic_op_for_code): New. Return an opcode table entry.
2822 (maybe_emit_op): New. Try to emit a fetch op.
2823 (expand_atomic_fetch_op): New.
2824 (expand_val_compare_and_swap_1): Remove.
2825 (expand_val_compare_and_swap, expand_bool_compare_and_swap): Remove.
2826 (expand_atomic_compare_and_swap): Rename from
2827 expand_atomic_compare_exchange. Rewrite to return both success and
2828 oldval return values; expand via both atomic and sync optabs.
2829 (can_compare_and_swap_p): New.
2830 (expand_compare_and_swap_loop): Use expand_atomic_compare_and_swap.
2831 (maybe_gen_insn): Handle 7 and 8 operands.
2832 * omp-low.c (expand_omp_atomic_fetch_op): Don't test individual
2833 fetch_op optabs, only test can_compare_and_swap_p. Use __atomic
2834 builtins instead of __sync builtins.
2835 (expand_omp_atomic_pipeline): Use can_compare_and_swap_p.
2836 * doc/extend.texi: Document __atomic built-in functions.
2837 * doc/invoke.texi: Document data race parameters.
2838 * doc/md.texi: Document atomic patterns.
2839 * config/i386/i386.md (UNSPEC_MOVA): New.
2840 (UNSPECV_CMPXCHG): Split into ...
2841 (UNSPECV_CMPXCHG_1, UNSPECV_CMPXCHG_2,
2842 UNSPECV_CMPXCHG_3, UNSPECV_CMPXCHG_4): New.
2843 * config/i386/sync.md (ATOMIC): New mode iterator.
2844 (atomic_load<ATOMIC>, atomic_store<ATOMIC>): New.
2845 (atomic_loaddi_fpu, atomic_storedi_fpu, movdi_via_fpu): New.
2846 (mem_thread_fence): Rename from memory_barrier.
2847 Handle the added memory model parameter.
2848 (mfence_nosse): Rename from memory_barrier_nosse.
2849 (sync_compare_and_swap<CASMODE>): Split into ...
2850 (atomic_compare_and_swap<SWI124>): this and ...
2851 (atomic_compare_and_swap<CASMODE>): this. Handle the new parameters.
2852 (atomic_compare_and_swap_single<SWI>): Rename from
2853 sync_compare_and_swap<SWI>; rewrite to use split unspecs.
2854 (atomic_compare_and_swap_double<DCASMODE>): Rename from
2855 sync_double_compare_and_swap<DCASMODE>; rewrite to use split unspecs.
2856 (*atomic_compare_and_swap_doubledi_pic): Rename from
2857 sync_double_compare_and_swapdi_pic; rewrite to use split unspecs.
2858 (atomic_fetch_add<SWI>): Rename from sync_old_add<SWI>; add memory
2860 (*atomic_fetch_add_cmp<SWI>): Similarly.
2861 (atomic_add<SWI>, atomic<any_logic><SWI>): Similarly.
2862 (atomic_sub<SWI>): Similarly. Use x86_maybe_negate_const_int.
2863 (sync_lock_test_and_set<SWI>): Merge with ...
2864 (atomic_exchange<SWI>): ... this.
2866 2011-11-6 Richard Guenther <rguenther@suse.de>
2868 * ipa-prop.c (ipa_modify_call_arguments): Re-compute inlinable flag.
2870 2011-11-06 Ira Rosen <ira.rosen@linaro.org>
2872 * tree-vectorizer.h (vectorizable_condition): Add argument.
2873 * tree-vect-loop.c (vectorizable_reduction): Fail for condition
2874 in SLP. Update calls to vectorizable_condition.
2875 * tree-vect-stmts.c (vect_is_simple_cond): Add basic block info to
2876 the arguments. Pass it to vect_is_simple_use_1.
2877 (vectorizable_condition): Add slp_node to the arguments. Support
2878 vectorization of basic blocks. Fail for reduction in SLP. Update
2879 calls to vect_is_simple_cond and vect_is_simple_use. Support SLP:
2880 call vect_get_slp_defs to get vector operands.
