gcc/ChangeLog:
[official-gcc.git] / gcc / ChangeLog
blob01aa9dc59c117731801b6d38ade51a2e320eefdf
1 2007-02-05  Alexandre Oliva  <aoliva@redhat.com>
3         PR debug/30189
4         * dwarf2out.c (modified_type_die): Follow DECL_ORIGINAL_TYPE
5         even if cv-qualification is the same.
7 2007-02-05  Geoffrey Keating  <geoffk@apple.com>
9         * config/rs6000/darwin-tramp.asm (__trampoline_setup): Call
10         __enable_execute_stack on completion.
12 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
14         * config/i386/athlon.md (athlon_fldxf_k8, athlon_fld_k8,
15         athlon_fstxf_k8, athlon_fst_k8, athlon_fist, athlon_fmov,
16         athlon_fadd_load, athlon_fadd_load_k8, athlon_fadd, athlon_fmul,
17         athlon_fmul_load, athlon_fmul_load_k8, athlon_fsgn,
18         athlon_fdiv_load, athlon_fdiv_load_k8, athlon_fdiv_k8,
19         athlon_fpspc_load, athlon_fpspc, athlon_fcmov_load,
20         athlon_fcmov_load_k8, athlon_fcmov_k8, athlon_fcomi_load_k8,
21         athlon_fcomi, athlon_fcom_load_k8, athlon_fcom): Added amdfam10.
23 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
25         * config/i386/i386.md (x86_sahf_1, cmpfp_i_mixed, cmpfp_i_sse,
26         cmpfp_i_i387, cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387,
27         swapsi, swaphi_1, swapqi_1, swapdi_rex64, fix_truncsfdi_sse,
28         fix_truncdfdi_sse, fix_truncsfsi_sse, fix_truncdfsi_sse,
29         x86_fldcw_1, floatsisf2_mixed, floatsisf2_sse, floatdisf2_mixed,
30         floatdisf2_sse, floatsidf2_mixed, floatsidf2_sse,
31         floatdidf2_mixed, floatdidf2_sse, muldi3_1_rex64, mulsi3_1,
32         mulsi3_1_zext, mulhi3_1, mulqi3_1, umulqihi3_1, mulqihi3_insn,
33         umulditi3_insn, umulsidi3_insn, mulditi3_insn, mulsidi3_insn,
34         umuldi3_highpart_rex64, umulsi3_highpart_insn,
35         umulsi3_highpart_zext, smuldi3_highpart_rex64,
36         smulsi3_highpart_insn, smulsi3_highpart_zext, x86_64_shld,
37         x86_shld_1, x86_64_shrd, sqrtsf2_mixed, sqrtsf2_sse,
38         sqrtsf2_i387, sqrtdf2_mixed, sqrtdf2_sse, sqrtdf2_i387,
39         sqrtextendsfdf2_i387, sqrtxf2, sqrtextendsfxf2_i387,
40         sqrtextenddfxf2_i387): Added amdfam10_decode.
41         
42         * config/i386/athlon.md (athlon_idirect_amdfam10,
43         athlon_ivector_amdfam10, athlon_idirect_load_amdfam10,
44         athlon_ivector_load_amdfam10, athlon_idirect_both_amdfam10,
45         athlon_ivector_both_amdfam10, athlon_idirect_store_amdfam10,
46         athlon_ivector_store_amdfam10): New define_insn_reservation.
47         (athlon_idirect_loadmov, athlon_idirect_movstore): Added
48         amdfam10.
50 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
52         * config/i386/athlon.md (athlon_call_amdfam10,
53         athlon_pop_amdfam10, athlon_lea_amdfam10): New
54         define_insn_reservation.
55         (athlon_branch, athlon_push, athlon_leave_k8, athlon_imul_k8,
56         athlon_imul_k8_DI, athlon_imul_mem_k8, athlon_imul_mem_k8_DI,
57         athlon_idiv, athlon_idiv_mem, athlon_str): Added amdfam10.
59 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
61         * config/i386/athlon.md (athlon_sseld_amdfam10,
62         athlon_mmxld_amdfam10, athlon_ssest_amdfam10,
63         athlon_mmxssest_short_amdfam10): New define_insn_reservation.
65 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
67         * config/i386/athlon.md (athlon_sseins_amdfam10): New
68         define_insn_reservation.
69         * config/i386/i386.md (sseins): Added sseins to define_attr type
70         and define_attr unit.
71         * config/i386/sse.md: Set type attribute to sseins for insertq
72         and insertqi.
74 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
76         * config/i386/athlon.md (sselog_load_amdfam10, sselog_amdfam10,
77         ssecmpvector_load_amdfam10, ssecmpvector_amdfam10,
78         ssecomi_load_amdfam10, ssecomi_amdfam10,
79         sseaddvector_load_amdfam10, sseaddvector_amdfam10): New
80         define_insn_reservation.
81         (ssecmp_load_k8, ssecmp, sseadd_load_k8, seadd): Added amdfam10.
83 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
85         * config/i386/athlon.md (cvtss2sd_load_amdfam10,
86         cvtss2sd_amdfam10, cvtps2pd_load_amdfam10, cvtps2pd_amdfam10,
87         cvtsi2sd_load_amdfam10, cvtsi2ss_load_amdfam10,
88         cvtsi2sd_amdfam10, cvtsi2ss_amdfam10, cvtsd2ss_load_amdfam10,
89         cvtsd2ss_amdfam10, cvtpd2ps_load_amdfam10, cvtpd2ps_amdfam10,
90         cvtsX2si_load_amdfam10, cvtsX2si_amdfam10): New 
91         define_insn_reservation.
93         * config/i386/sse.md (cvtsi2ss, cvtsi2ssq, cvtss2si,
94         cvtss2siq, cvttss2si, cvttss2siq, cvtsi2sd, cvtsi2sdq,
95         cvtsd2si, cvtsd2siq, cvttsd2si, cvttsd2siq,
96         cvtpd2dq, cvttpd2dq, cvtsd2ss, cvtss2sd,
97         cvtpd2ps, cvtps2pd): Added amdfam10_decode attribute.
99 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
101         * config/i386/athlon.md (athlon_ssedivvector_amdfam10,
102         athlon_ssedivvector_load_amdfam10, athlon_ssemulvector_amdfam10,
103         athlon_ssemulvector_load_amdfam10): New define_insn_reservation.
104         (athlon_ssediv, athlon_ssediv_load_k8, athlon_ssemul,
105         athlon_ssemul_load_k8): Added amdfam10.
107 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
109         * config/i386/i386.h (TARGET_SSE_UNALIGNED_MOVE_OPTIMAL): New macro.
110         (x86_sse_unaligned_move_optimal): New variable.
111         
112         * config/i386/i386.c (x86_sse_unaligned_move_optimal): Enable for  
113         m_AMDFAM10.
114         (ix86_expand_vector_move_misalign): Add code to generate movupd/movups
115         for unaligned vector SSE double/single precision loads for AMDFAM10.
117 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
119         * config/i386/i386.h (TARGET_AMDFAM10): New macro.
120         (TARGET_CPU_CPP_BUILTINS): Add code for amdfam10.
121         Define TARGET_CPU_DEFAULT_amdfam10.
122         (TARGET_CPU_DEFAULT_NAMES): Add amdfam10.
123         (processor_type): Add PROCESSOR_AMDFAM10.       
124         
125         * config/i386/i386.md: Add amdfam10 as a new cpu attribute to match
126         processor_type in config/i386/i386.h.
127         Enable imul peepholes for TARGET_AMDFAM10.
128         
129         * config.gcc: Add support for --with-cpu option for amdfam10.
130         
131         * config/i386/i386.c (amdfam10_cost): New variable.
132         (m_AMDFAM10): New macro.
133         (m_ATHLON_K8_AMDFAM10): New macro.
134         (x86_use_leave, x86_push_memory, x86_movx, x86_unroll_strlen,
135         x86_cmove, x86_3dnow_a, x86_deep_branch, x86_use_simode_fiop,
136         x86_promote_QImode, x86_integer_DFmode_moves,
137         x86_partial_reg_dependency, x86_memory_mismatch_stall, 
138         x86_accumulate_outgoing_args, x86_arch_always_fancy_math_387,
139         x86_sse_partial_reg_dependency, x86_sse_typeless_stores,
140         x86_use_ffreep, x86_use_incdec, x86_four_jump_limit,
141         x86_schedule, x86_use_bt, x86_cmpxchg16b, x86_pad_returns):
142         Enable/disable for amdfam10.
143         (override_options): Add amdfam10_cost to processor_target_table.
144         Set up PROCESSOR_AMDFAM10 for amdfam10 entry in 
145         processor_alias_table.
146         (ix86_issue_rate): Add PROCESSOR_AMDFAM10.
147         (ix86_adjust_cost): Add code for amdfam10.
149 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
150         
151         * config/i386/i386.opt: Add new Advanced Bit Manipulation (-mabm)
152         instruction set feature flag. Add new (-mpopcnt) flag for popcnt 
153         instruction. Add new SSE4A (-msse4a) instruction set feature flag.
154         * config/i386/i386.h: Add builtin definition for SSE4A.
155         * config/i386/i386.md: Add support for ABM instructions 
156         (popcnt and lzcnt).
157         * config/i386/sse.md: Add support for SSE4A instructions
158         (movntss, movntsd, extrq, insertq).
159         * config/i386/i386.c: Add support for ABM and SSE4A builtins.
160         Add -march=amdfam10 flag.
161         * config/i386/ammintrin.h: Add support for SSE4A intrinsics.
162         * doc/invoke.texi: Add documentation on flags for sse4a, abm, popcnt
163         and amdfam10.
164         * doc/extend.texi: Add documentation for SSE4A builtins.
166 2007-02-05  Bob Wilson  <bob.wilson@acm.org>
168         * config/xtensa/xtensa.c (constantpool_mem_p): Skip over SUBREGs.
170 2007-02-05  Richard Guenther  <rguenther@suse.de>
172         * tree-vectorizer.h (vectorizable_function): Add argument type
173         argument, change return type.
174         * tree-vect-patterns.c (vect_recog_pow_pattern): Adjust caller.
175         * tree-vect-transform.c (vectorizable_function): Handle extra
176         argument, return vectorized function decl.
177         (build_vectorized_function_call): Remove.
178         (vectorizable_call): Handle calls with result and argument types
179         differing.  Handle loop vectorization factor correctly.
180         * targhooks.c (default_builtin_vectorized_function): Adjust for
181         extra argument.
182         * targhooks.h (default_builtin_vectorized_function): Likewise.
183         * target.h (builtin_vectorized_function): Add argument type
184         argument.
185         * config/i386/i386.c (ix86_builtin_vectorized_function): Handle
186         extra argument, allow vectorizing of lrintf.
187         * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION): Adjust
188         documentation of target hook.
190 2007-02-05  Hans-Peter Nilsson  <hp@axis.com>
192         PR target/30665
193         * config/cris/cris.md ("*andsi_movu", "*andsi_clear", "*andhi_movu")
194         ("*andhi_clear", andu (casesi+45)): For size-changed operand where
195         memory is allowed, require !side_effects_p, not just !MEM_VOLATILE_P.
197 2007-02-05  Roger Sayle  <roger@eyesopen.com>
199         * fold-const.c (fold_unary) <REAL_PART>: Test for availability of
200         BUILT_IN_COS before simplifying REAL_PART(CEXPI)) to COS.
201         <IMAG_PART>: Likewise, check for availability of BUILT_IN_SIN.
202         * builtins.c (fold_builtin_sincos): Check for TARGET_C99_FUNCTIONS
203         before canonicalizing sincos to cexpi.
204         (fold_builtin_cexp): Likewise, for canonicalizing cexp to cexpi.
206 2007-02-05  Roger Sayle  <roger@eyesopen.com>
208         * config/alpha/alpha.c (alpha_add_builtins): New Helper function.
209         Set TREE_READONLY and TREE_NOTHROW directly, not via attributes.
210         (alpha_init_builtins): Use alpha_add_builtins to process tables.
212 2007-02-05  Roger Sayle  <roger@eyesopen.com>
214         * mips-tfile.c (initialize_init_file): Correct endianness test.
216 2007-02-05  Kazu Hirata  <kazu@codesourcery.com>
218         * config/m68k/m68k.md (pushdi-1, pushdi, movsi+1): Don't use
219         the 'y' constraint.
221 2007-02-05  Richard Sandiford  <richard@codesourcery.com>
223         * dwarf2out.c (dwarf2out_frame_debug_expr): Record the register
224         saves in a PARALLEL before the register assignments.
226 2007-02-05  Richard Sandiford  <richard@codesourcery.com>
228         * doc/tm.texi (DWARF_ALT_FRAME_RETURN_COLUMN): Do not require
229         DWARF_FRAME_RETURN_COLUMN to be a general register.
230         * dwarf2out.c (init_return_column_size): New function, split from...
231         (expand_builtin_init_dwarf_reg_sizes): ...here.  Allow both
232         DWARF_FRAME_RETURN_COLUMN and DWARF_ALT_FRAME_RETURN_COLUMN
233         to be nongeneral registers.
234         * config/m68k/m68k.h (DWARF_FRAME_REGNUM): Only map FP and
235         integer registers.
236         (DWARF_FRAME_REGISTERS, DWARF_FRAME_RETURN_COLUMN): Define.
237         (DWARF_ALT_FRAME_RETURN_COLUMN): Define.
239 2007-02-04  Zdenek Dvorak <dvorakz@suse.cz>
241         * cfgcleanup.c (try_optimize_cfg): Avoid removing ENTRY_BLOCK_PTR.
243 2007-02-04  Zdenek Dvorak <dvorakz@suse.cz>
245         * cfgloopmanip.c (loop_delete_branch_edge): Removed.
246         (remove_path): Use can_remove_branch_p and remove_branch instead
247         of loop_delete_branch_edge.
248         * tree-ssa-loop-manip.c (scale_dominated_blocks_in_loop): New function.
249         (tree_transform_and_unroll_loop): Remove dead branches immediately.
250         Update profile using scale_dominated_blocks_in_loop.
251         * cfghooks.c (can_remove_branch_p, remove_branch): New functions.
252         * cfghooks.h (struct cfg_hooks): Add can_remove_branch_p.
253         (can_remove_branch_p, remove_branch): Declare.
254         * tree-cfg.c (tree_can_remove_branch_p): New function.
255         (tree_cfg_hooks): Add tree_can_remove_branch_p.
256         * cfgrtl.c (rtl_can_remove_branch_p): New function.
257         (rtl_cfg_hooks, cfg_layout_rtl_cfg_hook): Add rtl_can_remove_branch_p.
259 2007-02-05  Jan Hubicka  <jh@suse.cz>
261         PR middle-end/30696
262         * ipa-inline.c (cgraph_clone_inlined_nodes): When there are unanalyzed
263         nodes in cgraph, don't remove offline copy of the function.
265 2007-02-04  Jan Hubicka  <jh@suse.cz>
267         * tree-sra.c (sra_walk_expr): Add linebreaks. BITFIELD_REFs into
268         vectors might cause maybe_lookup_element_for_expr to be called
269         on non-sra-candidate.
271 2007-02-04  Kazu Hirata  <kazu@codesourcery.com>
273         * config/bfin/bfin-modes.def, config/bfin/bfin.c,
274         config/bfin/bfin.md, config/bfin/predicates.md: Follow
275         spelling conventions.
277 2007-02-04  Richard Guenther  <rguenther@suse.de>
279         PR middle-end/30636
280         * fold-const.c (try_move_mult_to_index): Make sure to not
281         overflow one dimension of a multi-dimensional array access.
283 2007-02-04  Jan Hubicka  <jh@suse.cz>
285         * passes.c (init_optimization_passes): Reindent.
287 2007-02-04  Jan Hubicka  <jh@suse.cz>
288             Eric Botcazou  <ebotcazou@adacore.com>
290         * tree-optimize.c (has_abnormal_outgoing_edge_p): Move to...
291         (execute_fixup_cfg): Break out the abnormal goto code.
292         * tree-inline.c (has_abnormal_outgoing_edge_p): ...here.
293         (make_nonlocal_label_edges): Move here from execute_fixup_cfg.
294         (optimize_inline_calls): Call make_nonlocal_label_edges.
296 2007-02-04  Jan Hubicka  <jh@suse.cz>
298         * tree-ssa-copyrename.c (copy_rename_partition_coalesce): Return
299         true when something was changed.
300         (rename_ssa_copies): When something was changed, do
301         TODO_remove_unused_locals.
302         * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
303         add TODO_remove_unused_locals when instruction was removed.
305 2007-02-04  Jan Hubicka  <jh@suse.cz>
307         * ipa-inline.c (try_inline): Improve debug output; work on already
308         inline edges too.
309         (cgraph_decide_inlining_incrementally): Indent; improve debug output;
310         call try_inline for already inlined edges too when flattening; 
311         inline also functions that make callee growth but overall unit size
312         reduce.
314 2007-02-04  Kazu Hirata  <kazu@codesourcery.com>
316         * config/m32c/bitops.md, config/m32c/jump.md,
317         config/m32c/m32c.c, config/m32c/m32c.h, config/m32r/m32r.c,
318         config/m32r/m32r.h, config/m32r/m32r.md,
319         config/m32r/predicates.md, config/m68hc11/larith.asm,
320         config/m68hc11/m68hc11.c, config/m68hc11/m68hc11.h,
321         config/m68k/m68k.h, config/mcore/mcore.md, config/mips/4k.md,
322         config/mips/mips-protos.h, config/mips/mips.c,
323         config/mips/mips.h, config/mips/mips.md, config/mips/mips16.S,
324         config/mn10300/mn10300.h, config/mn10300/predicates.md,
325         config/mt/mt.c, config/mt/mt.h, config/mt/mt.md: Follow
326         spelling conventions.
328         * config/v850/v850.c, config/v850/v850.h, config/v850/v850.md:
329         Follow spelling conventions.
331 2007-02-03  Douglas Gregor  <doug.gregor@gmail.com>
333        * c-opts.c (c_common_post_options): If C++0x mode is enabled, don't
334        warn about C++0x compatibility.
335         
336 2007-02-04  Kazu Hirata  <kazu@codesourcery.com>
338         * config/h8300/h8300.c, config/h8300/h8300.h,
339         config/h8300/h8300.md: Follow spelling conventions.
341 2007-02-03  Uros Bizjak  <ubizjak@gmail.com>
343         PR middle-end/30667
344         * combine.c (try_combine): Do not substitute source operand
345         with constants wider than 2 * HOST_BITS_PER_WIDE_INT.
347 2007-02-03  Jan Hubicka  <jh@suse.cz>
349         PR gcov-profile/30650
350         * value-prof.c (stringop_block_profile): Fix handling of size counter;
351         do not divide by zero for never executed counters.
352         (tree_find_values_to_profile): Fix counters.
353         * gcov-ui.h (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): Fix comments.
355 2007-02-03  Ian Lance Taylor  <iant@google.com>
357         * lower-subreg.c (simple_move_operand): New static function,
358         broken out of simple_move.  Reject LABEL_REF, SYMBOL_REF, and HIGH
359         operands.
360         (simple_move): Call simple_move_operand.
361         (find_decomposable_subregs): Add special handling of MEMs.
362         (can_decompose_p): Rename from cannot_decompose_p.  Reverse
363         meaning of return value.  If we see a hard register, test whether
364         it can store a word_mode value.  Change all callers.
366 2007-02-03  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
368         * pa.md (addvdi3, addvsi3, subvdi3, subvsi3, negvdi2, negvsi2): New
369         ftrapv insns and expanders.
370         (subdi3): Change define_expand operand 1 to arith11_operand, and
371         operand 2 to reg_or_0_operand.  Change constraints of 64-bit insn
372         pattern to handle reg_or_0 operands.  Revise 32-bit insn pattern to
373         handle 11-bit constants and reg_or_0 operands in operands 1 and 2,
374         respectively.
376         PR middle-end/30174
377         * varasm.c (notice_global_symbol): Treat global objects as weak when
378         flag_shlib is true.
380 2007-02-03  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
382         * emit-rtl.c (dconstpi): Delete.
383         (dconstsqrt2): New.
384         (init_emit_once): Delete dconstpi and init dconstsqrt2.
385         * real.h (dconstpi): Delete.
386         (dconstsqrt2): New.
387         * builtins.c (fold_builtin_cabs): Use dconstsqrt2.
388         (fold_builtin_hypot): Likewise.
