OSDN Git Service

8badc160f2caf591032ca461f2606f9e7e60d2d5
[pf3gnuchains/gcc-fork.git] / gcc / ChangeLog
1 2007-02-06  Dorit Nuzman  <dorit@il.ibm.com>
2             Victor Kaplansky  <victork@il.ibm.com>
3
4         * tree-vectorizer.c (vect_is_simple_use): Support induction.
5         (vect_is_simple_reduction): Support reduction with induction as
6         one of the operands.
7         (vect_is_simple_iv_evolution): Fix formatting.
8         * tree-vect-analyze.c (vect_mark_stmts_to_be_vectorized): Fix 
9         formatting.  Don't mark induction phis for vectorization.
10         (vect_analyze_scalar_cycles): Analyze all inductions, then reductions.
11         * tree-vect-transform.c (get_initial_def_for_induction): New function.
12         (vect_get_vec_def_for_operand): Support induction.
13         (vect_get_vec_def_for_stmt_copy): Fix formatting and add check for
14         induction case.
15         (vectorizable_reduction): Support reduction with induction as one of 
16         the operands. 
17         (vectorizable_type_demotion): Use def-type of stmt argument rather
18         than dummy def-type.
19
20         * tree-ssa-loop.c (gate_scev_const_prop): Return the value of
21         flag_tree_scev_cprop.
22         * common.opt (tree-scev-cprop): New flag.
23
24         * tree-vect-transform.c (vect_create_destination_var): Use 'kind' in
25         call to vect_get_new_vect_var.
26
27 2007-02-06  Ira Rosen  <irar@il.ibm.com>
28
29         * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Check that 
30         vectype is not NULL.
31         (vect_pattern_recog_1): Likewise.
32
33 2007-02-05  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
34
35         * fold-const.c (negate_expr_p): Handle CONJ_EXPR.
36         (fold_negate_expr): Likewise.
37
38 2007-02-05  Alexandre Oliva  <aoliva@redhat.com>
39
40         PR debug/30189
41         * dwarf2out.c (modified_type_die): Follow DECL_ORIGINAL_TYPE
42         even if cv-qualification is the same.
43
44 2007-02-05  Geoffrey Keating  <geoffk@apple.com>
45
46         * config/rs6000/darwin-tramp.asm (__trampoline_setup): Call
47         __enable_execute_stack on completion.
48
49 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
50
51         * config/i386/athlon.md (athlon_fldxf_k8, athlon_fld_k8,
52         athlon_fstxf_k8, athlon_fst_k8, athlon_fist, athlon_fmov,
53         athlon_fadd_load, athlon_fadd_load_k8, athlon_fadd, athlon_fmul,
54         athlon_fmul_load, athlon_fmul_load_k8, athlon_fsgn,
55         athlon_fdiv_load, athlon_fdiv_load_k8, athlon_fdiv_k8,
56         athlon_fpspc_load, athlon_fpspc, athlon_fcmov_load,
57         athlon_fcmov_load_k8, athlon_fcmov_k8, athlon_fcomi_load_k8,
58         athlon_fcomi, athlon_fcom_load_k8, athlon_fcom): Added amdfam10.
59
60 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
61
62         * config/i386/i386.md (x86_sahf_1, cmpfp_i_mixed, cmpfp_i_sse,
63         cmpfp_i_i387, cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387,
64         swapsi, swaphi_1, swapqi_1, swapdi_rex64, fix_truncsfdi_sse,
65         fix_truncdfdi_sse, fix_truncsfsi_sse, fix_truncdfsi_sse,
66         x86_fldcw_1, floatsisf2_mixed, floatsisf2_sse, floatdisf2_mixed,
67         floatdisf2_sse, floatsidf2_mixed, floatsidf2_sse,
68         floatdidf2_mixed, floatdidf2_sse, muldi3_1_rex64, mulsi3_1,
69         mulsi3_1_zext, mulhi3_1, mulqi3_1, umulqihi3_1, mulqihi3_insn,
70         umulditi3_insn, umulsidi3_insn, mulditi3_insn, mulsidi3_insn,
71         umuldi3_highpart_rex64, umulsi3_highpart_insn,
72         umulsi3_highpart_zext, smuldi3_highpart_rex64,
73         smulsi3_highpart_insn, smulsi3_highpart_zext, x86_64_shld,
74         x86_shld_1, x86_64_shrd, sqrtsf2_mixed, sqrtsf2_sse,
75         sqrtsf2_i387, sqrtdf2_mixed, sqrtdf2_sse, sqrtdf2_i387,
76         sqrtextendsfdf2_i387, sqrtxf2, sqrtextendsfxf2_i387,
77         sqrtextenddfxf2_i387): Added amdfam10_decode.
78         
79         * config/i386/athlon.md (athlon_idirect_amdfam10,
80         athlon_ivector_amdfam10, athlon_idirect_load_amdfam10,
81         athlon_ivector_load_amdfam10, athlon_idirect_both_amdfam10,
82         athlon_ivector_both_amdfam10, athlon_idirect_store_amdfam10,
83         athlon_ivector_store_amdfam10): New define_insn_reservation.
84         (athlon_idirect_loadmov, athlon_idirect_movstore): Added
85         amdfam10.
86
87 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
88
89         * config/i386/athlon.md (athlon_call_amdfam10,
90         athlon_pop_amdfam10, athlon_lea_amdfam10): New
91         define_insn_reservation.
92         (athlon_branch, athlon_push, athlon_leave_k8, athlon_imul_k8,
93         athlon_imul_k8_DI, athlon_imul_mem_k8, athlon_imul_mem_k8_DI,
94         athlon_idiv, athlon_idiv_mem, athlon_str): Added amdfam10.
95
96 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
97
98         * config/i386/athlon.md (athlon_sseld_amdfam10,
99         athlon_mmxld_amdfam10, athlon_ssest_amdfam10,
100         athlon_mmxssest_short_amdfam10): New define_insn_reservation.
101
102 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
103
104         * config/i386/athlon.md (athlon_sseins_amdfam10): New
105         define_insn_reservation.
106         * config/i386/i386.md (sseins): Added sseins to define_attr type
107         and define_attr unit.
108         * config/i386/sse.md: Set type attribute to sseins for insertq
109         and insertqi.
110
111 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
112
113         * config/i386/athlon.md (sselog_load_amdfam10, sselog_amdfam10,
114         ssecmpvector_load_amdfam10, ssecmpvector_amdfam10,
115         ssecomi_load_amdfam10, ssecomi_amdfam10,
116         sseaddvector_load_amdfam10, sseaddvector_amdfam10): New
117         define_insn_reservation.
118         (ssecmp_load_k8, ssecmp, sseadd_load_k8, seadd): Added amdfam10.
119
120 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
121
122         * config/i386/athlon.md (cvtss2sd_load_amdfam10,
123         cvtss2sd_amdfam10, cvtps2pd_load_amdfam10, cvtps2pd_amdfam10,
124         cvtsi2sd_load_amdfam10, cvtsi2ss_load_amdfam10,
125         cvtsi2sd_amdfam10, cvtsi2ss_amdfam10, cvtsd2ss_load_amdfam10,
126         cvtsd2ss_amdfam10, cvtpd2ps_load_amdfam10, cvtpd2ps_amdfam10,
127         cvtsX2si_load_amdfam10, cvtsX2si_amdfam10): New 
128         define_insn_reservation.
129
130         * config/i386/sse.md (cvtsi2ss, cvtsi2ssq, cvtss2si,
131         cvtss2siq, cvttss2si, cvttss2siq, cvtsi2sd, cvtsi2sdq,
132         cvtsd2si, cvtsd2siq, cvttsd2si, cvttsd2siq,
133         cvtpd2dq, cvttpd2dq, cvtsd2ss, cvtss2sd,
134         cvtpd2ps, cvtps2pd): Added amdfam10_decode attribute.
135
136 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
137
138         * config/i386/athlon.md (athlon_ssedivvector_amdfam10,
139         athlon_ssedivvector_load_amdfam10, athlon_ssemulvector_amdfam10,
140         athlon_ssemulvector_load_amdfam10): New define_insn_reservation.
141         (athlon_ssediv, athlon_ssediv_load_k8, athlon_ssemul,
142         athlon_ssemul_load_k8): Added amdfam10.
143
144 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
145
146         * config/i386/i386.h (TARGET_SSE_UNALIGNED_MOVE_OPTIMAL): New macro.
147         (x86_sse_unaligned_move_optimal): New variable.
148         
149         * config/i386/i386.c (x86_sse_unaligned_move_optimal): Enable for  
150         m_AMDFAM10.
151         (ix86_expand_vector_move_misalign): Add code to generate movupd/movups
152         for unaligned vector SSE double/single precision loads for AMDFAM10.
153
154 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
155
156         * config/i386/i386.h (TARGET_AMDFAM10): New macro.
157         (TARGET_CPU_CPP_BUILTINS): Add code for amdfam10.
158         Define TARGET_CPU_DEFAULT_amdfam10.
159         (TARGET_CPU_DEFAULT_NAMES): Add amdfam10.
160         (processor_type): Add PROCESSOR_AMDFAM10.       
161         
162         * config/i386/i386.md: Add amdfam10 as a new cpu attribute to match
163         processor_type in config/i386/i386.h.
164         Enable imul peepholes for TARGET_AMDFAM10.
165         
166         * config.gcc: Add support for --with-cpu option for amdfam10.
167         
168         * config/i386/i386.c (amdfam10_cost): New variable.
169         (m_AMDFAM10): New macro.
170         (m_ATHLON_K8_AMDFAM10): New macro.
171         (x86_use_leave, x86_push_memory, x86_movx, x86_unroll_strlen,
172         x86_cmove, x86_3dnow_a, x86_deep_branch, x86_use_simode_fiop,
173         x86_promote_QImode, x86_integer_DFmode_moves,
174         x86_partial_reg_dependency, x86_memory_mismatch_stall, 
175         x86_accumulate_outgoing_args, x86_arch_always_fancy_math_387,
176         x86_sse_partial_reg_dependency, x86_sse_typeless_stores,
177         x86_use_ffreep, x86_use_incdec, x86_four_jump_limit,
178         x86_schedule, x86_use_bt, x86_cmpxchg16b, x86_pad_returns):
179         Enable/disable for amdfam10.
180         (override_options): Add amdfam10_cost to processor_target_table.
181         Set up PROCESSOR_AMDFAM10 for amdfam10 entry in 
182         processor_alias_table.
183         (ix86_issue_rate): Add PROCESSOR_AMDFAM10.
184         (ix86_adjust_cost): Add code for amdfam10.
185
186 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
187         
188         * config/i386/i386.opt: Add new Advanced Bit Manipulation (-mabm)
189         instruction set feature flag. Add new (-mpopcnt) flag for popcnt 
190         instruction. Add new SSE4A (-msse4a) instruction set feature flag.
191         * config/i386/i386.h: Add builtin definition for SSE4A.
192         * config/i386/i386.md: Add support for ABM instructions 
193         (popcnt and lzcnt).
194         * config/i386/sse.md: Add support for SSE4A instructions
195         (movntss, movntsd, extrq, insertq).
196         * config/i386/i386.c: Add support for ABM and SSE4A builtins.
197         Add -march=amdfam10 flag.
198         * config/i386/ammintrin.h: Add support for SSE4A intrinsics.
199         * doc/invoke.texi: Add documentation on flags for sse4a, abm, popcnt
200         and amdfam10.
201         * doc/extend.texi: Add documentation for SSE4A builtins.
202
203 2007-02-05  Bob Wilson  <bob.wilson@acm.org>
204
205         * config/xtensa/xtensa.c (constantpool_mem_p): Skip over SUBREGs.
206
207 2007-02-05  Richard Guenther  <rguenther@suse.de>
208
209         * tree-vectorizer.h (vectorizable_function): Add argument type
210         argument, change return type.
211         * tree-vect-patterns.c (vect_recog_pow_pattern): Adjust caller.
212         * tree-vect-transform.c (vectorizable_function): Handle extra
213         argument, return vectorized function decl.
214         (build_vectorized_function_call): Remove.
215         (vectorizable_call): Handle calls with result and argument types
216         differing.  Handle loop vectorization factor correctly.
217         * targhooks.c (default_builtin_vectorized_function): Adjust for
218         extra argument.
219         * targhooks.h (default_builtin_vectorized_function): Likewise.
220         * target.h (builtin_vectorized_function): Add argument type
221         argument.
222         * config/i386/i386.c (ix86_builtin_vectorized_function): Handle
223         extra argument, allow vectorizing of lrintf.
224         * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION): Adjust
225         documentation of target hook.
226
227 2007-02-05  Hans-Peter Nilsson  <hp@axis.com>
228
229         PR target/30665
230         * config/cris/cris.md ("*andsi_movu", "*andsi_clear", "*andhi_movu")
231         ("*andhi_clear", andu (casesi+45)): For size-changed operand where
232         memory is allowed, require !side_effects_p, not just !MEM_VOLATILE_P.
233
234 2007-02-05  Roger Sayle  <roger@eyesopen.com>
235
236         * fold-const.c (fold_unary) <REAL_PART>: Test for availability of
237         BUILT_IN_COS before simplifying REAL_PART(CEXPI)) to COS.
238         <IMAG_PART>: Likewise, check for availability of BUILT_IN_SIN.
239         * builtins.c (fold_builtin_sincos): Check for TARGET_C99_FUNCTIONS
240         before canonicalizing sincos to cexpi.
241         (fold_builtin_cexp): Likewise, for canonicalizing cexp to cexpi.
242
243 2007-02-05  Roger Sayle  <roger@eyesopen.com>
244
245         * config/alpha/alpha.c (alpha_add_builtins): New Helper function.
246         Set TREE_READONLY and TREE_NOTHROW directly, not via attributes.
247         (alpha_init_builtins): Use alpha_add_builtins to process tables.
248
249 2007-02-05  Roger Sayle  <roger@eyesopen.com>
250
251         * mips-tfile.c (initialize_init_file): Correct endianness test.
252
253 2007-02-05  Kazu Hirata  <kazu@codesourcery.com>
254
255         * config/m68k/m68k.md (pushdi-1, pushdi, movsi+1): Don't use
256         the 'y' constraint.
257
258 2007-02-05  Richard Sandiford  <richard@codesourcery.com>
259
260         * dwarf2out.c (dwarf2out_frame_debug_expr): Record the register
261         saves in a PARALLEL before the register assignments.
262
263 2007-02-05  Richard Sandiford  <richard@codesourcery.com>
264
265         * doc/tm.texi (DWARF_ALT_FRAME_RETURN_COLUMN): Do not require
266         DWARF_FRAME_RETURN_COLUMN to be a general register.
267         * dwarf2out.c (init_return_column_size): New function, split from...
268         (expand_builtin_init_dwarf_reg_sizes): ...here.  Allow both
269         DWARF_FRAME_RETURN_COLUMN and DWARF_ALT_FRAME_RETURN_COLUMN
270         to be nongeneral registers.
271         * config/m68k/m68k.h (DWARF_FRAME_REGNUM): Only map FP and
272         integer registers.
273         (DWARF_FRAME_REGISTERS, DWARF_FRAME_RETURN_COLUMN): Define.
274         (DWARF_ALT_FRAME_RETURN_COLUMN): Define.
275
276 2007-02-04  Zdenek Dvorak <dvorakz@suse.cz>
277
278         * cfgcleanup.c (try_optimize_cfg): Avoid removing ENTRY_BLOCK_PTR.
279
280 2007-02-04  Zdenek Dvorak <dvorakz@suse.cz>
281
282         * cfgloopmanip.c (loop_delete_branch_edge): Removed.
283         (remove_path): Use can_remove_branch_p and remove_branch instead
284         of loop_delete_branch_edge.
285         * tree-ssa-loop-manip.c (scale_dominated_blocks_in_loop): New function.
286         (tree_transform_and_unroll_loop): Remove dead branches immediately.
287         Update profile using scale_dominated_blocks_in_loop.
288         * cfghooks.c (can_remove_branch_p, remove_branch): New functions.
289         * cfghooks.h (struct cfg_hooks): Add can_remove_branch_p.
290         (can_remove_branch_p, remove_branch): Declare.
291         * tree-cfg.c (tree_can_remove_branch_p): New function.
292         (tree_cfg_hooks): Add tree_can_remove_branch_p.
293         * cfgrtl.c (rtl_can_remove_branch_p): New function.
294         (rtl_cfg_hooks, cfg_layout_rtl_cfg_hook): Add rtl_can_remove_branch_p.
295
296 2007-02-05  Jan Hubicka  <jh@suse.cz>
297
298         PR middle-end/30696
299         * ipa-inline.c (cgraph_clone_inlined_nodes): When there are unanalyzed
300         nodes in cgraph, don't remove offline copy of the function.
301
302 2007-02-04  Jan Hubicka  <jh@suse.cz>
303
304         * tree-sra.c (sra_walk_expr): Add linebreaks. BITFIELD_REFs into
305         vectors might cause maybe_lookup_element_for_expr to be called
306         on non-sra-candidate.
307
308 2007-02-04  Kazu Hirata  <kazu@codesourcery.com>
309
310         * config/bfin/bfin-modes.def, config/bfin/bfin.c,
311         config/bfin/bfin.md, config/bfin/predicates.md: Follow
312         spelling conventions.
313
314 2007-02-04  Richard Guenther  <rguenther@suse.de>
315
316         PR middle-end/30636
317         * fold-const.c (try_move_mult_to_index): Make sure to not
318         overflow one dimension of a multi-dimensional array access.
319
320 2007-02-04  Jan Hubicka  <jh@suse.cz>
321
322         * passes.c (init_optimization_passes): Reindent.
323
324 2007-02-04  Jan Hubicka  <jh@suse.cz>
325             Eric Botcazou  <ebotcazou@adacore.com>
326
327         * tree-optimize.c (has_abnormal_outgoing_edge_p): Move to...
328         (execute_fixup_cfg): Break out the abnormal goto code.
329         * tree-inline.c (has_abnormal_outgoing_edge_p): ...here.
330         (make_nonlocal_label_edges): Move here from execute_fixup_cfg.
331         (optimize_inline_calls): Call make_nonlocal_label_edges.
332
333 2007-02-04  Jan Hubicka  <jh@suse.cz>
334
335         * tree-ssa-copyrename.c (copy_rename_partition_coalesce): Return
336         true when something was changed.
337         (rename_ssa_copies): When something was changed, do
338         TODO_remove_unused_locals.
339         * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
340         add TODO_remove_unused_locals when instruction was removed.
341
342 2007-02-04  Jan Hubicka  <jh@suse.cz>
343
344         * ipa-inline.c (try_inline): Improve debug output; work on already
345         inline edges too.
346         (cgraph_decide_inlining_incrementally): Indent; improve debug output;
347         call try_inline for already inlined edges too when flattening; 
348         inline also functions that make callee growth but overall unit size
349         reduce.
350
351 2007-02-04  Kazu Hirata  <kazu@codesourcery.com>
352
353         * config/m32c/bitops.md, config/m32c/jump.md,
354         config/m32c/m32c.c, config/m32c/m32c.h, config/m32r/m32r.c,
355         config/m32r/m32r.h, config/m32r/m32r.md,
356         config/m32r/predicates.md, config/m68hc11/larith.asm,
357         config/m68hc11/m68hc11.c, config/m68hc11/m68hc11.h,
358         config/m68k/m68k.h, config/mcore/mcore.md, config/mips/4k.md,
359         config/mips/mips-protos.h, config/mips/mips.c,
360         config/mips/mips.h, config/mips/mips.md, config/mips/mips16.S,
361         config/mn10300/mn10300.h, config/mn10300/predicates.md,
362         config/mt/mt.c, config/mt/mt.h, config/mt/mt.md: Follow
363         spelling conventions.
364
365         * config/v850/v850.c, config/v850/v850.h, config/v850/v850.md:
366         Follow spelling conventions.
367
368 2007-02-03  Douglas Gregor  <doug.gregor@gmail.com>
369
370        * c-opts.c (c_common_post_options): If C++0x mode is enabled, don't
371        warn about C++0x compatibility.