2881 (vect_analyze_stmt): Update calls to vectorizable_condition.
2882 (vect_transform_stmt): Likewise.
2883 * tree-vect-slp.c (vect_create_new_slp_node): Handle COND_EXPR.
2884 (vect_get_and_check_slp_defs): Handle COND_EXPR. Allow pattern
2886 (vect_build_slp_tree): Handle COND_EXPR.
2887 (vect_analyze_slp_instance): Push pattern statements to root node.
2888 (vect_get_constant_vectors): Fix comments. Handle COND_EXPR.
2890 2011-11-05 David S. Miller <davem@davemloft.net>
2892 * config/sparc/sparc.md (UNSPEC_SHORT_LOAD): New unspec.
2893 (zero-extend_v8qi_vis, zero_extend_v4hi_vis): New expanders.
2894 (*zero_extend_v8qi_<P:mode>_insn,
2895 *zero_extend_v4hi_<P:mode>_insn): New insns.
2896 * config/sparc/sparc.c (vector_init_move_words,
2897 vector_init_prepare_elts, sparc_expand_vector_init_vis2,
2898 sparc_expand_vector_init_vis1): New functions.
2899 (vector_init_bshuffle): Rewrite to handle more cases and make use
2900 of locs[] array prepared by vector_init_prepare_elts.
2901 (vector_init_fpmerge, vector_init_faligndata): Delete.
2902 (sparc_expand_vector_init): Rewrite using new infrastructure.
2904 2011-11-05 Joern Rennecke <joern.rennecke@embecosm.com>
2906 * config.gcc (epiphany-*-*): New architecture.
2907 (epiphany-*-elf): New configuration.
2908 * config/epiphany, common/config/epiphany : New directories.
2909 * doc/extend.texi (disinterrupt attribute): Add Epiphany.
2910 (interrupt attribute): Add Epiphany.
2911 (long_call, short_call attribute): Add Epiphany.
2912 * doc/invoke.texi (Options): Add Epiphany options.
2913 * doc/md.texi (Machine Constraints): Add Epiphany constraints.
2914 * doc/install.texi (Options specification):
2915 Add --with-stack-offset=@var{num} description.
2916 (host/target specific issues): Add epiphany-*-elf.
2917 * doc/contrib.texi (Contributors): Mention Epiphany port.
2919 2011-11-05 Jakub Jelinek <jakub@redhat.com>
2921 PR tree-optimization/50693
2922 * tree-cfg.c (gimple_can_merge_blocks_p): Allow merging with
2923 non-forced user labels.
2924 (gimple_merge_blocks): Turn non-forced user labels into
2925 debug bind stmt with the label as first operand and reset value.
2926 (gimple_duplicate_bb): Don't duplicate label debug stmts.
2927 * dwarf2out.c (gen_label_die): Handle NOTE_INSN_DELETED_DEBUG_LABEL.
2928 * final.c (final_scan_insn): Likewise.
2929 (rest_of_clean_state): Don't dump NOTE_INSN_DELETED_DEBUG_LABEL.
2930 * var-tracking.c (debug_label_num): New variable.
2931 (delete_debug_insns): Don't delete DEBUG_INSNs for LABEL_DECLs,
2932 instead turn them into NOTE_INSN_DELETED_DEBUG_LABEL notes.
2933 * cfglayout.c (skip_insns_after_block, duplicate_insn_chain): Handle
2934 NOTE_INSN_DELETED_DEBUG_LABEL.
2935 (duplicate_insn_chain): Don't duplicate LABEL_DECL DEBUG_INSNs.
2936 * insn-notes.def (DELETED_DEBUG_LABEL): New note kind.
2937 * print-rtl.c (print_rtx): Handle NOTE_INSN_DELETED_DEBUG_LABEL.
2938 * gengtype.c (adjust_field_rtx_def): Likewise.
2939 * config/i386/i386.c (ix86_output_function_epilogue): For MachO
2940 clear CODE_LABEL_NUMBER of NOTE_INSN_DELETED_DEBUG_LABEL
2941 if their are at the end of function and nop hasn't been emitted.
2942 * config/rs6000/rs6000.c (rs6000_output_function_epilogue): Likewise.
