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