372         
373 2007-02-04  Kazu Hirata  <kazu@codesourcery.com>
374
375         * config/h8300/h8300.c, config/h8300/h8300.h,
376         config/h8300/h8300.md: Follow spelling conventions.
377
378 2007-02-03  Uros Bizjak  <ubizjak@gmail.com>
379
380         PR middle-end/30667
381         * combine.c (try_combine): Do not substitute source operand
382         with constants wider than 2 * HOST_BITS_PER_WIDE_INT.
383
384 2007-02-03  Jan Hubicka  <jh@suse.cz>
385
386         PR gcov-profile/30650
387         * value-prof.c (stringop_block_profile): Fix handling of size counter;
388         do not divide by zero for never executed counters.
389         (tree_find_values_to_profile): Fix counters.
390         * gcov-ui.h (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): Fix comments.
391
392 2007-02-03  Ian Lance Taylor  <iant@google.com>
393
394         * lower-subreg.c (simple_move_operand): New static function,
395         broken out of simple_move.  Reject LABEL_REF, SYMBOL_REF, and HIGH
396         operands.
397         (simple_move): Call simple_move_operand.
398         (find_decomposable_subregs): Add special handling of MEMs.
399         (can_decompose_p): Rename from cannot_decompose_p.  Reverse
400         meaning of return value.  If we see a hard register, test whether
401         it can store a word_mode value.  Change all callers.
402
403 2007-02-03  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
404
405         * pa.md (addvdi3, addvsi3, subvdi3, subvsi3, negvdi2, negvsi2): New
406         ftrapv insns and expanders.
407         (subdi3): Change define_expand operand 1 to arith11_operand, and
408         operand 2 to reg_or_0_operand.  Change constraints of 64-bit insn
409         pattern to handle reg_or_0 operands.  Revise 32-bit insn pattern to
410         handle 11-bit constants and reg_or_0 operands in operands 1 and 2,
411         respectively.
412
413         PR middle-end/30174
414         * varasm.c (notice_global_symbol): Treat global objects as weak when
415         flag_shlib is true.
416
417 2007-02-03  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
418
419         * emit-rtl.c (dconstpi): Delete.
420         (dconstsqrt2): New.
421         (init_emit_once): Delete dconstpi and init dconstsqrt2.
422         * real.h (dconstpi): Delete.
423         (dconstsqrt2): New.
424         * builtins.c (fold_builtin_cabs): Use dconstsqrt2.
425         (fold_builtin_hypot): Likewise.
426
427 2007-02-03  Tom Tromey  <tromey@redhat.com>
428
429         PR driver/30246
430         * gcc.c (cpp_unique_options): Any of -ggdb3, -gstabs3,
431         -gcoff3, -gxcoff3, -gvms3 implies -dD.
432
433 2007-02-03  Kazu Hirata  <kazu@codesourcery.com>
434
435         * c-decl.c, config/avr/avr.c, config/avr/avr.h,
436         config/m68k/m68k.c, config/m68k/netbsd-elf.h,
437         config/mn10300/mn10300.c, config/pdp11/pdp11.h,
438         config/rs6000/cell.md, config/rs6000/darwin.h,
439         config/sh/sh.md, config/sh/sh4-300.md, config/spu/spu.c,
440         config/spu/spu.md, cselib.c, expr.c, haifa-sched.c, hwint.h,
441         jump.c, reload.c, sched-deps.c, sched-int.h, tree-inline.c,
442         tree-profile.c, tree-ssa-live.h, tree-vrp.c: Fix comment
443         typos.  Follow spelling conventions.
444         * doc/invoke.texi: Follow spelling conventions.
445
446 2007-02-03  Roger Sayle  <roger@eyesopen.com>
447
448         * simplify-rtx.c (simplify_relational_operation_1): Implement some
449         canonicalization transformations that attempt to simplify integer
450         constant comparisons to become comparisons against zero.
451
452 2007-02-03  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
453
454         * builtins.c (fold_builtin_cabs): Fold cabs(x+xi) into
455         fabs(x)*sqrt(2).
456         * fold-const.c (fold_binary): Fix comment typos.  Fold complex
457         (x,0)-(0,y) into (x,-y).  Likewise (0,y)-(x,0) into (-x,y).
458
459 2007-02-02  Mike Stump  <mrs@apple.com>
460
461         * config/darwin9.h (DARWIN_LINKER_GENERATES_ISLANDS): Add.
462         * config/rs6000/rs6000.c (DARWIN_GENERATE_ISLANDS): Add.
463         (output_call): Use DARWIN_GENERATE_ISLANDS to decide when to
464         generate a branch island.
465
466 2007-02-02  Bob Wilson  <bob.wilson@acm.org>
467
468         * config/xtensa/xtensa.c (smalloffset_mem_p): Use BASE_REG_P.
469         (xtensa_legitimate_address_p): New.
470         (xtensa_legitimize_address): New.
471         (xtensa_output_addr_const_extra): New.
472         * config/xtensa/xtensa.h (REG_OK_STRICT_FLAG): Define.
473         (BASE_REG_P): New.
474         (REG_OK_FOR_BASE_P): Use BASE_REG_P.
475         (GO_IF_LEGITIMATE_ADDRESS): Move code to xtensa_legitimate_address_p.
476         (LEGITIMIZE_ADDRESS): Move code to xtensa_legitimize_address.
477         (OUTPUT_ADDR_CONST_EXTRA): Move code to xtensa_output_addr_const_extra.
478         * config/xtensa/xtensa-protos.h (xtensa_legitimate_address_p): New.
479         (xtensa_legitimize_address): New.
480         (xtensa_output_addr_const_extra): New.
481
482 2007-02-02  Steve Ellcey  <sje@cup.hp.com>
483
484         * config/ia64/ia64.c (ia64_print_operand): Fix compare strings.
485
486 2007-02-02  Ian Lance Taylor  <iant@google.com>
487
488         * expmed.c (expand_divmod): Add comment.
489
490 2007-02-02  Kazu Hirata  <kazu@codesourcery.com>
491
492         * emit-rtl.c (renumber_insns): Remove.
493         * flags.h: Remove the extern for flag_renumber_insns.
494         * rtl.h: Remove the prototype for renumber_insns.
495         * toplev.c (flag_renumber_insns): Remove.
496
497 2007-02-02  Hui-May Chang  <hm.chang@apple.com>
498
499         Revert for x86 darwin:
500         2005-06-19  Uros Bizjak  <uros@kss-loka.si>
501         
502         * config/i386/i386.c (ix86_function_arg_regno_p): Put back the
503         code before the following patch under TARGET_MACHO.
504         (ix86_function_value_regno_p): Likewise.
505
506 2007-02-02  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
507
508         * fold-const.c (negate_expr_p, fold_negate_expr): Handle
509         COMPLEX_EXPR.
510
511 2007-02-02  Maxim Kuvyrkov  <mkuvyrkov@ispras.ru>
512
513         PR target/29682
514         * config/ia64/ia64.c (ia64_speculate_insn): Restrict to memory loads to
515         general or fp registers.  Add comments.
516         * config/ia64/ia64.md (reg_pred_prefix): Add comment.
517
518 2007-02-02  Paolo Bonzini  <bonzini@gnu.org>
519
520         * pointer-set.c (insert_aux): Only return insertion slot.
521         (pointer_set_insert): Adjust.
522         (pointer_set_traverse, struct pointer_map_t, pointer_map_create,
523         pointer_map_destroy, pointer_map_insert, pointer_map_contains,
524         pointer_map_traverse): New.
525         * pointer-set.h (pointer_set_traverse, struct pointer_map_t,
526         pointer_map_create,  pointer_map_destroy, pointer_map_insert,
527         pointer_map_contains, pointer_map_traverse): Declare.
528
529 2007-02-02  Jakub Jelinek  <jakub@redhat.com>
530
531         PR middle-end/30473
532         * builtins.c (fold_builtin_sprintf): Do not attempt to optimize
533         sprintf (str, "%s").  Do not optimize sprintf (str, "nopercent", p++).
534
535 2007-02-02  Maxim Kuvyrkov  <mkuvyrkov@ispras.ru>
536
537         * sched-int.h (ds_to_dk, dk_to_ds): Declare functions.
538         
539         (struct _dep): New type.
540         (dep_t): New typedef.
541         (DEP_PRO, DEP_CON, DEP_KIND): New access macros.
542         (DEP_STATUS): New access macro.  The macro with the same name was
543         renamed to DEP_LINK_STATUS.
544         (dep_init): Declare function
545
546         (struct _dep_link): New type.
547         (dep_link_t): New typedef.
548         (DEP_LINK_NODE, DEP_LINK_NEXT, DEP_LINK_PREV_NEXTP): New access macros.
549         (DEP_LINK_DEP, DEP_LINK_PRO, DEP_LINK_CON, DEP_LINK_KIND): New macros.
550         (DEP_LINK_STATUS): New macro.
551         (debug_dep_links): New debug function.
552
553         (struct _deps_list): New type.
554         (deps_list_t): New typedef.
555         (DEPS_LIST_FIRST): New access macro.
556         (FOR_EACH_DEP_LINK): New cycle macro.
557         (create_deps_list, free_deps_list, delete_deps_list): Declare
558         functions.
559         (deps_list_empty_p, debug_deps_list, add_back_dep_to_deps_list): Ditto.
560         (find_link_by_pro_in_deps_list, find_link_by_con_in_deps_list): Ditto.
561         (copy_deps_list_change_con): Ditto.
562
563         (move_dep_link): Declare function.
564
565         (struct _dep_node): New type.
566         (dep_node_t): New typedef.
567         (DEP_NODE_BACK, DEP_NODE_DEP, DEP_NODE_FORW): New access macros.
568
569         (struct haifa_insn_data.back_deps): New field to hold backward
570         dependencies of the insn.
571         (struct haifa_insn_data.depend): Rename to forw_deps.  Change its type
572         to deps_list_t.
573         (struct haifa_insn_data.resolved_deps): Rename to resolved_back_deps.
574         Change its type to deps_list_t.
575         (INSN_BACK_DEPS): New access macro to use instead of LOG_LINKS.
576         (INSN_DEPEND): Rename to INSN_FORW_DEPS.
577         (RESOLVED_DEPS): Rename to INSN_RESOLVED_BACK_DEPS.
578
579         (INSN_COST): Move to haifa-sched.c.  Use insn_cost () instead.
580         
581         (DEP_STATUS): Rename to DEP_LINK_STATUS.  Fix typo in the comment.
582
583         (add_forw_dep, delete_back_forw_dep, insn_cost): Update declaration and
584         all callers.
585         (dep_cost): Declare.
586         
587         * sched-deps.c (CHECK): New macro to (en/dis)able sanity checks.
588         (ds_to_dk, dk_to_ds): New functions.
589         
590         (init_dep_1): New static function.
591         (init_dep): New function.
592         (copy_dep): New static function.
593         
594         (dep_link_consistent_p, attach_dep_link, add_to_deps_list): New static
595         functions.
596         (detach_dep_link): New static function.
597         (move_dep_link): New function.
598         
599         (dep_links_consistent_p, dump_dep_links): New static functions.
600         (debug_dep_links): New debugging function.
601         
602         (deps_obstack, dl_obstack, dn_obstack): New static variables.
603         
604         (alloc_deps_list, init_deps_list): New static functions.
605         (create_deps_list): New function.
606         (clear_deps_list): New static function.
607         (free_deps_list, delete_deps_list, deps_list_empty_p): New functions.
608         (deps_list_consistent_p, dump_deps_list): New static functions.
609         (debug_deps_list): New function.
610         (add_back_dep_to_deps_list, find_link_by_pro_in_deps_list): New
611         functions.
612         (find_link_by_con_in_deps_list, copy_deps_list_change_con): Ditto.
613
614         (maybe_add_or_update_back_dep_1, add_or_update_back_dep_1): Update to
615         use new scheduler dependencies lists.
616         (add_back_dep, delete_all_dependences, fixup_sched_groups): Ditto.
617         (sched_analyze): Ditto.  Initialize dependencies lists.
618         (add_forw_dep, compute_forward_dependences): Update to use new
619         scheduler dependencies lists.
620         
621         (init_dependency_caches): Init deps_obstack.
622         (free_dependency_caches): Free deps_obstack.
623         
624         (adjust_add_sorted_back_dep, adjust_back_add_forw_dep): Update to use
625         new scheduler dependencies lists.
626         (delete_forw_dep, add_or_update_back_forw_dep): Ditto.
627         (add_back_forw_dep, delete_back_forw_dep): Ditto.
628
629         * sched-rgn.c (set_spec_fed, find_conditional_protection, is_pfree):
630         Update to use new scheduler dependencies lists.
631         (is_conditionally_protected, is_prisky, add_branch_dependences): Ditto.
632         (debug_dependencies): Ditto.
633         (schedule_region): Update comments.
634         
635         * sched-ebb.c (earliest_block_with_similiar_load): Update to use new
636         scheduler dependencies lists.
637         (schedule_ebb): Update comments.
638         
639         * rtl.def (DEPS_LIST): Remove.
640         
641         * lists.c (unused_deps_list): Remove.
642         (free_list): Update assertions.
643         
644         (alloc_DEPS_LIST, free_DEPS_LIST_list, free_DEPS_LIST_node): Remove.
645         (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
646
647         * rtl.h (free_DEPS_LIST_list, alloc_DEPS_LIST): Remove declarations.
648         (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
649         
650         * haifa-sched.c (comments): Update.
651         (insn_cost1): Remove.  Inline the code into insn_cost ().
652         (insn_cost): Update to use new scheduler dependencies lists.  Move
653         processing of the dependency cost to dep_cost ().
654         (dep_cost): New function.  Use it instead of insn_cost () when
655         evaluating cost of the dependency.  Use compatible interface to
656         interact with the target.
657         (priority): Update to use new scheduler dependencies lists.
658         (rank_for_schedule): Ditto.  Optimize heuristic that prefers the insn
659         with greater number of insns that depend on the insn.
660         (schedule_insn): Update to use new scheduler dependencies lists.  Add
661         code to free backward dependencies lists.  Inline and optimize code
662         from resolve_dep () - see PR28071.
663         (ok_for_early_queue_removal): Update to use new scheduler dependencies
664         lists.  Update call to targetm.sched.is_costly_dependence hook.
665         
666         (fix_inter_tick, try_ready, fix_tick_ready): Update to use new
667         scheduler dependencies lists.
668         
669         (resolve_dep): Remove.  Move the logic to schedule_insn ().
670         (init_h_i_d): Initialize dependencies lists.
671         
672         (process_insn_depend_be_in_spec): Rename to
673         process_insn_forw_deps_be_in_spec.  Update to use new scheduler
674         dependencies lists.
675         (add_to_speculative_block, create_check_block_twin, fix_recovery_deps):
676         Update to use new scheduler dependencies lists.
677         (clear_priorities, calc_priorities, add_jump_dependencies): Ditto.
678         
679         * ddg.c (create_ddg_dependence, create_ddg_dep_no_link): Update to use
680         new scheduler dependencies lists.
681         (build_intra_loop_deps): Ditto.
682         
683         * target.h (struct _dep): Declare to use in
684         gcc_target.sched.is_costly_dependence.
685         (struct gcc_target.sched.adjust_cost): Fix typo.
686         (struct gcc_target.sched.is_costly_dependence): Change signature to use
687         single dep_t parameter instead of an equivalent triad.
688         (struct gcc_target.sched.adjust_cost_2): Remove.
689
690         * target-def.h (TARGET_SCHED_ADJUST_COST_2): Remove.
691
692         * reg-notes.def (DEP_TRUE, DEP_OUTPUT, DEP_ANTI): Update comments.
693
694         * doc/tm.texi (TARGET_SCHED_IS_COSTLY_DEPENDENCE): Update
695         documentation.
696         (TARGET_SCHED_ADJUST_COST_2): Remove documentation.
697
698         * doc/rtl.texi (LOG_LINKS): Remove part about instruction scheduler.
699         (REG_DEP_TRUE): Document.
700         
701         * config/ia64/ia64.c (ia64_adjust_cost_2): Rename to ia64_adjust_cost.
702         Change signature to correspond to the targetm.sched.adjust_cost hook.  
703         Update use in TARGET_SCHED_ADJUST_COST_2.
704         (TARGET_SCHED_ADJUST_COST_2): Rename to TARGET_SCHED_ADJUST_COST.
705         (ia64_dependencies_evaluation_hook, ia64_dfa_new_cycle): Update to use
706         new scheduler dependencies lists.
707         (ia64_gen_check): Ditto.
708
709         * config/mips/mips.c (vr4130_swap_insns_p): Update to use new scheduler
710         dependencies lists.
711         
712         * config/rs6000/rs6000.c (rs6000_is_costly_dependence): Change
713         signature to correspond to the targetm.sched.is_costly_dependence hook.
714         (is_costly_group): Update to use new scheduler dependencies lists.
715
716         * config/spu/spu.c (spu_sched_adjust_cost): Use insn_cost () function
717         instead of INSN_COST () macro.
718
719 2007-02-01  Ian Lance Taylor  <iant@google.com>
720
721         * lower-subreg.c (resolve_clobber): Handle a subreg of a concatn.
722
723 2007-02-01  Guy Martin  <gmsoft@gentoo.org>
724
725         * pa.md (tp_load): Correct mfctl instruction syntax.
726
727 2007-02-01  Geoffrey Keating  <geoffk@apple.com>
728
729         * config/rs6000/rs6000.c (rs6000_stack_info): Correct
730         altivec_padding_size calculation on AIX.  Improve comment, add
731         assert to verify that it's right.
732
733         * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
734         64-bit.
735
736 2007-2-01   Seongbae Park <seongbae.park@gmail.com>
737
738         PR inline-asm/28686
739         * global.c (global_alloc): Add mising initialization of
740         ELIMINABLE_REGSET.
741
742 2007-02-01  Roger Sayle  <roger@eyesopen.com>
743
744         * alias.c (init_alias_analysis): Correct whitespace.
745         * bb-reorder.c (fix_edges_for_rarely_executed_code,
746         partition_hot_cold_basic_blocks): Likewise.
747         * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
748         expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
749         maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
750         fold_builtin_snprintf_chk, fold_builtin_printf,
751         fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
752         do_mpfr_arg3, do_mpfr_sincos): Likewise.
753         * cfgcleanup.c (cleanup_cfg): Likewise.
754         * cfgexpand.c (tree_expand_cfg): Likewise.
755         * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
756         * function.c (get_next_funcdef_no): Likewise.
757         * gengtype.c (main): Likewise.
758         * genmodes.c (main): Likewise.
759         * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
760         * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
761         * ifcvt.c (noce_emit_move_insn): Likewise.
762         * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
763         Likewise.
764         * stor-layout.c (get_best_mode): Likewise.
765         * tree-ssa-loop-niter.c (get_val_for): Likewise.
766         * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
767         scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
768         perform_var_substitution, solve_graph): Likewise.
769         * tree-vrp.c (vrp_finalize): Likewise.
770
771 2007-02-01  Ian Lance Taylor  <iant@google.com>
772
773         * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
774         high part of a paradoxical subreg, return a constant zero.
775
776 2007-02-01  Zdenek Dvorak <dvorakz@suse.cz>
777
778         * toplev.c (lang_dependent_init): Call init_set_costs.
779         * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
780
781 2007-02-01  Richard Guenther  <rguenther@suse.de>
782
783         PR middle-end/30656
784         * fold-const.c (fold_negate_expr): Allow negating a
785         constant if overflow does not change.
786
787 2007-02-01  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
788
789         * doc/c-tree.texi (Expression trees): Improve markup.
790         * doc/tm.texi (Register Classes, Addressing Modes)
791         (Floating Point): Fix spacing after abbreviations.  Fix some
792         typos.
793
794 2007-02-01  Ben Elliston  <bje@au.ibm.com>
795
796         * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
797         * doc/contrib.texi: Likewise.
798         * doc/install.texi: Likewise.
