1 2007-02-08 DJ Delorie <dj@redhat.com>
3 * config/m32c/m32c.c (m32c_unpend_compare): Add default to silence
5 (legal_subregs): Use unsigned char, make const.
6 (m32c_illegal_subreg_p): Use ARRAY_SIZE. Delete unused variables.
8 2007-02-08 Paul Brook <paul@codesourcery.com>
10 * config/arm/lib1funcs.asm (RETLDM): Pop directly into PC when no
11 special interworking needed.
13 2007-02-08 Harsha Jagasia <harsha.jagasia@amd.com>
15 * config/i386/xmmintrin.h: Make inclusion of emmintrin.h
16 conditional to __SSE2__.
17 (Entries below should have been added to first ChangeLog
18 entry for amdfam10 dated 2007-02-05)
19 * config/i386/emmintrin.h: Generate #error if __SSE2__ is not
21 * config/i386/pmmintrin.h: Generate #error if __SSE3__ is not
23 * config/i386/tmmintrin.h: Generate #error if __SSSE3__ is not
26 2007-02-08 DJ Delorie <dj@redhat.com>
28 * config/m32c/m32c-protos.h (m32c_illegal_subreg_p): New.
29 * config/m32c/m32c.c (legal_subregs): New.
30 (m32c_illegal_subreg_p): New.
31 * config/m32c/predicates.md (m32c_any_operand): Use it to reject
32 unsupported subregs of hard regs.
34 2007-02-08 Jan Hubicka <jh@suse.cz>
36 * tree-cfg.c (bsi_replace): Shortcut when replacing the statement with
37 the same one; always update histograms.
39 2007-02-08 Diego Novillo <dnovillo@redhat.com>
41 * passes.c (init_optimization_passes): Tidy comment.
43 2007-02-08 Roger Sayle <roger@eyesopen.com>
45 * simplify-rtx.c (simplify_unary_operation_1) <POPCOUNT>: We can
46 strip zero_extend, bswap and rotates from POCOUNT's argument.
47 <PARITY>: Likewise, we can strip not, bswap, sign_extend,
48 zero_extend and rotates from PARITY's argument.
49 <BSWAP>: A byte-swap followed by a byte-swap is an identity.
50 (simplify_const_unary_operation) <BSWAP>: Evaluate the byte-swap
51 of an integer constant at compile-time.
53 2007-02-08 Diego Novillo <dnovillo@redhat.com>
56 * tree-flow.h (struct var_ann_d): Remove field 'is_used'.
58 * tree-ssa-alias.c (compute_is_aliased): Remove. Update all
61 * tree-ssa-live.c (remove_unused_locals): Do not remove
62 TREE_ADDRESSABLE variables.
63 * tree-ssa-structalias.c (compute_points_to_sets): Tidy.
64 * tree-ssa-operands.c (add_virtual_operand): Remove argument
65 FOR_CLOBBER. Update all users.
66 If VAR has an associated alias set, add a virtual operand for
67 it if no alias is found to conflict with the memory reference.
69 2007-02-07 Jan Hubicka <jh@suse.cz>
70 Robert Kidd <rkidd@crhc.uiuc.edu>
72 * value-prof.c (visit_hist, free_hist): Return 1 instead of 0.
74 2007-02-07 Ian Lance Taylor <iant@google.com>
76 * lower-subreg.c (simple_move): Reject PARTIAL_INT modes.
78 2007-02-07 Roger Sayle <roger@eyesopen.com>
80 * config/rs6000/rs6000.md (ctz<mode>2, ffs<mode>2, popcount<mode>2,
81 parity<mode>2, smulsi3_highpart, abstf2_internal, allocate_stack,
82 tablejumpdi, movsi_to_cr_one): Remove constraints from
83 define_expand's match_operands.
85 2007-02-07 Roger Sayle <roger@eyesopen.com>
87 * global.c (compute_regsets): Move declatation of "i" inside of
88 #ifdef ELIMINABLE_REGS to avoid unused variable bootstrap failure.
90 2007-02-07 Jakub Jelinek <jakub@redhat.com>
93 * gimplify.c (gimplify_scan_omp_clauses): Remove special casing
94 of INDIRECT_REF <RESULT_DECL>.
96 * config/i386/i386.c (override_options): Set PTA_SSSE3 for core2.
98 2007-02-06 J"orn Rennecke <joern.rennecke@arc.com>
99 Kaz Kojima <kkojima@gcc.gnu.org>
102 * config/sh/sh.c (expand_cbranchdi4): Use scratch register
104 (sh_initialize_trampoline): Add parentheses to avoid a warning.
106 2007-02-06 Zdenek Dvorak <dvorakz@suse.cz>
108 * doc/loop.texi: Document possibility not to perform disambiguation
109 of loops with multiple latches.
110 * cfgloopmanip.c (alp_enum_p): Removed.
111 (add_loop): Handle subloops. Use get_loop_body_with_size.
112 (create_preheader): Do not allow ENTRY_BLOCK_PTR to be preheader.
113 * cfghooks.c (redirect_edge_and_branch_force): Set dominator for
114 the new forwarder block.
115 (make_forwarder_block): Only call new_bb_cbk if it is not NULL.
116 Handle the case latch is NULL.
117 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Avoid cfg modifications
118 when marking loop exits.
119 * ifcvt.c (if_convert): Ditto. Mark loop exits even if cfg cannot
121 * loop-init.c (loop_optimizer_init): Do not modify cfg. Call
122 disambiguate_loops_with_multiple_latches.
123 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Calculate dominators
124 before fix_loop_structure.
125 * cfgloop.c: Include pointer-set.h and output.h.
126 (canonicalize_loop_headers, HEADER_BLOCK, LATCH_EDGE,
127 update_latch_info, mfb_keep_just, mfb_keep_nonlatch): Removed.
128 (get_loop_latch_edges, find_subloop_latch_edge_by_profile,
129 find_subloop_latch_edge_by_ivs, find_subloop_latch_edge,
130 mfb_redirect_edges_in_set, form_subloop, merge_latch_edges,
131 disambiguate_multiple_latches, get_loop_body_with_size,
132 disambiguate_loops_with_multiple_latches): New functions.
133 (flow_loop_dump): Dump multiple latch edges.
134 (flow_loop_nodes_find): Handle loops with multiple latches.
135 (flow_loops_find): Ditto. Do not call canonicalize_loop_headers.
136 (glb_enum_p): Modified.
137 (get_loop_body): Use get_loop_body_with_size.
138 * cfgloop.h (LOOPS_HAVE_RECORDED_EXITS): New flag.
139 (AVOID_CFG_MODIFICATIONS): New constant.
140 (disambiguate_loops_with_multiple_latches, add_loop,
141 get_loop_body_with_size): Declare.
142 * Makefile.in (cfgloop.o): Add pointer-set.h and output.h.
144 2007-02-06 Seongbae Park <seongbae.park@gmail.com>
147 * global.c (compute_regsets): New function.
148 (global_alloc): Refactored ELIMINABLE_REGSET
149 and NO_GLOBAL_ALLOC_REGS computation out.
150 (rest_of_handle_global_alloc): Call compute_regsets()
151 for non-optimizing case.
153 2007-02-06 Richard Henderson <rth@redhat.com>
155 * config/i386/constraints.md (Y2): Rename from Y.
156 (Yi): New constraint.
157 * config/i386/i386.md (movsi_1, movdi_2, pushdf_nointeger,
158 pushdf_integer, movdf_nointeger, movdf_integer, zero_extendsidi2_32,
159 zero_extendsidi2_rex64, truncxfdf2_mixed): Change Y constraints to Y2.
160 (extendsfdf2_mixed, extendsfdf2_sse, truncdfsf_fast_mixed,
161 truncdfsf_fast_sse, truncdfsf_mixed, fix_truncdfdi_sse,
162 fix_truncdfsi_sse, floatsidf2_mixed, floatsidf2_sse,
163 floatdidf2_mixed, floatdidf2_sse, absnegdf2_mixed,
164 absnegdf2_sse, sse_setccdf, fop_df_comm_mixed, fop_df_comm_sse,
165 fop_df_1_mixed, fop_df_1_sse): Change Y constraints to x.
166 * config/i386/mmx.md (mov<MMXMODEI>_internal_rex64,
167 mov<MMXMODEI>_internal, movv2sf_internal_rex64, movv2sf_internal,
168 vec_extractv2si_1): Change Y constraints to Y2.
169 * config/i386/sse.md (vec_setv4sf_0, vec_concatv2df, vec_dupv4si,
170 vec_dupv2di, sse2_concatv2si, vec_concatv4si_1, vec_concatv2di):
171 Change Y constraints to Y2.
172 (sse2_loadld): Change Y constraints to x.
174 2007-02-06 Roger Sayle <roger@eyesopen.com>
176 * config/rs6000/rs6000.md (popcount<mode>2): Rewrite.
177 (parity<mode>2): New define_expand using rs6000_emit_parity.
178 * config/rs6000/rs6000.c (rs6000_emit_popcount,
179 rs6000_emit_parity): New functions.
180 * config/rs6000/rs6000-protos.h (rs6000_emit_popcount,
181 rs6000_emit_parity): Prototype here.
183 2007-02-06 Ian Lance Taylor <iant@google.com>
185 * lower-subreg.c (simple_move_operand): Reject CONST.
186 (resolve_clobber): Call validate_change rather than directly
187 assigning to XEXP (pat, 0).
189 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
191 * Makefile.in (tree-ssa-loop-ivopts.o): Add pointer-set.h dependency.
192 (tree-ssa-reassoc.o): Add pointer-set.h dependency.
193 (tree-cfg.o): Remove hashtab.h dependency.
195 * tree-ssa-loop-ivopts.c: Include pointer-set.h.
196 (struct ivopts_data): Change niters to pointer_map_t.
197 (struct nfe_cache_elt, nfe_hash, nfe_eq): Delete.
198 (niter_for_exit): Create pointer_map on demand. Change for
200 (tree_ssa_iv_optimize_init): Initialize data->niters to NULL.
201 (free_loop_data): Destroy data->niters if created and reset field.
202 (tree_ssa_iv_optimize_finalize): Don't delete data->niters here.
203 (tree_ssa_iv_optimize_loop): Check for presence of stale data.
205 * tree-ssa-reassoc.c: Include pointer-set.h.
206 (bb_rank): Change to long *.
207 (operand_rank): Change to pointer_map_t.
208 (find_operand_rank): Return long, -1 if not found. Declare as inline.
209 (insert_operand_rank): Accept long.
210 (operand_entry_hash, operand_entry_eq): Remove.
211 (get_rank): Return long. Adjust for changes above.
212 (init_reassoc): Change rank type to long. Adjust creation of bb_rank
214 (fini_reassoc): Delete operand_rank with pointer_map_destroy.
216 * tree-ssa-structalias.c (vi_for_tree): Change to pointer_map.
217 (struct tree_vi, tree_vi_t, tree_vi_hash, tree_vi_eq): Delete.
218 (insert_vi_for_tree): Rewrite for pointer_map API. Assert argument
220 (lookup_vi_for_tree): Rewrite for pointer_map API. Return varinfo_t
221 directly since it cannot be NULL.
222 (get_vi_for_tree): Rewrite for pointer_map API.
223 (find_what_p_points_to): Adjust for change to lookup_vi_for_tree.
224 (init_alias_vars): Create vi_for_tree as pointer_map.
225 (delete_points_to_sets): Delete vi_for_tree using pointer_map_destroy.
227 * tree-cfg.c: Don't include hashtab.h.
228 (edge_to_cases): Declare as pointer_map.
229 (struct edge_to_cases_elt, edge_to_cases_hash, edge_to_cases_eq):
231 (edge_to_cases_cleanup): Rewrite as pointer_map_traverse callback.
232 (start_recording_case_labels): Create edge_to_cases as pointer_map.
233 (end_recoding_case_labels): Cleanup edge_to_cases manually before
235 (record_switch_edge): Delete.
236 (get_cases_for_edge): Adjust for pointer_map API, inline
237 record_switch_edge (rewritten for new API), remove goto.
239 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
241 * Makefile.in (tree-nested.o): Add pointer-set.h dependency.
242 * tree-nested.c: Include pointer-set.h.
243 (var_map_elt, var_map_eq, var_map_hash): Delete.
244 (struct nesting_info): Remove GTY marker. Change the two htab_t's
246 (nesting_info_bitmap_obstack): New.
247 (lookup_field_for_decl): Adjust for pointer_map API.
248 (lookup_tramp_for_decl): Adjust for pointer_map API.
249 (get_nonlocal_debug_decl): Adjust for pointer_map API.
250 (get_local_debug_decl): Adjust for pointer_map API.
251 (convert_nl_goto_reference): Adjust for pointer_map API.
252 (convert_nl_goto_receiver): Adjust for pointer_map API.
253 (create_nesting_tree): Create outside GGC space. Create bitmap on
254 the new obstack. Create field_map and var_map as pointer_maps.
255 (free_nesting_tree): Adjust for changes to create_nesting_tree.
257 (lower_nested_functions): Move root here, no need to NULL it.
258 Initialize and release the obstack.
260 2007-02-06 Paolo Bonzini <bonzini@gnu.org>
262 * tree.c (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p):
263 Remove prototypes and make them non-static.
264 (struct tree_int_map): Remove.
265 * tree.h (struct tree_int_map): Move here, turning TO into an
267 (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p): Declare.
269 * tree.h (TREE_COMPLEXITY): Remove.
270 (struct tree_exp): Remove complexity field.
271 * tree.c (build1_stat): Don't set it.
273 2007-02-06 Dorit Nuzman <dorit@il.ibm.com>
274 Victor Kaplansky <victork@il.ibm.com>
276 * tree-vectorizer.c (vect_is_simple_use): Support induction.
277 (vect_is_simple_reduction): Support reduction with induction as
279 (vect_is_simple_iv_evolution): Fix formatting.
280 * tree-vect-analyze.c (vect_mark_stmts_to_be_vectorized): Fix
281 formatting. Don't mark induction phis for vectorization.
282 (vect_analyze_scalar_cycles): Analyze all inductions, then reductions.
283 * tree-vect-transform.c (get_initial_def_for_induction): New function.
284 (vect_get_vec_def_for_operand): Support induction.
285 (vect_get_vec_def_for_stmt_copy): Fix formatting and add check for
287 (vectorizable_reduction): Support reduction with induction as one of
289 (vectorizable_type_demotion): Use def-type of stmt argument rather
292 * tree-ssa-loop.c (gate_scev_const_prop): Return the value of
293 flag_tree_scev_cprop.
294 * common.opt (tree-scev-cprop): New flag.
296 * tree-vect-transform.c (vect_create_destination_var): Use 'kind' in
297 call to vect_get_new_vect_var.
299 2007-02-06 Ira Rosen <irar@il.ibm.com>
301 * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Check that
303 (vect_pattern_recog_1): Likewise.
305 2007-02-05 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
307 * fold-const.c (negate_expr_p): Handle CONJ_EXPR.
308 (fold_negate_expr): Likewise.
310 2007-02-05 Alexandre Oliva <aoliva@redhat.com>
313 * dwarf2out.c (modified_type_die): Follow DECL_ORIGINAL_TYPE
314 even if cv-qualification is the same.
316 2007-02-05 Geoffrey Keating <geoffk@apple.com>
318 * config/rs6000/darwin-tramp.asm (__trampoline_setup): Call
319 __enable_execute_stack on completion.
321 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
323 * config/i386/athlon.md (athlon_fldxf_k8, athlon_fld_k8,
324 athlon_fstxf_k8, athlon_fst_k8, athlon_fist, athlon_fmov,
325 athlon_fadd_load, athlon_fadd_load_k8, athlon_fadd, athlon_fmul,
326 athlon_fmul_load, athlon_fmul_load_k8, athlon_fsgn,
327 athlon_fdiv_load, athlon_fdiv_load_k8, athlon_fdiv_k8,
328 athlon_fpspc_load, athlon_fpspc, athlon_fcmov_load,
329 athlon_fcmov_load_k8, athlon_fcmov_k8, athlon_fcomi_load_k8,
330 athlon_fcomi, athlon_fcom_load_k8, athlon_fcom): Added amdfam10.
332 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
334 * config/i386/i386.md (x86_sahf_1, cmpfp_i_mixed, cmpfp_i_sse,
335 cmpfp_i_i387, cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387,
336 swapsi, swaphi_1, swapqi_1, swapdi_rex64, fix_truncsfdi_sse,
337 fix_truncdfdi_sse, fix_truncsfsi_sse, fix_truncdfsi_sse,
338 x86_fldcw_1, floatsisf2_mixed, floatsisf2_sse, floatdisf2_mixed,
339 floatdisf2_sse, floatsidf2_mixed, floatsidf2_sse,
340 floatdidf2_mixed, floatdidf2_sse, muldi3_1_rex64, mulsi3_1,
341 mulsi3_1_zext, mulhi3_1, mulqi3_1, umulqihi3_1, mulqihi3_insn,
342 umulditi3_insn, umulsidi3_insn, mulditi3_insn, mulsidi3_insn,
343 umuldi3_highpart_rex64, umulsi3_highpart_insn,
344 umulsi3_highpart_zext, smuldi3_highpart_rex64,
345 smulsi3_highpart_insn, smulsi3_highpart_zext, x86_64_shld,
346 x86_shld_1, x86_64_shrd, sqrtsf2_mixed, sqrtsf2_sse,
347 sqrtsf2_i387, sqrtdf2_mixed, sqrtdf2_sse, sqrtdf2_i387,
348 sqrtextendsfdf2_i387, sqrtxf2, sqrtextendsfxf2_i387,
349 sqrtextenddfxf2_i387): Added amdfam10_decode.
351 * config/i386/athlon.md (athlon_idirect_amdfam10,
352 athlon_ivector_amdfam10, athlon_idirect_load_amdfam10,
353 athlon_ivector_load_amdfam10, athlon_idirect_both_amdfam10,
354 athlon_ivector_both_amdfam10, athlon_idirect_store_amdfam10,
355 athlon_ivector_store_amdfam10): New define_insn_reservation.
356 (athlon_idirect_loadmov, athlon_idirect_movstore): Added
359 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
361 * config/i386/athlon.md (athlon_call_amdfam10,
362 athlon_pop_amdfam10, athlon_lea_amdfam10): New
363 define_insn_reservation.
364 (athlon_branch, athlon_push, athlon_leave_k8, athlon_imul_k8,
365 athlon_imul_k8_DI, athlon_imul_mem_k8, athlon_imul_mem_k8_DI,
366 athlon_idiv, athlon_idiv_mem, athlon_str): Added amdfam10.
368 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
370 * config/i386/athlon.md (athlon_sseld_amdfam10,
371 athlon_mmxld_amdfam10, athlon_ssest_amdfam10,
372 athlon_mmxssest_short_amdfam10): New define_insn_reservation.
374 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
376 * config/i386/athlon.md (athlon_sseins_amdfam10): New
377 define_insn_reservation.
378 * config/i386/i386.md (sseins): Added sseins to define_attr type
379 and define_attr unit.
380 * config/i386/sse.md: Set type attribute to sseins for insertq
383 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
385 * config/i386/athlon.md (sselog_load_amdfam10, sselog_amdfam10,
386 ssecmpvector_load_amdfam10, ssecmpvector_amdfam10,
387 ssecomi_load_amdfam10, ssecomi_amdfam10,
388 sseaddvector_load_amdfam10, sseaddvector_amdfam10): New
389 define_insn_reservation.
390 (ssecmp_load_k8, ssecmp, sseadd_load_k8, seadd): Added amdfam10.
