1 2007-09-11 David Daney <ddaney@avtrex.com>
3 * doc/invoke.texi: Document new MIPS -mllsc and -mno-llsc options.
4 * doc/install.texi: Document new --with-llsc and --without-llsc
6 * config.gcc: Handle --with-llsc and --without-llsc configure options.
7 * config/mips/mips.md (sync, memory_barrier): Wrap sync instrunction
8 in %| and %- operand codes. Depend on GENERATE_SYNC instead of
10 (sync_compare_and_swap<mode>, sync_add<mode>, sync_sub<mode>,
11 sync_old_add<mode>, sync_old_sub<mode>, sync_new_add<mode>,
12 sync_new_sub<mode>, sync_<optab><mode>, sync_old_<optab><mode>,
13 sync_new_<optab><mode>, sync_nand<mode>, sync_old_nand<mode>,
14 sync_new_nand<mode>, sync_lock_test_and_set<mode>): Depend on
15 GENERATE_LL_SC instead of ISA_HAS_LL_SC.
16 * config/mips/mips.opt (mllsc): New option.
17 * config/mips/mips.c (mips_llsc): Define variable.
18 (mips_handle_option): Handle mllsc option.
19 (override_options): Set mips_print_operand_punct for '|' and '-'.
20 (print_operand): Add new %| and %- operand codes.
21 * config/mips/mips.h (mips_llsc_setting): New enum type.
23 (OPTION_DEFAULT_SPECS): Add llsc handling.
24 (GENERATE_SYNC): New macro.
25 (GENERATE_LL_SC): New macro.
26 (MIPS_COMPARE_AND_SWAP, MIPS_SYNC_OP, MIPS_SYNC_OLD_OP,
27 MIPS_SYNC_NEW_OP, MIPS_SYNC_NAND, MIPS_SYNC_OLD_NAND,
28 MIPS_SYNC_NEW_NAND, MIPS_SYNC_EXCHANGE): Wrap instructions
29 in %| and %- operand codes.
31 2007-09-11 Eric Botcazou <ebotcazou@adacore.com>
33 * tree-ssa-structalias.c (push_fields_onto_fieldstack): Deal with
34 TYPE_NONALIASED_COMPONENT like with DECL_NONADDRESSABLE_P.
36 2007-09-11 Jason Merrill <jason@redhat.com>
39 * stor-layout.c (layout_decl): Do pack variable size fields.
41 2007-09-11 Maxim Kuvyrkov <maxim@codesourcery.com>
43 * config/m68k/predicates.md (movsi_const0_operand,
44 non_symbolic_call_operand): New predicates.
46 * config/m68k/constraints.md: (Cs, Ci, C0, Cj, CQ, CW, CZ, CS, Ap, Ac):
48 * doc/md.texi (Constraints for Particular Machines: Motorola 680x0):
49 Document constraints N, O, P, R, S, T, Q, U, W, Cs, Ci, C0, Cj, CQ,
50 CW, CZ, CS, Ap and Ac.
52 * config/m68k/m68k.md (UNSPEC_IB): New constant.
53 (constraints.md): New include.
54 (cpu, type, type1, opx, opy, opx_type, opy_type, size, opx_access,
55 opx_mem, opy_mem, op_mem, guess, split): New attributes.
56 (movdf_internal): Name pattern. Fix to use alternatives. Add split.
59 (tstsi_internal): Name pattern. Fix to use alternatives. Specify
60 attributes. Split tstsi_internal_68020_cf from it.
61 (tstsi_internal_68020_cf): New pattern.
62 (tsthi_internal, tstqi_internal): Name pattern. Specify attributes.
63 (tst<mode>_cf): Specify attributea.
64 (cmpsi_cf): Name pattern. Specify attributes.
65 (cmp<mode>_68881, cmp<mode>_cf): Specify type attribute.
66 (pushexthisi_const): Fix to use alternatives. Specify
68 (movsi_const0): Split movsi_const0_68000_10 and movsi_const0_68040_60
69 from it. Fix to use alternatives. Specify attributes.
70 (movsi_const0_68040_10, movsi_const0_68040_60): New patterns.
71 (movsi_cf, movstrictqi_cf): Fix to use alternatives. Specify
73 (movsf_cf_soft): Specify attributes.
74 (movdf_cf_soft): Add split.
75 (pushasi, zero_extendhisi2_cf, zero_extendqisi2_cfv4,
76 cfv4_extendhisi2, 68k_extendhisi2, extendqihi2, cfv4_extendqisi2,
77 68k_extendqisi2, truncdfsf2_cf): Specify attributes.
78 (truncdfsf2_68881): Name pattern. Specify attributes.
79 (floatsi<mode>2_cf, floathi<mode>2_68881, floathi<mode>2_cf,
80 floatqi<mode>2_68881, floatqi<mode>2_cf, ftrunc<mode>2_cf,
81 fix<mode>qi2_cf, fix<mode>hi2_cf, fix<mode>si2_cf, adddi_dishl32):
83 (addsi3_5200): Fix to use alternatives. Specify attributes.
85 (add<mode>3_cf, subdi_dishl32): Specify attributes.
86 (subsi3): Add alternative for subq.l. Specify attributes.
87 (sub<mode>3_cf, mulhi3, mulhisi3): Specify attributes.
88 (mulhisisi3_s, mulsi3_68020, mulsi3_cf): Name pattern. Specify
90 (umulhisi3): Specify attributes.
91 (mulhisisi3_z): Name pattern. Specify attributes.
92 (fmul<mode>3_cf, div<mode>3_cf, negsi2_internal, negsi2_5200,
93 sqrt<mode>2_68881, clzsi2, one_cmplsi2_5200, subreghi1ashrdi_const32,
94 subregsi1ashrdi_const32, ashrsi3, subreg1lshrdi_const32, lshrsi3,
95 bsetmemqi): Specify attributes.
96 (bsetmemqi_ext): Name pattern. Specify attributes.
97 (bclrmemqi): Specify attributes.
98 (bclrmemqi_ext, scc, sls): Name pattern. Specify attributes.
99 (beq, bne, bgt, bgtu, blt, bltu, bge, bgeu, ble, bleu): Specify
101 (beq2, bne2, bgt2, bgtu2, blt2, bltu2, bge2, bgeu2, ble2, bleu2): Name
102 pattern. Specify attributes.
103 (jump): Specify attributes.
104 (tablejump_internal): Name pattern. Specify attributes.
105 (call_value): Split into non_symbolic_call_value,
106 symbolic_call_value_jsr, symbolic_call_value_bsr. Fix to use
107 alternatives. Specify attributes.
108 (non_symbolic_call_value, symbolic_call_value_jsr,
109 symbolic_call_value_bsr): New patterns.
110 (nop, return, unlink, indirect_jump): Specify attributes.
111 (trap): Fix condition. Specify attributes.
114 * config/m68k/m68k.c (m68k_symbolic_call_var): New variable.
115 (override_options): Initialize it. Initialize m68k_sched_cpu.
116 (CONST_METHOD): Rename to M68K_CONST_METHOD, move to m68k.h.
117 (const_method): Make global, rename to m68k_const_method.
118 (const_int_cost, output_move_const_into_data_reg): Update.
119 (output_move_double): Parametrize to emit rtl code, rename to
121 (output_reg_adjust, emit_reg_adjust, output_compadr, output_movsi,
122 emit_movsi): New static functions.
123 (output_move_double): New function with semantics of old
125 (m68k_emit_move_double): New function.
126 (m68k_sched_cpu): New variable.
127 (attr_op_type): New enum.
128 (sched_guess_p): New variable.
129 (sched_address_type, sched_operand_type, sched_attr_op_type):
130 New static functions.
131 (m68k_sched_attr_opx_type, m68k_sched_attr_opy_type,
132 m68k_sched_attr_size, m68k_sched_attr_op_mem): New functions.
133 (sched_branch_type): New static variable.
134 (m68k_sched_branch_type): New function.
135 * config/m68k/m68k.h (M68K_SYMBOLIC_CALL): New enum.
136 (m68k_symbolic_call_var): Declare.
137 (M68K_CONST_METHOD): Rename from CONST_METHOD. Move here from m68k.c.
138 (m68k_const_method, m68k_emit_move_double, m68k_sched_cpu,
139 m68k_sched_attr_opx_type, m68k_sched_attr_opy_type,
140 m68k_sched_attr_size, m68k_sched_attr_op_mem, m68k_sched_branch_type):
143 2007-09-11 Jakub Jelinek <jakub@redhat.com>
145 * builtins.def (BUILT_IN_VA_ARG_PACK_LEN): New builtin.