390 2007-02-03  Tom Tromey  <tromey@redhat.com>
392         PR driver/30246
393         * gcc.c (cpp_unique_options): Any of -ggdb3, -gstabs3,
394         -gcoff3, -gxcoff3, -gvms3 implies -dD.
396 2007-02-03  Kazu Hirata  <kazu@codesourcery.com>
398         * c-decl.c, config/avr/avr.c, config/avr/avr.h,
399         config/m68k/m68k.c, config/m68k/netbsd-elf.h,
400         config/mn10300/mn10300.c, config/pdp11/pdp11.h,
401         config/rs6000/cell.md, config/rs6000/darwin.h,
402         config/sh/sh.md, config/sh/sh4-300.md, config/spu/spu.c,
403         config/spu/spu.md, cselib.c, expr.c, haifa-sched.c, hwint.h,
404         jump.c, reload.c, sched-deps.c, sched-int.h, tree-inline.c,
405         tree-profile.c, tree-ssa-live.h, tree-vrp.c: Fix comment
406         typos.  Follow spelling conventions.
407         * doc/invoke.texi: Follow spelling conventions.
409 2007-02-03  Roger Sayle  <roger@eyesopen.com>
411         * simplify-rtx.c (simplify_relational_operation_1): Implement some
412         canonicalization transformations that attempt to simplify integer
413         constant comparisons to become comparisons against zero.
415 2007-02-03  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
417         * builtins.c (fold_builtin_cabs): Fold cabs(x+xi) into
418         fabs(x)*sqrt(2).
419         * fold-const.c (fold_binary): Fix comment typos.  Fold complex
420         (x,0)-(0,y) into (x,-y).  Likewise (0,y)-(x,0) into (-x,y).
422 2007-02-02  Mike Stump  <mrs@apple.com>
424         * config/darwin9.h (DARWIN_LINKER_GENERATES_ISLANDS): Add.
425         * config/rs6000/rs6000.c (DARWIN_GENERATE_ISLANDS): Add.
426         (output_call): Use DARWIN_GENERATE_ISLANDS to decide when to
427         generate a branch island.
429 2007-02-02  Bob Wilson  <bob.wilson@acm.org>
431         * config/xtensa/xtensa.c (smalloffset_mem_p): Use BASE_REG_P.
432         (xtensa_legitimate_address_p): New.
433         (xtensa_legitimize_address): New.
434         (xtensa_output_addr_const_extra): New.
435         * config/xtensa/xtensa.h (REG_OK_STRICT_FLAG): Define.
436         (BASE_REG_P): New.
437         (REG_OK_FOR_BASE_P): Use BASE_REG_P.
438         (GO_IF_LEGITIMATE_ADDRESS): Move code to xtensa_legitimate_address_p.
439         (LEGITIMIZE_ADDRESS): Move code to xtensa_legitimize_address.
440         (OUTPUT_ADDR_CONST_EXTRA): Move code to xtensa_output_addr_const_extra.
441         * config/xtensa/xtensa-protos.h (xtensa_legitimate_address_p): New.
442         (xtensa_legitimize_address): New.
443         (xtensa_output_addr_const_extra): New.
445 2007-02-02  Steve Ellcey  <sje@cup.hp.com>
447         * config/ia64/ia64.c (ia64_print_operand): Fix compare strings.
449 2007-02-02  Ian Lance Taylor  <iant@google.com>
451         * expmed.c (expand_divmod): Add comment.
453 2007-02-02  Kazu Hirata  <kazu@codesourcery.com>
455         * emit-rtl.c (renumber_insns): Remove.
456         * flags.h: Remove the extern for flag_renumber_insns.
457         * rtl.h: Remove the prototype for renumber_insns.
458         * toplev.c (flag_renumber_insns): Remove.
460 2007-02-02  Hui-May Chang  <hm.chang@apple.com>
462         Revert for x86 darwin:
463         2005-06-19  Uros Bizjak  <uros@kss-loka.si>
464         
465         * config/i386/i386.c (ix86_function_arg_regno_p): Put back the
466         code before the following patch under TARGET_MACHO.
467         (ix86_function_value_regno_p): Likewise.
469 2007-02-02  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
471         * fold-const.c (negate_expr_p, fold_negate_expr): Handle
472         COMPLEX_EXPR.
474 2007-02-02  Maxim Kuvyrkov  <mkuvyrkov@ispras.ru>
476         PR target/29682
477         * config/ia64/ia64.c (ia64_speculate_insn): Restrict to memory loads to
478         general or fp registers.  Add comments.
479         * config/ia64/ia64.md (reg_pred_prefix): Add comment.
481 2007-02-02  Paolo Bonzini  <bonzini@gnu.org>
483         * pointer-set.c (insert_aux): Only return insertion slot.
484         (pointer_set_insert): Adjust.
485         (pointer_set_traverse, struct pointer_map_t, pointer_map_create,
486         pointer_map_destroy, pointer_map_insert, pointer_map_contains,
487         pointer_map_traverse): New.
488         * pointer-set.h (pointer_set_traverse, struct pointer_map_t,
489         pointer_map_create,  pointer_map_destroy, pointer_map_insert,
490         pointer_map_contains, pointer_map_traverse): Declare.
492 2007-02-02  Jakub Jelinek  <jakub@redhat.com>
494         PR middle-end/30473
495         * builtins.c (fold_builtin_sprintf): Do not attempt to optimize
496         sprintf (str, "%s").  Do not optimize sprintf (str, "nopercent", p++).
498 2007-02-02  Maxim Kuvyrkov  <mkuvyrkov@ispras.ru>
500         * sched-int.h (ds_to_dk, dk_to_ds): Declare functions.
501         
502         (struct _dep): New type.
503         (dep_t): New typedef.
504         (DEP_PRO, DEP_CON, DEP_KIND): New access macros.
505         (DEP_STATUS): New access macro.  The macro with the same name was
506         renamed to DEP_LINK_STATUS.
507         (dep_init): Declare function
509         (struct _dep_link): New type.
510         (dep_link_t): New typedef.
511         (DEP_LINK_NODE, DEP_LINK_NEXT, DEP_LINK_PREV_NEXTP): New access macros.
512         (DEP_LINK_DEP, DEP_LINK_PRO, DEP_LINK_CON, DEP_LINK_KIND): New macros.
513         (DEP_LINK_STATUS): New macro.
514         (debug_dep_links): New debug function.
516         (struct _deps_list): New type.
517         (deps_list_t): New typedef.
518         (DEPS_LIST_FIRST): New access macro.
519         (FOR_EACH_DEP_LINK): New cycle macro.
520         (create_deps_list, free_deps_list, delete_deps_list): Declare
521         functions.
522         (deps_list_empty_p, debug_deps_list, add_back_dep_to_deps_list): Ditto.
523         (find_link_by_pro_in_deps_list, find_link_by_con_in_deps_list): Ditto.
524         (copy_deps_list_change_con): Ditto.
526         (move_dep_link): Declare function.
528         (struct _dep_node): New type.
529         (dep_node_t): New typedef.
530         (DEP_NODE_BACK, DEP_NODE_DEP, DEP_NODE_FORW): New access macros.
532         (struct haifa_insn_data.back_deps): New field to hold backward
533         dependencies of the insn.
534         (struct haifa_insn_data.depend): Rename to forw_deps.  Change its type
535         to deps_list_t.
536         (struct haifa_insn_data.resolved_deps): Rename to resolved_back_deps.
537         Change its type to deps_list_t.
538         (INSN_BACK_DEPS): New access macro to use instead of LOG_LINKS.
539         (INSN_DEPEND): Rename to INSN_FORW_DEPS.
540         (RESOLVED_DEPS): Rename to INSN_RESOLVED_BACK_DEPS.
542         (INSN_COST): Move to haifa-sched.c.  Use insn_cost () instead.
543         
544         (DEP_STATUS): Rename to DEP_LINK_STATUS.  Fix typo in the comment.
546         (add_forw_dep, delete_back_forw_dep, insn_cost): Update declaration and
547         all callers.
548         (dep_cost): Declare.
549         
550         * sched-deps.c (CHECK): New macro to (en/dis)able sanity checks.
551         (ds_to_dk, dk_to_ds): New functions.
552         
553         (init_dep_1): New static function.
554         (init_dep): New function.
555         (copy_dep): New static function.
556         
557         (dep_link_consistent_p, attach_dep_link, add_to_deps_list): New static
558         functions.
559         (detach_dep_link): New static function.
560         (move_dep_link): New function.
561         
562         (dep_links_consistent_p, dump_dep_links): New static functions.
563         (debug_dep_links): New debugging function.
564         
565         (deps_obstack, dl_obstack, dn_obstack): New static variables.
566         
567         (alloc_deps_list, init_deps_list): New static functions.
568         (create_deps_list): New function.
569         (clear_deps_list): New static function.
570         (free_deps_list, delete_deps_list, deps_list_empty_p): New functions.
571         (deps_list_consistent_p, dump_deps_list): New static functions.
572         (debug_deps_list): New function.
573         (add_back_dep_to_deps_list, find_link_by_pro_in_deps_list): New
574         functions.
575         (find_link_by_con_in_deps_list, copy_deps_list_change_con): Ditto.
577         (maybe_add_or_update_back_dep_1, add_or_update_back_dep_1): Update to
578         use new scheduler dependencies lists.
579         (add_back_dep, delete_all_dependences, fixup_sched_groups): Ditto.
580         (sched_analyze): Ditto.  Initialize dependencies lists.
581         (add_forw_dep, compute_forward_dependences): Update to use new
582         scheduler dependencies lists.
583         
584         (init_dependency_caches): Init deps_obstack.
585         (free_dependency_caches): Free deps_obstack.
586         
587         (adjust_add_sorted_back_dep, adjust_back_add_forw_dep): Update to use
588         new scheduler dependencies lists.
589         (delete_forw_dep, add_or_update_back_forw_dep): Ditto.
590         (add_back_forw_dep, delete_back_forw_dep): Ditto.
592         * sched-rgn.c (set_spec_fed, find_conditional_protection, is_pfree):
593         Update to use new scheduler dependencies lists.
594         (is_conditionally_protected, is_prisky, add_branch_dependences): Ditto.
595         (debug_dependencies): Ditto.
596         (schedule_region): Update comments.
597         
598         * sched-ebb.c (earliest_block_with_similiar_load): Update to use new
599         scheduler dependencies lists.
600         (schedule_ebb): Update comments.
601         
602         * rtl.def (DEPS_LIST): Remove.
603         
604         * lists.c (unused_deps_list): Remove.
605         (free_list): Update assertions.
606         
607         (alloc_DEPS_LIST, free_DEPS_LIST_list, free_DEPS_LIST_node): Remove.
608         (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
610         * rtl.h (free_DEPS_LIST_list, alloc_DEPS_LIST): Remove declarations.
611         (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
612         
613         * haifa-sched.c (comments): Update.
614         (insn_cost1): Remove.  Inline the code into insn_cost ().
615         (insn_cost): Update to use new scheduler dependencies lists.  Move
616         processing of the dependency cost to dep_cost ().
617         (dep_cost): New function.  Use it instead of insn_cost () when
618         evaluating cost of the dependency.  Use compatible interface to
619         interact with the target.
620         (priority): Update to use new scheduler dependencies lists.
621         (rank_for_schedule): Ditto.  Optimize heuristic that prefers the insn
622         with greater number of insns that depend on the insn.
623         (schedule_insn): Update to use new scheduler dependencies lists.  Add
624         code to free backward dependencies lists.  Inline and optimize code
625         from resolve_dep () - see PR28071.
626         (ok_for_early_queue_removal): Update to use new scheduler dependencies
627         lists.  Update call to targetm.sched.is_costly_dependence hook.
628         
629         (fix_inter_tick, try_ready, fix_tick_ready): Update to use new
630         scheduler dependencies lists.
631         
632         (resolve_dep): Remove.  Move the logic to schedule_insn ().
633         (init_h_i_d): Initialize dependencies lists.
634         
635         (process_insn_depend_be_in_spec): Rename to
636         process_insn_forw_deps_be_in_spec.  Update to use new scheduler
637         dependencies lists.
638         (add_to_speculative_block, create_check_block_twin, fix_recovery_deps):
639         Update to use new scheduler dependencies lists.
640         (clear_priorities, calc_priorities, add_jump_dependencies): Ditto.
641         
642         * ddg.c (create_ddg_dependence, create_ddg_dep_no_link): Update to use
643         new scheduler dependencies lists.
644         (build_intra_loop_deps): Ditto.
645         
646         * target.h (struct _dep): Declare to use in
647         gcc_target.sched.is_costly_dependence.
648         (struct gcc_target.sched.adjust_cost): Fix typo.
649         (struct gcc_target.sched.is_costly_dependence): Change signature to use
650         single dep_t parameter instead of an equivalent triad.
651         (struct gcc_target.sched.adjust_cost_2): Remove.
653         * target-def.h (TARGET_SCHED_ADJUST_COST_2): Remove.
655         * reg-notes.def (DEP_TRUE, DEP_OUTPUT, DEP_ANTI): Update comments.
657         * doc/tm.texi (TARGET_SCHED_IS_COSTLY_DEPENDENCE): Update
658         documentation.
659         (TARGET_SCHED_ADJUST_COST_2): Remove documentation.
661         * doc/rtl.texi (LOG_LINKS): Remove part about instruction scheduler.
662         (REG_DEP_TRUE): Document.
663         
664         * config/ia64/ia64.c (ia64_adjust_cost_2): Rename to ia64_adjust_cost.
665         Change signature to correspond to the targetm.sched.adjust_cost hook.  
666         Update use in TARGET_SCHED_ADJUST_COST_2.
667         (TARGET_SCHED_ADJUST_COST_2): Rename to TARGET_SCHED_ADJUST_COST.
668         (ia64_dependencies_evaluation_hook, ia64_dfa_new_cycle): Update to use
669         new scheduler dependencies lists.
670         (ia64_gen_check): Ditto.
672         * config/mips/mips.c (vr4130_swap_insns_p): Update to use new scheduler
673         dependencies lists.
674         
675         * config/rs6000/rs6000.c (rs6000_is_costly_dependence): Change
676         signature to correspond to the targetm.sched.is_costly_dependence hook.
677         (is_costly_group): Update to use new scheduler dependencies lists.
679         * config/spu/spu.c (spu_sched_adjust_cost): Use insn_cost () function
680         instead of INSN_COST () macro.
682 2007-02-01  Ian Lance Taylor  <iant@google.com>
684         * lower-subreg.c (resolve_clobber): Handle a subreg of a concatn.
686 2007-02-01  Guy Martin  <gmsoft@gentoo.org>
688         * pa.md (tp_load): Correct mfctl instruction syntax.
690 2007-02-01  Geoffrey Keating  <geoffk@apple.com>
692         * config/rs6000/rs6000.c (rs6000_stack_info): Correct
693         altivec_padding_size calculation on AIX.  Improve comment, add
694         assert to verify that it's right.
696         * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
697         64-bit.
699 2007-2-01   Seongbae Park <seongbae.park@gmail.com>
701         PR inline-asm/28686
702         * global.c (global_alloc): Add mising initialization of
703         ELIMINABLE_REGSET.
705 2007-02-01  Roger Sayle  <roger@eyesopen.com>
707         * alias.c (init_alias_analysis): Correct whitespace.
708         * bb-reorder.c (fix_edges_for_rarely_executed_code,
709         partition_hot_cold_basic_blocks): Likewise.
710         * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
711         expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
712         maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
713         fold_builtin_snprintf_chk, fold_builtin_printf,
714         fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
715         do_mpfr_arg3, do_mpfr_sincos): Likewise.
716         * cfgcleanup.c (cleanup_cfg): Likewise.
717         * cfgexpand.c (tree_expand_cfg): Likewise.
718         * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
719         * function.c (get_next_funcdef_no): Likewise.
720         * gengtype.c (main): Likewise.
721         * genmodes.c (main): Likewise.
722         * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
723         * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
724         * ifcvt.c (noce_emit_move_insn): Likewise.
725         * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
726         Likewise.
727         * stor-layout.c (get_best_mode): Likewise.
728         * tree-ssa-loop-niter.c (get_val_for): Likewise.
729         * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
730         scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
731         perform_var_substitution, solve_graph): Likewise.
732         * tree-vrp.c (vrp_finalize): Likewise.
734 2007-02-01  Ian Lance Taylor  <iant@google.com>
736         * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
737         high part of a paradoxical subreg, return a constant zero.
739 2007-02-01  Zdenek Dvorak <dvorakz@suse.cz>
741         * toplev.c (lang_dependent_init): Call init_set_costs.
742         * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
744 2007-02-01  Richard Guenther  <rguenther@suse.de>
746         PR middle-end/30656
747         * fold-const.c (fold_negate_expr): Allow negating a
748         constant if overflow does not change.
750 2007-02-01  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
752         * doc/c-tree.texi (Expression trees): Improve markup.
753         * doc/tm.texi (Register Classes, Addressing Modes)
754         (Floating Point): Fix spacing after abbreviations.  Fix some
755         typos.
757 2007-02-01  Ben Elliston  <bje@au.ibm.com>
759         * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
760         * doc/contrib.texi: Likewise.
761         * doc/install.texi: Likewise.
763 2007-01-31  Richard Henderson  <rth@redhat.com>
764             Ian Lance Taylor  <iant@google.com>
766         * lower-subreg.c: New file.
767         * rtl.def (CONCATN): Define.
768         * passes.c (init_optimization_passes): Add pass_lower_subreg and
769         pass_lower_subreg2.
770         * emit-rtl.c (update_reg_offset): New static function, broken out
771         of gen_rtx_REG_offset.
772         (gen_rtx_REG_offset): Call update_reg_offset.
773         (gen_reg_rtx_offset): New function.
774         * regclass.c: Revert patch of 2006-03-05, restoring
775         reg_scan_update.
776         (clear_reg_info_regno): New function.
777         * dwarf2out.c (concatn_loc_descriptor): New static function.
778         (loc_descriptor): Handle CONCATN.
779         * common.opt (fsplit_wide_types): New option.
780         * opts.c (decode_options): Set flag_split_wide_types when
781         optimizing.
782         * timevar.def (TV_LOWER_SUBREG): Define.
783         * rtl.h (gen_reg_rtx_offset): Declare.
784         (reg_scan_update): Declare.
785         * regs.h (clear_reg_info_regno): Declare.
786         * tree-pass.h (pass_lower_subreg): Declare.
787         (pass_lower_subreg2): Declare.
788         * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
789         (Optimize Options): Add -fsplit-wide-types to -O1 list.  Document
790         -fsplit-wide-types.
791         * doc/rtl.texi (Regs and Memory): Document concat and concatn.
792         * Makefile.in (OBJS-common): Add lower-subreg.o.
793         (lower-subreg.o): New target.
795 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
797         * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
799 2007-01-31  Anatoly Sokolov <aesok@post.ru>
801         PR target/19087
802         * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
804 2007-01-31  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
806         PR middle-end/29335
807         * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
809 2007-01-31  Zdenek Dvorak <dvorakz@suse.cz>
811         * cfgloop.h: Include vec-prim.h.
812         (enum li_flags): Remove LI_ONLY_OLD.
813         (loop_iterator): Changed.
814         (fel_next, fel_init): Iterate over loop tree.
815         (FOR_EACH_LOOP_BREAK): New macro.
816         * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
817         FOR_EACH_LOOP.
818         * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
819         * modulo-sched.c (sms_schedule): Ditto.
820         * tree-vectorizer.c (vectorize_loops): Ditto.
821         * doc/loop.texi: Update information on loop numbering and behavior of
822         FOR_EACH_LOOP wrto new loops.
823         * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
824         add_to_evolution_1): Test nestedness of loops instead of comparing
825         their numbers.
826         * tree-chrec.c (chrec_fold_plus_poly_poly,
827         chrec_fold_multiply_poly_poly, chrec_evaluate,
828         hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
829         reset_evolution_in_loop): Ditto.
830         * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
832 2007-01-31  Dirk Mueller  <dmueller@suse.de>
834         * c-common.c (warn_about_parentheses): Separate warning about
835         un-parenthized sequence of comparison operators from the one
836         which is supposed to warn about x <= y <= z.
838 2007-01-31  Uros Bizjak  <ubizjak@gmail.com>
840         * optabs.h (enum optab_index): Add new OTI_isinf.
841         (isinf_optab): Define corresponding macro.
842         * optabs.c (init_optabs): Initialize isinf_optab.
843         * genopinit.c (optabs): Implement isinf_optab using isinf?f2
844         patterns.
845         * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
846         (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
847         using isinf_optab.
848         (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
849         expand_builtin_interclass_mathfn.