392 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
394 * config/i386/athlon.md (cvtss2sd_load_amdfam10,
395 cvtss2sd_amdfam10, cvtps2pd_load_amdfam10, cvtps2pd_amdfam10,
396 cvtsi2sd_load_amdfam10, cvtsi2ss_load_amdfam10,
397 cvtsi2sd_amdfam10, cvtsi2ss_amdfam10, cvtsd2ss_load_amdfam10,
398 cvtsd2ss_amdfam10, cvtpd2ps_load_amdfam10, cvtpd2ps_amdfam10,
399 cvtsX2si_load_amdfam10, cvtsX2si_amdfam10): New
400 define_insn_reservation.
402 * config/i386/sse.md (cvtsi2ss, cvtsi2ssq, cvtss2si,
403 cvtss2siq, cvttss2si, cvttss2siq, cvtsi2sd, cvtsi2sdq,
404 cvtsd2si, cvtsd2siq, cvttsd2si, cvttsd2siq,
405 cvtpd2dq, cvttpd2dq, cvtsd2ss, cvtss2sd,
406 cvtpd2ps, cvtps2pd): Added amdfam10_decode attribute.
408 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
410 * config/i386/athlon.md (athlon_ssedivvector_amdfam10,
411 athlon_ssedivvector_load_amdfam10, athlon_ssemulvector_amdfam10,
412 athlon_ssemulvector_load_amdfam10): New define_insn_reservation.
413 (athlon_ssediv, athlon_ssediv_load_k8, athlon_ssemul,
414 athlon_ssemul_load_k8): Added amdfam10.
416 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
418 * config/i386/i386.h (TARGET_SSE_UNALIGNED_MOVE_OPTIMAL): New macro.
419 (x86_sse_unaligned_move_optimal): New variable.
421 * config/i386/i386.c (x86_sse_unaligned_move_optimal): Enable for
423 (ix86_expand_vector_move_misalign): Add code to generate movupd/movups
424 for unaligned vector SSE double/single precision loads for AMDFAM10.
426 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
428 * config/i386/i386.h (TARGET_AMDFAM10): New macro.
429 (TARGET_CPU_CPP_BUILTINS): Add code for amdfam10.
430 Define TARGET_CPU_DEFAULT_amdfam10.
431 (TARGET_CPU_DEFAULT_NAMES): Add amdfam10.
432 (processor_type): Add PROCESSOR_AMDFAM10.
434 * config/i386/i386.md: Add amdfam10 as a new cpu attribute to match
435 processor_type in config/i386/i386.h.
436 Enable imul peepholes for TARGET_AMDFAM10.
438 * config.gcc: Add support for --with-cpu option for amdfam10.
440 * config/i386/i386.c (amdfam10_cost): New variable.
441 (m_AMDFAM10): New macro.
442 (m_ATHLON_K8_AMDFAM10): New macro.
443 (x86_use_leave, x86_push_memory, x86_movx, x86_unroll_strlen,
444 x86_cmove, x86_3dnow_a, x86_deep_branch, x86_use_simode_fiop,
445 x86_promote_QImode, x86_integer_DFmode_moves,
446 x86_partial_reg_dependency, x86_memory_mismatch_stall,
447 x86_accumulate_outgoing_args, x86_arch_always_fancy_math_387,
448 x86_sse_partial_reg_dependency, x86_sse_typeless_stores,
449 x86_use_ffreep, x86_use_incdec, x86_four_jump_limit,
450 x86_schedule, x86_use_bt, x86_cmpxchg16b, x86_pad_returns):
451 Enable/disable for amdfam10.
452 (override_options): Add amdfam10_cost to processor_target_table.
453 Set up PROCESSOR_AMDFAM10 for amdfam10 entry in
454 processor_alias_table.
455 (ix86_issue_rate): Add PROCESSOR_AMDFAM10.
456 (ix86_adjust_cost): Add code for amdfam10.
458 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
460 * config/i386/i386.opt: Add new Advanced Bit Manipulation (-mabm)
461 instruction set feature flag. Add new (-mpopcnt) flag for popcnt
462 instruction. Add new SSE4A (-msse4a) instruction set feature flag.
463 * config/i386/i386.h: Add builtin definition for SSE4A.
464 * config/i386/i386.md: Add support for ABM instructions
466 * config/i386/sse.md: Add support for SSE4A instructions
467 (movntss, movntsd, extrq, insertq).
468 * config/i386/i386.c: Add support for ABM and SSE4A builtins.
469 Add -march=amdfam10 flag.
470 * config/i386/ammintrin.h: Add support for SSE4A intrinsics.
471 * doc/invoke.texi: Add documentation on flags for sse4a, abm, popcnt
473 * doc/extend.texi: Add documentation for SSE4A builtins.
475 2007-02-05 Bob Wilson <bob.wilson@acm.org>
477 * config/xtensa/xtensa.c (constantpool_mem_p): Skip over SUBREGs.
479 2007-02-05 Richard Guenther <rguenther@suse.de>
481 * tree-vectorizer.h (vectorizable_function): Add argument type
482 argument, change return type.
483 * tree-vect-patterns.c (vect_recog_pow_pattern): Adjust caller.
484 * tree-vect-transform.c (vectorizable_function): Handle extra
485 argument, return vectorized function decl.
486 (build_vectorized_function_call): Remove.
487 (vectorizable_call): Handle calls with result and argument types
488 differing. Handle loop vectorization factor correctly.
489 * targhooks.c (default_builtin_vectorized_function): Adjust for
491 * targhooks.h (default_builtin_vectorized_function): Likewise.
492 * target.h (builtin_vectorized_function): Add argument type
494 * config/i386/i386.c (ix86_builtin_vectorized_function): Handle
495 extra argument, allow vectorizing of lrintf.
496 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION): Adjust
497 documentation of target hook.
499 2007-02-05 Hans-Peter Nilsson <hp@axis.com>
502 * config/cris/cris.md ("*andsi_movu", "*andsi_clear", "*andhi_movu")
503 ("*andhi_clear", andu (casesi+45)): For size-changed operand where
504 memory is allowed, require !side_effects_p, not just !MEM_VOLATILE_P.
506 2007-02-05 Roger Sayle <roger@eyesopen.com>
508 * fold-const.c (fold_unary) <REAL_PART>: Test for availability of
509 BUILT_IN_COS before simplifying REAL_PART(CEXPI)) to COS.
510 <IMAG_PART>: Likewise, check for availability of BUILT_IN_SIN.
511 * builtins.c (fold_builtin_sincos): Check for TARGET_C99_FUNCTIONS
512 before canonicalizing sincos to cexpi.
513 (fold_builtin_cexp): Likewise, for canonicalizing cexp to cexpi.
515 2007-02-05 Roger Sayle <roger@eyesopen.com>
517 * config/alpha/alpha.c (alpha_add_builtins): New Helper function.
518 Set TREE_READONLY and TREE_NOTHROW directly, not via attributes.
519 (alpha_init_builtins): Use alpha_add_builtins to process tables.
521 2007-02-05 Roger Sayle <roger@eyesopen.com>
523 * mips-tfile.c (initialize_init_file): Correct endianness test.
525 2007-02-05 Kazu Hirata <kazu@codesourcery.com>
527 * config/m68k/m68k.md (pushdi-1, pushdi, movsi+1): Don't use
530 2007-02-05 Richard Sandiford <richard@codesourcery.com>
532 * dwarf2out.c (dwarf2out_frame_debug_expr): Record the register
533 saves in a PARALLEL before the register assignments.
535 2007-02-05 Richard Sandiford <richard@codesourcery.com>
537 * doc/tm.texi (DWARF_ALT_FRAME_RETURN_COLUMN): Do not require
538 DWARF_FRAME_RETURN_COLUMN to be a general register.
539 * dwarf2out.c (init_return_column_size): New function, split from...
540 (expand_builtin_init_dwarf_reg_sizes): ...here. Allow both
541 DWARF_FRAME_RETURN_COLUMN and DWARF_ALT_FRAME_RETURN_COLUMN
542 to be nongeneral registers.
543 * config/m68k/m68k.h (DWARF_FRAME_REGNUM): Only map FP and
545 (DWARF_FRAME_REGISTERS, DWARF_FRAME_RETURN_COLUMN): Define.
546 (DWARF_ALT_FRAME_RETURN_COLUMN): Define.
548 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
550 * cfgcleanup.c (try_optimize_cfg): Avoid removing ENTRY_BLOCK_PTR.
552 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
554 * cfgloopmanip.c (loop_delete_branch_edge): Removed.
555 (remove_path): Use can_remove_branch_p and remove_branch instead
556 of loop_delete_branch_edge.
557 * tree-ssa-loop-manip.c (scale_dominated_blocks_in_loop): New function.
558 (tree_transform_and_unroll_loop): Remove dead branches immediately.
559 Update profile using scale_dominated_blocks_in_loop.
560 * cfghooks.c (can_remove_branch_p, remove_branch): New functions.
561 * cfghooks.h (struct cfg_hooks): Add can_remove_branch_p.
562 (can_remove_branch_p, remove_branch): Declare.
563 * tree-cfg.c (tree_can_remove_branch_p): New function.
564 (tree_cfg_hooks): Add tree_can_remove_branch_p.
565 * cfgrtl.c (rtl_can_remove_branch_p): New function.
566 (rtl_cfg_hooks, cfg_layout_rtl_cfg_hook): Add rtl_can_remove_branch_p.
568 2007-02-05 Jan Hubicka <jh@suse.cz>
571 * ipa-inline.c (cgraph_clone_inlined_nodes): When there are unanalyzed
572 nodes in cgraph, don't remove offline copy of the function.
574 2007-02-04 Jan Hubicka <jh@suse.cz>
576 * tree-sra.c (sra_walk_expr): Add linebreaks. BITFIELD_REFs into
577 vectors might cause maybe_lookup_element_for_expr to be called
578 on non-sra-candidate.
580 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
582 * config/bfin/bfin-modes.def, config/bfin/bfin.c,
583 config/bfin/bfin.md, config/bfin/predicates.md: Follow
584 spelling conventions.
586 2007-02-04 Richard Guenther <rguenther@suse.de>
589 * fold-const.c (try_move_mult_to_index): Make sure to not
590 overflow one dimension of a multi-dimensional array access.
592 2007-02-04 Jan Hubicka <jh@suse.cz>
594 * passes.c (init_optimization_passes): Reindent.
596 2007-02-04 Jan Hubicka <jh@suse.cz>
597 Eric Botcazou <ebotcazou@adacore.com>
599 * tree-optimize.c (has_abnormal_outgoing_edge_p): Move to...
600 (execute_fixup_cfg): Break out the abnormal goto code.
601 * tree-inline.c (has_abnormal_outgoing_edge_p): ...here.
602 (make_nonlocal_label_edges): Move here from execute_fixup_cfg.
603 (optimize_inline_calls): Call make_nonlocal_label_edges.
605 2007-02-04 Jan Hubicka <jh@suse.cz>
607 * tree-ssa-copyrename.c (copy_rename_partition_coalesce): Return
608 true when something was changed.
609 (rename_ssa_copies): When something was changed, do
610 TODO_remove_unused_locals.
611 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
612 add TODO_remove_unused_locals when instruction was removed.
614 2007-02-04 Jan Hubicka <jh@suse.cz>
616 * ipa-inline.c (try_inline): Improve debug output; work on already
618 (cgraph_decide_inlining_incrementally): Indent; improve debug output;
619 call try_inline for already inlined edges too when flattening;
620 inline also functions that make callee growth but overall unit size
623 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
625 * config/m32c/bitops.md, config/m32c/jump.md,
626 config/m32c/m32c.c, config/m32c/m32c.h, config/m32r/m32r.c,
627 config/m32r/m32r.h, config/m32r/m32r.md,
628 config/m32r/predicates.md, config/m68hc11/larith.asm,
629 config/m68hc11/m68hc11.c, config/m68hc11/m68hc11.h,
630 config/m68k/m68k.h, config/mcore/mcore.md, config/mips/4k.md,
631 config/mips/mips-protos.h, config/mips/mips.c,
632 config/mips/mips.h, config/mips/mips.md, config/mips/mips16.S,
633 config/mn10300/mn10300.h, config/mn10300/predicates.md,
634 config/mt/mt.c, config/mt/mt.h, config/mt/mt.md: Follow
635 spelling conventions.
637 * config/v850/v850.c, config/v850/v850.h, config/v850/v850.md:
638 Follow spelling conventions.
640 2007-02-03 Douglas Gregor <doug.gregor@gmail.com>
642 * c-opts.c (c_common_post_options): If C++0x mode is enabled, don't
643 warn about C++0x compatibility.
645 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
647 * config/h8300/h8300.c, config/h8300/h8300.h,
648 config/h8300/h8300.md: Follow spelling conventions.
650 2007-02-03 Uros Bizjak <ubizjak@gmail.com>
653 * combine.c (try_combine): Do not substitute source operand
654 with constants wider than 2 * HOST_BITS_PER_WIDE_INT.
656 2007-02-03 Jan Hubicka <jh@suse.cz>
658 PR gcov-profile/30650
659 * value-prof.c (stringop_block_profile): Fix handling of size counter;
660 do not divide by zero for never executed counters.
661 (tree_find_values_to_profile): Fix counters.
662 * gcov-ui.h (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): Fix comments.
664 2007-02-03 Ian Lance Taylor <iant@google.com>
666 * lower-subreg.c (simple_move_operand): New static function,
667 broken out of simple_move. Reject LABEL_REF, SYMBOL_REF, and HIGH
669 (simple_move): Call simple_move_operand.
670 (find_decomposable_subregs): Add special handling of MEMs.
671 (can_decompose_p): Rename from cannot_decompose_p. Reverse
672 meaning of return value. If we see a hard register, test whether
673 it can store a word_mode value. Change all callers.
675 2007-02-03 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
677 * pa.md (addvdi3, addvsi3, subvdi3, subvsi3, negvdi2, negvsi2): New
678 ftrapv insns and expanders.
679 (subdi3): Change define_expand operand 1 to arith11_operand, and
680 operand 2 to reg_or_0_operand. Change constraints of 64-bit insn
681 pattern to handle reg_or_0 operands. Revise 32-bit insn pattern to
682 handle 11-bit constants and reg_or_0 operands in operands 1 and 2,
686 * varasm.c (notice_global_symbol): Treat global objects as weak when
689 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
691 * emit-rtl.c (dconstpi): Delete.
693 (init_emit_once): Delete dconstpi and init dconstsqrt2.
694 * real.h (dconstpi): Delete.
696 * builtins.c (fold_builtin_cabs): Use dconstsqrt2.
697 (fold_builtin_hypot): Likewise.
699 2007-02-03 Tom Tromey <tromey@redhat.com>
702 * gcc.c (cpp_unique_options): Any of -ggdb3, -gstabs3,
703 -gcoff3, -gxcoff3, -gvms3 implies -dD.
705 2007-02-03 Kazu Hirata <kazu@codesourcery.com>
707 * c-decl.c, config/avr/avr.c, config/avr/avr.h,
708 config/m68k/m68k.c, config/m68k/netbsd-elf.h,
709 config/mn10300/mn10300.c, config/pdp11/pdp11.h,
710 config/rs6000/cell.md, config/rs6000/darwin.h,
711 config/sh/sh.md, config/sh/sh4-300.md, config/spu/spu.c,
712 config/spu/spu.md, cselib.c, expr.c, haifa-sched.c, hwint.h,
713 jump.c, reload.c, sched-deps.c, sched-int.h, tree-inline.c,
714 tree-profile.c, tree-ssa-live.h, tree-vrp.c: Fix comment
715 typos. Follow spelling conventions.
716 * doc/invoke.texi: Follow spelling conventions.
718 2007-02-03 Roger Sayle <roger@eyesopen.com>
720 * simplify-rtx.c (simplify_relational_operation_1): Implement some
721 canonicalization transformations that attempt to simplify integer
722 constant comparisons to become comparisons against zero.
724 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
726 * builtins.c (fold_builtin_cabs): Fold cabs(x+xi) into
728 * fold-const.c (fold_binary): Fix comment typos. Fold complex
729 (x,0)-(0,y) into (x,-y). Likewise (0,y)-(x,0) into (-x,y).
731 2007-02-02 Mike Stump <mrs@apple.com>
733 * config/darwin9.h (DARWIN_LINKER_GENERATES_ISLANDS): Add.
734 * config/rs6000/rs6000.c (DARWIN_GENERATE_ISLANDS): Add.
735 (output_call): Use DARWIN_GENERATE_ISLANDS to decide when to
736 generate a branch island.
738 2007-02-02 Bob Wilson <bob.wilson@acm.org>
740 * config/xtensa/xtensa.c (smalloffset_mem_p): Use BASE_REG_P.
741 (xtensa_legitimate_address_p): New.
742 (xtensa_legitimize_address): New.
743 (xtensa_output_addr_const_extra): New.
744 * config/xtensa/xtensa.h (REG_OK_STRICT_FLAG): Define.
746 (REG_OK_FOR_BASE_P): Use BASE_REG_P.
747 (GO_IF_LEGITIMATE_ADDRESS): Move code to xtensa_legitimate_address_p.
748 (LEGITIMIZE_ADDRESS): Move code to xtensa_legitimize_address.
749 (OUTPUT_ADDR_CONST_EXTRA): Move code to xtensa_output_addr_const_extra.
750 * config/xtensa/xtensa-protos.h (xtensa_legitimate_address_p): New.
751 (xtensa_legitimize_address): New.
752 (xtensa_output_addr_const_extra): New.
754 2007-02-02 Steve Ellcey <sje@cup.hp.com>
756 * config/ia64/ia64.c (ia64_print_operand): Fix compare strings.
758 2007-02-02 Ian Lance Taylor <iant@google.com>
760 * expmed.c (expand_divmod): Add comment.
762 2007-02-02 Kazu Hirata <kazu@codesourcery.com>
764 * emit-rtl.c (renumber_insns): Remove.
765 * flags.h: Remove the extern for flag_renumber_insns.
766 * rtl.h: Remove the prototype for renumber_insns.
767 * toplev.c (flag_renumber_insns): Remove.
769 2007-02-02 Hui-May Chang <hm.chang@apple.com>
771 Revert for x86 darwin:
772 2005-06-19 Uros Bizjak <uros@kss-loka.si>
774 * config/i386/i386.c (ix86_function_arg_regno_p): Put back the
775 code before the following patch under TARGET_MACHO.
776 (ix86_function_value_regno_p): Likewise.
778 2007-02-02 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
780 * fold-const.c (negate_expr_p, fold_negate_expr): Handle
783 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
786 * config/ia64/ia64.c (ia64_speculate_insn): Restrict to memory loads to
787 general or fp registers. Add comments.
788 * config/ia64/ia64.md (reg_pred_prefix): Add comment.
790 2007-02-02 Paolo Bonzini <bonzini@gnu.org>
792 * pointer-set.c (insert_aux): Only return insertion slot.
793 (pointer_set_insert): Adjust.
794 (pointer_set_traverse, struct pointer_map_t, pointer_map_create,
795 pointer_map_destroy, pointer_map_insert, pointer_map_contains,
796 pointer_map_traverse): New.
797 * pointer-set.h (pointer_set_traverse, struct pointer_map_t,
798 pointer_map_create, pointer_map_destroy, pointer_map_insert,
799 pointer_map_contains, pointer_map_traverse): Declare.
801 2007-02-02 Jakub Jelinek <jakub@redhat.com>
804 * builtins.c (fold_builtin_sprintf): Do not attempt to optimize
805 sprintf (str, "%s"). Do not optimize sprintf (str, "nopercent", p++).
807 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
809 * sched-int.h (ds_to_dk, dk_to_ds): Declare functions.
811 (struct _dep): New type.
812 (dep_t): New typedef.
813 (DEP_PRO, DEP_CON, DEP_KIND): New access macros.