799
800 2007-01-31  Richard Henderson  <rth@redhat.com>
801             Ian Lance Taylor  <iant@google.com>
802
803         * lower-subreg.c: New file.
804         * rtl.def (CONCATN): Define.
805         * passes.c (init_optimization_passes): Add pass_lower_subreg and
806         pass_lower_subreg2.
807         * emit-rtl.c (update_reg_offset): New static function, broken out
808         of gen_rtx_REG_offset.
809         (gen_rtx_REG_offset): Call update_reg_offset.
810         (gen_reg_rtx_offset): New function.
811         * regclass.c: Revert patch of 2006-03-05, restoring
812         reg_scan_update.
813         (clear_reg_info_regno): New function.
814         * dwarf2out.c (concatn_loc_descriptor): New static function.
815         (loc_descriptor): Handle CONCATN.
816         * common.opt (fsplit_wide_types): New option.
817         * opts.c (decode_options): Set flag_split_wide_types when
818         optimizing.
819         * timevar.def (TV_LOWER_SUBREG): Define.
820         * rtl.h (gen_reg_rtx_offset): Declare.
821         (reg_scan_update): Declare.
822         * regs.h (clear_reg_info_regno): Declare.
823         * tree-pass.h (pass_lower_subreg): Declare.
824         (pass_lower_subreg2): Declare.
825         * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
826         (Optimize Options): Add -fsplit-wide-types to -O1 list.  Document
827         -fsplit-wide-types.
828         * doc/rtl.texi (Regs and Memory): Document concat and concatn.
829         * Makefile.in (OBJS-common): Add lower-subreg.o.
830         (lower-subreg.o): New target.
831
832 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
833
834         * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
835
836 2007-01-31  Anatoly Sokolov <aesok@post.ru>
837
838         PR target/19087
839         * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
840
841 2007-01-31  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
842
843         PR middle-end/29335
844         * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
845
846 2007-01-31  Zdenek Dvorak <dvorakz@suse.cz>
847
848         * cfgloop.h: Include vec-prim.h.
849         (enum li_flags): Remove LI_ONLY_OLD.
850         (loop_iterator): Changed.
851         (fel_next, fel_init): Iterate over loop tree.
852         (FOR_EACH_LOOP_BREAK): New macro.
853         * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
854         FOR_EACH_LOOP.
855         * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
856         * modulo-sched.c (sms_schedule): Ditto.
857         * tree-vectorizer.c (vectorize_loops): Ditto.
858         * doc/loop.texi: Update information on loop numbering and behavior of
859         FOR_EACH_LOOP wrto new loops.
860         * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
861         add_to_evolution_1): Test nestedness of loops instead of comparing
862         their numbers.
863         * tree-chrec.c (chrec_fold_plus_poly_poly,
864         chrec_fold_multiply_poly_poly, chrec_evaluate,
865         hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
866         reset_evolution_in_loop): Ditto.
867         * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
868
869 2007-01-31  Dirk Mueller  <dmueller@suse.de>
870
871         * c-common.c (warn_about_parentheses): Separate warning about
872         un-parenthized sequence of comparison operators from the one
873         which is supposed to warn about x <= y <= z.
874
875 2007-01-31  Uros Bizjak  <ubizjak@gmail.com>
876
877         * optabs.h (enum optab_index): Add new OTI_isinf.
878         (isinf_optab): Define corresponding macro.
879         * optabs.c (init_optabs): Initialize isinf_optab.
880         * genopinit.c (optabs): Implement isinf_optab using isinf?f2
881         patterns.
882         * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
883         (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
884         using isinf_optab.
885         (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
886         expand_builtin_interclass_mathfn.
887         * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
888         * config/i386/i386.md (UNSPEC_FXAM): New constant.
889         (fxam<mode>2_i387): New insn pattern.
890         (isinf<mode>2) New expander to implement isinf, isinff and isinfl
891         built-in functions as x87 inline asm.
892
893 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
894
895         * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
896
897 2007-01-30  Eric Christopher  <echristo@apple.com>
898
899         * config.gcc: Add geode.
900
901 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
902
903         * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
904         ipa-inline.c, tree-profile.c, tree-ssa-live.c,
905         tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
906         value-prof.c: Fix comment typos.
907
908 2007-01-30  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
909
910         PR c++/24745
911         * doc/invoke.texi (Wpointer-arith): Document warning.
912
913 2007-01-30  Janis Johnson
914
915         * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
916         the list of discrepancies from the draft TR.
917
918 2007-01-30  Dirk Mueller  <dmueller@suse.de>
919
920         PR c++/30601
921         * doc/invoke.texi (-Wreturn-type): Update description to
922         match new behavior.
923
924 2007-01-30  Richard Sandiford  <richard@codesourcery.com>
925
926         * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
927
928 2007-01-30  Uros Bizjak  <ubizjak@gmail.com>
929
930         * builtins.c (expand_builtin_int_interclass_roundingfn): New function
931         to handle optabs that operate on floating point input argument and
932         output to integer output.
933         (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
934         (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
935         (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
936         expand_builtin_interclass_mathfn ().
937         * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
938         (ilogbsi2): Remove.
939         (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
940         ilogbl built-in functions as x87 intrinsics.
941
942 2007-01-30  Richard Guenther  <rguenther@suse.de>
943
944         PR middle-end/27657
945         * dwarf2out.c (reference_to_unused): Query varpool if the
946         variable was output.
947
948 2007-01-30  Richard Guenther  <rguenther@suse.de>
949
950         PR middle-end/30313
951         * passes.c (execute_one_pass): Reset in_gimple_form to not
952         confuse non-unit-at-a-time mode.
953
954 2007-01-29  Roger Sayle  <roger@eyesopen.com>
955             Richard Guenther  <rguenther@suse.de>
956
957         * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
958         avoid undefined behaviour on overflow.  Use force_fit_type_double
959         to construct the constant with the specified TREE_OVERFLOW.
960
961 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
962
963         * config/dfp-bit.c: Add parameterized support for fp exceptions.
964         * config/dfp-bit.h: Ditto.
965
966 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
967
968         * c-decl.c (pop_scope): Replace warnings with call to
969         warn_for_unused_label.
970         * c-common.h (warn_for_unused_label): Declare.
971         * c-common.c (warn_for_unused_label): Define.
972
973 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
974
975         * tree-optimize.c (update_inlined_to_pointers): Delete unused
976         function.
977
978 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
979
980         * Makefile.in (USER_H): Remove decfloat.h.
981         * ginclude/decfloat.h: Delete, moving contents to ...
982         * ginclude/float.h: Add support for decimal floating point,
983         guarded by __STDC_WANT_DEC_FP__.
984
985 2007-01-29  Mike Stump  <mrs@apple.com>
986
987         * doc/gccint.texi (Top): Rename Loop Representation to Loop
988         Analysis and Representation to resolve case insensitive conflict.
989         * doc/loop.texi (Loop Analysis and Representation): Likewise.
990
991 2007-01-28  Daniel Berlin  <dberlin@dberlin.org>
992
993         * tree.h (struct tree_memory_tag): Add aliases member.
994         (MTAG_ALIASES): New macro.
995         * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
996         (add_may_alias): Remove pointer-set. Update for may_aliases being
997         a bitmap.
998         (mark_aliases_call_clobbered): Update for may_aliases being a
999         bitmap.
1000         (compute_tag_properties): Ditto.
1001         (create_partition_for): Ditto.
1002         (compute_memory_partitions): Ditto.
1003         (dump_may_aliases_for): Ditto.
1004         (is_aliased_with): Ditto.
1005         (add_may_alias_for_new_tag): Ditto.
1006         (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
1007         (compute_is_aliased): New function.
1008         (compute_may_aliases): Call compute_is_aliased).
1009         (init_alias_info): Initialize alias_bitmap_obstack.
1010         (union_alias_set_into): New function.
1011         (compute_flow_sensitive_aliasing): Use union_aliases_into.
1012         (have_common_aliases_p): Rewrite to take two bitmaps and use
1013         intersection.
1014         (compute_flow_insensitive_aliasing): Stop using pointer-sets.
1015         Update for bitmaps.
1016         (finalize_ref_all_pointers): Update for add_may_alias changes.
1017         (new_type_alias): Ditto.
1018         * tree-flow-inline.h (may_aliases): Return a bitmap.
1019         * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
1020         * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
1021         may_aliases being a bitmap.
1022         * tree-flow.h (struct var_ann_d): Remove may_aliases member.
1023         may_aliases now returns a bitmap.
1024         * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
1025         being a bitmap.
1026         * tree-ssa-operands.c (add_virtual_operand): Update for
1027         may_aliases being a bitmap.
1028
1029 2007-01-29  Daniel Berlin  <dberlin@dberlin.org>
1030
1031         PR tree-optimization/30630
1032         * tree-ssa-structalias.c (do_complex_constraint): Mark correct
1033         variable as changed.
1034
1035 2007-01-29  Simon Martin  <simartin@users.sourceforge.net>
1036
1037         PR c++/28266
1038         * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
1039         expanded only once even if an error occurs.
1040
1041 2007-01-29  Ben Elliston  <bje@au.ibm.com>
1042
1043         * gcov-io.h (__gcov_indirect_call_profiler): Declare.
1044         (__gcov_average_profiler): Likewise.
1045         (__gcov_ior_profiler): Likewise.
1046         (__gcov_merge_ior): Likewise.
1047
1048 2007-01-28  Jan Hubicka  <jh@suse.cz>
1049
1050         * builtins.c (expand_builtin_memset): Fix typo in my last patch.
1051         * value-prof.c (stringop_block_profile): Likewise.
1052
1053 2007-01-28  Jan Hubicka  <jh@suse.cz>
1054
1055         * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
1056         variant handling histograms; add wrapper.
1057         (clear_storage_via_libcall): Export.
1058         (emit_block_move_hints): Break out from ...; add histograms.
1059         (emit_block_move): ... this one.
1060         (clear_storage_hints): Break out from ...; add histograms.
1061         (clear_storage): ... this one.
1062         (set_storage_via_memset): Handle histogram.
1063         * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
1064         (clear_storage_hints, clear_storage_via_libcall): Declare.
1065         (set_storage_via_setmem): Update prototype.
1066         * doc/md.texi (movmem, setmem): Document new arguments.
1067
1068         * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
1069         new histograms.
1070         (stringop_block_profile): New global function.
1071         (tree_stringops_values_to_profile): Profile block size and alignment.
1072         * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
1073         HIST_TYPE_IOR.
1074         (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
1075         (stringop_block_profile): Declare.
1076         * builtins.c: Include value-prof.h.
1077         (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
1078         * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
1079         (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
1080         (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
1081         * profile.c (instrument_values): Add new counters.
1082         * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
1083         calls.
1084         * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
1085         (tree_init_edge_profiler): Build new profilers.
1086         (tree_gen_average_profiler, tree_gen_ior_profiler): New.
1087         (pass_tree_profile): Add dump.
1088         (tree_profile_hooks): Update.
1089         * Makefile.in (LIBGCOV): Add new constants.
1090         * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
1091         __gcov_ior_profiler): New.
1092         * i386.md (movmem/setmem expanders): Add new optional arguments.
1093
1094 2007-01-28  David Edelsohn  <edelsohn@gnu.org>
1095
1096         * doc/md.texi (Standard Pattern Names): Document blockage pattern.
1097
1098 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
1099
1100         * tree-ssa-loop-unswitch.c: Include tree-inline.h.
1101         (tree_unswitch_single_loop): Pass eni_size_weights to
1102         tree_num_loop_insns.
1103         * tree-ssa-loop-manip.c: Include tree-inline.h.
1104         (can_unroll_loop_p): Pass eni_size_weights to
1105         tree_num_loop_insns.
1106         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
1107         Pass eni_size_weights to estimate_num_insns.
1108         * tree.h (init_inline_once): Export.
1109         * toplev.c (backend_init): Call init_inline_once.
1110         * cgraphunit.c (cgraph_process_new_functions,
1111         cgraph_analyze_function): Pass eni_inlining_weights to
1112         estimate_num_insns.
1113         * ipa-inline.c (compute_inline_parameters): Ditto.
1114         * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
1115         to estimate_num_insns.
1116         (try_unroll_loop_completely): Pass eni_size_weights to
1117         tree_num_loop_insns.
1118         * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
1119         ot estimate_num_insns.
1120         * tree-ssa-loop-prefetch.c: Include tree-inline.h.
1121         (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
1122         * tree-inline.c (eni_inlining_weights, eni_size_weights,
1123         eni_time_weights): New variables.
1124         (init_inline_once): Initialize them.
1125         (struct eni_data): Mew.
1126         (estimate_num_insns_1, estimate_num_insns): Use weights.
1127         * tree-inline.h (struct eni_weights_d): New.
1128         (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
1129         (estimate_num_insns): Declaration changed.
1130         * cfgloop.h (tree_num_loop_insns): Declaration changed.
1131         * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
1132         tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
1133
1134 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
1135
1136         * tree-data-ref.c (conflict_fn): Assert that the number of affine
1137         relations in the conflict function is valid.
1138
1139 2007-01-27  Ian Lance Taylor  <iant@google.com>
1140
1141         * common.opt: Add fstrict-overflow.
1142         * opts.c (decode_options): Set flag_strict_overflow if -O2.
1143         * flags.h (TYPE_OVERFLOW_WRAPS): Define.
1144         (TYPE_OVERFLOW_UNDEFINED): Define.
1145         (TYPE_OVERFLOW_TRAPS): Define.  This replaces TYPE_TRAP_SIGNED.
1146         Replace all uses.
1147         * tree.h (TYPE_TRAP_SIGNED): Don't define.
1148         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
1149         (fold_negate_expr): Likewise.
1150         (make_range): Likewise.
1151         (extract_muldiv_1): Likewise.
1152         (maybe_canonicalize_comparison): Likewise.
1153         (fold_comparison): Likewise.
1154         (fold_binary): Likewise.
1155         (tree_expr_nonnegative_p): Likewise.
1156         (tree_expr_nonzero_p): Likewise.
1157         * tree-vrp.c (compare_values): Likewise.
1158         (extract_range_from_binary_expr): Likewise.
1159         (extract_range_from_unary_expr): Likewise.
1160         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
1161         Likewise.
1162         (nowrap_type_p): Likewise.
1163         * tree-scalar-evolution.c (simple_iv): Likewise.
1164         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
1165         (build_range_check): Likewise.
1166         (extract_muldiv_1): Likewise.
1167         (fold_comparison): Likewise.
1168         * tree-vrp.c (vrp_int_const_binop): Likewise.
1169         (extract_range_from_unary_expr): Likewise.
1170         * convert.c (convert_to_integer): Likewise.
1171         * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
1172         (fold_comparison): Likewise.
1173         (fold_binary): Likewise.
1174         * optabs.c (optab_for_tree_code): Likewise.
1175         * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
1176         * simplify-rtx.c (simplify_const_relational_operation): Check
1177         flag_strict_overflow and flag_trapv.
1178         (simplify_const_relational_operation): Likewise.
1179         * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
1180         (Optimize Options): Add -fstrict-overflow to -O2 list.  Document
1181         -fstrict-overflow.
1182
1183 2007-01-27  Roger Sayle  <roger@eyesopen.com>
1184
1185         * tree.c (tree_fold_gcd): Delete.
1186         * tree.h (tree_fold_gcd): Remove prototype.
1187         * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
1188         test whether one constant integer is a multiple of another.  Instead
1189         call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
1190         * fold-const.c (multiple_of_p):  We've determined both TOP and
1191         BOTTOM are integer constants so we can call int_const_binop directly
1192         instead of the more generic const_binop.
1193
1194 2007-01-27  Roger Sayle  <roger@eyesopen.com>
1195
1196         * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
1197         1*X check that the constant hasn't overflowed, to preserve the
1198         TREE_OVERFLOW bit.
1199         (round_up): Provide an efficient implementation when rouding-up an
1200         INTEGER_CST to a power-of-two.
1201
1202 2007-01-28  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
1203
1204         * doc/sourcebuild.texi: Add comma for clarity.
1205         * doc/extend.texi: Fix some typos.
1206         * doc/passes.texi: Likewise.
1207         * doc/cppinternals.texi: Likewise.
1208         * doc/c-tree.texi: Likewise.
1209         * doc/tree-ssa.texi: Likewise.
1210         * doc/install.texi: Likewise.
1211
1212 2007-01-27  Jan Hubicka  <jh@suse.cz>
1213
1214         * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
1215         (pass_sra): Do not require alias information.
1216         * passes.c (init_optimization_passes): Add SRA
1217
1218 2007-01-27  Steven Bosscher  <steven@gcc.gnu.org>
1219
1220         * tracer.c (rest_of_handle_tracer): We already cleaned
1221         up the CFG in tracer() so don't do it here again.
1222         * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
1223         cleanup_cfg here, either.  And don't call renumber_insns.
1224
1225         * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
1226         and BB_HEAD are in the insn stream here.  Instead make sure
1227         that BB_INSN is valid on all insns.  Also, do check here that
1228         there are no pending branch predictions...
1229         (rtl_verify_flow_info): ...instead of doing it here.  Checks
1230         for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
1231         here.
1232
1233 2007-01-26  Roger Sayle  <roger@eyesopen.com>
1234
1235         * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
1236         function to simplify/factorize operand order canonicalization.
1237         (ix86_fixup_binary_operands): Reorganize using the above function.
1238         (ix86_binary_operator_ok): Likewise.
1239
1240 2007-01-27  Jakub Jelinek  <jakub@redhat.com>
1241
1242         * genattrtab.c (struct attr_value_list, insn_code_values): Move to
1243         file scope from optimize_attrs.
1244         (simplify_test_exp): If insn_code_values is not NULL, use it to speed
1245         up search.
1246         (optimize_attrs): Clear insn_code_values after freeing it.
1247
1248 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
1249
1250         * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
1251         a parts.base assignment.
1252
1253 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
1254
1255         * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
1256         (compute_subscript_distance, initialize_data_dependence_relation,
1257         finalize_ddr_dependent, analyze_ziv_subscript,
1258         analyze_siv_subscript_cst_affine,
1259         compute_overlap_steps_for_affine_univar,
1260         compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
1261         analyze_siv_subscript, analyze_miv_subscript,
1262         analyze_overlapping_iterations, subscript_dependence_tester_1,
1263         compute_self_dependence, free_dependence_relation): Work
1264         with affine_fn instead of chrecs.
1265         (dump_affine_function, dump_conflict_function, affine_function_equal_p,
1266         common_affine_function, affine_function_base,
1267         affine_function_constant_p, affine_fn_op, affine_fn_plus,
1268         affine_fn_minus, affine_fn_free, conflict_fn_not_known,
1269         conflict_fn_no_dependence, free_conflict_function, free_subscripts,
1270         conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
1271         (all_chrecs_equal_p): Removed.
1272         * tree-data-ref.h (affine_fn, conflict_function): New types.
1273         (struct subscript): Change type of conflicting_iterations_in_a
1274         and conflicting_iterations_in_b.
1275
1276 2007-01-26  Steve Ellcey  <sje@cup.hp.com>
1277
1278         PR other/30182
1279         * config/pa/pa.h (TARGET_HPUX_11): New.
1280         * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
1281         * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
1282
1283 2007-01-26  Daniel Berlin  <dberlin@dberlin.org>
1284         Richard Guenther  <rguenther@suse.de>
1285
1286         * tree-ssa-structalias.c (solve_graph): Handle case
1287         we merged the variable to another.
1288
1289 2007-01-25  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
1290
1291         * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
1292         functions.
1293
1294         * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
1295         catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
1296
1297 2007-01-25  DJ Delorie  <dj@redhat.com>
1298
1299         * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
1300         allow changes to modes which don't fit in those registers.
1301
1302         * reload1.c (choose_reload_regs): Check for invalid subregs before
1303         computing their locations, not after.
1304
1305 2007-01-25  Geoffrey Keating  <geoffk@apple.com>
1306
1307         PR 25127
1308         * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
1309         save Altivec registers in an eh_return function.