850         * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
851         * config/i386/i386.md (UNSPEC_FXAM): New constant.
852         (fxam<mode>2_i387): New insn pattern.
853         (isinf<mode>2) New expander to implement isinf, isinff and isinfl
854         built-in functions as x87 inline asm.
856 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
858         * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
860 2007-01-30  Eric Christopher  <echristo@apple.com>
862         * config.gcc: Add geode.
864 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
866         * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
867         ipa-inline.c, tree-profile.c, tree-ssa-live.c,
868         tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
869         value-prof.c: Fix comment typos.
871 2007-01-30  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
873         PR c++/24745
874         * doc/invoke.texi (Wpointer-arith): Document warning.
876 2007-01-30  Janis Johnson
878         * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
879         the list of discrepancies from the draft TR.
881 2007-01-30  Dirk Mueller  <dmueller@suse.de>
883         PR c++/30601
884         * doc/invoke.texi (-Wreturn-type): Update description to
885         match new behavior.
887 2007-01-30  Richard Sandiford  <richard@codesourcery.com>
889         * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
891 2007-01-30  Uros Bizjak  <ubizjak@gmail.com>
893         * builtins.c (expand_builtin_int_interclass_roundingfn): New function
894         to handle optabs that operate on floating point input argument and
895         output to integer output.
896         (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
897         (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
898         (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
899         expand_builtin_interclass_mathfn ().
900         * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
901         (ilogbsi2): Remove.
902         (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
903         ilogbl built-in functions as x87 intrinsics.
905 2007-01-30  Richard Guenther  <rguenther@suse.de>
907         PR middle-end/27657
908         * dwarf2out.c (reference_to_unused): Query varpool if the
909         variable was output.
911 2007-01-30  Richard Guenther  <rguenther@suse.de>
913         PR middle-end/30313
914         * passes.c (execute_one_pass): Reset in_gimple_form to not
915         confuse non-unit-at-a-time mode.
917 2007-01-29  Roger Sayle  <roger@eyesopen.com>
918             Richard Guenther  <rguenther@suse.de>
920         * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
921         avoid undefined behaviour on overflow.  Use force_fit_type_double
922         to construct the constant with the specified TREE_OVERFLOW.
924 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
926         * config/dfp-bit.c: Add parameterized support for fp exceptions.
927         * config/dfp-bit.h: Ditto.
929 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
931         * c-decl.c (pop_scope): Replace warnings with call to
932         warn_for_unused_label.
933         * c-common.h (warn_for_unused_label): Declare.
934         * c-common.c (warn_for_unused_label): Define.
936 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
938         * tree-optimize.c (update_inlined_to_pointers): Delete unused
939         function.
941 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
943         * Makefile.in (USER_H): Remove decfloat.h.
944         * ginclude/decfloat.h: Delete, moving contents to ...
945         * ginclude/float.h: Add support for decimal floating point,
946         guarded by __STDC_WANT_DEC_FP__.
948 2007-01-29  Mike Stump  <mrs@apple.com>
950         * doc/gccint.texi (Top): Rename Loop Representation to Loop
951         Analysis and Representation to resolve case insensitive conflict.
952         * doc/loop.texi (Loop Analysis and Representation): Likewise.
954 2007-01-28  Daniel Berlin  <dberlin@dberlin.org>
956         * tree.h (struct tree_memory_tag): Add aliases member.
957         (MTAG_ALIASES): New macro.
958         * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
959         (add_may_alias): Remove pointer-set. Update for may_aliases being
960         a bitmap.
961         (mark_aliases_call_clobbered): Update for may_aliases being a
962         bitmap.
963         (compute_tag_properties): Ditto.
964         (create_partition_for): Ditto.
965         (compute_memory_partitions): Ditto.
966         (dump_may_aliases_for): Ditto.
967         (is_aliased_with): Ditto.
968         (add_may_alias_for_new_tag): Ditto.
969         (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
970         (compute_is_aliased): New function.
971         (compute_may_aliases): Call compute_is_aliased).
972         (init_alias_info): Initialize alias_bitmap_obstack.
973         (union_alias_set_into): New function.
974         (compute_flow_sensitive_aliasing): Use union_aliases_into.
975         (have_common_aliases_p): Rewrite to take two bitmaps and use
976         intersection.
977         (compute_flow_insensitive_aliasing): Stop using pointer-sets.
978         Update for bitmaps.
979         (finalize_ref_all_pointers): Update for add_may_alias changes.
980         (new_type_alias): Ditto.
981         * tree-flow-inline.h (may_aliases): Return a bitmap.
982         * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
983         * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
984         may_aliases being a bitmap.
985         * tree-flow.h (struct var_ann_d): Remove may_aliases member.
986         may_aliases now returns a bitmap.
987         * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
988         being a bitmap.
989         * tree-ssa-operands.c (add_virtual_operand): Update for
990         may_aliases being a bitmap.
992 2007-01-29  Daniel Berlin  <dberlin@dberlin.org>
994         PR tree-optimization/30630
995         * tree-ssa-structalias.c (do_complex_constraint): Mark correct
996         variable as changed.
998 2007-01-29  Simon Martin  <simartin@users.sourceforge.net>
1000         PR c++/28266
1001         * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
1002         expanded only once even if an error occurs.
1004 2007-01-29  Ben Elliston  <bje@au.ibm.com>
1006         * gcov-io.h (__gcov_indirect_call_profiler): Declare.
1007         (__gcov_average_profiler): Likewise.
1008         (__gcov_ior_profiler): Likewise.
1009         (__gcov_merge_ior): Likewise.
1011 2007-01-28  Jan Hubicka  <jh@suse.cz>
1013         * builtins.c (expand_builtin_memset): Fix typo in my last patch.
1014         * value-prof.c (stringop_block_profile): Likewise.
1016 2007-01-28  Jan Hubicka  <jh@suse.cz>
1018         * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
1019         variant handling histograms; add wrapper.
1020         (clear_storage_via_libcall): Export.
1021         (emit_block_move_hints): Break out from ...; add histograms.
1022         (emit_block_move): ... this one.
1023         (clear_storage_hints): Break out from ...; add histograms.
1024         (clear_storage): ... this one.
1025         (set_storage_via_memset): Handle histogram.
1026         * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
1027         (clear_storage_hints, clear_storage_via_libcall): Declare.
1028         (set_storage_via_setmem): Update prototype.
1029         * doc/md.texi (movmem, setmem): Document new arguments.
1031         * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
1032         new histograms.
1033         (stringop_block_profile): New global function.
1034         (tree_stringops_values_to_profile): Profile block size and alignment.
1035         * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
1036         HIST_TYPE_IOR.
1037         (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
1038         (stringop_block_profile): Declare.
1039         * builtins.c: Include value-prof.h.
1040         (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
1041         * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
1042         (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
1043         (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
1044         * profile.c (instrument_values): Add new counters.
1045         * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
1046         calls.
1047         * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
1048         (tree_init_edge_profiler): Build new profilers.
1049         (tree_gen_average_profiler, tree_gen_ior_profiler): New.
1050         (pass_tree_profile): Add dump.
1051         (tree_profile_hooks): Update.
1052         * Makefile.in (LIBGCOV): Add new constants.
1053         * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
1054         __gcov_ior_profiler): New.
1055         * i386.md (movmem/setmem expanders): Add new optional arguments.
1057 2007-01-28  David Edelsohn  <edelsohn@gnu.org>
1059         * doc/md.texi (Standard Pattern Names): Document blockage pattern.
1061 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
1063         * tree-ssa-loop-unswitch.c: Include tree-inline.h.
1064         (tree_unswitch_single_loop): Pass eni_size_weights to
1065         tree_num_loop_insns.
1066         * tree-ssa-loop-manip.c: Include tree-inline.h.
1067         (can_unroll_loop_p): Pass eni_size_weights to
1068         tree_num_loop_insns.
1069         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
1070         Pass eni_size_weights to estimate_num_insns.
1071         * tree.h (init_inline_once): Export.
1072         * toplev.c (backend_init): Call init_inline_once.
1073         * cgraphunit.c (cgraph_process_new_functions,
1074         cgraph_analyze_function): Pass eni_inlining_weights to
1075         estimate_num_insns.
1076         * ipa-inline.c (compute_inline_parameters): Ditto.
1077         * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
1078         to estimate_num_insns.
1079         (try_unroll_loop_completely): Pass eni_size_weights to
1080         tree_num_loop_insns.
1081         * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
1082         ot estimate_num_insns.
1083         * tree-ssa-loop-prefetch.c: Include tree-inline.h.
1084         (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
1085         * tree-inline.c (eni_inlining_weights, eni_size_weights,
1086         eni_time_weights): New variables.
1087         (init_inline_once): Initialize them.
1088         (struct eni_data): Mew.
1089         (estimate_num_insns_1, estimate_num_insns): Use weights.
1090         * tree-inline.h (struct eni_weights_d): New.
1091         (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
1092         (estimate_num_insns): Declaration changed.
1093         * cfgloop.h (tree_num_loop_insns): Declaration changed.
1094         * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
1095         tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
1097 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
1099         * tree-data-ref.c (conflict_fn): Assert that the number of affine
1100         relations in the conflict function is valid.
1102 2007-01-27  Ian Lance Taylor  <iant@google.com>
1104         * common.opt: Add fstrict-overflow.
1105         * opts.c (decode_options): Set flag_strict_overflow if -O2.
1106         * flags.h (TYPE_OVERFLOW_WRAPS): Define.
1107         (TYPE_OVERFLOW_UNDEFINED): Define.
1108         (TYPE_OVERFLOW_TRAPS): Define.  This replaces TYPE_TRAP_SIGNED.
1109         Replace all uses.
1110         * tree.h (TYPE_TRAP_SIGNED): Don't define.
1111         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
1112         (fold_negate_expr): Likewise.
1113         (make_range): Likewise.
1114         (extract_muldiv_1): Likewise.
1115         (maybe_canonicalize_comparison): Likewise.
1116         (fold_comparison): Likewise.
1117         (fold_binary): Likewise.
1118         (tree_expr_nonnegative_p): Likewise.
1119         (tree_expr_nonzero_p): Likewise.
1120         * tree-vrp.c (compare_values): Likewise.
1121         (extract_range_from_binary_expr): Likewise.
1122         (extract_range_from_unary_expr): Likewise.
1123         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
1124         Likewise.
1125         (nowrap_type_p): Likewise.
1126         * tree-scalar-evolution.c (simple_iv): Likewise.
1127         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
1128         (build_range_check): Likewise.
1129         (extract_muldiv_1): Likewise.
1130         (fold_comparison): Likewise.
1131         * tree-vrp.c (vrp_int_const_binop): Likewise.
1132         (extract_range_from_unary_expr): Likewise.
1133         * convert.c (convert_to_integer): Likewise.
1134         * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
1135         (fold_comparison): Likewise.
1136         (fold_binary): Likewise.
1137         * optabs.c (optab_for_tree_code): Likewise.
1138         * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
1139         * simplify-rtx.c (simplify_const_relational_operation): Check
1140         flag_strict_overflow and flag_trapv.
1141         (simplify_const_relational_operation): Likewise.
1142         * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
1143         (Optimize Options): Add -fstrict-overflow to -O2 list.  Document
1144         -fstrict-overflow.
1146 2007-01-27  Roger Sayle  <roger@eyesopen.com>
1148         * tree.c (tree_fold_gcd): Delete.
1149         * tree.h (tree_fold_gcd): Remove prototype.
1150         * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
1151         test whether one constant integer is a multiple of another.  Instead
1152         call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
1153         * fold-const.c (multiple_of_p):  We've determined both TOP and
1154         BOTTOM are integer constants so we can call int_const_binop directly
1155         instead of the more generic const_binop.
1157 2007-01-27  Roger Sayle  <roger@eyesopen.com>
1159         * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
1160         1*X check that the constant hasn't overflowed, to preserve the
1161         TREE_OVERFLOW bit.
1162         (round_up): Provide an efficient implementation when rouding-up an
1163         INTEGER_CST to a power-of-two.
1165 2007-01-28  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
1167         * doc/sourcebuild.texi: Add comma for clarity.
1168         * doc/extend.texi: Fix some typos.
1169         * doc/passes.texi: Likewise.
1170         * doc/cppinternals.texi: Likewise.
1171         * doc/c-tree.texi: Likewise.
1172         * doc/tree-ssa.texi: Likewise.
1173         * doc/install.texi: Likewise.
1175 2007-01-27  Jan Hubicka  <jh@suse.cz>
1177         * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
1178         (pass_sra): Do not require alias information.
1179         * passes.c (init_optimization_passes): Add SRA
1181 2007-01-27  Steven Bosscher  <steven@gcc.gnu.org>
1183         * tracer.c (rest_of_handle_tracer): We already cleaned
1184         up the CFG in tracer() so don't do it here again.
1185         * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
1186         cleanup_cfg here, either.  And don't call renumber_insns.
1188         * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
1189         and BB_HEAD are in the insn stream here.  Instead make sure
1190         that BB_INSN is valid on all insns.  Also, do check here that
1191         there are no pending branch predictions...
1192         (rtl_verify_flow_info): ...instead of doing it here.  Checks
1193         for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
1194         here.
1196 2007-01-26  Roger Sayle  <roger@eyesopen.com>
1198         * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
1199         function to simplify/factorize operand order canonicalization.
1200         (ix86_fixup_binary_operands): Reorganize using the above function.
1201         (ix86_binary_operator_ok): Likewise.
1203 2007-01-27  Jakub Jelinek  <jakub@redhat.com>
1205         * genattrtab.c (struct attr_value_list, insn_code_values): Move to
1206         file scope from optimize_attrs.
1207         (simplify_test_exp): If insn_code_values is not NULL, use it to speed
1208         up search.
1209         (optimize_attrs): Clear insn_code_values after freeing it.
1211 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
1213         * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
1214         a parts.base assignment.
1216 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
1218         * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
1219         (compute_subscript_distance, initialize_data_dependence_relation,
1220         finalize_ddr_dependent, analyze_ziv_subscript,
1221         analyze_siv_subscript_cst_affine,
1222         compute_overlap_steps_for_affine_univar,
1223         compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
1224         analyze_siv_subscript, analyze_miv_subscript,
1225         analyze_overlapping_iterations, subscript_dependence_tester_1,
1226         compute_self_dependence, free_dependence_relation): Work
1227         with affine_fn instead of chrecs.
1228         (dump_affine_function, dump_conflict_function, affine_function_equal_p,
1229         common_affine_function, affine_function_base,
1230         affine_function_constant_p, affine_fn_op, affine_fn_plus,
1231         affine_fn_minus, affine_fn_free, conflict_fn_not_known,
1232         conflict_fn_no_dependence, free_conflict_function, free_subscripts,
1233         conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
1234         (all_chrecs_equal_p): Removed.
1235         * tree-data-ref.h (affine_fn, conflict_function): New types.
1236         (struct subscript): Change type of conflicting_iterations_in_a
1237         and conflicting_iterations_in_b.
1239 2007-01-26  Steve Ellcey  <sje@cup.hp.com>
1241         PR other/30182
1242         * config/pa/pa.h (TARGET_HPUX_11): New.
1243         * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
1244         * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
1246 2007-01-26  Daniel Berlin  <dberlin@dberlin.org>
1247         Richard Guenther  <rguenther@suse.de>
1249         * tree-ssa-structalias.c (solve_graph): Handle case
1250         we merged the variable to another.
1252 2007-01-25  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
1254         * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
1255         functions.
1257         * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
1258         catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
1260 2007-01-25  DJ Delorie  <dj@redhat.com>
1262         * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
1263         allow changes to modes which don't fit in those registers.
1265         * reload1.c (choose_reload_regs): Check for invalid subregs before
1266         computing their locations, not after.
1268 2007-01-25  Geoffrey Keating  <geoffk@apple.com>
1270         PR 25127
1271         * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
1272         save Altivec registers in an eh_return function.
1273         (compute_vrsave_mask): Likewise.
1274         (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
1275         for saving Altivec registers.
1276         (rs6000_emit_prologue): Don't allocate stack twice in
1277         eh_return function.  Correct expected value of altivec_save_offset
1278         when using save_world.  Describe save of R0 to stack when using
1279         save_world.  Describe stack pointer adjustment when using
1280         save_world.  Remove duplicated eh_return parameter register saving.
1281         Update sp_offset variable after save_world.
1282         * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
1283         darwin-world.asm.
1284         (LIB2FUNCS_EXTRA): Add darwin-world.asm.
1285         * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
1286         implies Altivec.
1288 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
1290         * config.gcc (ia64*-*-hpux*): Make posix threads the default.
1292 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
1294         PR other/30182
1295         * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
1296         * config/ia64/ia64.c (ia64_init_builtins):  Ditto.
1298 2007-01-25  Richard Guenther  <rguenther@suse.de>
1300         * doc/invoke.texi (-Wcoverage-mismatch): Document.
1301         * common.opt (-Wcoverage-mismatch): New warning option.
1302         * coverage.c (get_coverage_counts): Ignore coverage mismatch
1303         if -Wcoverage-mismatch is given.
1305 2007-01-25  Razya Ladelsky  <razya@il.ibm.com>
1307         * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
1308         (ipcp_driver): Change to static definition.
1309         Add dumping of the ifunctions.
1310         (constant_val_insert): Remove unused parameter. Support for SSA.
1311         (ipcp_propagate_const): Support for SSA.
1312         (ipcp_profile_bb_print): Print only analyzed nodes.
1313         (ipcp_replace_map_create): Remove support for Fortran constant
1314         for now.
1315         * ipa-prop.c (ipa_method_modify_stmt,
1316         ipa_callsite_compute_param): Support for SSA.
1317         * ipa-prop.h (ipcp_driver): Remove declaration.
1318         (IS_VALID_TREE_MAP_INDEX): Add define.
1320 2007-01-24  Geoffrey Keating  <geoffk@apple.com>
1322         * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
1324 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
1326         PR middle-end/30447
1327         * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
1328         constant argument to cabs and do it without checking for
1329         -funsafe-math-optimizations.
1331 2007-01-24  Douglas Gregor  <dgregor@osl.iu.edu>
1333         * c-common.h (RID_FIRST_CXX0X): New.
1334         (RID_LAST_CXX0X): New.
1335         * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
1336         by -Wall.
1337         * c.opt (Wc++0x-compat): New.
1338         * doc/invoke.texi (-Wc++0x-compat): Document.
1340 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
1342         * builtins.c (fold_builtin_carg): New.
1343         (fold_builtin_1): Use it.
1345 2007-01-24  Jan Hubicka  <jh@suse.cz>
1347         * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
1349         * ipa-inline.c (initial_insns, max_insns): Delete.
1350         (compute_max_insns): New function.
1351         (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
1352         of insns as base for code growth.
1353         (cgraph_decide_inlining): Make initial_insns local; do not compute
1354         max_insns.
1355         * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
1356         * doc/invoke.texi (inline-unit-growth): Update docs.
1358 2007-01-24  Jakub Jelinek  <jakub@redhat.com>
1360         * config/i386/i386.h (x86_cmpxchg16b): Remove const.
1361         (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
1362         * config/i386/i386.c (x86_cmpxchg16b): Remove const.
1363         (override_options): Add PTA_CX16 flag.  Set x86_cmpxchg16b
1364         for CPUs that have PTA_CX16 set.
1366         PR middle-end/27416
1367         * gimplify.c (omp_check_private): New function.
1368         (gimplify_scan_omp_clauses): Use it for
1369         firstprivate/lastprivate/reduction.
1371         PR middle-end/30494
1372         * gimplify.c (omp_add_variable): Don't call omp_notice_variable
1373         on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
1375         PR middle-end/30421
1376         * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
1377         If lower_lastprivate_clauses emits some statements, append them
1378         to dlist rather than body_p and to body_p append an initializer.
1379         (lower_omp_for): Adjust caller.
1381 2007-01-24  Steve Ellcey  <sje@cup.hp.com>
1383         * target.h (globalize_decl_name): New.
1384         * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
1385         * output.h (default_globalize_decl_name): New.
1386         * varasm.c (asm_output_bss): Use globalize_decl_name instead of
1387         globalize_label.
1388         (globalize_decl): Ditto.
1389         (default_globalize_decl_name): New.
1390         * config/ia64/ia64.c (ia64_globalize_decl_name): New.
1391         (ia64_handle_version_id_attribute): New.
1392         (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
1393         (ia64_asm_output_external): Use globalize_decl_name instead
1394         of globalize_label.
1395         * doc/extend.texi (version_id): New pragma.
