* fold-const.c (negate_expr_p): Handle CONJ_EXPR.
[official-gcc.git] / gcc / ChangeLog
blob947df6f73c3274223cb37cc05f7978625130a1a1
1 2007-02-05  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
3         * fold-const.c (negate_expr_p): Handle CONJ_EXPR.
4         (fold_negate_expr): Likewise.
6 2007-02-05  Alexandre Oliva  <aoliva@redhat.com>
8         PR debug/30189
9         * dwarf2out.c (modified_type_die): Follow DECL_ORIGINAL_TYPE
10         even if cv-qualification is the same.
12 2007-02-05  Geoffrey Keating  <geoffk@apple.com>
14         * config/rs6000/darwin-tramp.asm (__trampoline_setup): Call
15         __enable_execute_stack on completion.
17 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
19         * config/i386/athlon.md (athlon_fldxf_k8, athlon_fld_k8,
20         athlon_fstxf_k8, athlon_fst_k8, athlon_fist, athlon_fmov,
21         athlon_fadd_load, athlon_fadd_load_k8, athlon_fadd, athlon_fmul,
22         athlon_fmul_load, athlon_fmul_load_k8, athlon_fsgn,
23         athlon_fdiv_load, athlon_fdiv_load_k8, athlon_fdiv_k8,
24         athlon_fpspc_load, athlon_fpspc, athlon_fcmov_load,
25         athlon_fcmov_load_k8, athlon_fcmov_k8, athlon_fcomi_load_k8,
26         athlon_fcomi, athlon_fcom_load_k8, athlon_fcom): Added amdfam10.
28 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
30         * config/i386/i386.md (x86_sahf_1, cmpfp_i_mixed, cmpfp_i_sse,
31         cmpfp_i_i387, cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387,
32         swapsi, swaphi_1, swapqi_1, swapdi_rex64, fix_truncsfdi_sse,
33         fix_truncdfdi_sse, fix_truncsfsi_sse, fix_truncdfsi_sse,
34         x86_fldcw_1, floatsisf2_mixed, floatsisf2_sse, floatdisf2_mixed,
35         floatdisf2_sse, floatsidf2_mixed, floatsidf2_sse,
36         floatdidf2_mixed, floatdidf2_sse, muldi3_1_rex64, mulsi3_1,
37         mulsi3_1_zext, mulhi3_1, mulqi3_1, umulqihi3_1, mulqihi3_insn,
38         umulditi3_insn, umulsidi3_insn, mulditi3_insn, mulsidi3_insn,
39         umuldi3_highpart_rex64, umulsi3_highpart_insn,
40         umulsi3_highpart_zext, smuldi3_highpart_rex64,
41         smulsi3_highpart_insn, smulsi3_highpart_zext, x86_64_shld,
42         x86_shld_1, x86_64_shrd, sqrtsf2_mixed, sqrtsf2_sse,
43         sqrtsf2_i387, sqrtdf2_mixed, sqrtdf2_sse, sqrtdf2_i387,
44         sqrtextendsfdf2_i387, sqrtxf2, sqrtextendsfxf2_i387,
45         sqrtextenddfxf2_i387): Added amdfam10_decode.
46         
47         * config/i386/athlon.md (athlon_idirect_amdfam10,
48         athlon_ivector_amdfam10, athlon_idirect_load_amdfam10,
49         athlon_ivector_load_amdfam10, athlon_idirect_both_amdfam10,
50         athlon_ivector_both_amdfam10, athlon_idirect_store_amdfam10,
51         athlon_ivector_store_amdfam10): New define_insn_reservation.
52         (athlon_idirect_loadmov, athlon_idirect_movstore): Added
53         amdfam10.
55 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
57         * config/i386/athlon.md (athlon_call_amdfam10,
58         athlon_pop_amdfam10, athlon_lea_amdfam10): New
59         define_insn_reservation.
60         (athlon_branch, athlon_push, athlon_leave_k8, athlon_imul_k8,
61         athlon_imul_k8_DI, athlon_imul_mem_k8, athlon_imul_mem_k8_DI,
62         athlon_idiv, athlon_idiv_mem, athlon_str): Added amdfam10.
64 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
66         * config/i386/athlon.md (athlon_sseld_amdfam10,
67         athlon_mmxld_amdfam10, athlon_ssest_amdfam10,
68         athlon_mmxssest_short_amdfam10): New define_insn_reservation.
70 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
72         * config/i386/athlon.md (athlon_sseins_amdfam10): New
73         define_insn_reservation.
74         * config/i386/i386.md (sseins): Added sseins to define_attr type
75         and define_attr unit.
76         * config/i386/sse.md: Set type attribute to sseins for insertq
77         and insertqi.
79 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
81         * config/i386/athlon.md (sselog_load_amdfam10, sselog_amdfam10,
82         ssecmpvector_load_amdfam10, ssecmpvector_amdfam10,
83         ssecomi_load_amdfam10, ssecomi_amdfam10,
84         sseaddvector_load_amdfam10, sseaddvector_amdfam10): New
85         define_insn_reservation.
86         (ssecmp_load_k8, ssecmp, sseadd_load_k8, seadd): Added amdfam10.
88 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
90         * config/i386/athlon.md (cvtss2sd_load_amdfam10,
91         cvtss2sd_amdfam10, cvtps2pd_load_amdfam10, cvtps2pd_amdfam10,
92         cvtsi2sd_load_amdfam10, cvtsi2ss_load_amdfam10,
93         cvtsi2sd_amdfam10, cvtsi2ss_amdfam10, cvtsd2ss_load_amdfam10,
94         cvtsd2ss_amdfam10, cvtpd2ps_load_amdfam10, cvtpd2ps_amdfam10,
95         cvtsX2si_load_amdfam10, cvtsX2si_amdfam10): New 
96         define_insn_reservation.
98         * config/i386/sse.md (cvtsi2ss, cvtsi2ssq, cvtss2si,
99         cvtss2siq, cvttss2si, cvttss2siq, cvtsi2sd, cvtsi2sdq,
100         cvtsd2si, cvtsd2siq, cvttsd2si, cvttsd2siq,
101         cvtpd2dq, cvttpd2dq, cvtsd2ss, cvtss2sd,
102         cvtpd2ps, cvtps2pd): Added amdfam10_decode attribute.
104 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
106         * config/i386/athlon.md (athlon_ssedivvector_amdfam10,
107         athlon_ssedivvector_load_amdfam10, athlon_ssemulvector_amdfam10,
108         athlon_ssemulvector_load_amdfam10): New define_insn_reservation.
109         (athlon_ssediv, athlon_ssediv_load_k8, athlon_ssemul,
110         athlon_ssemul_load_k8): Added amdfam10.
112 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
114         * config/i386/i386.h (TARGET_SSE_UNALIGNED_MOVE_OPTIMAL): New macro.
115         (x86_sse_unaligned_move_optimal): New variable.
116         
117         * config/i386/i386.c (x86_sse_unaligned_move_optimal): Enable for  
118         m_AMDFAM10.
119         (ix86_expand_vector_move_misalign): Add code to generate movupd/movups
120         for unaligned vector SSE double/single precision loads for AMDFAM10.
122 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
124         * config/i386/i386.h (TARGET_AMDFAM10): New macro.
125         (TARGET_CPU_CPP_BUILTINS): Add code for amdfam10.
126         Define TARGET_CPU_DEFAULT_amdfam10.
127         (TARGET_CPU_DEFAULT_NAMES): Add amdfam10.
128         (processor_type): Add PROCESSOR_AMDFAM10.       
129         
130         * config/i386/i386.md: Add amdfam10 as a new cpu attribute to match
131         processor_type in config/i386/i386.h.
132         Enable imul peepholes for TARGET_AMDFAM10.
133         
134         * config.gcc: Add support for --with-cpu option for amdfam10.
135         
136         * config/i386/i386.c (amdfam10_cost): New variable.
137         (m_AMDFAM10): New macro.
138         (m_ATHLON_K8_AMDFAM10): New macro.
139         (x86_use_leave, x86_push_memory, x86_movx, x86_unroll_strlen,
140         x86_cmove, x86_3dnow_a, x86_deep_branch, x86_use_simode_fiop,
141         x86_promote_QImode, x86_integer_DFmode_moves,
142         x86_partial_reg_dependency, x86_memory_mismatch_stall, 
143         x86_accumulate_outgoing_args, x86_arch_always_fancy_math_387,
144         x86_sse_partial_reg_dependency, x86_sse_typeless_stores,
145         x86_use_ffreep, x86_use_incdec, x86_four_jump_limit,
146         x86_schedule, x86_use_bt, x86_cmpxchg16b, x86_pad_returns):
147         Enable/disable for amdfam10.
148         (override_options): Add amdfam10_cost to processor_target_table.
149         Set up PROCESSOR_AMDFAM10 for amdfam10 entry in 
150         processor_alias_table.
151         (ix86_issue_rate): Add PROCESSOR_AMDFAM10.
152         (ix86_adjust_cost): Add code for amdfam10.
154 2007-02-05      Harsha Jagasia  <harsha.jagasia@amd.com>
155         
156         * config/i386/i386.opt: Add new Advanced Bit Manipulation (-mabm)
157         instruction set feature flag. Add new (-mpopcnt) flag for popcnt 
158         instruction. Add new SSE4A (-msse4a) instruction set feature flag.
159         * config/i386/i386.h: Add builtin definition for SSE4A.
160         * config/i386/i386.md: Add support for ABM instructions 
161         (popcnt and lzcnt).
162         * config/i386/sse.md: Add support for SSE4A instructions
163         (movntss, movntsd, extrq, insertq).
164         * config/i386/i386.c: Add support for ABM and SSE4A builtins.
165         Add -march=amdfam10 flag.
166         * config/i386/ammintrin.h: Add support for SSE4A intrinsics.
167         * doc/invoke.texi: Add documentation on flags for sse4a, abm, popcnt
168         and amdfam10.
169         * doc/extend.texi: Add documentation for SSE4A builtins.
171 2007-02-05  Bob Wilson  <bob.wilson@acm.org>
173         * config/xtensa/xtensa.c (constantpool_mem_p): Skip over SUBREGs.
175 2007-02-05  Richard Guenther  <rguenther@suse.de>
177         * tree-vectorizer.h (vectorizable_function): Add argument type
178         argument, change return type.
179         * tree-vect-patterns.c (vect_recog_pow_pattern): Adjust caller.
180         * tree-vect-transform.c (vectorizable_function): Handle extra
181         argument, return vectorized function decl.
182         (build_vectorized_function_call): Remove.
183         (vectorizable_call): Handle calls with result and argument types
184         differing.  Handle loop vectorization factor correctly.
185         * targhooks.c (default_builtin_vectorized_function): Adjust for
186         extra argument.
187         * targhooks.h (default_builtin_vectorized_function): Likewise.
188         * target.h (builtin_vectorized_function): Add argument type
189         argument.
190         * config/i386/i386.c (ix86_builtin_vectorized_function): Handle
191         extra argument, allow vectorizing of lrintf.
192         * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION): Adjust
193         documentation of target hook.
195 2007-02-05  Hans-Peter Nilsson  <hp@axis.com>
197         PR target/30665
198         * config/cris/cris.md ("*andsi_movu", "*andsi_clear", "*andhi_movu")
199         ("*andhi_clear", andu (casesi+45)): For size-changed operand where
200         memory is allowed, require !side_effects_p, not just !MEM_VOLATILE_P.
202 2007-02-05  Roger Sayle  <roger@eyesopen.com>
204         * fold-const.c (fold_unary) <REAL_PART>: Test for availability of
205         BUILT_IN_COS before simplifying REAL_PART(CEXPI)) to COS.
206         <IMAG_PART>: Likewise, check for availability of BUILT_IN_SIN.
207         * builtins.c (fold_builtin_sincos): Check for TARGET_C99_FUNCTIONS
208         before canonicalizing sincos to cexpi.
209         (fold_builtin_cexp): Likewise, for canonicalizing cexp to cexpi.
211 2007-02-05  Roger Sayle  <roger@eyesopen.com>
213         * config/alpha/alpha.c (alpha_add_builtins): New Helper function.
214         Set TREE_READONLY and TREE_NOTHROW directly, not via attributes.
215         (alpha_init_builtins): Use alpha_add_builtins to process tables.
217 2007-02-05  Roger Sayle  <roger@eyesopen.com>
219         * mips-tfile.c (initialize_init_file): Correct endianness test.
221 2007-02-05  Kazu Hirata  <kazu@codesourcery.com>
223         * config/m68k/m68k.md (pushdi-1, pushdi, movsi+1): Don't use
224         the 'y' constraint.
226 2007-02-05  Richard Sandiford  <richard@codesourcery.com>
228         * dwarf2out.c (dwarf2out_frame_debug_expr): Record the register
229         saves in a PARALLEL before the register assignments.
231 2007-02-05  Richard Sandiford  <richard@codesourcery.com>
233         * doc/tm.texi (DWARF_ALT_FRAME_RETURN_COLUMN): Do not require
234         DWARF_FRAME_RETURN_COLUMN to be a general register.
235         * dwarf2out.c (init_return_column_size): New function, split from...
236         (expand_builtin_init_dwarf_reg_sizes): ...here.  Allow both
237         DWARF_FRAME_RETURN_COLUMN and DWARF_ALT_FRAME_RETURN_COLUMN
238         to be nongeneral registers.
239         * config/m68k/m68k.h (DWARF_FRAME_REGNUM): Only map FP and
240         integer registers.
241         (DWARF_FRAME_REGISTERS, DWARF_FRAME_RETURN_COLUMN): Define.
242         (DWARF_ALT_FRAME_RETURN_COLUMN): Define.
244 2007-02-04  Zdenek Dvorak <dvorakz@suse.cz>
246         * cfgcleanup.c (try_optimize_cfg): Avoid removing ENTRY_BLOCK_PTR.
248 2007-02-04  Zdenek Dvorak <dvorakz@suse.cz>
250         * cfgloopmanip.c (loop_delete_branch_edge): Removed.
251         (remove_path): Use can_remove_branch_p and remove_branch instead
252         of loop_delete_branch_edge.
253         * tree-ssa-loop-manip.c (scale_dominated_blocks_in_loop): New function.
254         (tree_transform_and_unroll_loop): Remove dead branches immediately.
255         Update profile using scale_dominated_blocks_in_loop.
256         * cfghooks.c (can_remove_branch_p, remove_branch): New functions.
257         * cfghooks.h (struct cfg_hooks): Add can_remove_branch_p.
258         (can_remove_branch_p, remove_branch): Declare.
259         * tree-cfg.c (tree_can_remove_branch_p): New function.
260         (tree_cfg_hooks): Add tree_can_remove_branch_p.
261         * cfgrtl.c (rtl_can_remove_branch_p): New function.
262         (rtl_cfg_hooks, cfg_layout_rtl_cfg_hook): Add rtl_can_remove_branch_p.
264 2007-02-05  Jan Hubicka  <jh@suse.cz>
266         PR middle-end/30696
267         * ipa-inline.c (cgraph_clone_inlined_nodes): When there are unanalyzed
268         nodes in cgraph, don't remove offline copy of the function.
270 2007-02-04  Jan Hubicka  <jh@suse.cz>
272         * tree-sra.c (sra_walk_expr): Add linebreaks. BITFIELD_REFs into
273         vectors might cause maybe_lookup_element_for_expr to be called
274         on non-sra-candidate.
276 2007-02-04  Kazu Hirata  <kazu@codesourcery.com>
278         * config/bfin/bfin-modes.def, config/bfin/bfin.c,
279         config/bfin/bfin.md, config/bfin/predicates.md: Follow
280         spelling conventions.
282 2007-02-04  Richard Guenther  <rguenther@suse.de>
284         PR middle-end/30636
285         * fold-const.c (try_move_mult_to_index): Make sure to not
286         overflow one dimension of a multi-dimensional array access.
288 2007-02-04  Jan Hubicka  <jh@suse.cz>
290         * passes.c (init_optimization_passes): Reindent.
292 2007-02-04  Jan Hubicka  <jh@suse.cz>
293             Eric Botcazou  <ebotcazou@adacore.com>
295         * tree-optimize.c (has_abnormal_outgoing_edge_p): Move to...
296         (execute_fixup_cfg): Break out the abnormal goto code.
297         * tree-inline.c (has_abnormal_outgoing_edge_p): ...here.
298         (make_nonlocal_label_edges): Move here from execute_fixup_cfg.
299         (optimize_inline_calls): Call make_nonlocal_label_edges.
301 2007-02-04  Jan Hubicka  <jh@suse.cz>
303         * tree-ssa-copyrename.c (copy_rename_partition_coalesce): Return
304         true when something was changed.
305         (rename_ssa_copies): When something was changed, do
306         TODO_remove_unused_locals.
307         * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
308         add TODO_remove_unused_locals when instruction was removed.
310 2007-02-04  Jan Hubicka  <jh@suse.cz>
312         * ipa-inline.c (try_inline): Improve debug output; work on already
313         inline edges too.
314         (cgraph_decide_inlining_incrementally): Indent; improve debug output;
315         call try_inline for already inlined edges too when flattening; 
316         inline also functions that make callee growth but overall unit size
317         reduce.
319 2007-02-04  Kazu Hirata  <kazu@codesourcery.com>
321         * config/m32c/bitops.md, config/m32c/jump.md,
322         config/m32c/m32c.c, config/m32c/m32c.h, config/m32r/m32r.c,
323         config/m32r/m32r.h, config/m32r/m32r.md,
324         config/m32r/predicates.md, config/m68hc11/larith.asm,
325         config/m68hc11/m68hc11.c, config/m68hc11/m68hc11.h,
326         config/m68k/m68k.h, config/mcore/mcore.md, config/mips/4k.md,
327         config/mips/mips-protos.h, config/mips/mips.c,
328         config/mips/mips.h, config/mips/mips.md, config/mips/mips16.S,
329         config/mn10300/mn10300.h, config/mn10300/predicates.md,
330         config/mt/mt.c, config/mt/mt.h, config/mt/mt.md: Follow
331         spelling conventions.
333         * config/v850/v850.c, config/v850/v850.h, config/v850/v850.md:
334         Follow spelling conventions.
336 2007-02-03  Douglas Gregor  <doug.gregor@gmail.com>
338        * c-opts.c (c_common_post_options): If C++0x mode is enabled, don't
339        warn about C++0x compatibility.
340         
341 2007-02-04  Kazu Hirata  <kazu@codesourcery.com>
343         * config/h8300/h8300.c, config/h8300/h8300.h,
344         config/h8300/h8300.md: Follow spelling conventions.
346 2007-02-03  Uros Bizjak  <ubizjak@gmail.com>
348         PR middle-end/30667
349         * combine.c (try_combine): Do not substitute source operand
350         with constants wider than 2 * HOST_BITS_PER_WIDE_INT.
352 2007-02-03  Jan Hubicka  <jh@suse.cz>
354         PR gcov-profile/30650
355         * value-prof.c (stringop_block_profile): Fix handling of size counter;
356         do not divide by zero for never executed counters.
357         (tree_find_values_to_profile): Fix counters.
358         * gcov-ui.h (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): Fix comments.
360 2007-02-03  Ian Lance Taylor  <iant@google.com>
362         * lower-subreg.c (simple_move_operand): New static function,
363         broken out of simple_move.  Reject LABEL_REF, SYMBOL_REF, and HIGH
364         operands.
365         (simple_move): Call simple_move_operand.
366         (find_decomposable_subregs): Add special handling of MEMs.
367         (can_decompose_p): Rename from cannot_decompose_p.  Reverse
368         meaning of return value.  If we see a hard register, test whether
369         it can store a word_mode value.  Change all callers.
371 2007-02-03  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
373         * pa.md (addvdi3, addvsi3, subvdi3, subvsi3, negvdi2, negvsi2): New
374         ftrapv insns and expanders.
375         (subdi3): Change define_expand operand 1 to arith11_operand, and
376         operand 2 to reg_or_0_operand.  Change constraints of 64-bit insn
377         pattern to handle reg_or_0 operands.  Revise 32-bit insn pattern to
378         handle 11-bit constants and reg_or_0 operands in operands 1 and 2,
379         respectively.
381         PR middle-end/30174
382         * varasm.c (notice_global_symbol): Treat global objects as weak when
383         flag_shlib is true.
385 2007-02-03  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
387         * emit-rtl.c (dconstpi): Delete.
388         (dconstsqrt2): New.
389         (init_emit_once): Delete dconstpi and init dconstsqrt2.
390         * real.h (dconstpi): Delete.
391         (dconstsqrt2): New.