2944 2011-11-05 Georg-Johann Lay <avr@gjlay.de>
2946 PR rtl-optimization/50448
2947 * cprop.c (try_replace_reg): Also try to replace uses of FROM that
2950 2011-11-05 Peter Dufault <dufault@hda.com>,
2951 Sebastian Huber <sebastian.huber@embedded-brains.de>
2953 * config/rs6000/rtems.h (SUBSUBTARGET_OVERRIDE_OPTIONS):
2954 Extend rs6000_spe handling.
2956 2011-11-05 Ralf Corsépius <ralf.corsepius@rtems.org>
2958 * config/rs6000/t-rtems: Add -mcpu=8540/-mfloat-gprs=double multilib.
2959 Remove -mcpu=601 multilib.
2960 Remove -Dmpc8260 multilib.
2961 * config/rs6000/rtems.h: Allow --float-gprs=... to override grps
2964 2011-11-05 Quentin Neill <quentin.neill@amd.com>
2966 Piledriver f16cintrin.h fix.
2967 * config/i386/f16cintrin.h: Contents moved from immintrin.h.
2968 * config/i386/immintrin.h: Include f16cintrin.h.
2969 * config.gcc (i[34567]86-*-*, x86_64-*-*): Add f16cintrin.h.
2972 2011-11-04 Eric Botcazou <ebotcazou@adacore.com>
2975 * c-parser.c (c_parser_postfix_expression) <RID_OFFSETOF>: Adjust call
2977 * c-typeck.c (build_unary_op) <ADDR_EXPR>: Call fold_offsetof_1.
2979 2011-11-04 Alan Modra <amodra@gmail.com>
2981 * reload1.c (gen_reload): Don't use REGNO on SUBREGs.
2982 * print-rtl.c (print_rtx): Don't segfault on negative regno.
2984 2011-11-04 David S. Miller <davem@davemloft.net>
2987 * config/sparc/sparc.c (sparc_expand_conditional_move): Handle the
2988 fact that sparc_emit_float_lib_cmp modifies the comparison in
2991 2011-11-04 Ralf Corsépius <ralf.corsepius@rtems.org>
2993 * config/lm32/t-rtems: New.
2994 * config.gcc (lm32-*-rtems*): Add t-rtems.
2996 2011-11-04 Eric Botcazou <ebotcazou@adacore.com>
2999 * config/sparc/sparc.h (ASM_CPU_SPEC): Pass -Av8 if -mcpu=v8.
3001 2011-11-04 Jiangning Liu <jiangning.liu@arm.com>
3003 PR rtl-optimization/38644
3004 * config/arm/arm.c (thumb1_expand_epilogue): Add memory barrier
3005 for epilogue having stack adjustment.
3007 2011-11-04 Georg-Johann Lay <avr@gjlay.de>
3010 * config/avr/avr-modes.def: New file defining PSImode.
3011 * config/avr/avr-c.c (__INT24_MAX__, __INT24_MIN__,
3012 __UINT24_MAX__): New built-in defines.
3013 * config/avr/avr.md (adjust_len): Add tstpsi, mov24, reload_in24,
3014 ashlpsi, ashrpsi, lshrpsi.
3015 (QISI, QIDI, HISI, HIDI, MPUSH, rotx, rotsmode): Add PSI.
3016 (MOVMODE): New mode iterator.
3017 (movpsi): New expander.
3018 (movqi, movhi, movsi, movsf, movpsi): Write as one using MOVMODE.
3019 (*reload_inpsi, *movpsi): New insns.
3020 (*reload_inpsi): New RTL peephole.
3021 (addpsi3, *addpsi3_zero_extend.qi, *addpsi3_zero_extend.hi,
3022 *addpsi3_sign_extend.hi): New insns.
3023 (subpsi3, *subpsi3_zero_extend.qi, *subpsi3_zero_extend.hi,
3024 *subpsi3_sign_extend.hi): New insns.
3025 (divmodpsi4, udivmodpsi4): New define insn-and-split.
3026 (*divmodpsi4_call, *udivmodpsi4_call): New insns.