1396         * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
1398 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
1400         * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
1401         _uleb128_t and _Unwind_SWord with _sleb128_t.
1402         * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
1403         uw_frame_state_for, uw_update_context_1): Likewise.
1404         * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
1405         * unwind-pe.h (read_uleb128, read_sleb128,
1406         read_encoded_value_with_base): Likewise.
1407         * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
1409 2007-01-24  Richard Guenther  <rguenther@suse.de>
1411         * builtins.c (expand_builtin_cexpi): Get the fndecl
1412         for cexp in the correct way.
1414 2007-01-24  Jan Hubicka  <jh@suse.cz>
1416         * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
1418 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
1420         * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
1421         (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
1422         __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
1423         __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
1424         __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
1425         __SIZEOF_PTRDIFF_T__.
1426         * doc/cpp.texi: Documentation for the new builtin macros added.
1428 2007-01-24  Uros Bizjak  <ubizjak@gmail.com>
1430         * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
1431         log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
1432         instead of operands[2] to avoid access past the end of array.
1434 2007-01-24  Richard Sandiford  <richard@codesourcery.com>
1436         * reload1.c (emit_reload_insns): Pass the reload register
1437         for a non-spill output reload through forget_old_reloads_1.
1439 2007-01-23  Joseph Myers  <joseph@codesourcery.com>
1441         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
1442         _SOFT_DOUBLE if doubles use software floating-point.
1443         * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
1444         double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
1445         * config/rs6000/darwin-ldouble.c: Also compile functions for
1446         hard-float without FPRs.  Use fmsub function for all __NO_FPRS__
1447         cases.  Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
1448         * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
1449         commented-out long double override.
1450         (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
1451         * config/rs6000/eabispe.h: Likewise.
1452         * config/rs6000/rs6000.c (rs6000_override_options): Don't override
1453         long double for non-SPE.
1454         (rs6000_handle_option): Likewise.
1455         (invalid_e500_subreg): Disallow more subregs involding DImode,
1456         DFmode, TImode or TFmode.
1457         (rs6000_legitimate_offset_address_p): Check TFmode offsets for
1458         E500 double.
1459         (legitimate_lo_sum_address_p): Also check for TFmode for E500
1460         double.
1461         (rs6000_legitimize_address): Also handle TFmode for E500 double.
1462         (rs6000_legitimize_reload_address): Also handle TFmode for E500
1463         double.
1464         (rs6000_legitimate_address): Also check for TFmode for E500
1465         double.
1466         (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
1467         (spe_build_register_parallel): Handle TFmode and TCmode.
1468         (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
1469         double.
1470         (function_arg): Handle TFmode and TCmode for E500 double.
1471         (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
1472         in general.
1473         (print_operand): Handle TFmode and TImode for %y.
1474         (rs6000_generate_compare): Handle TFmode comparisons for E500
1475         double.
1476         (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
1477         (rs6000_function_value): Handle TFmode and TCmode for E500 double.
1478         (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
1479         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
1480         TFmode for E500 double.
1481         * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
1482         (floatsidf2): Enable for E500 double.
1483         (movtf_softfloat): Use rs6000_nonimmediate_operand.
1484         (extenddftf2): Change to extenddftf2_fprs.
1485         (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
1486         depending on TARGET_E500_DOUBLE.
1487         (extendsftf2): Enable for E500 double.
1488         (trunctfdf2): Enable for E500 double.
1489         (trunctfsf2): Change to trunctfsf2_fprs.
1490         (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
1491         depending on TARGET_E500_DOUBLE.
1492         (floatsitf2): Enable for E500 double.
1493         (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
1494         (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
1495         gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
1496         (negtf2): Change to negtf2_internal.
1497         (negtf2): New expander.
1498         (abstf2): Enable for E500 double.  Call gen_spe_abstf2_tst,
1499         gen_spe_abstf2_cmp or gen_abstf2_internal depending on
1500         TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
1501         (movdi_internal32): Use rs6000_nonimmediate_operand.
1502         (unnamed splitter): Likewise.
1503         * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
1504         TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
1505         (SPE64TF, DITI): New mode macros.
1506         (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
1507         modes.
1508         (frob_tf_ti): New.
1509         (frob_<mode>_di_2): New.
1510         (frob_tf_di_8_2): New.
1511         (frob_di_df): Change to frob_di_<mode>; allow more modes.
1512         (frob_ti_tf): New.
1513         (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
1514         more modes.
1515         (frob_ti_<mode>_8_2): New.
1516         (frob_ti_tf_2): New.
1517         (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
1518         mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
1519         TFmode.
1520         (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
1521         mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
1522         (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
1523         spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
1524         spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
1525         (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
1526         tsttflt_gp): New.
1528 2007-01-23  Ian Lance Taylor  <iant@google.com>
1530         * Makefile.in (OBJS-common): Reformat, alphabetize, but put
1531         insn-*.o first.
1532         (OBJS-archive): Reformat, alphabetize.
1533         (OBJS): Change out_object_file to OBJS-md.
1535 2007-01-23  Uros Bizjak  <ubizjak@gmail.com>
1537         * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
1538         JUMP_P and CALL_P predicates where applicable.
1539         * config/i386/i386.c: Ditto.
1540         * config/i386/i386.md: Ditto.
1541         * config/i386/mmx.md: Ditto.
1542         * config/i386/predicates.md: Ditto.
1544 2007-01-22  Andreas Schwab  <schwab@suse.de>
1546         * config/m68k/m68k.h: Fix comment.
1548 2007-01-22  Jan Hubicka  <jh@suse.cz>
1550         * passes.c (init_optimization_passes): Do not rerun
1551         pass_early_warn_uninitialized.
1553 2007-01-22  Richard Guenther  <rguenther@suse.de>
1555         PR tree-optimization/30038
1556         * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
1557         function.
1558         (execute_cse_sincos_1): Likewise.
1559         (execute_cse_sincos): Likewise.
1560         (gate_cse_sincos): Likewise.
1561         (pass_cse_sincos): New pass CSEing sin() and cos() calls using
1562         the cexpi() canonicalization of sincos().
1563         * tree-pass.h (pass_cse_sincos): Declare.
1564         * passes.c (init_optimization_passes): New pass pas_cse_sincos.
1566 2007-01-21  Eric Botcazou  <ebotcazou@libertysurf.fr>
1568         PR rtl-optimization/29329
1569         * combine.c (replaced_rhs_insn): Rename to i2mod.
1570         (replaced_rhs_value): Rename to i2mod_new_rhs.
1571         (i2mod_old_rhs): New global variable.
1572         (combine_instructions): Adjust for above change.  Save a copy of
1573         the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
1574         note are substituted in the second instruction.
1575         (distribute_notes) <REG_DEAD>: Adjust for above change.  Do not
1576         ditch the note if it pertains to the second eliminated register
1577         and this register is mentioned in i2mod_old_rhs.
1579         Revert:
1580         2006-09-12  Eric Botcazou  <ebotcazou@libertysurf.fr>
1582         * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
1583         the insn to which the note was originally attached.
1585 2007-01-21  Jan Hubicka  <jh@suse.cz>
1587         * ipa-inline.c (inlining_mode): Comment, move up.
1588         (cgraph_decide_inlining_incrementally): Do not perform inlining
1589         itself; fix handling of flattening of self recursive functions.
1590         (cgraph_find_cycles): Remove.
1591         (cgraph_flatten_node): Remove.
1592         (cgraph_decide_inlining): Use incremental inliner to handle flattening.
1593         (try_inline): New function.
1594         (cgraph_early_inlining): Update call of
1595         cgraph_decide_inlining_incrementally.  Apply inlining here.
1596         (apply_inline): Update call of cgraph_decide_inlining_incrementally.
1598 2007-01-21  Dirk Mueller  <dmueller@suse.de>
1600         PR bootstrap/30511
1601         * tree-vrp.c (check_array_bounds): do not warn
1602         about ADDR_EXPR's of ARRAY_REF's which are immediately
1603         used in binary expressions.
1605 2007-01-21  Ira Rosen  <irar@il.ibm.com>
1607         * tree-vectorizer.h (struct _stmt_vec_info): Add new field
1608         read_write_dep and macros for its access.
1609         * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
1610         * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
1611         argument, call vect_check_interleaving for every independent pair of
1612         data-refs. Mark loads that access the same memory location as a store
1613         in the loop.
1614         (vect_check_dependences): Remove.
1615         (vect_analyze_data_ref_dependences): Remove  vect_check_dependences
1616         call, fix the call to vect_analyze_data_ref_dependence.
1617         (vect_analyze_data_ref_access): For statements that access the same
1618         data-ref, check that they are not stores; for loads, check that there
1619         is no store that access the same location.
1621 2007-01-20  Roger Sayle  <roger@eyesopen.com>
1622             Joseph Myers  <joseph@codesourcery.com>
1624         * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
1625         (-fdump-rtl-stack): Likewise.
1626         (-fno-signed-zeros): Use @minus{} for a minus sign.  Correct typo.
1627         (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
1628         (-mpt-fixed): Use @minus{} for minus sign.
1629         (Using Precompiled Headers): Fix under/overfull hbox.
1631 2007-01-20  Jan Hubicka  <jh@suse.cz>
1633         * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
1634         place within annotation.
1636 2007-01-20  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
1638         * pa.c (output_move_double): Change array size of xoperands to 4.
1640 2007-01-20  Jan Hubicka  <jh@suse.cz>
1642         * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
1643         as temporary.
1645 2007-01-19  Ian Lance Taylor  <iant@google.com>
1647         * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
1648         constant to unsigned HOST_WIDE_INT before negating.
1650 2007-01-19  Ian Lance Taylor  <iant@google.com>
1652         * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
1653         to unsigned.
1654         (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
1655         (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
1656         (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
1657         (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
1658         (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
1659         (realloc_vdef, realloc_vuse): Change second parameter to
1660         unsigned.
1661         (ssa_operand_iterator_d): Change vuse_index and mayuse_index
1662         fields to unsigned.
1663         * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
1664         unsigned. Change x and lim locals to unsigned.
1665         (realloc_vdef, realloc_vuse): Change num_elem parameter to
1666         unsigned.
1667         (finalize_ssa_vuse_ops): Change old_i local to unsigned.
1668         (copy_virtual_operands): Change i and n locals to unsigned.
1670 2007-01-19  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
1672         PR middle-end/29335
1673         * builtins.c (fold_builtin_1): Handle builtin fdim.
1675 2007-01-20  Jan Hubicka  <jh@suse.cz>
1677         * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
1678         * tree-ssa-structalias.c (compute_points_to_sets): Do call
1679         init_alias_heapvars.
1680         (init_alias_heapvars): Initialize only when not already initialized.
1681         (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
1683 2007-01-19  Roger Sayle  <roger@eyesopen.com>
1685         * common.opt (fsigned-zeros): New command line option.
1686         * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
1687         of flag_unsafe_math_optimizations.
1688         * opts.c (set_fast_math_flags): The -ffast-math command line option
1689         implies -fno-signed-zeros.
1690         (fast_math_flags_set_p): Likewise.
1692         * doc/invoke.texi: Document new -fno-signed-zeros option, and update
1693         the documentation of -ffast-math appropriately.  Wrap long lines.
1695 2007-01-19  Steve Ellcey  <sje@cup.hp.com>
1697         * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
1698         * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
1700 2007-01-19  Tomas Bily  <tbily@suse.cz>
1702         * cgraphunit.c (cgraph_finalize_function): Updating of pid
1703         * tree-profile.c:
1704         (tree_init_ic_make_global_vars): New function
1705         (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
1706         (tree_gen_ic_profiler): New function
1707         (tree_gen_ic_func_profiler): New function
1708         (tree_profiling): Added calling of tree_gen_ic_func_profiler
1709         (tree_profile_hooks): Added hook for indirec/virtual calls
1710         * value-prof.c (tree_find_values_to_profile): New case for
1711         indirect calls
1712         (tree_values_to_profile): Call for determining indirect/virtual
1713         counters
1714         (tree_indirect_call_to_profile): New function
1715         (tree_ic_transform): New function
1716         (tree_ic): New function
1717         (find_func_by_pid): New function
1718         (init_pid_map): New function
1719         (tree_value_profile_transformations): Added check for
1720         indirect/virtual call transformation
1721         * value-prof.h (enum hist_type): New counter type for
1722         indirect/virtual calls
1723         (profile_hooks): Added new hook for profiling indirect/virtual
1724         calls
1725         * profile.c (instrument_values): New case for indirect/virtual
1726         call added
1727         * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
1728         (GCOV_COUNTER_V_INDIR): New counter type
1729         (GCOV_COUNTER_NAMES): New name of counter "indirect" added
1730         (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
1731         call added
1732         * cgraph.c: Definition of cgraph_max_pid
1733         (cgraph_create_node): Default init of pid attribute
1734         * cgraph.h: Declaration of cgraph_max_pid
1735         (struct cgraph_node): Added pid attribute
1736         * libgcov.c (__gcov_indirect_call_profiler): New function
1737         (__gcov_one_value_profiler_body): New function
1738         (__gcov_one_value_profiler): Body was moved to
1739         __gcov_one_value_profiler_body and calls it
1741 2007-01-19  Basile Starynkevitch  <basile@starynkevitch.net>
1743         * doc/gty.texi (Options): Document the mark_hook option to GTY.
1744         * gengtype.c (write_types_data, write_func_for_structure,
1745         write_types, ggc_wtd, pch_wtd): Add skip_hooks to
1746         write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
1747         (walk_type, write_func_for_structure): Generate the mark_hook if
1748         needed.
1750 2007-01-19  Jan Hubicka  <jh@suse.cz>
1752         * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
1753         'early' argument take inlining mode argument specifying whether to
1754         inline for size/speeed or all functions; add support for flattening;
1755         improve dumpting.
1756         (cgraph_early_inlining): Update call of decide_inlining_incrementally.
1758 2007-01-19  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
1760         PR c++/17947
1761         * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
1762         the name of the declared identifier.
1764 2007-01-19  Dirk Mueller  <dmueller@suse.de>
1766         * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
1767         result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
1768         duplicate evaluation.
1770 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
1772         * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
1773         Macroize expander using X87MODEF12 mode macro.  Extend operand 1
1774         to XFMode, use acosxf2 and truncate result to requested mode.
1775         Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1776         (asin<mode>2): Similarly, with asin expanders.
1777         (*fscalexf4_i387): Rename from *fscalexf4.
1778         (expNcorexf3): New expander.
1779         (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
1780         (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
1781         X87MODEF12 mode macro.  Extend operand 1 to XFMode, use expxf2 and
1782         truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
1783         patterns for SSE math.
1784         (exp10<mode>2): Similarly, with exp10 expanders.
1785         (exp2<mode>2): Similarly, with exp2 expanders.
1786         (expm1<mode>2): Similarly, with expm1 expanders.
1787         (ldexp<mode>3): Similarly, with ldexp expanders.
1788         (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
1789         Use gen_truncxf<mode>2_i387_noop to truncate result.
1791 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1793         * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
1794         for ColdFire targets.
1796 2007-01-19  Nathan Sidwell  <nathan@codesourcery.com>
1797             Richard Sandiford  <richard@codesourcery.com>
1799         * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
1800         (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
1801         (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
1802         * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
1803         (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1804         * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1806 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1808         * config/m68k/m68k.md (adddi_dilshr32): Rename to...
1809         (*adddi_dilshr32): ...this.  Fix formatting.  Remove commented-out
1810         non-canonical pattern.  Restrict to !TARGET_COLDFIRE.
1811         (*adddi_dilshr32_cf): New pattern.
1812         (adddi3, subdi3): Remove first alternatives.
1814 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1816         * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
1817         implemented using move.l, do not use its cc result for floating-point
1818         comparisons.
1820 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1822         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
1823         from accepting 's' constraints if flag_pic.
1825 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1827         * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
1828         (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
1829         (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
1830         condition from TARGET_68881 to TARGET_HARD_FLOAT.
1832 2007-01-19  Sandra Loosemore  <sandra@codesourcery.com>
1834         * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
1835         ColdFire alternatives.
1836         * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
1837         * config/m68k/m68k.md (clzsi2):  Define for ColdFire
1838         architectures that support ff1 instruction.
1840 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1841             Julian Brown  <julian@codesourcery.com>
1843         * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
1844         * config/m68k/m68k.md (*movsi_cfv4): Fold into...
1845         (*movsi_cf): ...here.  Remove unnecessary 'R' from 'Rg'.
1846         Add commentary.
1848 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1850         * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
1851         and return a bool.
1852         (output_move_const_into_data_reg, output_move_simode_const): Delete.
1853         * config/m68k/m68k.c (const_method, const_int_cost): Take a
1854         HOST_WIDE_INT instead of an rtx.
1855         (m68k_rtx_costs): Update call accordingly.
1856         (output_move_const_into_data_reg): Likewise.  Fix formatting.
1857         (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
1858         Return a bool.
1859         (output_move_simode_const): Update calls after above changes.
1860         Rework to use automatic variables and predicates like MEM_P.
1861         * config/m68k/m68k.md (pushexthisi_const): Update call to
1862         valid_mov3q_const.
1864 2007-01-19  Dirk Mueller  <dmueller@suse.de>
1866         * tree-ssa-alias.c (perform_var_substitution): Fix typo
1867         in dump_flags test.
1869 2007-01-19  Richard Guenther  <rguenther@suse.de>
1871         * builtins.c (expand_builtin_cexpi): Fall back to expanding
1872         via cexp in case sincos is not available.
1874 2007-01-19  Richard Guenther  <rguenther@suse.de>
1876         * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
1877         * defaults.h (TARGET_HAS_SINCOS): Default to off.
1878         * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
1879         * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
1880         * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
1881         * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
1882         * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
1883         * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
1885 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
1887         * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
1888         New insn patterns.
1889         (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
1890         (atan2xf3): Directly generate RTL pattern.
1891         (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
1892         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
1893         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1894         disable patterns for SSE math.
1895         (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
1896         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
1897         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1898         disable patterns for SSE math.
1900 2007-01-19  Alexandre Oliva  <aoliva@redhat.com>
1902         * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
1904 2007-01-18  Roger Sayle  <roger@eyesopen.com>
1906         * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
1907         VIEW_CONVERT_EXPR to the same type as it's operand.
1909 2007-01-18  David Edelsohn  <edelsohn@gnu.org>
1911         * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
1912         configured for long double 128.
1914 2007-01-18  Mike Stump  <mrs@apple.com>
1916         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
1917         error.
1919 2007-01-18  Michael Meissner  <michael.meissner@amd.com>
1921         * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
1922         type correct.
1924 2007-01-18  Jan Hubicka  <jh@suse.cz>
1926         * tree-ssa-operands.c (vop_free_bucket_size): Never return value
1927         greater than NUM_VOP_FREE_BUCKETS.
1929 2007-01-18  Daniel Berlin  <dberlin@dberlin.org>
1931         * tree-ssa-structalias.c: Update comments.
1932         (ptabitmap_obstack): Removed.
1933         (pta_obstack): New.
1934         (oldpta_obstack): Ditto.
1935         (stats): Add a few members.
1936         (struct variable_info): Remove node, complex, address_taken, and
1937         indirect_target members. Add oldsolution member.
1938         (new_var_info): Do not initialize removed members.
1939         (constraint_expr_type): Remove INCLUDES.
1940         (constraint_graph): Add size, implicit_preds, rep,
1941         indirect_cycles, eq_rep, label, direct_nodes, and complex members.
1942         (FIRST_REF_NODE): New macro.
1943         (LAST_REF_NODE): Ditto.
1944         (FIRST_ADDR_NODE): Ditto.
1945         (find): New function.
1946         (unite): Ditto.
1947         (dump_constraint): Do not handle INCLUDES.
1948         (insert_into_complex): Do not insert duplicate constraints.
1949         (condense_varmap_nodes): Renamed and rewritten into ...
1950         (merge_node_constraints): This. Also fix bug in handling of
1951         offseted copy constraints.
1952         (clear_edges_for_node): No longer need to deal with preds at all,
1953         or removing associated preds/succs.
1954         (merge_graph_nodes): Deal with indirect_cycles.
1955         Don't deal with predecessors.
1956         (add_implicit_graph_edge): New function.
1957         (add_pred_graph_edge): Ditto.
1958         (add_graph_edge): Don't deal with predecessors.
1959         (build_constraint_graph): Removed.
1960         (build_pred_graph): New function.
1961         (build_succ_graph): Ditto.
1962         (struct scc_info): Removed in_component. Added roots, dfs, and
1963         node_mapping. Remove visited_index, unification_queue.
