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>
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.
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
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
76 * config/i386/sse.md: Set type attribute to sseins for insertq
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.
117 * config/i386/i386.c (x86_sse_unaligned_move_optimal): Enable for
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.
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.
134 * config.gcc: Add support for --with-cpu option for amdfam10.
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>
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
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
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
187 * targhooks.h (default_builtin_vectorized_function): Likewise.
188 * target.h (builtin_vectorized_function): Add argument type
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>
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
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
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>
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>
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
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
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.
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>
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
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,
382 * varasm.c (notice_global_symbol): Treat global objects as weak when
385 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
387 * emit-rtl.c (dconstpi): Delete.
389 (init_emit_once): Delete dconstpi and init dconstsqrt2.
390 * real.h (dconstpi): Delete.
392 * builtins.c (fold_builtin_cabs): Use dconstsqrt2.
393 (fold_builtin_hypot): Likewise.
395 2007-02-03 Tom Tromey <tromey@redhat.com>
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
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.
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>
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
479 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
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>
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.
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
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
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.
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
555 * sched-deps.c (CHECK): New macro to (en/dis)able sanity checks.
556 (ds_to_dk, dk_to_ds): New functions.
558 (init_dep_1): New static function.
559 (init_dep): New function.
560 (copy_dep): New static function.
562 (dep_link_consistent_p, attach_dep_link, add_to_deps_list): New static
564 (detach_dep_link): New static function.
565 (move_dep_link): New function.
567 (dep_links_consistent_p, dump_dep_links): New static functions.
568 (debug_dep_links): New debugging function.
570 (deps_obstack, dl_obstack, dn_obstack): New static variables.
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
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.
589 (init_dependency_caches): Init deps_obstack.
590 (free_dependency_caches): Free deps_obstack.
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.
603 * sched-ebb.c (earliest_block_with_similiar_load): Update to use new
604 scheduler dependencies lists.
605 (schedule_ebb): Update comments.
607 * rtl.def (DEPS_LIST): Remove.
609 * lists.c (unused_deps_list): Remove.
610 (free_list): Update assertions.
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.
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.
634 (fix_inter_tick, try_ready, fix_tick_ready): Update to use new
635 scheduler dependencies lists.
637 (resolve_dep): Remove. Move the logic to schedule_insn ().
638 (init_h_i_d): Initialize dependencies lists.
640 (process_insn_depend_be_in_spec): Rename to
641 process_insn_forw_deps_be_in_spec. Update to use new scheduler
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.
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.
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
664 (TARGET_SCHED_ADJUST_COST_2): Remove documentation.
666 * doc/rtl.texi (LOG_LINKS): Remove part about instruction scheduler.
667 (REG_DEP_TRUE): Document.
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
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
704 2007-2-01 Seongbae Park <seongbae.park@gmail.com>
707 * global.c (global_alloc): Add mising initialization of
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):
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>
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
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
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
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
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
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>
807 * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
809 2007-01-31 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
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
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
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
850 * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
851 (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
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>
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>
889 * doc/invoke.texi (-Wreturn-type): Update description to
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.
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>
913 * dwarf2out.c (reference_to_unused): Query varpool if the
916 2007-01-30 Richard Guenther <rguenther@suse.de>
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
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
966 (mark_aliases_call_clobbered): Update for may_aliases being a
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
982 (compute_flow_insensitive_aliasing): Stop using pointer-sets.
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
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>
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
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
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
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
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.
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):
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
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
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
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
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>
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'
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>
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
1289 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
1290 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
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>
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
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>
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
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
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.
1372 * gimplify.c (omp_check_private): New function.
1373 (gimplify_scan_omp_clauses): Use it for
1374 firstprivate/lastprivate/reduction.
1377 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
1378 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
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
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
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
1464 (legitimate_lo_sum_address_p): Also check for TFmode for E500
1466 (rs6000_legitimize_address): Also handle TFmode for E500 double.
1467 (rs6000_legitimize_reload_address): Also handle TFmode for E500
1469 (rs6000_legitimate_address): Also check for TFmode for E500
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
1475 (function_arg): Handle TFmode and TCmode for E500 double.
1476 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
1478 (print_operand): Handle TFmode and TImode for %y.
1479 (rs6000_generate_compare): Handle TFmode comparisons for E500
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
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.
1518 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
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
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,
1533 2007-01-23 Ian Lance Taylor <iant@google.com>
1535 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
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
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.
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>
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
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
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
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
1666 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
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
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>
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
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
1717 (tree_values_to_profile): Call for determining indirect/virtual
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
1730 * profile.c (instrument_values): New case for indirect/virtual
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
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
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;
1761 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
1763 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
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
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'.
1853 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1855 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
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.
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
1869 2007-01-19 Dirk Mueller <dmueller@suse.de>
1871 * tree-ssa-alias.c (perform_var_substitution): Fix typo
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):
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
1924 2007-01-18 Michael Meissner <michael.meissner@amd.com>
1926 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
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.
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.
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.
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
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
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
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
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
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
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
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
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
2220 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
2221 for shared libraries and crt1.o for executables. Use crti.o and
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
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
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
2255 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
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.
2273 (m32r_output_function_epilogue): Don't output the function epilogue
2275 (m32r_expand_epilogue): New.
2276 (direct_return): Return false if the function has the interrupt
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>
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
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
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
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>
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
2342 * Makefile.in: make tree-vrp.o depend on toplev.h
2343 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
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
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
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
2432 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
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
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
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
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
2485 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
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.
2516 (__gcc_qunord): 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
2545 2007-01-16 Jan Hubicka <jh@suse.cz>
2547 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
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
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
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
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>
2631 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
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
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
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
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
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
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
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
2850 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
2851 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
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
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
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
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
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
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
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):
3017 * tree.def (TARGET_MEM_REF): Add comment on types of
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
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
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):
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.
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
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
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
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):
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
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
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>
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
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
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
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
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
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
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
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
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):
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>
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>
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>
3584 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
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.
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
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>
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>
3727 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
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
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
3745 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
3746 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
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
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
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 '(', ')', '.',
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
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.
3859 (UNSPEC_STACK_ALIGN, ce_count): New.
3860 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
3861 arm_zero_extendsidi2, arm_zero_extendqidi2): New
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
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>
3894 * gimplify.c (gimplify_modify_expr_complex_part): Move below
3895 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
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
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
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
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
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
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
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
3963 (do_early_tree_profiling, pass_early_tree_profile): Kill.
3964 * tree-cfg.c (update_modified_stmts): Do not update when operands are
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
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
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>
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>
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>
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
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
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
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):
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
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))
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>
4160 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
4161 REG is a hard register. Do not modify REG before calling
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>
4169 * gimplify (voidify_wrapper_expr): Update for
4172 2007-01-01 Andreas Schwab <schwab@suse.de>
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
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
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
4197 2007-01-01 Mike Stump <mrs@apple.com>
4199 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
4201 * config/darwin.c (machopic_select_rtx_section): Use
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.