392         * builtins.c (fold_builtin_cabs): Use dconstsqrt2.
393         (fold_builtin_hypot): Likewise.
395 2007-02-03  Tom Tromey  <tromey@redhat.com>
397         PR driver/30246
398         * gcc.c (cpp_unique_options): Any of -ggdb3, -gstabs3,
399         -gcoff3, -gxcoff3, -gvms3 implies -dD.
401 2007-02-03  Kazu Hirata  <kazu@codesourcery.com>
403         * c-decl.c, config/avr/avr.c, config/avr/avr.h,
404         config/m68k/m68k.c, config/m68k/netbsd-elf.h,
405         config/mn10300/mn10300.c, config/pdp11/pdp11.h,
406         config/rs6000/cell.md, config/rs6000/darwin.h,
407         config/sh/sh.md, config/sh/sh4-300.md, config/spu/spu.c,
408         config/spu/spu.md, cselib.c, expr.c, haifa-sched.c, hwint.h,
409         jump.c, reload.c, sched-deps.c, sched-int.h, tree-inline.c,
410         tree-profile.c, tree-ssa-live.h, tree-vrp.c: Fix comment
411         typos.  Follow spelling conventions.
412         * doc/invoke.texi: Follow spelling conventions.
414 2007-02-03  Roger Sayle  <roger@eyesopen.com>
416         * simplify-rtx.c (simplify_relational_operation_1): Implement some
417         canonicalization transformations that attempt to simplify integer
418         constant comparisons to become comparisons against zero.
420 2007-02-03  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
422         * builtins.c (fold_builtin_cabs): Fold cabs(x+xi) into
423         fabs(x)*sqrt(2).
424         * fold-const.c (fold_binary): Fix comment typos.  Fold complex
425         (x,0)-(0,y) into (x,-y).  Likewise (0,y)-(x,0) into (-x,y).
427 2007-02-02  Mike Stump  <mrs@apple.com>
429         * config/darwin9.h (DARWIN_LINKER_GENERATES_ISLANDS): Add.
430         * config/rs6000/rs6000.c (DARWIN_GENERATE_ISLANDS): Add.
431         (output_call): Use DARWIN_GENERATE_ISLANDS to decide when to
432         generate a branch island.
434 2007-02-02  Bob Wilson  <bob.wilson@acm.org>
436         * config/xtensa/xtensa.c (smalloffset_mem_p): Use BASE_REG_P.
437         (xtensa_legitimate_address_p): New.
438         (xtensa_legitimize_address): New.
439         (xtensa_output_addr_const_extra): New.
440         * config/xtensa/xtensa.h (REG_OK_STRICT_FLAG): Define.
441         (BASE_REG_P): New.
442         (REG_OK_FOR_BASE_P): Use BASE_REG_P.
443         (GO_IF_LEGITIMATE_ADDRESS): Move code to xtensa_legitimate_address_p.
444         (LEGITIMIZE_ADDRESS): Move code to xtensa_legitimize_address.
445         (OUTPUT_ADDR_CONST_EXTRA): Move code to xtensa_output_addr_const_extra.
446         * config/xtensa/xtensa-protos.h (xtensa_legitimate_address_p): New.
447         (xtensa_legitimize_address): New.
448         (xtensa_output_addr_const_extra): New.
450 2007-02-02  Steve Ellcey  <sje@cup.hp.com>
452         * config/ia64/ia64.c (ia64_print_operand): Fix compare strings.
454 2007-02-02  Ian Lance Taylor  <iant@google.com>
456         * expmed.c (expand_divmod): Add comment.
458 2007-02-02  Kazu Hirata  <kazu@codesourcery.com>
460         * emit-rtl.c (renumber_insns): Remove.
461         * flags.h: Remove the extern for flag_renumber_insns.
462         * rtl.h: Remove the prototype for renumber_insns.
463         * toplev.c (flag_renumber_insns): Remove.
465 2007-02-02  Hui-May Chang  <hm.chang@apple.com>
467         Revert for x86 darwin:
468         2005-06-19  Uros Bizjak  <uros@kss-loka.si>
469         
470         * config/i386/i386.c (ix86_function_arg_regno_p): Put back the
471         code before the following patch under TARGET_MACHO.
472         (ix86_function_value_regno_p): Likewise.
474 2007-02-02  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
476         * fold-const.c (negate_expr_p, fold_negate_expr): Handle
477         COMPLEX_EXPR.
479 2007-02-02  Maxim Kuvyrkov  <mkuvyrkov@ispras.ru>
481         PR target/29682
482         * config/ia64/ia64.c (ia64_speculate_insn): Restrict to memory loads to
483         general or fp registers.  Add comments.
484         * config/ia64/ia64.md (reg_pred_prefix): Add comment.
486 2007-02-02  Paolo Bonzini  <bonzini@gnu.org>
488         * pointer-set.c (insert_aux): Only return insertion slot.
489         (pointer_set_insert): Adjust.
490         (pointer_set_traverse, struct pointer_map_t, pointer_map_create,
491         pointer_map_destroy, pointer_map_insert, pointer_map_contains,
492         pointer_map_traverse): New.
493         * pointer-set.h (pointer_set_traverse, struct pointer_map_t,
494         pointer_map_create,  pointer_map_destroy, pointer_map_insert,
495         pointer_map_contains, pointer_map_traverse): Declare.
497 2007-02-02  Jakub Jelinek  <jakub@redhat.com>
499         PR middle-end/30473
500         * builtins.c (fold_builtin_sprintf): Do not attempt to optimize
501         sprintf (str, "%s").  Do not optimize sprintf (str, "nopercent", p++).
503 2007-02-02  Maxim Kuvyrkov  <mkuvyrkov@ispras.ru>
505         * sched-int.h (ds_to_dk, dk_to_ds): Declare functions.
506         
507         (struct _dep): New type.
508         (dep_t): New typedef.
509         (DEP_PRO, DEP_CON, DEP_KIND): New access macros.
510         (DEP_STATUS): New access macro.  The macro with the same name was
511         renamed to DEP_LINK_STATUS.
512         (dep_init): Declare function
514         (struct _dep_link): New type.
515         (dep_link_t): New typedef.
516         (DEP_LINK_NODE, DEP_LINK_NEXT, DEP_LINK_PREV_NEXTP): New access macros.
517         (DEP_LINK_DEP, DEP_LINK_PRO, DEP_LINK_CON, DEP_LINK_KIND): New macros.
518         (DEP_LINK_STATUS): New macro.
519         (debug_dep_links): New debug function.
521         (struct _deps_list): New type.
522         (deps_list_t): New typedef.
523         (DEPS_LIST_FIRST): New access macro.
524         (FOR_EACH_DEP_LINK): New cycle macro.
525         (create_deps_list, free_deps_list, delete_deps_list): Declare
526         functions.
527         (deps_list_empty_p, debug_deps_list, add_back_dep_to_deps_list): Ditto.
528         (find_link_by_pro_in_deps_list, find_link_by_con_in_deps_list): Ditto.
529         (copy_deps_list_change_con): Ditto.
531         (move_dep_link): Declare function.
533         (struct _dep_node): New type.
534         (dep_node_t): New typedef.
535         (DEP_NODE_BACK, DEP_NODE_DEP, DEP_NODE_FORW): New access macros.
537         (struct haifa_insn_data.back_deps): New field to hold backward
538         dependencies of the insn.
539         (struct haifa_insn_data.depend): Rename to forw_deps.  Change its type
540         to deps_list_t.
541         (struct haifa_insn_data.resolved_deps): Rename to resolved_back_deps.
542         Change its type to deps_list_t.
543         (INSN_BACK_DEPS): New access macro to use instead of LOG_LINKS.
544         (INSN_DEPEND): Rename to INSN_FORW_DEPS.
545         (RESOLVED_DEPS): Rename to INSN_RESOLVED_BACK_DEPS.
547         (INSN_COST): Move to haifa-sched.c.  Use insn_cost () instead.
548         
549         (DEP_STATUS): Rename to DEP_LINK_STATUS.  Fix typo in the comment.
551         (add_forw_dep, delete_back_forw_dep, insn_cost): Update declaration and
552         all callers.
553         (dep_cost): Declare.
554         
555         * sched-deps.c (CHECK): New macro to (en/dis)able sanity checks.
556         (ds_to_dk, dk_to_ds): New functions.
557         
558         (init_dep_1): New static function.
559         (init_dep): New function.
560         (copy_dep): New static function.
561         
562         (dep_link_consistent_p, attach_dep_link, add_to_deps_list): New static
563         functions.
564         (detach_dep_link): New static function.
565         (move_dep_link): New function.
566         
567         (dep_links_consistent_p, dump_dep_links): New static functions.
568         (debug_dep_links): New debugging function.
569         
570         (deps_obstack, dl_obstack, dn_obstack): New static variables.
571         
572         (alloc_deps_list, init_deps_list): New static functions.
573         (create_deps_list): New function.
574         (clear_deps_list): New static function.
575         (free_deps_list, delete_deps_list, deps_list_empty_p): New functions.
576         (deps_list_consistent_p, dump_deps_list): New static functions.
577         (debug_deps_list): New function.
578         (add_back_dep_to_deps_list, find_link_by_pro_in_deps_list): New
579         functions.
580         (find_link_by_con_in_deps_list, copy_deps_list_change_con): Ditto.
582         (maybe_add_or_update_back_dep_1, add_or_update_back_dep_1): Update to
583         use new scheduler dependencies lists.
584         (add_back_dep, delete_all_dependences, fixup_sched_groups): Ditto.
585         (sched_analyze): Ditto.  Initialize dependencies lists.
586         (add_forw_dep, compute_forward_dependences): Update to use new
587         scheduler dependencies lists.
588         
589         (init_dependency_caches): Init deps_obstack.
590         (free_dependency_caches): Free deps_obstack.
591         
592         (adjust_add_sorted_back_dep, adjust_back_add_forw_dep): Update to use
593         new scheduler dependencies lists.
594         (delete_forw_dep, add_or_update_back_forw_dep): Ditto.
595         (add_back_forw_dep, delete_back_forw_dep): Ditto.
597         * sched-rgn.c (set_spec_fed, find_conditional_protection, is_pfree):
598         Update to use new scheduler dependencies lists.
599         (is_conditionally_protected, is_prisky, add_branch_dependences): Ditto.
600         (debug_dependencies): Ditto.
601         (schedule_region): Update comments.
602         
603         * sched-ebb.c (earliest_block_with_similiar_load): Update to use new
604         scheduler dependencies lists.
605         (schedule_ebb): Update comments.
606         
607         * rtl.def (DEPS_LIST): Remove.
608         
609         * lists.c (unused_deps_list): Remove.
610         (free_list): Update assertions.
611         
612         (alloc_DEPS_LIST, free_DEPS_LIST_list, free_DEPS_LIST_node): Remove.
613         (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
615         * rtl.h (free_DEPS_LIST_list, alloc_DEPS_LIST): Remove declarations.
616         (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
617         
618         * haifa-sched.c (comments): Update.
619         (insn_cost1): Remove.  Inline the code into insn_cost ().
620         (insn_cost): Update to use new scheduler dependencies lists.  Move
621         processing of the dependency cost to dep_cost ().
622         (dep_cost): New function.  Use it instead of insn_cost () when
623         evaluating cost of the dependency.  Use compatible interface to
624         interact with the target.
625         (priority): Update to use new scheduler dependencies lists.
626         (rank_for_schedule): Ditto.  Optimize heuristic that prefers the insn
627         with greater number of insns that depend on the insn.
628         (schedule_insn): Update to use new scheduler dependencies lists.  Add
629         code to free backward dependencies lists.  Inline and optimize code
630         from resolve_dep () - see PR28071.
631         (ok_for_early_queue_removal): Update to use new scheduler dependencies
632         lists.  Update call to targetm.sched.is_costly_dependence hook.
633         
634         (fix_inter_tick, try_ready, fix_tick_ready): Update to use new
635         scheduler dependencies lists.
636         
637         (resolve_dep): Remove.  Move the logic to schedule_insn ().
638         (init_h_i_d): Initialize dependencies lists.
639         
640         (process_insn_depend_be_in_spec): Rename to
641         process_insn_forw_deps_be_in_spec.  Update to use new scheduler
642         dependencies lists.
643         (add_to_speculative_block, create_check_block_twin, fix_recovery_deps):
644         Update to use new scheduler dependencies lists.
645         (clear_priorities, calc_priorities, add_jump_dependencies): Ditto.
646         
647         * ddg.c (create_ddg_dependence, create_ddg_dep_no_link): Update to use
648         new scheduler dependencies lists.
649         (build_intra_loop_deps): Ditto.
650         
651         * target.h (struct _dep): Declare to use in
652         gcc_target.sched.is_costly_dependence.
653         (struct gcc_target.sched.adjust_cost): Fix typo.
654         (struct gcc_target.sched.is_costly_dependence): Change signature to use
655         single dep_t parameter instead of an equivalent triad.
656         (struct gcc_target.sched.adjust_cost_2): Remove.
658         * target-def.h (TARGET_SCHED_ADJUST_COST_2): Remove.
660         * reg-notes.def (DEP_TRUE, DEP_OUTPUT, DEP_ANTI): Update comments.
662         * doc/tm.texi (TARGET_SCHED_IS_COSTLY_DEPENDENCE): Update
663         documentation.
664         (TARGET_SCHED_ADJUST_COST_2): Remove documentation.
666         * doc/rtl.texi (LOG_LINKS): Remove part about instruction scheduler.
667         (REG_DEP_TRUE): Document.
668         
669         * config/ia64/ia64.c (ia64_adjust_cost_2): Rename to ia64_adjust_cost.
670         Change signature to correspond to the targetm.sched.adjust_cost hook.  
671         Update use in TARGET_SCHED_ADJUST_COST_2.
672         (TARGET_SCHED_ADJUST_COST_2): Rename to TARGET_SCHED_ADJUST_COST.
673         (ia64_dependencies_evaluation_hook, ia64_dfa_new_cycle): Update to use
674         new scheduler dependencies lists.
675         (ia64_gen_check): Ditto.
677         * config/mips/mips.c (vr4130_swap_insns_p): Update to use new scheduler
678         dependencies lists.
679         
680         * config/rs6000/rs6000.c (rs6000_is_costly_dependence): Change
681         signature to correspond to the targetm.sched.is_costly_dependence hook.
682         (is_costly_group): Update to use new scheduler dependencies lists.
684         * config/spu/spu.c (spu_sched_adjust_cost): Use insn_cost () function
685         instead of INSN_COST () macro.
687 2007-02-01  Ian Lance Taylor  <iant@google.com>
689         * lower-subreg.c (resolve_clobber): Handle a subreg of a concatn.
691 2007-02-01  Guy Martin  <gmsoft@gentoo.org>
693         * pa.md (tp_load): Correct mfctl instruction syntax.
695 2007-02-01  Geoffrey Keating  <geoffk@apple.com>
697         * config/rs6000/rs6000.c (rs6000_stack_info): Correct
698         altivec_padding_size calculation on AIX.  Improve comment, add
699         assert to verify that it's right.
701         * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
702         64-bit.
704 2007-2-01   Seongbae Park <seongbae.park@gmail.com>
706         PR inline-asm/28686
707         * global.c (global_alloc): Add mising initialization of
708         ELIMINABLE_REGSET.
710 2007-02-01  Roger Sayle  <roger@eyesopen.com>
712         * alias.c (init_alias_analysis): Correct whitespace.
713         * bb-reorder.c (fix_edges_for_rarely_executed_code,
714         partition_hot_cold_basic_blocks): Likewise.
715         * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
716         expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
717         maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
718         fold_builtin_snprintf_chk, fold_builtin_printf,
719         fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
720         do_mpfr_arg3, do_mpfr_sincos): Likewise.
721         * cfgcleanup.c (cleanup_cfg): Likewise.
722         * cfgexpand.c (tree_expand_cfg): Likewise.
723         * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
724         * function.c (get_next_funcdef_no): Likewise.
725         * gengtype.c (main): Likewise.
726         * genmodes.c (main): Likewise.
727         * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
728         * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
729         * ifcvt.c (noce_emit_move_insn): Likewise.
730         * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
731         Likewise.
732         * stor-layout.c (get_best_mode): Likewise.
733         * tree-ssa-loop-niter.c (get_val_for): Likewise.
734         * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
735         scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
736         perform_var_substitution, solve_graph): Likewise.
737         * tree-vrp.c (vrp_finalize): Likewise.
739 2007-02-01  Ian Lance Taylor  <iant@google.com>
741         * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
742         high part of a paradoxical subreg, return a constant zero.
744 2007-02-01  Zdenek Dvorak <dvorakz@suse.cz>
746         * toplev.c (lang_dependent_init): Call init_set_costs.
747         * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
749 2007-02-01  Richard Guenther  <rguenther@suse.de>
751         PR middle-end/30656
752         * fold-const.c (fold_negate_expr): Allow negating a
753         constant if overflow does not change.
755 2007-02-01  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
757         * doc/c-tree.texi (Expression trees): Improve markup.
758         * doc/tm.texi (Register Classes, Addressing Modes)
759         (Floating Point): Fix spacing after abbreviations.  Fix some
760         typos.
762 2007-02-01  Ben Elliston  <bje@au.ibm.com>
764         * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
765         * doc/contrib.texi: Likewise.
766         * doc/install.texi: Likewise.
768 2007-01-31  Richard Henderson  <rth@redhat.com>
769             Ian Lance Taylor  <iant@google.com>
771         * lower-subreg.c: New file.
772         * rtl.def (CONCATN): Define.
773         * passes.c (init_optimization_passes): Add pass_lower_subreg and
774         pass_lower_subreg2.
775         * emit-rtl.c (update_reg_offset): New static function, broken out
776         of gen_rtx_REG_offset.
777         (gen_rtx_REG_offset): Call update_reg_offset.
778         (gen_reg_rtx_offset): New function.
779         * regclass.c: Revert patch of 2006-03-05, restoring
780         reg_scan_update.
781         (clear_reg_info_regno): New function.
782         * dwarf2out.c (concatn_loc_descriptor): New static function.
783         (loc_descriptor): Handle CONCATN.
784         * common.opt (fsplit_wide_types): New option.
785         * opts.c (decode_options): Set flag_split_wide_types when
786         optimizing.
787         * timevar.def (TV_LOWER_SUBREG): Define.
788         * rtl.h (gen_reg_rtx_offset): Declare.
789         (reg_scan_update): Declare.
790         * regs.h (clear_reg_info_regno): Declare.
791         * tree-pass.h (pass_lower_subreg): Declare.
792         (pass_lower_subreg2): Declare.
793         * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
794         (Optimize Options): Add -fsplit-wide-types to -O1 list.  Document
795         -fsplit-wide-types.
796         * doc/rtl.texi (Regs and Memory): Document concat and concatn.
797         * Makefile.in (OBJS-common): Add lower-subreg.o.
798         (lower-subreg.o): New target.
800 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
802         * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
804 2007-01-31  Anatoly Sokolov <aesok@post.ru>
806         PR target/19087
807         * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
809 2007-01-31  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
811         PR middle-end/29335
812         * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
814 2007-01-31  Zdenek Dvorak <dvorakz@suse.cz>
816         * cfgloop.h: Include vec-prim.h.
817         (enum li_flags): Remove LI_ONLY_OLD.
818         (loop_iterator): Changed.
819         (fel_next, fel_init): Iterate over loop tree.
820         (FOR_EACH_LOOP_BREAK): New macro.
821         * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
822         FOR_EACH_LOOP.
823         * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
824         * modulo-sched.c (sms_schedule): Ditto.
825         * tree-vectorizer.c (vectorize_loops): Ditto.
826         * doc/loop.texi: Update information on loop numbering and behavior of
827         FOR_EACH_LOOP wrto new loops.
828         * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
829         add_to_evolution_1): Test nestedness of loops instead of comparing
830         their numbers.
831         * tree-chrec.c (chrec_fold_plus_poly_poly,
832         chrec_fold_multiply_poly_poly, chrec_evaluate,
833         hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
834         reset_evolution_in_loop): Ditto.
835         * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
837 2007-01-31  Dirk Mueller  <dmueller@suse.de>
839         * c-common.c (warn_about_parentheses): Separate warning about
840         un-parenthized sequence of comparison operators from the one
841         which is supposed to warn about x <= y <= z.
843 2007-01-31  Uros Bizjak  <ubizjak@gmail.com>
845         * optabs.h (enum optab_index): Add new OTI_isinf.
846         (isinf_optab): Define corresponding macro.
847         * optabs.c (init_optabs): Initialize isinf_optab.