1964         (scc_visit): Deal with union-find we do now.
1965         Deal with cycles with REF nodes.
1966         (collapse_nodes): Renamed and rewritten to ...
1967         (unify_nodes): This.
1968         (process_unification_queue): Removed.
1969         (topo_visit): Cleanup
1970         (do_da_constraint): Use find.
1971         (do_sd_constraint): Ditto.
1972         (do_ds_constraint): Ditto.
1973         (do_complex_constraint): Ditto.
1974         (init_scc_info): Update for removed and added members.
1975         (find_and_collapse_graph_cycles): Renamed and rewritten into ...
1976         (find_indirect_cycles): This.
1977         (equivalence_class): New variable.
1978         (label_visit): New function.
1979         (perform_variable_substitution): Rewritten.
1980         (free_var_substitution_info): New function.
1981         (find_equivalent_node): Ditto.
1982         (move_complex_constraints): Ditto.
1983         (eliminate_indirect_cycles): Ditto.
1984         (solve_graph): Only propagate changed bits.
1985         Use indirect cycle elimination.
1986         Use find.
1987         (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
1988         (tree_id_eq): Renamed to ...
1989         (tree_vi_eq): This. Update for member change
1990         (insert_id_for_tree): Renamed and rewritten to ...
1991         (insert_vi_for_tree): This.
1992         (lookup_id_for_tree): Renamed and rewritten to ...
1993         (lookup_vi_for_tree): This.
1994         (get_id_for_tree): Renamed and rewritten to ...
1995         (get_vi_for_tree): Ditto.
1996         (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
1997         (process_constraint): Don't handle INCLUDES.
1998         Remove special ADDRESSOF case.
1999         (find_func_aliases): Rewrite to use vi functions instead of id
2000         ones.
2001         (create_function_info_for): Ditto.
2002         (create_variable_info_for): Ditto.
2003         (intra_create_variable_infos): Ditto.
2004         (merge_smts_into): Ditto.
2005         (find_what_p_points_to): Ditto.
2006         (init_base_vars): Ditto.
2007         (init_alias_vars): Ditto.
2008         (remove_preds_and_fake_succs): New function.
2009         (dump_sa_points_to_info): Dump new stats.
2010         (dump_solution_for_var): Use find.
2011         (set_used_smts): Fix formatting.
2012         (compute_points_to_sets): Updated for new functions.
2013         (ipa_pta_execute): Ditto.
2015 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
2016             Richard Sandiford  <richard@codesourcery.com>
2018         * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
2019         parallels.
2020         * calls.c (expand_call): If the return value is a PARALLEL,
2021         extract its first member.
2022         * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
2023         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2024         use FUNCTION_EXTRA_EPILOGUE.
2025         (m68k_function_value): Return a PARALLEL if the return value
2026         is of a pointer type.
2027         * config/m68k/netbsd-elf.h (current_function_returns_pointer)
2028         (FUNCTION_EXTRA_EPILOGUE): Remove.
2029         * config/m68k/m68k.md (D0_REG): New constant.
2031 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
2033         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2034         output a NOP for empty epilogues.
2036 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2038         * config/m68k/m68k.c (m68k_use_return_insn): Update comments
2039         before function.  Extend register save check to include all
2040         registers, not just integer ones.
2042 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
2044         * config/m68k/m68k-protos.h (use_return_insn): Rename to...
2045         (m68k_use_return_insn): ...this.
2046         * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
2047         * config/m68k/m68k.c (use_return_insn): Rename to...
2048         (m68k_use_return_insn): ...this.
2049         * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
2050         of USE_RETURN_INSN.
2052 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
2054         * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
2055         nearest even, fix denormal rounding overflow.
2057 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2059         * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
2060         f%$smove and f%$move.
2061         (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
2062         (extendsfdf2_cf): Use fdmove instead of f%&move.
2063         (truncdfsf2_cf): Use fsmove instead of f%$smove.
2064         (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
2066 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2068         * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
2069         GPR<-GPR moves.
2071 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2073         * real.h (real_format): Add a canonical_nan_lsbs_set field.
2074         (coldfire_single_format): Declare.
2075         (coldfire_double_format): Likewise.
2076         * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
2077         of qnan_msb_set to determine the lower bits of a canonical
2078         NaN significand.
2079         (encode_ieee_double): Likewise.
2080         (encode_ieee_quad): Likewise.
2081         (ieee_single_format): Initialize canonical_nan_lsbs_set.
2082         (mips_single_format): Likewise.
2083         (ieee_double_format): Likewise.
2084         (mips_double_format): Likewise.
2085         (ieee_extended_motorola_format): Likewise.
2086         (ieee_extended_intel_96_format): Likewise.
2087         (ieee_extended_intel_128_format): Likewise.
2088         (ieee_extended_intel_96_round_53_format): Likewise.
2089         (ibm_extended_format): Likewise.
2090         (mips_extended_format): Likewise.
2091         (ieee_quad_format): Likewise.
2092         (mips_quad_format): Likewise.
2093         (vax_f_format): Likewise.
2094         (vax_d_format): Likewise.
2095         (vax_g_format): Likewise.
2096         (i370_single_format): Likewise.
2097         (i370_double_format): Likewise.
2098         (decimal_single_format): Likewise.
2099         (decimal_double_format): Likewise.
2100         (decimal_quad_format): Likewise.
2101         (c4x_single_format): Likewise.
2102         (c4x_extended_format): Likewise.
2103         (real_internal_format): Likewise.
2104         (coldfire_single_format): New real_format.
2105         (coldfire_double_format): Likewise.
2106         * config/pdp11/pdp11.c (pdp11_f_format): Initialize
2107         canonical_nan_lsbs_set.
2108         (pdp11_d_format): Likewise.
2109         * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
2110         if TARGET_COLDFIRE_CPU.
2112 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2114         * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
2115         (output_call, m68k_legitimize_call_address): Declare.
2116         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
2117         parenthesees.  Add support for a 'W' constraint.
2118         (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
2119         (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
2120         (m68k_symbolic_call, m68k_symbolic_jump): Declare.
2121         * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
2122         variables.
2123         (override_options): Initialize them.  Do not set flag_no_function_cse
2124         for TARGET_ID_SHARED_LIBRARY.
2125         (m68k_output_pic_call): Delete.
2126         (m68k_legitimize_call_address): New function.
2127         (print_operand): Remove the %o prefix.  Handle the %p prefix.
2128         (output_call): New function.
2129         (m68k_output_mi_thunk): Use m68k_symbolic_jump.  Always load the
2130         target address from the GOT if symbolic jumps are not allowed.
2131         * config/m68k/m68k.md (call, general_operand): Do not set
2132         SYMBOL_REF_FLAG.  Use m68k_legitimize_call_address instead.
2133         Merge separate flag_pic and !flag_pic define_insns into...
2134         (*call, *call_value): ...these new patterns.  Match the address
2135         rather than the containing MEM and require it to be a call_operand.
2136         Use output_call to generate the asm template.
2137         * config/m68k/predicates.md (const_call_operand): New predicate.
2138         (call_operand): Likewise.
2140 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
2142         * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
2143         (STACK_GROWS_DOWNWARD): Define to 1.
2144         (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
2145         comments, emphasizing that these values are only defaults.
2146         * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
2147         * config/m68k/m68k.c (const_method): Remove trailing whitespace.
2149 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2151         * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
2153 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
2154             Richard Sandiford  <richard@codesourcery.com>
2156         * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
2157         (FP_REGNO_P): New macros.
2158         (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
2159         (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
2160         (EH_RETURN_STACKADJ_RTX): Use A0_REG.
2161         * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
2163 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
2165         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2166         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
2167         to $tmake_file.
2168         * config/m68k/t-floatlib: New file, extracting common code from...
2169         * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
2170         * config/m68k/t-uclinux: Here.
2171         * config/m68k/fpgnulib.c: Do not compile extendeed precision
2172         routines on ColdFire targets.
2174 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
2176         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2177         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
2178         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
2179         (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
2180         explicitly set MOTOROLA to 1.
2181         * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
2183 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2184             Nathan Sidwell  <nathan@codesourcery.com>
2186         * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
2187         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
2188         * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
2189         * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
2190         * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
2191         * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
2192         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
2193         (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
2194         (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
2195         * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2196         * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
2197         (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
2198         (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
2199         (DBX_REGISTER_NUMBER): Delete.
2200         * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
2201         (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2202         * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
2204 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2206         * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
2207         * config/flat.h: New file.
2208         * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
2209         OBJECT_FORMAT_FLAT.
2210         * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
2211         indirect references for -msep-data or -mid-shared-library.
2212         Do not use PC-relative code addresses either.
2213         * config/m68k/m68k.c (override_options): Restrict -fPIC error
2214         to -mpcrel.
2215         * config/m68k/uclinux.h (STARTFILE_SPEC): Define.  Use Scrt1.o
2216         for shared libraries and crt1.o for executables.  Use crti.o and
2217         crtbegin.o.
2218         (ENDFILE_SPEC): Use crtend.o and crtn.o.
2219         (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
2220         Do not add -elf2flt or -shared-lib-id options here.
2221         (LINK_SPEC): Define.  Pass -elf2flt if no -elf2flt option is given.
2222         Pass -shared-lib-id if -mid-shared-library, taking the library
2223         identifier from -mshared-library-id if given, otherwise
2224         defaulting to 0.
2225         (EH_FRAME_IN_DATA_SECTION): Do not undefine.
2226         (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
2227         (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
2228         and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
2229         (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
2230         * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
2231         and crtend.o.
2232         * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
2233         jump sequence for ISA A and ISA A+.
2234         (PICJUMP): Likewise.
2236 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2238         * config.gcc (m68k-*-uclinux*): Base the port on the common
2239         and m68k GNU/Linux files rather than on the generic ELF ones.
2240         * config/m68k/uclinux.h (TARGET_VERSION): Override.
2241         (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
2243 2007-01-18  Julian Brown  <julian@codesourcery.com>
2244             Richard Sandiford  <richard@codesourcery.com>
2246         * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
2247         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2248         * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
2249         before redefining.
2250         * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
2251         80 unconditionally.
2252         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2254 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2256         * doc/install.texi: Document m68k-uclinuxoldabi.
2257         * config.gcc (m68k-*-uclinuxoldabi): New configuration.
2258         * config/m68k/uclinux-oldabi.h: New file, copied from
2259         config/m68k/uclinux.h.
2261 2007-01-18  Kaz Kojima  <kkojima@rr.iij4u.or.jp>
2263         * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
2264         * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
2265         (m32r_compute_frame_size): Use unsigned for regno.
2266         (m32r_reload_lr): Use gen_frame_mem.
2267         (pop): New.
2268         (m32r_output_function_epilogue): Don't output the function epilogue
2269         textually here.
2270         (m32r_expand_epilogue): New.
2271         (direct_return): Return false if the function has the interrupt
2272         attribute.
2273         (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
2274         * config/m32r/m32r.md (epilogue): New expander.
2275         (return_lr, return_rte): New insns.
2276         (return): Make it expander.
2277         (return_normal): New expander.
2279 2007-01-18  Josh Conner  <jconner@apple.com>
2281         PR target/30485
2282         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
2283         support for UNLE, UNLT, UNGE, and UNGT.
2285 2007-01-18  Jan Hubicka  <jh@suse.cz>
2287         * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
2288         * tree-tailcall (add_virtual_phis): Likewise.
2289         (optimize_tail_call): Return TODOs.
2290         (execute_tail_calls): Return TODOs.
2291         * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
2292         * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
2293         changed.
2294         * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
2295         Cleanup cfg using TODO.
2296         * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
2297         * passes.c (execute_function_todo): When cleanup did something, remove
2298         unused locals.
2299         * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
2300         (make_edges): Don't cleanup_cfg.
2302 2007-01-18  Uros Bizjak  <ubizjak@gmail.com>
2304         * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
2305         UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
2306         operate on the top of stack.
2307         [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
2308         register handling of instructions that output to the second
2309         stack slot.
2310         [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
2311         (move_for_stack_reg): Special-case check for dead destination
2312         stack slot for constant load of 1.0 inside UNSPEC_TAN.
2314         * config/i386/i386.md (UNSPEC_TAN): New constant.
2315         (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
2316         (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
2317         to correctly model move of constant 1.0 to top stack slot.
2318         (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
2319         (unnamed peephole2 pattern): Remove corresponding peephole2
2320         pattern that optimizes tan insn and loading of constant 1.0.
2321         (tanxf2): Use fptanxf4_i387.
2322         (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
2323         patterns using X87MODEF12 mode macro.  Use fptan_extend<mode>xf4_i387
2324         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
2325         disable patterns for SSE math.
2326         (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
2327         (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
2329 2007-01-18  Dirk Mueller  <dmueller@suse.de>
2330             Richard Guenther <rguenther@suse.de>
2332         PR diagnostic/8268
2333         * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
2334         * common.opt (Warray-bounds): Add new warning option.
2335         * c-opts.c (c_common_handle_option): Define -Warray-bounds
2336         if -Wall is given.
2337         * Makefile.in: make tree-vrp.o depend on toplev.h
2338         * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
2339         is enabled.
2340         (check_array_refs, check_array_bounds, check_array_ref): New.
2342 2007-01-18  Jan Hubicka  <jh@suse.cz>
2344         * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
2345         (execute_ssa_ccp): Return flags conditionally.
2346         * tree-ssa-propagate.c (substitue_and_fold): Return if something was
2347         changed.
2348         * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
2350 2007-01-18  Steven Bosscher  <steven@gcc.gnu.org>
2352         * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
2353         the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
2355         * Makefile.c (GTFILES): Add cfglayout.h.
2356         * gengtype.c (open_base_files): Likewise.
2357         * cfglayout.c (cfg_layout_function_footer,
2358         cfg_layout_function_header) Reindent to make gengtype happy.
2359         * cfglayout.h (cfg_layout_function_footer,
2360         cfg_layout_function_header): Add GTY(()) marker.
2362         * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
2364 2007-01-18  Ben Elliston  <bje@au.ibm.com>
2366         * genautomata.c (write_automata): Include xstrerror output in the
2367         error message if writing the DFA description file fails.
2369 2007-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2371         * config/mips/mips-protos.h (mips_output_external): Make it
2372         return void.
2373         * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
2374         * config/mips/mips.c (irix_output_external_libcall): Likewise.
2375         (extern_list): Likewise.
2376         (extern_head): Likewise.
2377         (TARGET_ASM_FILE_END): Likewise.
2378         (mips_file_end): Likewise.
2379         (mips_output_external): Rewritten.
2381 2007-01-18  Ben Elliston  <bje@au.ibm.com>
2383         * genpreds.c (write_insn_preds_c): Only write out the function
2384         body for regclass_for_constraint if we have register constraints.
2386 2007-01-17  Tom Tromey  <tromey@redhat.com>
2388         * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
2389         * doc/install.texi (Testing): Use sourceware.org.
2390         (Binaries): Likewise.
2391         (Specific): Likewise.
2392         * doc/contrib.texi (Contributors): Use sourceware.org.
2394 2007-01-17  Anatoly Sokolov <aesok@post.ru>
2396         * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
2397         (AVR_ENHANCED): Rename to ...
2398         (AVR_HAVE_MUL): ... new.
2399         (avr_enhanced_p): Rename to ...
2400         (avr_have_mul_p): ... new.
2401         (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
2402         'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
2403         * config/avr/avr.c (avr_enhanced_p): Rename to ...
2404         (avr_have_mul_p): ... new.
2405         (base_arch_s): Rename 'enhanced' to 'have_mul'.
2406         (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
2407         'avr_enhanced_p' and 'enhanced'.
2408         (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
2409         AVR_HAVE_MUL instead of AVR_ENHANCED.
2410         * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
2411         (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
2412         *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
2413         instead of AVR_ENHANCED.
2414         (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
2415         * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
2416         (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
2418 2007-01-17  Ian Lance Taylor  <iant@google.com>
2420         * vec.h (VEC_reserve_exact): Define.
2421         (vec_gc_p_reserve_exact): Declare.
2422         (vec_gc_o_reserve_exact): Declare.
2423         (vec_heap_p_reserve_exact): Declare.
2424         (vec_heap_o_reserve_exact): Declare.
2425         (VEC_OP (T,A,reserve_exact)): New static inline function, three
2426         versions.
2427         (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
2428         negative parameter.
2429         (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
2430         (VEC_OP (T,A,copy)) [all versions]: Likewise.
2431         (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
2432         * vec.c (calculate_allocation): Add exact parameter.  Change all
2433         callers.
2434         (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
2435         (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
2436         (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
2437         (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
2438         (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
2439         (vec_heap_p_reserve_exact): New function.
2440         (vec_heap_o_reserve_exact): New function.
2442 2007-01-17  Jan Hubicka  <jh@suse.cz>
2444         * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
2446 2007-01-17  Eric Christopher  <echristo@apple.com>
2448         * config.gcc: Support core2 processor.
2450 2007-01-16  Jan Hubicka  <jh@suse.cz>
2452         * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
2453         the names, just unlink the chain so we don't crash on dangling pointers
2454         to dead SSA names.
2456 2007-01-16  Jan Hubicka  <jh@suse.cz>
2458         * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
2459         * tree-pass.h: Reorder to make IPA passes appear toegher.
2460         (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
2461         * cgraphunit.c (cgraph_finalize_function): Do not compute inling
2462         parameters, do not call early inliner.
2463         * ipa-inline.c: Update comments.  Include tree-flow.h
2464         (cgraph_decide_inlining): Do not compute inlining parameters.
2465         (cgraph_decide_inlining_incrementally): Return TODOs; assume to
2466         be called with function context set up.
2467         (pass_ipa_inline): Remove unreachable functions before pass.
2468         (cgraph_early_inlining): Simplify assuming to be called from the
2469         PM as local pass.
2470         (pass_early_inline): New pass.
2471         (cgraph_gate_ipa_early_inlining): New gate.
2472         (pass_ipa_early_inline): Turn into simple wrapper.
2473         (compute_inline_parameters): New function.
2474         (gate_inline_passes): New gate.
2475         (pass_inline_parameters): New pass.
2476         (apply_inline): Move here from tree-optimize.c
2477         (pass_apply_inline): New pass.
2478         * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
2479         transforming.
2480         * tree-inline.c (optimize_inline_calls): Return TODOs rather than
2481         doing them by hand.
2482         (tree_function_versioning): Do not allocate dummy struct function.
2483         * tree-inline.h (optimize_inline_calls): Update prototype.
2484         * tree-optimize.c (execute_fixup_cfg): Export.
2485         (pass_fixup_cfg): Remove
2486         (tree_rest_of_compilation): Do not apply inlines.
2487         * tree-flow.h (execute_fixup_cfg): Declare.
2488         * Makefile.in (gt-passes.c): New.
2489         * passes.c: Include gt-passes.h
2490         (init_optimization_passes): New passes.
2491         (nnodes, order): New static vars.
2492         (do_per_function_toporder): New function.
2493         (execute_one_pass): Dump current pass here.
2494         (execute_ipa_pass_list): Don't dump current pass here.
2496 2007-01-16  Janis Johnson  <janis187@us.ibm.com>
2498         * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
2499         (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
2501 2007-01-16  David Edelsohn  <edelsohn@gnu.org>
2503         * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
2504         (strong_alias): Define.
2505         (__gcc_qmul): Provide non-FMA for soft-float.
2506         (__gcc_qdiv): Same.
2507         (__gcc_qneg): New.
2508         (__gcc_qeq): New.
2509         (__gcc_qle): New.
2510         (__gcc_qge): New.
2511         (__gcc_qunord): New.
2512         (__gcc_stoq): New.
2513         (__gcc_dtoq): New.
2514         (__gcc_qtos): New.
2515         (__gcc_qtod): New.
2516         (__gcc_qtoi): New.
2517         (__gcc_qtou): New.
2518         (__gcc_itoq): New.
2519         (__gcc_utoq): New.
2520         (fmsub): New.
2521         * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
2522         soft-float functions.
2523         * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
2524         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
2525         about long double soft float.
2527 2007-01-16  Dorit Nuzman  <dorit@il.ibm.com>
2528             Tehila Meyzels  <tehila@il.ibm.com>
2530         * tree-vectorizer.h (is_pattern_stmt_p): New.
2531         * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
2532         formatting (tabs instead of spaces). Cleanup and clarify setting
2533         of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
2534         * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
2535         (vectorizable_type_demotion): Check that types are integral.
2536         (vectorizable_type_promotion): Likewise.
2537         (vectorizable_store): Fix typo.  Eliminate new-line at end of
2538         comments.