3027 (andpsi3, iorpsi3, xorpsi3): New insns.
3028 (*rotlpsi2.1, *rotlpsi2.23): New insns.
3029 (*rotw<mode>): Insn condition only allow even-sized modes.
3030 (*rotb<mode>): Insn condition allows odd-sized modes.
3031 (ashlpsi3, ashrpsi3, lshrpsi3, *addpsi3.lt0): New insns.
3032 (negpsi2, one_cmplpsi2): New insns.
3033 (extendqipsi2, extendhipsi2, extendpsisi2): New insns.
3034 (zero_extendqipsi2, zero_extendhipsi2, zero_extendpsisi2): New
3036 (*cmppsi, *negated_tstpsi, *reversed_tstpsi): New insns.
3037 (cbranchpsi4): New expander.
3038 * config/avr/constraints.md (Ca3, Co3, Cx3): New constraints.
3039 * config/avr/avr-protos.h (avr_out_tstpsi, avr_out_movpsi,
3040 avr_out_ashlpsi3, avr_out_ashrpsi3, avr_out_lshrpsi3,
3041 avr_out_reload_inpsi): New prototypes.
3043 * config/avr/avr.c (TARGET_SCALAR_MODE_SUPPORTED_P): Define to...
3044 (avr_scalar_mode_supported_p): ...this new static function.
3045 (avr_asm_len): Always return "".
3046 (avr_out_load_psi, avr_out_store_psi): New static functions.
3047 (avr_out_movpsi, avr_out_reload_inpsi): New functions.
3048 (avr_out_tstpsi): New function.
3049 (avr_out_ashlpsi3, avr_out_ashrpsi3, avr_out_lshrpsi3): New functions.
3050 (avr_out_plus_1, output_reload_in_const): Handle 3-byte types.
3051 (avr_simplify_comparison_p): Ditto.
3052 (adjust_insn_length): Handle ADJUST_LEN_RELOAD_IN24,
3053 ADJUST_LEN_MOV24, ADJUST_LEN_TSTPSI, ADJUST_LEN_ASHLPSI,
3054 ADJUST_LEN_ASHRPSI, ADJUST_LEN_LSHRPSI.
3055 (avr_rtx_costs_1): Report PSI costs.
3056 (avr_libcall_value): Handle odd-sized parameters.
3057 (avr_init_builtin_int24): New static function to define built-in
3058 24-bit types __int24 and __uint24.
3059 (avr_init_builtins): Use it.
3061 2011-11-04 Thomas Doerfler <thomas.doerfler@embedded-brains.de>
3064 * config/arm/rtems-elf.h, config/arm/t-rtems: Add optional
3065 support for VFP floating point model.
3067 2011-11-04 Tristan Gingold <gingold@adacore.com>
3069 * config/alpha/vms.h (ASM_OUTPUT_DEF): Do not switch section.
3071 2011-11-04 Ira Rosen <ira.rosen@linaro.org>
3074 2011-10-24 Ira Rosen <ira.rosen@linaro.org>
3076 PR tree-optimization/50730
3077 * tree-vect-data-refs.c (vect_analyze_data_refs): Stop basic block
3078 analysis if encountered unsupported data-ref.
3080 2011-11-04 Jakub Jelinek <jakub@redhat.com>
3082 * config/i386/i386.c (ix86_expand_vector_convert_uns_vsivsf): New
3084 * config/i386/i386-protos.h (ix86_expand_vector_convert_uns_vsivsf):
3086 * config/i386/sse.md (floatuns<sseintvecmodelower><mode>2): Use it.
3087 For floatunsv8siv8sf2 require TARGET_AVX2.
3089 * config/i386/i386.c (ix86_expand_adjust_ufix_to_sfix_si): Add
3090 XORP argument. Subtract 0x1p31 instead of 0x1p32. Use normal
3091 signalling comparison instead of non-signalling. Store into
3092 *XORP pseudo holding 0x80000000 integers if 0x1p31 has been
3093 subtracted and 0 otherwise.