848         * genopinit.c (optabs): Implement isinf_optab using isinf?f2
849         patterns.
850         * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
851         (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
852         using isinf_optab.
853         (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
854         expand_builtin_interclass_mathfn.
855         * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
856         * config/i386/i386.md (UNSPEC_FXAM): New constant.
857         (fxam<mode>2_i387): New insn pattern.
858         (isinf<mode>2) New expander to implement isinf, isinff and isinfl
859         built-in functions as x87 inline asm.
861 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
863         * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
865 2007-01-30  Eric Christopher  <echristo@apple.com>
867         * config.gcc: Add geode.
869 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
871         * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
872         ipa-inline.c, tree-profile.c, tree-ssa-live.c,
873         tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
874         value-prof.c: Fix comment typos.
876 2007-01-30  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
878         PR c++/24745
879         * doc/invoke.texi (Wpointer-arith): Document warning.
881 2007-01-30  Janis Johnson
883         * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
884         the list of discrepancies from the draft TR.
886 2007-01-30  Dirk Mueller  <dmueller@suse.de>
888         PR c++/30601
889         * doc/invoke.texi (-Wreturn-type): Update description to
890         match new behavior.
892 2007-01-30  Richard Sandiford  <richard@codesourcery.com>
894         * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
896 2007-01-30  Uros Bizjak  <ubizjak@gmail.com>
898         * builtins.c (expand_builtin_int_interclass_roundingfn): New function
899         to handle optabs that operate on floating point input argument and
900         output to integer output.
901         (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
902         (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
903         (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
904         expand_builtin_interclass_mathfn ().
905         * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
906         (ilogbsi2): Remove.
907         (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
908         ilogbl built-in functions as x87 intrinsics.
910 2007-01-30  Richard Guenther  <rguenther@suse.de>
912         PR middle-end/27657
913         * dwarf2out.c (reference_to_unused): Query varpool if the
914         variable was output.
916 2007-01-30  Richard Guenther  <rguenther@suse.de>
918         PR middle-end/30313
919         * passes.c (execute_one_pass): Reset in_gimple_form to not
920         confuse non-unit-at-a-time mode.
922 2007-01-29  Roger Sayle  <roger@eyesopen.com>
923             Richard Guenther  <rguenther@suse.de>
925         * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
926         avoid undefined behaviour on overflow.  Use force_fit_type_double
927         to construct the constant with the specified TREE_OVERFLOW.
929 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
931         * config/dfp-bit.c: Add parameterized support for fp exceptions.
932         * config/dfp-bit.h: Ditto.
934 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
936         * c-decl.c (pop_scope): Replace warnings with call to
937         warn_for_unused_label.
938         * c-common.h (warn_for_unused_label): Declare.
939         * c-common.c (warn_for_unused_label): Define.
941 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
943         * tree-optimize.c (update_inlined_to_pointers): Delete unused
944         function.
946 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
948         * Makefile.in (USER_H): Remove decfloat.h.
949         * ginclude/decfloat.h: Delete, moving contents to ...
950         * ginclude/float.h: Add support for decimal floating point,
951         guarded by __STDC_WANT_DEC_FP__.
953 2007-01-29  Mike Stump  <mrs@apple.com>
955         * doc/gccint.texi (Top): Rename Loop Representation to Loop
956         Analysis and Representation to resolve case insensitive conflict.
957         * doc/loop.texi (Loop Analysis and Representation): Likewise.
959 2007-01-28  Daniel Berlin  <dberlin@dberlin.org>
961         * tree.h (struct tree_memory_tag): Add aliases member.
962         (MTAG_ALIASES): New macro.
963         * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
964         (add_may_alias): Remove pointer-set. Update for may_aliases being
965         a bitmap.
966         (mark_aliases_call_clobbered): Update for may_aliases being a
967         bitmap.
968         (compute_tag_properties): Ditto.
969         (create_partition_for): Ditto.
970         (compute_memory_partitions): Ditto.
971         (dump_may_aliases_for): Ditto.
972         (is_aliased_with): Ditto.
973         (add_may_alias_for_new_tag): Ditto.
974         (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
975         (compute_is_aliased): New function.
976         (compute_may_aliases): Call compute_is_aliased).
977         (init_alias_info): Initialize alias_bitmap_obstack.
978         (union_alias_set_into): New function.
979         (compute_flow_sensitive_aliasing): Use union_aliases_into.
980         (have_common_aliases_p): Rewrite to take two bitmaps and use
981         intersection.
982         (compute_flow_insensitive_aliasing): Stop using pointer-sets.
983         Update for bitmaps.
984         (finalize_ref_all_pointers): Update for add_may_alias changes.
985         (new_type_alias): Ditto.
986         * tree-flow-inline.h (may_aliases): Return a bitmap.
987         * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
988         * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
989         may_aliases being a bitmap.
990         * tree-flow.h (struct var_ann_d): Remove may_aliases member.
991         may_aliases now returns a bitmap.
992         * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
993         being a bitmap.
994         * tree-ssa-operands.c (add_virtual_operand): Update for
995         may_aliases being a bitmap.
997 2007-01-29  Daniel Berlin  <dberlin@dberlin.org>
999         PR tree-optimization/30630
1000         * tree-ssa-structalias.c (do_complex_constraint): Mark correct
1001         variable as changed.
1003 2007-01-29  Simon Martin  <simartin@users.sourceforge.net>
1005         PR c++/28266
1006         * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
1007         expanded only once even if an error occurs.
1009 2007-01-29  Ben Elliston  <bje@au.ibm.com>
1011         * gcov-io.h (__gcov_indirect_call_profiler): Declare.
1012         (__gcov_average_profiler): Likewise.
1013         (__gcov_ior_profiler): Likewise.
1014         (__gcov_merge_ior): Likewise.
1016 2007-01-28  Jan Hubicka  <jh@suse.cz>
1018         * builtins.c (expand_builtin_memset): Fix typo in my last patch.
1019         * value-prof.c (stringop_block_profile): Likewise.
1021 2007-01-28  Jan Hubicka  <jh@suse.cz>
1023         * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
1024         variant handling histograms; add wrapper.
1025         (clear_storage_via_libcall): Export.
1026         (emit_block_move_hints): Break out from ...; add histograms.
1027         (emit_block_move): ... this one.
1028         (clear_storage_hints): Break out from ...; add histograms.
1029         (clear_storage): ... this one.
1030         (set_storage_via_memset): Handle histogram.
1031         * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
1032         (clear_storage_hints, clear_storage_via_libcall): Declare.
1033         (set_storage_via_setmem): Update prototype.
1034         * doc/md.texi (movmem, setmem): Document new arguments.
1036         * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
1037         new histograms.
1038         (stringop_block_profile): New global function.
1039         (tree_stringops_values_to_profile): Profile block size and alignment.
1040         * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
1041         HIST_TYPE_IOR.
1042         (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
1043         (stringop_block_profile): Declare.
1044         * builtins.c: Include value-prof.h.
1045         (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
1046         * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
1047         (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
1048         (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
1049         * profile.c (instrument_values): Add new counters.
1050         * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
1051         calls.
1052         * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
1053         (tree_init_edge_profiler): Build new profilers.
1054         (tree_gen_average_profiler, tree_gen_ior_profiler): New.
1055         (pass_tree_profile): Add dump.
1056         (tree_profile_hooks): Update.
1057         * Makefile.in (LIBGCOV): Add new constants.
1058         * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
1059         __gcov_ior_profiler): New.
1060         * i386.md (movmem/setmem expanders): Add new optional arguments.
1062 2007-01-28  David Edelsohn  <edelsohn@gnu.org>
1064         * doc/md.texi (Standard Pattern Names): Document blockage pattern.
1066 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
1068         * tree-ssa-loop-unswitch.c: Include tree-inline.h.
1069         (tree_unswitch_single_loop): Pass eni_size_weights to
1070         tree_num_loop_insns.
1071         * tree-ssa-loop-manip.c: Include tree-inline.h.
1072         (can_unroll_loop_p): Pass eni_size_weights to
1073         tree_num_loop_insns.
1074         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
1075         Pass eni_size_weights to estimate_num_insns.
1076         * tree.h (init_inline_once): Export.
1077         * toplev.c (backend_init): Call init_inline_once.
1078         * cgraphunit.c (cgraph_process_new_functions,
1079         cgraph_analyze_function): Pass eni_inlining_weights to
1080         estimate_num_insns.
1081         * ipa-inline.c (compute_inline_parameters): Ditto.
1082         * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
1083         to estimate_num_insns.
1084         (try_unroll_loop_completely): Pass eni_size_weights to
1085         tree_num_loop_insns.
1086         * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
1087         ot estimate_num_insns.
1088         * tree-ssa-loop-prefetch.c: Include tree-inline.h.
1089         (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
1090         * tree-inline.c (eni_inlining_weights, eni_size_weights,
1091         eni_time_weights): New variables.
1092         (init_inline_once): Initialize them.
1093         (struct eni_data): Mew.
1094         (estimate_num_insns_1, estimate_num_insns): Use weights.
1095         * tree-inline.h (struct eni_weights_d): New.
1096         (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
1097         (estimate_num_insns): Declaration changed.
1098         * cfgloop.h (tree_num_loop_insns): Declaration changed.
1099         * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
1100         tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
1102 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
1104         * tree-data-ref.c (conflict_fn): Assert that the number of affine
1105         relations in the conflict function is valid.
1107 2007-01-27  Ian Lance Taylor  <iant@google.com>
1109         * common.opt: Add fstrict-overflow.
1110         * opts.c (decode_options): Set flag_strict_overflow if -O2.
1111         * flags.h (TYPE_OVERFLOW_WRAPS): Define.
1112         (TYPE_OVERFLOW_UNDEFINED): Define.
1113         (TYPE_OVERFLOW_TRAPS): Define.  This replaces TYPE_TRAP_SIGNED.
1114         Replace all uses.
1115         * tree.h (TYPE_TRAP_SIGNED): Don't define.
1116         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
1117         (fold_negate_expr): Likewise.
1118         (make_range): Likewise.
1119         (extract_muldiv_1): Likewise.
1120         (maybe_canonicalize_comparison): Likewise.
1121         (fold_comparison): Likewise.
1122         (fold_binary): Likewise.
1123         (tree_expr_nonnegative_p): Likewise.
1124         (tree_expr_nonzero_p): Likewise.
1125         * tree-vrp.c (compare_values): Likewise.
1126         (extract_range_from_binary_expr): Likewise.
1127         (extract_range_from_unary_expr): Likewise.
1128         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
1129         Likewise.
1130         (nowrap_type_p): Likewise.
1131         * tree-scalar-evolution.c (simple_iv): Likewise.
1132         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
1133         (build_range_check): Likewise.
1134         (extract_muldiv_1): Likewise.
1135         (fold_comparison): Likewise.
1136         * tree-vrp.c (vrp_int_const_binop): Likewise.
1137         (extract_range_from_unary_expr): Likewise.
1138         * convert.c (convert_to_integer): Likewise.
1139         * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
1140         (fold_comparison): Likewise.
1141         (fold_binary): Likewise.
1142         * optabs.c (optab_for_tree_code): Likewise.
1143         * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
1144         * simplify-rtx.c (simplify_const_relational_operation): Check
1145         flag_strict_overflow and flag_trapv.
1146         (simplify_const_relational_operation): Likewise.
1147         * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
1148         (Optimize Options): Add -fstrict-overflow to -O2 list.  Document
1149         -fstrict-overflow.
1151 2007-01-27  Roger Sayle  <roger@eyesopen.com>
1153         * tree.c (tree_fold_gcd): Delete.
1154         * tree.h (tree_fold_gcd): Remove prototype.
1155         * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
1156         test whether one constant integer is a multiple of another.  Instead
1157         call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
1158         * fold-const.c (multiple_of_p):  We've determined both TOP and
1159         BOTTOM are integer constants so we can call int_const_binop directly
1160         instead of the more generic const_binop.
1162 2007-01-27  Roger Sayle  <roger@eyesopen.com>
1164         * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
1165         1*X check that the constant hasn't overflowed, to preserve the
1166         TREE_OVERFLOW bit.
1167         (round_up): Provide an efficient implementation when rouding-up an
1168         INTEGER_CST to a power-of-two.
1170 2007-01-28  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
1172         * doc/sourcebuild.texi: Add comma for clarity.
1173         * doc/extend.texi: Fix some typos.
1174         * doc/passes.texi: Likewise.
1175         * doc/cppinternals.texi: Likewise.
1176         * doc/c-tree.texi: Likewise.
1177         * doc/tree-ssa.texi: Likewise.
1178         * doc/install.texi: Likewise.
1180 2007-01-27  Jan Hubicka  <jh@suse.cz>
1182         * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
1183         (pass_sra): Do not require alias information.
1184         * passes.c (init_optimization_passes): Add SRA
1186 2007-01-27  Steven Bosscher  <steven@gcc.gnu.org>
1188         * tracer.c (rest_of_handle_tracer): We already cleaned
1189         up the CFG in tracer() so don't do it here again.
1190         * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
1191         cleanup_cfg here, either.  And don't call renumber_insns.
1193         * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
1194         and BB_HEAD are in the insn stream here.  Instead make sure
1195         that BB_INSN is valid on all insns.  Also, do check here that
1196         there are no pending branch predictions...
1197         (rtl_verify_flow_info): ...instead of doing it here.  Checks
1198         for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
1199         here.
1201 2007-01-26  Roger Sayle  <roger@eyesopen.com>
1203         * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
1204         function to simplify/factorize operand order canonicalization.
1205         (ix86_fixup_binary_operands): Reorganize using the above function.
1206         (ix86_binary_operator_ok): Likewise.
1208 2007-01-27  Jakub Jelinek  <jakub@redhat.com>
1210         * genattrtab.c (struct attr_value_list, insn_code_values): Move to
1211         file scope from optimize_attrs.
1212         (simplify_test_exp): If insn_code_values is not NULL, use it to speed
1213         up search.
1214         (optimize_attrs): Clear insn_code_values after freeing it.
1216 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
1218         * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
1219         a parts.base assignment.
1221 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
1223         * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
1224         (compute_subscript_distance, initialize_data_dependence_relation,
1225         finalize_ddr_dependent, analyze_ziv_subscript,
1226         analyze_siv_subscript_cst_affine,
1227         compute_overlap_steps_for_affine_univar,
1228         compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
1229         analyze_siv_subscript, analyze_miv_subscript,
1230         analyze_overlapping_iterations, subscript_dependence_tester_1,
1231         compute_self_dependence, free_dependence_relation): Work
1232         with affine_fn instead of chrecs.
1233         (dump_affine_function, dump_conflict_function, affine_function_equal_p,
1234         common_affine_function, affine_function_base,
1235         affine_function_constant_p, affine_fn_op, affine_fn_plus,
1236         affine_fn_minus, affine_fn_free, conflict_fn_not_known,
1237         conflict_fn_no_dependence, free_conflict_function, free_subscripts,
1238         conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
1239         (all_chrecs_equal_p): Removed.
1240         * tree-data-ref.h (affine_fn, conflict_function): New types.
1241         (struct subscript): Change type of conflicting_iterations_in_a
1242         and conflicting_iterations_in_b.
1244 2007-01-26  Steve Ellcey  <sje@cup.hp.com>
1246         PR other/30182
1247         * config/pa/pa.h (TARGET_HPUX_11): New.
1248         * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
1249         * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
1251 2007-01-26  Daniel Berlin  <dberlin@dberlin.org>
1252         Richard Guenther  <rguenther@suse.de>
1254         * tree-ssa-structalias.c (solve_graph): Handle case
1255         we merged the variable to another.
1257 2007-01-25  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
1259         * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
1260         functions.
1262         * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
1263         catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
1265 2007-01-25  DJ Delorie  <dj@redhat.com>
1267         * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
1268         allow changes to modes which don't fit in those registers.
1270         * reload1.c (choose_reload_regs): Check for invalid subregs before
1271         computing their locations, not after.
1273 2007-01-25  Geoffrey Keating  <geoffk@apple.com>
1275         PR 25127
1276         * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
1277         save Altivec registers in an eh_return function.
1278         (compute_vrsave_mask): Likewise.
1279         (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
1280         for saving Altivec registers.
1281         (rs6000_emit_prologue): Don't allocate stack twice in
1282         eh_return function.  Correct expected value of altivec_save_offset
1283         when using save_world.  Describe save of R0 to stack when using
1284         save_world.  Describe stack pointer adjustment when using
1285         save_world.  Remove duplicated eh_return parameter register saving.
1286         Update sp_offset variable after save_world.
1287         * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
1288         darwin-world.asm.
1289         (LIB2FUNCS_EXTRA): Add darwin-world.asm.
1290         * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
1291         implies Altivec.
1293 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
1295         * config.gcc (ia64*-*-hpux*): Make posix threads the default.
1297 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
1299         PR other/30182
1300         * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
1301         * config/ia64/ia64.c (ia64_init_builtins):  Ditto.
1303 2007-01-25  Richard Guenther  <rguenther@suse.de>
1305         * doc/invoke.texi (-Wcoverage-mismatch): Document.
1306         * common.opt (-Wcoverage-mismatch): New warning option.
1307         * coverage.c (get_coverage_counts): Ignore coverage mismatch
1308         if -Wcoverage-mismatch is given.
1310 2007-01-25  Razya Ladelsky  <razya@il.ibm.com>
1312         * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
1313         (ipcp_driver): Change to static definition.
1314         Add dumping of the ifunctions.
1315         (constant_val_insert): Remove unused parameter. Support for SSA.
1316         (ipcp_propagate_const): Support for SSA.
1317         (ipcp_profile_bb_print): Print only analyzed nodes.
1318         (ipcp_replace_map_create): Remove support for Fortran constant
1319         for now.
1320         * ipa-prop.c (ipa_method_modify_stmt,
1321         ipa_callsite_compute_param): Support for SSA.
1322         * ipa-prop.h (ipcp_driver): Remove declaration.
1323         (IS_VALID_TREE_MAP_INDEX): Add define.
1325 2007-01-24  Geoffrey Keating  <geoffk@apple.com>
1327         * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
1329 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
1331         PR middle-end/30447
1332         * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
1333         constant argument to cabs and do it without checking for
1334         -funsafe-math-optimizations.
1336 2007-01-24  Douglas Gregor  <dgregor@osl.iu.edu>
1338         * c-common.h (RID_FIRST_CXX0X): New.
1339         (RID_LAST_CXX0X): New.
1340         * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
1341         by -Wall.
1342         * c.opt (Wc++0x-compat): New.
1343         * doc/invoke.texi (-Wc++0x-compat): Document.
1345 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
1347         * builtins.c (fold_builtin_carg): New.
1348         (fold_builtin_1): Use it.
1350 2007-01-24  Jan Hubicka  <jh@suse.cz>
1352         * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
1354         * ipa-inline.c (initial_insns, max_insns): Delete.
1355         (compute_max_insns): New function.
1356         (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
1357         of insns as base for code growth.
1358         (cgraph_decide_inlining): Make initial_insns local; do not compute
1359         max_insns.
1360         * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
1361         * doc/invoke.texi (inline-unit-growth): Update docs.
1363 2007-01-24  Jakub Jelinek  <jakub@redhat.com>
1365         * config/i386/i386.h (x86_cmpxchg16b): Remove const.
1366         (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
1367         * config/i386/i386.c (x86_cmpxchg16b): Remove const.
1368         (override_options): Add PTA_CX16 flag.  Set x86_cmpxchg16b
1369         for CPUs that have PTA_CX16 set.
1371         PR middle-end/27416
1372         * gimplify.c (omp_check_private): New function.
1373         (gimplify_scan_omp_clauses): Use it for
1374         firstprivate/lastprivate/reduction.
1376         PR middle-end/30494
1377         * gimplify.c (omp_add_variable): Don't call omp_notice_variable
1378         on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
1380         PR middle-end/30421
1381         * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
1382         If lower_lastprivate_clauses emits some statements, append them
1383         to dlist rather than body_p and to body_p append an initializer.
1384         (lower_omp_for): Adjust caller.
1386 2007-01-24  Steve Ellcey  <sje@cup.hp.com>
1388         * target.h (globalize_decl_name): New.
1389         * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
1390         * output.h (default_globalize_decl_name): New.
1391         * varasm.c (asm_output_bss): Use globalize_decl_name instead of
1392         globalize_label.
1393         (globalize_decl): Ditto.
1394         (default_globalize_decl_name): New.
1395         * config/ia64/ia64.c (ia64_globalize_decl_name): New.
1396         (ia64_handle_version_id_attribute): New.