2540 2007-01-16  Jan Hubicka  <jh@suse.cz>
2542         * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
2543         cgraph edges too.
2545 2007-01-15  Eric Christopher  <echristo@apple.com>
2547         * ifcvt.c: Include vec.h, vecprim.h.
2548         (check_cond_move_block): New argument regs.
2549         Reorganize. Add registers used to regs.
2550         (cond_move_process_if_block): Use regs set above as
2551         loop bounds.
2553 2007-01-15  Eric Christopher  <echristo@apple.com>
2555         * config/darwin.h: Update copyright.
2556         (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
2557         (LINK_COMMAND_SPEC): Add -u.
2558         (LINK_SPEC): Fix umbrella for above.
2560 2007-01-15  Joseph S. Myers  <joseph@codesourcery.com>
2562         * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
2563         glibc CVS.
2565 2007-01-15  Tom Tromey  <tromey@redhat.com>
2567         * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
2568         * doc/install.texi (Testing): Don't mention jacks.
2569         (Configuration): Document --enable-java-maintainer-mode.  Move
2570         --with-java-home to libgcj-specific section.  Document
2571         --with-ecj-jar.
2572         (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
2574 2007-01-15  Jan Hubicka  <jh@suse.cz>
2576         * tree-ssa-dce.c (DCE_TODOs): New.
2577         (propagate_necessity): Return if something changed.
2578         (eliminate_unnecessary_stmts): Likewise.
2579         (perform_tree_ssa_dce): Return TODO flags when needed.
2580         (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
2582 2007-01-15  Uros Bizjak  <ubizjak@gmail.com>
2584         * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
2585         (fyl2x_extend<mode>xf3_i387): New insn pattern.
2586         (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
2587         patterns using X87MODEF12 mode macro.  Extend operand 1
2588         to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
2589         (log10<mode>2): Ditto.
2590         (log2<mode>2): Ditto.
2591         (log1p<mode>2): Ditto.
2592         (logb<mode>2): Ditto.
2593         (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
2594         (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
2595         (*fxtractxf3_i387): Rename from *fxtractxf3.
2596         (fxtract_extend<mode>xf3_i387): New insn pattern.
2597         (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
2598         * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
2599         and gen_fyl2xxf3_i387().
2601 2007-01-14  Zdenek Dvorak <dvorakz@suse.cz>
2603         * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
2604         * cfgloopmanip.c (fix_loop_placement): Made static.  Use
2605         get_loop_exit_edges.  Changed return type to bool.
2606         * cfgloop.h (fix_loop_placement): Declaration removed.
2608 2007-01-14  Dorit Nuzman  <dorit@il.ibm.com>
2610         * param.h (MIN_VECT_LOOP_BOUND): New.
2611         * params.def (MIN_VECT_LOOP_BOUND): New.
2612         * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
2613         argument - minimum threshold for number of iterations.
2614         * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
2615         argument to declaration.
2616         * tree-vect-analyze.c (vect_analyze_operations): Check value of
2617         MIN_VECT_LOOP_BOUND.
2618         * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
2619         slpeel_tree_peel_loop_to_edge with additional argument.
2620         (vect_do_peeling_for_alignment): Likewise.
2621         * doc/invoke.texi (min-vect-loop-bound): Document new param option.
2623 2007-01-14  Uros Bizjak  <ubizjak@gmail.com>
2625         PR target/30413
2626         * config/i386/i386.c (print_operand) ['z']: Output 'b' for
2627         operands of size 1.
2629 2007-01-14  Jan Hubicka  <jh@suse.cz>
2631         * tree-dfa.c (remove_referenced_var): New function.
2632         * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
2633         prune referenced vars list too.
2634         * tree-flow.h (remove_referenced_var): Declare.
2636 2007-01-14  Jan Hubicka  <jh@suse.cz>
2638         * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
2639         separately.
2640         (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
2641         (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
2642         * except.h (verify_eh_throw_table_statements): Kill prototype.
2643         * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
2644         into call exrepssions.
2645         * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
2646         eh trhow verifier.
2647         * tree-cfg.c: Include pointer-set.h.
2648         (verify_node_sharing): Work on pointer set.
2649         (verify_eh_throw_stmt_node): New.
2650         (verify_stmts): Use pointers sets, verify throw_stmt.
2652 2007-01-13  Zdenek Dvorak <dvorakz@suse.cz>
2654         * ipa-reference.c (analyze_function): Consider also addresses taken
2655         in phi nodes.
2657 2007-01-12  Roger Sayle  <roger@eyesopen.com>
2659         * c-typeck.c (null_pointer_constant_p): Replace use of
2660         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2661         (build_c_cast): Likewise.
2663 2007-01-12  Roger Sayle  <roger@eyesopen.com>
2665         * tree.h (force_fit_type_double): Remove unused final argument.
2666         * c-common.c (constant_expression_warning): Replace use of
2667         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2668         (convert_and_check): Likewise.
2669         (shorten_compare): Update call to force_fit_type_double.
2670         (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
2671         * convert.c (convert_to_pointer): Update call to
2672         force_fit_type_double.
2673         * fold-const.c (force_fit_type_double): Remove overflowed_const
2674         argument.
2675         (int_const_binop, fold_convert_const_int_from_int,
2676         fold_convert_const_int_from_real, fold_div_compare,
2677         fold_sign_changed_comparison, fold_unary, fold_negate_const,
2678         fold_abs_const, fold_not_const): Remove the final argument from
2679         calls to force_fit_type_double.
2681 2007-01-12  Andrew Pinski  <andrew_pinski@playstation.sony.com>
2683         * configure.ac: Set insn to "nop" for spu-*-* also.
2684         * configure: Regenerate.
2686 2007-01-12  Olga Golovanevsky  <olga@il.ibm.com>
2688         * builtins.def : Add BUILT_IN_FREE.
2690 2007-01-12  Jan Hubicka  <jh@suse.cz>
2692         PR tree-optimization/30443
2693         * tree-inline.c (tree_function_versioning): Do not optimize when
2694         cloning for inlining.
2696 2007-01-12  Zdenek Dvorak <dvorakz@suse.cz>
2698         * doc/loop.texi: Document recording of loop exits.
2699         * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
2700         (update_single_exits_after_duplication,
2701         update_single_exit_for_duplicated_loop,
2702         update_single_exit_for_duplicated_loops): Removed.
2703         (duplicate_loop_to_header_edge): Do not call
2704         update_single_exits_after_duplication and
2705         update_single_exit_for_duplicated_loops.
2706         (loop_version): Do not update single_exit information.
2707         (fix_loop_structure): Use record_loop_exits instead of
2708         mark_single_exit_loops.
2709         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
2710         the lists of loop exits.
2711         * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
2712         split_edge, merge_blocks): Update the lists of loop exits.
2713         * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
2714         loop_optimizer_init.
2715         * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
2716         of mark_single_exit_loops.
2717         (loop_optimizer_finalize): Call release_recorded_exits.
2718         * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
2719         LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
2720         * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
2721         update single exit information.
2722         * lambda-code.c (perfect_nestify): Ditto.
2723         * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
2724         (mark_single_exit_loops): Removed.
2725         (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
2726         get_exit_descriptions, rescan_loop_exit, record_loop_exits,
2727         dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
2728         functions.
2729         (get_loop_exit_edges, single_exit): Use recorded exit lists.
2730         (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
2731         (verify_loop_structure): Verify consistency of the exit lists.
2732         (flow_loops_find): Use alloc_loop.  Initialize exits hash.
2733         (set_single_exit): Removed.
2734         * cfgloop.h (struct loop_exit): New function.
2735         (struct loop): single_exit_ field replaced by exits field.
2736         (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
2737         (struct loops): Added exits hash.
2738         (mark_single_exit_loops, set_single_exit): Declaration removed.
2739         (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
2741 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2743         * doc/invoke.texi: Avoid use of @headitem.
2745 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2747         * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
2749 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2751         * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
2752         list of targets.
2754 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
2755             Richard Sandiford  <richard@codesourcery.com>
2757         * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
2758         * config/m68k/m68k.opt: Resort options.
2759         (mbitfield, mrtd, mshort): Remove RejectNegative properties.
2761 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
2762             Richard Sandiford  <richard@codesourcery.com>
2764         * doc/invoke.texi: Document the macros that are defined by
2765         m68k's -mtune and -mhard-float options.
2766         * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
2767         (m68k_cpp_cpu_family): Likewise.
2768         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
2769         of __ucfv*__ macros.  Define __mcffpu__ if generating code for
2770         ColdFire FPUs.  Define __mcf_cpu_* and __mcf_family_* macros.
2771         * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
2772         (m68k_cpp_cpu_family): Likewise.
2774 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2776         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
2777         macros besides mc68000 as tuning macros.  Use a switch statement
2778         to set them and mcpu32.
2780 2007-01-12  Julian Brown  <julian@codesourcery.com>
2782         * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
2783         (TARGET_68040_ONLY): Rename to...
2784         (TARGET_68040): ...this.
2785         * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
2786         * config/m68k/m68k.md: Likewise.
2788 2007-01-12  Julian Brown  <julian@codesourcery.com>
2789             Nathan Sidwell  <nathan@codesourcery.com>
2790             Richard Sandiford  <richard@codesourcery.com>
2792         * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
2793         the -mcpu= argument associated with the --with-cpu setting.
2794         Define M68K_DEFAULT_TUNE to the default -mtune= option,
2795         if different from the one implied by the -mcpu setting.
2796         Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
2797         using mcpu=FOO as the default CPU option.  Set target_cpu_default2.
2798         * doc/invoke.texi: Mention ColdFire in the introduction to the
2799         m68k options.  Document the new -march, -mcpu, -mtune, -mdiv,
2800         -mno-div and -mhard-float options.  Make -m68881 a synonym for
2801         -mhard-float.  Document the previously-undocumented -m5206e,
2802         -m528x, -m5307 and -m5407 options.  Tweak the existing option
2803         documentation for consistency.
2804         * doc/install.texi: Mention new --with-cpu arguments.
2805         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
2806         default CPU if neither -mcpu nor -march are specified.
2807         (ASM_CPU_SPEC): Pass down -mcpu and -march options.
2808         (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
2809         TARGET_ISA*.  Set the legacy __mcf*__ cpu macros in the same way,
2810         using m68k_tune to decide between families that implement the
2811         same ISA.  Use m68k_tune to set __mcfv4e__.
2812         (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
2813         (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
2814         (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
2815         (FL_ISA_C, FL_ISA_MMU): New macros.
2816         (MASK_COLDFIRE): Delete.
2817         (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
2818         (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
2819         (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
2820         (TARGET_HARD_FLOAT): Do not define here.
2821         (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
2822         (TUNE_68000): New macro.
2823         (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
2824         (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
2825         (TUNE_CFV2): Redefine in terms of m68k_tune.
2826         (uarch_type, target_device, fpu_type): New enums.
2827         (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
2828         * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
2829         (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
2830         (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
2831         (FL_FOR_isa_c): New macros.
2832         (m68k_isa): New enum.
2833         (m68k_target_selection): New structure.
2834         (all_devices, all_isas, all_microarchs): New tables.
2835         (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
2836         (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
2837         (MASK_ALL_CPU_BITS): Delete.
2838         (m68k_find_selection): New function.
2839         (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
2840         Map the legacy target options to a combination of the new ones.
2841         (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
2842         m68k_cpu_flags.  Handle M68K_DEFAULT_TUNE.  Use m68k_cpu_flags
2843         to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
2844         settings.
2845         * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
2846         (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
2847         properties.
2848         (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
2849         (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
2850         * config/m68k/m68k-devices.def: New file.
2852 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2853             Nathan Sidwell  <nathan@codesourcery.com>
2855         * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
2856         (SUBTARGET_EXTRA_SPECS): New macros.
2857         * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
2858         use %(asm_cpu_spec) instead.
2859         * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
2860         * config/m68k/openbsd.h (ASM_SPEC): Likewise.
2861         * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
2862         (EXTRA_SPECS): Rename to...
2863         (SUBTARGET_EXTRA_SPECS): ...this.
2865 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
2866             Richard Sandiford  <richard@codesourcery.com>
2867             Julian Brown  <julian@codesourcery.com>
2869         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2870         (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
2871         to the configuration's default CPU.
2872         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
2873         Remove default masks.
2874         (m680[012]0-*-*): Set the default with_cpu to the first part of
2875         the target name.
2876         (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
2877         (m68k*-*-linux): Extend the --with-cpu handling to...
2878         (m680[012]0-*-*, m68k*-*-*): ...these configurations.  Allow m68000
2879         and m68010.  Don't set target_cpu_default2.
2880         * doc/install.texi: Document --with-cpu for m68k.
2881         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
2882         * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
2883         (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
2884         (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
2885         (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
2886         (ASM_SPEC): Remove use of %(asm_cpu_default).
2887         (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
2888         * config/m68k/linux.h (TARGET_DEFAULT): Delete.
2889         (CPP_SPEC): Merge definitions.  Do not handle __HAVE_68881__ here.
2890         * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
2891         __HAVE_FPU__ if TARGET_HARD_FLOAT.
2892         (TARGET_DEFAULT): Delete.
2893         (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
2894         cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
2895         (CPP_CPU_SPEC): Delete.
2896         (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
2897         the appropriate string.
2898         (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
2899         (CPP_SPEC): Define to NETBSD_CPP_SPEC.
2900         (ASM_SPEC): Don't use %(asm_default_spec).
2901         * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
2902         TARGET_DEFAULT and add MASK_68881.
2903         * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
2905 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2907         * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
2908         (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
2909         MASK_68010 alongside MASK_68020.
2910         * doc/invoke.texi: Document -m68010.
2911         * config/m68k/m68k.opt (m68010): New.
2912         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
2913         if TUNE_68010.
2914         (TUNE_68010): New macro.
2915         * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
2916         (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
2917         (M68K_CPU_m68332): Add MASK_68010.
2918         * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
2919         fallback definition.
2920         * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
2921         defines.
2922         * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
2923         (m68k_handle_option): Handle OPT_m68010.  Add MASK_68010
2924         to all entries that use MASK_68020.
2925         (output_move_simode_const, output_move_himode, output_move_qimode)
2926         (output_move_stricthi, output_move_strictqi): Use TARGET_68010
2927         instead of TARGET_68020 to select clr behavior.  Remove comment
2928         about there being no TARGET_68010.
2929         * config/m68k/m68k.md: Likewise throughout.
2931 2007-01-12  Julian Brown  <julian@codesourcery.com>
2933         * config/m68k/m68k.h (TARGET_ISAB): New macro.
2934         * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
2935         * config/m68k/m68k.md: Likewise.
2937 2007-01-12  Julian Brown  <julian@codesourcery.com>
2939         * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
2940         TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
2942 2007-01-12  Julian Brown  <julian@codesourcery.com>
2944         * config/m68k/m68k.h (TUNE_68040_60): New macro.
2945         * config/m68k/m68k.c (standard_68881_constant_p): Use it.
2946         * config/m68k/m68k.md: Likewise.
2948 2007-01-12  Julian Brown  <julian@codesourcery.com>
2949             Richard Sandiford  <richard@codesourcery.com>
2951         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
2952         instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
2953         TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
2954         TARGET_CPU32.
2955         (TARGET_CPU32): Rename to...
2956         (TUNE_CPU32): ...this.
2957         (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
2958         (TUNE_CFV2): New macros.
2959         * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
2960         remove conditions that are implied by TARGET_68020.
2961         * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
2962         instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
2963         (m68k_output_function_epilogue): Likewise.
2964         (m68k_rtx_costs): Likewise.  Use TUNE_68060 instead of TARGET_68060
2965         and TUNE_CFV2 instead of TARGET_5200.  Use TUNE_68000_10 instead of
2966         "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
2967         non-68000 timings.  Refactor multiplication and division costs.
2968         (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
2969         TUNE_CPU32 instead of TARGET_CPU32.
2970         (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
2971         and TUNE_68060 instead of TARGET_68060.
2972         * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
2973         TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
2974         TARGET_CPU32.
2975         (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
2976         && !TARGET_COLDFIRE" to choose between moveq and clr.
2977         Likewise in the unnamed movsf pattern.
2978         (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
2979         "!TARGET_68020 && !TARGET_COLDFIRE".  Likewise the unnamed
2980         ashiftrt pattern.
2982 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2984         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
2985         of tabbing before backslashes.
2987 2007-01-11  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
2989         * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
2990         * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
2991         (ASM_OUTPUT_INTERNAL_LABEL): Define.
2993 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
2995         * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
2996         find_interesting_uses_cond.
2997         (find_interesting_uses_cond): Use extract_cond_operands.
2998         (rewrite_use_compare): Use extract_cond_operands and
2999         force_gimple_operand_bsi.  Do not call update_stmt.
3000         (determine_use_iv_cost_condition): Use extract_cond_operands.
3001         Return cheaper of using original bound and changing the exit bound.
3003 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
3005         PR tree-optimization/29516
3006         * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
3007         most_expensive_mult_to_index, addr_to_parts,
3008         create_mem_ref, maybe_fold_tmr): Make the type of
3009         fields of TARGET_MEM_REF sizetype.
3010         (move_fixed_address_to_symbol, move_pointer_to_base):
3011         New functions.
3012         * tree.def (TARGET_MEM_REF): Add comment on types of
3013         the operands.
3015 2007-01-11  Joseph Myers  <joseph@codesourcery.com>
3017         * c-common.c (vector_types_convertible_p): Treat opaque types as
3018         always convertible if they have the same size, but not otherwise.
3020 2007-01-11  Steven Bosscher  <steven@gcc.gnu.org>
3022         * ifcvt.c (struct noce_if_info): Add comments to the fields.
3023         Remove the b_unconditional field.
3024         (noce_try_sign_mask): Do not look at b_unconditional.
3025         (noce_process_if_block): Do not use merge_if_blocks.  Update
3026         the CFG here.  Do not set b_unconditional.
3027         (cond_move_process_if_block): Likewise.
3028         (find_cond_trap): Likewise.
3029         (check_cond_move_block): Require simple jump insns at the end
3030         of the basic block.
3032 2007-01-11  Jan Hubicka  <jh@suse.cz>
3034         PR tree-optimization/1046
3035         * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
3036         when alias info is not ready.
3037         (pass_tail_recursion): Do not require aliasing.
3038         * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
3039         * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
3040         * tree-ssa-copy.c (pass_copy_prop): Likewise.
3041         * tree-ssa-forwprop.c (pass_forwprop): Likewise.
3042         * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
3043         * passes.c (init_optimization_passes): Execute rename_ssa_copies,
3044         ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
3045         before inlining.
3046         * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
3047         When aliasing is not build, mark statement as volatile.
3049 2007-01-11  Tom Tromey  <tromey@redhat.com>
3051         PR preprocessor/15185, PR preprocessor/20989:
3052         * doc/cppopts.texi <-MT>: Update description of algorithm for
3053         computing default target.
3054         <-M, -MD>: Reword "basename" text.
3056 2007-01-11  Roger Sayle  <roger@eyesopen.com>
3058         * builtins.c (expand_builtin_pow, expand_builtin_powi,
3059         fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
3060         fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
3061         fold_builtin_int_int_roundingfn, fold_builtin_bitop,
3062         fold_builtin_bswap, real_constp, fold_builtin_pow,
3063         fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
3064         do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
3065         uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3066         * convert.c (convert_to_pointer): Likewise.
3067         * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
3068         * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
3069         const_binop, fold_convert_const_int_from_int,
3070         fold_convert_const_int_from_real,
3071         fold_convert_const_real_from_real, sign_bit_p,
3072         optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
3073         fold_sign_changed_comparison, fold_unary, fold_comparison,
3074         fold_binary, multiple_of_p, tree_Expr_non_zero_p,
3075         fold_negate_const, fold_abs_const, fold_not_const): Likewise.
3076         * print-tree.c (print_node_brief, print_node): Likewise.
3077         * stor-layout.c (place_field, layout_type): Likewise.
3078         * tree-chrec.c (keep_cast): Likewise.
3079         * tree.c (build_vector, build_real, build_real_from_int_cst,
3080         build_complex): Likewise.
3082 2007-01-11  Roger Sayle  <roger@eyesopen.com>
3084         * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete.  For the time being,
3085         treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
3087 2007-01-11  Paolo Bonzini  <bonzini@gnu.org>
3089         * configure.ac (strict1_warn): Rename to strict_warn.
3090         (WERROR, --enable-werror, symlink hacks, stage1_cflags,
3091         cc_set_by_configure, quoted_cc_set_by_configure,
3092         stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
3093         all_boot_languages, all_stagestuff): Remove.