146 * builtins.c (expand_builtin) <case BUILT_IN_VA_ARG_PACK_LEN>: Issue
147 error if __builtin_va_arg_pack_len () wasn't optimized out during
149 * tree-inline.c (copy_bb): Replace __builtin_va_arg_pack_len ()
150 with the number of inline's anonymous arguments.
151 * doc/extend.texi: Document __builtin_va_arg_pack_len ().
153 2007-09-11 Zdenek Dvorak <ook@ucw.cz>
155 * fold-const.c (extract_muldiv_1): Do not simplify
158 2007-09-11 Jan Hubicka <jh@suse.cz>
160 * i386.h (ix86_tune_indices): Add X86_TUNE_INTER_UNIT_CONVERSIONS.
161 (TARGET_INTER_UNIT_CONVERSIONS): New.
162 * i386.md (floatsi expanders): Remove redundant check for SImode
163 source; offload to memory when asked for.
164 (floatsisf2_mixed, floatsisf2_sse, floatsidf2_mixed, floatsidf2_sse
165 floatdisf2_mixed, floatsisf2_sse, floatsidf2_mixed, floatsidf2_sse):
167 (floatsisf2_mixed_memory, floatsisf2_sse_memory,
168 floatsidf2_mixed_memory, floatsidf2_sse_memory
169 floatdisf2_mixed_memory, floatsisf2_sse_memory,
170 floatsidf2_mixed_memory, floatsidf2_sse_memory): New.
172 2007-09-11 Jan Hubicka <jh@suse.cz>
174 * toplev.c (process_options): all frontends now do unit-at-a-time.
175 * cgraphunit.c: update comments.
176 (cgraph_expand_function): call passmanager dirrectly; emit thunks.
177 * c-decl.c (finish_function): use cgraph_add_new_function.
178 * function.c (expand_function_end): We are always unit-at-a-time.
180 2007-09-11 Richard Sandiford <richard@codesourcery.com>
182 * config/mips/mips.c (mips_set_mips16_mode): Use separate anchor
184 (mips_use_anchors_for_symbol_p): Use default_use_anchors_for_symbol_p.
186 2007-09-11 Richard Sandiford <richard@codesourcery.com>
188 * config/mips/mips.c (mips_symbol_insns_1): Allow LEAs of
189 SYMBOL_FORCE_TO_MEM constants.
190 (mips_rtx_costs): Give a cost of 1 to force_to_mem_operands.
191 (mips16_rewrite_pool_refs_info): New structure.
192 (mips16_rewrite_pool_constant): New function, split out from...
193 (mips16_rewrite_pool_refs): ...here. Take a pointer to a
194 mips16_rewrite_pool_refs_info structure rather than a pointer
195 to a constant pool. Force force_to_mem_operands into memory.
196 (mips16_lay_out_constants): Update call to mips16_rewrite_pool_refs.
197 * config/mips/predicates.md (force_to_mem_operand): New predicate.
198 * config/mips/constraints.md (kf): New constraint.
199 * config/mips/mips.md (*movdi_64bit_mips16): Add a d <- kf alternative.
200 (*movsi_mips16): Likewise.
202 2007-09-11 Richard Sandiford <richard@codesourcery.com>
203 Nigel Stephens <nigel@mips.com>
204 David Ung <davidu@mips.com>
206 * config/mips/mips.h (CONSTANT_POOL_COST): Move to...
207 * config/mips/mips.c: ...here and set to 4 for TARGET_MIPS16.
208 (mips16_constant_cost, mips_immediate_operand_p, mips_binary_cost)
209 (mips_fp_mult_cost, mips_fp_div_cost, mips_sign_extend_cost)
210 (mips_zero_extend_cost): New functions.
211 (mips_rtx_costs): Treat COMPARE constants as having zero cost.
212 Use the new functions. Tweak many cost estimates, both here
213 and in the new subroutines. Return false when the cost of the
214 operands has not been calculated. Check for *clear_upper32.
215 Check for floating-point multiply-add, reciprocal and rsqrt
216 patterns. Handle comparison and rotation codes.
218 2007-09-11 Danny Smith <dannysmith@users.sourceforge.net>
220 * config/i386/cygming.h (TARGET_STRIP_NAME_ENCODING): Don't
222 * config/i386/i386.c (get_dllimport_decl): Don't strip
225 2007-09-10 Janis Johnson <janis187@us.ibm.com>
228 * config/dfp-bit.c: Don't skip TFmode conversions; move strto*
230 (DFP_TO_BFP): Use for either XFmode or TFmode.
231 (BFP_TO_DFP): Use for either XFmode or TFmode; always use cast
233 * config/dfp-bit.h: Include float.h.
234 (LONG_DOUBLE_HAS_XF_MODE, LONG_DOUBLE_HAS_TF_MODE): Define if long
235 double is one of these modes, rather than using LIBGCC_HAS_*F_MODE
236 which doesn't mean the same thing.
237 (BFP_KIND): Use 4 to mean TFmode.
238 (BFP_FMT): Specify the number of decimal digits based on the
239 number of mantissa digits.
240 (BFP_VIA_TYPE): Binary float type to use as cast for sprintf.
241 (BFP_TO_DFP, DFP_TO_BFP): Define names for TFmode variants.
242 (STR_TO_BFP): Use strtold for XFmode or TFmode.
243 (TFtype): Define if TFmode is supported.
244 * doc/libgcc.texi (Decimal float library routines): Document
245 TF conversion functions.
247 2007-09-10 Chao-ying Fu <fu@mips.com>
249 * config/mips/mips.c (mips_scalar_mode_supported_p): Declare.
250 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
251 (mips_emit_compare): Process fixed-point modes.
252 (mips_pad_arg_upward): Support fixed-point types.
253 (override_options): Allow fixed-point modes in accumulators.
254 (mips_pass_by_reference): Pass DQ, UDQ, DA, and UDA modes in registers.
255 (mips_vector_mode_supported_p): Support V2HQmode, V2UHQmode, V2HAmode,
256 V2UHAmode, V4QQmode, and V4UQQmode when TARGET_DSP.
257 (mips_scalar_mode_supported_p): New function to accept fixed-point
258 modes if the width is not greater than two BITS_PER_WORD.
259 * config/mips/mips.h (SHORT_FRACT_TYPE_SIZE, FRACT_TYPE_SIZE,
260 LONG_FRACT_TYPE_SIZE, LONG_LONG_FRACT_TYPE_SIZE,
261 SHORT_ACCUM_TYPE_SIZE, ACCUM_TYPE_SIZE, LONG_ACCUM_TYPE_SIZE,
262 LONG_LONG_ACCUM_TYPE_SIZE): Define.
263 * config/mips/mips.md ("d"): Update mode attribute for fixed-point
265 ("IMODE"): New mode attribute.
266 (mips-fixed.md): Include.
267 * config/mips/mips-modes.def: Create VECTOR_MODES for FRACT, UFRACT,
269 * config/mips/mips-fixed.md: New file.
271 2007-09-11 Ben Elliston <bje@au.ibm.com>
273 * config/spu/spu.md: Formatting fixes.
275 2007-09-10 Janis Johnson <janis187@us.ibm.com>
277 * config/dfp-bit.c (dfp_unary_func): Delete.
278 (dfp_unary_op): Delete.
279 (dfp_binary_op): Use decFloat functions instead of decNumber
280 functions for binary operations.
281 (d32_binary_op): Convert 32-bit operands to 64 bits for evaluation.
282 (dnn_binary_op): Call dfp_binary_op with decFloat rather than
284 (dfp_compare_op): Use decFloat functions insteadof decNumber
285 functions for comparisons.
286 (d32_compare_op): Convert 32-bit operands to 64 bits for evaluation.
287 (dnn_binary_op): Call dfp_compare_op with decFloat rather than
289 (DFP_ADD, DFP_SUB, DFP_MULTIPLE, DFP_DIVIDE): Use macros for
290 call to dxx_binary_op and decFloat function.
291 (DFP_EQ, DFP_NE, DFP_LT, DFP_GT, DFP_LE, DFP_GE): Use macros for
292 calls to dxx_binary_op and decFloat function.
293 * config/dfp-bit.h: Include decFloat header files.