1397         (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
1398         (ia64_asm_output_external): Use globalize_decl_name instead
1399         of globalize_label.
1400         * doc/extend.texi (version_id): New pragma.
1401         * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
1403 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
1405         * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
1406         _uleb128_t and _Unwind_SWord with _sleb128_t.
1407         * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
1408         uw_frame_state_for, uw_update_context_1): Likewise.
1409         * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
1410         * unwind-pe.h (read_uleb128, read_sleb128,
1411         read_encoded_value_with_base): Likewise.
1412         * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
1414 2007-01-24  Richard Guenther  <rguenther@suse.de>
1416         * builtins.c (expand_builtin_cexpi): Get the fndecl
1417         for cexp in the correct way.
1419 2007-01-24  Jan Hubicka  <jh@suse.cz>
1421         * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
1423 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
1425         * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
1426         (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
1427         __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
1428         __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
1429         __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
1430         __SIZEOF_PTRDIFF_T__.
1431         * doc/cpp.texi: Documentation for the new builtin macros added.
1433 2007-01-24  Uros Bizjak  <ubizjak@gmail.com>
1435         * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
1436         log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
1437         instead of operands[2] to avoid access past the end of array.
1439 2007-01-24  Richard Sandiford  <richard@codesourcery.com>
1441         * reload1.c (emit_reload_insns): Pass the reload register
1442         for a non-spill output reload through forget_old_reloads_1.
1444 2007-01-23  Joseph Myers  <joseph@codesourcery.com>
1446         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
1447         _SOFT_DOUBLE if doubles use software floating-point.
1448         * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
1449         double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
1450         * config/rs6000/darwin-ldouble.c: Also compile functions for
1451         hard-float without FPRs.  Use fmsub function for all __NO_FPRS__
1452         cases.  Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
1453         * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
1454         commented-out long double override.
1455         (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
1456         * config/rs6000/eabispe.h: Likewise.
1457         * config/rs6000/rs6000.c (rs6000_override_options): Don't override
1458         long double for non-SPE.
1459         (rs6000_handle_option): Likewise.
1460         (invalid_e500_subreg): Disallow more subregs involding DImode,
1461         DFmode, TImode or TFmode.
1462         (rs6000_legitimate_offset_address_p): Check TFmode offsets for
1463         E500 double.
1464         (legitimate_lo_sum_address_p): Also check for TFmode for E500
1465         double.
1466         (rs6000_legitimize_address): Also handle TFmode for E500 double.
1467         (rs6000_legitimize_reload_address): Also handle TFmode for E500
1468         double.
1469         (rs6000_legitimate_address): Also check for TFmode for E500
1470         double.
1471         (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
1472         (spe_build_register_parallel): Handle TFmode and TCmode.
1473         (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
1474         double.
1475         (function_arg): Handle TFmode and TCmode for E500 double.
1476         (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
1477         in general.
1478         (print_operand): Handle TFmode and TImode for %y.
1479         (rs6000_generate_compare): Handle TFmode comparisons for E500
1480         double.
1481         (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
1482         (rs6000_function_value): Handle TFmode and TCmode for E500 double.
1483         (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
1484         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
1485         TFmode for E500 double.
1486         * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
1487         (floatsidf2): Enable for E500 double.
1488         (movtf_softfloat): Use rs6000_nonimmediate_operand.
1489         (extenddftf2): Change to extenddftf2_fprs.
1490         (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
1491         depending on TARGET_E500_DOUBLE.
1492         (extendsftf2): Enable for E500 double.
1493         (trunctfdf2): Enable for E500 double.
1494         (trunctfsf2): Change to trunctfsf2_fprs.
1495         (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
1496         depending on TARGET_E500_DOUBLE.
1497         (floatsitf2): Enable for E500 double.
1498         (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
1499         (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
1500         gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
1501         (negtf2): Change to negtf2_internal.
1502         (negtf2): New expander.
1503         (abstf2): Enable for E500 double.  Call gen_spe_abstf2_tst,
1504         gen_spe_abstf2_cmp or gen_abstf2_internal depending on
1505         TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
1506         (movdi_internal32): Use rs6000_nonimmediate_operand.
1507         (unnamed splitter): Likewise.
1508         * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
1509         TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
1510         (SPE64TF, DITI): New mode macros.
1511         (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
1512         modes.
1513         (frob_tf_ti): New.
1514         (frob_<mode>_di_2): New.
1515         (frob_tf_di_8_2): New.
1516         (frob_di_df): Change to frob_di_<mode>; allow more modes.
1517         (frob_ti_tf): New.
1518         (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
1519         more modes.
1520         (frob_ti_<mode>_8_2): New.
1521         (frob_ti_tf_2): New.
1522         (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
1523         mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
1524         TFmode.
1525         (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
1526         mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
1527         (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
1528         spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
1529         spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
1530         (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
1531         tsttflt_gp): New.
1533 2007-01-23  Ian Lance Taylor  <iant@google.com>
1535         * Makefile.in (OBJS-common): Reformat, alphabetize, but put
1536         insn-*.o first.
1537         (OBJS-archive): Reformat, alphabetize.
1538         (OBJS): Change out_object_file to OBJS-md.
1540 2007-01-23  Uros Bizjak  <ubizjak@gmail.com>
1542         * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
1543         JUMP_P and CALL_P predicates where applicable.
1544         * config/i386/i386.c: Ditto.
1545         * config/i386/i386.md: Ditto.
1546         * config/i386/mmx.md: Ditto.
1547         * config/i386/predicates.md: Ditto.
1549 2007-01-22  Andreas Schwab  <schwab@suse.de>
1551         * config/m68k/m68k.h: Fix comment.
1553 2007-01-22  Jan Hubicka  <jh@suse.cz>
1555         * passes.c (init_optimization_passes): Do not rerun
1556         pass_early_warn_uninitialized.
1558 2007-01-22  Richard Guenther  <rguenther@suse.de>
1560         PR tree-optimization/30038
1561         * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
1562         function.
1563         (execute_cse_sincos_1): Likewise.
1564         (execute_cse_sincos): Likewise.
1565         (gate_cse_sincos): Likewise.
1566         (pass_cse_sincos): New pass CSEing sin() and cos() calls using
1567         the cexpi() canonicalization of sincos().
1568         * tree-pass.h (pass_cse_sincos): Declare.
1569         * passes.c (init_optimization_passes): New pass pas_cse_sincos.
1571 2007-01-21  Eric Botcazou  <ebotcazou@libertysurf.fr>
1573         PR rtl-optimization/29329
1574         * combine.c (replaced_rhs_insn): Rename to i2mod.
1575         (replaced_rhs_value): Rename to i2mod_new_rhs.
1576         (i2mod_old_rhs): New global variable.
1577         (combine_instructions): Adjust for above change.  Save a copy of
1578         the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
1579         note are substituted in the second instruction.
1580         (distribute_notes) <REG_DEAD>: Adjust for above change.  Do not
1581         ditch the note if it pertains to the second eliminated register
1582         and this register is mentioned in i2mod_old_rhs.
1584         Revert:
1585         2006-09-12  Eric Botcazou  <ebotcazou@libertysurf.fr>
1587         * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
1588         the insn to which the note was originally attached.
1590 2007-01-21  Jan Hubicka  <jh@suse.cz>
1592         * ipa-inline.c (inlining_mode): Comment, move up.
1593         (cgraph_decide_inlining_incrementally): Do not perform inlining
1594         itself; fix handling of flattening of self recursive functions.
1595         (cgraph_find_cycles): Remove.
1596         (cgraph_flatten_node): Remove.
1597         (cgraph_decide_inlining): Use incremental inliner to handle flattening.
1598         (try_inline): New function.
1599         (cgraph_early_inlining): Update call of
1600         cgraph_decide_inlining_incrementally.  Apply inlining here.
1601         (apply_inline): Update call of cgraph_decide_inlining_incrementally.
1603 2007-01-21  Dirk Mueller  <dmueller@suse.de>
1605         PR bootstrap/30511
1606         * tree-vrp.c (check_array_bounds): do not warn
1607         about ADDR_EXPR's of ARRAY_REF's which are immediately
1608         used in binary expressions.
1610 2007-01-21  Ira Rosen  <irar@il.ibm.com>
1612         * tree-vectorizer.h (struct _stmt_vec_info): Add new field
1613         read_write_dep and macros for its access.
1614         * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
1615         * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
1616         argument, call vect_check_interleaving for every independent pair of
1617         data-refs. Mark loads that access the same memory location as a store
1618         in the loop.
1619         (vect_check_dependences): Remove.
1620         (vect_analyze_data_ref_dependences): Remove  vect_check_dependences
1621         call, fix the call to vect_analyze_data_ref_dependence.
1622         (vect_analyze_data_ref_access): For statements that access the same
1623         data-ref, check that they are not stores; for loads, check that there
1624         is no store that access the same location.
1626 2007-01-20  Roger Sayle  <roger@eyesopen.com>
1627             Joseph Myers  <joseph@codesourcery.com>
1629         * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
1630         (-fdump-rtl-stack): Likewise.
1631         (-fno-signed-zeros): Use @minus{} for a minus sign.  Correct typo.
1632         (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
1633         (-mpt-fixed): Use @minus{} for minus sign.
1634         (Using Precompiled Headers): Fix under/overfull hbox.
1636 2007-01-20  Jan Hubicka  <jh@suse.cz>
1638         * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
1639         place within annotation.
1641 2007-01-20  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
1643         * pa.c (output_move_double): Change array size of xoperands to 4.
1645 2007-01-20  Jan Hubicka  <jh@suse.cz>
1647         * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
1648         as temporary.
1650 2007-01-19  Ian Lance Taylor  <iant@google.com>
1652         * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
1653         constant to unsigned HOST_WIDE_INT before negating.
1655 2007-01-19  Ian Lance Taylor  <iant@google.com>
1657         * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
1658         to unsigned.
1659         (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
1660         (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
1661         (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
1662         (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
1663         (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
1664         (realloc_vdef, realloc_vuse): Change second parameter to
1665         unsigned.
1666         (ssa_operand_iterator_d): Change vuse_index and mayuse_index
1667         fields to unsigned.
1668         * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
1669         unsigned. Change x and lim locals to unsigned.
1670         (realloc_vdef, realloc_vuse): Change num_elem parameter to
1671         unsigned.
1672         (finalize_ssa_vuse_ops): Change old_i local to unsigned.
1673         (copy_virtual_operands): Change i and n locals to unsigned.
1675 2007-01-19  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
1677         PR middle-end/29335
1678         * builtins.c (fold_builtin_1): Handle builtin fdim.
1680 2007-01-20  Jan Hubicka  <jh@suse.cz>
1682         * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
1683         * tree-ssa-structalias.c (compute_points_to_sets): Do call
1684         init_alias_heapvars.
1685         (init_alias_heapvars): Initialize only when not already initialized.
1686         (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
1688 2007-01-19  Roger Sayle  <roger@eyesopen.com>
1690         * common.opt (fsigned-zeros): New command line option.
1691         * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
1692         of flag_unsafe_math_optimizations.
1693         * opts.c (set_fast_math_flags): The -ffast-math command line option
1694         implies -fno-signed-zeros.
1695         (fast_math_flags_set_p): Likewise.
1697         * doc/invoke.texi: Document new -fno-signed-zeros option, and update
1698         the documentation of -ffast-math appropriately.  Wrap long lines.
1700 2007-01-19  Steve Ellcey  <sje@cup.hp.com>
1702         * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
1703         * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
1705 2007-01-19  Tomas Bily  <tbily@suse.cz>
1707         * cgraphunit.c (cgraph_finalize_function): Updating of pid
1708         * tree-profile.c:
1709         (tree_init_ic_make_global_vars): New function
1710         (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
1711         (tree_gen_ic_profiler): New function
1712         (tree_gen_ic_func_profiler): New function
1713         (tree_profiling): Added calling of tree_gen_ic_func_profiler
1714         (tree_profile_hooks): Added hook for indirec/virtual calls
1715         * value-prof.c (tree_find_values_to_profile): New case for
1716         indirect calls
1717         (tree_values_to_profile): Call for determining indirect/virtual
1718         counters
1719         (tree_indirect_call_to_profile): New function
1720         (tree_ic_transform): New function
1721         (tree_ic): New function
1722         (find_func_by_pid): New function
1723         (init_pid_map): New function
1724         (tree_value_profile_transformations): Added check for
1725         indirect/virtual call transformation
1726         * value-prof.h (enum hist_type): New counter type for
1727         indirect/virtual calls
1728         (profile_hooks): Added new hook for profiling indirect/virtual
1729         calls
1730         * profile.c (instrument_values): New case for indirect/virtual
1731         call added
1732         * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
1733         (GCOV_COUNTER_V_INDIR): New counter type
1734         (GCOV_COUNTER_NAMES): New name of counter "indirect" added
1735         (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
1736         call added
1737         * cgraph.c: Definition of cgraph_max_pid
1738         (cgraph_create_node): Default init of pid attribute
1739         * cgraph.h: Declaration of cgraph_max_pid
1740         (struct cgraph_node): Added pid attribute
1741         * libgcov.c (__gcov_indirect_call_profiler): New function
1742         (__gcov_one_value_profiler_body): New function
1743         (__gcov_one_value_profiler): Body was moved to
1744         __gcov_one_value_profiler_body and calls it
1746 2007-01-19  Basile Starynkevitch  <basile@starynkevitch.net>
1748         * doc/gty.texi (Options): Document the mark_hook option to GTY.
1749         * gengtype.c (write_types_data, write_func_for_structure,
1750         write_types, ggc_wtd, pch_wtd): Add skip_hooks to
1751         write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
1752         (walk_type, write_func_for_structure): Generate the mark_hook if
1753         needed.
1755 2007-01-19  Jan Hubicka  <jh@suse.cz>
1757         * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
1758         'early' argument take inlining mode argument specifying whether to
1759         inline for size/speeed or all functions; add support for flattening;
1760         improve dumpting.
1761         (cgraph_early_inlining): Update call of decide_inlining_incrementally.
1763 2007-01-19  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
1765         PR c++/17947
1766         * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
1767         the name of the declared identifier.
1769 2007-01-19  Dirk Mueller  <dmueller@suse.de>
1771         * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
1772         result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
1773         duplicate evaluation.
1775 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
1777         * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
1778         Macroize expander using X87MODEF12 mode macro.  Extend operand 1
1779         to XFMode, use acosxf2 and truncate result to requested mode.
1780         Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1781         (asin<mode>2): Similarly, with asin expanders.
1782         (*fscalexf4_i387): Rename from *fscalexf4.
1783         (expNcorexf3): New expander.
1784         (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
1785         (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
1786         X87MODEF12 mode macro.  Extend operand 1 to XFMode, use expxf2 and
1787         truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
1788         patterns for SSE math.
1789         (exp10<mode>2): Similarly, with exp10 expanders.
1790         (exp2<mode>2): Similarly, with exp2 expanders.
1791         (expm1<mode>2): Similarly, with expm1 expanders.
1792         (ldexp<mode>3): Similarly, with ldexp expanders.
1793         (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
1794         Use gen_truncxf<mode>2_i387_noop to truncate result.
1796 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1798         * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
1799         for ColdFire targets.
1801 2007-01-19  Nathan Sidwell  <nathan@codesourcery.com>
1802             Richard Sandiford  <richard@codesourcery.com>
1804         * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
1805         (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
1806         (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
1807         * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
1808         (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1809         * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1811 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1813         * config/m68k/m68k.md (adddi_dilshr32): Rename to...
1814         (*adddi_dilshr32): ...this.  Fix formatting.  Remove commented-out
1815         non-canonical pattern.  Restrict to !TARGET_COLDFIRE.
1816         (*adddi_dilshr32_cf): New pattern.
1817         (adddi3, subdi3): Remove first alternatives.
1819 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1821         * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
1822         implemented using move.l, do not use its cc result for floating-point
1823         comparisons.
1825 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1827         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
1828         from accepting 's' constraints if flag_pic.
1830 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1832         * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
1833         (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
1834         (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
1835         condition from TARGET_68881 to TARGET_HARD_FLOAT.
1837 2007-01-19  Sandra Loosemore  <sandra@codesourcery.com>
1839         * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
1840         ColdFire alternatives.
1841         * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
1842         * config/m68k/m68k.md (clzsi2):  Define for ColdFire
1843         architectures that support ff1 instruction.
1845 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1846             Julian Brown  <julian@codesourcery.com>
1848         * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
1849         * config/m68k/m68k.md (*movsi_cfv4): Fold into...
1850         (*movsi_cf): ...here.  Remove unnecessary 'R' from 'Rg'.
1851         Add commentary.
1853 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1855         * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
1856         and return a bool.
1857         (output_move_const_into_data_reg, output_move_simode_const): Delete.
1858         * config/m68k/m68k.c (const_method, const_int_cost): Take a
1859         HOST_WIDE_INT instead of an rtx.
1860         (m68k_rtx_costs): Update call accordingly.
1861         (output_move_const_into_data_reg): Likewise.  Fix formatting.
1862         (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
1863         Return a bool.
1864         (output_move_simode_const): Update calls after above changes.
1865         Rework to use automatic variables and predicates like MEM_P.
1866         * config/m68k/m68k.md (pushexthisi_const): Update call to
1867         valid_mov3q_const.
1869 2007-01-19  Dirk Mueller  <dmueller@suse.de>
1871         * tree-ssa-alias.c (perform_var_substitution): Fix typo
1872         in dump_flags test.
1874 2007-01-19  Richard Guenther  <rguenther@suse.de>
1876         * builtins.c (expand_builtin_cexpi): Fall back to expanding
1877         via cexp in case sincos is not available.
1879 2007-01-19  Richard Guenther  <rguenther@suse.de>
1881         * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
1882         * defaults.h (TARGET_HAS_SINCOS): Default to off.
1883         * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
1884         * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
1885         * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
1886         * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
1887         * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
1888         * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
1890 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
1892         * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
1893         New insn patterns.
1894         (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
1895         (atan2xf3): Directly generate RTL pattern.
1896         (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
1897         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
1898         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1899         disable patterns for SSE math.
1900         (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
1901         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
1902         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1903         disable patterns for SSE math.
1905 2007-01-19  Alexandre Oliva  <aoliva@redhat.com>
1907         * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
1909 2007-01-18  Roger Sayle  <roger@eyesopen.com>
1911         * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
1912         VIEW_CONVERT_EXPR to the same type as it's operand.
1914 2007-01-18  David Edelsohn  <edelsohn@gnu.org>
1916         * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
1917         configured for long double 128.
1919 2007-01-18  Mike Stump  <mrs@apple.com>
1921         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
1922         error.
1924 2007-01-18  Michael Meissner  <michael.meissner@amd.com>
1926         * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
1927         type correct.
1929 2007-01-18  Jan Hubicka  <jh@suse.cz>
1931         * tree-ssa-operands.c (vop_free_bucket_size): Never return value
1932         greater than NUM_VOP_FREE_BUCKETS.
1934 2007-01-18  Daniel Berlin  <dberlin@dberlin.org>
1936         * tree-ssa-structalias.c: Update comments.
1937         (ptabitmap_obstack): Removed.
1938         (pta_obstack): New.
1939         (oldpta_obstack): Ditto.
1940         (stats): Add a few members.
1941         (struct variable_info): Remove node, complex, address_taken, and
1942         indirect_target members. Add oldsolution member.
1943         (new_var_info): Do not initialize removed members.
1944         (constraint_expr_type): Remove INCLUDES.
1945         (constraint_graph): Add size, implicit_preds, rep,
1946         indirect_cycles, eq_rep, label, direct_nodes, and complex members.
1947         (FIRST_REF_NODE): New macro.
1948         (LAST_REF_NODE): Ditto.
1949         (FIRST_ADDR_NODE): Ditto.
1950         (find): New function.
1951         (unite): Ditto.
1952         (dump_constraint): Do not handle INCLUDES.
1953         (insert_into_complex): Do not insert duplicate constraints.
1954         (condense_varmap_nodes): Renamed and rewritten into ...
1955         (merge_node_constraints): This. Also fix bug in handling of
1956         offseted copy constraints.
1957         (clear_edges_for_node): No longer need to deal with preds at all,
1958         or removing associated preds/succs.
1959         (merge_graph_nodes): Deal with indirect_cycles.
1960         Don't deal with predecessors.
1961         (add_implicit_graph_edge): New function.
1962         (add_pred_graph_edge): Ditto.
1963         (add_graph_edge): Don't deal with predecessors.
1964         (build_constraint_graph): Removed.
1965         (build_pred_graph): New function.
1966         (build_succ_graph): Ditto.