3094         (target_list): Remove bootstrap targets.
3095         * Makefile.in (quickstrap): Unconditionally make a synonym of all.
3096         (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
3097         REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
3098         SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
3099         VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
3100         STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
3101         stage1_copy, stage2_build, stage2_copy, stageprofile_build,
3102         stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
3103         stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
3104         bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
3105         bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
3106         unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
3107         restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
3108         cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
3109         *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
3110         stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
3111         stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
3112         risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
3113         (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
3114         (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
3115         (mostlyclean): Adjust.
3116         (clean, distclean): Don't mention bootstrap stuff.
3117         * configure: Regenerate.
3118         * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
3119         java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
3120         treelang/config-lang.in (stagestuff): Remove.
3121         * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
3122         stageprofile, stagefeedback, stagestuff): Remove mention.
3124 2007-01-11  Nick Clifton  <nickc@redhat.com>
3126         * config/mcore/predicates.md (mcore_general_movesrc_operand):
3127         Accept CONSTs.
3128         (mcore_general_movdst_operand): Do not accept CONST_INTs.
3129         (mcore_arith_K_S_operand): Run the test for the S constraint not
3130         the test for the M constraint.
3131         (mcore_addsub_operand): Do not accept integer values that are
3132         larger than 32 bits.
3133         * config/mcore/mcore.md: Remove unused constraints from split.
3134         (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
3135         (addsi3): Likewise.
3136         (allocate_stack): Likewise.
3137         * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
3138         operands to 32 bits.
3139         (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
3140         INTVAL.
3141         (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
3142         mcore_const_ok_for_inline, mcore_const_trick_uses_not,
3143         try_constant_tricks, mcore_num_ones, mcore_num_zeros,
3144         mcore_output_bclri, mcore_output_andn, output_inline_const,
3145         mcore_output_move, mcore_output_movedouble): Likewise.
3146         (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
3147         (output_inline_const): Likewise.
3148         (output_inline_const): Fix format strings used in sprintf
3149         statements.
3150         * config/mcore/mcore-protos.h: Update prototypes for changed
3151         functions in mcore.c.
3152         * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
3153         HOST_WIDE_INT and not int.
3154         (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
3155         CONST_OK_FOR_N): Likewise.
3156         (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
3157         (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
3158         an INTVAL.
3160 2007-01-10  Jan Hubicka  <jh@suse.cz>
3162         * tree-vrp.c (remove_range_assertions): Release defs.
3163         * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
3164         * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
3166 2007-01-10  Paul Brook  <paul@codesourcery.com>
3168         * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
3169         * config/arm/arm.md (mulsi3subsi): New insn.
3171 2007-01-10  Zdenek Dvorak <dvorakz@suse.cz>
3173         * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
3174         (tree_transform_and_unroll_loop): New.
3175         * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
3176         Declare.
3178 2007-01-10  Robert Kennedy <jimbob@google.com>
3180         * fold-const.c (fold_comparison): Fold comparisons like (x *
3181         1000 < 0) to (x < 0).
3183 2007-01-10  Ian Lance Taylor  <iant@google.com>
3185         * tree-pretty-print.c (dump_generic_node): Print parentheses when
3186         operands have the same priority.
3188 2007-01-10  Tom Tromey  <tromey@redhat.com>
3190         * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
3191         (fold_binary): Likewise.
3192         * langhooks.c (lhd_can_use_bit_fields_p): Removed.
3193         * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
3194         (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
3195         (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
3196         * langhooks.h (struct lang_hooks): Removed field
3197         'can_use_bit_fields_p'.
3199 2007-01-10  Ralf Corsépius <ralf.corsepius@rtems.org>
3201         * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
3203 2007-01-10  Razya Ladelsky  <razya@il.ibm.com>
3205         * function.c (get_last_funcdef_no): New function.
3206         * function.h (get_last_funcdef_no): Declare.
3207         * tree-inline.c (initialize_cfun): Add initialization.
3208         (tree_function_versioning): Cleanup.
3210 2007-01-10  Jan Hubicka  <jh@suse.cz>
3212         * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
3213         PHIs.
3215 2007-01-10  Sa Liu  <saliu@de.ibm.com>
3216             Ben Elliston  <bje@au.ibm.com>
3218         * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
3219         for SAVE_NONLOCAL and Pmode for any other save level.
3220         * spu-protos.h (spu_restore_stack_block): Declare.
3221         * spu.md (save_stack_block): Remove.
3222         (restore_stack_block): Call spu_restore_stack_block.
3223         * spu.c (spu_restore_stack_block): New function.
3224         (spu_expand_epilogue): Remove old comment.
3226 2007-01-09  Zdenek Dvorak <dvorakz@suse.cz>
3228         PR tree-optimization/30322
3229         * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
3230         (cand_value_at): Return the value as aff_tree.
3231         (may_eliminate_iv): Convert the bound from aff_tree to tree.
3232         * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
3233         aff_combination_mult): New functions.
3234         (aff_combination_add): Use aff_combination_add_cst.
3235         (aff_combination_convert): Allow conversions to a wider type.
3236         (tree_to_aff_combination): Handle BIT_NOT_EXPR.
3237         * tree-affine.h (aff_combination_mult): Declare.
3239 2007-01-09  Carlos O'Donell  <carlos@codesourcery.com>
3241         * doc/tm.texi: Update documentation to reflect reality of exec
3242         and start file search behaviours. Update copyright year.
3243         * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
3244         header file directories.
3246 2007-01-09  Uros Bizjak  <ubizjak@gmail.com>
3248         * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
3249         (*cosxf2): Rename to cosxf2_i387.
3250         (*sindf2, *sinsf2): Extend operand 1 to XFmode.  Macroize patterns
3251         using X87MODEF12 mode macro. Rename patterns to
3252         *sin_extend<mode>xf2_i387.  Use SSE_FLOAT_MODE_P to disable patterns
3253         for SSE math.
3254         (*cosdf2, *cossf2): Ditto.
3255         (sincosdf3, sincossf3): Ditto.  Rewrite corresponding splitters
3256         to match extended input operands.
3257         (sincos<mode>3): New expander.
3258         (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
3259         insn patterns and corresponding splitters.
3261 2007-01-09  Kaz Kojima  <kkojima@gcc.gnu.org>
3263         * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
3264         (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
3266 2007-01-09  Nicolas Pitre  <nico@cam.org>
3268         PR target/30173
3269         * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
3271 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
3273         * target.h (struct gcc_target): New field library_rtti_comdat.
3274         * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
3275         (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
3276         * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
3277         * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
3279 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
3281         * doc/invoke.texi (Optimize Options): Correct description of -O0.
3283 2007-01-08  Richard Guenther  <rguenther@suse.de>
3285         * tree.h (force_fit_type_double): Export.
3286         (force_fit_type): Remove.
3287         * fold-const.c (force_fit_type_double): New function.
3288         (force_fit_type): Remove.
3289         (int_const_binop): Use it.
3290         (fold_convert_const_int_from_int): Likewise.
3291         (fold_convert_const_int_from_real): Likewise.
3292         (fold_div_compare): Likewise.
3293         (fold_sign_changed_comparison): Likewise.
3294         (fold_unary): Likewise.
3295         (fold_negate_const): Likewise.
3296         (fold_abs_const): Likewise.
3297         (fold_not_const): Likewise.
3298         * c-common.c (shorten_compare): Use force_fit_type_double.
3299         * convert.c (convert_to_pointer): Likewise.
3301 2007-01-08  Richard Guenther  <rguenther@suse.de>
3303         * tree.h (build_int_cst_wide_type): Export.
3304         * tree.c (build_int_cst_wide_type): New function.
3305         (build_int_cst_wide): Fix comment.
3306         * builtins.c (fold_builtin_object_size): Use build_int_cst
3307         to build -1 or 0 of the correct type.  Use fit_double_type
3308         to check for overflow.
3309         * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
3310         to build the mask.
3311         (decode_field_reference): Likewise.
3312         (all_ones_mask_p): Likewise.
3313         (native_interpret_int): Use build_int_cst_wide_type.
3314         (fold_binary): Use build_int_cst_type to build an all-ones
3315         value.
3316         * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
3318 2007-01-08  Daniel Jacobowitz  <dan@codesourcery.com>
3320         * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
3322 2007-01-09  Ben Elliston  <bje@au.ibm.com>
3324         * genautomata.c (STATS_OPTION): New option.
3325         (stats_flag): New flag.
3326         (gen_automata_option): Handle it.
3327         (initiate_automaton_gen): Ditto.
3328         (write_automata): Output statistics only if stats_flag is
3329         set. Likewise, output time statistics only if time_flag is set.
3330         * doc/md.texi (Processor pipeline description): Document new flag.
3332 2007-01-08  Richard Guenther  <rguenther@suse.de>
3334         * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
3335         * tree.c (build_int_cst_type): Likewise.
3336         (size_in_bytes): Don't call force_fit_type on the result.
3337         (int_fits_type_p): Use fit_double_type.
3338         * fold-const.c (fit_double_type): New function.
3339         (force_fit_type): Use it.
3340         * tree.h (fit_double_type): Export.
3342 2007-01-08  Jan Hubicka  <jh@suse.cz>
3344         * tree-vectorizer.c (gate_increase_alignment): Fix return type.
3345         * ipa.c (function_and_variable_visibility): Fix return type.
3347 2007-01-08  Richard Guenther  <rguenther@suse.de>
3349         * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
3350         of offset to build the index.
3351         * tree-pretty-print.c (dump_generic_node): Don't build negated
3352         const just for printing.
3353         * c-pretty-print.c (pp_c_integer_constant): Likewise.
3354         * builtins.c (fold_builtin_int_roundingfn): Check if result
3355         fits the type by using force_fit_type and comparing the result.
3356         * predict.c (predict_loops): Use compare_tree_int for comparison.
3357         * tree.c (build_int_cst): Fall back to integer_type_node for
3358         NULL_TREE type.
3359         (build_int_cst_wide): Assert type is non-null.
3361 2007-01-08  Roberto Costa  <roberto.costa@st.com>
3363         * tree-vrp.c (extract_range_from_cond_expr): New.
3364         (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
3365         * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
3366         as expressions.
3367         (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
3368         * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
3369         tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
3370         used as expressions.
3371         * tree-object-size.c (cond_expr_object_size): New.
3372         (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
3374 2007-01-08  Jan Hubicka  <jh@suse.cz>
3376         * tree-ssa-forwprop.c (forward_propagate_into_cond,
3377         tree_ssa_forward_propagate_single_use_va): Release defs of propagated
3378         statement.
3380 2007-01-08  Richard Guenther  <rguenther@suse.de>
3382         PR tree-optimization/23603
3383         * tree-vrp.c (set_value_range_to_truthvalue): New function.
3384         (extract_range_from_binary): Fall back to truthvalue instead of
3385         varying for TRUTH_*_EXPR.
3386         (extract_range_from_comparison): Fall back to truthvalue instead of
3387         varying.
3388         (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
3389         if all visited PHI values were constant.
3391 2007-01-08  Jan Hubicka  <jh@suse.cz>
3393         * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
3394         (cgraph_analyze_function): break out from ...
3395         (cgraph_finalize_compilation_unit): ... here.
3396         (cgraph_expand_function): Remove forgoten commented out line.
3397         (cgraph_optimize): Analyze functions.
3399 2007-01-08  Jan Hubicka  <jh@suse.cz>
3401         * tree-pas.h (TODO_remove_function): New flag.
3402         (TODO_update*): Renumber.
3403         (pass_ipa_increase_alignment,
3404         pass_ipa_function_and_variable_visibility): New passes.
3405         * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
3406         (cgraph_function_and_variable_visibility): Move to ipa.c
3407         (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
3408         cgraph_increase_alignment.
3409         * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
3410         (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
3411         calling tree-inline.
3412         (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
3413         (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
3414         * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
3415         (gate_increase_alignment): New function.
3416         (pass_ipa_increase_alignment): New pass.
3417         * ipa.c: Inline tree-pass.h and timevar.h
3418         (function_and_variable_visibility): Move here from cgraphunit.c
3419         * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
3420         * passes.c (init_optimization_passes): Add the two new passes.
3421         (execute_todo): Handle cgraph_remove_functions.
3423 2007-01-08  Nick Clifton  <nickc@redhat.com>
3425         * config/frv/predicates.md (reg_or_0_operand): Accept
3426         CONST_DOUBLEs.
3428 2007-01-08  Ralf Corsépius <ralf.corsepius@rtems.org>
3430         * config/bfin/rtems.h, config/bfin/t-rtems: New.
3431         * config.gcc: Add bfin*-rtems*.
3433 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
3435         * c.opt: Add -flax-vector-conversions.
3436         * c-typeck.c (convert_for_assignment): Pass flag to
3437         vector_types_convertible_p to allow emission of note.
3438         (digest_init): Likewise.
3439         * c-opts.c: Handle -flax-vector-conversions.
3440         * c-common.c (flag_lax_vector_conversions): New.
3441         (vector_types_convertible_p): Unless -flax-vector conversions
3442         has been passed, disallow conversions between vectors with
3443         differing numbers of subparts and/or element types.  If such
3444         a conversion is disallowed, possibly emit a note on the first
3445         occasion only to inform the user of -flax-vector-conversions.
3446         The new last argument specifies this.
3447         * c-common.h (flag_lax_vector_conversions): New.
3448         (vector_types_convertible_p): Add extra argument.
3449         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
3450         char_type_node for V*QI type vectors.
3451         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
3452         Update to satisfy new typechecking rules.
3453         * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
3454         C and C++ variants.
3455         * doc/invoke.texi (C Dialect Options): Document
3456         -flax-vector-conversions.
3458 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
3460         PR tree-optimization/29877
3461         * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
3462         a register variable on the RHS to not be replaceable.
3464 2007-01-08  Chen Liqin  <liqin@sunnorth.com.cn>
3465         * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
3466         * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
3467         const_simm15, const_pow2, const_npow2): Added.
3468         * config/score/misc.md (insv, extv, extzv, movmemsi,
3469         move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
3470         * config/score/score.c (score_address_cost, score_select_cc_mode):
3471         Added.
3472         Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
3473         Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
3474         Update score_print_operand.
3475         * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
3476         Adjust register allocate order and update some macro define.
3477         * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
3478         mdx_block_move_straight, mdx_block_move_loop_head,
3479         mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
3480         mdx_block_move): Added.
3481         (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
3482         mdp_limm): Updated and fix some bug and typo.
3483         * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
3484         (movsf, movdf, doloop_end): Added.
3486 2007-01-08  Kazu Hirata  <kazu@codesourcery.com>
3488         * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
3489         config/arm/thumb2.md: Fix comment typos.
3490         * doc/extend.texi: Fix a typo.
3492 2007-01-07  Eric Christopher  <echristo@apple.com>
3494         * configure.ac: Check for __stack_chk_fail for darwin.
3495         * configure: Regenerate.
3497 2007-01-07  Richard Guenther  <rguenther@suse.de>
3499         * tree-vrp.c (extract_range_from_assert): CSE calls to
3500         compare_values where possible.
3501         (extract_range_from_unary_expr): Likewise.
3503 2007-01-07  Anatoly Sokolov <aesok@post.ru>
3505         * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
3506         * config/avr/avr.c (call_insn_operand): Delete function.
3507         * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
3508         instead of register number. Use predicates.
3509         * config/avr/predicates.md (const0_operand, reg_or_0_operand,
3510         call_insn_operand): Add.
3512 2007-01-06  Jan Hubicka  <jh@suse.cz>
3514         * tree-pass.h (pass_build_cgraph_edges): Declare.
3515         * cgraphunit.c (record_refernece): Move to cgraphbuild.c
3516         (visited_nodes): Remove.
3517         (cgraph_create_edges): Move to cgraphbuild.c; rename to
3518         build_cgrpah_edges; make visited_nodes local.
3519         (cgraph_process_new_functions): DO not call initialize_inline_failed.
3520         (record_references_in_initializer): Move to cgraphbuild.c
3521         (initialize_inline_failed, rebuild_cgraph_edges,
3522         pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
3523         (verify_cgraph_node): Make visited_nodes local.
3524         (cgraph_analyze_function): Do not call cgraph_create_edges and
3525         initialize_inline_failed.
3526         (cgraph_expand_function): Do not call cgraph_lower_function;
3527         assert that function is already lowered.
3528         * Makefile.in (cgraphbuild.o): New.
3529         * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
3530         at the end of lowering passes.
3532 2007-01-06  Steven Bosscher  <steven@gcc.gnu.org>
3534         * ifcvt.c (cond_move_convert_if_block): New function, code
3535         factored out from...
3536         (cond_move_process_if_block): ...here.  Call the new function
3537         on the THEN and ELSE blocks.
3538         (merge_if_block): Do not copy global_live_at_end, merge_blocks
3539         already takes care of this.
3541 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
3543         PR c/19978
3544         * tree.h (TREE_OVERFLOW_P): New.
3545         * c-typeck.c (parser_build_unary_op): Warn only if result
3546         overflowed and operands did not.
3547         (parser_build_binary_op): Likewise.
3548         (convert_for_assignment): Remove redundant overflow_warning.
3549         * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
3551 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
3553         * c-typeck.c (store_init_value): Split over two lines to follow
3554         the GNU coding style.
3556 2007-01-05  Benjamin Kosnik  <bkoz@redhat.com>
3558         * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
3559         __GXX_EXPERIMENTAL_CXX0X__.
3560         * doc/cpp.texi: Same.
3562 2007-01-05  Richard Guenther  <rguenther@suse.de>
3564         PR middle-end/27826
3565         * tree.c (get_narrower): Do not construct COMPONENT_REFs
3566         with mismatched types.  Instead explicitly build a
3567         conversion NOP_EXPR.
3569 2007-01-05  Ian Lance Taylor  <iant@google.com>
3571         * c-common.c (decl_with_nonnull_addr_p): New function.
3572         (c_common_truthvalue_conversion): Call it.
3573         * c-typeck.c (build_binary_op): Likewise.
3574         * c-common.h (decl_with_nonnull_addr_p): Declare.
3576 2007-01-05  Jakub Jelinek  <jakub@redhat.com>
3578         PR c/30360
3579         * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
3580         denom against 0.0.
3582 2007-01-05  Joel Brobecker  <brobecker@adacore.com>
3584         * doc/install.texi (Final install): Document the fact that
3585         the GNAT runtime should not be stripped.
3587 2007-01-04  Jan Hubicka  <jh@suse.cz>
3589         * tree-inline.c (fold_marked_statements): Update operand caches
3590         and EH after folding
3592 2007-01-04  Ian Lance Taylor  <iant@google.com>
3594         * c-common.c (check_function_nonnull): Whitespace fix.
3596 2007-01-04  Jan Hubicka  <jh@suse.cz>
3598         * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
3599         comitted older version of patch.
3600         (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
3601         function, celanup cfg and collect garbage.
3603 2007-01-04  Mike Stump  <mrs@apple.com>
3605         * Makefile.in (mostlyclean): Don't remove libgcc anymore.
3606         (clean): Likewise.
3608 2007-01-04  Eric Christopher  <echristo@apple.com>
3610         * libgcc2.c (__bswapsi2): Use SItype.
3611         (__bswapdi2): Use DItype.
3612         * libgcc2.h: Update for above.
3614 2007-01-04  Paul Brook  <paul@codesourcery.com>
3616         * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
3617         mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
3618         mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
3619         umulsidi3, umulsidi3adddi, smulsi3_highpart,
3620         umulsi3_highpart): Make conditional on !arm_arch6.
3621         (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
3622         mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
3623         mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
3624         umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
3625         umulsi3_highpart_v6): New insns.
3627 2007-01-04  Roger Sayle  <roger@eyesopen.com>
3629         * fold-const.c (fold_convert): When casting an expression to void,
3630         fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
3631         have a type.  Instead of attempting to build a NOP_EXPR, return
3632         these "special" trees directly.
3634 2007-01-04  Joseph Myers  <joseph@codesourcery.com>
3636         * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
3637         MULT inside MINUS as either argument.  Use rs6000_cost->dmul -
3638         rs6000_cost->fp not 0 as adjustment for outer NEG.
3640 2007-01-04  Jan Hubicka  <jh@suse.cz>
3642         * cgraph.c (cgraph_release_function_body): New function.
3643         (cgraph_remove_node): Use it.
3644         * cgraph.h (cgraph_release_function_body): Declare.
3645         * cgraphunit.c (cgraph_expand_function): Use it.
3646         * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
3647         * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
3648         is initialized and while compilation of other function is running.