814 (DEP_STATUS): New access macro. The macro with the same name was
815 renamed to DEP_LINK_STATUS.
816 (dep_init): Declare function
818 (struct _dep_link): New type.
819 (dep_link_t): New typedef.
820 (DEP_LINK_NODE, DEP_LINK_NEXT, DEP_LINK_PREV_NEXTP): New access macros.
821 (DEP_LINK_DEP, DEP_LINK_PRO, DEP_LINK_CON, DEP_LINK_KIND): New macros.
822 (DEP_LINK_STATUS): New macro.
823 (debug_dep_links): New debug function.
825 (struct _deps_list): New type.
826 (deps_list_t): New typedef.
827 (DEPS_LIST_FIRST): New access macro.
828 (FOR_EACH_DEP_LINK): New cycle macro.
829 (create_deps_list, free_deps_list, delete_deps_list): Declare
831 (deps_list_empty_p, debug_deps_list, add_back_dep_to_deps_list): Ditto.
832 (find_link_by_pro_in_deps_list, find_link_by_con_in_deps_list): Ditto.
833 (copy_deps_list_change_con): Ditto.
835 (move_dep_link): Declare function.
837 (struct _dep_node): New type.
838 (dep_node_t): New typedef.
839 (DEP_NODE_BACK, DEP_NODE_DEP, DEP_NODE_FORW): New access macros.
841 (struct haifa_insn_data.back_deps): New field to hold backward
842 dependencies of the insn.
843 (struct haifa_insn_data.depend): Rename to forw_deps. Change its type
845 (struct haifa_insn_data.resolved_deps): Rename to resolved_back_deps.
846 Change its type to deps_list_t.
847 (INSN_BACK_DEPS): New access macro to use instead of LOG_LINKS.
848 (INSN_DEPEND): Rename to INSN_FORW_DEPS.
849 (RESOLVED_DEPS): Rename to INSN_RESOLVED_BACK_DEPS.
851 (INSN_COST): Move to haifa-sched.c. Use insn_cost () instead.
853 (DEP_STATUS): Rename to DEP_LINK_STATUS. Fix typo in the comment.
855 (add_forw_dep, delete_back_forw_dep, insn_cost): Update declaration and
859 * sched-deps.c (CHECK): New macro to (en/dis)able sanity checks.
860 (ds_to_dk, dk_to_ds): New functions.
862 (init_dep_1): New static function.
863 (init_dep): New function.
864 (copy_dep): New static function.
866 (dep_link_consistent_p, attach_dep_link, add_to_deps_list): New static
868 (detach_dep_link): New static function.
869 (move_dep_link): New function.
871 (dep_links_consistent_p, dump_dep_links): New static functions.
872 (debug_dep_links): New debugging function.
874 (deps_obstack, dl_obstack, dn_obstack): New static variables.
876 (alloc_deps_list, init_deps_list): New static functions.
877 (create_deps_list): New function.
878 (clear_deps_list): New static function.
879 (free_deps_list, delete_deps_list, deps_list_empty_p): New functions.
880 (deps_list_consistent_p, dump_deps_list): New static functions.
881 (debug_deps_list): New function.
882 (add_back_dep_to_deps_list, find_link_by_pro_in_deps_list): New
884 (find_link_by_con_in_deps_list, copy_deps_list_change_con): Ditto.
886 (maybe_add_or_update_back_dep_1, add_or_update_back_dep_1): Update to
887 use new scheduler dependencies lists.
888 (add_back_dep, delete_all_dependences, fixup_sched_groups): Ditto.
889 (sched_analyze): Ditto. Initialize dependencies lists.
890 (add_forw_dep, compute_forward_dependences): Update to use new
891 scheduler dependencies lists.
893 (init_dependency_caches): Init deps_obstack.
894 (free_dependency_caches): Free deps_obstack.
896 (adjust_add_sorted_back_dep, adjust_back_add_forw_dep): Update to use
897 new scheduler dependencies lists.
898 (delete_forw_dep, add_or_update_back_forw_dep): Ditto.
899 (add_back_forw_dep, delete_back_forw_dep): Ditto.
901 * sched-rgn.c (set_spec_fed, find_conditional_protection, is_pfree):
902 Update to use new scheduler dependencies lists.
903 (is_conditionally_protected, is_prisky, add_branch_dependences): Ditto.
904 (debug_dependencies): Ditto.
905 (schedule_region): Update comments.
907 * sched-ebb.c (earliest_block_with_similiar_load): Update to use new
908 scheduler dependencies lists.
909 (schedule_ebb): Update comments.
911 * rtl.def (DEPS_LIST): Remove.
913 * lists.c (unused_deps_list): Remove.
914 (free_list): Update assertions.
916 (alloc_DEPS_LIST, free_DEPS_LIST_list, free_DEPS_LIST_node): Remove.
917 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
919 * rtl.h (free_DEPS_LIST_list, alloc_DEPS_LIST): Remove declarations.
920 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
922 * haifa-sched.c (comments): Update.
923 (insn_cost1): Remove. Inline the code into insn_cost ().
924 (insn_cost): Update to use new scheduler dependencies lists. Move
925 processing of the dependency cost to dep_cost ().
926 (dep_cost): New function. Use it instead of insn_cost () when
927 evaluating cost of the dependency. Use compatible interface to
928 interact with the target.
929 (priority): Update to use new scheduler dependencies lists.
930 (rank_for_schedule): Ditto. Optimize heuristic that prefers the insn
931 with greater number of insns that depend on the insn.
932 (schedule_insn): Update to use new scheduler dependencies lists. Add
933 code to free backward dependencies lists. Inline and optimize code
934 from resolve_dep () - see PR28071.
935 (ok_for_early_queue_removal): Update to use new scheduler dependencies
936 lists. Update call to targetm.sched.is_costly_dependence hook.
938 (fix_inter_tick, try_ready, fix_tick_ready): Update to use new
939 scheduler dependencies lists.
941 (resolve_dep): Remove. Move the logic to schedule_insn ().
942 (init_h_i_d): Initialize dependencies lists.
944 (process_insn_depend_be_in_spec): Rename to
945 process_insn_forw_deps_be_in_spec. Update to use new scheduler
947 (add_to_speculative_block, create_check_block_twin, fix_recovery_deps):
948 Update to use new scheduler dependencies lists.
949 (clear_priorities, calc_priorities, add_jump_dependencies): Ditto.
951 * ddg.c (create_ddg_dependence, create_ddg_dep_no_link): Update to use
952 new scheduler dependencies lists.
953 (build_intra_loop_deps): Ditto.
955 * target.h (struct _dep): Declare to use in
956 gcc_target.sched.is_costly_dependence.
957 (struct gcc_target.sched.adjust_cost): Fix typo.
958 (struct gcc_target.sched.is_costly_dependence): Change signature to use
959 single dep_t parameter instead of an equivalent triad.
960 (struct gcc_target.sched.adjust_cost_2): Remove.
962 * target-def.h (TARGET_SCHED_ADJUST_COST_2): Remove.
964 * reg-notes.def (DEP_TRUE, DEP_OUTPUT, DEP_ANTI): Update comments.
966 * doc/tm.texi (TARGET_SCHED_IS_COSTLY_DEPENDENCE): Update
968 (TARGET_SCHED_ADJUST_COST_2): Remove documentation.
970 * doc/rtl.texi (LOG_LINKS): Remove part about instruction scheduler.
971 (REG_DEP_TRUE): Document.
973 * config/ia64/ia64.c (ia64_adjust_cost_2): Rename to ia64_adjust_cost.
974 Change signature to correspond to the targetm.sched.adjust_cost hook.
975 Update use in TARGET_SCHED_ADJUST_COST_2.
976 (TARGET_SCHED_ADJUST_COST_2): Rename to TARGET_SCHED_ADJUST_COST.
977 (ia64_dependencies_evaluation_hook, ia64_dfa_new_cycle): Update to use
978 new scheduler dependencies lists.
979 (ia64_gen_check): Ditto.
981 * config/mips/mips.c (vr4130_swap_insns_p): Update to use new scheduler
984 * config/rs6000/rs6000.c (rs6000_is_costly_dependence): Change
985 signature to correspond to the targetm.sched.is_costly_dependence hook.
986 (is_costly_group): Update to use new scheduler dependencies lists.
988 * config/spu/spu.c (spu_sched_adjust_cost): Use insn_cost () function
989 instead of INSN_COST () macro.
991 2007-02-01 Ian Lance Taylor <iant@google.com>
993 * lower-subreg.c (resolve_clobber): Handle a subreg of a concatn.
995 2007-02-01 Guy Martin <gmsoft@gentoo.org>
997 * pa.md (tp_load): Correct mfctl instruction syntax.
999 2007-02-01 Geoffrey Keating <geoffk@apple.com>
1001 * config/rs6000/rs6000.c (rs6000_stack_info): Correct
1002 altivec_padding_size calculation on AIX. Improve comment, add
1003 assert to verify that it's right.
1005 * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
1008 2007-2-01 Seongbae Park <seongbae.park@gmail.com>
1011 * global.c (global_alloc): Add mising initialization of
1014 2007-02-01 Roger Sayle <roger@eyesopen.com>
1016 * alias.c (init_alias_analysis): Correct whitespace.
1017 * bb-reorder.c (fix_edges_for_rarely_executed_code,
1018 partition_hot_cold_basic_blocks): Likewise.
1019 * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
1020 expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
1021 maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
1022 fold_builtin_snprintf_chk, fold_builtin_printf,
1023 fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
1024 do_mpfr_arg3, do_mpfr_sincos): Likewise.
1025 * cfgcleanup.c (cleanup_cfg): Likewise.
1026 * cfgexpand.c (tree_expand_cfg): Likewise.
1027 * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
1028 * function.c (get_next_funcdef_no): Likewise.
1029 * gengtype.c (main): Likewise.
1030 * genmodes.c (main): Likewise.
1031 * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
1032 * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
1033 * ifcvt.c (noce_emit_move_insn): Likewise.
1034 * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
1036 * stor-layout.c (get_best_mode): Likewise.
1037 * tree-ssa-loop-niter.c (get_val_for): Likewise.
1038 * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
1039 scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
1040 perform_var_substitution, solve_graph): Likewise.
1041 * tree-vrp.c (vrp_finalize): Likewise.
1043 2007-02-01 Ian Lance Taylor <iant@google.com>
1045 * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
1046 high part of a paradoxical subreg, return a constant zero.
1048 2007-02-01 Zdenek Dvorak <dvorakz@suse.cz>
1050 * toplev.c (lang_dependent_init): Call init_set_costs.
1051 * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
1053 2007-02-01 Richard Guenther <rguenther@suse.de>
1056 * fold-const.c (fold_negate_expr): Allow negating a
1057 constant if overflow does not change.
1059 2007-02-01 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1061 * doc/c-tree.texi (Expression trees): Improve markup.
1062 * doc/tm.texi (Register Classes, Addressing Modes)
1063 (Floating Point): Fix spacing after abbreviations. Fix some
1066 2007-02-01 Ben Elliston <bje@au.ibm.com>
1068 * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
1069 * doc/contrib.texi: Likewise.
1070 * doc/install.texi: Likewise.
1072 2007-01-31 Richard Henderson <rth@redhat.com>
1073 Ian Lance Taylor <iant@google.com>
1075 * lower-subreg.c: New file.
1076 * rtl.def (CONCATN): Define.
1077 * passes.c (init_optimization_passes): Add pass_lower_subreg and
1079 * emit-rtl.c (update_reg_offset): New static function, broken out
1080 of gen_rtx_REG_offset.
1081 (gen_rtx_REG_offset): Call update_reg_offset.
1082 (gen_reg_rtx_offset): New function.
1083 * regclass.c: Revert patch of 2006-03-05, restoring
1085 (clear_reg_info_regno): New function.
1086 * dwarf2out.c (concatn_loc_descriptor): New static function.
1087 (loc_descriptor): Handle CONCATN.
1088 * common.opt (fsplit_wide_types): New option.
1089 * opts.c (decode_options): Set flag_split_wide_types when
1091 * timevar.def (TV_LOWER_SUBREG): Define.
1092 * rtl.h (gen_reg_rtx_offset): Declare.
1093 (reg_scan_update): Declare.
1094 * regs.h (clear_reg_info_regno): Declare.
1095 * tree-pass.h (pass_lower_subreg): Declare.
1096 (pass_lower_subreg2): Declare.
1097 * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
1098 (Optimize Options): Add -fsplit-wide-types to -O1 list. Document
1100 * doc/rtl.texi (Regs and Memory): Document concat and concatn.
1101 * Makefile.in (OBJS-common): Add lower-subreg.o.
1102 (lower-subreg.o): New target.
1104 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1106 * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
1108 2007-01-31 Anatoly Sokolov <aesok@post.ru>
1111 * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
1113 2007-01-31 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1116 * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
1118 2007-01-31 Zdenek Dvorak <dvorakz@suse.cz>
1120 * cfgloop.h: Include vec-prim.h.
1121 (enum li_flags): Remove LI_ONLY_OLD.
1122 (loop_iterator): Changed.
1123 (fel_next, fel_init): Iterate over loop tree.
1124 (FOR_EACH_LOOP_BREAK): New macro.
1125 * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
1127 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
1128 * modulo-sched.c (sms_schedule): Ditto.
1129 * tree-vectorizer.c (vectorize_loops): Ditto.
1130 * doc/loop.texi: Update information on loop numbering and behavior of
1131 FOR_EACH_LOOP wrto new loops.
1132 * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
1133 add_to_evolution_1): Test nestedness of loops instead of comparing
1135 * tree-chrec.c (chrec_fold_plus_poly_poly,
1136 chrec_fold_multiply_poly_poly, chrec_evaluate,
1137 hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
1138 reset_evolution_in_loop): Ditto.
1139 * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
1141 2007-01-31 Dirk Mueller <dmueller@suse.de>
1143 * c-common.c (warn_about_parentheses): Separate warning about
1144 un-parenthized sequence of comparison operators from the one
1145 which is supposed to warn about x <= y <= z.
1147 2007-01-31 Uros Bizjak <ubizjak@gmail.com>
1149 * optabs.h (enum optab_index): Add new OTI_isinf.
1150 (isinf_optab): Define corresponding macro.
1151 * optabs.c (init_optabs): Initialize isinf_optab.
1152 * genopinit.c (optabs): Implement isinf_optab using isinf?f2
1154 * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
1155 (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
1157 (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
1158 expand_builtin_interclass_mathfn.
1159 * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
1160 * config/i386/i386.md (UNSPEC_FXAM): New constant.
1161 (fxam<mode>2_i387): New insn pattern.
1162 (isinf<mode>2) New expander to implement isinf, isinff and isinfl
1163 built-in functions as x87 inline asm.
1165 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1167 * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
1169 2007-01-30 Eric Christopher <echristo@apple.com>
1171 * config.gcc: Add geode.
1173 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1175 * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
1176 ipa-inline.c, tree-profile.c, tree-ssa-live.c,
1177 tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
1178 value-prof.c: Fix comment typos.
1180 2007-01-30 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1183 * doc/invoke.texi (Wpointer-arith): Document warning.
1185 2007-01-30 Janis Johnson
1187 * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
1188 the list of discrepancies from the draft TR.
1190 2007-01-30 Dirk Mueller <dmueller@suse.de>
1193 * doc/invoke.texi (-Wreturn-type): Update description to
1196 2007-01-30 Richard Sandiford <richard@codesourcery.com>
1198 * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
1200 2007-01-30 Uros Bizjak <ubizjak@gmail.com>
1202 * builtins.c (expand_builtin_int_interclass_roundingfn): New function
1203 to handle optabs that operate on floating point input argument and
1204 output to integer output.
1205 (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
1206 (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
1207 (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
1208 expand_builtin_interclass_mathfn ().
1209 * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
1211 (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
1212 ilogbl built-in functions as x87 intrinsics.
1214 2007-01-30 Richard Guenther <rguenther@suse.de>
1217 * dwarf2out.c (reference_to_unused): Query varpool if the
1218 variable was output.
1220 2007-01-30 Richard Guenther <rguenther@suse.de>
1223 * passes.c (execute_one_pass): Reset in_gimple_form to not
1224 confuse non-unit-at-a-time mode.
1226 2007-01-29 Roger Sayle <roger@eyesopen.com>
1227 Richard Guenther <rguenther@suse.de>
1229 * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
1230 avoid undefined behaviour on overflow. Use force_fit_type_double
1231 to construct the constant with the specified TREE_OVERFLOW.
1233 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1235 * config/dfp-bit.c: Add parameterized support for fp exceptions.
1236 * config/dfp-bit.h: Ditto.
1238 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1240 * c-decl.c (pop_scope): Replace warnings with call to
1241 warn_for_unused_label.
1242 * c-common.h (warn_for_unused_label): Declare.
1243 * c-common.c (warn_for_unused_label): Define.
1245 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1247 * tree-optimize.c (update_inlined_to_pointers): Delete unused
1250 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1252 * Makefile.in (USER_H): Remove decfloat.h.
1253 * ginclude/decfloat.h: Delete, moving contents to ...
1254 * ginclude/float.h: Add support for decimal floating point,
1255 guarded by __STDC_WANT_DEC_FP__.
1257 2007-01-29 Mike Stump <mrs@apple.com>
1259 * doc/gccint.texi (Top): Rename Loop Representation to Loop
1260 Analysis and Representation to resolve case insensitive conflict.
1261 * doc/loop.texi (Loop Analysis and Representation): Likewise.
1263 2007-01-28 Daniel Berlin <dberlin@dberlin.org>
1265 * tree.h (struct tree_memory_tag): Add aliases member.
1266 (MTAG_ALIASES): New macro.
1267 * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
1268 (add_may_alias): Remove pointer-set. Update for may_aliases being
1270 (mark_aliases_call_clobbered): Update for may_aliases being a
1272 (compute_tag_properties): Ditto.
1273 (create_partition_for): Ditto.
1274 (compute_memory_partitions): Ditto.
1275 (dump_may_aliases_for): Ditto.
1276 (is_aliased_with): Ditto.
1277 (add_may_alias_for_new_tag): Ditto.
1278 (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
1279 (compute_is_aliased): New function.
1280 (compute_may_aliases): Call compute_is_aliased).
1281 (init_alias_info): Initialize alias_bitmap_obstack.
1282 (union_alias_set_into): New function.
1283 (compute_flow_sensitive_aliasing): Use union_aliases_into.
1284 (have_common_aliases_p): Rewrite to take two bitmaps and use
1286 (compute_flow_insensitive_aliasing): Stop using pointer-sets.
1288 (finalize_ref_all_pointers): Update for add_may_alias changes.
1289 (new_type_alias): Ditto.
1290 * tree-flow-inline.h (may_aliases): Return a bitmap.
1291 * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
1292 * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
1293 may_aliases being a bitmap.
1294 * tree-flow.h (struct var_ann_d): Remove may_aliases member.
1295 may_aliases now returns a bitmap.
1296 * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
1298 * tree-ssa-operands.c (add_virtual_operand): Update for
1299 may_aliases being a bitmap.
1301 2007-01-29 Daniel Berlin <dberlin@dberlin.org>
1303 PR tree-optimization/30630
1304 * tree-ssa-structalias.c (do_complex_constraint): Mark correct
1305 variable as changed.
1307 2007-01-29 Simon Martin <simartin@users.sourceforge.net>
1310 * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
1311 expanded only once even if an error occurs.
1313 2007-01-29 Ben Elliston <bje@au.ibm.com>
1315 * gcov-io.h (__gcov_indirect_call_profiler): Declare.
1316 (__gcov_average_profiler): Likewise.
1317 (__gcov_ior_profiler): Likewise.
1318 (__gcov_merge_ior): Likewise.