1310         (compute_vrsave_mask): Likewise.
1311         (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
1312         for saving Altivec registers.
1313         (rs6000_emit_prologue): Don't allocate stack twice in
1314         eh_return function.  Correct expected value of altivec_save_offset
1315         when using save_world.  Describe save of R0 to stack when using
1316         save_world.  Describe stack pointer adjustment when using
1317         save_world.  Remove duplicated eh_return parameter register saving.
1318         Update sp_offset variable after save_world.
1319         * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
1320         darwin-world.asm.
1321         (LIB2FUNCS_EXTRA): Add darwin-world.asm.
1322         * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
1323         implies Altivec.
1324
1325 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
1326
1327         * config.gcc (ia64*-*-hpux*): Make posix threads the default.
1328
1329 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
1330
1331         PR other/30182
1332         * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
1333         * config/ia64/ia64.c (ia64_init_builtins):  Ditto.
1334
1335 2007-01-25  Richard Guenther  <rguenther@suse.de>
1336
1337         * doc/invoke.texi (-Wcoverage-mismatch): Document.
1338         * common.opt (-Wcoverage-mismatch): New warning option.
1339         * coverage.c (get_coverage_counts): Ignore coverage mismatch
1340         if -Wcoverage-mismatch is given.
1341
1342 2007-01-25  Razya Ladelsky  <razya@il.ibm.com>
1343
1344         * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
1345         (ipcp_driver): Change to static definition.
1346         Add dumping of the ifunctions.
1347         (constant_val_insert): Remove unused parameter. Support for SSA.
1348         (ipcp_propagate_const): Support for SSA.
1349         (ipcp_profile_bb_print): Print only analyzed nodes.
1350         (ipcp_replace_map_create): Remove support for Fortran constant
1351         for now.
1352         * ipa-prop.c (ipa_method_modify_stmt,
1353         ipa_callsite_compute_param): Support for SSA.
1354         * ipa-prop.h (ipcp_driver): Remove declaration.
1355         (IS_VALID_TREE_MAP_INDEX): Add define.
1356
1357 2007-01-24  Geoffrey Keating  <geoffk@apple.com>
1358
1359         * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
1360
1361 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
1362
1363         PR middle-end/30447
1364         * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
1365         constant argument to cabs and do it without checking for
1366         -funsafe-math-optimizations.
1367
1368 2007-01-24  Douglas Gregor  <dgregor@osl.iu.edu>
1369
1370         * c-common.h (RID_FIRST_CXX0X): New.
1371         (RID_LAST_CXX0X): New.
1372         * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
1373         by -Wall.
1374         * c.opt (Wc++0x-compat): New.
1375         * doc/invoke.texi (-Wc++0x-compat): Document.
1376
1377 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
1378
1379         * builtins.c (fold_builtin_carg): New.
1380         (fold_builtin_1): Use it.
1381
1382 2007-01-24  Jan Hubicka  <jh@suse.cz>
1383
1384         * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
1385
1386         * ipa-inline.c (initial_insns, max_insns): Delete.
1387         (compute_max_insns): New function.
1388         (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
1389         of insns as base for code growth.
1390         (cgraph_decide_inlining): Make initial_insns local; do not compute
1391         max_insns.
1392         * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
1393         * doc/invoke.texi (inline-unit-growth): Update docs.
1394
1395 2007-01-24  Jakub Jelinek  <jakub@redhat.com>
1396
1397         * config/i386/i386.h (x86_cmpxchg16b): Remove const.
1398         (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
1399         * config/i386/i386.c (x86_cmpxchg16b): Remove const.
1400         (override_options): Add PTA_CX16 flag.  Set x86_cmpxchg16b
1401         for CPUs that have PTA_CX16 set.
1402
1403         PR middle-end/27416
1404         * gimplify.c (omp_check_private): New function.
1405         (gimplify_scan_omp_clauses): Use it for
1406         firstprivate/lastprivate/reduction.
1407
1408         PR middle-end/30494
1409         * gimplify.c (omp_add_variable): Don't call omp_notice_variable
1410         on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
1411
1412         PR middle-end/30421
1413         * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
1414         If lower_lastprivate_clauses emits some statements, append them
1415         to dlist rather than body_p and to body_p append an initializer.
1416         (lower_omp_for): Adjust caller.
1417
1418 2007-01-24  Steve Ellcey  <sje@cup.hp.com>
1419
1420         * target.h (globalize_decl_name): New.
1421         * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
1422         * output.h (default_globalize_decl_name): New.
1423         * varasm.c (asm_output_bss): Use globalize_decl_name instead of
1424         globalize_label.
1425         (globalize_decl): Ditto.
1426         (default_globalize_decl_name): New.
1427         * config/ia64/ia64.c (ia64_globalize_decl_name): New.
1428         (ia64_handle_version_id_attribute): New.
1429         (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
1430         (ia64_asm_output_external): Use globalize_decl_name instead
1431         of globalize_label.
1432         * doc/extend.texi (version_id): New pragma.
1433         * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
1434
1435 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
1436
1437         * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
1438         _uleb128_t and _Unwind_SWord with _sleb128_t.
1439         * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
1440         uw_frame_state_for, uw_update_context_1): Likewise.
1441         * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
1442         * unwind-pe.h (read_uleb128, read_sleb128,
1443         read_encoded_value_with_base): Likewise.
1444         * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
1445
1446 2007-01-24  Richard Guenther  <rguenther@suse.de>
1447
1448         * builtins.c (expand_builtin_cexpi): Get the fndecl
1449         for cexp in the correct way.
1450
1451 2007-01-24  Jan Hubicka  <jh@suse.cz>
1452
1453         * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
1454
1455 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
1456
1457         * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
1458         (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
1459         __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
1460         __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
1461         __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
1462         __SIZEOF_PTRDIFF_T__.
1463         * doc/cpp.texi: Documentation for the new builtin macros added.
1464
1465 2007-01-24  Uros Bizjak  <ubizjak@gmail.com>
1466
1467         * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
1468         log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
1469         instead of operands[2] to avoid access past the end of array.
1470
1471 2007-01-24  Richard Sandiford  <richard@codesourcery.com>
1472
1473         * reload1.c (emit_reload_insns): Pass the reload register
1474         for a non-spill output reload through forget_old_reloads_1.
1475
1476 2007-01-23  Joseph Myers  <joseph@codesourcery.com>
1477
1478         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
1479         _SOFT_DOUBLE if doubles use software floating-point.
1480         * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
1481         double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
1482         * config/rs6000/darwin-ldouble.c: Also compile functions for
1483         hard-float without FPRs.  Use fmsub function for all __NO_FPRS__
1484         cases.  Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
1485         * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
1486         commented-out long double override.
1487         (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
1488         * config/rs6000/eabispe.h: Likewise.
1489         * config/rs6000/rs6000.c (rs6000_override_options): Don't override
1490         long double for non-SPE.
1491         (rs6000_handle_option): Likewise.
1492         (invalid_e500_subreg): Disallow more subregs involding DImode,
1493         DFmode, TImode or TFmode.
1494         (rs6000_legitimate_offset_address_p): Check TFmode offsets for
1495         E500 double.
1496         (legitimate_lo_sum_address_p): Also check for TFmode for E500
1497         double.
1498         (rs6000_legitimize_address): Also handle TFmode for E500 double.
1499         (rs6000_legitimize_reload_address): Also handle TFmode for E500
1500         double.
1501         (rs6000_legitimate_address): Also check for TFmode for E500
1502         double.
1503         (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
1504         (spe_build_register_parallel): Handle TFmode and TCmode.
1505         (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
1506         double.
1507         (function_arg): Handle TFmode and TCmode for E500 double.
1508         (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
1509         in general.
1510         (print_operand): Handle TFmode and TImode for %y.
1511         (rs6000_generate_compare): Handle TFmode comparisons for E500
1512         double.
1513         (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
1514         (rs6000_function_value): Handle TFmode and TCmode for E500 double.
1515         (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
1516         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
1517         TFmode for E500 double.
1518         * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
1519         (floatsidf2): Enable for E500 double.
1520         (movtf_softfloat): Use rs6000_nonimmediate_operand.
1521         (extenddftf2): Change to extenddftf2_fprs.
1522         (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
1523         depending on TARGET_E500_DOUBLE.
1524         (extendsftf2): Enable for E500 double.
1525         (trunctfdf2): Enable for E500 double.
1526         (trunctfsf2): Change to trunctfsf2_fprs.
1527         (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
1528         depending on TARGET_E500_DOUBLE.
1529         (floatsitf2): Enable for E500 double.
1530         (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
1531         (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
1532         gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
1533         (negtf2): Change to negtf2_internal.
1534         (negtf2): New expander.
1535         (abstf2): Enable for E500 double.  Call gen_spe_abstf2_tst,
1536         gen_spe_abstf2_cmp or gen_abstf2_internal depending on
1537         TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
1538         (movdi_internal32): Use rs6000_nonimmediate_operand.
1539         (unnamed splitter): Likewise.
1540         * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
1541         TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
1542         (SPE64TF, DITI): New mode macros.
1543         (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
1544         modes.
1545         (frob_tf_ti): New.
1546         (frob_<mode>_di_2): New.
1547         (frob_tf_di_8_2): New.
1548         (frob_di_df): Change to frob_di_<mode>; allow more modes.
1549         (frob_ti_tf): New.
1550         (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
1551         more modes.
1552         (frob_ti_<mode>_8_2): New.
1553         (frob_ti_tf_2): New.
1554         (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
1555         mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
1556         TFmode.
1557         (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
1558         mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
1559         (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
1560         spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
1561         spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
1562         (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
1563         tsttflt_gp): New.
1564
1565 2007-01-23  Ian Lance Taylor  <iant@google.com>
1566
1567         * Makefile.in (OBJS-common): Reformat, alphabetize, but put
1568         insn-*.o first.
1569         (OBJS-archive): Reformat, alphabetize.
1570         (OBJS): Change out_object_file to OBJS-md.
1571
1572 2007-01-23  Uros Bizjak  <ubizjak@gmail.com>
1573
1574         * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
1575         JUMP_P and CALL_P predicates where applicable.
1576         * config/i386/i386.c: Ditto.
1577         * config/i386/i386.md: Ditto.
1578         * config/i386/mmx.md: Ditto.
1579         * config/i386/predicates.md: Ditto.
1580
1581 2007-01-22  Andreas Schwab  <schwab@suse.de>
1582
1583         * config/m68k/m68k.h: Fix comment.
1584
1585 2007-01-22  Jan Hubicka  <jh@suse.cz>
1586
1587         * passes.c (init_optimization_passes): Do not rerun
1588         pass_early_warn_uninitialized.
1589
1590 2007-01-22  Richard Guenther  <rguenther@suse.de>
1591
1592         PR tree-optimization/30038
1593         * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
1594         function.
1595         (execute_cse_sincos_1): Likewise.
1596         (execute_cse_sincos): Likewise.
1597         (gate_cse_sincos): Likewise.
1598         (pass_cse_sincos): New pass CSEing sin() and cos() calls using
1599         the cexpi() canonicalization of sincos().
1600         * tree-pass.h (pass_cse_sincos): Declare.
1601         * passes.c (init_optimization_passes): New pass pas_cse_sincos.
1602
1603 2007-01-21  Eric Botcazou  <ebotcazou@libertysurf.fr>
1604
1605         PR rtl-optimization/29329
1606         * combine.c (replaced_rhs_insn): Rename to i2mod.
1607         (replaced_rhs_value): Rename to i2mod_new_rhs.
1608         (i2mod_old_rhs): New global variable.
1609         (combine_instructions): Adjust for above change.  Save a copy of
1610         the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
1611         note are substituted in the second instruction.
1612         (distribute_notes) <REG_DEAD>: Adjust for above change.  Do not
1613         ditch the note if it pertains to the second eliminated register
1614         and this register is mentioned in i2mod_old_rhs.
1615
1616         Revert:
1617         2006-09-12  Eric Botcazou  <ebotcazou@libertysurf.fr>
1618
1619         * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
1620         the insn to which the note was originally attached.
1621
1622 2007-01-21  Jan Hubicka  <jh@suse.cz>
1623
1624         * ipa-inline.c (inlining_mode): Comment, move up.
1625         (cgraph_decide_inlining_incrementally): Do not perform inlining
1626         itself; fix handling of flattening of self recursive functions.
1627         (cgraph_find_cycles): Remove.
1628         (cgraph_flatten_node): Remove.
1629         (cgraph_decide_inlining): Use incremental inliner to handle flattening.
1630         (try_inline): New function.
1631         (cgraph_early_inlining): Update call of
1632         cgraph_decide_inlining_incrementally.  Apply inlining here.
1633         (apply_inline): Update call of cgraph_decide_inlining_incrementally.
1634
1635 2007-01-21  Dirk Mueller  <dmueller@suse.de>
1636
1637         PR bootstrap/30511
1638         * tree-vrp.c (check_array_bounds): do not warn
1639         about ADDR_EXPR's of ARRAY_REF's which are immediately
1640         used in binary expressions.
1641
1642 2007-01-21  Ira Rosen  <irar@il.ibm.com>
1643
1644         * tree-vectorizer.h (struct _stmt_vec_info): Add new field
1645         read_write_dep and macros for its access.
1646         * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
1647         * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
1648         argument, call vect_check_interleaving for every independent pair of
1649         data-refs. Mark loads that access the same memory location as a store
1650         in the loop.
1651         (vect_check_dependences): Remove.
1652         (vect_analyze_data_ref_dependences): Remove  vect_check_dependences
1653         call, fix the call to vect_analyze_data_ref_dependence.
1654         (vect_analyze_data_ref_access): For statements that access the same
1655         data-ref, check that they are not stores; for loads, check that there
1656         is no store that access the same location.
1657
1658 2007-01-20  Roger Sayle  <roger@eyesopen.com>
1659             Joseph Myers  <joseph@codesourcery.com>
1660
1661         * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
1662         (-fdump-rtl-stack): Likewise.
1663         (-fno-signed-zeros): Use @minus{} for a minus sign.  Correct typo.
1664         (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
1665         (-mpt-fixed): Use @minus{} for minus sign.
1666         (Using Precompiled Headers): Fix under/overfull hbox.
1667
1668 2007-01-20  Jan Hubicka  <jh@suse.cz>
1669
1670         * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
1671         place within annotation.
1672
1673 2007-01-20  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
1674
1675         * pa.c (output_move_double): Change array size of xoperands to 4.
1676
1677 2007-01-20  Jan Hubicka  <jh@suse.cz>
1678
1679         * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
1680         as temporary.
1681
1682 2007-01-19  Ian Lance Taylor  <iant@google.com>
1683
1684         * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
1685         constant to unsigned HOST_WIDE_INT before negating.
1686
1687 2007-01-19  Ian Lance Taylor  <iant@google.com>
1688
1689         * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
1690         to unsigned.
1691         (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
1692         (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
1693         (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
1694         (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
1695         (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
1696         (realloc_vdef, realloc_vuse): Change second parameter to
1697         unsigned.
1698         (ssa_operand_iterator_d): Change vuse_index and mayuse_index
1699         fields to unsigned.
1700         * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
1701         unsigned. Change x and lim locals to unsigned.
1702         (realloc_vdef, realloc_vuse): Change num_elem parameter to
1703         unsigned.
1704         (finalize_ssa_vuse_ops): Change old_i local to unsigned.
1705         (copy_virtual_operands): Change i and n locals to unsigned.
1706
1707 2007-01-19  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
1708
1709         PR middle-end/29335
1710         * builtins.c (fold_builtin_1): Handle builtin fdim.
1711
1712 2007-01-20  Jan Hubicka  <jh@suse.cz>
1713
1714         * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
1715         * tree-ssa-structalias.c (compute_points_to_sets): Do call
1716         init_alias_heapvars.
1717         (init_alias_heapvars): Initialize only when not already initialized.
1718         (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
1719
1720 2007-01-19  Roger Sayle  <roger@eyesopen.com>
1721
1722         * common.opt (fsigned-zeros): New command line option.
1723         * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
1724         of flag_unsafe_math_optimizations.
1725         * opts.c (set_fast_math_flags): The -ffast-math command line option
1726         implies -fno-signed-zeros.
1727         (fast_math_flags_set_p): Likewise.
1728
1729         * doc/invoke.texi: Document new -fno-signed-zeros option, and update
1730         the documentation of -ffast-math appropriately.  Wrap long lines.
1731
1732 2007-01-19  Steve Ellcey  <sje@cup.hp.com>
1733
1734         * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
1735         * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
1736
1737 2007-01-19  Tomas Bily  <tbily@suse.cz>
1738
1739         * cgraphunit.c (cgraph_finalize_function): Updating of pid
1740         * tree-profile.c:
1741         (tree_init_ic_make_global_vars): New function
1742         (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
1743         (tree_gen_ic_profiler): New function
1744         (tree_gen_ic_func_profiler): New function
1745         (tree_profiling): Added calling of tree_gen_ic_func_profiler
1746         (tree_profile_hooks): Added hook for indirec/virtual calls
1747         * value-prof.c (tree_find_values_to_profile): New case for
1748         indirect calls
1749         (tree_values_to_profile): Call for determining indirect/virtual
1750         counters
1751         (tree_indirect_call_to_profile): New function
1752         (tree_ic_transform): New function
1753         (tree_ic): New function
1754         (find_func_by_pid): New function
1755         (init_pid_map): New function
1756         (tree_value_profile_transformations): Added check for
1757         indirect/virtual call transformation
1758         * value-prof.h (enum hist_type): New counter type for
1759         indirect/virtual calls
1760         (profile_hooks): Added new hook for profiling indirect/virtual
1761         calls
1762         * profile.c (instrument_values): New case for indirect/virtual
1763         call added
1764         * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
1765         (GCOV_COUNTER_V_INDIR): New counter type
1766         (GCOV_COUNTER_NAMES): New name of counter "indirect" added
1767         (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
1768         call added
1769         * cgraph.c: Definition of cgraph_max_pid
1770         (cgraph_create_node): Default init of pid attribute
1771         * cgraph.h: Declaration of cgraph_max_pid
1772         (struct cgraph_node): Added pid attribute
1773         * libgcov.c (__gcov_indirect_call_profiler): New function
1774         (__gcov_one_value_profiler_body): New function
1775         (__gcov_one_value_profiler): Body was moved to
1776         __gcov_one_value_profiler_body and calls it
1777
1778 2007-01-19  Basile Starynkevitch  <basile@starynkevitch.net>
1779
1780         * doc/gty.texi (Options): Document the mark_hook option to GTY.
1781         * gengtype.c (write_types_data, write_func_for_structure,
1782         write_types, ggc_wtd, pch_wtd): Add skip_hooks to
1783         write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
1784         (walk_type, write_func_for_structure): Generate the mark_hook if
1785         needed.
1786
1787 2007-01-19  Jan Hubicka  <jh@suse.cz>
1788
1789         * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
1790         'early' argument take inlining mode argument specifying whether to
1791         inline for size/speeed or all functions; add support for flattening;
1792         improve dumpting.
1793         (cgraph_early_inlining): Update call of decide_inlining_incrementally.
1794
1795 2007-01-19  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
1796
1797         PR c++/17947
1798         * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
1799         the name of the declared identifier.
1800
1801 2007-01-19  Dirk Mueller  <dmueller@suse.de>
1802
1803         * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
1804         result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
1805         duplicate evaluation.
1806
1807 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
1808
1809         * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
1810         Macroize expander using X87MODEF12 mode macro.  Extend operand 1
1811         to XFMode, use acosxf2 and truncate result to requested mode.
1812         Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1813         (asin<mode>2): Similarly, with asin expanders.
1814         (*fscalexf4_i387): Rename from *fscalexf4.
1815         (expNcorexf3): New expander.
1816         (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
1817         (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
1818         X87MODEF12 mode macro.  Extend operand 1 to XFMode, use expxf2 and
1819         truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
1820         patterns for SSE math.
1821         (exp10<mode>2): Similarly, with exp10 expanders.