1967         (struct scc_info): Removed in_component. Added roots, dfs, and
1968         node_mapping. Remove visited_index, unification_queue.
1969         (scc_visit): Deal with union-find we do now.
1970         Deal with cycles with REF nodes.
1971         (collapse_nodes): Renamed and rewritten to ...
1972         (unify_nodes): This.
1973         (process_unification_queue): Removed.
1974         (topo_visit): Cleanup
1975         (do_da_constraint): Use find.
1976         (do_sd_constraint): Ditto.
1977         (do_ds_constraint): Ditto.
1978         (do_complex_constraint): Ditto.
1979         (init_scc_info): Update for removed and added members.
1980         (find_and_collapse_graph_cycles): Renamed and rewritten into ...
1981         (find_indirect_cycles): This.
1982         (equivalence_class): New variable.
1983         (label_visit): New function.
1984         (perform_variable_substitution): Rewritten.
1985         (free_var_substitution_info): New function.
1986         (find_equivalent_node): Ditto.
1987         (move_complex_constraints): Ditto.
1988         (eliminate_indirect_cycles): Ditto.
1989         (solve_graph): Only propagate changed bits.
1990         Use indirect cycle elimination.
1991         Use find.
1992         (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
1993         (tree_id_eq): Renamed to ...
1994         (tree_vi_eq): This. Update for member change
1995         (insert_id_for_tree): Renamed and rewritten to ...
1996         (insert_vi_for_tree): This.
1997         (lookup_id_for_tree): Renamed and rewritten to ...
1998         (lookup_vi_for_tree): This.
1999         (get_id_for_tree): Renamed and rewritten to ...
2000         (get_vi_for_tree): Ditto.
2001         (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
2002         (process_constraint): Don't handle INCLUDES.
2003         Remove special ADDRESSOF case.
2004         (find_func_aliases): Rewrite to use vi functions instead of id
2005         ones.
2006         (create_function_info_for): Ditto.
2007         (create_variable_info_for): Ditto.
2008         (intra_create_variable_infos): Ditto.
2009         (merge_smts_into): Ditto.
2010         (find_what_p_points_to): Ditto.
2011         (init_base_vars): Ditto.
2012         (init_alias_vars): Ditto.
2013         (remove_preds_and_fake_succs): New function.
2014         (dump_sa_points_to_info): Dump new stats.
2015         (dump_solution_for_var): Use find.
2016         (set_used_smts): Fix formatting.
2017         (compute_points_to_sets): Updated for new functions.
2018         (ipa_pta_execute): Ditto.
2020 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
2021             Richard Sandiford  <richard@codesourcery.com>
2023         * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
2024         parallels.
2025         * calls.c (expand_call): If the return value is a PARALLEL,
2026         extract its first member.
2027         * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
2028         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2029         use FUNCTION_EXTRA_EPILOGUE.
2030         (m68k_function_value): Return a PARALLEL if the return value
2031         is of a pointer type.
2032         * config/m68k/netbsd-elf.h (current_function_returns_pointer)
2033         (FUNCTION_EXTRA_EPILOGUE): Remove.
2034         * config/m68k/m68k.md (D0_REG): New constant.
2036 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
2038         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2039         output a NOP for empty epilogues.
2041 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2043         * config/m68k/m68k.c (m68k_use_return_insn): Update comments
2044         before function.  Extend register save check to include all
2045         registers, not just integer ones.
2047 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
2049         * config/m68k/m68k-protos.h (use_return_insn): Rename to...
2050         (m68k_use_return_insn): ...this.
2051         * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
2052         * config/m68k/m68k.c (use_return_insn): Rename to...
2053         (m68k_use_return_insn): ...this.
2054         * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
2055         of USE_RETURN_INSN.
2057 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
2059         * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
2060         nearest even, fix denormal rounding overflow.
2062 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2064         * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
2065         f%$smove and f%$move.
2066         (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
2067         (extendsfdf2_cf): Use fdmove instead of f%&move.
2068         (truncdfsf2_cf): Use fsmove instead of f%$smove.
2069         (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
2071 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2073         * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
2074         GPR<-GPR moves.
2076 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2078         * real.h (real_format): Add a canonical_nan_lsbs_set field.
2079         (coldfire_single_format): Declare.
2080         (coldfire_double_format): Likewise.
2081         * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
2082         of qnan_msb_set to determine the lower bits of a canonical
2083         NaN significand.
2084         (encode_ieee_double): Likewise.
2085         (encode_ieee_quad): Likewise.
2086         (ieee_single_format): Initialize canonical_nan_lsbs_set.
2087         (mips_single_format): Likewise.
2088         (ieee_double_format): Likewise.
2089         (mips_double_format): Likewise.
2090         (ieee_extended_motorola_format): Likewise.
2091         (ieee_extended_intel_96_format): Likewise.
2092         (ieee_extended_intel_128_format): Likewise.
2093         (ieee_extended_intel_96_round_53_format): Likewise.
2094         (ibm_extended_format): Likewise.
2095         (mips_extended_format): Likewise.
2096         (ieee_quad_format): Likewise.
2097         (mips_quad_format): Likewise.
2098         (vax_f_format): Likewise.
2099         (vax_d_format): Likewise.
2100         (vax_g_format): Likewise.
2101         (i370_single_format): Likewise.
2102         (i370_double_format): Likewise.
2103         (decimal_single_format): Likewise.
2104         (decimal_double_format): Likewise.
2105         (decimal_quad_format): Likewise.
2106         (c4x_single_format): Likewise.
2107         (c4x_extended_format): Likewise.
2108         (real_internal_format): Likewise.
2109         (coldfire_single_format): New real_format.
2110         (coldfire_double_format): Likewise.
2111         * config/pdp11/pdp11.c (pdp11_f_format): Initialize
2112         canonical_nan_lsbs_set.
2113         (pdp11_d_format): Likewise.
2114         * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
2115         if TARGET_COLDFIRE_CPU.
2117 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2119         * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
2120         (output_call, m68k_legitimize_call_address): Declare.
2121         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
2122         parenthesees.  Add support for a 'W' constraint.
2123         (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
2124         (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
2125         (m68k_symbolic_call, m68k_symbolic_jump): Declare.
2126         * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
2127         variables.
2128         (override_options): Initialize them.  Do not set flag_no_function_cse
2129         for TARGET_ID_SHARED_LIBRARY.
2130         (m68k_output_pic_call): Delete.
2131         (m68k_legitimize_call_address): New function.
2132         (print_operand): Remove the %o prefix.  Handle the %p prefix.
2133         (output_call): New function.
2134         (m68k_output_mi_thunk): Use m68k_symbolic_jump.  Always load the
2135         target address from the GOT if symbolic jumps are not allowed.
2136         * config/m68k/m68k.md (call, general_operand): Do not set
2137         SYMBOL_REF_FLAG.  Use m68k_legitimize_call_address instead.
2138         Merge separate flag_pic and !flag_pic define_insns into...
2139         (*call, *call_value): ...these new patterns.  Match the address
2140         rather than the containing MEM and require it to be a call_operand.
2141         Use output_call to generate the asm template.
2142         * config/m68k/predicates.md (const_call_operand): New predicate.
2143         (call_operand): Likewise.
2145 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
2147         * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
2148         (STACK_GROWS_DOWNWARD): Define to 1.
2149         (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
2150         comments, emphasizing that these values are only defaults.
2151         * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
2152         * config/m68k/m68k.c (const_method): Remove trailing whitespace.
2154 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2156         * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
2158 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
2159             Richard Sandiford  <richard@codesourcery.com>
2161         * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
2162         (FP_REGNO_P): New macros.
2163         (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
2164         (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
2165         (EH_RETURN_STACKADJ_RTX): Use A0_REG.
2166         * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
2168 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
2170         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2171         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
2172         to $tmake_file.
2173         * config/m68k/t-floatlib: New file, extracting common code from...
2174         * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
2175         * config/m68k/t-uclinux: Here.
2176         * config/m68k/fpgnulib.c: Do not compile extendeed precision
2177         routines on ColdFire targets.
2179 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
2181         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2182         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
2183         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
2184         (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
2185         explicitly set MOTOROLA to 1.
2186         * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
2188 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2189             Nathan Sidwell  <nathan@codesourcery.com>
2191         * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
2192         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
2193         * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
2194         * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
2195         * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
2196         * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
2197         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
2198         (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
2199         (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
2200         * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2201         * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
2202         (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
2203         (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
2204         (DBX_REGISTER_NUMBER): Delete.
2205         * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
2206         (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2207         * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
2209 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2211         * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
2212         * config/flat.h: New file.
2213         * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
2214         OBJECT_FORMAT_FLAT.
2215         * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
2216         indirect references for -msep-data or -mid-shared-library.
2217         Do not use PC-relative code addresses either.
2218         * config/m68k/m68k.c (override_options): Restrict -fPIC error
2219         to -mpcrel.
2220         * config/m68k/uclinux.h (STARTFILE_SPEC): Define.  Use Scrt1.o
2221         for shared libraries and crt1.o for executables.  Use crti.o and
2222         crtbegin.o.
2223         (ENDFILE_SPEC): Use crtend.o and crtn.o.
2224         (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
2225         Do not add -elf2flt or -shared-lib-id options here.
2226         (LINK_SPEC): Define.  Pass -elf2flt if no -elf2flt option is given.
2227         Pass -shared-lib-id if -mid-shared-library, taking the library
2228         identifier from -mshared-library-id if given, otherwise
2229         defaulting to 0.
2230         (EH_FRAME_IN_DATA_SECTION): Do not undefine.
2231         (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
2232         (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
2233         and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
2234         (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
2235         * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
2236         and crtend.o.
2237         * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
2238         jump sequence for ISA A and ISA A+.
2239         (PICJUMP): Likewise.
2241 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2243         * config.gcc (m68k-*-uclinux*): Base the port on the common
2244         and m68k GNU/Linux files rather than on the generic ELF ones.
2245         * config/m68k/uclinux.h (TARGET_VERSION): Override.
2246         (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
2248 2007-01-18  Julian Brown  <julian@codesourcery.com>
2249             Richard Sandiford  <richard@codesourcery.com>
2251         * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
2252         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2253         * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
2254         before redefining.
2255         * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
2256         80 unconditionally.
2257         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2259 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2261         * doc/install.texi: Document m68k-uclinuxoldabi.
2262         * config.gcc (m68k-*-uclinuxoldabi): New configuration.
2263         * config/m68k/uclinux-oldabi.h: New file, copied from
2264         config/m68k/uclinux.h.
2266 2007-01-18  Kaz Kojima  <kkojima@rr.iij4u.or.jp>
2268         * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
2269         * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
2270         (m32r_compute_frame_size): Use unsigned for regno.
2271         (m32r_reload_lr): Use gen_frame_mem.
2272         (pop): New.
2273         (m32r_output_function_epilogue): Don't output the function epilogue
2274         textually here.
2275         (m32r_expand_epilogue): New.
2276         (direct_return): Return false if the function has the interrupt
2277         attribute.
2278         (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
2279         * config/m32r/m32r.md (epilogue): New expander.
2280         (return_lr, return_rte): New insns.
2281         (return): Make it expander.
2282         (return_normal): New expander.
2284 2007-01-18  Josh Conner  <jconner@apple.com>
2286         PR target/30485
2287         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
2288         support for UNLE, UNLT, UNGE, and UNGT.
2290 2007-01-18  Jan Hubicka  <jh@suse.cz>
2292         * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
2293         * tree-tailcall (add_virtual_phis): Likewise.
2294         (optimize_tail_call): Return TODOs.
2295         (execute_tail_calls): Return TODOs.
2296         * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
2297         * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
2298         changed.
2299         * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
2300         Cleanup cfg using TODO.
2301         * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
2302         * passes.c (execute_function_todo): When cleanup did something, remove
2303         unused locals.
2304         * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
2305         (make_edges): Don't cleanup_cfg.
2307 2007-01-18  Uros Bizjak  <ubizjak@gmail.com>
2309         * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
2310         UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
2311         operate on the top of stack.
2312         [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
2313         register handling of instructions that output to the second
2314         stack slot.
2315         [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
2316         (move_for_stack_reg): Special-case check for dead destination
2317         stack slot for constant load of 1.0 inside UNSPEC_TAN.
2319         * config/i386/i386.md (UNSPEC_TAN): New constant.
2320         (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
2321         (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
2322         to correctly model move of constant 1.0 to top stack slot.
2323         (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
2324         (unnamed peephole2 pattern): Remove corresponding peephole2
2325         pattern that optimizes tan insn and loading of constant 1.0.
2326         (tanxf2): Use fptanxf4_i387.
2327         (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
2328         patterns using X87MODEF12 mode macro.  Use fptan_extend<mode>xf4_i387
2329         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
2330         disable patterns for SSE math.
2331         (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
2332         (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
2334 2007-01-18  Dirk Mueller  <dmueller@suse.de>
2335             Richard Guenther <rguenther@suse.de>
2337         PR diagnostic/8268
2338         * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
2339         * common.opt (Warray-bounds): Add new warning option.
2340         * c-opts.c (c_common_handle_option): Define -Warray-bounds
2341         if -Wall is given.
2342         * Makefile.in: make tree-vrp.o depend on toplev.h
2343         * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
2344         is enabled.
2345         (check_array_refs, check_array_bounds, check_array_ref): New.
2347 2007-01-18  Jan Hubicka  <jh@suse.cz>
2349         * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
2350         (execute_ssa_ccp): Return flags conditionally.
2351         * tree-ssa-propagate.c (substitue_and_fold): Return if something was
2352         changed.
2353         * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
2355 2007-01-18  Steven Bosscher  <steven@gcc.gnu.org>
2357         * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
2358         the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
2360         * Makefile.c (GTFILES): Add cfglayout.h.
2361         * gengtype.c (open_base_files): Likewise.
2362         * cfglayout.c (cfg_layout_function_footer,
2363         cfg_layout_function_header) Reindent to make gengtype happy.
2364         * cfglayout.h (cfg_layout_function_footer,
2365         cfg_layout_function_header): Add GTY(()) marker.
2367         * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
2369 2007-01-18  Ben Elliston  <bje@au.ibm.com>
2371         * genautomata.c (write_automata): Include xstrerror output in the
2372         error message if writing the DFA description file fails.
2374 2007-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2376         * config/mips/mips-protos.h (mips_output_external): Make it
2377         return void.
2378         * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
2379         * config/mips/mips.c (irix_output_external_libcall): Likewise.
2380         (extern_list): Likewise.
2381         (extern_head): Likewise.
2382         (TARGET_ASM_FILE_END): Likewise.
2383         (mips_file_end): Likewise.
2384         (mips_output_external): Rewritten.
2386 2007-01-18  Ben Elliston  <bje@au.ibm.com>
2388         * genpreds.c (write_insn_preds_c): Only write out the function
2389         body for regclass_for_constraint if we have register constraints.
2391 2007-01-17  Tom Tromey  <tromey@redhat.com>
2393         * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
2394         * doc/install.texi (Testing): Use sourceware.org.
2395         (Binaries): Likewise.
2396         (Specific): Likewise.
2397         * doc/contrib.texi (Contributors): Use sourceware.org.
2399 2007-01-17  Anatoly Sokolov <aesok@post.ru>
2401         * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
2402         (AVR_ENHANCED): Rename to ...
2403         (AVR_HAVE_MUL): ... new.
2404         (avr_enhanced_p): Rename to ...
2405         (avr_have_mul_p): ... new.
2406         (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
2407         'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
2408         * config/avr/avr.c (avr_enhanced_p): Rename to ...
2409         (avr_have_mul_p): ... new.
2410         (base_arch_s): Rename 'enhanced' to 'have_mul'.
2411         (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
2412         'avr_enhanced_p' and 'enhanced'.
2413         (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
2414         AVR_HAVE_MUL instead of AVR_ENHANCED.
2415         * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
2416         (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
2417         *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
2418         instead of AVR_ENHANCED.
2419         (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
2420         * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
2421         (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
2423 2007-01-17  Ian Lance Taylor  <iant@google.com>
2425         * vec.h (VEC_reserve_exact): Define.
2426         (vec_gc_p_reserve_exact): Declare.
2427         (vec_gc_o_reserve_exact): Declare.
2428         (vec_heap_p_reserve_exact): Declare.
2429         (vec_heap_o_reserve_exact): Declare.
2430         (VEC_OP (T,A,reserve_exact)): New static inline function, three
2431         versions.
2432         (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
2433         negative parameter.
2434         (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
2435         (VEC_OP (T,A,copy)) [all versions]: Likewise.
2436         (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
2437         * vec.c (calculate_allocation): Add exact parameter.  Change all
2438         callers.
2439         (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
2440         (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
2441         (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
2442         (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
2443         (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
2444         (vec_heap_p_reserve_exact): New function.
2445         (vec_heap_o_reserve_exact): New function.
2447 2007-01-17  Jan Hubicka  <jh@suse.cz>
2449         * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
2451 2007-01-17  Eric Christopher  <echristo@apple.com>
2453         * config.gcc: Support core2 processor.
2455 2007-01-16  Jan Hubicka  <jh@suse.cz>
2457         * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
2458         the names, just unlink the chain so we don't crash on dangling pointers
2459         to dead SSA names.
2461 2007-01-16  Jan Hubicka  <jh@suse.cz>
2463         * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
2464         * tree-pass.h: Reorder to make IPA passes appear toegher.
2465         (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
2466         * cgraphunit.c (cgraph_finalize_function): Do not compute inling
2467         parameters, do not call early inliner.
2468         * ipa-inline.c: Update comments.  Include tree-flow.h
2469         (cgraph_decide_inlining): Do not compute inlining parameters.
2470         (cgraph_decide_inlining_incrementally): Return TODOs; assume to
2471         be called with function context set up.
2472         (pass_ipa_inline): Remove unreachable functions before pass.
2473         (cgraph_early_inlining): Simplify assuming to be called from the
2474         PM as local pass.
2475         (pass_early_inline): New pass.
2476         (cgraph_gate_ipa_early_inlining): New gate.
2477         (pass_ipa_early_inline): Turn into simple wrapper.
2478         (compute_inline_parameters): New function.
2479         (gate_inline_passes): New gate.
2480         (pass_inline_parameters): New pass.
2481         (apply_inline): Move here from tree-optimize.c
2482         (pass_apply_inline): New pass.
2483         * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
2484         transforming.
2485         * tree-inline.c (optimize_inline_calls): Return TODOs rather than
2486         doing them by hand.
2487         (tree_function_versioning): Do not allocate dummy struct function.
2488         * tree-inline.h (optimize_inline_calls): Update prototype.
2489         * tree-optimize.c (execute_fixup_cfg): Export.
2490         (pass_fixup_cfg): Remove
2491         (tree_rest_of_compilation): Do not apply inlines.
2492         * tree-flow.h (execute_fixup_cfg): Declare.
2493         * Makefile.in (gt-passes.c): New.
2494         * passes.c: Include gt-passes.h
2495         (init_optimization_passes): New passes.
2496         (nnodes, order): New static vars.
2497         (do_per_function_toporder): New function.
2498         (execute_one_pass): Dump current pass here.
2499         (execute_ipa_pass_list): Don't dump current pass here.
2501 2007-01-16  Janis Johnson  <janis187@us.ibm.com>
2503         * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
2504         (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
2506 2007-01-16  David Edelsohn  <edelsohn@gnu.org>
2508         * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
2509         (strong_alias): Define.
2510         (__gcc_qmul): Provide non-FMA for soft-float.
2511         (__gcc_qdiv): Same.
2512         (__gcc_qneg): New.
2513         (__gcc_qeq): New.
2514         (__gcc_qle): New.
2515         (__gcc_qge): New.
2516         (__gcc_qunord): New.
2517         (__gcc_stoq): New.
2518         (__gcc_dtoq): New.
2519         (__gcc_qtos): New.
2520         (__gcc_qtod): New.
2521         (__gcc_qtoi): New.
2522         (__gcc_qtou): New.
2523         (__gcc_itoq): New.
2524         (__gcc_utoq): New.
2525         (fmsub): New.
2526         * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
2527         soft-float functions.
2528         * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
2529         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
2530         about long double soft float.
2532 2007-01-16  Dorit Nuzman  <dorit@il.ibm.com>
2533             Tehila Meyzels  <tehila@il.ibm.com>
2535         * tree-vectorizer.h (is_pattern_stmt_p): New.
2536         * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
2537         formatting (tabs instead of spaces). Cleanup and clarify setting
2538         of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
2539         * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
2540         (vectorizable_type_demotion): Check that types are integral.
2541         (vectorizable_type_promotion): Likewise.