1320 2007-01-28 Jan Hubicka <jh@suse.cz>
1322 * builtins.c (expand_builtin_memset): Fix typo in my last patch.
1323 * value-prof.c (stringop_block_profile): Likewise.
1325 2007-01-28 Jan Hubicka <jh@suse.cz>
1327 * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
1328 variant handling histograms; add wrapper.
1329 (clear_storage_via_libcall): Export.
1330 (emit_block_move_hints): Break out from ...; add histograms.
1331 (emit_block_move): ... this one.
1332 (clear_storage_hints): Break out from ...; add histograms.
1333 (clear_storage): ... this one.
1334 (set_storage_via_memset): Handle histogram.
1335 * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
1336 (clear_storage_hints, clear_storage_via_libcall): Declare.
1337 (set_storage_via_setmem): Update prototype.
1338 * doc/md.texi (movmem, setmem): Document new arguments.
1340 * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
1342 (stringop_block_profile): New global function.
1343 (tree_stringops_values_to_profile): Profile block size and alignment.
1344 * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
1346 (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
1347 (stringop_block_profile): Declare.
1348 * builtins.c: Include value-prof.h.
1349 (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
1350 * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
1351 (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
1352 (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
1353 * profile.c (instrument_values): Add new counters.
1354 * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
1356 * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
1357 (tree_init_edge_profiler): Build new profilers.
1358 (tree_gen_average_profiler, tree_gen_ior_profiler): New.
1359 (pass_tree_profile): Add dump.
1360 (tree_profile_hooks): Update.
1361 * Makefile.in (LIBGCOV): Add new constants.
1362 * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
1363 __gcov_ior_profiler): New.
1364 * i386.md (movmem/setmem expanders): Add new optional arguments.
1366 2007-01-28 David Edelsohn <edelsohn@gnu.org>
1368 * doc/md.texi (Standard Pattern Names): Document blockage pattern.
1370 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1372 * tree-ssa-loop-unswitch.c: Include tree-inline.h.
1373 (tree_unswitch_single_loop): Pass eni_size_weights to
1374 tree_num_loop_insns.
1375 * tree-ssa-loop-manip.c: Include tree-inline.h.
1376 (can_unroll_loop_p): Pass eni_size_weights to
1377 tree_num_loop_insns.
1378 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
1379 Pass eni_size_weights to estimate_num_insns.
1380 * tree.h (init_inline_once): Export.
1381 * toplev.c (backend_init): Call init_inline_once.
1382 * cgraphunit.c (cgraph_process_new_functions,
1383 cgraph_analyze_function): Pass eni_inlining_weights to
1385 * ipa-inline.c (compute_inline_parameters): Ditto.
1386 * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
1387 to estimate_num_insns.
1388 (try_unroll_loop_completely): Pass eni_size_weights to
1389 tree_num_loop_insns.
1390 * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
1391 ot estimate_num_insns.
1392 * tree-ssa-loop-prefetch.c: Include tree-inline.h.
1393 (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
1394 * tree-inline.c (eni_inlining_weights, eni_size_weights,
1395 eni_time_weights): New variables.
1396 (init_inline_once): Initialize them.
1397 (struct eni_data): Mew.
1398 (estimate_num_insns_1, estimate_num_insns): Use weights.
1399 * tree-inline.h (struct eni_weights_d): New.
1400 (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
1401 (estimate_num_insns): Declaration changed.
1402 * cfgloop.h (tree_num_loop_insns): Declaration changed.
1403 * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
1404 tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
1406 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1408 * tree-data-ref.c (conflict_fn): Assert that the number of affine
1409 relations in the conflict function is valid.
1411 2007-01-27 Ian Lance Taylor <iant@google.com>
1413 * common.opt: Add fstrict-overflow.
1414 * opts.c (decode_options): Set flag_strict_overflow if -O2.
1415 * flags.h (TYPE_OVERFLOW_WRAPS): Define.
1416 (TYPE_OVERFLOW_UNDEFINED): Define.
1417 (TYPE_OVERFLOW_TRAPS): Define. This replaces TYPE_TRAP_SIGNED.
1419 * tree.h (TYPE_TRAP_SIGNED): Don't define.
1420 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
1421 (fold_negate_expr): Likewise.
1422 (make_range): Likewise.
1423 (extract_muldiv_1): Likewise.
1424 (maybe_canonicalize_comparison): Likewise.
1425 (fold_comparison): Likewise.
1426 (fold_binary): Likewise.
1427 (tree_expr_nonnegative_p): Likewise.
1428 (tree_expr_nonzero_p): Likewise.
1429 * tree-vrp.c (compare_values): Likewise.
1430 (extract_range_from_binary_expr): Likewise.
1431 (extract_range_from_unary_expr): Likewise.
1432 * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
1434 (nowrap_type_p): Likewise.
1435 * tree-scalar-evolution.c (simple_iv): Likewise.
1436 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
1437 (build_range_check): Likewise.
1438 (extract_muldiv_1): Likewise.
1439 (fold_comparison): Likewise.
1440 * tree-vrp.c (vrp_int_const_binop): Likewise.
1441 (extract_range_from_unary_expr): Likewise.
1442 * convert.c (convert_to_integer): Likewise.
1443 * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
1444 (fold_comparison): Likewise.
1445 (fold_binary): Likewise.
1446 * optabs.c (optab_for_tree_code): Likewise.
1447 * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
1448 * simplify-rtx.c (simplify_const_relational_operation): Check
1449 flag_strict_overflow and flag_trapv.
1450 (simplify_const_relational_operation): Likewise.
1451 * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
1452 (Optimize Options): Add -fstrict-overflow to -O2 list. Document
1455 2007-01-27 Roger Sayle <roger@eyesopen.com>
1457 * tree.c (tree_fold_gcd): Delete.
1458 * tree.h (tree_fold_gcd): Remove prototype.
1459 * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
1460 test whether one constant integer is a multiple of another. Instead
1461 call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
1462 * fold-const.c (multiple_of_p): We've determined both TOP and
1463 BOTTOM are integer constants so we can call int_const_binop directly
1464 instead of the more generic const_binop.
1466 2007-01-27 Roger Sayle <roger@eyesopen.com>
1468 * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
1469 1*X check that the constant hasn't overflowed, to preserve the
1471 (round_up): Provide an efficient implementation when rouding-up an
1472 INTEGER_CST to a power-of-two.
1474 2007-01-28 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1476 * doc/sourcebuild.texi: Add comma for clarity.
1477 * doc/extend.texi: Fix some typos.
1478 * doc/passes.texi: Likewise.
1479 * doc/cppinternals.texi: Likewise.
1480 * doc/c-tree.texi: Likewise.
1481 * doc/tree-ssa.texi: Likewise.
1482 * doc/install.texi: Likewise.
1484 2007-01-27 Jan Hubicka <jh@suse.cz>
1486 * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
1487 (pass_sra): Do not require alias information.
1488 * passes.c (init_optimization_passes): Add SRA
1490 2007-01-27 Steven Bosscher <steven@gcc.gnu.org>
1492 * tracer.c (rest_of_handle_tracer): We already cleaned
1493 up the CFG in tracer() so don't do it here again.
1494 * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
1495 cleanup_cfg here, either. And don't call renumber_insns.
1497 * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
1498 and BB_HEAD are in the insn stream here. Instead make sure
1499 that BB_INSN is valid on all insns. Also, do check here that
1500 there are no pending branch predictions...
1501 (rtl_verify_flow_info): ...instead of doing it here. Checks
1502 for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
1505 2007-01-26 Roger Sayle <roger@eyesopen.com>
1507 * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
1508 function to simplify/factorize operand order canonicalization.
1509 (ix86_fixup_binary_operands): Reorganize using the above function.
1510 (ix86_binary_operator_ok): Likewise.
1512 2007-01-27 Jakub Jelinek <jakub@redhat.com>
1514 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
1515 file scope from optimize_attrs.
1516 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
1518 (optimize_attrs): Clear insn_code_values after freeing it.
1520 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
1522 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
1523 a parts.base assignment.
1525 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
1527 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
1528 (compute_subscript_distance, initialize_data_dependence_relation,
1529 finalize_ddr_dependent, analyze_ziv_subscript,
1530 analyze_siv_subscript_cst_affine,
1531 compute_overlap_steps_for_affine_univar,
1532 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
1533 analyze_siv_subscript, analyze_miv_subscript,
1534 analyze_overlapping_iterations, subscript_dependence_tester_1,
1535 compute_self_dependence, free_dependence_relation): Work
1536 with affine_fn instead of chrecs.
1537 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
1538 common_affine_function, affine_function_base,
1539 affine_function_constant_p, affine_fn_op, affine_fn_plus,
1540 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
1541 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
1542 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
1543 (all_chrecs_equal_p): Removed.
1544 * tree-data-ref.h (affine_fn, conflict_function): New types.
1545 (struct subscript): Change type of conflicting_iterations_in_a
1546 and conflicting_iterations_in_b.
1548 2007-01-26 Steve Ellcey <sje@cup.hp.com>
1551 * config/pa/pa.h (TARGET_HPUX_11): New.
1552 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
1553 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
1555 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
1556 Richard Guenther <rguenther@suse.de>
1558 * tree-ssa-structalias.c (solve_graph): Handle case
1559 we merged the variable to another.
1561 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1563 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
1566 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
1567 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
1569 2007-01-25 DJ Delorie <dj@redhat.com>
1571 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
1572 allow changes to modes which don't fit in those registers.
1574 * reload1.c (choose_reload_regs): Check for invalid subregs before
1575 computing their locations, not after.
1577 2007-01-25 Geoffrey Keating <geoffk@apple.com>
1580 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
1581 save Altivec registers in an eh_return function.
1582 (compute_vrsave_mask): Likewise.
1583 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
1584 for saving Altivec registers.
1585 (rs6000_emit_prologue): Don't allocate stack twice in
1586 eh_return function. Correct expected value of altivec_save_offset
1587 when using save_world. Describe save of R0 to stack when using
1588 save_world. Describe stack pointer adjustment when using
1589 save_world. Remove duplicated eh_return parameter register saving.
1590 Update sp_offset variable after save_world.
1591 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
1593 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
1594 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
1597 2007-01-25 Steve Ellcey <sje@cup.hp.com>
1599 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
1601 2007-01-25 Steve Ellcey <sje@cup.hp.com>
1604 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
1605 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
1607 2007-01-25 Richard Guenther <rguenther@suse.de>
1609 * doc/invoke.texi (-Wcoverage-mismatch): Document.
1610 * common.opt (-Wcoverage-mismatch): New warning option.
1611 * coverage.c (get_coverage_counts): Ignore coverage mismatch
1612 if -Wcoverage-mismatch is given.
1614 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
1616 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
1617 (ipcp_driver): Change to static definition.
1618 Add dumping of the ifunctions.
1619 (constant_val_insert): Remove unused parameter. Support for SSA.
1620 (ipcp_propagate_const): Support for SSA.
1621 (ipcp_profile_bb_print): Print only analyzed nodes.
1622 (ipcp_replace_map_create): Remove support for Fortran constant
1624 * ipa-prop.c (ipa_method_modify_stmt,
1625 ipa_callsite_compute_param): Support for SSA.
1626 * ipa-prop.h (ipcp_driver): Remove declaration.
1627 (IS_VALID_TREE_MAP_INDEX): Add define.
1629 2007-01-24 Geoffrey Keating <geoffk@apple.com>
1631 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
1633 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1636 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
1637 constant argument to cabs and do it without checking for
1638 -funsafe-math-optimizations.
1640 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
1642 * c-common.h (RID_FIRST_CXX0X): New.
1643 (RID_LAST_CXX0X): New.
1644 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
1646 * c.opt (Wc++0x-compat): New.
1647 * doc/invoke.texi (-Wc++0x-compat): Document.
1649 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1651 * builtins.c (fold_builtin_carg): New.
1652 (fold_builtin_1): Use it.
1654 2007-01-24 Jan Hubicka <jh@suse.cz>
1656 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
1658 * ipa-inline.c (initial_insns, max_insns): Delete.
1659 (compute_max_insns): New function.
1660 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
1661 of insns as base for code growth.
1662 (cgraph_decide_inlining): Make initial_insns local; do not compute
1664 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
1665 * doc/invoke.texi (inline-unit-growth): Update docs.
1667 2007-01-24 Jakub Jelinek <jakub@redhat.com>
1669 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
1670 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
1671 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
1672 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
1673 for CPUs that have PTA_CX16 set.
1676 * gimplify.c (omp_check_private): New function.
1677 (gimplify_scan_omp_clauses): Use it for
1678 firstprivate/lastprivate/reduction.
1681 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
1682 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
1685 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
1686 If lower_lastprivate_clauses emits some statements, append them
1687 to dlist rather than body_p and to body_p append an initializer.
1688 (lower_omp_for): Adjust caller.
1690 2007-01-24 Steve Ellcey <sje@cup.hp.com>
1692 * target.h (globalize_decl_name): New.
1693 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
1694 * output.h (default_globalize_decl_name): New.
1695 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
1697 (globalize_decl): Ditto.
1698 (default_globalize_decl_name): New.
1699 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
1700 (ia64_handle_version_id_attribute): New.
1701 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
1702 (ia64_asm_output_external): Use globalize_decl_name instead
1704 * doc/extend.texi (version_id): New pragma.
1705 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
1707 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
1709 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
1710 _uleb128_t and _Unwind_SWord with _sleb128_t.
1711 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
1712 uw_frame_state_for, uw_update_context_1): Likewise.
1713 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
1714 * unwind-pe.h (read_uleb128, read_sleb128,
1715 read_encoded_value_with_base): Likewise.
1716 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
1718 2007-01-24 Richard Guenther <rguenther@suse.de>
1720 * builtins.c (expand_builtin_cexpi): Get the fndecl
1721 for cexp in the correct way.
1723 2007-01-24 Jan Hubicka <jh@suse.cz>
1725 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
1727 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
1729 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
1730 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
1731 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
1732 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
1733 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
1734 __SIZEOF_PTRDIFF_T__.
1735 * doc/cpp.texi: Documentation for the new builtin macros added.
1737 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
1739 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
1740 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
1741 instead of operands[2] to avoid access past the end of array.
1743 2007-01-24 Richard Sandiford <richard@codesourcery.com>
1745 * reload1.c (emit_reload_insns): Pass the reload register
1746 for a non-spill output reload through forget_old_reloads_1.
1748 2007-01-23 Joseph Myers <joseph@codesourcery.com>
1750 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
1751 _SOFT_DOUBLE if doubles use software floating-point.
1752 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
1753 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
1754 * config/rs6000/darwin-ldouble.c: Also compile functions for
1755 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
1756 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
1757 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
1758 commented-out long double override.
1759 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
1760 * config/rs6000/eabispe.h: Likewise.
1761 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
1762 long double for non-SPE.
1763 (rs6000_handle_option): Likewise.
1764 (invalid_e500_subreg): Disallow more subregs involding DImode,
1765 DFmode, TImode or TFmode.
1766 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
1768 (legitimate_lo_sum_address_p): Also check for TFmode for E500
1770 (rs6000_legitimize_address): Also handle TFmode for E500 double.
1771 (rs6000_legitimize_reload_address): Also handle TFmode for E500
1773 (rs6000_legitimate_address): Also check for TFmode for E500
1775 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
1776 (spe_build_register_parallel): Handle TFmode and TCmode.
1777 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
1779 (function_arg): Handle TFmode and TCmode for E500 double.
1780 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
1782 (print_operand): Handle TFmode and TImode for %y.
1783 (rs6000_generate_compare): Handle TFmode comparisons for E500
1785 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
1786 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
1787 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
1788 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
1789 TFmode for E500 double.
1790 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
1791 (floatsidf2): Enable for E500 double.
1792 (movtf_softfloat): Use rs6000_nonimmediate_operand.
1793 (extenddftf2): Change to extenddftf2_fprs.
1794 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
1795 depending on TARGET_E500_DOUBLE.
1796 (extendsftf2): Enable for E500 double.
1797 (trunctfdf2): Enable for E500 double.
1798 (trunctfsf2): Change to trunctfsf2_fprs.
1799 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
1800 depending on TARGET_E500_DOUBLE.
1801 (floatsitf2): Enable for E500 double.
1802 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
1803 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
1804 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
1805 (negtf2): Change to negtf2_internal.
1806 (negtf2): New expander.
1807 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
1808 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
1809 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
1810 (movdi_internal32): Use rs6000_nonimmediate_operand.
1811 (unnamed splitter): Likewise.
1812 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
1813 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
1814 (SPE64TF, DITI): New mode macros.
1815 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
1818 (frob_<mode>_di_2): New.
1819 (frob_tf_di_8_2): New.
1820 (frob_di_df): Change to frob_di_<mode>; allow more modes.
1822 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
1824 (frob_ti_<mode>_8_2): New.
1825 (frob_ti_tf_2): New.
1826 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
1827 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
1829 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
1830 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
1831 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
1832 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
1833 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
1834 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
1837 2007-01-23 Ian Lance Taylor <iant@google.com>
1839 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
1841 (OBJS-archive): Reformat, alphabetize.
1842 (OBJS): Change out_object_file to OBJS-md.
1844 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
1846 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
1847 JUMP_P and CALL_P predicates where applicable.
1848 * config/i386/i386.c: Ditto.
1849 * config/i386/i386.md: Ditto.
1850 * config/i386/mmx.md: Ditto.
1851 * config/i386/predicates.md: Ditto.
1853 2007-01-22 Andreas Schwab <schwab@suse.de>
1855 * config/m68k/m68k.h: Fix comment.
1857 2007-01-22 Jan Hubicka <jh@suse.cz>
1859 * passes.c (init_optimization_passes): Do not rerun
1860 pass_early_warn_uninitialized.
1862 2007-01-22 Richard Guenther <rguenther@suse.de>
1864 PR tree-optimization/30038
1865 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
1867 (execute_cse_sincos_1): Likewise.
1868 (execute_cse_sincos): Likewise.
1869 (gate_cse_sincos): Likewise.
1870 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
1871 the cexpi() canonicalization of sincos().
1872 * tree-pass.h (pass_cse_sincos): Declare.
1873 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
1875 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
1877 PR rtl-optimization/29329
1878 * combine.c (replaced_rhs_insn): Rename to i2mod.
1879 (replaced_rhs_value): Rename to i2mod_new_rhs.
1880 (i2mod_old_rhs): New global variable.
1881 (combine_instructions): Adjust for above change. Save a copy of
1882 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
1883 note are substituted in the second instruction.
1884 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
1885 ditch the note if it pertains to the second eliminated register
1886 and this register is mentioned in i2mod_old_rhs.
1889 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
1891 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
1892 the insn to which the note was originally attached.
1894 2007-01-21 Jan Hubicka <jh@suse.cz>
1896 * ipa-inline.c (inlining_mode): Comment, move up.
1897 (cgraph_decide_inlining_incrementally): Do not perform inlining
1898 itself; fix handling of flattening of self recursive functions.
1899 (cgraph_find_cycles): Remove.
1900 (cgraph_flatten_node): Remove.
1901 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
1902 (try_inline): New function.
1903 (cgraph_early_inlining): Update call of
1904 cgraph_decide_inlining_incrementally. Apply inlining here.
1905 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
1907 2007-01-21 Dirk Mueller <dmueller@suse.de>
1910 * tree-vrp.c (check_array_bounds): do not warn
1911 about ADDR_EXPR's of ARRAY_REF's which are immediately
1912 used in binary expressions.
1914 2007-01-21 Ira Rosen <irar@il.ibm.com>
1916 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
1917 read_write_dep and macros for its access.
1918 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
1919 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
1920 argument, call vect_check_interleaving for every independent pair of
1921 data-refs. Mark loads that access the same memory location as a store
1923 (vect_check_dependences): Remove.