1822         (exp2<mode>2): Similarly, with exp2 expanders.
1823         (expm1<mode>2): Similarly, with expm1 expanders.
1824         (ldexp<mode>3): Similarly, with ldexp expanders.
1825         (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
1826         Use gen_truncxf<mode>2_i387_noop to truncate result.
1827
1828 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1829
1830         * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
1831         for ColdFire targets.
1832
1833 2007-01-19  Nathan Sidwell  <nathan@codesourcery.com>
1834             Richard Sandiford  <richard@codesourcery.com>
1835
1836         * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
1837         (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
1838         (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
1839         * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
1840         (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1841         * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1842
1843 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1844
1845         * config/m68k/m68k.md (adddi_dilshr32): Rename to...
1846         (*adddi_dilshr32): ...this.  Fix formatting.  Remove commented-out
1847         non-canonical pattern.  Restrict to !TARGET_COLDFIRE.
1848         (*adddi_dilshr32_cf): New pattern.
1849         (adddi3, subdi3): Remove first alternatives.
1850
1851 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1852
1853         * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
1854         implemented using move.l, do not use its cc result for floating-point
1855         comparisons.
1856
1857 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1858
1859         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
1860         from accepting 's' constraints if flag_pic.
1861
1862 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1863
1864         * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
1865         (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
1866         (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
1867         condition from TARGET_68881 to TARGET_HARD_FLOAT.
1868
1869 2007-01-19  Sandra Loosemore  <sandra@codesourcery.com>
1870
1871         * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
1872         ColdFire alternatives.
1873         * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
1874         * config/m68k/m68k.md (clzsi2):  Define for ColdFire
1875         architectures that support ff1 instruction.
1876
1877 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1878             Julian Brown  <julian@codesourcery.com>
1879
1880         * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
1881         * config/m68k/m68k.md (*movsi_cfv4): Fold into...
1882         (*movsi_cf): ...here.  Remove unnecessary 'R' from 'Rg'.
1883         Add commentary.
1884
1885 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1886
1887         * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
1888         and return a bool.
1889         (output_move_const_into_data_reg, output_move_simode_const): Delete.
1890         * config/m68k/m68k.c (const_method, const_int_cost): Take a
1891         HOST_WIDE_INT instead of an rtx.
1892         (m68k_rtx_costs): Update call accordingly.
1893         (output_move_const_into_data_reg): Likewise.  Fix formatting.
1894         (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
1895         Return a bool.
1896         (output_move_simode_const): Update calls after above changes.
1897         Rework to use automatic variables and predicates like MEM_P.
1898         * config/m68k/m68k.md (pushexthisi_const): Update call to
1899         valid_mov3q_const.
1900
1901 2007-01-19  Dirk Mueller  <dmueller@suse.de>
1902
1903         * tree-ssa-alias.c (perform_var_substitution): Fix typo
1904         in dump_flags test.
1905
1906 2007-01-19  Richard Guenther  <rguenther@suse.de>
1907
1908         * builtins.c (expand_builtin_cexpi): Fall back to expanding
1909         via cexp in case sincos is not available.
1910
1911 2007-01-19  Richard Guenther  <rguenther@suse.de>
1912
1913         * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
1914         * defaults.h (TARGET_HAS_SINCOS): Default to off.
1915         * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
1916         * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
1917         * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
1918         * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
1919         * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
1920         * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
1921
1922 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
1923
1924         * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
1925         New insn patterns.
1926         (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
1927         (atan2xf3): Directly generate RTL pattern.
1928         (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
1929         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
1930         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1931         disable patterns for SSE math.
1932         (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
1933         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
1934         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1935         disable patterns for SSE math.
1936
1937 2007-01-19  Alexandre Oliva  <aoliva@redhat.com>
1938
1939         * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
1940
1941 2007-01-18  Roger Sayle  <roger@eyesopen.com>
1942
1943         * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
1944         VIEW_CONVERT_EXPR to the same type as it's operand.
1945
1946 2007-01-18  David Edelsohn  <edelsohn@gnu.org>
1947
1948         * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
1949         configured for long double 128.
1950
1951 2007-01-18  Mike Stump  <mrs@apple.com>
1952
1953         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
1954         error.
1955
1956 2007-01-18  Michael Meissner  <michael.meissner@amd.com>
1957
1958         * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
1959         type correct.
1960
1961 2007-01-18  Jan Hubicka  <jh@suse.cz>
1962
1963         * tree-ssa-operands.c (vop_free_bucket_size): Never return value
1964         greater than NUM_VOP_FREE_BUCKETS.
1965
1966 2007-01-18  Daniel Berlin  <dberlin@dberlin.org>
1967
1968         * tree-ssa-structalias.c: Update comments.
1969         (ptabitmap_obstack): Removed.
1970         (pta_obstack): New.
1971         (oldpta_obstack): Ditto.
1972         (stats): Add a few members.
1973         (struct variable_info): Remove node, complex, address_taken, and
1974         indirect_target members. Add oldsolution member.
1975         (new_var_info): Do not initialize removed members.
1976         (constraint_expr_type): Remove INCLUDES.
1977         (constraint_graph): Add size, implicit_preds, rep,
1978         indirect_cycles, eq_rep, label, direct_nodes, and complex members.
1979         (FIRST_REF_NODE): New macro.
1980         (LAST_REF_NODE): Ditto.
1981         (FIRST_ADDR_NODE): Ditto.
1982         (find): New function.
1983         (unite): Ditto.
1984         (dump_constraint): Do not handle INCLUDES.
1985         (insert_into_complex): Do not insert duplicate constraints.
1986         (condense_varmap_nodes): Renamed and rewritten into ...
1987         (merge_node_constraints): This. Also fix bug in handling of
1988         offseted copy constraints.
1989         (clear_edges_for_node): No longer need to deal with preds at all,
1990         or removing associated preds/succs.
1991         (merge_graph_nodes): Deal with indirect_cycles.
1992         Don't deal with predecessors.
1993         (add_implicit_graph_edge): New function.
1994         (add_pred_graph_edge): Ditto.
1995         (add_graph_edge): Don't deal with predecessors.
1996         (build_constraint_graph): Removed.
1997         (build_pred_graph): New function.
1998         (build_succ_graph): Ditto.
1999         (struct scc_info): Removed in_component. Added roots, dfs, and
2000         node_mapping. Remove visited_index, unification_queue.
2001         (scc_visit): Deal with union-find we do now.
2002         Deal with cycles with REF nodes.
2003         (collapse_nodes): Renamed and rewritten to ...
2004         (unify_nodes): This.
2005         (process_unification_queue): Removed.
2006         (topo_visit): Cleanup
2007         (do_da_constraint): Use find.
2008         (do_sd_constraint): Ditto.
2009         (do_ds_constraint): Ditto.
2010         (do_complex_constraint): Ditto.
2011         (init_scc_info): Update for removed and added members.
2012         (find_and_collapse_graph_cycles): Renamed and rewritten into ...
2013         (find_indirect_cycles): This.
2014         (equivalence_class): New variable.
2015         (label_visit): New function.
2016         (perform_variable_substitution): Rewritten.
2017         (free_var_substitution_info): New function.
2018         (find_equivalent_node): Ditto.
2019         (move_complex_constraints): Ditto.
2020         (eliminate_indirect_cycles): Ditto.
2021         (solve_graph): Only propagate changed bits.
2022         Use indirect cycle elimination.
2023         Use find.
2024         (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
2025         (tree_id_eq): Renamed to ...
2026         (tree_vi_eq): This. Update for member change
2027         (insert_id_for_tree): Renamed and rewritten to ...
2028         (insert_vi_for_tree): This.
2029         (lookup_id_for_tree): Renamed and rewritten to ...
2030         (lookup_vi_for_tree): This.
2031         (get_id_for_tree): Renamed and rewritten to ...
2032         (get_vi_for_tree): Ditto.
2033         (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
2034         (process_constraint): Don't handle INCLUDES.
2035         Remove special ADDRESSOF case.
2036         (find_func_aliases): Rewrite to use vi functions instead of id
2037         ones.
2038         (create_function_info_for): Ditto.
2039         (create_variable_info_for): Ditto.
2040         (intra_create_variable_infos): Ditto.
2041         (merge_smts_into): Ditto.
2042         (find_what_p_points_to): Ditto.
2043         (init_base_vars): Ditto.
2044         (init_alias_vars): Ditto.
2045         (remove_preds_and_fake_succs): New function.
2046         (dump_sa_points_to_info): Dump new stats.
2047         (dump_solution_for_var): Use find.
2048         (set_used_smts): Fix formatting.
2049         (compute_points_to_sets): Updated for new functions.
2050         (ipa_pta_execute): Ditto.
2051
2052 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
2053             Richard Sandiford  <richard@codesourcery.com>
2054
2055         * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
2056         parallels.
2057         * calls.c (expand_call): If the return value is a PARALLEL,
2058         extract its first member.
2059         * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
2060         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2061         use FUNCTION_EXTRA_EPILOGUE.
2062         (m68k_function_value): Return a PARALLEL if the return value
2063         is of a pointer type.
2064         * config/m68k/netbsd-elf.h (current_function_returns_pointer)
2065         (FUNCTION_EXTRA_EPILOGUE): Remove.
2066         * config/m68k/m68k.md (D0_REG): New constant.
2067
2068 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
2069
2070         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2071         output a NOP for empty epilogues.
2072
2073 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2074
2075         * config/m68k/m68k.c (m68k_use_return_insn): Update comments
2076         before function.  Extend register save check to include all
2077         registers, not just integer ones.
2078
2079 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
2080
2081         * config/m68k/m68k-protos.h (use_return_insn): Rename to...
2082         (m68k_use_return_insn): ...this.
2083         * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
2084         * config/m68k/m68k.c (use_return_insn): Rename to...
2085         (m68k_use_return_insn): ...this.
2086         * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
2087         of USE_RETURN_INSN.
2088
2089 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
2090
2091         * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
2092         nearest even, fix denormal rounding overflow.
2093
2094 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2095
2096         * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
2097         f%$smove and f%$move.
2098         (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
2099         (extendsfdf2_cf): Use fdmove instead of f%&move.
2100         (truncdfsf2_cf): Use fsmove instead of f%$smove.
2101         (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
2102
2103 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2104
2105         * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
2106         GPR<-GPR moves.
2107
2108 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2109
2110         * real.h (real_format): Add a canonical_nan_lsbs_set field.
2111         (coldfire_single_format): Declare.
2112         (coldfire_double_format): Likewise.
2113         * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
2114         of qnan_msb_set to determine the lower bits of a canonical
2115         NaN significand.
2116         (encode_ieee_double): Likewise.
2117         (encode_ieee_quad): Likewise.
2118         (ieee_single_format): Initialize canonical_nan_lsbs_set.
2119         (mips_single_format): Likewise.
2120         (ieee_double_format): Likewise.
2121         (mips_double_format): Likewise.
2122         (ieee_extended_motorola_format): Likewise.
2123         (ieee_extended_intel_96_format): Likewise.
2124         (ieee_extended_intel_128_format): Likewise.
2125         (ieee_extended_intel_96_round_53_format): Likewise.
2126         (ibm_extended_format): Likewise.
2127         (mips_extended_format): Likewise.
2128         (ieee_quad_format): Likewise.
2129         (mips_quad_format): Likewise.
2130         (vax_f_format): Likewise.
2131         (vax_d_format): Likewise.
2132         (vax_g_format): Likewise.
2133         (i370_single_format): Likewise.
2134         (i370_double_format): Likewise.
2135         (decimal_single_format): Likewise.
2136         (decimal_double_format): Likewise.
2137         (decimal_quad_format): Likewise.
2138         (c4x_single_format): Likewise.
2139         (c4x_extended_format): Likewise.
2140         (real_internal_format): Likewise.
2141         (coldfire_single_format): New real_format.
2142         (coldfire_double_format): Likewise.
2143         * config/pdp11/pdp11.c (pdp11_f_format): Initialize
2144         canonical_nan_lsbs_set.
2145         (pdp11_d_format): Likewise.
2146         * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
2147         if TARGET_COLDFIRE_CPU.
2148
2149 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2150
2151         * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
2152         (output_call, m68k_legitimize_call_address): Declare.
2153         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
2154         parenthesees.  Add support for a 'W' constraint.
2155         (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
2156         (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
2157         (m68k_symbolic_call, m68k_symbolic_jump): Declare.
2158         * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
2159         variables.
2160         (override_options): Initialize them.  Do not set flag_no_function_cse
2161         for TARGET_ID_SHARED_LIBRARY.
2162         (m68k_output_pic_call): Delete.
2163         (m68k_legitimize_call_address): New function.
2164         (print_operand): Remove the %o prefix.  Handle the %p prefix.
2165         (output_call): New function.
2166         (m68k_output_mi_thunk): Use m68k_symbolic_jump.  Always load the
2167         target address from the GOT if symbolic jumps are not allowed.
2168         * config/m68k/m68k.md (call, general_operand): Do not set
2169         SYMBOL_REF_FLAG.  Use m68k_legitimize_call_address instead.
2170         Merge separate flag_pic and !flag_pic define_insns into...
2171         (*call, *call_value): ...these new patterns.  Match the address
2172         rather than the containing MEM and require it to be a call_operand.
2173         Use output_call to generate the asm template.
2174         * config/m68k/predicates.md (const_call_operand): New predicate.
2175         (call_operand): Likewise.
2176
2177 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
2178
2179         * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
2180         (STACK_GROWS_DOWNWARD): Define to 1.
2181         (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
2182         comments, emphasizing that these values are only defaults.
2183         * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
2184         * config/m68k/m68k.c (const_method): Remove trailing whitespace.
2185
2186 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2187
2188         * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
2189
2190 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
2191             Richard Sandiford  <richard@codesourcery.com>
2192
2193         * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
2194         (FP_REGNO_P): New macros.
2195         (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
2196         (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
2197         (EH_RETURN_STACKADJ_RTX): Use A0_REG.
2198         * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
2199
2200 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
2201
2202         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2203         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
2204         to $tmake_file.
2205         * config/m68k/t-floatlib: New file, extracting common code from...
2206         * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
2207         * config/m68k/t-uclinux: Here.
2208         * config/m68k/fpgnulib.c: Do not compile extendeed precision
2209         routines on ColdFire targets.
2210
2211 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
2212
2213         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2214         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
2215         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
2216         (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
2217         explicitly set MOTOROLA to 1.
2218         * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
2219
2220 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2221             Nathan Sidwell  <nathan@codesourcery.com>
2222
2223         * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
2224         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
2225         * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
2226         * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
2227         * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
2228         * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
2229         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
2230         (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
2231         (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
2232         * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2233         * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
2234         (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
2235         (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
2236         (DBX_REGISTER_NUMBER): Delete.
2237         * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
2238         (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2239         * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
2240
2241 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2242
2243         * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
2244         * config/flat.h: New file.
2245         * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
2246         OBJECT_FORMAT_FLAT.
2247         * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
2248         indirect references for -msep-data or -mid-shared-library.
2249         Do not use PC-relative code addresses either.
2250         * config/m68k/m68k.c (override_options): Restrict -fPIC error
2251         to -mpcrel.
2252         * config/m68k/uclinux.h (STARTFILE_SPEC): Define.  Use Scrt1.o
2253         for shared libraries and crt1.o for executables.  Use crti.o and
2254         crtbegin.o.
2255         (ENDFILE_SPEC): Use crtend.o and crtn.o.
2256         (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
2257         Do not add -elf2flt or -shared-lib-id options here.
2258         (LINK_SPEC): Define.  Pass -elf2flt if no -elf2flt option is given.
2259         Pass -shared-lib-id if -mid-shared-library, taking the library
2260         identifier from -mshared-library-id if given, otherwise
2261         defaulting to 0.
2262         (EH_FRAME_IN_DATA_SECTION): Do not undefine.
2263         (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
2264         (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
2265         and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
2266         (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
2267         * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
2268         and crtend.o.
2269         * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
2270         jump sequence for ISA A and ISA A+.
2271         (PICJUMP): Likewise.
2272
2273 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2274
2275         * config.gcc (m68k-*-uclinux*): Base the port on the common
2276         and m68k GNU/Linux files rather than on the generic ELF ones.
2277         * config/m68k/uclinux.h (TARGET_VERSION): Override.
2278         (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
2279
2280 2007-01-18  Julian Brown  <julian@codesourcery.com>
2281             Richard Sandiford  <richard@codesourcery.com>
2282
2283         * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
2284         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2285         * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
2286         before redefining.
2287         * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
2288         80 unconditionally.
2289         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2290
2291 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2292
2293         * doc/install.texi: Document m68k-uclinuxoldabi.
2294         * config.gcc (m68k-*-uclinuxoldabi): New configuration.
2295         * config/m68k/uclinux-oldabi.h: New file, copied from
2296         config/m68k/uclinux.h.
2297
2298 2007-01-18  Kaz Kojima  <kkojima@rr.iij4u.or.jp>
2299
2300         * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
2301         * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
2302         (m32r_compute_frame_size): Use unsigned for regno.
2303         (m32r_reload_lr): Use gen_frame_mem.
2304         (pop): New.
2305         (m32r_output_function_epilogue): Don't output the function epilogue
2306         textually here.
2307         (m32r_expand_epilogue): New.
2308         (direct_return): Return false if the function has the interrupt
2309         attribute.
2310         (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
2311         * config/m32r/m32r.md (epilogue): New expander.
2312         (return_lr, return_rte): New insns.
2313         (return): Make it expander.
2314         (return_normal): New expander.
2315
2316 2007-01-18  Josh Conner  <jconner@apple.com>
2317
2318         PR target/30485
2319         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
2320         support for UNLE, UNLT, UNGE, and UNGT.
2321
2322 2007-01-18  Jan Hubicka  <jh@suse.cz>
2323
2324         * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
2325         * tree-tailcall (add_virtual_phis): Likewise.
2326         (optimize_tail_call): Return TODOs.
2327         (execute_tail_calls): Return TODOs.
2328         * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
2329         * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
2330         changed.
2331         * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
2332         Cleanup cfg using TODO.
2333         * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
2334         * passes.c (execute_function_todo): When cleanup did something, remove
2335         unused locals.
2336         * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
2337         (make_edges): Don't cleanup_cfg.
2338
2339 2007-01-18  Uros Bizjak  <ubizjak@gmail.com>
2340
2341         * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
2342         UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
2343         operate on the top of stack.
2344         [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
2345         register handling of instructions that output to the second
2346         stack slot.
2347         [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
2348         (move_for_stack_reg): Special-case check for dead destination
2349         stack slot for constant load of 1.0 inside UNSPEC_TAN.
2350
2351         * config/i386/i386.md (UNSPEC_TAN): New constant.
2352         (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
2353         (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
2354         to correctly model move of constant 1.0 to top stack slot.
2355         (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
2356         (unnamed peephole2 pattern): Remove corresponding peephole2
2357         pattern that optimizes tan insn and loading of constant 1.0.
2358         (tanxf2): Use fptanxf4_i387.
2359         (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
2360         patterns using X87MODEF12 mode macro.  Use fptan_extend<mode>xf4_i387
2361         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
2362         disable patterns for SSE math.
2363         (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
2364         (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
2365
2366 2007-01-18  Dirk Mueller  <dmueller@suse.de>
2367             Richard Guenther <rguenther@suse.de>
2368
2369         PR diagnostic/8268
2370         * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
2371         * common.opt (Warray-bounds): Add new warning option.
2372         * c-opts.c (c_common_handle_option): Define -Warray-bounds
2373         if -Wall is given.
2374         * Makefile.in: make tree-vrp.o depend on toplev.h
2375         * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
2376         is enabled.
2377         (check_array_refs, check_array_bounds, check_array_ref): New.
2378
2379 2007-01-18  Jan Hubicka  <jh@suse.cz>
2380
2381         * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
2382         (execute_ssa_ccp): Return flags conditionally.
2383         * tree-ssa-propagate.c (substitue_and_fold): Return if something was
2384         changed.