2542         (vectorizable_store): Fix typo.  Eliminate new-line at end of
2543         comments.
2545 2007-01-16  Jan Hubicka  <jh@suse.cz>
2547         * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
2548         cgraph edges too.
2550 2007-01-15  Eric Christopher  <echristo@apple.com>
2552         * ifcvt.c: Include vec.h, vecprim.h.
2553         (check_cond_move_block): New argument regs.
2554         Reorganize. Add registers used to regs.
2555         (cond_move_process_if_block): Use regs set above as
2556         loop bounds.
2558 2007-01-15  Eric Christopher  <echristo@apple.com>
2560         * config/darwin.h: Update copyright.
2561         (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
2562         (LINK_COMMAND_SPEC): Add -u.
2563         (LINK_SPEC): Fix umbrella for above.
2565 2007-01-15  Joseph S. Myers  <joseph@codesourcery.com>
2567         * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
2568         glibc CVS.
2570 2007-01-15  Tom Tromey  <tromey@redhat.com>
2572         * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
2573         * doc/install.texi (Testing): Don't mention jacks.
2574         (Configuration): Document --enable-java-maintainer-mode.  Move
2575         --with-java-home to libgcj-specific section.  Document
2576         --with-ecj-jar.
2577         (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
2579 2007-01-15  Jan Hubicka  <jh@suse.cz>
2581         * tree-ssa-dce.c (DCE_TODOs): New.
2582         (propagate_necessity): Return if something changed.
2583         (eliminate_unnecessary_stmts): Likewise.
2584         (perform_tree_ssa_dce): Return TODO flags when needed.
2585         (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
2587 2007-01-15  Uros Bizjak  <ubizjak@gmail.com>
2589         * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
2590         (fyl2x_extend<mode>xf3_i387): New insn pattern.
2591         (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
2592         patterns using X87MODEF12 mode macro.  Extend operand 1
2593         to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
2594         (log10<mode>2): Ditto.
2595         (log2<mode>2): Ditto.
2596         (log1p<mode>2): Ditto.
2597         (logb<mode>2): Ditto.
2598         (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
2599         (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
2600         (*fxtractxf3_i387): Rename from *fxtractxf3.
2601         (fxtract_extend<mode>xf3_i387): New insn pattern.
2602         (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
2603         * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
2604         and gen_fyl2xxf3_i387().
2606 2007-01-14  Zdenek Dvorak <dvorakz@suse.cz>
2608         * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
2609         * cfgloopmanip.c (fix_loop_placement): Made static.  Use
2610         get_loop_exit_edges.  Changed return type to bool.
2611         * cfgloop.h (fix_loop_placement): Declaration removed.
2613 2007-01-14  Dorit Nuzman  <dorit@il.ibm.com>
2615         * param.h (MIN_VECT_LOOP_BOUND): New.
2616         * params.def (MIN_VECT_LOOP_BOUND): New.
2617         * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
2618         argument - minimum threshold for number of iterations.
2619         * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
2620         argument to declaration.
2621         * tree-vect-analyze.c (vect_analyze_operations): Check value of
2622         MIN_VECT_LOOP_BOUND.
2623         * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
2624         slpeel_tree_peel_loop_to_edge with additional argument.
2625         (vect_do_peeling_for_alignment): Likewise.
2626         * doc/invoke.texi (min-vect-loop-bound): Document new param option.
2628 2007-01-14  Uros Bizjak  <ubizjak@gmail.com>
2630         PR target/30413
2631         * config/i386/i386.c (print_operand) ['z']: Output 'b' for
2632         operands of size 1.
2634 2007-01-14  Jan Hubicka  <jh@suse.cz>
2636         * tree-dfa.c (remove_referenced_var): New function.
2637         * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
2638         prune referenced vars list too.
2639         * tree-flow.h (remove_referenced_var): Declare.
2641 2007-01-14  Jan Hubicka  <jh@suse.cz>
2643         * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
2644         separately.
2645         (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
2646         (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
2647         * except.h (verify_eh_throw_table_statements): Kill prototype.
2648         * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
2649         into call exrepssions.
2650         * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
2651         eh trhow verifier.
2652         * tree-cfg.c: Include pointer-set.h.
2653         (verify_node_sharing): Work on pointer set.
2654         (verify_eh_throw_stmt_node): New.
2655         (verify_stmts): Use pointers sets, verify throw_stmt.
2657 2007-01-13  Zdenek Dvorak <dvorakz@suse.cz>
2659         * ipa-reference.c (analyze_function): Consider also addresses taken
2660         in phi nodes.
2662 2007-01-12  Roger Sayle  <roger@eyesopen.com>
2664         * c-typeck.c (null_pointer_constant_p): Replace use of
2665         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2666         (build_c_cast): Likewise.
2668 2007-01-12  Roger Sayle  <roger@eyesopen.com>
2670         * tree.h (force_fit_type_double): Remove unused final argument.
2671         * c-common.c (constant_expression_warning): Replace use of
2672         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2673         (convert_and_check): Likewise.
2674         (shorten_compare): Update call to force_fit_type_double.
2675         (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
2676         * convert.c (convert_to_pointer): Update call to
2677         force_fit_type_double.
2678         * fold-const.c (force_fit_type_double): Remove overflowed_const
2679         argument.
2680         (int_const_binop, fold_convert_const_int_from_int,
2681         fold_convert_const_int_from_real, fold_div_compare,
2682         fold_sign_changed_comparison, fold_unary, fold_negate_const,
2683         fold_abs_const, fold_not_const): Remove the final argument from
2684         calls to force_fit_type_double.
2686 2007-01-12  Andrew Pinski  <andrew_pinski@playstation.sony.com>
2688         * configure.ac: Set insn to "nop" for spu-*-* also.
2689         * configure: Regenerate.
2691 2007-01-12  Olga Golovanevsky  <olga@il.ibm.com>
2693         * builtins.def : Add BUILT_IN_FREE.
2695 2007-01-12  Jan Hubicka  <jh@suse.cz>
2697         PR tree-optimization/30443
2698         * tree-inline.c (tree_function_versioning): Do not optimize when
2699         cloning for inlining.
2701 2007-01-12  Zdenek Dvorak <dvorakz@suse.cz>
2703         * doc/loop.texi: Document recording of loop exits.
2704         * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
2705         (update_single_exits_after_duplication,
2706         update_single_exit_for_duplicated_loop,
2707         update_single_exit_for_duplicated_loops): Removed.
2708         (duplicate_loop_to_header_edge): Do not call
2709         update_single_exits_after_duplication and
2710         update_single_exit_for_duplicated_loops.
2711         (loop_version): Do not update single_exit information.
2712         (fix_loop_structure): Use record_loop_exits instead of
2713         mark_single_exit_loops.
2714         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
2715         the lists of loop exits.
2716         * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
2717         split_edge, merge_blocks): Update the lists of loop exits.
2718         * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
2719         loop_optimizer_init.
2720         * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
2721         of mark_single_exit_loops.
2722         (loop_optimizer_finalize): Call release_recorded_exits.
2723         * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
2724         LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
2725         * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
2726         update single exit information.
2727         * lambda-code.c (perfect_nestify): Ditto.
2728         * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
2729         (mark_single_exit_loops): Removed.
2730         (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
2731         get_exit_descriptions, rescan_loop_exit, record_loop_exits,
2732         dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
2733         functions.
2734         (get_loop_exit_edges, single_exit): Use recorded exit lists.
2735         (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
2736         (verify_loop_structure): Verify consistency of the exit lists.
2737         (flow_loops_find): Use alloc_loop.  Initialize exits hash.
2738         (set_single_exit): Removed.
2739         * cfgloop.h (struct loop_exit): New function.
2740         (struct loop): single_exit_ field replaced by exits field.
2741         (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
2742         (struct loops): Added exits hash.
2743         (mark_single_exit_loops, set_single_exit): Declaration removed.
2744         (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
2746 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2748         * doc/invoke.texi: Avoid use of @headitem.
2750 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2752         * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
2754 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2756         * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
2757         list of targets.
2759 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
2760             Richard Sandiford  <richard@codesourcery.com>
2762         * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
2763         * config/m68k/m68k.opt: Resort options.
2764         (mbitfield, mrtd, mshort): Remove RejectNegative properties.
2766 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
2767             Richard Sandiford  <richard@codesourcery.com>
2769         * doc/invoke.texi: Document the macros that are defined by
2770         m68k's -mtune and -mhard-float options.
2771         * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
2772         (m68k_cpp_cpu_family): Likewise.
2773         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
2774         of __ucfv*__ macros.  Define __mcffpu__ if generating code for
2775         ColdFire FPUs.  Define __mcf_cpu_* and __mcf_family_* macros.
2776         * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
2777         (m68k_cpp_cpu_family): Likewise.
2779 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2781         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
2782         macros besides mc68000 as tuning macros.  Use a switch statement
2783         to set them and mcpu32.
2785 2007-01-12  Julian Brown  <julian@codesourcery.com>
2787         * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
2788         (TARGET_68040_ONLY): Rename to...
2789         (TARGET_68040): ...this.
2790         * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
2791         * config/m68k/m68k.md: Likewise.
2793 2007-01-12  Julian Brown  <julian@codesourcery.com>
2794             Nathan Sidwell  <nathan@codesourcery.com>
2795             Richard Sandiford  <richard@codesourcery.com>
2797         * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
2798         the -mcpu= argument associated with the --with-cpu setting.
2799         Define M68K_DEFAULT_TUNE to the default -mtune= option,
2800         if different from the one implied by the -mcpu setting.
2801         Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
2802         using mcpu=FOO as the default CPU option.  Set target_cpu_default2.
2803         * doc/invoke.texi: Mention ColdFire in the introduction to the
2804         m68k options.  Document the new -march, -mcpu, -mtune, -mdiv,
2805         -mno-div and -mhard-float options.  Make -m68881 a synonym for
2806         -mhard-float.  Document the previously-undocumented -m5206e,
2807         -m528x, -m5307 and -m5407 options.  Tweak the existing option
2808         documentation for consistency.
2809         * doc/install.texi: Mention new --with-cpu arguments.
2810         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
2811         default CPU if neither -mcpu nor -march are specified.
2812         (ASM_CPU_SPEC): Pass down -mcpu and -march options.
2813         (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
2814         TARGET_ISA*.  Set the legacy __mcf*__ cpu macros in the same way,
2815         using m68k_tune to decide between families that implement the
2816         same ISA.  Use m68k_tune to set __mcfv4e__.
2817         (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
2818         (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
2819         (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
2820         (FL_ISA_C, FL_ISA_MMU): New macros.
2821         (MASK_COLDFIRE): Delete.
2822         (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
2823         (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
2824         (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
2825         (TARGET_HARD_FLOAT): Do not define here.
2826         (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
2827         (TUNE_68000): New macro.
2828         (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
2829         (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
2830         (TUNE_CFV2): Redefine in terms of m68k_tune.
2831         (uarch_type, target_device, fpu_type): New enums.
2832         (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
2833         * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
2834         (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
2835         (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
2836         (FL_FOR_isa_c): New macros.
2837         (m68k_isa): New enum.
2838         (m68k_target_selection): New structure.
2839         (all_devices, all_isas, all_microarchs): New tables.
2840         (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
2841         (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
2842         (MASK_ALL_CPU_BITS): Delete.
2843         (m68k_find_selection): New function.
2844         (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
2845         Map the legacy target options to a combination of the new ones.
2846         (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
2847         m68k_cpu_flags.  Handle M68K_DEFAULT_TUNE.  Use m68k_cpu_flags
2848         to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
2849         settings.
2850         * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
2851         (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
2852         properties.
2853         (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
2854         (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
2855         * config/m68k/m68k-devices.def: New file.
2857 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2858             Nathan Sidwell  <nathan@codesourcery.com>
2860         * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
2861         (SUBTARGET_EXTRA_SPECS): New macros.
2862         * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
2863         use %(asm_cpu_spec) instead.
2864         * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
2865         * config/m68k/openbsd.h (ASM_SPEC): Likewise.
2866         * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
2867         (EXTRA_SPECS): Rename to...
2868         (SUBTARGET_EXTRA_SPECS): ...this.
2870 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
2871             Richard Sandiford  <richard@codesourcery.com>
2872             Julian Brown  <julian@codesourcery.com>
2874         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2875         (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
2876         to the configuration's default CPU.
2877         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
2878         Remove default masks.
2879         (m680[012]0-*-*): Set the default with_cpu to the first part of
2880         the target name.
2881         (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
2882         (m68k*-*-linux): Extend the --with-cpu handling to...
2883         (m680[012]0-*-*, m68k*-*-*): ...these configurations.  Allow m68000
2884         and m68010.  Don't set target_cpu_default2.
2885         * doc/install.texi: Document --with-cpu for m68k.
2886         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
2887         * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
2888         (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
2889         (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
2890         (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
2891         (ASM_SPEC): Remove use of %(asm_cpu_default).
2892         (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
2893         * config/m68k/linux.h (TARGET_DEFAULT): Delete.
2894         (CPP_SPEC): Merge definitions.  Do not handle __HAVE_68881__ here.
2895         * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
2896         __HAVE_FPU__ if TARGET_HARD_FLOAT.
2897         (TARGET_DEFAULT): Delete.
2898         (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
2899         cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
2900         (CPP_CPU_SPEC): Delete.
2901         (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
2902         the appropriate string.
2903         (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
2904         (CPP_SPEC): Define to NETBSD_CPP_SPEC.
2905         (ASM_SPEC): Don't use %(asm_default_spec).
2906         * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
2907         TARGET_DEFAULT and add MASK_68881.
2908         * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
2910 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2912         * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
2913         (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
2914         MASK_68010 alongside MASK_68020.
2915         * doc/invoke.texi: Document -m68010.
2916         * config/m68k/m68k.opt (m68010): New.
2917         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
2918         if TUNE_68010.
2919         (TUNE_68010): New macro.
2920         * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
2921         (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
2922         (M68K_CPU_m68332): Add MASK_68010.
2923         * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
2924         fallback definition.
2925         * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
2926         defines.
2927         * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
2928         (m68k_handle_option): Handle OPT_m68010.  Add MASK_68010
2929         to all entries that use MASK_68020.
2930         (output_move_simode_const, output_move_himode, output_move_qimode)
2931         (output_move_stricthi, output_move_strictqi): Use TARGET_68010
2932         instead of TARGET_68020 to select clr behavior.  Remove comment
2933         about there being no TARGET_68010.
2934         * config/m68k/m68k.md: Likewise throughout.
2936 2007-01-12  Julian Brown  <julian@codesourcery.com>
2938         * config/m68k/m68k.h (TARGET_ISAB): New macro.
2939         * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
2940         * config/m68k/m68k.md: Likewise.
2942 2007-01-12  Julian Brown  <julian@codesourcery.com>
2944         * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
2945         TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
2947 2007-01-12  Julian Brown  <julian@codesourcery.com>
2949         * config/m68k/m68k.h (TUNE_68040_60): New macro.
2950         * config/m68k/m68k.c (standard_68881_constant_p): Use it.
2951         * config/m68k/m68k.md: Likewise.
2953 2007-01-12  Julian Brown  <julian@codesourcery.com>
2954             Richard Sandiford  <richard@codesourcery.com>
2956         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
2957         instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
2958         TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
2959         TARGET_CPU32.
2960         (TARGET_CPU32): Rename to...
2961         (TUNE_CPU32): ...this.
2962         (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
2963         (TUNE_CFV2): New macros.
2964         * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
2965         remove conditions that are implied by TARGET_68020.
2966         * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
2967         instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
2968         (m68k_output_function_epilogue): Likewise.
2969         (m68k_rtx_costs): Likewise.  Use TUNE_68060 instead of TARGET_68060
2970         and TUNE_CFV2 instead of TARGET_5200.  Use TUNE_68000_10 instead of
2971         "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
2972         non-68000 timings.  Refactor multiplication and division costs.
2973         (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
2974         TUNE_CPU32 instead of TARGET_CPU32.
2975         (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
2976         and TUNE_68060 instead of TARGET_68060.
2977         * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
2978         TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
2979         TARGET_CPU32.
2980         (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
2981         && !TARGET_COLDFIRE" to choose between moveq and clr.
2982         Likewise in the unnamed movsf pattern.
2983         (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
2984         "!TARGET_68020 && !TARGET_COLDFIRE".  Likewise the unnamed
2985         ashiftrt pattern.
2987 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2989         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
2990         of tabbing before backslashes.
2992 2007-01-11  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
2994         * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
2995         * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
2996         (ASM_OUTPUT_INTERNAL_LABEL): Define.
2998 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
3000         * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
3001         find_interesting_uses_cond.
3002         (find_interesting_uses_cond): Use extract_cond_operands.
3003         (rewrite_use_compare): Use extract_cond_operands and
3004         force_gimple_operand_bsi.  Do not call update_stmt.
3005         (determine_use_iv_cost_condition): Use extract_cond_operands.
3006         Return cheaper of using original bound and changing the exit bound.
3008 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
3010         PR tree-optimization/29516
3011         * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
3012         most_expensive_mult_to_index, addr_to_parts,
3013         create_mem_ref, maybe_fold_tmr): Make the type of
3014         fields of TARGET_MEM_REF sizetype.
3015         (move_fixed_address_to_symbol, move_pointer_to_base):
3016         New functions.
3017         * tree.def (TARGET_MEM_REF): Add comment on types of
3018         the operands.
3020 2007-01-11  Joseph Myers  <joseph@codesourcery.com>
3022         * c-common.c (vector_types_convertible_p): Treat opaque types as
3023         always convertible if they have the same size, but not otherwise.
3025 2007-01-11  Steven Bosscher  <steven@gcc.gnu.org>
3027         * ifcvt.c (struct noce_if_info): Add comments to the fields.
3028         Remove the b_unconditional field.
3029         (noce_try_sign_mask): Do not look at b_unconditional.
3030         (noce_process_if_block): Do not use merge_if_blocks.  Update
3031         the CFG here.  Do not set b_unconditional.
3032         (cond_move_process_if_block): Likewise.
3033         (find_cond_trap): Likewise.
3034         (check_cond_move_block): Require simple jump insns at the end
3035         of the basic block.
3037 2007-01-11  Jan Hubicka  <jh@suse.cz>
3039         PR tree-optimization/1046
3040         * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
3041         when alias info is not ready.
3042         (pass_tail_recursion): Do not require aliasing.
3043         * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
3044         * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
3045         * tree-ssa-copy.c (pass_copy_prop): Likewise.
3046         * tree-ssa-forwprop.c (pass_forwprop): Likewise.
3047         * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
3048         * passes.c (init_optimization_passes): Execute rename_ssa_copies,
3049         ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
3050         before inlining.
3051         * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
3052         When aliasing is not build, mark statement as volatile.
3054 2007-01-11  Tom Tromey  <tromey@redhat.com>
3056         PR preprocessor/15185, PR preprocessor/20989:
3057         * doc/cppopts.texi <-MT>: Update description of algorithm for
3058         computing default target.
3059         <-M, -MD>: Reword "basename" text.
3061 2007-01-11  Roger Sayle  <roger@eyesopen.com>
3063         * builtins.c (expand_builtin_pow, expand_builtin_powi,
3064         fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
3065         fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
3066         fold_builtin_int_int_roundingfn, fold_builtin_bitop,
3067         fold_builtin_bswap, real_constp, fold_builtin_pow,
3068         fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
3069         do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
3070         uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3071         * convert.c (convert_to_pointer): Likewise.
3072         * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
3073         * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
3074         const_binop, fold_convert_const_int_from_int,
3075         fold_convert_const_int_from_real,
3076         fold_convert_const_real_from_real, sign_bit_p,
3077         optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
3078         fold_sign_changed_comparison, fold_unary, fold_comparison,
3079         fold_binary, multiple_of_p, tree_Expr_non_zero_p,
3080         fold_negate_const, fold_abs_const, fold_not_const): Likewise.
3081         * print-tree.c (print_node_brief, print_node): Likewise.
3082         * stor-layout.c (place_field, layout_type): Likewise.
3083         * tree-chrec.c (keep_cast): Likewise.
3084         * tree.c (build_vector, build_real, build_real_from_int_cst,
3085         build_complex): Likewise.
3087 2007-01-11  Roger Sayle  <roger@eyesopen.com>
3089         * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete.  For the time being,
3090         treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
3092 2007-01-11  Paolo Bonzini  <bonzini@gnu.org>
3094         * configure.ac (strict1_warn): Rename to strict_warn.