1924 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
1925 call, fix the call to vect_analyze_data_ref_dependence.
1926 (vect_analyze_data_ref_access): For statements that access the same
1927 data-ref, check that they are not stores; for loads, check that there
1928 is no store that access the same location.
1930 2007-01-20 Roger Sayle <roger@eyesopen.com>
1931 Joseph Myers <joseph@codesourcery.com>
1933 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
1934 (-fdump-rtl-stack): Likewise.
1935 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
1936 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
1937 (-mpt-fixed): Use @minus{} for minus sign.
1938 (Using Precompiled Headers): Fix under/overfull hbox.
1940 2007-01-20 Jan Hubicka <jh@suse.cz>
1942 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
1943 place within annotation.
1945 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1947 * pa.c (output_move_double): Change array size of xoperands to 4.
1949 2007-01-20 Jan Hubicka <jh@suse.cz>
1951 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
1954 2007-01-19 Ian Lance Taylor <iant@google.com>
1956 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
1957 constant to unsigned HOST_WIDE_INT before negating.
1959 2007-01-19 Ian Lance Taylor <iant@google.com>
1961 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
1963 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
1964 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
1965 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
1966 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
1967 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
1968 (realloc_vdef, realloc_vuse): Change second parameter to
1970 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
1972 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
1973 unsigned. Change x and lim locals to unsigned.
1974 (realloc_vdef, realloc_vuse): Change num_elem parameter to
1976 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
1977 (copy_virtual_operands): Change i and n locals to unsigned.
1979 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1982 * builtins.c (fold_builtin_1): Handle builtin fdim.
1984 2007-01-20 Jan Hubicka <jh@suse.cz>
1986 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
1987 * tree-ssa-structalias.c (compute_points_to_sets): Do call
1988 init_alias_heapvars.
1989 (init_alias_heapvars): Initialize only when not already initialized.
1990 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
1992 2007-01-19 Roger Sayle <roger@eyesopen.com>
1994 * common.opt (fsigned-zeros): New command line option.
1995 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
1996 of flag_unsafe_math_optimizations.
1997 * opts.c (set_fast_math_flags): The -ffast-math command line option
1998 implies -fno-signed-zeros.
1999 (fast_math_flags_set_p): Likewise.
2001 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
2002 the documentation of -ffast-math appropriately. Wrap long lines.
2004 2007-01-19 Steve Ellcey <sje@cup.hp.com>
2006 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
2007 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
2009 2007-01-19 Tomas Bily <tbily@suse.cz>
2011 * cgraphunit.c (cgraph_finalize_function): Updating of pid
2013 (tree_init_ic_make_global_vars): New function
2014 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
2015 (tree_gen_ic_profiler): New function
2016 (tree_gen_ic_func_profiler): New function
2017 (tree_profiling): Added calling of tree_gen_ic_func_profiler
2018 (tree_profile_hooks): Added hook for indirec/virtual calls
2019 * value-prof.c (tree_find_values_to_profile): New case for
2021 (tree_values_to_profile): Call for determining indirect/virtual
2023 (tree_indirect_call_to_profile): New function
2024 (tree_ic_transform): New function
2025 (tree_ic): New function
2026 (find_func_by_pid): New function
2027 (init_pid_map): New function
2028 (tree_value_profile_transformations): Added check for
2029 indirect/virtual call transformation
2030 * value-prof.h (enum hist_type): New counter type for
2031 indirect/virtual calls
2032 (profile_hooks): Added new hook for profiling indirect/virtual
2034 * profile.c (instrument_values): New case for indirect/virtual
2036 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
2037 (GCOV_COUNTER_V_INDIR): New counter type
2038 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
2039 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
2041 * cgraph.c: Definition of cgraph_max_pid
2042 (cgraph_create_node): Default init of pid attribute
2043 * cgraph.h: Declaration of cgraph_max_pid
2044 (struct cgraph_node): Added pid attribute
2045 * libgcov.c (__gcov_indirect_call_profiler): New function
2046 (__gcov_one_value_profiler_body): New function
2047 (__gcov_one_value_profiler): Body was moved to
2048 __gcov_one_value_profiler_body and calls it
2050 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
2052 * doc/gty.texi (Options): Document the mark_hook option to GTY.
2053 * gengtype.c (write_types_data, write_func_for_structure,
2054 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
2055 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
2056 (walk_type, write_func_for_structure): Generate the mark_hook if
2059 2007-01-19 Jan Hubicka <jh@suse.cz>
2061 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
2062 'early' argument take inlining mode argument specifying whether to
2063 inline for size/speeed or all functions; add support for flattening;
2065 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
2067 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2070 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
2071 the name of the declared identifier.
2073 2007-01-19 Dirk Mueller <dmueller@suse.de>
2075 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
2076 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
2077 duplicate evaluation.
2079 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2081 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
2082 Macroize expander using X87MODEF12 mode macro. Extend operand 1
2083 to XFMode, use acosxf2 and truncate result to requested mode.
2084 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
2085 (asin<mode>2): Similarly, with asin expanders.
2086 (*fscalexf4_i387): Rename from *fscalexf4.
2087 (expNcorexf3): New expander.
2088 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
2089 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
2090 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
2091 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
2092 patterns for SSE math.
2093 (exp10<mode>2): Similarly, with exp10 expanders.
2094 (exp2<mode>2): Similarly, with exp2 expanders.
2095 (expm1<mode>2): Similarly, with expm1 expanders.
2096 (ldexp<mode>3): Similarly, with ldexp expanders.
2097 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
2098 Use gen_truncxf<mode>2_i387_noop to truncate result.
2100 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2102 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
2103 for ColdFire targets.
2105 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
2106 Richard Sandiford <richard@codesourcery.com>
2108 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
2109 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
2110 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
2111 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
2112 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2113 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2115 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2117 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
2118 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
2119 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
2120 (*adddi_dilshr32_cf): New pattern.
2121 (adddi3, subdi3): Remove first alternatives.
2123 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2125 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
2126 implemented using move.l, do not use its cc result for floating-point
2129 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2131 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
2132 from accepting 's' constraints if flag_pic.
2134 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2136 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
2137 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
2138 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
2139 condition from TARGET_68881 to TARGET_HARD_FLOAT.
2141 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
2143 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
2144 ColdFire alternatives.
2145 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
2146 * config/m68k/m68k.md (clzsi2): Define for ColdFire
2147 architectures that support ff1 instruction.
2149 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2150 Julian Brown <julian@codesourcery.com>
2152 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
2153 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
2154 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
2157 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2159 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
2161 (output_move_const_into_data_reg, output_move_simode_const): Delete.
2162 * config/m68k/m68k.c (const_method, const_int_cost): Take a
2163 HOST_WIDE_INT instead of an rtx.
2164 (m68k_rtx_costs): Update call accordingly.
2165 (output_move_const_into_data_reg): Likewise. Fix formatting.
2166 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
2168 (output_move_simode_const): Update calls after above changes.
2169 Rework to use automatic variables and predicates like MEM_P.
2170 * config/m68k/m68k.md (pushexthisi_const): Update call to
2173 2007-01-19 Dirk Mueller <dmueller@suse.de>
2175 * tree-ssa-alias.c (perform_var_substitution): Fix typo
2178 2007-01-19 Richard Guenther <rguenther@suse.de>
2180 * builtins.c (expand_builtin_cexpi): Fall back to expanding
2181 via cexp in case sincos is not available.
2183 2007-01-19 Richard Guenther <rguenther@suse.de>
2185 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
2186 * defaults.h (TARGET_HAS_SINCOS): Default to off.
2187 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
2188 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
2189 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
2190 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
2191 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
2192 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
2194 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2196 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
2198 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
2199 (atan2xf3): Directly generate RTL pattern.
2200 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
2201 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2202 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2203 disable patterns for SSE math.
2204 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
2205 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2206 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2207 disable patterns for SSE math.
2209 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
2211 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
2213 2007-01-18 Roger Sayle <roger@eyesopen.com>
2215 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
2216 VIEW_CONVERT_EXPR to the same type as it's operand.
2218 2007-01-18 David Edelsohn <edelsohn@gnu.org>
2220 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
2221 configured for long double 128.
2223 2007-01-18 Mike Stump <mrs@apple.com>
2225 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
2228 2007-01-18 Michael Meissner <michael.meissner@amd.com>
2230 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
2233 2007-01-18 Jan Hubicka <jh@suse.cz>
2235 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
2236 greater than NUM_VOP_FREE_BUCKETS.
2238 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
2240 * tree-ssa-structalias.c: Update comments.
2241 (ptabitmap_obstack): Removed.
2243 (oldpta_obstack): Ditto.
2244 (stats): Add a few members.
2245 (struct variable_info): Remove node, complex, address_taken, and
2246 indirect_target members. Add oldsolution member.
2247 (new_var_info): Do not initialize removed members.
2248 (constraint_expr_type): Remove INCLUDES.
2249 (constraint_graph): Add size, implicit_preds, rep,
2250 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
2251 (FIRST_REF_NODE): New macro.
2252 (LAST_REF_NODE): Ditto.
2253 (FIRST_ADDR_NODE): Ditto.
2254 (find): New function.
2256 (dump_constraint): Do not handle INCLUDES.
2257 (insert_into_complex): Do not insert duplicate constraints.
2258 (condense_varmap_nodes): Renamed and rewritten into ...
2259 (merge_node_constraints): This. Also fix bug in handling of
2260 offseted copy constraints.
2261 (clear_edges_for_node): No longer need to deal with preds at all,
2262 or removing associated preds/succs.
2263 (merge_graph_nodes): Deal with indirect_cycles.
2264 Don't deal with predecessors.
2265 (add_implicit_graph_edge): New function.
2266 (add_pred_graph_edge): Ditto.
2267 (add_graph_edge): Don't deal with predecessors.
2268 (build_constraint_graph): Removed.
2269 (build_pred_graph): New function.
2270 (build_succ_graph): Ditto.
2271 (struct scc_info): Removed in_component. Added roots, dfs, and
2272 node_mapping. Remove visited_index, unification_queue.
2273 (scc_visit): Deal with union-find we do now.
2274 Deal with cycles with REF nodes.
2275 (collapse_nodes): Renamed and rewritten to ...
2276 (unify_nodes): This.
2277 (process_unification_queue): Removed.
2278 (topo_visit): Cleanup
2279 (do_da_constraint): Use find.
2280 (do_sd_constraint): Ditto.
2281 (do_ds_constraint): Ditto.
2282 (do_complex_constraint): Ditto.
2283 (init_scc_info): Update for removed and added members.
2284 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
2285 (find_indirect_cycles): This.
2286 (equivalence_class): New variable.
2287 (label_visit): New function.
2288 (perform_variable_substitution): Rewritten.
2289 (free_var_substitution_info): New function.
2290 (find_equivalent_node): Ditto.
2291 (move_complex_constraints): Ditto.
2292 (eliminate_indirect_cycles): Ditto.
2293 (solve_graph): Only propagate changed bits.
2294 Use indirect cycle elimination.
2296 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
2297 (tree_id_eq): Renamed to ...
2298 (tree_vi_eq): This. Update for member change
2299 (insert_id_for_tree): Renamed and rewritten to ...
2300 (insert_vi_for_tree): This.
2301 (lookup_id_for_tree): Renamed and rewritten to ...
2302 (lookup_vi_for_tree): This.
2303 (get_id_for_tree): Renamed and rewritten to ...
2304 (get_vi_for_tree): Ditto.
2305 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
2306 (process_constraint): Don't handle INCLUDES.
2307 Remove special ADDRESSOF case.
2308 (find_func_aliases): Rewrite to use vi functions instead of id
2310 (create_function_info_for): Ditto.
2311 (create_variable_info_for): Ditto.
2312 (intra_create_variable_infos): Ditto.
2313 (merge_smts_into): Ditto.
2314 (find_what_p_points_to): Ditto.
2315 (init_base_vars): Ditto.
2316 (init_alias_vars): Ditto.
2317 (remove_preds_and_fake_succs): New function.
2318 (dump_sa_points_to_info): Dump new stats.
2319 (dump_solution_for_var): Use find.
2320 (set_used_smts): Fix formatting.
2321 (compute_points_to_sets): Updated for new functions.
2322 (ipa_pta_execute): Ditto.
2324 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2325 Richard Sandiford <richard@codesourcery.com>
2327 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
2329 * calls.c (expand_call): If the return value is a PARALLEL,
2330 extract its first member.
2331 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
2332 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2333 use FUNCTION_EXTRA_EPILOGUE.
2334 (m68k_function_value): Return a PARALLEL if the return value
2335 is of a pointer type.
2336 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
2337 (FUNCTION_EXTRA_EPILOGUE): Remove.
2338 * config/m68k/m68k.md (D0_REG): New constant.
2340 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2342 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2343 output a NOP for empty epilogues.
2345 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2347 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
2348 before function. Extend register save check to include all
2349 registers, not just integer ones.
2351 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2353 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
2354 (m68k_use_return_insn): ...this.
2355 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
2356 * config/m68k/m68k.c (use_return_insn): Rename to...
2357 (m68k_use_return_insn): ...this.
2358 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
2361 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2363 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
2364 nearest even, fix denormal rounding overflow.
2366 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2368 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
2369 f%$smove and f%$move.
2370 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
2371 (extendsfdf2_cf): Use fdmove instead of f%&move.
2372 (truncdfsf2_cf): Use fsmove instead of f%$smove.
2373 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
2375 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2377 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
2380 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2382 * real.h (real_format): Add a canonical_nan_lsbs_set field.
2383 (coldfire_single_format): Declare.
2384 (coldfire_double_format): Likewise.
2385 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
2386 of qnan_msb_set to determine the lower bits of a canonical
2388 (encode_ieee_double): Likewise.
2389 (encode_ieee_quad): Likewise.
2390 (ieee_single_format): Initialize canonical_nan_lsbs_set.
2391 (mips_single_format): Likewise.
2392 (ieee_double_format): Likewise.
2393 (mips_double_format): Likewise.
2394 (ieee_extended_motorola_format): Likewise.
2395 (ieee_extended_intel_96_format): Likewise.
2396 (ieee_extended_intel_128_format): Likewise.
2397 (ieee_extended_intel_96_round_53_format): Likewise.
2398 (ibm_extended_format): Likewise.
2399 (mips_extended_format): Likewise.
2400 (ieee_quad_format): Likewise.
2401 (mips_quad_format): Likewise.
2402 (vax_f_format): Likewise.
2403 (vax_d_format): Likewise.
2404 (vax_g_format): Likewise.
2405 (i370_single_format): Likewise.
2406 (i370_double_format): Likewise.
2407 (decimal_single_format): Likewise.
2408 (decimal_double_format): Likewise.
2409 (decimal_quad_format): Likewise.
2410 (c4x_single_format): Likewise.
2411 (c4x_extended_format): Likewise.
2412 (real_internal_format): Likewise.
2413 (coldfire_single_format): New real_format.
2414 (coldfire_double_format): Likewise.
2415 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
2416 canonical_nan_lsbs_set.
2417 (pdp11_d_format): Likewise.
2418 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
2419 if TARGET_COLDFIRE_CPU.
2421 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2423 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
2424 (output_call, m68k_legitimize_call_address): Declare.
2425 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
2426 parenthesees. Add support for a 'W' constraint.
2427 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
2428 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
2429 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
2430 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
2432 (override_options): Initialize them. Do not set flag_no_function_cse
2433 for TARGET_ID_SHARED_LIBRARY.
2434 (m68k_output_pic_call): Delete.
2435 (m68k_legitimize_call_address): New function.
2436 (print_operand): Remove the %o prefix. Handle the %p prefix.
2437 (output_call): New function.
2438 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
2439 target address from the GOT if symbolic jumps are not allowed.
2440 * config/m68k/m68k.md (call, general_operand): Do not set
2441 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
2442 Merge separate flag_pic and !flag_pic define_insns into...
2443 (*call, *call_value): ...these new patterns. Match the address
2444 rather than the containing MEM and require it to be a call_operand.
2445 Use output_call to generate the asm template.
2446 * config/m68k/predicates.md (const_call_operand): New predicate.
2447 (call_operand): Likewise.
2449 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2451 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
2452 (STACK_GROWS_DOWNWARD): Define to 1.
2453 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
2454 comments, emphasizing that these values are only defaults.
2455 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
2456 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
2458 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2460 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
2462 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2463 Richard Sandiford <richard@codesourcery.com>
2465 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
2466 (FP_REGNO_P): New macros.
2467 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
2468 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
2469 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
2470 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
2472 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2474 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2475 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
2477 * config/m68k/t-floatlib: New file, extracting common code from...
2478 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
2479 * config/m68k/t-uclinux: Here.
2480 * config/m68k/fpgnulib.c: Do not compile extendeed precision
2481 routines on ColdFire targets.
2483 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2485 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2486 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
2487 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
2488 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
2489 explicitly set MOTOROLA to 1.
2490 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
2492 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2493 Nathan Sidwell <nathan@codesourcery.com>
2495 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
2496 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
2497 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
2498 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
2499 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
2500 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
2501 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
2502 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
2503 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
2504 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2505 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
2506 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
2507 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
2508 (DBX_REGISTER_NUMBER): Delete.
2509 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
2510 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2511 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
2513 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2515 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
2516 * config/flat.h: New file.
2517 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
2519 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
2520 indirect references for -msep-data or -mid-shared-library.
2521 Do not use PC-relative code addresses either.
2522 * config/m68k/m68k.c (override_options): Restrict -fPIC error
2524 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
2525 for shared libraries and crt1.o for executables. Use crti.o and
2527 (ENDFILE_SPEC): Use crtend.o and crtn.o.
2528 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
2529 Do not add -elf2flt or -shared-lib-id options here.
2530 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
2531 Pass -shared-lib-id if -mid-shared-library, taking the library
2532 identifier from -mshared-library-id if given, otherwise
2534 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
2535 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
2536 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
2537 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
2538 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
2539 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
2541 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
2542 jump sequence for ISA A and ISA A+.
2543 (PICJUMP): Likewise.
2545 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2547 * config.gcc (m68k-*-uclinux*): Base the port on the common
2548 and m68k GNU/Linux files rather than on the generic ELF ones.
2549 * config/m68k/uclinux.h (TARGET_VERSION): Override.
2550 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
2552 2007-01-18 Julian Brown <julian@codesourcery.com>
2553 Richard Sandiford <richard@codesourcery.com>
2555 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
2556 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2557 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
2559 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
2561 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2563 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2565 * doc/install.texi: Document m68k-uclinuxoldabi.
2566 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
2567 * config/m68k/uclinux-oldabi.h: New file, copied from
2568 config/m68k/uclinux.h.
2570 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
2572 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
2573 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
2574 (m32r_compute_frame_size): Use unsigned for regno.
2575 (m32r_reload_lr): Use gen_frame_mem.
2577 (m32r_output_function_epilogue): Don't output the function epilogue
2579 (m32r_expand_epilogue): New.
2580 (direct_return): Return false if the function has the interrupt
2582 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
2583 * config/m32r/m32r.md (epilogue): New expander.
2584 (return_lr, return_rte): New insns.
2585 (return): Make it expander.
2586 (return_normal): New expander.
2588 2007-01-18 Josh Conner <jconner@apple.com>
2591 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
2592 support for UNLE, UNLT, UNGE, and UNGT.
2594 2007-01-18 Jan Hubicka <jh@suse.cz>
2596 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
2597 * tree-tailcall (add_virtual_phis): Likewise.
2598 (optimize_tail_call): Return TODOs.
2599 (execute_tail_calls): Return TODOs.
2600 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
2601 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
2603 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
2604 Cleanup cfg using TODO.
2605 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
2606 * passes.c (execute_function_todo): When cleanup did something, remove
2608 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
2609 (make_edges): Don't cleanup_cfg.