294 (decFloat, DFP_BINARY_OP, DFP_COMPARE_OP, DEC_FLOAT_ADD,
295 DEC_FLOAT_SUBTRACT, DEC_FLOAT_MULTIPLY, DEC_FLOAT_DIVIDE,
296 DEC_FLOAT_COMPARE, DEC_FLOAT_IS_ZERO, DEC_FLOAT_IS_NAN,
297 DEC_FLOAT_IS_SIGNED: Define for each of 3 operand widths.
299 2007-09-10 Harsha Jagasia <harsha.jagasia@amd.com>
300 Jan Sjodin <jan.sjodin@amd.com>
302 * tree-vect-analyze.c (vect_analyze_operations): Change
303 comparison of loop iterations with threshold to less than
304 or equal to instead of less than. Reduce
305 min_scalar_loop_bound by one.
306 * tree-vect-transform.c (vect_estimate_min_profitable_iters):
307 Change prologue and epilogue iterations estimate to vf/2,
308 when unknown at compile-time. Change versioning guard
309 cost to taken_branch_cost. If peeling for alignment is
310 unknown at compile-time, change peel guard costs to one
311 taken branch and one not-taken branch per peeled loop.
312 If peeling for alignment is known but number of scalar loop
313 iterations is unknown at compile-time, change peel guard
314 costs to one taken branch per peeled loop. Change the cost
315 model equation to consider vector iterations as the loop
316 iterations less the prologue and epilogue iterations.
317 Change outside vector cost check to less than or equal to
318 zero instead of equal to zero.
319 (vect_do_peeling_for_loop_bound): Reduce
320 min_scalar_loop_bound by one.
321 * tree-vectorizer.h: Add TARG_COND_TAKEN_BRANCH_COST and
322 TARG_COND_NOT_TAKEN_BRANCH_COST.
323 * config/i386/i386.h (processor_costs): Add
324 scalar_stmt_cost, scalar_load_cost, scalar_store_cost,
325 vec_stmt_cost, vec_to_scalar_cost, scalar_to_vec_cost,
326 vec_align_load_cost, vect_unalign_load_cost,
327 vec_store_cost, cond_taken_branch_cost,
328 cond_not_taken_branch_cost.
329 Define macros for x86 costs.
330 * config/i386/i386.c:
331 (size_cost): Set scalar_stmt_cost, scalar_load_cost,
332 scalar_store_cost, vec_stmt_cost, vec_to_scalar_cost,
333 scalar_to_vec_cost, vec_align_load_cost,
334 vect_unalign_load_cost, vec_store_cost,
335 cond_taken_branch_cost, cond_not_taken_branch_cost to one.
336 (i386_cost, i486_cost, pentium_cost, pentiumpro_cost,
337 geode_cost, k6_cost, athlon_cost, pentium4_cost, nocona_cost,
338 core2_cost, generic64_cost, generic32_cost): Set to default
340 (k8_cost, amdfam10_cost): Costs for vectorization tuned.
341 (x86_builtin_vectorization_cost): New.
343 2007-09-10 Janis Johnson <janis187@us.ibm.com>
344 Ben Elliston <bje@au.ibm.com>
346 * dfp.c: Include decimal128Local.h;
347 (dfp_byte_swap): Remove.
348 (encode_decimal32, decode_decimal32): Don't handle endianness.
349 (encode_decimal64, decode_decimal64): Ditto.
350 (encode_decimal128, decode_decimal128): Ditto.
351 * config/dfp-bit.c (host_to_ieee32, ieee_to_host_32): Ditto.
353 (host_to_ieee_64, ieee_to_host_64): Don't handle endianness.
355 (host_to_ieee_128, ieee_to_host_128): Don't handle endianness.
356 * Makefile.in (DECNUM_H): Add decimal128Local.h.
358 2007-09-10 David Daney <ddaney@avtrex.com>
360 * config/mips/mips.md (UNSPEC_MEMORY_BARRIER): New entry in
362 (memory_barrier): Rewrote as an insn that clobbers memory.
364 2007-09-10 Richard Sandiford <richard@codesourcery.com>
366 * config/mips/mips.c (mips_global_pointer): Check
367 call_really_used_regs instead of call_used_regs.
368 (mips_save_reg_p): Likewise. Save all call-saved registers
369 if current_function_saves_all_registers. Fix indentation.
370 No longer treat $18 as a special case.
371 (compute_frame_size): Guard FPR loop with TARGET_HARD_FLOAT.
373 2007-09-10 Richard Sandiford <richard@codesourcery.com>
375 * config/mips/mips.h (MIPS_ARCH_FLOAT_SPEC): New macro.
376 * config/mips/mips.c (mips_cpu_info_table): Mention it in the
377 the introductory comment.
378 (MIPS_MARCH_CONTROLS_SOFT_FLOAT): Delete.
379 (override_options): Don't test for it.
380 * config/mips/sde.h (MIPS_MARCH_CONTROLS_SOFT_FLOAT): Delete.
381 (DRIVER_SELF_SPECS): Add MIPS_ARCH_FLOAT_SPEC.
382 * config/mips/vr.h: As for config/mips/sde.h.
384 2007-09-10 Trevor Smigiel <trevor_smigiel@playstation.sony.com>
385 Revital Eres <eres@il.ibm.com>
387 * target.h (struct gcc_target.sched): New field: sms_res_mii.
388 (struct ddg): Define.
389 * target-def.h (TARGET_SCHED_SMS_RES_MII): Define.
390 (TARGET_SCHED): Add TARGET_SCHED_SMS_RES_MII.
391 * config/spu/spu.c: Include ddg.h.
392 (TARGET_SCHED_SMS_RES_MII): Define.
393 (spu_sms_res_mii): New function to calculate mii.
394 * modulo-sched (res_MII): Use it.
395 * doc/tm.texi: Document TARGET_SCHED_SMS_RES_MII.
397 2007-09-10 Andreas Krebbel <krebbel1@de.ibm.com>
399 * config/s390/s390.c (s390_dump_pool): Create copy of constant
400 pool entries since they might hold values that must not be shared.
402 2007-09-10 Uros Bizjak <ubizjak@gmail.com>
405 * gcc/config/i386/sse.md (ashr<mode>3): Change op2 mode to SImode.
406 Use 'N' operand constraint for op2.
407 (lshr<mode>3): Ditto.
408 (ashl<mode>3): Ditto.
409 (vec_shl_<mode>): Use const_0_to_255_mul_8_operand predicate for op2.
410 (vec_shr_<mode>): Use const_0_to_255_mul_8_operand predicate for op2.
412 * gcc/config/i386/i386.c (ix86_expand_builtin) [IX86_BUILTIN_PSLL?128,
413 IX86_BUILTIN_PSRA*?128, IX86_BUILTIN_PSRL?128]: Convert op1 to SImode.
415 2007-09-10 Andreas Krebbel <krebbel1@de.ibm.com>
417 * config/s390/s390.md ("fixuns_trunc<BFP:mode><GPR:mode>2"):
418 Change mode macro in the last real_2expN parameter to uppercase.
420 2007-09-10 Michael Matz <matz@suse.de>
422 * tree-pass.h (pass_cselim): Declare new pass.
423 * passes.c (init_optimization_passes): Link in pass_cselim.
424 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Renamed from
425 tree_ssa_phiopt; add do_store_elim parameter, handle it by calling
426 cond_store_replacement.
427 (condstoretemp): New static variable.
428 (cond_store_replacement): New function.
429 (tree_ssa_phiopt, tree_ssa_cs_elim): New wrappers around
430 tree_ssa_phiopt_worker.
431 (struct name_to_bb): New.
432 (get_non_trapping, name_to_bb_hash, name_to_bb_eq, add_or_mark_expr,
433 nt_init_block, nt_fini_block): New static functions.
434 (seen_ssa_names, nontrap_set): New static variables.
435 (gate_cselim, pass_cselim): Define new pass.
436 * common.opt (ftree-cselim): New flag.
437 * toplev.c (process_options): Set flag_tree_cselim if required.
439 2007-09-10 Hans-Peter Nilsson <hp@axis.com>
441 * simplify-rtx.c (simplify_relational_operation_1): For recent
442 canonicalization, don't recurse if op1 equals both PLUS arguments.
444 2007-09-09 David Daney <ddaney@avtrex.com>
446 * optabs.c (expand_sync_operation): Use plus insn if minus
448 (expand_sync_fetch_operation): Ditto.
450 2007-09-09 H.J. Lu <hongjiu.lu@intel.com>
452 * i386.md (*floatsisf2_mixed_vector): Use cvtdq2ps instead
454 (*floatsisf2_sse_vector): Likewise.