3095         (WERROR, --enable-werror, symlink hacks, stage1_cflags,
3096         cc_set_by_configure, quoted_cc_set_by_configure,
3097         stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
3098         all_boot_languages, all_stagestuff): Remove.
3099         (target_list): Remove bootstrap targets.
3100         * Makefile.in (quickstrap): Unconditionally make a synonym of all.
3101         (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
3102         REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
3103         SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
3104         VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
3105         STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
3106         stage1_copy, stage2_build, stage2_copy, stageprofile_build,
3107         stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
3108         stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
3109         bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
3110         bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
3111         unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
3112         restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
3113         cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
3114         *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
3115         stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
3116         stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
3117         risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
3118         (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
3119         (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
3120         (mostlyclean): Adjust.
3121         (clean, distclean): Don't mention bootstrap stuff.
3122         * configure: Regenerate.
3123         * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
3124         java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
3125         treelang/config-lang.in (stagestuff): Remove.
3126         * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
3127         stageprofile, stagefeedback, stagestuff): Remove mention.
3129 2007-01-11  Nick Clifton  <nickc@redhat.com>
3131         * config/mcore/predicates.md (mcore_general_movesrc_operand):
3132         Accept CONSTs.
3133         (mcore_general_movdst_operand): Do not accept CONST_INTs.
3134         (mcore_arith_K_S_operand): Run the test for the S constraint not
3135         the test for the M constraint.
3136         (mcore_addsub_operand): Do not accept integer values that are
3137         larger than 32 bits.
3138         * config/mcore/mcore.md: Remove unused constraints from split.
3139         (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
3140         (addsi3): Likewise.
3141         (allocate_stack): Likewise.
3142         * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
3143         operands to 32 bits.
3144         (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
3145         INTVAL.
3146         (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
3147         mcore_const_ok_for_inline, mcore_const_trick_uses_not,
3148         try_constant_tricks, mcore_num_ones, mcore_num_zeros,
3149         mcore_output_bclri, mcore_output_andn, output_inline_const,
3150         mcore_output_move, mcore_output_movedouble): Likewise.
3151         (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
3152         (output_inline_const): Likewise.
3153         (output_inline_const): Fix format strings used in sprintf
3154         statements.
3155         * config/mcore/mcore-protos.h: Update prototypes for changed
3156         functions in mcore.c.
3157         * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
3158         HOST_WIDE_INT and not int.
3159         (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
3160         CONST_OK_FOR_N): Likewise.
3161         (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
3162         (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
3163         an INTVAL.
3165 2007-01-10  Jan Hubicka  <jh@suse.cz>
3167         * tree-vrp.c (remove_range_assertions): Release defs.
3168         * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
3169         * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
3171 2007-01-10  Paul Brook  <paul@codesourcery.com>
3173         * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
3174         * config/arm/arm.md (mulsi3subsi): New insn.
3176 2007-01-10  Zdenek Dvorak <dvorakz@suse.cz>
3178         * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
3179         (tree_transform_and_unroll_loop): New.
3180         * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
3181         Declare.
3183 2007-01-10  Robert Kennedy <jimbob@google.com>
3185         * fold-const.c (fold_comparison): Fold comparisons like (x *
3186         1000 < 0) to (x < 0).
3188 2007-01-10  Ian Lance Taylor  <iant@google.com>
3190         * tree-pretty-print.c (dump_generic_node): Print parentheses when
3191         operands have the same priority.
3193 2007-01-10  Tom Tromey  <tromey@redhat.com>
3195         * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
3196         (fold_binary): Likewise.
3197         * langhooks.c (lhd_can_use_bit_fields_p): Removed.
3198         * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
3199         (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
3200         (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
3201         * langhooks.h (struct lang_hooks): Removed field
3202         'can_use_bit_fields_p'.
3204 2007-01-10  Ralf Corsépius <ralf.corsepius@rtems.org>
3206         * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
3208 2007-01-10  Razya Ladelsky  <razya@il.ibm.com>
3210         * function.c (get_last_funcdef_no): New function.
3211         * function.h (get_last_funcdef_no): Declare.
3212         * tree-inline.c (initialize_cfun): Add initialization.
3213         (tree_function_versioning): Cleanup.
3215 2007-01-10  Jan Hubicka  <jh@suse.cz>
3217         * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
3218         PHIs.
3220 2007-01-10  Sa Liu  <saliu@de.ibm.com>
3221             Ben Elliston  <bje@au.ibm.com>
3223         * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
3224         for SAVE_NONLOCAL and Pmode for any other save level.
3225         * spu-protos.h (spu_restore_stack_block): Declare.
3226         * spu.md (save_stack_block): Remove.
3227         (restore_stack_block): Call spu_restore_stack_block.
3228         * spu.c (spu_restore_stack_block): New function.
3229         (spu_expand_epilogue): Remove old comment.
3231 2007-01-09  Zdenek Dvorak <dvorakz@suse.cz>
3233         PR tree-optimization/30322
3234         * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
3235         (cand_value_at): Return the value as aff_tree.
3236         (may_eliminate_iv): Convert the bound from aff_tree to tree.
3237         * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
3238         aff_combination_mult): New functions.
3239         (aff_combination_add): Use aff_combination_add_cst.
3240         (aff_combination_convert): Allow conversions to a wider type.
3241         (tree_to_aff_combination): Handle BIT_NOT_EXPR.
3242         * tree-affine.h (aff_combination_mult): Declare.
3244 2007-01-09  Carlos O'Donell  <carlos@codesourcery.com>
3246         * doc/tm.texi: Update documentation to reflect reality of exec
3247         and start file search behaviours. Update copyright year.
3248         * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
3249         header file directories.
3251 2007-01-09  Uros Bizjak  <ubizjak@gmail.com>
3253         * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
3254         (*cosxf2): Rename to cosxf2_i387.
3255         (*sindf2, *sinsf2): Extend operand 1 to XFmode.  Macroize patterns
3256         using X87MODEF12 mode macro. Rename patterns to
3257         *sin_extend<mode>xf2_i387.  Use SSE_FLOAT_MODE_P to disable patterns
3258         for SSE math.
3259         (*cosdf2, *cossf2): Ditto.
3260         (sincosdf3, sincossf3): Ditto.  Rewrite corresponding splitters
3261         to match extended input operands.
3262         (sincos<mode>3): New expander.
3263         (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
3264         insn patterns and corresponding splitters.
3266 2007-01-09  Kaz Kojima  <kkojima@gcc.gnu.org>
3268         * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
3269         (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
3271 2007-01-09  Nicolas Pitre  <nico@cam.org>
3273         PR target/30173
3274         * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
3276 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
3278         * target.h (struct gcc_target): New field library_rtti_comdat.
3279         * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
3280         (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
3281         * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
3282         * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
3284 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
3286         * doc/invoke.texi (Optimize Options): Correct description of -O0.
3288 2007-01-08  Richard Guenther  <rguenther@suse.de>
3290         * tree.h (force_fit_type_double): Export.
3291         (force_fit_type): Remove.
3292         * fold-const.c (force_fit_type_double): New function.
3293         (force_fit_type): Remove.
3294         (int_const_binop): Use it.
3295         (fold_convert_const_int_from_int): Likewise.
3296         (fold_convert_const_int_from_real): Likewise.
3297         (fold_div_compare): Likewise.
3298         (fold_sign_changed_comparison): Likewise.
3299         (fold_unary): Likewise.
3300         (fold_negate_const): Likewise.
3301         (fold_abs_const): Likewise.
3302         (fold_not_const): Likewise.
3303         * c-common.c (shorten_compare): Use force_fit_type_double.
3304         * convert.c (convert_to_pointer): Likewise.
3306 2007-01-08  Richard Guenther  <rguenther@suse.de>
3308         * tree.h (build_int_cst_wide_type): Export.
3309         * tree.c (build_int_cst_wide_type): New function.
3310         (build_int_cst_wide): Fix comment.
3311         * builtins.c (fold_builtin_object_size): Use build_int_cst
3312         to build -1 or 0 of the correct type.  Use fit_double_type
3313         to check for overflow.
3314         * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
3315         to build the mask.
3316         (decode_field_reference): Likewise.
3317         (all_ones_mask_p): Likewise.
3318         (native_interpret_int): Use build_int_cst_wide_type.
3319         (fold_binary): Use build_int_cst_type to build an all-ones
3320         value.
3321         * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
3323 2007-01-08  Daniel Jacobowitz  <dan@codesourcery.com>
3325         * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
3327 2007-01-09  Ben Elliston  <bje@au.ibm.com>
3329         * genautomata.c (STATS_OPTION): New option.
3330         (stats_flag): New flag.
3331         (gen_automata_option): Handle it.
3332         (initiate_automaton_gen): Ditto.
3333         (write_automata): Output statistics only if stats_flag is
3334         set. Likewise, output time statistics only if time_flag is set.
3335         * doc/md.texi (Processor pipeline description): Document new flag.
3337 2007-01-08  Richard Guenther  <rguenther@suse.de>
3339         * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
3340         * tree.c (build_int_cst_type): Likewise.
3341         (size_in_bytes): Don't call force_fit_type on the result.
3342         (int_fits_type_p): Use fit_double_type.
3343         * fold-const.c (fit_double_type): New function.
3344         (force_fit_type): Use it.
3345         * tree.h (fit_double_type): Export.
3347 2007-01-08  Jan Hubicka  <jh@suse.cz>
3349         * tree-vectorizer.c (gate_increase_alignment): Fix return type.
3350         * ipa.c (function_and_variable_visibility): Fix return type.
3352 2007-01-08  Richard Guenther  <rguenther@suse.de>
3354         * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
3355         of offset to build the index.
3356         * tree-pretty-print.c (dump_generic_node): Don't build negated
3357         const just for printing.
3358         * c-pretty-print.c (pp_c_integer_constant): Likewise.
3359         * builtins.c (fold_builtin_int_roundingfn): Check if result
3360         fits the type by using force_fit_type and comparing the result.
3361         * predict.c (predict_loops): Use compare_tree_int for comparison.
3362         * tree.c (build_int_cst): Fall back to integer_type_node for
3363         NULL_TREE type.
3364         (build_int_cst_wide): Assert type is non-null.
3366 2007-01-08  Roberto Costa  <roberto.costa@st.com>
3368         * tree-vrp.c (extract_range_from_cond_expr): New.
3369         (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
3370         * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
3371         as expressions.
3372         (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
3373         * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
3374         tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
3375         used as expressions.
3376         * tree-object-size.c (cond_expr_object_size): New.
3377         (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
3379 2007-01-08  Jan Hubicka  <jh@suse.cz>
3381         * tree-ssa-forwprop.c (forward_propagate_into_cond,
3382         tree_ssa_forward_propagate_single_use_va): Release defs of propagated
3383         statement.
3385 2007-01-08  Richard Guenther  <rguenther@suse.de>
3387         PR tree-optimization/23603
3388         * tree-vrp.c (set_value_range_to_truthvalue): New function.
3389         (extract_range_from_binary): Fall back to truthvalue instead of
3390         varying for TRUTH_*_EXPR.
3391         (extract_range_from_comparison): Fall back to truthvalue instead of
3392         varying.
3393         (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
3394         if all visited PHI values were constant.
3396 2007-01-08  Jan Hubicka  <jh@suse.cz>
3398         * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
3399         (cgraph_analyze_function): break out from ...
3400         (cgraph_finalize_compilation_unit): ... here.
3401         (cgraph_expand_function): Remove forgoten commented out line.
3402         (cgraph_optimize): Analyze functions.
3404 2007-01-08  Jan Hubicka  <jh@suse.cz>
3406         * tree-pas.h (TODO_remove_function): New flag.
3407         (TODO_update*): Renumber.
3408         (pass_ipa_increase_alignment,
3409         pass_ipa_function_and_variable_visibility): New passes.
3410         * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
3411         (cgraph_function_and_variable_visibility): Move to ipa.c
3412         (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
3413         cgraph_increase_alignment.
3414         * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
3415         (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
3416         calling tree-inline.
3417         (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
3418         (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
3419         * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
3420         (gate_increase_alignment): New function.
3421         (pass_ipa_increase_alignment): New pass.
3422         * ipa.c: Inline tree-pass.h and timevar.h
3423         (function_and_variable_visibility): Move here from cgraphunit.c
3424         * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
3425         * passes.c (init_optimization_passes): Add the two new passes.
3426         (execute_todo): Handle cgraph_remove_functions.
3428 2007-01-08  Nick Clifton  <nickc@redhat.com>
3430         * config/frv/predicates.md (reg_or_0_operand): Accept
3431         CONST_DOUBLEs.
3433 2007-01-08  Ralf Corsépius <ralf.corsepius@rtems.org>
3435         * config/bfin/rtems.h, config/bfin/t-rtems: New.
3436         * config.gcc: Add bfin*-rtems*.
3438 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
3440         * c.opt: Add -flax-vector-conversions.
3441         * c-typeck.c (convert_for_assignment): Pass flag to
3442         vector_types_convertible_p to allow emission of note.
3443         (digest_init): Likewise.
3444         * c-opts.c: Handle -flax-vector-conversions.
3445         * c-common.c (flag_lax_vector_conversions): New.
3446         (vector_types_convertible_p): Unless -flax-vector conversions
3447         has been passed, disallow conversions between vectors with
3448         differing numbers of subparts and/or element types.  If such
3449         a conversion is disallowed, possibly emit a note on the first
3450         occasion only to inform the user of -flax-vector-conversions.
3451         The new last argument specifies this.
3452         * c-common.h (flag_lax_vector_conversions): New.
3453         (vector_types_convertible_p): Add extra argument.
3454         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
3455         char_type_node for V*QI type vectors.
3456         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
3457         Update to satisfy new typechecking rules.
3458         * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
3459         C and C++ variants.
3460         * doc/invoke.texi (C Dialect Options): Document
3461         -flax-vector-conversions.
3463 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
3465         PR tree-optimization/29877
3466         * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
3467         a register variable on the RHS to not be replaceable.
3469 2007-01-08  Chen Liqin  <liqin@sunnorth.com.cn>
3470         * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
3471         * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
3472         const_simm15, const_pow2, const_npow2): Added.
3473         * config/score/misc.md (insv, extv, extzv, movmemsi,
3474         move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
3475         * config/score/score.c (score_address_cost, score_select_cc_mode):
3476         Added.
3477         Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
3478         Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
3479         Update score_print_operand.
3480         * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
3481         Adjust register allocate order and update some macro define.
3482         * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
3483         mdx_block_move_straight, mdx_block_move_loop_head,
3484         mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
3485         mdx_block_move): Added.
3486         (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
3487         mdp_limm): Updated and fix some bug and typo.
3488         * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
3489         (movsf, movdf, doloop_end): Added.
3491 2007-01-08  Kazu Hirata  <kazu@codesourcery.com>
3493         * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
3494         config/arm/thumb2.md: Fix comment typos.
3495         * doc/extend.texi: Fix a typo.
3497 2007-01-07  Eric Christopher  <echristo@apple.com>
3499         * configure.ac: Check for __stack_chk_fail for darwin.
3500         * configure: Regenerate.
3502 2007-01-07  Richard Guenther  <rguenther@suse.de>
3504         * tree-vrp.c (extract_range_from_assert): CSE calls to
3505         compare_values where possible.
3506         (extract_range_from_unary_expr): Likewise.
3508 2007-01-07  Anatoly Sokolov <aesok@post.ru>
3510         * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
3511         * config/avr/avr.c (call_insn_operand): Delete function.
3512         * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
3513         instead of register number. Use predicates.
3514         * config/avr/predicates.md (const0_operand, reg_or_0_operand,
3515         call_insn_operand): Add.
3517 2007-01-06  Jan Hubicka  <jh@suse.cz>
3519         * tree-pass.h (pass_build_cgraph_edges): Declare.
3520         * cgraphunit.c (record_refernece): Move to cgraphbuild.c
3521         (visited_nodes): Remove.
3522         (cgraph_create_edges): Move to cgraphbuild.c; rename to
3523         build_cgrpah_edges; make visited_nodes local.
3524         (cgraph_process_new_functions): DO not call initialize_inline_failed.
3525         (record_references_in_initializer): Move to cgraphbuild.c
3526         (initialize_inline_failed, rebuild_cgraph_edges,
3527         pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
3528         (verify_cgraph_node): Make visited_nodes local.
3529         (cgraph_analyze_function): Do not call cgraph_create_edges and
3530         initialize_inline_failed.
3531         (cgraph_expand_function): Do not call cgraph_lower_function;
3532         assert that function is already lowered.
3533         * Makefile.in (cgraphbuild.o): New.
3534         * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
3535         at the end of lowering passes.
3537 2007-01-06  Steven Bosscher  <steven@gcc.gnu.org>
3539         * ifcvt.c (cond_move_convert_if_block): New function, code
3540         factored out from...
3541         (cond_move_process_if_block): ...here.  Call the new function
3542         on the THEN and ELSE blocks.
3543         (merge_if_block): Do not copy global_live_at_end, merge_blocks
3544         already takes care of this.
3546 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
3548         PR c/19978
3549         * tree.h (TREE_OVERFLOW_P): New.
3550         * c-typeck.c (parser_build_unary_op): Warn only if result
3551         overflowed and operands did not.
3552         (parser_build_binary_op): Likewise.
3553         (convert_for_assignment): Remove redundant overflow_warning.
3554         * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
3556 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
3558         * c-typeck.c (store_init_value): Split over two lines to follow
3559         the GNU coding style.
3561 2007-01-05  Benjamin Kosnik  <bkoz@redhat.com>
3563         * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
3564         __GXX_EXPERIMENTAL_CXX0X__.
3565         * doc/cpp.texi: Same.
3567 2007-01-05  Richard Guenther  <rguenther@suse.de>
3569         PR middle-end/27826
3570         * tree.c (get_narrower): Do not construct COMPONENT_REFs
3571         with mismatched types.  Instead explicitly build a
3572         conversion NOP_EXPR.
3574 2007-01-05  Ian Lance Taylor  <iant@google.com>
3576         * c-common.c (decl_with_nonnull_addr_p): New function.
3577         (c_common_truthvalue_conversion): Call it.
3578         * c-typeck.c (build_binary_op): Likewise.
3579         * c-common.h (decl_with_nonnull_addr_p): Declare.
3581 2007-01-05  Jakub Jelinek  <jakub@redhat.com>
3583         PR c/30360
3584         * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
3585         denom against 0.0.
3587 2007-01-05  Joel Brobecker  <brobecker@adacore.com>
3589         * doc/install.texi (Final install): Document the fact that
3590         the GNAT runtime should not be stripped.
3592 2007-01-04  Jan Hubicka  <jh@suse.cz>
3594         * tree-inline.c (fold_marked_statements): Update operand caches
3595         and EH after folding
3597 2007-01-04  Ian Lance Taylor  <iant@google.com>
3599         * c-common.c (check_function_nonnull): Whitespace fix.
3601 2007-01-04  Jan Hubicka  <jh@suse.cz>
3603         * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
3604         comitted older version of patch.
3605         (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
3606         function, celanup cfg and collect garbage.
3608 2007-01-04  Mike Stump  <mrs@apple.com>
3610         * Makefile.in (mostlyclean): Don't remove libgcc anymore.
3611         (clean): Likewise.
3613 2007-01-04  Eric Christopher  <echristo@apple.com>
3615         * libgcc2.c (__bswapsi2): Use SItype.
3616         (__bswapdi2): Use DItype.
3617         * libgcc2.h: Update for above.
3619 2007-01-04  Paul Brook  <paul@codesourcery.com>
3621         * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
3622         mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
3623         mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
3624         umulsidi3, umulsidi3adddi, smulsi3_highpart,
3625         umulsi3_highpart): Make conditional on !arm_arch6.
3626         (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
3627         mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
3628         mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
3629         umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
3630         umulsi3_highpart_v6): New insns.
3632 2007-01-04  Roger Sayle  <roger@eyesopen.com>
3634         * fold-const.c (fold_convert): When casting an expression to void,
3635         fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
3636         have a type.  Instead of attempting to build a NOP_EXPR, return
3637         these "special" trees directly.
3639 2007-01-04  Joseph Myers  <joseph@codesourcery.com>
3641         * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
3642         MULT inside MINUS as either argument.  Use rs6000_cost->dmul -
3643         rs6000_cost->fp not 0 as adjustment for outer NEG.
3645 2007-01-04  Jan Hubicka  <jh@suse.cz>
3647         * cgraph.c (cgraph_release_function_body): New function.
3648         (cgraph_remove_node): Use it.
3649         * cgraph.h (cgraph_release_function_body): Declare.
3650         * cgraphunit.c (cgraph_expand_function): Use it.