2611 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
2613 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
2614 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
2615 operate on the top of stack.
2616 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
2617 register handling of instructions that output to the second
2619 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
2620 (move_for_stack_reg): Special-case check for dead destination
2621 stack slot for constant load of 1.0 inside UNSPEC_TAN.
2623 * config/i386/i386.md (UNSPEC_TAN): New constant.
2624 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
2625 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
2626 to correctly model move of constant 1.0 to top stack slot.
2627 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
2628 (unnamed peephole2 pattern): Remove corresponding peephole2
2629 pattern that optimizes tan insn and loading of constant 1.0.
2630 (tanxf2): Use fptanxf4_i387.
2631 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
2632 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
2633 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2634 disable patterns for SSE math.
2635 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
2636 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
2638 2007-01-18 Dirk Mueller <dmueller@suse.de>
2639 Richard Guenther <rguenther@suse.de>
2642 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
2643 * common.opt (Warray-bounds): Add new warning option.
2644 * c-opts.c (c_common_handle_option): Define -Warray-bounds
2646 * Makefile.in: make tree-vrp.o depend on toplev.h
2647 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
2649 (check_array_refs, check_array_bounds, check_array_ref): New.
2651 2007-01-18 Jan Hubicka <jh@suse.cz>
2653 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
2654 (execute_ssa_ccp): Return flags conditionally.
2655 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
2657 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
2659 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
2661 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
2662 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
2664 * Makefile.c (GTFILES): Add cfglayout.h.
2665 * gengtype.c (open_base_files): Likewise.
2666 * cfglayout.c (cfg_layout_function_footer,
2667 cfg_layout_function_header) Reindent to make gengtype happy.
2668 * cfglayout.h (cfg_layout_function_footer,
2669 cfg_layout_function_header): Add GTY(()) marker.
2671 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
2673 2007-01-18 Ben Elliston <bje@au.ibm.com>
2675 * genautomata.c (write_automata): Include xstrerror output in the
2676 error message if writing the DFA description file fails.
2678 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
2680 * config/mips/mips-protos.h (mips_output_external): Make it
2682 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
2683 * config/mips/mips.c (irix_output_external_libcall): Likewise.
2684 (extern_list): Likewise.
2685 (extern_head): Likewise.
2686 (TARGET_ASM_FILE_END): Likewise.
2687 (mips_file_end): Likewise.
2688 (mips_output_external): Rewritten.
2690 2007-01-18 Ben Elliston <bje@au.ibm.com>
2692 * genpreds.c (write_insn_preds_c): Only write out the function
2693 body for regclass_for_constraint if we have register constraints.
2695 2007-01-17 Tom Tromey <tromey@redhat.com>
2697 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
2698 * doc/install.texi (Testing): Use sourceware.org.
2699 (Binaries): Likewise.
2700 (Specific): Likewise.
2701 * doc/contrib.texi (Contributors): Use sourceware.org.
2703 2007-01-17 Anatoly Sokolov <aesok@post.ru>
2705 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
2706 (AVR_ENHANCED): Rename to ...
2707 (AVR_HAVE_MUL): ... new.
2708 (avr_enhanced_p): Rename to ...
2709 (avr_have_mul_p): ... new.
2710 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
2711 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
2712 * config/avr/avr.c (avr_enhanced_p): Rename to ...
2713 (avr_have_mul_p): ... new.
2714 (base_arch_s): Rename 'enhanced' to 'have_mul'.
2715 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
2716 'avr_enhanced_p' and 'enhanced'.
2717 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
2718 AVR_HAVE_MUL instead of AVR_ENHANCED.
2719 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
2720 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
2721 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
2722 instead of AVR_ENHANCED.
2723 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
2724 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
2725 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
2727 2007-01-17 Ian Lance Taylor <iant@google.com>
2729 * vec.h (VEC_reserve_exact): Define.
2730 (vec_gc_p_reserve_exact): Declare.
2731 (vec_gc_o_reserve_exact): Declare.
2732 (vec_heap_p_reserve_exact): Declare.
2733 (vec_heap_o_reserve_exact): Declare.
2734 (VEC_OP (T,A,reserve_exact)): New static inline function, three
2736 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
2738 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
2739 (VEC_OP (T,A,copy)) [all versions]: Likewise.
2740 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
2741 * vec.c (calculate_allocation): Add exact parameter. Change all
2743 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
2744 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
2745 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
2746 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
2747 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
2748 (vec_heap_p_reserve_exact): New function.
2749 (vec_heap_o_reserve_exact): New function.
2751 2007-01-17 Jan Hubicka <jh@suse.cz>
2753 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
2755 2007-01-17 Eric Christopher <echristo@apple.com>
2757 * config.gcc: Support core2 processor.
2759 2007-01-16 Jan Hubicka <jh@suse.cz>
2761 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
2762 the names, just unlink the chain so we don't crash on dangling pointers
2765 2007-01-16 Jan Hubicka <jh@suse.cz>
2767 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
2768 * tree-pass.h: Reorder to make IPA passes appear toegher.
2769 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
2770 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
2771 parameters, do not call early inliner.
2772 * ipa-inline.c: Update comments. Include tree-flow.h
2773 (cgraph_decide_inlining): Do not compute inlining parameters.
2774 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
2775 be called with function context set up.
2776 (pass_ipa_inline): Remove unreachable functions before pass.
2777 (cgraph_early_inlining): Simplify assuming to be called from the
2779 (pass_early_inline): New pass.
2780 (cgraph_gate_ipa_early_inlining): New gate.
2781 (pass_ipa_early_inline): Turn into simple wrapper.
2782 (compute_inline_parameters): New function.
2783 (gate_inline_passes): New gate.
2784 (pass_inline_parameters): New pass.
2785 (apply_inline): Move here from tree-optimize.c
2786 (pass_apply_inline): New pass.
2787 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
2789 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
2791 (tree_function_versioning): Do not allocate dummy struct function.
2792 * tree-inline.h (optimize_inline_calls): Update prototype.
2793 * tree-optimize.c (execute_fixup_cfg): Export.
2794 (pass_fixup_cfg): Remove
2795 (tree_rest_of_compilation): Do not apply inlines.
2796 * tree-flow.h (execute_fixup_cfg): Declare.
2797 * Makefile.in (gt-passes.c): New.
2798 * passes.c: Include gt-passes.h
2799 (init_optimization_passes): New passes.
2800 (nnodes, order): New static vars.
2801 (do_per_function_toporder): New function.
2802 (execute_one_pass): Dump current pass here.
2803 (execute_ipa_pass_list): Don't dump current pass here.
2805 2007-01-16 Janis Johnson <janis187@us.ibm.com>
2807 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
2808 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
2810 2007-01-16 David Edelsohn <edelsohn@gnu.org>
2812 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
2813 (strong_alias): Define.
2814 (__gcc_qmul): Provide non-FMA for soft-float.
2820 (__gcc_qunord): New.
2830 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
2831 soft-float functions.
2832 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
2833 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
2834 about long double soft float.
2836 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
2837 Tehila Meyzels <tehila@il.ibm.com>
2839 * tree-vectorizer.h (is_pattern_stmt_p): New.
2840 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
2841 formatting (tabs instead of spaces). Cleanup and clarify setting
2842 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
2843 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
2844 (vectorizable_type_demotion): Check that types are integral.
2845 (vectorizable_type_promotion): Likewise.
2846 (vectorizable_store): Fix typo. Eliminate new-line at end of
2849 2007-01-16 Jan Hubicka <jh@suse.cz>
2851 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
2854 2007-01-15 Eric Christopher <echristo@apple.com>
2856 * ifcvt.c: Include vec.h, vecprim.h.
2857 (check_cond_move_block): New argument regs.
2858 Reorganize. Add registers used to regs.
2859 (cond_move_process_if_block): Use regs set above as
2862 2007-01-15 Eric Christopher <echristo@apple.com>
2864 * config/darwin.h: Update copyright.
2865 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
2866 (LINK_COMMAND_SPEC): Add -u.
2867 (LINK_SPEC): Fix umbrella for above.
2869 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
2871 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
2874 2007-01-15 Tom Tromey <tromey@redhat.com>
2876 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
2877 * doc/install.texi (Testing): Don't mention jacks.
2878 (Configuration): Document --enable-java-maintainer-mode. Move
2879 --with-java-home to libgcj-specific section. Document
2881 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
2883 2007-01-15 Jan Hubicka <jh@suse.cz>
2885 * tree-ssa-dce.c (DCE_TODOs): New.
2886 (propagate_necessity): Return if something changed.
2887 (eliminate_unnecessary_stmts): Likewise.
2888 (perform_tree_ssa_dce): Return TODO flags when needed.
2889 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
2891 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
2893 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
2894 (fyl2x_extend<mode>xf3_i387): New insn pattern.
2895 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
2896 patterns using X87MODEF12 mode macro. Extend operand 1
2897 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
2898 (log10<mode>2): Ditto.
2899 (log2<mode>2): Ditto.
2900 (log1p<mode>2): Ditto.
2901 (logb<mode>2): Ditto.
2902 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
2903 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
2904 (*fxtractxf3_i387): Rename from *fxtractxf3.
2905 (fxtract_extend<mode>xf3_i387): New insn pattern.
2906 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
2907 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
2908 and gen_fyl2xxf3_i387().
2910 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
2912 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
2913 * cfgloopmanip.c (fix_loop_placement): Made static. Use
2914 get_loop_exit_edges. Changed return type to bool.
2915 * cfgloop.h (fix_loop_placement): Declaration removed.
2917 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
2919 * param.h (MIN_VECT_LOOP_BOUND): New.
2920 * params.def (MIN_VECT_LOOP_BOUND): New.
2921 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
2922 argument - minimum threshold for number of iterations.
2923 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
2924 argument to declaration.
2925 * tree-vect-analyze.c (vect_analyze_operations): Check value of
2926 MIN_VECT_LOOP_BOUND.
2927 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
2928 slpeel_tree_peel_loop_to_edge with additional argument.
2929 (vect_do_peeling_for_alignment): Likewise.
2930 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
2932 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
2935 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
2938 2007-01-14 Jan Hubicka <jh@suse.cz>
2940 * tree-dfa.c (remove_referenced_var): New function.
2941 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
2942 prune referenced vars list too.
2943 * tree-flow.h (remove_referenced_var): Declare.
2945 2007-01-14 Jan Hubicka <jh@suse.cz>
2947 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
2949 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
2950 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
2951 * except.h (verify_eh_throw_table_statements): Kill prototype.
2952 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
2953 into call exrepssions.
2954 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
2956 * tree-cfg.c: Include pointer-set.h.
2957 (verify_node_sharing): Work on pointer set.
2958 (verify_eh_throw_stmt_node): New.
2959 (verify_stmts): Use pointers sets, verify throw_stmt.
2961 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
2963 * ipa-reference.c (analyze_function): Consider also addresses taken
2966 2007-01-12 Roger Sayle <roger@eyesopen.com>
2968 * c-typeck.c (null_pointer_constant_p): Replace use of
2969 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2970 (build_c_cast): Likewise.
2972 2007-01-12 Roger Sayle <roger@eyesopen.com>
2974 * tree.h (force_fit_type_double): Remove unused final argument.
2975 * c-common.c (constant_expression_warning): Replace use of
2976 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2977 (convert_and_check): Likewise.
2978 (shorten_compare): Update call to force_fit_type_double.
2979 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
2980 * convert.c (convert_to_pointer): Update call to
2981 force_fit_type_double.
2982 * fold-const.c (force_fit_type_double): Remove overflowed_const
2984 (int_const_binop, fold_convert_const_int_from_int,
2985 fold_convert_const_int_from_real, fold_div_compare,
2986 fold_sign_changed_comparison, fold_unary, fold_negate_const,
2987 fold_abs_const, fold_not_const): Remove the final argument from
2988 calls to force_fit_type_double.
2990 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
2992 * configure.ac: Set insn to "nop" for spu-*-* also.
2993 * configure: Regenerate.
2995 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
2997 * builtins.def : Add BUILT_IN_FREE.
2999 2007-01-12 Jan Hubicka <jh@suse.cz>
3001 PR tree-optimization/30443
3002 * tree-inline.c (tree_function_versioning): Do not optimize when
3003 cloning for inlining.
3005 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
3007 * doc/loop.texi: Document recording of loop exits.
3008 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
3009 (update_single_exits_after_duplication,
3010 update_single_exit_for_duplicated_loop,
3011 update_single_exit_for_duplicated_loops): Removed.
3012 (duplicate_loop_to_header_edge): Do not call
3013 update_single_exits_after_duplication and
3014 update_single_exit_for_duplicated_loops.
3015 (loop_version): Do not update single_exit information.
3016 (fix_loop_structure): Use record_loop_exits instead of
3017 mark_single_exit_loops.
3018 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
3019 the lists of loop exits.
3020 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
3021 split_edge, merge_blocks): Update the lists of loop exits.
3022 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
3023 loop_optimizer_init.
3024 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
3025 of mark_single_exit_loops.
3026 (loop_optimizer_finalize): Call release_recorded_exits.
3027 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
3028 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
3029 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
3030 update single exit information.
3031 * lambda-code.c (perfect_nestify): Ditto.
3032 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
3033 (mark_single_exit_loops): Removed.
3034 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
3035 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
3036 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
3038 (get_loop_exit_edges, single_exit): Use recorded exit lists.
3039 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
3040 (verify_loop_structure): Verify consistency of the exit lists.
3041 (flow_loops_find): Use alloc_loop. Initialize exits hash.
3042 (set_single_exit): Removed.
3043 * cfgloop.h (struct loop_exit): New function.
3044 (struct loop): single_exit_ field replaced by exits field.
3045 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
3046 (struct loops): Added exits hash.
3047 (mark_single_exit_loops, set_single_exit): Declaration removed.
3048 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
3050 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3052 * doc/invoke.texi: Avoid use of @headitem.
3054 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3056 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
3058 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3060 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
3063 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3064 Richard Sandiford <richard@codesourcery.com>
3066 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
3067 * config/m68k/m68k.opt: Resort options.
3068 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
3070 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3071 Richard Sandiford <richard@codesourcery.com>
3073 * doc/invoke.texi: Document the macros that are defined by
3074 m68k's -mtune and -mhard-float options.
3075 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
3076 (m68k_cpp_cpu_family): Likewise.
3077 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
3078 of __ucfv*__ macros. Define __mcffpu__ if generating code for
3079 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
3080 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
3081 (m68k_cpp_cpu_family): Likewise.
3083 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3085 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
3086 macros besides mc68000 as tuning macros. Use a switch statement
3087 to set them and mcpu32.
3089 2007-01-12 Julian Brown <julian@codesourcery.com>
3091 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
3092 (TARGET_68040_ONLY): Rename to...
3093 (TARGET_68040): ...this.
3094 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
3095 * config/m68k/m68k.md: Likewise.
3097 2007-01-12 Julian Brown <julian@codesourcery.com>
3098 Nathan Sidwell <nathan@codesourcery.com>
3099 Richard Sandiford <richard@codesourcery.com>
3101 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
3102 the -mcpu= argument associated with the --with-cpu setting.
3103 Define M68K_DEFAULT_TUNE to the default -mtune= option,
3104 if different from the one implied by the -mcpu setting.
3105 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
3106 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
3107 * doc/invoke.texi: Mention ColdFire in the introduction to the
3108 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
3109 -mno-div and -mhard-float options. Make -m68881 a synonym for
3110 -mhard-float. Document the previously-undocumented -m5206e,
3111 -m528x, -m5307 and -m5407 options. Tweak the existing option
3112 documentation for consistency.
3113 * doc/install.texi: Mention new --with-cpu arguments.
3114 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
3115 default CPU if neither -mcpu nor -march are specified.
3116 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
3117 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
3118 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
3119 using m68k_tune to decide between families that implement the
3120 same ISA. Use m68k_tune to set __mcfv4e__.
3121 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
3122 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
3123 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
3124 (FL_ISA_C, FL_ISA_MMU): New macros.
3125 (MASK_COLDFIRE): Delete.
3126 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
3127 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
3128 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
3129 (TARGET_HARD_FLOAT): Do not define here.
3130 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
3131 (TUNE_68000): New macro.
3132 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
3133 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
3134 (TUNE_CFV2): Redefine in terms of m68k_tune.
3135 (uarch_type, target_device, fpu_type): New enums.
3136 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
3137 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
3138 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
3139 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
3140 (FL_FOR_isa_c): New macros.
3141 (m68k_isa): New enum.
3142 (m68k_target_selection): New structure.
3143 (all_devices, all_isas, all_microarchs): New tables.
3144 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
3145 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
3146 (MASK_ALL_CPU_BITS): Delete.
3147 (m68k_find_selection): New function.
3148 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
3149 Map the legacy target options to a combination of the new ones.
3150 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
3151 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
3152 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
3154 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
3155 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
3157 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
3158 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
3159 * config/m68k/m68k-devices.def: New file.
3161 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3162 Nathan Sidwell <nathan@codesourcery.com>
3164 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
3165 (SUBTARGET_EXTRA_SPECS): New macros.
3166 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
3167 use %(asm_cpu_spec) instead.
3168 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
3169 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
3170 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
3171 (EXTRA_SPECS): Rename to...
3172 (SUBTARGET_EXTRA_SPECS): ...this.
3174 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3175 Richard Sandiford <richard@codesourcery.com>
3176 Julian Brown <julian@codesourcery.com>
3178 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3179 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
3180 to the configuration's default CPU.
3181 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
3182 Remove default masks.
3183 (m680[012]0-*-*): Set the default with_cpu to the first part of
3185 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
3186 (m68k*-*-linux): Extend the --with-cpu handling to...
3187 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
3188 and m68010. Don't set target_cpu_default2.
3189 * doc/install.texi: Document --with-cpu for m68k.
3190 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
3191 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
3192 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
3193 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
3194 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
3195 (ASM_SPEC): Remove use of %(asm_cpu_default).
3196 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
3197 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
3198 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
3199 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
3200 __HAVE_FPU__ if TARGET_HARD_FLOAT.
3201 (TARGET_DEFAULT): Delete.
3202 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
3203 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
3204 (CPP_CPU_SPEC): Delete.
3205 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
3206 the appropriate string.
3207 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
3208 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
3209 (ASM_SPEC): Don't use %(asm_default_spec).
3210 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
3211 TARGET_DEFAULT and add MASK_68881.
3212 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
3214 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3216 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
3217 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
3218 MASK_68010 alongside MASK_68020.
3219 * doc/invoke.texi: Document -m68010.
3220 * config/m68k/m68k.opt (m68010): New.
3221 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
3223 (TUNE_68010): New macro.
3224 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
3225 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
3226 (M68K_CPU_m68332): Add MASK_68010.
3227 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
3228 fallback definition.
3229 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
3231 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
3232 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
3233 to all entries that use MASK_68020.
3234 (output_move_simode_const, output_move_himode, output_move_qimode)
3235 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
3236 instead of TARGET_68020 to select clr behavior. Remove comment
3237 about there being no TARGET_68010.
3238 * config/m68k/m68k.md: Likewise throughout.
3240 2007-01-12 Julian Brown <julian@codesourcery.com>
3242 * config/m68k/m68k.h (TARGET_ISAB): New macro.
3243 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
3244 * config/m68k/m68k.md: Likewise.
3246 2007-01-12 Julian Brown <julian@codesourcery.com>
3248 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
3249 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
3251 2007-01-12 Julian Brown <julian@codesourcery.com>
3253 * config/m68k/m68k.h (TUNE_68040_60): New macro.
3254 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
3255 * config/m68k/m68k.md: Likewise.