3094 * config/i386/i386-protos.h (ix86_expand_adjust_ufix_to_sfix_si):
3096 * config/i386/sse.md (fixuns_trunc<mode><sseintvecmodelower>2): Enable
3097 already for TARGET_SSE2. Xor in vector initialized by
3098 ix86_expand_adjust_ufix_to_sfix_si at the end.
3099 (vec_pack_ufix_trunc_<mode>): Likewise.
3101 * tree-vect-stmts.c (vectorizable_conversion): Rewritten to handle
3102 not just FLOAT_EXPR and FIX_TRUNC_EXPR, but also CONVERT_EXPR_CODE_P,
3103 WIDEN_MULT_EXPR and WIDEN_LSHIFT_EXPR to handle what
3104 vectorizable_type_demotion and vectorizable_type_promotion did.
3105 Additionally handle FLOAT_EXPR and FIX_TRUNC_EXPR where the integer
3106 is {,un}signed {char,short}.
3107 (vect_create_vectorized_demotion_stmts): Fix comment typo. For
3108 recursive calls unconditionally use VEC_PACK_TRUNC_EXPR.
3109 Push vec_dest back to the vec_dsts vector at the end.
3110 (vect_create_vectorized_promotion_stmts): Don't recurse, do just
3111 one step. Removed multi_step_cvt, vec_dsts, slp_node and
3112 prev_stmt_info arguments, add vec_dest argument. Push always
3113 into vec_tmp, not just when multi_step_cvt != 0, replace *vec_oprdn0
3114 with vec_tmp at the end after freeing old *vec_oprnd0 vector.
3115 (vectorizable_type_demotion, vectorizable_type_promotion): Removed.
3116 (vect_analyze_stmt): Don't call vectorizable_type_demotion and
3117 vectorizable_type_promotion. Call vectorizable_conversion even
3118 for SLP bb vectorization.
3119 (vect_transform_stmt): Call vectorizable_conversion instead of
3120 vectorizable_type_demotion and vectorizable_type_promotion.
3121 (supportable_widening_operation): Clear *multi_step_cvt first,
3122 simplify c1/c2 computation, free *interm_types vector on failure.
3123 (supportable_narrowing_operation): Clear *multi_step_cvt first,
3124 free *interm_types vector on failure, handle multi-step
3127 2011-11-04 Tristan Gingold <gingold@adacore.com>
3129 * config/alpha/alpha.c (alpha_write_linkage): Remove fundecl
3130 argument. Conditionally generate crash debug info. Adjust
3131 for alpha_funcs_tree removal.
3132 (machine_function): Add links field.
3133 (alpha_start_function): Conditionally generate crash debug info.
3134 (alpha_end_function): Adjust call to alpha_write_linkage.
3135 (alpha_funcs): Remove.
3136 (links_kind): Remove.
3137 (alpha_links): Remove num, target and lkind field. Add func field.
3138 (alpha_links_tree): Remove.
3139 (alpha_funcs_tree): Remove.
3140 (alpha_need_linkage): Remove.
3141 (alpha_use_linkage): Change prototype. Adjust.
3142 (alpha_write_one_linkage): Use ASM_OUTPUT_INTERNAL_LABEL.
3143 Use SYMBOL_REF_EXTERNAL_P and SYMBOL_REF_LOCAL_P macro.
3144 * config/alpha/alpha-protos.h (alpha_use_linkage): Update.
3145 (alpha_need_linkage): Remove.
3146 * config/alpha/alpha.md: Update calls to alpha_use_linkage.
3147 Adjust calls to alpha_need_linkage.
3149 2011-11-03 Uros Bizjak <ubizjak@gmail.com>
3151 * sched-vis.c (print_value): Handle STRICT_LOW_PART.
3153 2011-11-03 Uros Bizjak <ubizjak@gmail.com>
3155 * config/i386/i386.md (lround<X87MODEF:mode><SWI248x:mode>2,
3156 rint<mode>2, floor<mode>2, lfloor<MODEF:mode><SWI48:mode>2,
3157 btrunc<mode>2, lwp_lwpval<mode>3): Use operands[N] instead of operandN.
3159 2011-11-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3162 * config/arm/t-bpabi: New file.