3651         * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
3652         * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
3653         is initialized and while compilation of other function is running.
3654         * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
3655         statement CFG annotations from here to ...
3656         * tree-cfg.c (delete_tree_cfg_annotations): ... here.
3658 2007-01-04  Zdenek Dvorak <dvorakz@suse.cz>
3660         * cfgloop.h (enum li_flags): Make the constants powers of two.
3662 2007-01-04  Jan Hubicka  <jh@suse.cz>
3664         * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
3665         set.
3666         (fold_marked_statements): New function.
3667         (optimize_inline_calls, tree_function_versioning): Fold new statements.
3668         * tree-inline.h (copy_body_data): Add statements_to_fold.
3670 2007-01-03  Daniel Jacobowitz  <dan@codesourcery.com>
3672         * config.gcc: Mention libgcc/config.host.
3673         * Makefile.in: Update comments mentioning libgcc.
3674         (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
3675         (all.cross, start.encap, rest.encap, rest.cross): Update
3676         dependencies for libgcc move.
3677         (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
3678         (clean-target, clean-target-libgcc): Delete.
3679         (srcdirify, GCC_EXTRA_PARTS): New macros.
3680         (libgcc-support, libgcc.mvars): New rules.
3681         (distclean): Remove mention of mklibgcc.
3682         (install): Don't reference INSTALL_LIBGCC.
3683         (install-common): Don't reference EXTRA_PARTS.
3684         (install-libgcc, install-multilib): Delete rules.
3685         * mklibgcc.in: Delete file.
3686         * doc/configfiles.texi: Don't mention mklibgcc.
3688         * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
3689         * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
3690         * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
3691         * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
3692         (mklibgcc, ldblspecs): Likewise.
3694         * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
3695         (posixpre.def): Use $(T).
3696         (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
3697         (SHLIB_SRC, SHLIB_INSTALL): Delete.
3698         (SHLIB_LINK): Make dummy.
3699         * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
3701         * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
3703         * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
3704         * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
3705         config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
3706         config/rs6000/t-newas, config/sparc/t-elf: Likewise.
3708         * configure.ac (all_outputs): Remove mklibgcc.
3709         * configure: Regenerated.
3711 2007-01-03  Josh Conner  <jconner@apple.com>
3713         PR middle-end/29683
3714         * calls.c (compute_argument_addresses): Set stack and stack_slot
3715         for partial args, too.
3716         (store_one_arg): Use locate.size.constant for the size when
3717         generating a save_area.
3719 2007-01-03  Robert Kennedy <jimbob@google.com>
3721         * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
3722         whose definitions are deleted due to basic block merging.
3724 2007-01-03  Paul Brook  <paul@codesourcery.com>
3726         PR target/16634
3727         * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
3728         functions.
3729         (use_return_insn): Return 0 for Thumb interrupt functions.
3730         (print_multi_reg): Add rfe argument for IRQ returns.
3731         (arm_output_epilogue): Pop interrupt return address directly into PC.
3732         (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
3734 2007-01-03  Paul Brook  <paul@codesourcery.com>
3736         Merge from sourcerygxx-4_1.
3737         * config/arm/thumb2.md: New file.
3738         * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
3739         Thumb-2.
3740         * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
3741         * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
3742         (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
3743         * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
3744         tables.
3745         (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
3746         * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
3747         comptibility.
3748         * config/arm/ieee754-sf.S: Ditto.
3749         * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
3750         (thumb1_base_register_rtx_p): ... to this.
3751         (thumb_index_register_rtx_p): Rename...
3752         (thumb1_index_register_rtx_p): ... to this.
3753         (thumb_output_function_prologue): Rename...
3754         (thumb1_output_function_prologue): ... to this.
3755         (thumb_legitimate_address_p): Rename...
3756         (thumb1_legitimate_address_p): ... to this.
3757         (thumb_rtx_costs): Rename...
3758         (thumb1_rtx_costs): ... to this.
3759         (thumb_compute_save_reg_mask): Rename...
3760         (thumb1_compute_save_reg_mask): ... to this.
3761         (thumb_final_prescan_insn): Rename...
3762         (thumb1_final_prescan_insn): ... to this.
3763         (thumb_expand_epilogue): Rename...
3764         (thumb1_expand_epilogue): ... to this.
3765         (arm_unwind_emit_stm): Rename...
3766         (arm_unwind_emit_sequence): ... to this.
3767         (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
3768         thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
3769         thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
3770         arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
3771         thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
3772         thumb2_output_casesi): New functions.
3773         (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
3774         (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
3775         FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
3776         THUMB2_WORK_REGS): Define.
3777         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
3778         arm_condexec_mask, arm_condexec_masklen)): New variables.
3779         (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
3780         (arm_override_options): Check new CPU capabilities.
3781         Set new architecture flag variables.
3782         (arm_isr_value): Handle v7m interrupt functions.
3783         (user_return_insn): Return 0 for v7m interrupt functions.  Handle
3784         Thumb-2.
3785         (const_ok_for_arm): Handle Thumb-2 constants.
3786         (arm_gen_constant): Ditto.  Use movw when available.
3787         (arm_function_ok_for_sibcall): Return false for v7m interrupt
3788         functions.
3789         (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
3790         (thumb_find_work_register, arm_load_pic_register,
3791         legitimize_tls_address, arm_address_cost, load_multiple_sequence,
3792         emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
3793         print_multi_reg, output_mov_long_double_fpa_from_arm,
3794         output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
3795         output_mov_double_fpa_from_arm, output_move_double,
3796         arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
3797         output_return_instruction, arm_output_function_prologue,
3798         arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
3799         arm_output_mi_thunk, thumb_set_return_address): Ditto.
3800         (arm_expand_prologue): Handle Thumb-2.  Use arm_save_coproc_regs.
3801         (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
3802         (arithmetic_instr, shift_op): Use arm_shift_nmem.
3803         (arm_print_operand): Use arm_print_condition.  Handle '(', ')', '.',
3804         '!' and 'L'.
3805         (arm_final_prescan_insn): Use extract_constrain_insn_cached.
3806         (thumb_expand_prologue): Use thumb_set_frame_pointer.
3807         (arm_file_start): Output directive for unified syntax.
3808         (arm_unwind_emit_set): Handle stack alignment instruction.
3809         * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
3810         Add v6t2, v7, v7a, v7r and v7m.
3811         (RETLDM): Add Thumb-2 code.
3812         (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
3813         * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
3814         (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
3815         TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
3816         THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
3817         ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
3818         THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
3819         CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
3820         ADJUST_INSN_LENGTH): Define.
3821         (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
3822         STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
3823         BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
3824         PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
3825         SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
3826         TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
3827         HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
3828         HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
3829         REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
3830         REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
3831         LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
3832         GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
3833         ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
3834         FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
3835         PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
3836         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
3837         * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
3838         cortex-m3.
3839         * config/arm/arm-tune.md: Regenerate.
3840         * config/arm/arm-protos.h: Update prototypes.
3841         * config/arm/vfp.md: Enable patterns for Thumb-2.
3842         (arm_movsi_vfp): Add movw alternative.  Use output_move_vfp.
3843         (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
3844         (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
3845         thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
3846         * config/arm/libunwind.S: Add Thumb-2 code.
3847         * config/arm/constraints.md: Update include Thumb-2.
3848         * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
3849         * config/arm/ieee754-df.S: Ditto.
3850         * config/arm/bpabi.S: Ditto.
3851         * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
3852         * config/arm/predicates.md (low_register_operand,
3853         low_reg_or_int_operand, thumb_16bit_operator): New.
3854         (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
3855         (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
3856         * config/arm/t-arm-elf: Add armv7 multilib.
3857         * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
3858         Include thumb2.md.
3859         (UNSPEC_STACK_ALIGN, ce_count): New.
3860         (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
3861         arm_zero_extendsidi2, arm_zero_extendqidi2): New
3862         insns/expanders.
3863         * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
3864         (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
3865         thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
3866         * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
3867         (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
3868         thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
3869         insns.
3870         * doc/extend.texi: Document ARMv7-M interrupt functions.
3871         * doc/invoke.texi: Document Thumb-2 new cores+architectures.
3873 2007-01-03  Jakub Jelinek  <jakub@redhat.com>
3875         * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
3876         (struct _Unwind_Context): Rename args_size to flags, remove
3877         signal_frame field, add a new args_size field and version field.
3878         (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
3879         _Unwind_IsExtendedContext): New inline functions.
3880         (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
3881         Assume by_value array is only present if _Unwind_IsExtendedContext.
3882         (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
3883         _Unwind_IsSignalFrame.
3884         (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
3885         (uw_update_context_1): Use _Unwind_SetSignalFrame.
3886         (uw_init_context_1): Initialize context->flags to
3887         EXTENDED_CONTEXT_BIT.
3888         * config/rs6000/linux-unwind.h (frob_update_context): Use
3889         _Unwind_SetSignalFrame.
3891 2007-01-03  Andrew Pinski  <andrew_pinski@playstation.sony.com>
3893         PR middle-end/30353
3894         * gimplify.c (gimplify_modify_expr_complex_part): Move below
3895         tree_to_gimple_tuple.  Call tree_to_gimple_tuple when we need
3896         the value.
3898 2007-01-03  Kazu Hirata  <kazu@codesourcery.com>
3900         * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
3901         REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
3902         RET, AT_SP): Remove.
3903         * config/i386/i386.md (*sse_prologue_save_insn): Use return
3904         instead of RET.
3906         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3907         * cfgbuild.c (find_basic_blocks): Likewise.
3908         * cfgrtl.c (rtl_create_basic_block): Likewise.
3909         * function.c (temp_slots_at_level): Likewise.
3910         * reg-stack.c (stack_regs_mentioned): Likewise.
3911         * regclass.c (allocate_reg_info): Likewise.
3912         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3913         set_bb_for_stmt, move_block_to_fn): Likewise.
3914         * tree-complex.c (tree_lower_complex): Likewise.
3915         * vec.h (VEC_safe_grow_cleared): New.
3917         * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
3919 2007-01-03  Zdenek Dvorak <dvorakz@suse.cz>
3921         * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
3922         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
3923         to loop_version.
3924         * cfgloopmanip.c (scale_loop_frequencies): Export.
3925         (loopify): Scale the frequencies by prescribed coefficients.
3926         (set_zero_probability): New function.
3927         (duplicate_loop_to_header_edge): Improve updating of frequencies.
3928         (lv_adjust_loop_entry_edge, loop_version): Set probabilities
3929         and frequencies according to arguments.
3930         * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
3931         correctly.
3932         * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
3933         * modulo-sched.c (sms_schedule): Set probabilities for entering
3934         versioned loop correctly.
3935         * tree-vect-transform.c (vect_transform_loop): Ditto.
3936         * cfgloop.h (loopify, loop_version): Declaration changed.
3937         (scale_loop_frequencies): Declared.
3939 2007-01-02  Jan Hubicka  <jh@suse.cz>
3941         * cgraph.c: Include tree-flow.h
3942         (cgraph_add_new-function): Handle IPA_SSA mode; execute
3943         early_local_passes.
3944         * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
3945         * tree-pass.h (pass_all_early_optimizations): Declare.
3946         * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
3947         early_local_passes.
3948         (cgraph_analyze_function): Do early_local_passes.
3949         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3950         Do not add referenced vars.
3951         * tree-optimize.c (gate_all_optimizations): Do not execute when not in
3952         SSA form.
3953         (gate_all_early_local_passes): New gate.
3954         (pass_early_local_passes): Use new gate.
3955         (execute_early_local_optimizations): New functions.
3956         (gate_all_early_optimizations): New gate.
3957         (pass_all_early_optimizations): New pass.
3958         (execute_free_datastructures): Free SSA only when initialized.
3959         (gate_init_datastructures): Init only when optimizing.
3960         (tree_lowering_passes): Do early local passes when called late.
3961         * tree-profile.c (do_tree_profiling): Don't profile functions added
3962         late.
3963         (do_early_tree_profiling, pass_early_tree_profile): Kill.
3964         * tree-cfg.c (update_modified_stmts): Do not update when operands are
3965         not active.
3966         * passes.c (init_optimizations_passes): Reorder so we go into SSA
3967         during early_local_passes.
3968         * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
3971 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
3973         * Makefile.in: Update copyright year.
3975 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
3977         * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
3978         in $(lang_checks) and check-consistency targets.
3980 2007-01-02  Jan Hubicka  <jh@suse.cz>
3982         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3983         Do not add referenced vars.
3984         * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
3985         are not active.
3986         * passes.c (init_optimization_passes): Put mudflap_2 after
3987         free_datastructures.
3989 2007-01-02  Jan Hubicka  <jh@suse.cz>
3991         * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
3992         Set NOTHROW flag on call statements proved to be nothrow.
3993         Update statement of local calls so new pure/const functions are
3994         updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
3995         goto receivers.
3996         (tree_rest_of_compilation): Register hooks and initialize bitmap
3997         early. Do not set after_inlining flag.
3999 2007-01-02  Steve Ellcey  <sje@cup.hp.com>
4001         * sbitmap.c (HOST_BITS_PER_LONG_LONG):  Change to
4002         HOST_BITS_PER_LONGLONG
4004 2007-01-02  Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4006         PR c/19977
4007         * c-typeck.c (store_init_value): Don't emit pedantic overflow
4008         warning for non-static initializers.
4010 2007-01-02  Steven Bosscher  <steven@gcc.gnu.org>
4012         * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
4013         m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
4014         rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
4015         Always use set_unique_reg_note to add REG_EQUAL notes.
4017 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
4019         Revert:
4020         2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
4022         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4023         * cfgbuild.c (find_basic_blocks): Likewise.
4024         * cfgrtl.c (rtl_create_basic_block): Likewise.
4025         * function.c (temp_slots_at_level): Likewise.
4026         * reg-stack.c (stack_regs_mentioned): Likewise.
4027         * regclass.c (allocate_reg_info): Likewise.
4028         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4029         set_bb_for_stmt, move_block_to_fn): Likewise.
4030         * tree-complex.c (tree_lower_complex): Likewise.
4031         * vec.h (VEC_safe_grow_cleared): New.
4033 2007-01-02  Ian Lance Taylor  <iant@google.com>
4035         * c-common.c (c_common_truthvalue_conversion): When warning about
4036         using an assignment as a truth value, set TREE_NO_WARNING.
4038 2007-01-02  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
4040         PR middle-end/7651
4041         * c.opt (Wold-style-declaration): New.
4042         * doc/invoke.texi (C-only Warning Options): New.
4043         (Wold-style-declaration): Document it.
4044         (Wextra): Enabled by -Wextra.
4045         * c-opts.c (c_common_post_options): Enabled by -Wextra.
4046         * c-decl.c (declspecs_add_scspec): Replace -Wextra with
4047         -Wold-style-declaration.
4049 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
4051         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4052         * cfgbuild.c (find_basic_blocks): Likewise.
4053         * cfgrtl.c (rtl_create_basic_block): Likewise.
4054         * function.c (temp_slots_at_level): Likewise.
4055         * reg-stack.c (stack_regs_mentioned): Likewise.
4056         * regclass.c (allocate_reg_info): Likewise.
4057         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4058         set_bb_for_stmt, move_block_to_fn): Likewise.
4059         * tree-complex.c (tree_lower_complex): Likewise.
4060         * vec.h (VEC_safe_grow_cleared): New.
4062 2007-01-02  Douglas Gregor  <doug.gregor@gmail.com>
4064         * c-common.c (c_common_nodes_and_builtins): Since variants of
4065         void_type_node get built before it is given a name, we need to
4066         give those variants the name, too.
4067         (complete_array_type): We need to work with the canonical main
4068         type of the array, from which we will build the qualified version.
4069         * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
4070         * print-tree.c (print_node): Display canonical type information
4071         for each type.
4072         * stor-layout.c (layout_type): When we don't know the
4073         alignment of a type for which we're building an array, we end up
4074         guessing wrong, so make the type require structural equality.
4075         * tree.c (make_node_stat): When we build a new type, it is its
4076         own canonical type.
4077         (build_type_attribute_qual_variant): When building an attribute
4078         variant, its canonical type is the non-attribute variant. However,
4079         if the attributes are target-dependent and they differ, we need to
4080         use structural equality checks for this type.
4081         (build_qualified_type): A qualified type is not equivalent to its
4082         unqualified variant; set the canonical type appropriately.
4083         (build_distinct_type_copy): When building a distinct type from
4084         another type, the new type is its own canonical type.
4085         (build_variant_type_copy): When building a new type variant, we
4086         assume that it is equivalent to the original type.
4087         (build_pointer_type_for_mode): When building a pointer type, also
4088         build a canonical type pointer.
4089         (build_reference_type_for_mode): When building a reference type,
4090         also build a canonical type reference.
4091         (build_index_type): When we can't hash an index type (e.g.,
4092         because its maximum value is negative), the index type requires
4093         structural equality tests.
4094         (build_array_type): Build the canonical form of an array type.
4095         (build_function_type): Function types require structural equality,
4096         because they contain default arguments, attributes, etc.
4097         (build_method_type_directly): Ditto for method types.
4098         (build_offset_type): Build the canonical offset type.
4099         (build_complex_type): Build the canonical vector type.
4100         (make_vector_type): Build the canonical vector type.
4101         * tree.h (TYPE_CANONICAL): New.
4102         (TYPE_STRUCTURAL_EQUALITY_P): New.
4103         (SET_TYPE_STRUCTURAL_EQUALITY): New.
4104         (struct tree_type): Added "canonical" field.
4105         * params.h (VERIFY_CANONICAL_TYPES): New.
4106         * doc/c-tree.texi (TYPE_CANONICAL): Document.
4107         (TYPE_STRUCTURAL_EQUALITY_P): Document.
4108         (SET_TYPE_STRUCTURAL_EQUALITY): Document.
4109         * doc/invoke.texi (verify-canonical-types): Document --param
4110         parameter for verifying canonical types.
4112 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
4114         * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
4115         powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
4116         powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
4117         powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
4118         tm_file.
4119         * config/rs6000/e500.h: New.
4120         * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
4121         TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
4122         Remove.
4123         * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
4124         TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
4125         TARGET_E500_DOUBLE): Remove.
4126         * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
4127         TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
4128         * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
4129         * config/rs6000/rs6000.c (rs6000_override_options): Use
4130         CHECK_E500_OPTIONS.
4132 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
4134         * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
4135         TARGET_E500_DOUBLE), not TARGET_E500, for %y.
4136         (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
4137         rs6000_emit_cmove): Don't check TARGET_E500.
4138         * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
4139         sordered): Don't check TARGET_E500.
4141 2007-01-01  Eric Christopher  <echristo@apple.com>
4143         * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
4144         HARD_REGISTER_NUM_P.
4146 2007-01-01  Roger Sayle  <roger@eyesopen.com>
4148         * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
4149         "X eq/ne (C1^C2)".  Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
4150         has no side-effects.  Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
4151         eq/ne Y".
4153 2007-01-01  Mike Stump  <mrs@apple.com>
4155         * configure.ac: Remove support for building with Apple's gcc-3.1.
4157 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
4159         PR middle-end/30311
4160         * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
4161         REG is a hard register.  Do not modify REG before calling
4162         subreg_nregs.
4163         * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
4164         multiple of the size of YMODE for certain lowpart cases.
4166 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
4168         PR middle-end/30253
4169         * gimplify (voidify_wrapper_expr): Update for
4170         GIMPLE_MODIFY_STMT.
4172 2007-01-01  Andreas Schwab  <schwab@suse.de>
4174         PR target/29166
4175         * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
4176         save of BR0 in extra_spill_size instead of spill_size.
4177         (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
4178         area.
4179         (ia64_expand_epilogue): Restore BR0 from its new location.
4181 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
4183         * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
4184         Use a temporary variable if the left hand side is not a gimple
4185         register.
4187 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
4189         * gimplify.c (gimplify_return_expr): Make the temporary variable
4190         for the return expression, a gimple register variable.
4192 2007-01-01  Jan Hubicka  <jh@suse.cz>
4194         * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
4195         INSN_LIST.
4197 2007-01-01  Mike Stump  <mrs@apple.com>
4199         * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
4200         .literal16.
4201         * config/darwin.c (machopic_select_rtx_section): Use
4202         HAVE_GAS_LITERAL16.
4203         (darwin_mergeable_constant_section): Likewise.
4204         * configure: Regenerate.
4205         * config.in: Regenerate.
4207 2007-01-01  Jan Hubicka  <jh@suse.cz>
4208             Andrew Pinski  <pinskia@gmail.com>
4210         * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
4211         before starting IPA passes.