2385         * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
2386
2387 2007-01-18  Steven Bosscher  <steven@gcc.gnu.org>
2388
2389         * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
2390         the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
2391
2392         * Makefile.c (GTFILES): Add cfglayout.h.
2393         * gengtype.c (open_base_files): Likewise.
2394         * cfglayout.c (cfg_layout_function_footer,
2395         cfg_layout_function_header) Reindent to make gengtype happy.
2396         * cfglayout.h (cfg_layout_function_footer,
2397         cfg_layout_function_header): Add GTY(()) marker.
2398
2399         * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
2400
2401 2007-01-18  Ben Elliston  <bje@au.ibm.com>
2402
2403         * genautomata.c (write_automata): Include xstrerror output in the
2404         error message if writing the DFA description file fails.
2405
2406 2007-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2407
2408         * config/mips/mips-protos.h (mips_output_external): Make it
2409         return void.
2410         * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
2411         * config/mips/mips.c (irix_output_external_libcall): Likewise.
2412         (extern_list): Likewise.
2413         (extern_head): Likewise.
2414         (TARGET_ASM_FILE_END): Likewise.
2415         (mips_file_end): Likewise.
2416         (mips_output_external): Rewritten.
2417
2418 2007-01-18  Ben Elliston  <bje@au.ibm.com>
2419
2420         * genpreds.c (write_insn_preds_c): Only write out the function
2421         body for regclass_for_constraint if we have register constraints.
2422
2423 2007-01-17  Tom Tromey  <tromey@redhat.com>
2424
2425         * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
2426         * doc/install.texi (Testing): Use sourceware.org.
2427         (Binaries): Likewise.
2428         (Specific): Likewise.
2429         * doc/contrib.texi (Contributors): Use sourceware.org.
2430
2431 2007-01-17  Anatoly Sokolov <aesok@post.ru>
2432
2433         * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
2434         (AVR_ENHANCED): Rename to ...
2435         (AVR_HAVE_MUL): ... new.
2436         (avr_enhanced_p): Rename to ...
2437         (avr_have_mul_p): ... new.
2438         (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
2439         'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
2440         * config/avr/avr.c (avr_enhanced_p): Rename to ...
2441         (avr_have_mul_p): ... new.
2442         (base_arch_s): Rename 'enhanced' to 'have_mul'.
2443         (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
2444         'avr_enhanced_p' and 'enhanced'.
2445         (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
2446         AVR_HAVE_MUL instead of AVR_ENHANCED.
2447         * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
2448         (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
2449         *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
2450         instead of AVR_ENHANCED.
2451         (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
2452         * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
2453         (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
2454
2455 2007-01-17  Ian Lance Taylor  <iant@google.com>
2456
2457         * vec.h (VEC_reserve_exact): Define.
2458         (vec_gc_p_reserve_exact): Declare.
2459         (vec_gc_o_reserve_exact): Declare.
2460         (vec_heap_p_reserve_exact): Declare.
2461         (vec_heap_o_reserve_exact): Declare.
2462         (VEC_OP (T,A,reserve_exact)): New static inline function, three
2463         versions.
2464         (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
2465         negative parameter.
2466         (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
2467         (VEC_OP (T,A,copy)) [all versions]: Likewise.
2468         (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
2469         * vec.c (calculate_allocation): Add exact parameter.  Change all
2470         callers.
2471         (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
2472         (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
2473         (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
2474         (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
2475         (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
2476         (vec_heap_p_reserve_exact): New function.
2477         (vec_heap_o_reserve_exact): New function.
2478
2479 2007-01-17  Jan Hubicka  <jh@suse.cz>
2480
2481         * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
2482
2483 2007-01-17  Eric Christopher  <echristo@apple.com>
2484
2485         * config.gcc: Support core2 processor.
2486
2487 2007-01-16  Jan Hubicka  <jh@suse.cz>
2488
2489         * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
2490         the names, just unlink the chain so we don't crash on dangling pointers
2491         to dead SSA names.
2492
2493 2007-01-16  Jan Hubicka  <jh@suse.cz>
2494
2495         * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
2496         * tree-pass.h: Reorder to make IPA passes appear toegher.
2497         (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
2498         * cgraphunit.c (cgraph_finalize_function): Do not compute inling
2499         parameters, do not call early inliner.
2500         * ipa-inline.c: Update comments.  Include tree-flow.h
2501         (cgraph_decide_inlining): Do not compute inlining parameters.
2502         (cgraph_decide_inlining_incrementally): Return TODOs; assume to
2503         be called with function context set up.
2504         (pass_ipa_inline): Remove unreachable functions before pass.
2505         (cgraph_early_inlining): Simplify assuming to be called from the
2506         PM as local pass.
2507         (pass_early_inline): New pass.
2508         (cgraph_gate_ipa_early_inlining): New gate.
2509         (pass_ipa_early_inline): Turn into simple wrapper.
2510         (compute_inline_parameters): New function.
2511         (gate_inline_passes): New gate.
2512         (pass_inline_parameters): New pass.
2513         (apply_inline): Move here from tree-optimize.c
2514         (pass_apply_inline): New pass.
2515         * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
2516         transforming.
2517         * tree-inline.c (optimize_inline_calls): Return TODOs rather than
2518         doing them by hand.
2519         (tree_function_versioning): Do not allocate dummy struct function.
2520         * tree-inline.h (optimize_inline_calls): Update prototype.
2521         * tree-optimize.c (execute_fixup_cfg): Export.
2522         (pass_fixup_cfg): Remove
2523         (tree_rest_of_compilation): Do not apply inlines.
2524         * tree-flow.h (execute_fixup_cfg): Declare.
2525         * Makefile.in (gt-passes.c): New.
2526         * passes.c: Include gt-passes.h
2527         (init_optimization_passes): New passes.
2528         (nnodes, order): New static vars.
2529         (do_per_function_toporder): New function.
2530         (execute_one_pass): Dump current pass here.
2531         (execute_ipa_pass_list): Don't dump current pass here.
2532
2533 2007-01-16  Janis Johnson  <janis187@us.ibm.com>
2534
2535         * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
2536         (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
2537
2538 2007-01-16  David Edelsohn  <edelsohn@gnu.org>
2539
2540         * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
2541         (strong_alias): Define.
2542         (__gcc_qmul): Provide non-FMA for soft-float.
2543         (__gcc_qdiv): Same.
2544         (__gcc_qneg): New.
2545         (__gcc_qeq): New.
2546         (__gcc_qle): New.
2547         (__gcc_qge): New.
2548         (__gcc_qunord): New.
2549         (__gcc_stoq): New.
2550         (__gcc_dtoq): New.
2551         (__gcc_qtos): New.
2552         (__gcc_qtod): New.
2553         (__gcc_qtoi): New.
2554         (__gcc_qtou): New.
2555         (__gcc_itoq): New.
2556         (__gcc_utoq): New.
2557         (fmsub): New.
2558         * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
2559         soft-float functions.
2560         * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
2561         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
2562         about long double soft float.
2563
2564 2007-01-16  Dorit Nuzman  <dorit@il.ibm.com>
2565             Tehila Meyzels  <tehila@il.ibm.com>
2566
2567         * tree-vectorizer.h (is_pattern_stmt_p): New.
2568         * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
2569         formatting (tabs instead of spaces). Cleanup and clarify setting
2570         of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
2571         * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
2572         (vectorizable_type_demotion): Check that types are integral.
2573         (vectorizable_type_promotion): Likewise.
2574         (vectorizable_store): Fix typo.  Eliminate new-line at end of
2575         comments.
2576
2577 2007-01-16  Jan Hubicka  <jh@suse.cz>
2578
2579         * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
2580         cgraph edges too.
2581
2582 2007-01-15  Eric Christopher  <echristo@apple.com>
2583
2584         * ifcvt.c: Include vec.h, vecprim.h.
2585         (check_cond_move_block): New argument regs.
2586         Reorganize. Add registers used to regs.
2587         (cond_move_process_if_block): Use regs set above as
2588         loop bounds.
2589
2590 2007-01-15  Eric Christopher  <echristo@apple.com>
2591
2592         * config/darwin.h: Update copyright.
2593         (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
2594         (LINK_COMMAND_SPEC): Add -u.
2595         (LINK_SPEC): Fix umbrella for above.
2596
2597 2007-01-15  Joseph S. Myers  <joseph@codesourcery.com>
2598
2599         * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
2600         glibc CVS.
2601
2602 2007-01-15  Tom Tromey  <tromey@redhat.com>
2603
2604         * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
2605         * doc/install.texi (Testing): Don't mention jacks.
2606         (Configuration): Document --enable-java-maintainer-mode.  Move
2607         --with-java-home to libgcj-specific section.  Document
2608         --with-ecj-jar.
2609         (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
2610
2611 2007-01-15  Jan Hubicka  <jh@suse.cz>
2612
2613         * tree-ssa-dce.c (DCE_TODOs): New.
2614         (propagate_necessity): Return if something changed.
2615         (eliminate_unnecessary_stmts): Likewise.
2616         (perform_tree_ssa_dce): Return TODO flags when needed.
2617         (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
2618
2619 2007-01-15  Uros Bizjak  <ubizjak@gmail.com>
2620
2621         * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
2622         (fyl2x_extend<mode>xf3_i387): New insn pattern.
2623         (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
2624         patterns using X87MODEF12 mode macro.  Extend operand 1
2625         to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
2626         (log10<mode>2): Ditto.
2627         (log2<mode>2): Ditto.
2628         (log1p<mode>2): Ditto.
2629         (logb<mode>2): Ditto.
2630         (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
2631         (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
2632         (*fxtractxf3_i387): Rename from *fxtractxf3.
2633         (fxtract_extend<mode>xf3_i387): New insn pattern.
2634         (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
2635         * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
2636         and gen_fyl2xxf3_i387().
2637
2638 2007-01-14  Zdenek Dvorak <dvorakz@suse.cz>
2639
2640         * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
2641         * cfgloopmanip.c (fix_loop_placement): Made static.  Use
2642         get_loop_exit_edges.  Changed return type to bool.
2643         * cfgloop.h (fix_loop_placement): Declaration removed.
2644
2645 2007-01-14  Dorit Nuzman  <dorit@il.ibm.com>
2646
2647         * param.h (MIN_VECT_LOOP_BOUND): New.
2648         * params.def (MIN_VECT_LOOP_BOUND): New.
2649         * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
2650         argument - minimum threshold for number of iterations.
2651         * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
2652         argument to declaration.
2653         * tree-vect-analyze.c (vect_analyze_operations): Check value of
2654         MIN_VECT_LOOP_BOUND.
2655         * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
2656         slpeel_tree_peel_loop_to_edge with additional argument.
2657         (vect_do_peeling_for_alignment): Likewise.
2658         * doc/invoke.texi (min-vect-loop-bound): Document new param option.
2659
2660 2007-01-14  Uros Bizjak  <ubizjak@gmail.com>
2661
2662         PR target/30413
2663         * config/i386/i386.c (print_operand) ['z']: Output 'b' for
2664         operands of size 1.
2665
2666 2007-01-14  Jan Hubicka  <jh@suse.cz>
2667
2668         * tree-dfa.c (remove_referenced_var): New function.
2669         * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
2670         prune referenced vars list too.
2671         * tree-flow.h (remove_referenced_var): Declare.
2672
2673 2007-01-14  Jan Hubicka  <jh@suse.cz>
2674
2675         * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
2676         separately.
2677         (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
2678         (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
2679         * except.h (verify_eh_throw_table_statements): Kill prototype.
2680         * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
2681         into call exrepssions.
2682         * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
2683         eh trhow verifier.
2684         * tree-cfg.c: Include pointer-set.h.
2685         (verify_node_sharing): Work on pointer set.
2686         (verify_eh_throw_stmt_node): New.
2687         (verify_stmts): Use pointers sets, verify throw_stmt.
2688
2689 2007-01-13  Zdenek Dvorak <dvorakz@suse.cz>
2690
2691         * ipa-reference.c (analyze_function): Consider also addresses taken
2692         in phi nodes.
2693
2694 2007-01-12  Roger Sayle  <roger@eyesopen.com>
2695
2696         * c-typeck.c (null_pointer_constant_p): Replace use of
2697         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2698         (build_c_cast): Likewise.
2699
2700 2007-01-12  Roger Sayle  <roger@eyesopen.com>
2701
2702         * tree.h (force_fit_type_double): Remove unused final argument.
2703         * c-common.c (constant_expression_warning): Replace use of
2704         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2705         (convert_and_check): Likewise.
2706         (shorten_compare): Update call to force_fit_type_double.
2707         (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
2708         * convert.c (convert_to_pointer): Update call to
2709         force_fit_type_double.
2710         * fold-const.c (force_fit_type_double): Remove overflowed_const
2711         argument.
2712         (int_const_binop, fold_convert_const_int_from_int,
2713         fold_convert_const_int_from_real, fold_div_compare,
2714         fold_sign_changed_comparison, fold_unary, fold_negate_const,
2715         fold_abs_const, fold_not_const): Remove the final argument from
2716         calls to force_fit_type_double.
2717
2718 2007-01-12  Andrew Pinski  <andrew_pinski@playstation.sony.com>
2719
2720         * configure.ac: Set insn to "nop" for spu-*-* also.
2721         * configure: Regenerate.
2722
2723 2007-01-12  Olga Golovanevsky  <olga@il.ibm.com>
2724
2725         * builtins.def : Add BUILT_IN_FREE.
2726
2727 2007-01-12  Jan Hubicka  <jh@suse.cz>
2728
2729         PR tree-optimization/30443
2730         * tree-inline.c (tree_function_versioning): Do not optimize when
2731         cloning for inlining.
2732
2733 2007-01-12  Zdenek Dvorak <dvorakz@suse.cz>
2734
2735         * doc/loop.texi: Document recording of loop exits.
2736         * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
2737         (update_single_exits_after_duplication,
2738         update_single_exit_for_duplicated_loop,
2739         update_single_exit_for_duplicated_loops): Removed.
2740         (duplicate_loop_to_header_edge): Do not call
2741         update_single_exits_after_duplication and
2742         update_single_exit_for_duplicated_loops.
2743         (loop_version): Do not update single_exit information.
2744         (fix_loop_structure): Use record_loop_exits instead of
2745         mark_single_exit_loops.
2746         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
2747         the lists of loop exits.
2748         * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
2749         split_edge, merge_blocks): Update the lists of loop exits.
2750         * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
2751         loop_optimizer_init.
2752         * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
2753         of mark_single_exit_loops.
2754         (loop_optimizer_finalize): Call release_recorded_exits.
2755         * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
2756         LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
2757         * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
2758         update single exit information.
2759         * lambda-code.c (perfect_nestify): Ditto.
2760         * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
2761         (mark_single_exit_loops): Removed.
2762         (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
2763         get_exit_descriptions, rescan_loop_exit, record_loop_exits,
2764         dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
2765         functions.
2766         (get_loop_exit_edges, single_exit): Use recorded exit lists.
2767         (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
2768         (verify_loop_structure): Verify consistency of the exit lists.
2769         (flow_loops_find): Use alloc_loop.  Initialize exits hash.
2770         (set_single_exit): Removed.
2771         * cfgloop.h (struct loop_exit): New function.
2772         (struct loop): single_exit_ field replaced by exits field.
2773         (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
2774         (struct loops): Added exits hash.
2775         (mark_single_exit_loops, set_single_exit): Declaration removed.
2776         (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
2777
2778 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2779
2780         * doc/invoke.texi: Avoid use of @headitem.
2781
2782 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2783
2784         * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
2785
2786 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2787
2788         * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
2789         list of targets.
2790
2791 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
2792             Richard Sandiford  <richard@codesourcery.com>
2793
2794         * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
2795         * config/m68k/m68k.opt: Resort options.
2796         (mbitfield, mrtd, mshort): Remove RejectNegative properties.
2797
2798 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
2799             Richard Sandiford  <richard@codesourcery.com>
2800
2801         * doc/invoke.texi: Document the macros that are defined by
2802         m68k's -mtune and -mhard-float options.
2803         * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
2804         (m68k_cpp_cpu_family): Likewise.
2805         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
2806         of __ucfv*__ macros.  Define __mcffpu__ if generating code for
2807         ColdFire FPUs.  Define __mcf_cpu_* and __mcf_family_* macros.
2808         * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
2809         (m68k_cpp_cpu_family): Likewise.
2810
2811 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2812
2813         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
2814         macros besides mc68000 as tuning macros.  Use a switch statement
2815         to set them and mcpu32.
2816
2817 2007-01-12  Julian Brown  <julian@codesourcery.com>
2818
2819         * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
2820         (TARGET_68040_ONLY): Rename to...
2821         (TARGET_68040): ...this.
2822         * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
2823         * config/m68k/m68k.md: Likewise.
2824
2825 2007-01-12  Julian Brown  <julian@codesourcery.com>
2826             Nathan Sidwell  <nathan@codesourcery.com>
2827             Richard Sandiford  <richard@codesourcery.com>
2828
2829         * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
2830         the -mcpu= argument associated with the --with-cpu setting.
2831         Define M68K_DEFAULT_TUNE to the default -mtune= option,
2832         if different from the one implied by the -mcpu setting.
2833         Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
2834         using mcpu=FOO as the default CPU option.  Set target_cpu_default2.
2835         * doc/invoke.texi: Mention ColdFire in the introduction to the
2836         m68k options.  Document the new -march, -mcpu, -mtune, -mdiv,
2837         -mno-div and -mhard-float options.  Make -m68881 a synonym for
2838         -mhard-float.  Document the previously-undocumented -m5206e,
2839         -m528x, -m5307 and -m5407 options.  Tweak the existing option
2840         documentation for consistency.
2841         * doc/install.texi: Mention new --with-cpu arguments.
2842         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
2843         default CPU if neither -mcpu nor -march are specified.
2844         (ASM_CPU_SPEC): Pass down -mcpu and -march options.
2845         (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
2846         TARGET_ISA*.  Set the legacy __mcf*__ cpu macros in the same way,
2847         using m68k_tune to decide between families that implement the
2848         same ISA.  Use m68k_tune to set __mcfv4e__.
2849         (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
2850         (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
2851         (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
2852         (FL_ISA_C, FL_ISA_MMU): New macros.
2853         (MASK_COLDFIRE): Delete.
2854         (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
2855         (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
2856         (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
2857         (TARGET_HARD_FLOAT): Do not define here.
2858         (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
2859         (TUNE_68000): New macro.
2860         (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
2861         (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
2862         (TUNE_CFV2): Redefine in terms of m68k_tune.
2863         (uarch_type, target_device, fpu_type): New enums.
2864         (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
2865         * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
2866         (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
2867         (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
2868         (FL_FOR_isa_c): New macros.
2869         (m68k_isa): New enum.
2870         (m68k_target_selection): New structure.
2871         (all_devices, all_isas, all_microarchs): New tables.
2872         (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
2873         (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
2874         (MASK_ALL_CPU_BITS): Delete.
2875         (m68k_find_selection): New function.
2876         (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
2877         Map the legacy target options to a combination of the new ones.
2878         (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
2879         m68k_cpu_flags.  Handle M68K_DEFAULT_TUNE.  Use m68k_cpu_flags
2880         to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
2881         settings.
2882         * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
2883         (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
2884         properties.
2885         (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
2886         (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
2887         * config/m68k/m68k-devices.def: New file.
2888
2889 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2890             Nathan Sidwell  <nathan@codesourcery.com>
2891
2892         * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
2893         (SUBTARGET_EXTRA_SPECS): New macros.
2894         * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
2895         use %(asm_cpu_spec) instead.
2896         * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
2897         * config/m68k/openbsd.h (ASM_SPEC): Likewise.
2898         * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
2899         (EXTRA_SPECS): Rename to...
2900         (SUBTARGET_EXTRA_SPECS): ...this.
2901
2902 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
2903             Richard Sandiford  <richard@codesourcery.com>
2904             Julian Brown  <julian@codesourcery.com>
2905
2906         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2907         (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
2908         to the configuration's default CPU.