456 2007-09-09 Krister Walfridsson <cato@df.lth.se>
458 * config/netbsd.h (HANDLE_PRAGMA_PACK_PUSH_POP): Define to 1.
460 2007-09-09 H.J. Lu <hongjiu.lu@intel.com>
462 * config/i386/i386.h (ix86_tune_indices): Rename
463 X86_USE_VECTOR_CONVERTS to X86_TUNE_USE_VECTOR_CONVERTS.
464 (TARGET_USE_VECTOR_CONVERTS): Updated.
465 * config/i386/i386.c: Likewise.
467 2007-09-09 Sandra Loosemore <sandra@codesourcery.com>
468 Nigel Stephens <nigel@mips.com>
470 * doc/invoke.texi (Overall Options): Add .sx file extension
472 * cppspec.c (known_suffixes): Likewise.
473 * gcc.c (default_compilers): Likewise.
475 2007-09-09 Rask Ingemann Lambertsen <rask@sygehus.dk>
478 * config/i386/i386.h (CANONICALIZE_COMPARISON): Delete.
479 * simplify-rtx.c (simplify_relational_operation_1): Add the
480 canonicalization from i386.h.
481 * doc/md.texi (Canonicalization of Instructions): Document it.
483 2007-09-09 Jan Hubicka <jh@suse.cz>
484 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com>
486 * i386.h (ix86_tune_indices): Add X86_USE_VECTOR_CONVERTS.
487 (TARGET_USE_VECTOR_CONVERTS): New.
488 * i386.md: New post-reload splitters for converting SF to DF and DF to
490 (floatsi* expander): Special case vector conversions.
491 (floatsisf2_mixed_vector, floatsisf2_sse_vector_nointernunit,
492 floatsisf2_sse_vector_internunit, floatsisf2_sse_vector,
493 floatsidf2_mixed_vector, floatsidf2_sse_vector): New.
494 (floatsisf2_mixed, floatsisf2_sse, floatsidf2_mixed, floatsidf2_sse):
495 Disable when doing vector converts.
496 (floatsi<mode>2_i387): Disable when
497 * sse.md (vec_dupv2df): Export.
498 * i386.c (ix86_tune_features): Enable SSE conversions.
500 2007-09-09 Richard Guenther <rguenther@suse.de>
502 * tree-ssa-operands.c (add_virtual_operand): Only mark
503 stores as has_volatile_ops if alias information is not available.
505 2007-09-09 Revital Eres <eres@il.ibm.com>
507 * doc/contrib.texi: Add myself.
509 2007-09-09 Ira Rosen <irar@il.ibm.com>
511 * tree-vectorizer.h (stmt_vinfo_set_inside_of_loop_cost,
512 stmt_vinfo_set_outside_of_loop_cost): New functions.
513 * tree-vect-transform.c (vect_get_cost_fields): Remove.
514 (vect_model_simple_cost): Call
515 stmt_vinfo_set_inside/outside_of_loop_cost to set the relevant cost
516 field instead of calling vect_get_cost_fields.
517 (vect_model_store_cost, vect_model_load_cost): Likewise.
519 2007-09-09 Revital Eres <eres@il.ibm.com>
521 * config/rs6000/rs6000.c (paired_init_builtins): Add const
522 declaration to bdesc_paired_preds variable.
523 (paired_expand_builtin): Likewise.
525 2007-09-09 Revital Eres <eres@il.ibm.com>
527 * dbgcnt.def (sms_sched_loop): New counter.
528 * modulo-sched.c: Use sms_sched_loop instead of
529 MAX_SMS_LOOP_NUMBER to determine the maximum number of loops to
530 perform swing modulo scheduling on. Include dbgcnt.h.
531 * Makefile.in: Add DBGCNT_H to modulo-sched.o.
532 * params.def: Remove PARAM_MAX_SMS_LOOP_NUMBER.
534 2007-09-09 Uros Bizjak <ubizjak@gmail.com>
536 * config/i386/i386.md (X87MODEF12, SSEMODEF): Remove mode iterators.
537 Substitute all uses with ...
538 (MODEF): New mode iterator.
540 (fix_trunc<mode>_fisttp_i387_1): Remove operand constraints
541 from pre-regalloc define_insn_and_split splitter pattern.
542 (*fix_trunc<mode>_i387_1): Ditto.
544 (*fist<mode>2_1): Ditto.
545 (frndintxf2_floor): Ditto.
546 (*fist<mode>2_floor_1): Ditto.
547 (frndintxf2_ceil): Ditto.
548 (*fist<mode>2_ceil_1): Ditto.
549 (frndintxf2_trunc): Ditto.
550 (frndintxf2_mask_pm): Ditto.
552 (prologue): Use (const_int 0) as never generated filler insn.
554 (sibcall_epilogue): Ditto.
555 (eh_return_si): Ditto.
556 (eh_return_di): Ditto.
558 (add<mode>3): Rename from adddf3 and addsf3. Macroize expander
559 using MODEF mode iterator.
560 (sub<mode>3): Rename from subdf3 and subsf3. Macroize expander
561 using MODEF mode iterator.
562 (mul<mode>3): Rename from muldf3 and mulsf3. Macroize expander
563 using MODEF mode iterator.
564 (nearbyint<mode>2): Rename from nearbyintdf2 and nearbyintsf2.
565 Macroize expander using MODEF mode iterator.
567 (zero_extendsidi2): Remove operand constraints from expander.
568 (smuldi3_highpart): Ditto.
569 (indirect_jump): Ditto.
572 * config/i386/sse.md (storentv4sf): Ditto.
573 (storentv2df): Ditto.
574 (storentv2di): Ditto.
576 (sse2_cvtpd2ps): Ditto.
577 (vec_interleave_highv16qi): Ditto.
578 (vec_interleave_lowv16qi): Ditto.
579 (vec_interleave_highv8hi): Ditto.
580 (vec_interleave_lowv8hi): Ditto.
581 (vec_interleave_highv4si): Ditto.
582 (vec_interleave_lowv4si): Ditto.
583 (vec_interleave_highv2di): Ditto.
584 (vec_interleave_lowv2di): Ditto.
585 (sse2_maskmovdqu): Ditto.
586 * config/i386/mmx.md (mmx_maskmovq): Ditto.
588 2007-09-09 Ira Rosen <irar@il.ibm.com>
590 * tree-vectorizer.h (enum vect_def_type): Start enumeration from 1.
591 (struct _slp_tree, struct _slp_instance): Define new data structures
592 along macros for their access.
593 (struct _loop_vec_info): Define new fields: strided_stores,
594 slp_instances, and slp_unrolling_factor along macros for their access.
595 (enum slp_vect_type): New.
596 (struct _stmt_vec_info): Define new field, slp_type, and macros for its
598 (STMT_VINFO_STRIDED_ACCESS): New macro.
599 (vect_free_slp_tree): Declare.
600 (vectorizable_load): Add an argument of type slp_tree.
601 (vectorizable_store, vectorizable_operation, vectorizable_conversion,
602 vectorizable_assignment): Likewise.
603 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
604 Declare (make extern).
605 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
606 (new_loop_vec_info): Likewise.
607 (destroy_loop_vec_info): Free memory allocated for SLP structures.
608 * tree-vect-analyze.c: Include recog.h.
609 (vect_update_slp_costs_according_to_vf): New.
610 (vect_analyze_operations): Add argument for calls to vectorizable_ ()
611 functions. For not pure SLP stmts with strided access check that the
612 group size is power of 2. Update the vectorization factor according to
613 SLP. Call vect_update_slp_costs_according_to_vf.
614 (vect_analyze_group_access): New.
615 (vect_analyze_data_ref_access): Call vect_analyze_group_access.
616 (vect_free_slp_tree): New functions.
617 (vect_get_and_check_slp_defs, vect_build_slp_tree, vect_print_slp_tree,
618 vect_mark_slp_stmts, vect_analyze_slp_instance, vect_analyze_slp,
619 vect_make_slp_decision, vect_detect_hybrid_slp_stmts,
620 vect_detect_hybrid_slp): Likewise.
621 (vect_analyze_loop): Call vect_analyze_slp, vect_make_slp_decision
622 and vect_detect_hybrid_slp.
623 * tree-vect-transform.c (vect_estimate_min_profitable_iters): Take
624 SLP costs into account.
625 (vect_get_cost_fields): New function.