3163 * config.gcc (arm*-*-linux*): Add arm/t-bpabi to tmake_file for
3165 (arm*-*-uclinux*): Add arm/t-bpabi to tmake_file for
3166 arm*-*-uclinux*eabi.
3167 (arm*-*-eabi*, arm*-*-symbianelf*): Add arm/t-bpabi to tmake_file
3170 2011-11-03 Michael Matz <matz@suse.de>
3173 * configure.ac: Check for -fno-exceptions -fno-rtti.
3174 * configure: Regenerate.
3175 * Makefile.in (NOEXCEPTION_FLAGS): New flag.
3176 (ALL_CXXFLAGS): Use it.
3178 2011-11-03 Uros Bizjak <ubizjak@gmail.com>
3180 * config/i386/i386.md: Use {} for multi-line preparation statements.
3182 2011-11-03 Eric Botcazou <ebotcazou@adacore.com>
3184 * config/sparc/sparc.md (movtf_insn_sp32_no_fpu): Consolidate into...
3185 (movtf_insn_sp32): ...this.
3186 (movtf_insn_sp64_no_fpu): Consolidate into...
3187 (movtf_insn_sp64): ...this.
3188 (movtf_insn_sp64_hq): Do not test TARGET_FPU.
3189 * config/sparc/sparc.c (sparc_legitimate_address_p): Likewise.
3191 2011-11-03 Tristan Gingold <gingold@adacore.com>
3193 * config/vms/vms.c (vms_patch_builtins): Fix typo.
3195 2011-11-03 Richard Guenther <rguenther@suse.de>
3198 * lto-opts.c: Re-implement.
3199 * lto-streamer.h (lto_register_user_option): Remove.
3200 (lto_read_file_options): Likewise.
3201 (lto_reissue_options): Likewise.
3202 (lto_clear_user_options): Likewise.
3203 (lto_clear_file_options): Likewise.
3204 * opts-global.c (post_handling_callback): Remove.
3205 (set_default_handlers): Do not set post_handling_callback.
3206 (decode_options): Remove LTO specific code.
3207 * lto-wrapper.c (merge_and_complain): New function.
3208 (run_gcc): Read all input file options and
3209 prepend a merged set before the linker driver options.
3210 * gcc.c (driver_post_handling_callback): Remove.
3211 (set_option_handlers): Do not set post_handling_callback.
3212 * opts-common.c (handle_option): Do not call post_handling_callback.
3213 * opts.h (struct cl_option_handlers): Remove post_handling_callback.
3215 2011-11-03 Richard Guenther <rguenther@suse.de>
3217 * collect2.c (main): Guard object_nbr variable with TARGET_AIX_VERSION.
3219 2011-11-03 Martin Jambor <mjambor@suse.cz>
3221 * ipa-prop.c (type_change_info): New fields offset, object,
3222 known_current_type and multiple_types_encountered.
3223 (extr_type_from_vtbl_ptr_store): New function.
3224 (check_stmt_for_type_change): Use it, set multiple_types_encountered if
3225 the result is different from the previous one.
3226 (detect_type_change): Renamed to detect_type_change_1. New parameter
3227 comp_type. Set up new fields in tci, build known type jump
3228 functions if the new type can be identified.
3229 (detect_type_change): New function.
3230 * tree.h (DECL_CONTEXT): Comment new use.
3232 2011-11-03 Richard Guenther <rguenther@suse.de>
3235 * lto-wrapper.c (get_options_from_collect_gcc_options): Properly
3236 decode an encoded literal '.
3238 2011-11-03 Tristan Gingold <gingold@adacore.com>
3240 * collect2.c (main): Add support of -f (response file) on AIX.
3242 2011-11-03 Ira Rosen <ira.rosen@linaro.org>
3244 PR tree-optimization/50912
3245 * tree-vectorizer.h (slp_void_p): New.
3246 (struct _slp_tree): Replace left and right with children. Update
3248 (struct _slp_oprnd_info): New.
3249 (vect_get_vec_defs): Declare.
3250 (vect_get_slp_defs): Update arguments.