2909         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
2910         Remove default masks.
2911         (m680[012]0-*-*): Set the default with_cpu to the first part of
2912         the target name.
2913         (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
2914         (m68k*-*-linux): Extend the --with-cpu handling to...
2915         (m680[012]0-*-*, m68k*-*-*): ...these configurations.  Allow m68000
2916         and m68010.  Don't set target_cpu_default2.
2917         * doc/install.texi: Document --with-cpu for m68k.
2918         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
2919         * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
2920         (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
2921         (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
2922         (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
2923         (ASM_SPEC): Remove use of %(asm_cpu_default).
2924         (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
2925         * config/m68k/linux.h (TARGET_DEFAULT): Delete.
2926         (CPP_SPEC): Merge definitions.  Do not handle __HAVE_68881__ here.
2927         * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
2928         __HAVE_FPU__ if TARGET_HARD_FLOAT.
2929         (TARGET_DEFAULT): Delete.
2930         (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
2931         cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
2932         (CPP_CPU_SPEC): Delete.
2933         (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
2934         the appropriate string.
2935         (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
2936         (CPP_SPEC): Define to NETBSD_CPP_SPEC.
2937         (ASM_SPEC): Don't use %(asm_default_spec).
2938         * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
2939         TARGET_DEFAULT and add MASK_68881.
2940         * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
2941
2942 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2943
2944         * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
2945         (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
2946         MASK_68010 alongside MASK_68020.
2947         * doc/invoke.texi: Document -m68010.
2948         * config/m68k/m68k.opt (m68010): New.
2949         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
2950         if TUNE_68010.
2951         (TUNE_68010): New macro.
2952         * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
2953         (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
2954         (M68K_CPU_m68332): Add MASK_68010.
2955         * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
2956         fallback definition.
2957         * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
2958         defines.
2959         * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
2960         (m68k_handle_option): Handle OPT_m68010.  Add MASK_68010
2961         to all entries that use MASK_68020.
2962         (output_move_simode_const, output_move_himode, output_move_qimode)
2963         (output_move_stricthi, output_move_strictqi): Use TARGET_68010
2964         instead of TARGET_68020 to select clr behavior.  Remove comment
2965         about there being no TARGET_68010.
2966         * config/m68k/m68k.md: Likewise throughout.
2967
2968 2007-01-12  Julian Brown  <julian@codesourcery.com>
2969
2970         * config/m68k/m68k.h (TARGET_ISAB): New macro.
2971         * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
2972         * config/m68k/m68k.md: Likewise.
2973
2974 2007-01-12  Julian Brown  <julian@codesourcery.com>
2975
2976         * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
2977         TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
2978
2979 2007-01-12  Julian Brown  <julian@codesourcery.com>
2980
2981         * config/m68k/m68k.h (TUNE_68040_60): New macro.
2982         * config/m68k/m68k.c (standard_68881_constant_p): Use it.
2983         * config/m68k/m68k.md: Likewise.
2984
2985 2007-01-12  Julian Brown  <julian@codesourcery.com>
2986             Richard Sandiford  <richard@codesourcery.com>
2987
2988         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
2989         instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
2990         TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
2991         TARGET_CPU32.
2992         (TARGET_CPU32): Rename to...
2993         (TUNE_CPU32): ...this.
2994         (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
2995         (TUNE_CFV2): New macros.
2996         * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
2997         remove conditions that are implied by TARGET_68020.
2998         * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
2999         instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
3000         (m68k_output_function_epilogue): Likewise.
3001         (m68k_rtx_costs): Likewise.  Use TUNE_68060 instead of TARGET_68060
3002         and TUNE_CFV2 instead of TARGET_5200.  Use TUNE_68000_10 instead of
3003         "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
3004         non-68000 timings.  Refactor multiplication and division costs.
3005         (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
3006         TUNE_CPU32 instead of TARGET_CPU32.
3007         (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
3008         and TUNE_68060 instead of TARGET_68060.
3009         * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
3010         TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
3011         TARGET_CPU32.
3012         (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
3013         && !TARGET_COLDFIRE" to choose between moveq and clr.
3014         Likewise in the unnamed movsf pattern.
3015         (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
3016         "!TARGET_68020 && !TARGET_COLDFIRE".  Likewise the unnamed
3017         ashiftrt pattern.
3018
3019 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
3020
3021         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
3022         of tabbing before backslashes.
3023
3024 2007-01-11  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
3025
3026         * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
3027         * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
3028         (ASM_OUTPUT_INTERNAL_LABEL): Define.
3029
3030 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
3031
3032         * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
3033         find_interesting_uses_cond.
3034         (find_interesting_uses_cond): Use extract_cond_operands.
3035         (rewrite_use_compare): Use extract_cond_operands and
3036         force_gimple_operand_bsi.  Do not call update_stmt.
3037         (determine_use_iv_cost_condition): Use extract_cond_operands.
3038         Return cheaper of using original bound and changing the exit bound.
3039
3040 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
3041
3042         PR tree-optimization/29516
3043         * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
3044         most_expensive_mult_to_index, addr_to_parts,
3045         create_mem_ref, maybe_fold_tmr): Make the type of
3046         fields of TARGET_MEM_REF sizetype.
3047         (move_fixed_address_to_symbol, move_pointer_to_base):
3048         New functions.
3049         * tree.def (TARGET_MEM_REF): Add comment on types of
3050         the operands.
3051
3052 2007-01-11  Joseph Myers  <joseph@codesourcery.com>
3053
3054         * c-common.c (vector_types_convertible_p): Treat opaque types as
3055         always convertible if they have the same size, but not otherwise.
3056
3057 2007-01-11  Steven Bosscher  <steven@gcc.gnu.org>
3058
3059         * ifcvt.c (struct noce_if_info): Add comments to the fields.
3060         Remove the b_unconditional field.
3061         (noce_try_sign_mask): Do not look at b_unconditional.
3062         (noce_process_if_block): Do not use merge_if_blocks.  Update
3063         the CFG here.  Do not set b_unconditional.
3064         (cond_move_process_if_block): Likewise.
3065         (find_cond_trap): Likewise.
3066         (check_cond_move_block): Require simple jump insns at the end
3067         of the basic block.
3068
3069 2007-01-11  Jan Hubicka  <jh@suse.cz>
3070
3071         PR tree-optimization/1046
3072         * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
3073         when alias info is not ready.
3074         (pass_tail_recursion): Do not require aliasing.
3075         * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
3076         * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
3077         * tree-ssa-copy.c (pass_copy_prop): Likewise.
3078         * tree-ssa-forwprop.c (pass_forwprop): Likewise.
3079         * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
3080         * passes.c (init_optimization_passes): Execute rename_ssa_copies,
3081         ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
3082         before inlining.
3083         * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
3084         When aliasing is not build, mark statement as volatile.
3085
3086 2007-01-11  Tom Tromey  <tromey@redhat.com>
3087
3088         PR preprocessor/15185, PR preprocessor/20989:
3089         * doc/cppopts.texi <-MT>: Update description of algorithm for
3090         computing default target.
3091         <-M, -MD>: Reword "basename" text.
3092
3093 2007-01-11  Roger Sayle  <roger@eyesopen.com>
3094
3095         * builtins.c (expand_builtin_pow, expand_builtin_powi,
3096         fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
3097         fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
3098         fold_builtin_int_int_roundingfn, fold_builtin_bitop,
3099         fold_builtin_bswap, real_constp, fold_builtin_pow,
3100         fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
3101         do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
3102         uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3103         * convert.c (convert_to_pointer): Likewise.
3104         * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
3105         * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
3106         const_binop, fold_convert_const_int_from_int,
3107         fold_convert_const_int_from_real,
3108         fold_convert_const_real_from_real, sign_bit_p,
3109         optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
3110         fold_sign_changed_comparison, fold_unary, fold_comparison,
3111         fold_binary, multiple_of_p, tree_Expr_non_zero_p,
3112         fold_negate_const, fold_abs_const, fold_not_const): Likewise.
3113         * print-tree.c (print_node_brief, print_node): Likewise.
3114         * stor-layout.c (place_field, layout_type): Likewise.
3115         * tree-chrec.c (keep_cast): Likewise.
3116         * tree.c (build_vector, build_real, build_real_from_int_cst,
3117         build_complex): Likewise.
3118
3119 2007-01-11  Roger Sayle  <roger@eyesopen.com>
3120
3121         * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete.  For the time being,
3122         treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
3123
3124 2007-01-11  Paolo Bonzini  <bonzini@gnu.org>
3125
3126         * configure.ac (strict1_warn): Rename to strict_warn.
3127         (WERROR, --enable-werror, symlink hacks, stage1_cflags,
3128         cc_set_by_configure, quoted_cc_set_by_configure,
3129         stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
3130         all_boot_languages, all_stagestuff): Remove.
3131         (target_list): Remove bootstrap targets.
3132         * Makefile.in (quickstrap): Unconditionally make a synonym of all.
3133         (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
3134         REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
3135         SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
3136         VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
3137         STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
3138         stage1_copy, stage2_build, stage2_copy, stageprofile_build,
3139         stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
3140         stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
3141         bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
3142         bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
3143         unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
3144         restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
3145         cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
3146         *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
3147         stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
3148         stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
3149         risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
3150         (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
3151         (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
3152         (mostlyclean): Adjust.
3153         (clean, distclean): Don't mention bootstrap stuff.
3154         * configure: Regenerate.
3155         * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
3156         java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
3157         treelang/config-lang.in (stagestuff): Remove.
3158         * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
3159         stageprofile, stagefeedback, stagestuff): Remove mention.
3160
3161 2007-01-11  Nick Clifton  <nickc@redhat.com>
3162
3163         * config/mcore/predicates.md (mcore_general_movesrc_operand):
3164         Accept CONSTs.
3165         (mcore_general_movdst_operand): Do not accept CONST_INTs.
3166         (mcore_arith_K_S_operand): Run the test for the S constraint not
3167         the test for the M constraint.
3168         (mcore_addsub_operand): Do not accept integer values that are
3169         larger than 32 bits.
3170         * config/mcore/mcore.md: Remove unused constraints from split.
3171         (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
3172         (addsi3): Likewise.
3173         (allocate_stack): Likewise.
3174         * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
3175         operands to 32 bits.
3176         (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
3177         INTVAL.
3178         (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
3179         mcore_const_ok_for_inline, mcore_const_trick_uses_not,
3180         try_constant_tricks, mcore_num_ones, mcore_num_zeros,
3181         mcore_output_bclri, mcore_output_andn, output_inline_const,
3182         mcore_output_move, mcore_output_movedouble): Likewise.
3183         (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
3184         (output_inline_const): Likewise.
3185         (output_inline_const): Fix format strings used in sprintf
3186         statements.
3187         * config/mcore/mcore-protos.h: Update prototypes for changed
3188         functions in mcore.c.
3189         * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
3190         HOST_WIDE_INT and not int.
3191         (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
3192         CONST_OK_FOR_N): Likewise.
3193         (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
3194         (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
3195         an INTVAL.
3196
3197 2007-01-10  Jan Hubicka  <jh@suse.cz>
3198
3199         * tree-vrp.c (remove_range_assertions): Release defs.
3200         * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
3201         * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
3202
3203 2007-01-10  Paul Brook  <paul@codesourcery.com>
3204
3205         * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
3206         * config/arm/arm.md (mulsi3subsi): New insn.
3207
3208 2007-01-10  Zdenek Dvorak <dvorakz@suse.cz>
3209
3210         * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
3211         (tree_transform_and_unroll_loop): New.
3212         * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
3213         Declare.
3214
3215 2007-01-10  Robert Kennedy <jimbob@google.com>
3216
3217         * fold-const.c (fold_comparison): Fold comparisons like (x *
3218         1000 < 0) to (x < 0).
3219
3220 2007-01-10  Ian Lance Taylor  <iant@google.com>
3221
3222         * tree-pretty-print.c (dump_generic_node): Print parentheses when
3223         operands have the same priority.
3224
3225 2007-01-10  Tom Tromey  <tromey@redhat.com>
3226
3227         * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
3228         (fold_binary): Likewise.
3229         * langhooks.c (lhd_can_use_bit_fields_p): Removed.
3230         * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
3231         (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
3232         (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
3233         * langhooks.h (struct lang_hooks): Removed field
3234         'can_use_bit_fields_p'.
3235
3236 2007-01-10  Ralf Cors├ępius <ralf.corsepius@rtems.org>
3237
3238         * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
3239
3240 2007-01-10  Razya Ladelsky  <razya@il.ibm.com>
3241
3242         * function.c (get_last_funcdef_no): New function.
3243         * function.h (get_last_funcdef_no): Declare.
3244         * tree-inline.c (initialize_cfun): Add initialization.
3245         (tree_function_versioning): Cleanup.
3246
3247 2007-01-10  Jan Hubicka  <jh@suse.cz>
3248
3249         * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
3250         PHIs.
3251
3252 2007-01-10  Sa Liu  <saliu@de.ibm.com>
3253             Ben Elliston  <bje@au.ibm.com>
3254
3255         * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
3256         for SAVE_NONLOCAL and Pmode for any other save level.
3257         * spu-protos.h (spu_restore_stack_block): Declare.
3258         * spu.md (save_stack_block): Remove.
3259         (restore_stack_block): Call spu_restore_stack_block.
3260         * spu.c (spu_restore_stack_block): New function.
3261         (spu_expand_epilogue): Remove old comment.
3262
3263 2007-01-09  Zdenek Dvorak <dvorakz@suse.cz>
3264
3265         PR tree-optimization/30322
3266         * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
3267         (cand_value_at): Return the value as aff_tree.
3268         (may_eliminate_iv): Convert the bound from aff_tree to tree.
3269         * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
3270         aff_combination_mult): New functions.
3271         (aff_combination_add): Use aff_combination_add_cst.
3272         (aff_combination_convert): Allow conversions to a wider type.
3273         (tree_to_aff_combination): Handle BIT_NOT_EXPR.
3274         * tree-affine.h (aff_combination_mult): Declare.
3275
3276 2007-01-09  Carlos O'Donell  <carlos@codesourcery.com>
3277
3278         * doc/tm.texi: Update documentation to reflect reality of exec
3279         and start file search behaviours. Update copyright year.
3280         * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
3281         header file directories.
3282
3283 2007-01-09  Uros Bizjak  <ubizjak@gmail.com>
3284
3285         * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
3286         (*cosxf2): Rename to cosxf2_i387.
3287         (*sindf2, *sinsf2): Extend operand 1 to XFmode.  Macroize patterns
3288         using X87MODEF12 mode macro. Rename patterns to
3289         *sin_extend<mode>xf2_i387.  Use SSE_FLOAT_MODE_P to disable patterns
3290         for SSE math.
3291         (*cosdf2, *cossf2): Ditto.
3292         (sincosdf3, sincossf3): Ditto.  Rewrite corresponding splitters
3293         to match extended input operands.
3294         (sincos<mode>3): New expander.
3295         (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
3296         insn patterns and corresponding splitters.
3297
3298 2007-01-09  Kaz Kojima  <kkojima@gcc.gnu.org>
3299
3300         * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
3301         (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
3302
3303 2007-01-09  Nicolas Pitre  <nico@cam.org>
3304
3305         PR target/30173
3306         * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
3307
3308 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
3309
3310         * target.h (struct gcc_target): New field library_rtti_comdat.
3311         * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
3312         (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
3313         * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
3314         * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
3315
3316 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
3317
3318         * doc/invoke.texi (Optimize Options): Correct description of -O0.
3319
3320 2007-01-08  Richard Guenther  <rguenther@suse.de>
3321
3322         * tree.h (force_fit_type_double): Export.
3323         (force_fit_type): Remove.
3324         * fold-const.c (force_fit_type_double): New function.
3325         (force_fit_type): Remove.
3326         (int_const_binop): Use it.
3327         (fold_convert_const_int_from_int): Likewise.
3328         (fold_convert_const_int_from_real): Likewise.
3329         (fold_div_compare): Likewise.
3330         (fold_sign_changed_comparison): Likewise.
3331         (fold_unary): Likewise.
3332         (fold_negate_const): Likewise.
3333         (fold_abs_const): Likewise.
3334         (fold_not_const): Likewise.
3335         * c-common.c (shorten_compare): Use force_fit_type_double.
3336         * convert.c (convert_to_pointer): Likewise.
3337
3338 2007-01-08  Richard Guenther  <rguenther@suse.de>
3339
3340         * tree.h (build_int_cst_wide_type): Export.
3341         * tree.c (build_int_cst_wide_type): New function.
3342         (build_int_cst_wide): Fix comment.
3343         * builtins.c (fold_builtin_object_size): Use build_int_cst
3344         to build -1 or 0 of the correct type.  Use fit_double_type
3345         to check for overflow.
3346         * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
3347         to build the mask.
3348         (decode_field_reference): Likewise.
3349         (all_ones_mask_p): Likewise.
3350         (native_interpret_int): Use build_int_cst_wide_type.
3351         (fold_binary): Use build_int_cst_type to build an all-ones
3352         value.
3353         * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
3354
3355 2007-01-08  Daniel Jacobowitz  <dan@codesourcery.com>
3356
3357         * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
3358
3359 2007-01-09  Ben Elliston  <bje@au.ibm.com>
3360
3361         * genautomata.c (STATS_OPTION): New option.
3362         (stats_flag): New flag.
3363         (gen_automata_option): Handle it.
3364         (initiate_automaton_gen): Ditto.
3365         (write_automata): Output statistics only if stats_flag is
3366         set. Likewise, output time statistics only if time_flag is set.
3367         * doc/md.texi (Processor pipeline description): Document new flag.
3368
3369 2007-01-08  Richard Guenther  <rguenther@suse.de>
3370
3371         * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
3372         * tree.c (build_int_cst_type): Likewise.
3373         (size_in_bytes): Don't call force_fit_type on the result.
3374         (int_fits_type_p): Use fit_double_type.
3375         * fold-const.c (fit_double_type): New function.
3376         (force_fit_type): Use it.
3377         * tree.h (fit_double_type): Export.
3378
3379 2007-01-08  Jan Hubicka  <jh@suse.cz>
3380
3381         * tree-vectorizer.c (gate_increase_alignment): Fix return type.
3382         * ipa.c (function_and_variable_visibility): Fix return type.
3383
3384 2007-01-08  Richard Guenther  <rguenther@suse.de>
3385
3386         * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
3387         of offset to build the index.
3388         * tree-pretty-print.c (dump_generic_node): Don't build negated
3389         const just for printing.
3390         * c-pretty-print.c (pp_c_integer_constant): Likewise.
3391         * builtins.c (fold_builtin_int_roundingfn): Check if result
3392         fits the type by using force_fit_type and comparing the result.
3393         * predict.c (predict_loops): Use compare_tree_int for comparison.
3394         * tree.c (build_int_cst): Fall back to integer_type_node for
3395         NULL_TREE type.
3396         (build_int_cst_wide): Assert type is non-null.
3397
3398 2007-01-08  Roberto Costa  <roberto.costa@st.com>
3399
3400         * tree-vrp.c (extract_range_from_cond_expr): New.
3401         (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
3402         * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
3403         as expressions.
3404         (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
3405         * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
3406         tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
3407         used as expressions.
3408         * tree-object-size.c (cond_expr_object_size): New.
3409         (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
3410
3411 2007-01-08  Jan Hubicka  <jh@suse.cz>
3412
3413         * tree-ssa-forwprop.c (forward_propagate_into_cond,
3414         tree_ssa_forward_propagate_single_use_va): Release defs of propagated
3415         statement.
3416
3417 2007-01-08  Richard Guenther  <rguenther@suse.de>
3418
3419         PR tree-optimization/23603
3420         * tree-vrp.c (set_value_range_to_truthvalue): New function.
3421         (extract_range_from_binary): Fall back to truthvalue instead of
3422         varying for TRUTH_*_EXPR.
3423         (extract_range_from_comparison): Fall back to truthvalue instead of
3424         varying.
3425         (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
3426         if all visited PHI values were constant.