3649         * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
3650         statement CFG annotations from here to ...
3651         * tree-cfg.c (delete_tree_cfg_annotations): ... here.
3653 2007-01-04  Zdenek Dvorak <dvorakz@suse.cz>
3655         * cfgloop.h (enum li_flags): Make the constants powers of two.
3657 2007-01-04  Jan Hubicka  <jh@suse.cz>
3659         * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
3660         set.
3661         (fold_marked_statements): New function.
3662         (optimize_inline_calls, tree_function_versioning): Fold new statements.
3663         * tree-inline.h (copy_body_data): Add statements_to_fold.
3665 2007-01-03  Daniel Jacobowitz  <dan@codesourcery.com>
3667         * config.gcc: Mention libgcc/config.host.
3668         * Makefile.in: Update comments mentioning libgcc.
3669         (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
3670         (all.cross, start.encap, rest.encap, rest.cross): Update
3671         dependencies for libgcc move.
3672         (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
3673         (clean-target, clean-target-libgcc): Delete.
3674         (srcdirify, GCC_EXTRA_PARTS): New macros.
3675         (libgcc-support, libgcc.mvars): New rules.
3676         (distclean): Remove mention of mklibgcc.
3677         (install): Don't reference INSTALL_LIBGCC.
3678         (install-common): Don't reference EXTRA_PARTS.
3679         (install-libgcc, install-multilib): Delete rules.
3680         * mklibgcc.in: Delete file.
3681         * doc/configfiles.texi: Don't mention mklibgcc.
3683         * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
3684         * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
3685         * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
3686         * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
3687         (mklibgcc, ldblspecs): Likewise.
3689         * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
3690         (posixpre.def): Use $(T).
3691         (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
3692         (SHLIB_SRC, SHLIB_INSTALL): Delete.
3693         (SHLIB_LINK): Make dummy.
3694         * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
3696         * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
3698         * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
3699         * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
3700         config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
3701         config/rs6000/t-newas, config/sparc/t-elf: Likewise.
3703         * configure.ac (all_outputs): Remove mklibgcc.
3704         * configure: Regenerated.
3706 2007-01-03  Josh Conner  <jconner@apple.com>
3708         PR middle-end/29683
3709         * calls.c (compute_argument_addresses): Set stack and stack_slot
3710         for partial args, too.
3711         (store_one_arg): Use locate.size.constant for the size when
3712         generating a save_area.
3714 2007-01-03  Robert Kennedy <jimbob@google.com>
3716         * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
3717         whose definitions are deleted due to basic block merging.
3719 2007-01-03  Paul Brook  <paul@codesourcery.com>
3721         PR target/16634
3722         * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
3723         functions.
3724         (use_return_insn): Return 0 for Thumb interrupt functions.
3725         (print_multi_reg): Add rfe argument for IRQ returns.
3726         (arm_output_epilogue): Pop interrupt return address directly into PC.
3727         (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
3729 2007-01-03  Paul Brook  <paul@codesourcery.com>
3731         Merge from sourcerygxx-4_1.
3732         * config/arm/thumb2.md: New file.
3733         * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
3734         Thumb-2.
3735         * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
3736         * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
3737         (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
3738         * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
3739         tables.
3740         (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
3741         * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
3742         comptibility.
3743         * config/arm/ieee754-sf.S: Ditto.
3744         * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
3745         (thumb1_base_register_rtx_p): ... to this.
3746         (thumb_index_register_rtx_p): Rename...
3747         (thumb1_index_register_rtx_p): ... to this.
3748         (thumb_output_function_prologue): Rename...
3749         (thumb1_output_function_prologue): ... to this.
3750         (thumb_legitimate_address_p): Rename...
3751         (thumb1_legitimate_address_p): ... to this.
3752         (thumb_rtx_costs): Rename...
3753         (thumb1_rtx_costs): ... to this.
3754         (thumb_compute_save_reg_mask): Rename...
3755         (thumb1_compute_save_reg_mask): ... to this.
3756         (thumb_final_prescan_insn): Rename...
3757         (thumb1_final_prescan_insn): ... to this.
3758         (thumb_expand_epilogue): Rename...
3759         (thumb1_expand_epilogue): ... to this.
3760         (arm_unwind_emit_stm): Rename...
3761         (arm_unwind_emit_sequence): ... to this.
3762         (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
3763         thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
3764         thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
3765         arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
3766         thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
3767         thumb2_output_casesi): New functions.
3768         (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
3769         (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
3770         FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
3771         THUMB2_WORK_REGS): Define.
3772         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
3773         arm_condexec_mask, arm_condexec_masklen)): New variables.
3774         (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
3775         (arm_override_options): Check new CPU capabilities.
3776         Set new architecture flag variables.
3777         (arm_isr_value): Handle v7m interrupt functions.
3778         (user_return_insn): Return 0 for v7m interrupt functions.  Handle
3779         Thumb-2.
3780         (const_ok_for_arm): Handle Thumb-2 constants.
3781         (arm_gen_constant): Ditto.  Use movw when available.
3782         (arm_function_ok_for_sibcall): Return false for v7m interrupt
3783         functions.
3784         (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
3785         (thumb_find_work_register, arm_load_pic_register,
3786         legitimize_tls_address, arm_address_cost, load_multiple_sequence,
3787         emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
3788         print_multi_reg, output_mov_long_double_fpa_from_arm,
3789         output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
3790         output_mov_double_fpa_from_arm, output_move_double,
3791         arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
3792         output_return_instruction, arm_output_function_prologue,
3793         arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
3794         arm_output_mi_thunk, thumb_set_return_address): Ditto.
3795         (arm_expand_prologue): Handle Thumb-2.  Use arm_save_coproc_regs.
3796         (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
3797         (arithmetic_instr, shift_op): Use arm_shift_nmem.
3798         (arm_print_operand): Use arm_print_condition.  Handle '(', ')', '.',
3799         '!' and 'L'.
3800         (arm_final_prescan_insn): Use extract_constrain_insn_cached.
3801         (thumb_expand_prologue): Use thumb_set_frame_pointer.
3802         (arm_file_start): Output directive for unified syntax.
3803         (arm_unwind_emit_set): Handle stack alignment instruction.
3804         * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
3805         Add v6t2, v7, v7a, v7r and v7m.
3806         (RETLDM): Add Thumb-2 code.
3807         (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
3808         * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
3809         (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
3810         TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
3811         THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
3812         ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
3813         THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
3814         CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
3815         ADJUST_INSN_LENGTH): Define.
3816         (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
3817         STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
3818         BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
3819         PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
3820         SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
3821         TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
3822         HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
3823         HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
3824         REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
3825         REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
3826         LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
3827         GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
3828         ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
3829         FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
3830         PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
3831         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
3832         * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
3833         cortex-m3.
3834         * config/arm/arm-tune.md: Regenerate.
3835         * config/arm/arm-protos.h: Update prototypes.
3836         * config/arm/vfp.md: Enable patterns for Thumb-2.
3837         (arm_movsi_vfp): Add movw alternative.  Use output_move_vfp.
3838         (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
3839         (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
3840         thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
3841         * config/arm/libunwind.S: Add Thumb-2 code.
3842         * config/arm/constraints.md: Update include Thumb-2.
3843         * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
3844         * config/arm/ieee754-df.S: Ditto.
3845         * config/arm/bpabi.S: Ditto.
3846         * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
3847         * config/arm/predicates.md (low_register_operand,
3848         low_reg_or_int_operand, thumb_16bit_operator): New.
3849         (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
3850         (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
3851         * config/arm/t-arm-elf: Add armv7 multilib.
3852         * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
3853         Include thumb2.md.
3854         (UNSPEC_STACK_ALIGN, ce_count): New.
3855         (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
3856         arm_zero_extendsidi2, arm_zero_extendqidi2): New
3857         insns/expanders.
3858         * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
3859         (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
3860         thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
3861         * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
3862         (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
3863         thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
3864         insns.
3865         * doc/extend.texi: Document ARMv7-M interrupt functions.
3866         * doc/invoke.texi: Document Thumb-2 new cores+architectures.
3868 2007-01-03  Jakub Jelinek  <jakub@redhat.com>
3870         * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
3871         (struct _Unwind_Context): Rename args_size to flags, remove
3872         signal_frame field, add a new args_size field and version field.
3873         (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
3874         _Unwind_IsExtendedContext): New inline functions.
3875         (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
3876         Assume by_value array is only present if _Unwind_IsExtendedContext.
3877         (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
3878         _Unwind_IsSignalFrame.
3879         (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
3880         (uw_update_context_1): Use _Unwind_SetSignalFrame.
3881         (uw_init_context_1): Initialize context->flags to
3882         EXTENDED_CONTEXT_BIT.
3883         * config/rs6000/linux-unwind.h (frob_update_context): Use
3884         _Unwind_SetSignalFrame.
3886 2007-01-03  Andrew Pinski  <andrew_pinski@playstation.sony.com>
3888         PR middle-end/30353
3889         * gimplify.c (gimplify_modify_expr_complex_part): Move below
3890         tree_to_gimple_tuple.  Call tree_to_gimple_tuple when we need
3891         the value.
3893 2007-01-03  Kazu Hirata  <kazu@codesourcery.com>
3895         * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
3896         REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
3897         RET, AT_SP): Remove.
3898         * config/i386/i386.md (*sse_prologue_save_insn): Use return
3899         instead of RET.
3901         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3902         * cfgbuild.c (find_basic_blocks): Likewise.
3903         * cfgrtl.c (rtl_create_basic_block): Likewise.
3904         * function.c (temp_slots_at_level): Likewise.
3905         * reg-stack.c (stack_regs_mentioned): Likewise.
3906         * regclass.c (allocate_reg_info): Likewise.
3907         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3908         set_bb_for_stmt, move_block_to_fn): Likewise.
3909         * tree-complex.c (tree_lower_complex): Likewise.
3910         * vec.h (VEC_safe_grow_cleared): New.
3912         * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
3914 2007-01-03  Zdenek Dvorak <dvorakz@suse.cz>
3916         * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
3917         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
3918         to loop_version.
3919         * cfgloopmanip.c (scale_loop_frequencies): Export.
3920         (loopify): Scale the frequencies by prescribed coefficients.
3921         (set_zero_probability): New function.
3922         (duplicate_loop_to_header_edge): Improve updating of frequencies.
3923         (lv_adjust_loop_entry_edge, loop_version): Set probabilities
3924         and frequencies according to arguments.
3925         * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
3926         correctly.
3927         * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
3928         * modulo-sched.c (sms_schedule): Set probabilities for entering
3929         versioned loop correctly.
3930         * tree-vect-transform.c (vect_transform_loop): Ditto.
3931         * cfgloop.h (loopify, loop_version): Declaration changed.
3932         (scale_loop_frequencies): Declared.
3934 2007-01-02  Jan Hubicka  <jh@suse.cz>
3936         * cgraph.c: Include tree-flow.h
3937         (cgraph_add_new-function): Handle IPA_SSA mode; execute
3938         early_local_passes.
3939         * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
3940         * tree-pass.h (pass_all_early_optimizations): Declare.
3941         * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
3942         early_local_passes.
3943         (cgraph_analyze_function): Do early_local_passes.
3944         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3945         Do not add referenced vars.
3946         * tree-optimize.c (gate_all_optimizations): Do not execute when not in
3947         SSA form.
3948         (gate_all_early_local_passes): New gate.
3949         (pass_early_local_passes): Use new gate.
3950         (execute_early_local_optimizations): New functions.
3951         (gate_all_early_optimizations): New gate.
3952         (pass_all_early_optimizations): New pass.
3953         (execute_free_datastructures): Free SSA only when initialized.
3954         (gate_init_datastructures): Init only when optimizing.
3955         (tree_lowering_passes): Do early local passes when called late.
3956         * tree-profile.c (do_tree_profiling): Don't profile functions added
3957         late.
3958         (do_early_tree_profiling, pass_early_tree_profile): Kill.
3959         * tree-cfg.c (update_modified_stmts): Do not update when operands are
3960         not active.
3961         * passes.c (init_optimizations_passes): Reorder so we go into SSA
3962         during early_local_passes.
3963         * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
3966 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
3968         * Makefile.in: Update copyright year.
3970 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
3972         * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
3973         in $(lang_checks) and check-consistency targets.
3975 2007-01-02  Jan Hubicka  <jh@suse.cz>
3977         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3978         Do not add referenced vars.
3979         * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
3980         are not active.
3981         * passes.c (init_optimization_passes): Put mudflap_2 after
3982         free_datastructures.
3984 2007-01-02  Jan Hubicka  <jh@suse.cz>
3986         * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
3987         Set NOTHROW flag on call statements proved to be nothrow.
3988         Update statement of local calls so new pure/const functions are
3989         updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
3990         goto receivers.
3991         (tree_rest_of_compilation): Register hooks and initialize bitmap
3992         early. Do not set after_inlining flag.
3994 2007-01-02  Steve Ellcey  <sje@cup.hp.com>
3996         * sbitmap.c (HOST_BITS_PER_LONG_LONG):  Change to
3997         HOST_BITS_PER_LONGLONG
3999 2007-01-02  Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4001         PR c/19977
4002         * c-typeck.c (store_init_value): Don't emit pedantic overflow
4003         warning for non-static initializers.
4005 2007-01-02  Steven Bosscher  <steven@gcc.gnu.org>
4007         * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
4008         m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
4009         rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
4010         Always use set_unique_reg_note to add REG_EQUAL notes.
4012 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
4014         Revert:
4015         2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
4017         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4018         * cfgbuild.c (find_basic_blocks): Likewise.
4019         * cfgrtl.c (rtl_create_basic_block): Likewise.
4020         * function.c (temp_slots_at_level): Likewise.
4021         * reg-stack.c (stack_regs_mentioned): Likewise.
4022         * regclass.c (allocate_reg_info): Likewise.
4023         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4024         set_bb_for_stmt, move_block_to_fn): Likewise.
4025         * tree-complex.c (tree_lower_complex): Likewise.
4026         * vec.h (VEC_safe_grow_cleared): New.
4028 2007-01-02  Ian Lance Taylor  <iant@google.com>
4030         * c-common.c (c_common_truthvalue_conversion): When warning about
4031         using an assignment as a truth value, set TREE_NO_WARNING.
4033 2007-01-02  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
4035         PR middle-end/7651
4036         * c.opt (Wold-style-declaration): New.
4037         * doc/invoke.texi (C-only Warning Options): New.
4038         (Wold-style-declaration): Document it.
4039         (Wextra): Enabled by -Wextra.
4040         * c-opts.c (c_common_post_options): Enabled by -Wextra.
4041         * c-decl.c (declspecs_add_scspec): Replace -Wextra with
4042         -Wold-style-declaration.
4044 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
4046         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4047         * cfgbuild.c (find_basic_blocks): Likewise.
4048         * cfgrtl.c (rtl_create_basic_block): Likewise.
4049         * function.c (temp_slots_at_level): Likewise.
4050         * reg-stack.c (stack_regs_mentioned): Likewise.
4051         * regclass.c (allocate_reg_info): Likewise.
4052         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4053         set_bb_for_stmt, move_block_to_fn): Likewise.
4054         * tree-complex.c (tree_lower_complex): Likewise.
4055         * vec.h (VEC_safe_grow_cleared): New.
4057 2007-01-02  Douglas Gregor  <doug.gregor@gmail.com>
4059         * c-common.c (c_common_nodes_and_builtins): Since variants of
4060         void_type_node get built before it is given a name, we need to
4061         give those variants the name, too.
4062         (complete_array_type): We need to work with the canonical main
4063         type of the array, from which we will build the qualified version.
4064         * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
4065         * print-tree.c (print_node): Display canonical type information
4066         for each type.
4067         * stor-layout.c (layout_type): When we don't know the
4068         alignment of a type for which we're building an array, we end up
4069         guessing wrong, so make the type require structural equality.
4070         * tree.c (make_node_stat): When we build a new type, it is its
4071         own canonical type.
4072         (build_type_attribute_qual_variant): When building an attribute
4073         variant, its canonical type is the non-attribute variant. However,
4074         if the attributes are target-dependent and they differ, we need to
4075         use structural equality checks for this type.
4076         (build_qualified_type): A qualified type is not equivalent to its
4077         unqualified variant; set the canonical type appropriately.
4078         (build_distinct_type_copy): When building a distinct type from
4079         another type, the new type is its own canonical type.
4080         (build_variant_type_copy): When building a new type variant, we
4081         assume that it is equivalent to the original type.
4082         (build_pointer_type_for_mode): When building a pointer type, also
4083         build a canonical type pointer.
4084         (build_reference_type_for_mode): When building a reference type,
4085         also build a canonical type reference.
4086         (build_index_type): When we can't hash an index type (e.g.,
4087         because its maximum value is negative), the index type requires
4088         structural equality tests.
4089         (build_array_type): Build the canonical form of an array type.
4090         (build_function_type): Function types require structural equality,
4091         because they contain default arguments, attributes, etc.
4092         (build_method_type_directly): Ditto for method types.
4093         (build_offset_type): Build the canonical offset type.
4094         (build_complex_type): Build the canonical vector type.
4095         (make_vector_type): Build the canonical vector type.
4096         * tree.h (TYPE_CANONICAL): New.
4097         (TYPE_STRUCTURAL_EQUALITY_P): New.
4098         (SET_TYPE_STRUCTURAL_EQUALITY): New.
4099         (struct tree_type): Added "canonical" field.
4100         * params.h (VERIFY_CANONICAL_TYPES): New.
4101         * doc/c-tree.texi (TYPE_CANONICAL): Document.
4102         (TYPE_STRUCTURAL_EQUALITY_P): Document.
4103         (SET_TYPE_STRUCTURAL_EQUALITY): Document.
4104         * doc/invoke.texi (verify-canonical-types): Document --param
4105         parameter for verifying canonical types.
4107 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
4109         * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
4110         powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
4111         powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
4112         powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
4113         tm_file.
4114         * config/rs6000/e500.h: New.
4115         * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
4116         TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
4117         Remove.
4118         * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
4119         TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
4120         TARGET_E500_DOUBLE): Remove.
4121         * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
4122         TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
4123         * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
4124         * config/rs6000/rs6000.c (rs6000_override_options): Use
4125         CHECK_E500_OPTIONS.
4127 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
4129         * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
4130         TARGET_E500_DOUBLE), not TARGET_E500, for %y.
4131         (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
4132         rs6000_emit_cmove): Don't check TARGET_E500.
4133         * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
4134         sordered): Don't check TARGET_E500.
4136 2007-01-01  Eric Christopher  <echristo@apple.com>
4138         * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
4139         HARD_REGISTER_NUM_P.
4141 2007-01-01  Roger Sayle  <roger@eyesopen.com>
4143         * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
4144         "X eq/ne (C1^C2)".  Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
4145         has no side-effects.  Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
4146         eq/ne Y".
4148 2007-01-01  Mike Stump  <mrs@apple.com>
4150         * configure.ac: Remove support for building with Apple's gcc-3.1.
4152 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
4154         PR middle-end/30311
4155         * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
4156         REG is a hard register.  Do not modify REG before calling
4157         subreg_nregs.
4158         * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
4159         multiple of the size of YMODE for certain lowpart cases.
4161 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
4163         PR middle-end/30253
4164         * gimplify (voidify_wrapper_expr): Update for
4165         GIMPLE_MODIFY_STMT.
4167 2007-01-01  Andreas Schwab  <schwab@suse.de>
4169         PR target/29166
4170         * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
4171         save of BR0 in extra_spill_size instead of spill_size.
4172         (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
4173         area.
4174         (ia64_expand_epilogue): Restore BR0 from its new location.
4176 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
4178         * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
4179         Use a temporary variable if the left hand side is not a gimple
4180         register.
4182 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
4184         * gimplify.c (gimplify_return_expr): Make the temporary variable
4185         for the return expression, a gimple register variable.
4187 2007-01-01  Jan Hubicka  <jh@suse.cz>
4189         * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
4190         INSN_LIST.
4192 2007-01-01  Mike Stump  <mrs@apple.com>
4194         * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
4195         .literal16.
4196         * config/darwin.c (machopic_select_rtx_section): Use
4197         HAVE_GAS_LITERAL16.
4198         (darwin_mergeable_constant_section): Likewise.
4199         * configure: Regenerate.
4200         * config.in: Regenerate.
4202 2007-01-01  Jan Hubicka  <jh@suse.cz>
4203             Andrew Pinski  <pinskia@gmail.com>
4205         * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
4206         before starting IPA passes.