3251 * tree-vect-loop.c (vect_create_epilog_for_reduction): Call
3252 vect_get_vec_defs instead of vect_get_slp_defs.
3253 (vectorizable_reduction): Likewise.
3254 * tree-vect-stmts.c (vect_get_vec_defs): Remove static, add argument.
3255 Update call to vect_get_slp_defs.
3256 (vectorizable_conversion): Update call to vect_get_vec_defs.
3257 (vectorizable_assignment, vectorizable_shift,
3258 vectorizable_operation): Likewise.
3259 (vectorizable_type_demotion): Call vect_get_vec_defs instead of
3261 (vectorizable_type_promotion, vectorizable_store): Likewise.
3262 (vect_analyze_stmt): Fix typo.
3263 * tree-vect-slp.c (vect_free_slp_tree): Update SLP tree traversal.
3264 (vect_print_slp_tree, vect_mark_slp_stmts,
3265 vect_mark_slp_stmts_relevant, vect_slp_rearrange_stmts,
3266 vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations,
3267 vect_schedule_slp_instance): Likewise.
3268 (vect_create_new_slp_node): New.
3269 (vect_create_oprnd_info, vect_free_oprnd_info): Likewise.
3270 (vect_get_and_check_slp_defs): Pass information about defs using
3271 oprnds_info, allow any number of operands.
3272 (vect_build_slp_tree): Likewise. Update calls to
3273 vect_get_and_check_slp_defs. Fix comments.
3274 (vect_analyze_slp_instance): Move node creation to
3275 vect_create_new_slp_node.
3276 (vect_get_slp_defs): Allow any number of operands.
3278 2011-11-02 Peter Bergner <bergner@vnet.ibm.com>
3279 Iain Sandoe <iains@gcc.gnu.org>
3281 * config/rs6000/rs6000.c (USE_HIDDEN_LINKONCE): New define.
3282 (get_ppc476_thunk_name): Use it.
3283 (rs6000_code_end): Likewise.
3284 (macho_branch_islands): Fix typo.
3286 2011-11-02 Paolo Carlini <paolo.carlini@oracle.com>
3287 Jason Merrill <jason@redhat.com>
3290 * configure.ac: Add -Wno-narrowing to warning options.
3291 * doc/invoke.texi ([-Wnarrowing], [-Wc++0x-compat]): Update.
3293 2011-11-02 Eric Botcazou <ebotcazou@adacore.com>
3296 * config/sparc/sparc.md (movsf_insn): Reindent constraints.
3297 (movdf_insn_sp32): Likewise. Remove redundant G constraint.
3298 (movdf_insn_sp64): Likewise.
3299 (DFmode splitter): Do not test TARGET_FPU.
3300 (movtf_insn_sp32): Reindent constraints.
3301 (movtf_insn_sp32_no_fpu): Likewise.
3302 (movtf_insn_sp64): Likewise.
3303 (movtf_insn_sp64_hq): Likewise.
3304 (movtf_insn_sp64_no_fpu): Likewise.
3306 2011-11-02 Paolo Carlini <paolo.carlini@oracle.com>
3309 * builtins.c (fold_builtin_memchr): Fix cast.
3311 2011-11-02 Teresa Johnson <tejohnson@google.com>
3313 * config/i386/predicates.md (promotable_binary_operator): Add minus
3314 to the list of promotable operators.
3316 2011-11-02 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3318 * gthr-single.h, gthr.h: Move to ../libgcc.
3319 * gthr-aix.h: Move to ../libgcc/config/rs6000.
3320 * gthr-dce.h: Move to ../libgcc/config/pa.
3321 * gthr-lynx.h: Move to ../libgcc/config.
3322 * gthr-mipssde.h: Move to ../libgcc/config/mips.
3323 * gthr-posix.h: Move to ../libgcc/config.
3324 * gthr-rtems.h: Likewise.
3325 * gthr-tpf.h: Move to ../libgcc/config/s390.
3326 * gthr-vxworks.h: Move to ../libgcc/config.
3327 * gthr-win32.h: Move to ../libgcc/config/i386.
3328 * configure.ac (gthread_flags): Remove
3329 (gthr-default.h): Don't create.