626 (vect_model_simple_cost): Make extern, add SLP parameter and handle
628 (vect_model_store_cost, vect_model_load_cost): Likewise.
629 (vect_get_constant_vectors): New function.
630 (vect_get_slp_vect_defs, vect_get_slp_defs,
631 vect_get_vec_defs_for_stmt_copy, vect_get_vec_defs_for_stmt_copy,
632 vect_get_vec_defs): Likewise.
633 (vectorizable_reduction): Don't handle SLP for now.
634 (vectorizable_call): Don't handle SLP for now. Add argument to
635 vect_model_simple_cost.
636 (vectorizable_conversion): Handle SLP (call vect_get_vec_defs to
637 get SLPed and vectorized defs). Fix indentation and spacing.
638 (vectorizable_assignment): Handle SLP.
639 (vectorizable_induction): Don't handle SLP for now.
640 (vectorizable_operation): Likewise.
641 (vectorizable_type_demotion): Add argument to
642 vect_model_simple_cost.
643 (vectorizable_type_promotion): Likewise.
644 (vectorizable_store, vectorizable_load): Handle SLP.
645 (vectorizable_condition): Don't handle SLP for now.
646 (vect_transform_stmt): Add a new argument for SLP. Check that there is
647 no SLP transformation required for unsupported cases. Add SLP
648 argument for supported cases.
649 (vect_remove_stores): New function.
650 (vect_schedule_slp_instance, vect_schedule_slp): Likewise.
651 (vect_transform_loop): Schedule SLP instances.
652 * Makefile.in: (tree-vect-analyze.o): Depend on recog.h.
654 2007-09-09 Andrew Haley <aph@redhat.com>
656 * optabs.c (sign_expand_binop): Set libcall_gen = NULL in the
659 2007-09-09 Hans-Peter Nilsson <hp@axis.com>
661 Divide REG_LABEL notes into REG_LABEL_OPERAND and REG_LABEL_TARGET.
662 * doc/rtl.texi (Insns): Specify when a label_ref makes a jump_insn.
663 Similar for what label_refs can go in the JUMP_TARGET field. Split
664 REG_LABEL documentation into REG_LABEL_TARGET and REG_LABEL_OPERAND.
665 * reload.c (find_reloads): Generate REG_LABEL_OPERAND, not
666 REG_LABEL when replacing an operand with a LABEL_REF for a
668 (subst_reloads): When replacing a LABEL_REG with a register,
669 instead of generating a REG_LABEL note, assert that there already
670 is one or that the label is a known target for the insn.
671 * rtlanal.c (computed_jump_p): Instead of looking for a REG_LABEL
672 note, check the JUMP_LABEL field. Remove "else" after return.
673 * reorg.c (emit_delay_sequence): Replace case for REG_LABEL with
674 cases for REG_LABEL_OPERAND and REG_LABEL_TARGET.
675 (fill_slots_from_thread): Handle both REG_LABEL_OPERAND and
676 REG_LABEL_TARGET notes, including the JUMP_TARGET field on JUMP_P
677 insns. Iterate over all notes; don't assume there's only one.
678 * cse.c (recorded_label_ref): Adjust comment to refer to
680 (cse_extended_basic_block): Do LABEL_REF check for all INSN_P
681 insns, not just NONJUMP_INSN_P.
682 (check_for_label_ref): For JUMP_P insns, check that the LABEL_REF
684 * jump.c (rebuild_jump_labels): Adjust head comment.
685 (init_label_info): Ditto. Remove REG_LABEL_OPERAND notes only;
686 don't reset REG_LABEL_TARGET notes, including the JUMP_LABEL field.
687 (mark_all_labels): For JUMP_P insns without a target, check if the
688 the target is noted on the previous nonjump insn.
689 (mark_jump_label_1): New function, guts from mark_jump_label.
690 <case IF_THEN_ELSE>: Handle first operand as a non-target when
691 marking jump target labels.
692 <case LABEL_REF>: Adjust for whether to generate a
693 REG_LABEL_TARGET or a REG_LABEL_OPERAND note.
694 For 'E' format rtl, iterate in descending element order.
695 (delete_related_insns): Handle both REG_LABEL_TARGET and
696 REG_LABEL_OPERAND notes. For JUMP_P insns with labels with zero
697 reference count, delete and fallthrough. Move finding-next-
698 non-deleted insn last in the function. Look at all INSN_P insns
699 for REG_LABEL_OPERAND notes.
700 (redirect_jump_2): Assert that OLABEL equals the old JUMP_LABEL of
702 * print-rtl.c (print_rtx): For JUMP_P insns and a non-NULL
703 JUMP_LABEL, output the INSN_UID of it.
704 * gcse.c: Adjust comments as appropriate to say REG_LABEL_OPERAND
705 and/or REG_LABEL_TARGET.
706 (add_label_notes): Only add REG_LABEL_OPERAND notes. Put in line
707 with jump.c copy by only adding notes for labels actually
708 referenced in the insn.
709 * emit-rtl.c (try_split): Don't assume only NONJUMP_INSN_P need
710 usage count increment; handle all INSN_P trial insns.
711 (emit_copy_of_insn_after): Change to not copy REG_LABEL_OPERAND
713 * rtl.h (struct rtx_def) <volatil>: Adjust to mention
714 REG_LABEL_TARGET and REG_LABEL_OPERAND.
715 (LABEL_REF_NONLOCAL_P): Allow REG_LABEL_TARGET and
717 * combine.c (distribute_notes): Adjust for REG_LABEL_TARGET on
718 JUMP_P insns and REG_LABEL_OPERAND everywhere.
719 * sched-rgn.c (is_cfg_nonregular): Check for REG_LABEL_OPERANDS
721 * reg-notes.def (LABEL_TARGET, LABEL_OPERAND): Split from LABEL.
722 * cfgrtl.c (delete_insn): Adjust to handle REG_LABEL_TARGET and
723 REG_LABEL_OPERAND notes.
724 * reload1.c (calculate_needs_all_insns): Adjust comments.
725 (set_label_offsets): Adjust to look for REG_LABEL_OPERAND notes.
726 * config/alpha/alpha.md (split for load of an address into a
727 four-insn sequence on Unicos/Mk): Adjust to use
729 * config/sh/sh.md (sh_reorg, final_prescan_insn): Ditto.
731 2007-09-09 Laurynas Biveinis <laurynas.biveinis@gmail.com>
734 2007-09-05 Laurynas Biveinis <laurynas.biveinis@gmail.com>
735 * regrename.c (copyprop_hardreg_forward_1): New variable next. Use
736 FOR_BB_INSNS_SAFE instead of for loop.
737 * cse.c (cse_extended_basic_block): Likewise.
738 * postreload.c (reload_cse_regs_1): New variable next. Make sure
739 that the for loop does not invoke NEXT_INSN on a deleted insn.
740 * function.c (instantiate_virtual_regs): Likewise.
741 * lower-subreg.c (remove_retval_note): Likewise.
742 (decompose_multiword_subregs): Use FOR_BB_INSNS_SAFE instead of
744 * emit-rtl.c (remove_insn): Set NEXT_INSN and PREV_INSN to NULL on
746 * cfgrtl.c (delete_insn): Set JUMP_LABEL to NULL on a deleted
747 insn, if it's a jump.
748 (try_redirect_by_replacing_jump): New variable jump_p. Call
749 tablejump_p before delete_insn_chain.
750 * reload1.c (reload): New variable next. Make sure that the for
751 loop does not invoke NEXT_INSN on a deleted insn.
752 (fixup_eh_region_note): Make the loop terminate if i becomes NULL.
753 (delete_output_reload): New variable prev. Make sure the the for
754 loops do not invoke PREV_INSN on a deleted insn.
756 2007-09-08 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
758 * pa/constraints.md: Remove 'X' from unused letters comment.
760 2007-09-08 Richard Guenther <rguenther@suse.de>
762 * tree-tailcall.c (find_tail_calls): If we don't have aliases
763 computed check stmt_ann->references_memory instead of counting
766 2007-09-08 Segher Boessenkool <segher@kernel.crashing.org>
768 * cse.c (fold_rtx): Use validate_unshare_change() instead of
769 validate_change() in one more case.
771 2007-09-08 Zdenek Dvorak <ook@ucw.cz>
773 PR tree-optimization/32283
774 * tree-ssa-loop-ivopts.c (may_eliminate_iv): Use
775 estimated_loop_iterations.