3427
3428 2007-01-08  Jan Hubicka  <jh@suse.cz>
3429
3430         * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
3431         (cgraph_analyze_function): break out from ...
3432         (cgraph_finalize_compilation_unit): ... here.
3433         (cgraph_expand_function): Remove forgoten commented out line.
3434         (cgraph_optimize): Analyze functions.
3435
3436 2007-01-08  Jan Hubicka  <jh@suse.cz>
3437
3438         * tree-pas.h (TODO_remove_function): New flag.
3439         (TODO_update*): Renumber.
3440         (pass_ipa_increase_alignment,
3441         pass_ipa_function_and_variable_visibility): New passes.
3442         * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
3443         (cgraph_function_and_variable_visibility): Move to ipa.c
3444         (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
3445         cgraph_increase_alignment.
3446         * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
3447         (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
3448         calling tree-inline.
3449         (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
3450         (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
3451         * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
3452         (gate_increase_alignment): New function.
3453         (pass_ipa_increase_alignment): New pass.
3454         * ipa.c: Inline tree-pass.h and timevar.h
3455         (function_and_variable_visibility): Move here from cgraphunit.c
3456         * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
3457         * passes.c (init_optimization_passes): Add the two new passes.
3458         (execute_todo): Handle cgraph_remove_functions.
3459
3460 2007-01-08  Nick Clifton  <nickc@redhat.com>
3461
3462         * config/frv/predicates.md (reg_or_0_operand): Accept
3463         CONST_DOUBLEs.
3464
3465 2007-01-08  Ralf Cors├ępius <ralf.corsepius@rtems.org>
3466
3467         * config/bfin/rtems.h, config/bfin/t-rtems: New.
3468         * config.gcc: Add bfin*-rtems*.
3469
3470 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
3471
3472         * c.opt: Add -flax-vector-conversions.
3473         * c-typeck.c (convert_for_assignment): Pass flag to
3474         vector_types_convertible_p to allow emission of note.
3475         (digest_init): Likewise.
3476         * c-opts.c: Handle -flax-vector-conversions.
3477         * c-common.c (flag_lax_vector_conversions): New.
3478         (vector_types_convertible_p): Unless -flax-vector conversions
3479         has been passed, disallow conversions between vectors with
3480         differing numbers of subparts and/or element types.  If such
3481         a conversion is disallowed, possibly emit a note on the first
3482         occasion only to inform the user of -flax-vector-conversions.
3483         The new last argument specifies this.
3484         * c-common.h (flag_lax_vector_conversions): New.
3485         (vector_types_convertible_p): Add extra argument.
3486         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
3487         char_type_node for V*QI type vectors.
3488         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
3489         Update to satisfy new typechecking rules.
3490         * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
3491         C and C++ variants.
3492         * doc/invoke.texi (C Dialect Options): Document
3493         -flax-vector-conversions.
3494
3495 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
3496
3497         PR tree-optimization/29877
3498         * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
3499         a register variable on the RHS to not be replaceable.
3500
3501 2007-01-08  Chen Liqin  <liqin@sunnorth.com.cn>
3502         * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
3503         * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
3504         const_simm15, const_pow2, const_npow2): Added.
3505         * config/score/misc.md (insv, extv, extzv, movmemsi,
3506         move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
3507         * config/score/score.c (score_address_cost, score_select_cc_mode):
3508         Added.
3509         Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
3510         Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
3511         Update score_print_operand.
3512         * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
3513         Adjust register allocate order and update some macro define.
3514         * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
3515         mdx_block_move_straight, mdx_block_move_loop_head,
3516         mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
3517         mdx_block_move): Added.
3518         (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
3519         mdp_limm): Updated and fix some bug and typo.
3520         * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
3521         (movsf, movdf, doloop_end): Added.
3522
3523 2007-01-08  Kazu Hirata  <kazu@codesourcery.com>
3524
3525         * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
3526         config/arm/thumb2.md: Fix comment typos.
3527         * doc/extend.texi: Fix a typo.
3528
3529 2007-01-07  Eric Christopher  <echristo@apple.com>
3530
3531         * configure.ac: Check for __stack_chk_fail for darwin.
3532         * configure: Regenerate.
3533
3534 2007-01-07  Richard Guenther  <rguenther@suse.de>
3535
3536         * tree-vrp.c (extract_range_from_assert): CSE calls to
3537         compare_values where possible.
3538         (extract_range_from_unary_expr): Likewise.
3539
3540 2007-01-07  Anatoly Sokolov <aesok@post.ru>
3541
3542         * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
3543         * config/avr/avr.c (call_insn_operand): Delete function.
3544         * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
3545         instead of register number. Use predicates.
3546         * config/avr/predicates.md (const0_operand, reg_or_0_operand,
3547         call_insn_operand): Add.
3548
3549 2007-01-06  Jan Hubicka  <jh@suse.cz>
3550
3551         * tree-pass.h (pass_build_cgraph_edges): Declare.
3552         * cgraphunit.c (record_refernece): Move to cgraphbuild.c
3553         (visited_nodes): Remove.
3554         (cgraph_create_edges): Move to cgraphbuild.c; rename to
3555         build_cgrpah_edges; make visited_nodes local.
3556         (cgraph_process_new_functions): DO not call initialize_inline_failed.
3557         (record_references_in_initializer): Move to cgraphbuild.c
3558         (initialize_inline_failed, rebuild_cgraph_edges,
3559         pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
3560         (verify_cgraph_node): Make visited_nodes local.
3561         (cgraph_analyze_function): Do not call cgraph_create_edges and
3562         initialize_inline_failed.
3563         (cgraph_expand_function): Do not call cgraph_lower_function;
3564         assert that function is already lowered.
3565         * Makefile.in (cgraphbuild.o): New.
3566         * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
3567         at the end of lowering passes.
3568
3569 2007-01-06  Steven Bosscher  <steven@gcc.gnu.org>
3570
3571         * ifcvt.c (cond_move_convert_if_block): New function, code
3572         factored out from...
3573         (cond_move_process_if_block): ...here.  Call the new function
3574         on the THEN and ELSE blocks.
3575         (merge_if_block): Do not copy global_live_at_end, merge_blocks
3576         already takes care of this.
3577
3578 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
3579
3580         PR c/19978
3581         * tree.h (TREE_OVERFLOW_P): New.
3582         * c-typeck.c (parser_build_unary_op): Warn only if result
3583         overflowed and operands did not.
3584         (parser_build_binary_op): Likewise.
3585         (convert_for_assignment): Remove redundant overflow_warning.
3586         * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
3587
3588 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
3589
3590         * c-typeck.c (store_init_value): Split over two lines to follow
3591         the GNU coding style.
3592
3593 2007-01-05  Benjamin Kosnik  <bkoz@redhat.com>
3594
3595         * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
3596         __GXX_EXPERIMENTAL_CXX0X__.
3597         * doc/cpp.texi: Same.
3598
3599 2007-01-05  Richard Guenther  <rguenther@suse.de>
3600
3601         PR middle-end/27826
3602         * tree.c (get_narrower): Do not construct COMPONENT_REFs
3603         with mismatched types.  Instead explicitly build a
3604         conversion NOP_EXPR.
3605
3606 2007-01-05  Ian Lance Taylor  <iant@google.com>
3607
3608         * c-common.c (decl_with_nonnull_addr_p): New function.
3609         (c_common_truthvalue_conversion): Call it.
3610         * c-typeck.c (build_binary_op): Likewise.
3611         * c-common.h (decl_with_nonnull_addr_p): Declare.
3612
3613 2007-01-05  Jakub Jelinek  <jakub@redhat.com>
3614
3615         PR c/30360
3616         * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
3617         denom against 0.0.
3618
3619 2007-01-05  Joel Brobecker  <brobecker@adacore.com>
3620
3621         * doc/install.texi (Final install): Document the fact that
3622         the GNAT runtime should not be stripped.
3623
3624 2007-01-04  Jan Hubicka  <jh@suse.cz>
3625
3626         * tree-inline.c (fold_marked_statements): Update operand caches
3627         and EH after folding
3628
3629 2007-01-04  Ian Lance Taylor  <iant@google.com>
3630
3631         * c-common.c (check_function_nonnull): Whitespace fix.
3632
3633 2007-01-04  Jan Hubicka  <jh@suse.cz>
3634
3635         * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
3636         comitted older version of patch.
3637         (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
3638         function, celanup cfg and collect garbage.
3639
3640 2007-01-04  Mike Stump  <mrs@apple.com>
3641
3642         * Makefile.in (mostlyclean): Don't remove libgcc anymore.
3643         (clean): Likewise.
3644
3645 2007-01-04  Eric Christopher  <echristo@apple.com>
3646
3647         * libgcc2.c (__bswapsi2): Use SItype.
3648         (__bswapdi2): Use DItype.
3649         * libgcc2.h: Update for above.
3650
3651 2007-01-04  Paul Brook  <paul@codesourcery.com>
3652
3653         * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
3654         mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
3655         mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
3656         umulsidi3, umulsidi3adddi, smulsi3_highpart,
3657         umulsi3_highpart): Make conditional on !arm_arch6.
3658         (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
3659         mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
3660         mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
3661         umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
3662         umulsi3_highpart_v6): New insns.
3663
3664 2007-01-04  Roger Sayle  <roger@eyesopen.com>
3665
3666         * fold-const.c (fold_convert): When casting an expression to void,
3667         fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
3668         have a type.  Instead of attempting to build a NOP_EXPR, return
3669         these "special" trees directly.
3670
3671 2007-01-04  Joseph Myers  <joseph@codesourcery.com>
3672
3673         * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
3674         MULT inside MINUS as either argument.  Use rs6000_cost->dmul -
3675         rs6000_cost->fp not 0 as adjustment for outer NEG.
3676
3677 2007-01-04  Jan Hubicka  <jh@suse.cz>
3678
3679         * cgraph.c (cgraph_release_function_body): New function.
3680         (cgraph_remove_node): Use it.
3681         * cgraph.h (cgraph_release_function_body): Declare.
3682         * cgraphunit.c (cgraph_expand_function): Use it.
3683         * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
3684         * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
3685         is initialized and while compilation of other function is running.
3686         * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
3687         statement CFG annotations from here to ...
3688         * tree-cfg.c (delete_tree_cfg_annotations): ... here.
3689
3690 2007-01-04  Zdenek Dvorak <dvorakz@suse.cz>
3691
3692         * cfgloop.h (enum li_flags): Make the constants powers of two.
3693
3694 2007-01-04  Jan Hubicka  <jh@suse.cz>
3695
3696         * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
3697         set.
3698         (fold_marked_statements): New function.
3699         (optimize_inline_calls, tree_function_versioning): Fold new statements.
3700         * tree-inline.h (copy_body_data): Add statements_to_fold.
3701
3702 2007-01-03  Daniel Jacobowitz  <dan@codesourcery.com>
3703
3704         * config.gcc: Mention libgcc/config.host.
3705         * Makefile.in: Update comments mentioning libgcc.
3706         (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
3707         (all.cross, start.encap, rest.encap, rest.cross): Update
3708         dependencies for libgcc move.
3709         (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
3710         (clean-target, clean-target-libgcc): Delete.
3711         (srcdirify, GCC_EXTRA_PARTS): New macros.
3712         (libgcc-support, libgcc.mvars): New rules.
3713         (distclean): Remove mention of mklibgcc.
3714         (install): Don't reference INSTALL_LIBGCC.
3715         (install-common): Don't reference EXTRA_PARTS.
3716         (install-libgcc, install-multilib): Delete rules.
3717         * mklibgcc.in: Delete file.
3718         * doc/configfiles.texi: Don't mention mklibgcc.
3719
3720         * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
3721         * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
3722         * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
3723         * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
3724         (mklibgcc, ldblspecs): Likewise.
3725
3726         * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
3727         (posixpre.def): Use $(T).
3728         (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
3729         (SHLIB_SRC, SHLIB_INSTALL): Delete.
3730         (SHLIB_LINK): Make dummy.
3731         * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
3732
3733         * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
3734
3735         * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
3736         * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
3737         config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
3738         config/rs6000/t-newas, config/sparc/t-elf: Likewise.
3739
3740         * configure.ac (all_outputs): Remove mklibgcc.
3741         * configure: Regenerated.
3742
3743 2007-01-03  Josh Conner  <jconner@apple.com>
3744
3745         PR middle-end/29683
3746         * calls.c (compute_argument_addresses): Set stack and stack_slot
3747         for partial args, too.
3748         (store_one_arg): Use locate.size.constant for the size when
3749         generating a save_area.
3750
3751 2007-01-03  Robert Kennedy <jimbob@google.com>
3752
3753         * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
3754         whose definitions are deleted due to basic block merging.
3755
3756 2007-01-03  Paul Brook  <paul@codesourcery.com>
3757
3758         PR target/16634
3759         * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
3760         functions.
3761         (use_return_insn): Return 0 for Thumb interrupt functions.
3762         (print_multi_reg): Add rfe argument for IRQ returns.
3763         (arm_output_epilogue): Pop interrupt return address directly into PC.
3764         (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
3765
3766 2007-01-03  Paul Brook  <paul@codesourcery.com>
3767
3768         Merge from sourcerygxx-4_1.
3769         * config/arm/thumb2.md: New file.
3770         * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
3771         Thumb-2.
3772         * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
3773         * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
3774         (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
3775         * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
3776         tables.
3777         (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
3778         * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
3779         comptibility.
3780         * config/arm/ieee754-sf.S: Ditto.
3781         * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
3782         (thumb1_base_register_rtx_p): ... to this.
3783         (thumb_index_register_rtx_p): Rename...
3784         (thumb1_index_register_rtx_p): ... to this.
3785         (thumb_output_function_prologue): Rename...
3786         (thumb1_output_function_prologue): ... to this.
3787         (thumb_legitimate_address_p): Rename...
3788         (thumb1_legitimate_address_p): ... to this.
3789         (thumb_rtx_costs): Rename...
3790         (thumb1_rtx_costs): ... to this.
3791         (thumb_compute_save_reg_mask): Rename...
3792         (thumb1_compute_save_reg_mask): ... to this.
3793         (thumb_final_prescan_insn): Rename...
3794         (thumb1_final_prescan_insn): ... to this.
3795         (thumb_expand_epilogue): Rename...
3796         (thumb1_expand_epilogue): ... to this.
3797         (arm_unwind_emit_stm): Rename...
3798         (arm_unwind_emit_sequence): ... to this.
3799         (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
3800         thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
3801         thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
3802         arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
3803         thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
3804         thumb2_output_casesi): New functions.
3805         (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
3806         (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
3807         FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
3808         THUMB2_WORK_REGS): Define.
3809         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
3810         arm_condexec_mask, arm_condexec_masklen)): New variables.
3811         (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
3812         (arm_override_options): Check new CPU capabilities.
3813         Set new architecture flag variables.
3814         (arm_isr_value): Handle v7m interrupt functions.
3815         (user_return_insn): Return 0 for v7m interrupt functions.  Handle
3816         Thumb-2.
3817         (const_ok_for_arm): Handle Thumb-2 constants.
3818         (arm_gen_constant): Ditto.  Use movw when available.
3819         (arm_function_ok_for_sibcall): Return false for v7m interrupt
3820         functions.
3821         (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
3822         (thumb_find_work_register, arm_load_pic_register,
3823         legitimize_tls_address, arm_address_cost, load_multiple_sequence,
3824         emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
3825         print_multi_reg, output_mov_long_double_fpa_from_arm,
3826         output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
3827         output_mov_double_fpa_from_arm, output_move_double,
3828         arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
3829         output_return_instruction, arm_output_function_prologue,
3830         arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
3831         arm_output_mi_thunk, thumb_set_return_address): Ditto.
3832         (arm_expand_prologue): Handle Thumb-2.  Use arm_save_coproc_regs.
3833         (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
3834         (arithmetic_instr, shift_op): Use arm_shift_nmem.
3835         (arm_print_operand): Use arm_print_condition.  Handle '(', ')', '.',
3836         '!' and 'L'.
3837         (arm_final_prescan_insn): Use extract_constrain_insn_cached.
3838         (thumb_expand_prologue): Use thumb_set_frame_pointer.
3839         (arm_file_start): Output directive for unified syntax.
3840         (arm_unwind_emit_set): Handle stack alignment instruction.
3841         * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
3842         Add v6t2, v7, v7a, v7r and v7m.
3843         (RETLDM): Add Thumb-2 code.
3844         (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
3845         * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
3846         (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
3847         TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
3848         THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
3849         ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
3850         THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
3851         CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
3852         ADJUST_INSN_LENGTH): Define.
3853         (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
3854         STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
3855         BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
3856         PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
3857         SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
3858         TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
3859         HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
3860         HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
3861         REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
3862         REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
3863         LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
3864         GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
3865         ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
3866         FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
3867         PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
3868         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
3869         * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
3870         cortex-m3.
3871         * config/arm/arm-tune.md: Regenerate.
3872         * config/arm/arm-protos.h: Update prototypes.
3873         * config/arm/vfp.md: Enable patterns for Thumb-2.
3874         (arm_movsi_vfp): Add movw alternative.  Use output_move_vfp.
3875         (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
3876         (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
3877         thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
3878         * config/arm/libunwind.S: Add Thumb-2 code.
3879         * config/arm/constraints.md: Update include Thumb-2.
3880         * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
3881         * config/arm/ieee754-df.S: Ditto.
3882         * config/arm/bpabi.S: Ditto.
3883         * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
3884         * config/arm/predicates.md (low_register_operand,
3885         low_reg_or_int_operand, thumb_16bit_operator): New.
3886         (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
3887         (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
3888         * config/arm/t-arm-elf: Add armv7 multilib.
3889         * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
3890         Include thumb2.md.
3891         (UNSPEC_STACK_ALIGN, ce_count): New.
3892         (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
3893         arm_zero_extendsidi2, arm_zero_extendqidi2): New
3894         insns/expanders.
3895         * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
3896         (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
3897         thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
3898         * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
3899         (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
3900         thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
3901         insns.
3902         * doc/extend.texi: Document ARMv7-M interrupt functions.
3903         * doc/invoke.texi: Document Thumb-2 new cores+architectures.
3904
3905 2007-01-03  Jakub Jelinek  <jakub@redhat.com>
3906
3907         * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
3908         (struct _Unwind_Context): Rename args_size to flags, remove
3909         signal_frame field, add a new args_size field and version field.
3910         (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
3911         _Unwind_IsExtendedContext): New inline functions.
3912         (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
3913         Assume by_value array is only present if _Unwind_IsExtendedContext.
3914         (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
3915         _Unwind_IsSignalFrame.
3916         (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
3917         (uw_update_context_1): Use _Unwind_SetSignalFrame.
3918         (uw_init_context_1): Initialize context->flags to
3919         EXTENDED_CONTEXT_BIT.
3920         * config/rs6000/linux-unwind.h (frob_update_context): Use
3921         _Unwind_SetSignalFrame.
3922
3923 2007-01-03  Andrew Pinski  <andrew_pinski@playstation.sony.com>
3924
3925         PR middle-end/30353
3926         * gimplify.c (gimplify_modify_expr_complex_part): Move below
3927         tree_to_gimple_tuple.  Call tree_to_gimple_tuple when we need
3928         the value.
3929
3930 2007-01-03  Kazu Hirata  <kazu@codesourcery.com>
3931
3932         * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
3933         REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
3934         RET, AT_SP): Remove.
3935         * config/i386/i386.md (*sse_prologue_save_insn): Use return
3936         instead of RET.
3937
3938         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3939         * cfgbuild.c (find_basic_blocks): Likewise.
3940         * cfgrtl.c (rtl_create_basic_block): Likewise.
3941         * function.c (temp_slots_at_level): Likewise.
3942         * reg-stack.c (stack_regs_mentioned): Likewise.
3943         * regclass.c (allocate_reg_info): Likewise.
3944         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3945         set_bb_for_stmt, move_block_to_fn): Likewise.
3946         * tree-complex.c (tree_lower_complex): Likewise.
3947         * vec.h (VEC_safe_grow_cleared): New.