3330 (thread_file): Don't substitute.
3331 * configure: Regenerate.
3332 * Makefile.in (GCC_THREAD_FILE): Remove.
3333 (GTHREAD_FLAGS): Remove.
3334 (libgcc.mvars): Remove GTHREAD_FLAGS.
3335 * config/t-vxworks (EXTRA_HEADERS): Remove.
3337 2011-11-02 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3338 Paolo Bonzini <bonzini@gnu.org>
3340 * configure.ac (libgcc_tm_file_list, libgcc_tm_include_list): Remove.
3341 * configure: Regenerate.
3342 * Makefile.in (libgcc_tm_file_list, libgcc_tm_include_list): Remove.
3343 (TM_H): Remove libgcc_tm.h, $(libgcc_tm_file_list).
3344 (libgcc_tm.h, cs-libgcc_tm.h): Remove.
3345 (clean): Remove libgcc_tm.h
3346 * mkconfig.sh: Don't include libgcc_tm.h in tm.h.
3347 * config.gcc (libgcc_tm_file): Remove.
3348 (arm*-*-linux*): Remove libgcc_tm_file for arm*-*-linux-*eabi.
3349 (arm*-*-uclinux*): Remove libgcc_tm_file for arm*-*-uclinux*eabi.
3350 (arm*-*-eabi*, arm*-*-symbianelf*): Remove libgcc_tm_file.
3351 (avr-*-rtems*): Likewise.
3352 (avr-*-*): Likewise.
3353 (frv-*-elf): Likewise.
3354 (frv-*-*linux*): Likewise.
3355 (h8300-*-rtems*): Likewise.
3356 (h8300-*-elf*): Likewise.
3357 (i[34567]86-*-darwin*): Likewise.
3358 (x86_64-*-darwin*): Likewise.
3359 (rx-*-elf*): Likewise.
3360 (tic6x-*-elf): Likewise.
3361 (tic6x-*-uclinux): Likewise.
3362 (i[34567]86-*-linux*, x86_64-*-linux*): Likewise.
3364 2011-11-02 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3366 * Makefile.in (LIBGCC2_DEBUG_CFLAGS LIBGCC2_CFLAGS)
3367 (LIBGCC2_INCLUDES, TARGET_LIBGCC2_CFLAGS, LIB2FUNCS_EXTRA)
3368 (LIB2FUNCS_STATIC_EXTRA, LIB2FUNCS_EXCLUDE, T, T_TARGET)
3369 (INCLUDES_FOR_TARGET): Remove.
3370 (LIBGCC2_CFLAGS): Don't export.
3371 (LIB2FUNCS_ST, LIB2_DIVMOD_FUNCS, LIB2ADD, LIB2ADD_ST, srcdirify):
3373 (libgcc-support): Remove $(LIB2ADD), $(LIB2ADD_ST) dependencies.
3374 (libgcc.mvars): Likewise.
3375 Don't emit LIB2FUNCS_ST, LIB2FUNCS_EXCLUDE, LIB2ADD, LIB2ADD_ST,
3376 LIB2_SIDITI_CONV_FUNCS, LIB2_DIVMOD_FUNCS, LIBGCC2_CFLAGS,
3377 TARGET_LIBGCC2_CFLAGS.
3379 * libgcc2.c, libgcc2.h, gbl-ctors.h, longlong.h: Move to ../libgcc.
3380 * config/darwin-64.c: Move to ../libgcc/config.
3381 * config/divmod.c, config/floatunsidf.c, config/floatunsisf.c,
3382 config/floatunsitf.c, config/floatunsixf.c, config/udivmod.c,
3383 config/udivmodsi4.c: Move to ../libgcc/config.
3384 * config/gthr-posix.c: Move to ../libgcc/config/alpha.
3385 * config/memcmp.c, config/memcpy.c, config/memmove.c,
3386 config/memset.c: Move to ../libgcc/config.
3387 * config/t-darwin (TARGET_LIBGCC2_CFLAGS): Remove.
3388 * config/t-freebsd: Remove.
3389 * config/t-freebsd-thread: Move to ../libgcc/config.
OSDN Git Service