776 (determine_use_iv_cost_condition): Decrease cost of expressions
777 used in iv elimination.
779 2007-09-08 Richard Guenther <rguenther@suse.de>
781 * tree-cfg.c (verify_gimple_expr): Avoid building new
782 pointer types, use TYPE_POINTER_TO if available instead.
784 2007-09-08 Uros Bizjak <ubizjak@gmail.com>
787 PR rtl-optimization/26449
788 * config/i386/sse.md (mulv4si3): Do not expand sse2 sequence.
789 (*sse2_mulv4si3): New define_insn_and_split pattern. Split insn in
791 (mulv16qi3): Implement as define_insn_and_split pattern instead of
792 define_expand. Split insn in split1 pass.
795 2007-09-08 Dorit Nuzman <dorit@il.ibm.com>
797 PR tree-optimization/33301
798 * tree-vect-analyze (analyze_operations): Look at the type of the rhs
801 2007-09-07 Zdenek Dvorak <ook@ucw.cz>
803 PR tree-optimization/32183
804 * Makefile.in (tree-ssa-reassoc.o): Also depend on $(CFGLOOP_H).
806 * tree-ssa-reassoc.c: Include cfgloop.h.
807 (is_reassociable_op): Add a loop argument and return true only
809 (linearize_expr): Updated.
810 (should_break_up_subtract): Likewise.
811 (linearize_expr_tree): Likewise.
812 (init_reassoc): Call loop_optimizer_init with
813 AVOID_CFG_MODIFICATIONS. Remove calculate_dominance_info call
815 (fini_reassoc): Call loop_optimizer_finalize.
817 2007-09-07 Sterling Augustine <sterling@tensilica.com>
819 * config/xtensa/lib2funcs.S (__xtensa_sync_caches): Use an ISYNC even
820 if there is no i-cache.
822 2007-09-07 Richard Guenther <rguenther@suse.de>
825 2007-09-06 Richard Guenther <rguenther@suse.de>
827 PR tree-optimization/32586
828 * tree-ssa-sccvn.c (simplify_binary_expression): Avoid
829 folding if nothing changed.
830 (simplify_unary_expression): New function. Do tree combining
831 on conversion like codes.
832 (try_to_simplify): Call it.
834 2007-09-07 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
837 * gthr-posix.h (__gthread_active_p): Add implementation for hppa-hpux.
838 (__gthread_active,__gthread_start, __gthread_active_init): New.
839 * gthr-posix95.h: Likewise.
841 2007-09-07 Roman Zippel <zippel@linux-m68k.org>
843 * function.h (struct function): Rename calls_unwind_init
844 to saves_all_registers.
845 (current_function_saves_all_registers): Renamed from
846 current_function_calls_unwind_init.
847 * reload1.c (reload): Set current_function_saves_all_registers.
848 * except.c (expand_builtin_unwind_init): Likewise.
849 * config/m68k/m68k.c (m68k_save_reg): Use
850 current_function_saves_all_registers to save pic register.
852 2007-09-07 Janis Johnson <janis187@us.ibm.com>
854 config/m68k/m68k.c (floating_exact_log2): Update call to real_2expN.
855 config/s390/s390.md (fixuns_trunc<BFP:mode><GPR:mode>2): Ditto.
857 Reapply reverted changes:
859 2007-09-06 Jan Hubicka <jh@suse.cz>
861 * config/i386.c (ix86_expand_lround, ix86_expand_round): Update call of
864 2007-09-06 Richard Sandiford <richard@codesourcery.com>
866 * config/mips/mips.md (fixuns_truncdfsi2, fixuns_truncdfdi2)
867 (fixuns_truncsfsi2, fixuns_truncsfdi2): Update calls to real_2expN.
869 2007-09-05 Janis Johnson <janis187@us.ibm.com>
871 * optabs.c (expand_float): Convert unsigned integer as signed only
872 if it provides sufficient accuracy; add mode argument to real_2expN.
873 (expand_fix): Fix comment typos; extend binary float into mode
874 wider than destination for converion to unsigned integer; add mode
875 argument to real_2expN.
876 * real.c (real_2expN): Add mode argument to special-case decimal
878 * real.h (real_2expN): Ditto.
879 * fixed-value.c (check_real_for_fixed_mode): Add mode argument to
881 (fixed_from_string): Ditto.
882 (fixed_to_decimal): Ditto.
883 (fixed_convert_from_real): Ditto.
884 (real_convert_from_fixed): Ditto.
885 * config/rs6000/rs6000.md (FP): Include DD and TD modes.
886 * config/rs6000/dfp.md (extendddtd2, adddd3, addtd3, subdd3, subtd3,
887 muldd3, multd3, divdd3, divtd3, cmpdd_internal1, cmptd_internal1,
888 floatditd2, ftruncdd2, fixdddi2, ftrunctd2, fixddi2): New.
890 2007-09-07 Diego Novillo <dnovillo@google.com>
892 * tree-flow.h (const_block_stmt_iterator): Remove.
893 Update all users to use block_stmt_iterator.
894 * tree-iterator.h (const_tree_stmt_iterator): Remove.
895 Update all users to use tree_stmt_iterator.
897 2007-09-07 Sandra Loosemore <sandra@codesourcery.com>
899 * config/mips/mips.c (mips_set_current_function): Temporarily
900 make this a no-op to fix bootstrap errors, pending rewrite.
902 2007-09-07 Jan Hubicka <jh@suse.cz>
904 * reorg.c (dbr_schedule): Move code removing placeholder USEs later
907 2007-09-07 Dorit Nuzman <dorit@il.ibm.com>
909 PR tree-optimization/33299
910 * tree-vect-transform.c (vect_create_epilog_for_reduction): Update
911 uses for all relevant loop-exit phis, not just the first.
913 2007-09-07 Richard Guenther <rguenther@suse.de>
916 * tree-ssa-operands.c (access_can_touch_variable): An access
917 of the form (*p)[0] can touch a variable of same size.
919 2007-09-07 Jan Hubicka <jh@suse.cz>
921 * passes.c (init_optimization_passes): Add simple dce and addressable
923 * tree-ssa.c (execute_update_addresses_taken): New function.
924 (pass_update_address_taken): New.
925 * tree-ssa-dse.c (execute_simple_dse): New function.
926 (pass_simple_dse): New.
927 * tree-pass.h (pass_simple_dse, pass_update_address_taken): Declare.
929 2007-09-07 Tobias Burnus <burnus@net-b.de>
932 * doc/invoke.texi: Fix -Wstrict-overflow= table.
934 2007-09-07 Richard Guenther <rguenther@suse.de>
937 2007-09-06 Richard Guenther <rguenther@suse.de>
939 PR tree-optimization/32586
940 * tree-ssa-sccvn.c (simplify_binary_expression): Avoid
941 folding if nothing changed.
942 (simplify_unary_expression): New function. Do tree combining
943 on conversion like codes.
944 (try_to_simplify): Call it.
946 2007-09-07 Richard Guenther <rguenther@suse.de>
947 Uros Bizjak <ubizjak@gmail.com>
949 PR tree-optimization/32821
950 * tree_if_conv.c (combine_blocks): Use alloc_stmt_list instead of
951 NULL in the call to set_bb_stmt_list.
953 2007-09-07 Richard Sandiford <richard@codesourcery.com>
955 * config/mips/mips.c (build_mips16_call_stub): Emit all direct
956 float calls here, rather than leaving some to the caller.
957 Use call_internal_direct and call_value_internal_direct.
958 * config/mips/mips.md (call_internal_direct): New pattern.
959 (call_value_internal_direct): Likewise.
961 2007-09-07 Richard Sandiford <richard@codesourcery.com>
963 * config/mips/mips.c (mips_base_move_loop_invariants): New variable.
964 (mips_set_mips16_mode): Restore flag_move_loop_invariants, then set
966 (override_options): Set mips_base_move_loop_invariants.
968 2007-09-07 Richard Sandiford <richard@codesourcery.com>
970 * config/mips/mips.opt (mcode-readable=): Move to keep list
973 2007-09-07 Richard Sandiford <richard@codesourcery.com>
975 * doc/invoke.texi: Document the MIPS -mlocal-sdata, -mextern-sdata
976 and -mgpopt options. Adjust the -G documentation to match.
977 * config/mips/mips.opt (mextern-data, mgpopt, mlocal-data): New.
978 * config/mips/mips.c (mips_rtx_constant_in_small_data_p): New
979 function, split out from mips_classify_symbol. Return false for
981 (mips_classify_symbol): Call mips_rtx_constant_in_small_data_p.