3257 2007-01-12 Julian Brown <julian@codesourcery.com>
3258 Richard Sandiford <richard@codesourcery.com>
3260 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
3261 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
3262 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
3264 (TARGET_CPU32): Rename to...
3265 (TUNE_CPU32): ...this.
3266 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
3267 (TUNE_CFV2): New macros.
3268 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
3269 remove conditions that are implied by TARGET_68020.
3270 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
3271 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
3272 (m68k_output_function_epilogue): Likewise.
3273 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
3274 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
3275 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
3276 non-68000 timings. Refactor multiplication and division costs.
3277 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
3278 TUNE_CPU32 instead of TARGET_CPU32.
3279 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
3280 and TUNE_68060 instead of TARGET_68060.
3281 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
3282 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
3284 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
3285 && !TARGET_COLDFIRE" to choose between moveq and clr.
3286 Likewise in the unnamed movsf pattern.
3287 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
3288 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
3291 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3293 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
3294 of tabbing before backslashes.
3296 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3298 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
3299 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
3300 (ASM_OUTPUT_INTERNAL_LABEL): Define.
3302 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3304 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
3305 find_interesting_uses_cond.
3306 (find_interesting_uses_cond): Use extract_cond_operands.
3307 (rewrite_use_compare): Use extract_cond_operands and
3308 force_gimple_operand_bsi. Do not call update_stmt.
3309 (determine_use_iv_cost_condition): Use extract_cond_operands.
3310 Return cheaper of using original bound and changing the exit bound.
3312 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3314 PR tree-optimization/29516
3315 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
3316 most_expensive_mult_to_index, addr_to_parts,
3317 create_mem_ref, maybe_fold_tmr): Make the type of
3318 fields of TARGET_MEM_REF sizetype.
3319 (move_fixed_address_to_symbol, move_pointer_to_base):
3321 * tree.def (TARGET_MEM_REF): Add comment on types of
3324 2007-01-11 Joseph Myers <joseph@codesourcery.com>
3326 * c-common.c (vector_types_convertible_p): Treat opaque types as
3327 always convertible if they have the same size, but not otherwise.
3329 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
3331 * ifcvt.c (struct noce_if_info): Add comments to the fields.
3332 Remove the b_unconditional field.
3333 (noce_try_sign_mask): Do not look at b_unconditional.
3334 (noce_process_if_block): Do not use merge_if_blocks. Update
3335 the CFG here. Do not set b_unconditional.
3336 (cond_move_process_if_block): Likewise.
3337 (find_cond_trap): Likewise.
3338 (check_cond_move_block): Require simple jump insns at the end
3341 2007-01-11 Jan Hubicka <jh@suse.cz>
3343 PR tree-optimization/1046
3344 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
3345 when alias info is not ready.
3346 (pass_tail_recursion): Do not require aliasing.
3347 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
3348 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
3349 * tree-ssa-copy.c (pass_copy_prop): Likewise.
3350 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
3351 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
3352 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
3353 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
3355 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
3356 When aliasing is not build, mark statement as volatile.
3358 2007-01-11 Tom Tromey <tromey@redhat.com>
3360 PR preprocessor/15185, PR preprocessor/20989:
3361 * doc/cppopts.texi <-MT>: Update description of algorithm for
3362 computing default target.
3363 <-M, -MD>: Reword "basename" text.
3365 2007-01-11 Roger Sayle <roger@eyesopen.com>
3367 * builtins.c (expand_builtin_pow, expand_builtin_powi,
3368 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
3369 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
3370 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
3371 fold_builtin_bswap, real_constp, fold_builtin_pow,
3372 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
3373 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
3374 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3375 * convert.c (convert_to_pointer): Likewise.
3376 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
3377 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
3378 const_binop, fold_convert_const_int_from_int,
3379 fold_convert_const_int_from_real,
3380 fold_convert_const_real_from_real, sign_bit_p,
3381 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
3382 fold_sign_changed_comparison, fold_unary, fold_comparison,
3383 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
3384 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
3385 * print-tree.c (print_node_brief, print_node): Likewise.
3386 * stor-layout.c (place_field, layout_type): Likewise.
3387 * tree-chrec.c (keep_cast): Likewise.
3388 * tree.c (build_vector, build_real, build_real_from_int_cst,
3389 build_complex): Likewise.
3391 2007-01-11 Roger Sayle <roger@eyesopen.com>
3393 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
3394 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
3396 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
3398 * configure.ac (strict1_warn): Rename to strict_warn.
3399 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
3400 cc_set_by_configure, quoted_cc_set_by_configure,
3401 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
3402 all_boot_languages, all_stagestuff): Remove.
3403 (target_list): Remove bootstrap targets.
3404 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
3405 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
3406 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
3407 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
3408 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
3409 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
3410 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
3411 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
3412 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
3413 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
3414 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
3415 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
3416 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
3417 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
3418 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
3419 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
3420 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
3421 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
3422 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
3423 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
3424 (mostlyclean): Adjust.
3425 (clean, distclean): Don't mention bootstrap stuff.
3426 * configure: Regenerate.
3427 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
3428 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
3429 treelang/config-lang.in (stagestuff): Remove.
3430 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
3431 stageprofile, stagefeedback, stagestuff): Remove mention.
3433 2007-01-11 Nick Clifton <nickc@redhat.com>
3435 * config/mcore/predicates.md (mcore_general_movesrc_operand):
3437 (mcore_general_movdst_operand): Do not accept CONST_INTs.
3438 (mcore_arith_K_S_operand): Run the test for the S constraint not
3439 the test for the M constraint.
3440 (mcore_addsub_operand): Do not accept integer values that are
3441 larger than 32 bits.
3442 * config/mcore/mcore.md: Remove unused constraints from split.
3443 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
3445 (allocate_stack): Likewise.
3446 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
3447 operands to 32 bits.
3448 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
3450 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
3451 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
3452 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
3453 mcore_output_bclri, mcore_output_andn, output_inline_const,
3454 mcore_output_move, mcore_output_movedouble): Likewise.
3455 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
3456 (output_inline_const): Likewise.
3457 (output_inline_const): Fix format strings used in sprintf
3459 * config/mcore/mcore-protos.h: Update prototypes for changed
3460 functions in mcore.c.
3461 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
3462 HOST_WIDE_INT and not int.
3463 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
3464 CONST_OK_FOR_N): Likewise.
3465 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
3466 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
3469 2007-01-10 Jan Hubicka <jh@suse.cz>
3471 * tree-vrp.c (remove_range_assertions): Release defs.
3472 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
3473 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
3475 2007-01-10 Paul Brook <paul@codesourcery.com>
3477 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
3478 * config/arm/arm.md (mulsi3subsi): New insn.
3480 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
3482 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
3483 (tree_transform_and_unroll_loop): New.
3484 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
3487 2007-01-10 Robert Kennedy <jimbob@google.com>
3489 * fold-const.c (fold_comparison): Fold comparisons like (x *
3490 1000 < 0) to (x < 0).
3492 2007-01-10 Ian Lance Taylor <iant@google.com>
3494 * tree-pretty-print.c (dump_generic_node): Print parentheses when
3495 operands have the same priority.
3497 2007-01-10 Tom Tromey <tromey@redhat.com>
3499 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
3500 (fold_binary): Likewise.
3501 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
3502 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
3503 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
3504 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
3505 * langhooks.h (struct lang_hooks): Removed field
3506 'can_use_bit_fields_p'.
3508 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
3510 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
3512 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
3514 * function.c (get_last_funcdef_no): New function.
3515 * function.h (get_last_funcdef_no): Declare.
3516 * tree-inline.c (initialize_cfun): Add initialization.
3517 (tree_function_versioning): Cleanup.
3519 2007-01-10 Jan Hubicka <jh@suse.cz>
3521 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
3524 2007-01-10 Sa Liu <saliu@de.ibm.com>
3525 Ben Elliston <bje@au.ibm.com>
3527 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
3528 for SAVE_NONLOCAL and Pmode for any other save level.
3529 * spu-protos.h (spu_restore_stack_block): Declare.
3530 * spu.md (save_stack_block): Remove.
3531 (restore_stack_block): Call spu_restore_stack_block.
3532 * spu.c (spu_restore_stack_block): New function.
3533 (spu_expand_epilogue): Remove old comment.
3535 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
3537 PR tree-optimization/30322
3538 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
3539 (cand_value_at): Return the value as aff_tree.
3540 (may_eliminate_iv): Convert the bound from aff_tree to tree.
3541 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
3542 aff_combination_mult): New functions.
3543 (aff_combination_add): Use aff_combination_add_cst.
3544 (aff_combination_convert): Allow conversions to a wider type.
3545 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
3546 * tree-affine.h (aff_combination_mult): Declare.
3548 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
3550 * doc/tm.texi: Update documentation to reflect reality of exec
3551 and start file search behaviours. Update copyright year.
3552 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
3553 header file directories.
3555 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
3557 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
3558 (*cosxf2): Rename to cosxf2_i387.
3559 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
3560 using X87MODEF12 mode macro. Rename patterns to
3561 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
3563 (*cosdf2, *cossf2): Ditto.
3564 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
3565 to match extended input operands.
3566 (sincos<mode>3): New expander.
3567 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
3568 insn patterns and corresponding splitters.
3570 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
3572 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
3573 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
3575 2007-01-09 Nicolas Pitre <nico@cam.org>
3578 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
3580 2007-01-08 Geoffrey Keating <geoffk@apple.com>
3582 * target.h (struct gcc_target): New field library_rtti_comdat.
3583 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
3584 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
3585 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
3586 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
3588 2007-01-08 Geoffrey Keating <geoffk@apple.com>
3590 * doc/invoke.texi (Optimize Options): Correct description of -O0.
3592 2007-01-08 Richard Guenther <rguenther@suse.de>
3594 * tree.h (force_fit_type_double): Export.
3595 (force_fit_type): Remove.
3596 * fold-const.c (force_fit_type_double): New function.
3597 (force_fit_type): Remove.
3598 (int_const_binop): Use it.
3599 (fold_convert_const_int_from_int): Likewise.
3600 (fold_convert_const_int_from_real): Likewise.
3601 (fold_div_compare): Likewise.
3602 (fold_sign_changed_comparison): Likewise.
3603 (fold_unary): Likewise.
3604 (fold_negate_const): Likewise.
3605 (fold_abs_const): Likewise.
3606 (fold_not_const): Likewise.
3607 * c-common.c (shorten_compare): Use force_fit_type_double.
3608 * convert.c (convert_to_pointer): Likewise.
3610 2007-01-08 Richard Guenther <rguenther@suse.de>
3612 * tree.h (build_int_cst_wide_type): Export.
3613 * tree.c (build_int_cst_wide_type): New function.
3614 (build_int_cst_wide): Fix comment.
3615 * builtins.c (fold_builtin_object_size): Use build_int_cst
3616 to build -1 or 0 of the correct type. Use fit_double_type
3617 to check for overflow.
3618 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
3620 (decode_field_reference): Likewise.
3621 (all_ones_mask_p): Likewise.
3622 (native_interpret_int): Use build_int_cst_wide_type.
3623 (fold_binary): Use build_int_cst_type to build an all-ones
3625 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
3627 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
3629 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
3631 2007-01-09 Ben Elliston <bje@au.ibm.com>
3633 * genautomata.c (STATS_OPTION): New option.
3634 (stats_flag): New flag.
3635 (gen_automata_option): Handle it.
3636 (initiate_automaton_gen): Ditto.
3637 (write_automata): Output statistics only if stats_flag is
3638 set. Likewise, output time statistics only if time_flag is set.
3639 * doc/md.texi (Processor pipeline description): Document new flag.
3641 2007-01-08 Richard Guenther <rguenther@suse.de>
3643 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
3644 * tree.c (build_int_cst_type): Likewise.
3645 (size_in_bytes): Don't call force_fit_type on the result.
3646 (int_fits_type_p): Use fit_double_type.
3647 * fold-const.c (fit_double_type): New function.
3648 (force_fit_type): Use it.
3649 * tree.h (fit_double_type): Export.
3651 2007-01-08 Jan Hubicka <jh@suse.cz>
3653 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
3654 * ipa.c (function_and_variable_visibility): Fix return type.
3656 2007-01-08 Richard Guenther <rguenther@suse.de>
3658 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
3659 of offset to build the index.
3660 * tree-pretty-print.c (dump_generic_node): Don't build negated
3661 const just for printing.
3662 * c-pretty-print.c (pp_c_integer_constant): Likewise.
3663 * builtins.c (fold_builtin_int_roundingfn): Check if result
3664 fits the type by using force_fit_type and comparing the result.
3665 * predict.c (predict_loops): Use compare_tree_int for comparison.
3666 * tree.c (build_int_cst): Fall back to integer_type_node for
3668 (build_int_cst_wide): Assert type is non-null.
3670 2007-01-08 Roberto Costa <roberto.costa@st.com>
3672 * tree-vrp.c (extract_range_from_cond_expr): New.
3673 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
3674 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
3676 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
3677 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
3678 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
3679 used as expressions.
3680 * tree-object-size.c (cond_expr_object_size): New.
3681 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
3683 2007-01-08 Jan Hubicka <jh@suse.cz>
3685 * tree-ssa-forwprop.c (forward_propagate_into_cond,
3686 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
3689 2007-01-08 Richard Guenther <rguenther@suse.de>
3691 PR tree-optimization/23603
3692 * tree-vrp.c (set_value_range_to_truthvalue): New function.
3693 (extract_range_from_binary): Fall back to truthvalue instead of
3694 varying for TRUTH_*_EXPR.
3695 (extract_range_from_comparison): Fall back to truthvalue instead of
3697 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
3698 if all visited PHI values were constant.
3700 2007-01-08 Jan Hubicka <jh@suse.cz>
3702 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
3703 (cgraph_analyze_function): break out from ...
3704 (cgraph_finalize_compilation_unit): ... here.
3705 (cgraph_expand_function): Remove forgoten commented out line.
3706 (cgraph_optimize): Analyze functions.
3708 2007-01-08 Jan Hubicka <jh@suse.cz>
3710 * tree-pas.h (TODO_remove_function): New flag.
3711 (TODO_update*): Renumber.
3712 (pass_ipa_increase_alignment,
3713 pass_ipa_function_and_variable_visibility): New passes.
3714 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
3715 (cgraph_function_and_variable_visibility): Move to ipa.c
3716 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
3717 cgraph_increase_alignment.
3718 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
3719 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
3720 calling tree-inline.
3721 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
3722 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
3723 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
3724 (gate_increase_alignment): New function.
3725 (pass_ipa_increase_alignment): New pass.
3726 * ipa.c: Inline tree-pass.h and timevar.h
3727 (function_and_variable_visibility): Move here from cgraphunit.c
3728 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
3729 * passes.c (init_optimization_passes): Add the two new passes.
3730 (execute_todo): Handle cgraph_remove_functions.
3732 2007-01-08 Nick Clifton <nickc@redhat.com>
3734 * config/frv/predicates.md (reg_or_0_operand): Accept
3737 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
3739 * config/bfin/rtems.h, config/bfin/t-rtems: New.
3740 * config.gcc: Add bfin*-rtems*.
3742 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
3744 * c.opt: Add -flax-vector-conversions.
3745 * c-typeck.c (convert_for_assignment): Pass flag to
3746 vector_types_convertible_p to allow emission of note.
3747 (digest_init): Likewise.
3748 * c-opts.c: Handle -flax-vector-conversions.
3749 * c-common.c (flag_lax_vector_conversions): New.
3750 (vector_types_convertible_p): Unless -flax-vector conversions
3751 has been passed, disallow conversions between vectors with
3752 differing numbers of subparts and/or element types. If such
3753 a conversion is disallowed, possibly emit a note on the first
3754 occasion only to inform the user of -flax-vector-conversions.
3755 The new last argument specifies this.
3756 * c-common.h (flag_lax_vector_conversions): New.
3757 (vector_types_convertible_p): Add extra argument.
3758 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
3759 char_type_node for V*QI type vectors.
3760 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
3761 Update to satisfy new typechecking rules.
3762 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
3764 * doc/invoke.texi (C Dialect Options): Document
3765 -flax-vector-conversions.
3767 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
3769 PR tree-optimization/29877
3770 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
3771 a register variable on the RHS to not be replaceable.
3773 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
3774 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
3775 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
3776 const_simm15, const_pow2, const_npow2): Added.
3777 * config/score/misc.md (insv, extv, extzv, movmemsi,
3778 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
3779 * config/score/score.c (score_address_cost, score_select_cc_mode):
3781 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
3782 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
3783 Update score_print_operand.
3784 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
3785 Adjust register allocate order and update some macro define.
3786 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
3787 mdx_block_move_straight, mdx_block_move_loop_head,
3788 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
3789 mdx_block_move): Added.
3790 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
3791 mdp_limm): Updated and fix some bug and typo.
3792 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
3793 (movsf, movdf, doloop_end): Added.
3795 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
3797 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
3798 config/arm/thumb2.md: Fix comment typos.
3799 * doc/extend.texi: Fix a typo.
3801 2007-01-07 Eric Christopher <echristo@apple.com>
3803 * configure.ac: Check for __stack_chk_fail for darwin.
3804 * configure: Regenerate.
3806 2007-01-07 Richard Guenther <rguenther@suse.de>
3808 * tree-vrp.c (extract_range_from_assert): CSE calls to
3809 compare_values where possible.
3810 (extract_range_from_unary_expr): Likewise.
3812 2007-01-07 Anatoly Sokolov <aesok@post.ru>
3814 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
3815 * config/avr/avr.c (call_insn_operand): Delete function.
3816 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
3817 instead of register number. Use predicates.
3818 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
3819 call_insn_operand): Add.
3821 2007-01-06 Jan Hubicka <jh@suse.cz>
3823 * tree-pass.h (pass_build_cgraph_edges): Declare.
3824 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
3825 (visited_nodes): Remove.
3826 (cgraph_create_edges): Move to cgraphbuild.c; rename to
3827 build_cgrpah_edges; make visited_nodes local.
3828 (cgraph_process_new_functions): DO not call initialize_inline_failed.
3829 (record_references_in_initializer): Move to cgraphbuild.c
3830 (initialize_inline_failed, rebuild_cgraph_edges,
3831 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
3832 (verify_cgraph_node): Make visited_nodes local.
3833 (cgraph_analyze_function): Do not call cgraph_create_edges and
3834 initialize_inline_failed.
3835 (cgraph_expand_function): Do not call cgraph_lower_function;
3836 assert that function is already lowered.
3837 * Makefile.in (cgraphbuild.o): New.
3838 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
3839 at the end of lowering passes.
3841 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
3843 * ifcvt.c (cond_move_convert_if_block): New function, code
3844 factored out from...
3845 (cond_move_process_if_block): ...here. Call the new function
3846 on the THEN and ELSE blocks.
3847 (merge_if_block): Do not copy global_live_at_end, merge_blocks
3848 already takes care of this.
3850 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3853 * tree.h (TREE_OVERFLOW_P): New.
3854 * c-typeck.c (parser_build_unary_op): Warn only if result
3855 overflowed and operands did not.
3856 (parser_build_binary_op): Likewise.
3857 (convert_for_assignment): Remove redundant overflow_warning.
3858 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
3860 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3862 * c-typeck.c (store_init_value): Split over two lines to follow
3863 the GNU coding style.
3865 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
3867 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
3868 __GXX_EXPERIMENTAL_CXX0X__.
3869 * doc/cpp.texi: Same.
3871 2007-01-05 Richard Guenther <rguenther@suse.de>
3874 * tree.c (get_narrower): Do not construct COMPONENT_REFs
3875 with mismatched types. Instead explicitly build a
3876 conversion NOP_EXPR.
3878 2007-01-05 Ian Lance Taylor <iant@google.com>
3880 * c-common.c (decl_with_nonnull_addr_p): New function.