982 Only use GP-relative accesses for non-pool symbols if TARGET_GPOPT.
983 (override_options): Check whether the -mgpopt setting is consistent
984 with the other flags.
985 (symbolic_expression_p): Delete.
986 (mips_select_rtx_section): Use mips_rtx_constant_in_small_data_p
987 and default_elf_select_rtx_section.
988 (mips_in_small_data_p): Honor section attributes for MIPS16 too.
989 Return false for local data unless TARGET_LOCAL_SDATA. Likewise
990 external data and TARGET_EXTERN_SDATA.
992 2007-09-07 Tobias Burnus <burnus@net-b.de>
995 * doc/cpp.texi (Common Predefined Macros): Add __GFORTRAN__.
997 2007-09-05 Laurynas Biveinis <laurynas.biveinis@gmail.com>
999 * regrename.c (copyprop_hardreg_forward_1): New variable next. Use
1000 FOR_BB_INSNS_SAFE instead of for loop.
1001 * cse.c (cse_extended_basic_block): Likewise.
1002 * postreload.c (reload_cse_regs_1): New variable next. Make sure
1003 that the for loop does not invoke NEXT_INSN on a deleted insn.
1004 * function.c (instantiate_virtual_regs): Likewise.
1005 * lower-subreg.c (remove_retval_note): Likewise.
1006 (decompose_multiword_subregs): Use FOR_BB_INSNS_SAFE instead of
1008 * emit-rtl.c (remove_insn): Set NEXT_INSN and PREV_INSN to NULL on
1010 * cfgrtl.c (delete_insn): Set JUMP_LABEL to NULL on a deleted
1011 insn, if it's a jump.
1012 (try_redirect_by_replacing_jump): New variable jump_p. Call
1013 tablejump_p before delete_insn_chain.
1014 * reload1.c (reload): New variable next. Make sure that the for
1015 loop does not invoke NEXT_INSN on a deleted insn.
1016 (fixup_eh_region_note): Make the loop terminate if i becomes NULL.
1017 (delete_output_reload): New variable prev. Make sure the the for
1018 loops do not invoke PREV_INSN on a deleted insn.
1020 2007-09-06 Zdenek Dvorak <ook@ucw.cz>
1022 * cgraphbuild.c (rebuild_cgraph_edges): Export.
1023 * cgraph.h (rebuild_cgraph_edges): Declare.
1024 * tree-pass.h (pass_expand_omp_ssa): New.
1025 * omp-low.c (find_omp_clause): Export.
1026 (copy_var_decl): Split from omp_copy_decl_2.
1027 (build_omp_barrier): Return the call to emit instead of emitting
1029 (lower_rec_input_clauses, expand_omp_single): Gimplify the result of
1031 (extract_omp_for_data, expand_parallel_call, expand_omp_parallel,
1032 expand_omp_for_generic, expand_omp_for_static_nochunk,
1033 expand_omp_for_static_chunk, expand_omp_for, expand_omp_sections):
1034 Adapted to work on SSA form.
1035 (execute_expand_omp): Do not invalidate dominance information.
1036 (gate_expand_omp): Do not run with -fopenmp-ssa flag.
1037 (gate_expand_omp_ssa, pass_expand_omp_ssa): New.
1038 * gimplify.c (gimplify_omp_for): Ensure that the control variable is
1040 (force_gimple_operand): Allow gimplifying code expressions without
1042 * tree-predcom.c (mark_virtual_ops_for_renaming): Handle phi nodes.
1043 * common.opt (fopenmp-ssa): New.
1044 * tree-flow.h (find_omp_clause, copy_var_decl): Declare.
1045 * Makefile.in (tree-cfg.o): Add TREE_INLINE_H dependency.
1046 * tree-cfg.c: Include tree-inline.h.
1047 (struct move_stmt_d): Replace vars_to_remove by vars_map field.
1048 (replace_by_duplicate_decl, replace_ssa_name,
1049 mark_virtual_ops_in_region): New functions.
1050 (move_stmt_r, move_block_to_fn, move_sese_region_to_fn): Adapted
1051 to work on SSA form.
1052 * passes.c (init_optimization_passes): Add pass_expand_omp_ssa pass.
1053 * tree-ssa-operands.c (get_expr_operands): Handle operands of OMP
1056 2007-09-06 Laurynas Biveinis <laurynas.biveinis@gmail.com>
1058 * tree-loop-linear.c: Include obstack.h.
1059 (linear_transform_loops): New obstack lambda_obstack.
1060 Initialize it, pass it to gcc_loopnest_to_lambda_loopnest,
1061 lambda_loopnest_transform, lambda_loopnest_to_gcc_loopnest calls
1062 and free afterwards.
1063 * lambda.h (struct obstack): New forward declaration.
1064 (lambda_linear_expression_new): New parameter of type struct
1066 (lambda_loopnest_new): Likewise.
1067 (lambda_loopnest_transform): Likewise.
1068 (lambda_body_vector_new): Likewise.
1069 (lambda_body_vector_compute_new): Likewise.
1070 (gcc_loopnest_to_lambda_loopnest): Likewise.
1071 (lambda_loopnest_to_gcc_loopnest): Likewise.
1072 * lambda-code.c: Include obstack.h.
1073 (lambda_lattice_new): New parameter lambda_obstack. Use it for
1075 (lambda_body_vector_new): Likewise.
1076 (lambda_linear_expression_new): Likewise.
1077 (lambda_lattice_new): Likewise.
1078 (lambda_loopnest_new): Likewise. Additionally use obstack to
1079 allocate LN_LOOPS(ret).
1080 (lambda_lattice_compute_base): New parameter lambda_obstack. Pass
1081 it to lambda_lattice_new.
1082 (lambda_body_vector_compute_new): New parameter lambda_obstack.
1083 Pass it to lambda_body_vector_new.
1084 (lambda_lattice_compute_base): New paramater lambda_obstack. Pass
1085 it to lambda_lattice_new.
1086 (compute_nest_using_fourier_motzkin): New parameter lambda_obstack.
1087 Pass it to lambda_loopnest_new, lambda_linear_expression_new.
1088 (lambda_compute_target_space): Likewise.
1089 (lambda_compute_auxillary_space): New parameter lambda_obstack.
1090 Pass it to lambda_lattice_compute_base and
1091 compute_nest_using_fourieer_motzkin.
1092 (lambda_loopnest_transform): New parameter lambda_obstack. Pass
1093 it to lambda_lattice_compute_base, lambda_lattice_auxillary_space
1094 and lambda_lattice_compute_target_space.
1095 (gcc_tree_to_linear_expression): Nex parameter lambda_obstack.
1096 Pass it to lambda_linear_expression_new.
1097 (gcc_loop_to_lambda_loop): New parameter lambda_obstack. Pass it
1098 to gcc_tree_to_linear_expression.
1099 (gcc_loopnest_to_lambda_loopnest): New parameter lambda_obstack.
1100 Pass it to gcc_loop_to_lambda_loop and lambda_loopnest_new.
1101 (lambda_loopnest_to_gcc_loopnest): New parameter lambda_obstack.
1102 Pass it to lambda_body_vector_new and
1103 lambda_body_vector_compute_new.
1104 * Makefile.in (tree-loop-linear.o): Add $(OBSTACK_H) dependency.
1105 (lambda-code.o): Likewise.
1107 2007-09-06 Chao-ying Fu <fu@mips.com>
1109 * ginclude/stdfix.h: New file.
1110 * Makefile.in (USER_H): Add $(srcdir)/ginclude/stdfix.h.
1111 (convert.o): Add dependence on fixed-value.h.
1112 * c-convert.c (convert): Support FIXED_POINT_TYPE.
1113 * c-cppbuiltin.c (builtin_define_fixed_point_constants): New function
1114 to define fixed-point constants.
1115 (c_cpp_builtins): Define fixed-point constants.
1116 * convert.c (fixed-value.h): New include.
1117 (convert_to_real): Update comment to include fixed-point.
1118 Support FIXED_POINT_TYPE.
1119 (convert_to_integer): Update comment to include fixed-point.
1120 Support FIXED_POINT_TYPE.
1121 (convert_to_complex): Support FIXED_POINT_TYPE.
1122 (convert_to_fixed): New function.
1123 * convert.h (convert_to_fixed): Declare.
1124 * genopinit.c: Add comment about $Q for only fixed-point modes.