3881 (c_common_truthvalue_conversion): Call it.
3882 * c-typeck.c (build_binary_op): Likewise.
3883 * c-common.h (decl_with_nonnull_addr_p): Declare.
3885 2007-01-05 Jakub Jelinek <jakub@redhat.com>
3888 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
3891 2007-01-05 Joel Brobecker <brobecker@adacore.com>
3893 * doc/install.texi (Final install): Document the fact that
3894 the GNAT runtime should not be stripped.
3896 2007-01-04 Jan Hubicka <jh@suse.cz>
3898 * tree-inline.c (fold_marked_statements): Update operand caches
3899 and EH after folding
3901 2007-01-04 Ian Lance Taylor <iant@google.com>
3903 * c-common.c (check_function_nonnull): Whitespace fix.
3905 2007-01-04 Jan Hubicka <jh@suse.cz>
3907 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
3908 comitted older version of patch.
3909 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
3910 function, celanup cfg and collect garbage.
3912 2007-01-04 Mike Stump <mrs@apple.com>
3914 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
3917 2007-01-04 Eric Christopher <echristo@apple.com>
3919 * libgcc2.c (__bswapsi2): Use SItype.
3920 (__bswapdi2): Use DItype.
3921 * libgcc2.h: Update for above.
3923 2007-01-04 Paul Brook <paul@codesourcery.com>
3925 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
3926 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
3927 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
3928 umulsidi3, umulsidi3adddi, smulsi3_highpart,
3929 umulsi3_highpart): Make conditional on !arm_arch6.
3930 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
3931 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
3932 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
3933 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
3934 umulsi3_highpart_v6): New insns.
3936 2007-01-04 Roger Sayle <roger@eyesopen.com>
3938 * fold-const.c (fold_convert): When casting an expression to void,
3939 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
3940 have a type. Instead of attempting to build a NOP_EXPR, return
3941 these "special" trees directly.
3943 2007-01-04 Joseph Myers <joseph@codesourcery.com>
3945 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
3946 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
3947 rs6000_cost->fp not 0 as adjustment for outer NEG.
3949 2007-01-04 Jan Hubicka <jh@suse.cz>
3951 * cgraph.c (cgraph_release_function_body): New function.
3952 (cgraph_remove_node): Use it.
3953 * cgraph.h (cgraph_release_function_body): Declare.
3954 * cgraphunit.c (cgraph_expand_function): Use it.
3955 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
3956 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
3957 is initialized and while compilation of other function is running.
3958 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
3959 statement CFG annotations from here to ...
3960 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
3962 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
3964 * cfgloop.h (enum li_flags): Make the constants powers of two.
3966 2007-01-04 Jan Hubicka <jh@suse.cz>
3968 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
3970 (fold_marked_statements): New function.
3971 (optimize_inline_calls, tree_function_versioning): Fold new statements.
3972 * tree-inline.h (copy_body_data): Add statements_to_fold.
3974 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
3976 * config.gcc: Mention libgcc/config.host.
3977 * Makefile.in: Update comments mentioning libgcc.
3978 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
3979 (all.cross, start.encap, rest.encap, rest.cross): Update
3980 dependencies for libgcc move.
3981 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
3982 (clean-target, clean-target-libgcc): Delete.
3983 (srcdirify, GCC_EXTRA_PARTS): New macros.
3984 (libgcc-support, libgcc.mvars): New rules.
3985 (distclean): Remove mention of mklibgcc.
3986 (install): Don't reference INSTALL_LIBGCC.
3987 (install-common): Don't reference EXTRA_PARTS.
3988 (install-libgcc, install-multilib): Delete rules.
3989 * mklibgcc.in: Delete file.
3990 * doc/configfiles.texi: Don't mention mklibgcc.
3992 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
3993 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
3994 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
3995 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
3996 (mklibgcc, ldblspecs): Likewise.
3998 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
3999 (posixpre.def): Use $(T).
4000 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
4001 (SHLIB_SRC, SHLIB_INSTALL): Delete.
4002 (SHLIB_LINK): Make dummy.
4003 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
4005 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
4007 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
4008 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
4009 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
4010 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
4012 * configure.ac (all_outputs): Remove mklibgcc.
4013 * configure: Regenerated.
4015 2007-01-03 Josh Conner <jconner@apple.com>
4018 * calls.c (compute_argument_addresses): Set stack and stack_slot
4019 for partial args, too.
4020 (store_one_arg): Use locate.size.constant for the size when
4021 generating a save_area.
4023 2007-01-03 Robert Kennedy <jimbob@google.com>
4025 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
4026 whose definitions are deleted due to basic block merging.
4028 2007-01-03 Paul Brook <paul@codesourcery.com>
4031 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
4033 (use_return_insn): Return 0 for Thumb interrupt functions.
4034 (print_multi_reg): Add rfe argument for IRQ returns.
4035 (arm_output_epilogue): Pop interrupt return address directly into PC.
4036 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
4038 2007-01-03 Paul Brook <paul@codesourcery.com>
4040 Merge from sourcerygxx-4_1.
4041 * config/arm/thumb2.md: New file.
4042 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
4044 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
4045 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4046 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
4047 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
4049 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4050 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
4052 * config/arm/ieee754-sf.S: Ditto.
4053 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
4054 (thumb1_base_register_rtx_p): ... to this.
4055 (thumb_index_register_rtx_p): Rename...
4056 (thumb1_index_register_rtx_p): ... to this.
4057 (thumb_output_function_prologue): Rename...
4058 (thumb1_output_function_prologue): ... to this.
4059 (thumb_legitimate_address_p): Rename...
4060 (thumb1_legitimate_address_p): ... to this.
4061 (thumb_rtx_costs): Rename...
4062 (thumb1_rtx_costs): ... to this.
4063 (thumb_compute_save_reg_mask): Rename...
4064 (thumb1_compute_save_reg_mask): ... to this.
4065 (thumb_final_prescan_insn): Rename...
4066 (thumb1_final_prescan_insn): ... to this.
4067 (thumb_expand_epilogue): Rename...
4068 (thumb1_expand_epilogue): ... to this.
4069 (arm_unwind_emit_stm): Rename...
4070 (arm_unwind_emit_sequence): ... to this.
4071 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
4072 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
4073 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
4074 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
4075 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
4076 thumb2_output_casesi): New functions.
4077 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
4078 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
4079 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
4080 THUMB2_WORK_REGS): Define.
4081 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
4082 arm_condexec_mask, arm_condexec_masklen)): New variables.
4083 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
4084 (arm_override_options): Check new CPU capabilities.
4085 Set new architecture flag variables.
4086 (arm_isr_value): Handle v7m interrupt functions.
4087 (user_return_insn): Return 0 for v7m interrupt functions. Handle
4089 (const_ok_for_arm): Handle Thumb-2 constants.
4090 (arm_gen_constant): Ditto. Use movw when available.
4091 (arm_function_ok_for_sibcall): Return false for v7m interrupt
4093 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
4094 (thumb_find_work_register, arm_load_pic_register,
4095 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
4096 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
4097 print_multi_reg, output_mov_long_double_fpa_from_arm,
4098 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
4099 output_mov_double_fpa_from_arm, output_move_double,
4100 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
4101 output_return_instruction, arm_output_function_prologue,
4102 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
4103 arm_output_mi_thunk, thumb_set_return_address): Ditto.
4104 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
4105 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
4106 (arithmetic_instr, shift_op): Use arm_shift_nmem.
4107 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
4109 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
4110 (thumb_expand_prologue): Use thumb_set_frame_pointer.
4111 (arm_file_start): Output directive for unified syntax.
4112 (arm_unwind_emit_set): Handle stack alignment instruction.
4113 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
4114 Add v6t2, v7, v7a, v7r and v7m.
4115 (RETLDM): Add Thumb-2 code.
4116 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
4117 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
4118 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
4119 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
4120 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
4121 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
4122 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
4123 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
4124 ADJUST_INSN_LENGTH): Define.
4125 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
4126 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
4127 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
4128 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
4129 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
4130 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
4131 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
4132 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
4133 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
4134 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
4135 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
4136 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
4137 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
4138 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
4139 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
4140 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
4141 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
4143 * config/arm/arm-tune.md: Regenerate.
4144 * config/arm/arm-protos.h: Update prototypes.
4145 * config/arm/vfp.md: Enable patterns for Thumb-2.
4146 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
4147 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
4148 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
4149 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
4150 * config/arm/libunwind.S: Add Thumb-2 code.
4151 * config/arm/constraints.md: Update include Thumb-2.
4152 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
4153 * config/arm/ieee754-df.S: Ditto.
4154 * config/arm/bpabi.S: Ditto.
4155 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
4156 * config/arm/predicates.md (low_register_operand,
4157 low_reg_or_int_operand, thumb_16bit_operator): New.
4158 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
4159 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
4160 * config/arm/t-arm-elf: Add armv7 multilib.
4161 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
4163 (UNSPEC_STACK_ALIGN, ce_count): New.
4164 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
4165 arm_zero_extendsidi2, arm_zero_extendqidi2): New
4167 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
4168 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
4169 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
4170 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
4171 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
4172 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
4174 * doc/extend.texi: Document ARMv7-M interrupt functions.
4175 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
4177 2007-01-03 Jakub Jelinek <jakub@redhat.com>
4179 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
4180 (struct _Unwind_Context): Rename args_size to flags, remove
4181 signal_frame field, add a new args_size field and version field.
4182 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
4183 _Unwind_IsExtendedContext): New inline functions.
4184 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
4185 Assume by_value array is only present if _Unwind_IsExtendedContext.
4186 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
4187 _Unwind_IsSignalFrame.
4188 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
4189 (uw_update_context_1): Use _Unwind_SetSignalFrame.
4190 (uw_init_context_1): Initialize context->flags to
4191 EXTENDED_CONTEXT_BIT.
4192 * config/rs6000/linux-unwind.h (frob_update_context): Use
4193 _Unwind_SetSignalFrame.
4195 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
4198 * gimplify.c (gimplify_modify_expr_complex_part): Move below
4199 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
4202 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
4204 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
4205 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
4206 RET, AT_SP): Remove.
4207 * config/i386/i386.md (*sse_prologue_save_insn): Use return
4210 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4211 * cfgbuild.c (find_basic_blocks): Likewise.
4212 * cfgrtl.c (rtl_create_basic_block): Likewise.
4213 * function.c (temp_slots_at_level): Likewise.
4214 * reg-stack.c (stack_regs_mentioned): Likewise.
4215 * regclass.c (allocate_reg_info): Likewise.
4216 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4217 set_bb_for_stmt, move_block_to_fn): Likewise.
4218 * tree-complex.c (tree_lower_complex): Likewise.
4219 * vec.h (VEC_safe_grow_cleared): New.
4221 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
4223 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
4225 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
4226 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
4228 * cfgloopmanip.c (scale_loop_frequencies): Export.
4229 (loopify): Scale the frequencies by prescribed coefficients.
4230 (set_zero_probability): New function.
4231 (duplicate_loop_to_header_edge): Improve updating of frequencies.
4232 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
4233 and frequencies according to arguments.
4234 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
4236 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
4237 * modulo-sched.c (sms_schedule): Set probabilities for entering
4238 versioned loop correctly.
4239 * tree-vect-transform.c (vect_transform_loop): Ditto.
4240 * cfgloop.h (loopify, loop_version): Declaration changed.
4241 (scale_loop_frequencies): Declared.
4243 2007-01-02 Jan Hubicka <jh@suse.cz>
4245 * cgraph.c: Include tree-flow.h
4246 (cgraph_add_new-function): Handle IPA_SSA mode; execute
4248 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
4249 * tree-pass.h (pass_all_early_optimizations): Declare.
4250 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
4252 (cgraph_analyze_function): Do early_local_passes.
4253 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4254 Do not add referenced vars.
4255 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
4257 (gate_all_early_local_passes): New gate.
4258 (pass_early_local_passes): Use new gate.
4259 (execute_early_local_optimizations): New functions.
4260 (gate_all_early_optimizations): New gate.
4261 (pass_all_early_optimizations): New pass.
4262 (execute_free_datastructures): Free SSA only when initialized.
4263 (gate_init_datastructures): Init only when optimizing.
4264 (tree_lowering_passes): Do early local passes when called late.
4265 * tree-profile.c (do_tree_profiling): Don't profile functions added
4267 (do_early_tree_profiling, pass_early_tree_profile): Kill.
4268 * tree-cfg.c (update_modified_stmts): Do not update when operands are
4270 * passes.c (init_optimizations_passes): Reorder so we go into SSA
4271 during early_local_passes.
4272 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
4275 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4277 * Makefile.in: Update copyright year.
4279 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4281 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
4282 in $(lang_checks) and check-consistency targets.
4284 2007-01-02 Jan Hubicka <jh@suse.cz>
4286 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4287 Do not add referenced vars.
4288 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
4290 * passes.c (init_optimization_passes): Put mudflap_2 after
4291 free_datastructures.
4293 2007-01-02 Jan Hubicka <jh@suse.cz>
4295 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
4296 Set NOTHROW flag on call statements proved to be nothrow.
4297 Update statement of local calls so new pure/const functions are
4298 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
4300 (tree_rest_of_compilation): Register hooks and initialize bitmap
4301 early. Do not set after_inlining flag.
4303 2007-01-02 Steve Ellcey <sje@cup.hp.com>
4305 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
4306 HOST_BITS_PER_LONGLONG
4308 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4311 * c-typeck.c (store_init_value): Don't emit pedantic overflow
4312 warning for non-static initializers.
4314 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
4316 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
4317 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
4318 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
4319 Always use set_unique_reg_note to add REG_EQUAL notes.
4321 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4324 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4326 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4327 * cfgbuild.c (find_basic_blocks): Likewise.
4328 * cfgrtl.c (rtl_create_basic_block): Likewise.
4329 * function.c (temp_slots_at_level): Likewise.
4330 * reg-stack.c (stack_regs_mentioned): Likewise.
4331 * regclass.c (allocate_reg_info): Likewise.
4332 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4333 set_bb_for_stmt, move_block_to_fn): Likewise.
4334 * tree-complex.c (tree_lower_complex): Likewise.
4335 * vec.h (VEC_safe_grow_cleared): New.
4337 2007-01-02 Ian Lance Taylor <iant@google.com>
4339 * c-common.c (c_common_truthvalue_conversion): When warning about
4340 using an assignment as a truth value, set TREE_NO_WARNING.
4342 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4345 * c.opt (Wold-style-declaration): New.
4346 * doc/invoke.texi (C-only Warning Options): New.
4347 (Wold-style-declaration): Document it.
4348 (Wextra): Enabled by -Wextra.
4349 * c-opts.c (c_common_post_options): Enabled by -Wextra.
4350 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
4351 -Wold-style-declaration.
4353 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4355 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4356 * cfgbuild.c (find_basic_blocks): Likewise.
4357 * cfgrtl.c (rtl_create_basic_block): Likewise.
4358 * function.c (temp_slots_at_level): Likewise.
4359 * reg-stack.c (stack_regs_mentioned): Likewise.
4360 * regclass.c (allocate_reg_info): Likewise.
4361 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4362 set_bb_for_stmt, move_block_to_fn): Likewise.
4363 * tree-complex.c (tree_lower_complex): Likewise.
4364 * vec.h (VEC_safe_grow_cleared): New.
4366 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
4368 * c-common.c (c_common_nodes_and_builtins): Since variants of
4369 void_type_node get built before it is given a name, we need to
4370 give those variants the name, too.
4371 (complete_array_type): We need to work with the canonical main
4372 type of the array, from which we will build the qualified version.
4373 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
4374 * print-tree.c (print_node): Display canonical type information
4376 * stor-layout.c (layout_type): When we don't know the
4377 alignment of a type for which we're building an array, we end up
4378 guessing wrong, so make the type require structural equality.
4379 * tree.c (make_node_stat): When we build a new type, it is its
4381 (build_type_attribute_qual_variant): When building an attribute
4382 variant, its canonical type is the non-attribute variant. However,
4383 if the attributes are target-dependent and they differ, we need to
4384 use structural equality checks for this type.
4385 (build_qualified_type): A qualified type is not equivalent to its
4386 unqualified variant; set the canonical type appropriately.
4387 (build_distinct_type_copy): When building a distinct type from
4388 another type, the new type is its own canonical type.
4389 (build_variant_type_copy): When building a new type variant, we
4390 assume that it is equivalent to the original type.
4391 (build_pointer_type_for_mode): When building a pointer type, also
4392 build a canonical type pointer.
4393 (build_reference_type_for_mode): When building a reference type,
4394 also build a canonical type reference.
4395 (build_index_type): When we can't hash an index type (e.g.,
4396 because its maximum value is negative), the index type requires
4397 structural equality tests.
4398 (build_array_type): Build the canonical form of an array type.
4399 (build_function_type): Function types require structural equality,
4400 because they contain default arguments, attributes, etc.
4401 (build_method_type_directly): Ditto for method types.
4402 (build_offset_type): Build the canonical offset type.
4403 (build_complex_type): Build the canonical vector type.
4404 (make_vector_type): Build the canonical vector type.
4405 * tree.h (TYPE_CANONICAL): New.
4406 (TYPE_STRUCTURAL_EQUALITY_P): New.
4407 (SET_TYPE_STRUCTURAL_EQUALITY): New.
4408 (struct tree_type): Added "canonical" field.
4409 * params.h (VERIFY_CANONICAL_TYPES): New.
4410 * doc/c-tree.texi (TYPE_CANONICAL): Document.
4411 (TYPE_STRUCTURAL_EQUALITY_P): Document.
4412 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
4413 * doc/invoke.texi (verify-canonical-types): Document --param
4414 parameter for verifying canonical types.
4416 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4418 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
4419 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
4420 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
4421 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
4423 * config/rs6000/e500.h: New.
4424 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
4425 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
4427 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
4428 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
4429 TARGET_E500_DOUBLE): Remove.
4430 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
4431 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
4432 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
4433 * config/rs6000/rs6000.c (rs6000_override_options): Use
4436 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4438 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
4439 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
4440 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
4441 rs6000_emit_cmove): Don't check TARGET_E500.
4442 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
4443 sordered): Don't check TARGET_E500.
4445 2007-01-01 Eric Christopher <echristo@apple.com>
4447 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
4448 HARD_REGISTER_NUM_P.
4450 2007-01-01 Roger Sayle <roger@eyesopen.com>
4452 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
4453 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
4454 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
4457 2007-01-01 Mike Stump <mrs@apple.com>
4459 * configure.ac: Remove support for building with Apple's gcc-3.1.
4461 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4464 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
4465 REG is a hard register. Do not modify REG before calling
4467 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
4468 multiple of the size of YMODE for certain lowpart cases.
4470 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4473 * gimplify (voidify_wrapper_expr): Update for
4476 2007-01-01 Andreas Schwab <schwab@suse.de>
4479 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
4480 save of BR0 in extra_spill_size instead of spill_size.
4481 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
4483 (ia64_expand_epilogue): Restore BR0 from its new location.
4485 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4487 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
4488 Use a temporary variable if the left hand side is not a gimple
4491 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4493 * gimplify.c (gimplify_return_expr): Make the temporary variable
4494 for the return expression, a gimple register variable.
4496 2007-01-01 Jan Hubicka <jh@suse.cz>
4498 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
4501 2007-01-01 Mike Stump <mrs@apple.com>
4503 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
4505 * config/darwin.c (machopic_select_rtx_section): Use
4507 (darwin_mergeable_constant_section): Likewise.
4508 * configure: Regenerate.
4509 * config.in: Regenerate.
4511 2007-01-01 Jan Hubicka <jh@suse.cz>
4512 Andrew Pinski <pinskia@gmail.com>
4514 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
4515 before starting IPA passes.