1125 (optabs): Add fract_optab, fractuns_optab, satfract_optab,
1126 satfractuns_optab, add_optab, ssadd_optab, usadd_optab, sub_optab,
1127 sssub_optab, ussub_optab, smul_optab, ssmul_optab, usmul_optab,
1128 ssmadd_widen_optab, usmadd_widen_optab, ssdiv_optab, udiv_optab,
1129 usdiv_optab, ssashl_optab, usashl_optab, neg_optab, ssneg_optab,
1130 usneg_optab for fixed-point modes.
1131 (gen_insn): Add force_fixed to track the $Q format for all fixed-point
1133 * optabs.c (optab_for_tree_code): For *DIV_EXPR, LSHIFT_EXPR,
1134 PLUS_EXPR, MINUS_EXPR, MULT_EXPR, NEGATE_EXPR, return signed or
1135 unsigned saturation optabs, when type is saturating.
1136 (shift_optab_p): Return true for SS_ASHIFT or US_ASHIFT.
1137 (expand_fixed_convert): New function.
1138 (gen_fixed_libfunc, gen_signed_fixed_libfunc,
1139 gen_unsigned_fixed_libfunc, gen_int_fp_fixed_libfunc,
1140 gen_int_fp_signed_fixed_libfunc, gen_int_fixed_libfunc,
1141 gen_int_signed_fixed_libfunc, gen_int_unsigned_fixed_libfunc,
1142 gen_fract_conv_libfunc, gen_fractuns_conv_libfunc,
1143 gen_satfract_conv_libfunc, gen_satfractuns_conv_libfunc): New
1145 (init_optabs): Initialize ssadd_optab, usadd_optab, sssub_optab,
1146 ussub_optab, ssmul_optab, usmul_optab, ssmadd_widen_optab,
1147 usmadd_widen_optab, ssmsub_widen_optab, usmsub_widen_optab,
1148 ssdiv_optab, usdiv_optab, ssashl_optab, usashl_optab, ssneg_optab,
1149 usneg_optab, fract_optab, fractuns_optab, satfract_optab,
1151 Initialize fixed-point libraries, including add, ssadd, usadd, sub,
1152 sssub, ussub, mul, ssmul, usmul, div, ssdiv, udiv, usdiv, ashl,
1153 ssashl, usashl, ashr, lshr, neg, ssneg, usneg, cmp, fract, satfract,
1154 fractuns, satfractuns.
1155 * optabs.h (enum optab_index): Add OTI_ssadd, OTI_usadd, OTI_sssub,
1156 OTI_ussub, OTI_ssmul, OTI_usmul, OTI_ssdiv, OTI_usdiv, OTI_ssneg,
1157 OTI_usneg, OTI_ssashl, OTI_usashl, OTI_ssmadd_widen, OTI_usmadd_widen,
1158 OTI_ssmsub_widen, OTI_usmsub_widen.
1159 (ssadd_optab, usadd_optab, sssub_optab, ussub_optab, ssmul_optab,
1160 usmul_optab, ssdiv_optab, usdiv_optab, ssneg_optab, usneg_optab,
1161 ssashl_optab, usashl_optab, ssmadd_widen_optab, usmadd_widen_optab,
1162 umsub_widen_optab, usmsub_widen_optab): Define.
1163 (enum convert_optab_index): Add COI_fract, COI_fractuns, COI_satfract,
1165 (fract_optab, fractuns_optab, satfract_optab, satfractuns_optab):
1167 (expand_fixed_convert): Declare.
1168 * expr.c (convert_move): Support the move of fixed-point modes.
1169 (emit_move_insn_1): Handle fixed-point mode to move via integer.
1170 (categorize_ctor_elements_1): Handle FIXED_CST.
1171 (count_type_elements): Handle FIXED_POINT_TYPE.
1172 (expand_expr_real_1): For VECTOR_CST, check MODE_VECTOR_FRACT,
1173 MODE_VECTOR_UFRACT, MODE_VECTOR_ACCUM, MODE_VECTOR_UACCUM.
1175 For PLUS_EXPR and MINUS_EXPR, support saturating and non-saturating
1176 multiply and add/subtract for fixed-point types.
1177 For MULT_EXPR, *DIV_EXPR, *SHIFT_EXPR, if the mode if a fixed-point
1178 mode, we jump to binop directly.
1179 Support FIXED_CONVERT_EXPR.
1180 (do_store_flag): Check FIXED_CST to put a constant second.
1181 (vector_mode_valid_p): Handle MODE_VECTOR_FRACT,
1182 MODE_VECTOR_UFRACT, MODE_VECTOR_ACCUM, MODE_VECTOR_UACCUM.
1183 (const_vector_from_tree): Support FIXED_CST.
1184 * doc/extend.texi (Fixed-Point): New node.
1185 * doc/md.texi (ssadd, usadd, sssub, ussub, ssmul, usmul, ssdiv, usdiv,
1186 ssmadd, usmadd, ssmsub, usmsub, ssashl, usashl, ssneg, usneg, fract,
1187 satfract, fractuns, satfractuns): Document them.
1189 2007-09-07 Bernd Schmidt <bernd.schmidt@analog.com>
1191 * config/bfin/bfin.h (PREFERRED_RELOAD_CLASS): Don't reload autoinc
1192 addresses into I registers.
1194 2007-09-06 Jan Hubicka <jh@suse.cz>
1195 Andreas Tobler <a.tobler@schweiz.org>
1197 * config/darwin.c (machopic_indirect_data_reference): Avoid invalid
1199 (machopic_legitimize_pic_address): Likewise.
1201 2007-09-06 Andrew Pinski <andrew_pinski@playstation.sony.com>
1202 Jan Hubicka <jh@suse.cz>
1204 * config/spu/spu.md (floatsidf2): Use convert_optab_libfunc
1205 instead of ufloat_optab->handlers directly.
1206 (floatdidf2): Likewise.
1208 2007-09-06 Sandra Loosemore <sandra@codesourcery.com>
1210 * config/mips/mips.c: Include diagnostic.h.
1211 (mips_set_current_function): Check errorcount and sorrycount
1212 before generating RTL.
1214 2007-09-06 Richard Sandiford <richard@codesourcery.com>
1217 * config/mips/mips.c (mips_classify_symbolic_expression): New function.
1218 (mips_classify_address): Use it instead of mips_symbolic_constant_p.
1219 (print_operand_reloc): Likewise.
1221 2007-09-06 Janis Johnson <janis187@us.ibm.com>
1225 2007-09-06 Jan Hubicka <jh@suse.cz>
1227 * i386.c (ix86_expand_lround, ix86_expand_round): Update call of
1230 2007-09-06 Richard Sandiford <richard@codesourcery.com>
1232 * config/mips/mips.md (fixuns_truncdfsi2, fixuns_truncdfdi2)
1233 (fixuns_truncsfsi2, fixuns_truncsfdi2): Update calls to real_2expN.
1235 2007-09-05 Janis Johnson <janis187@us.ibm.com>
1237 * optabs.c (expand_float): Convert unsigned integer as signed only
1238 if it provides sufficient accuracy; add mode argument to real_2expN.
1239 (expand_fix): Fix comment typos; extend binary float into mode
1240 wider than destination for converion to unsigned integer; add mode
1241 argument to real_2expN.
1242 * real.c (real_2expN): Add mode argument to special-case decimal
1244 * real.h (real_2expN): Ditto.
1245 * fixed-value.c (check_real_for_fixed_mode): Add mode argument to
1247 (fixed_from_string): Ditto.
1248 (fixed_to_decimal): Ditto.
1249 (fixed_convert_from_real): Ditto.
1250 (real_convert_from_fixed): Ditto.
1251 * config/rs6000/rs6000.md (FP): Include DD and TD modes.
1252 * config/rs6000/dfp.md (extendddtd2, adddd3, addtd3, subdd3, subtd3,
1253 muldd3, multd3, divdd3, divtd3, cmpdd_internal1, cmptd_internal1,
1254 floatditd2, ftruncdd2, fixdddi2, ftrunctd2, fixddi2): New.
1256 2007-09-06 Tom Tromey <tromey@redhat.com>
1258 * tree-cfg.c (remove_bb): Only warn if line is non-zero.
1259 * c-pch.c (c_common_read_pch): Restore current location after
1261 * tree.c (expand_location): Update.
1262 (expr_filename): Changed return type. Unified the two cases.
1263 (expr_lineno): Likewise.
1264 (annotate_with_file_line): Don't use EXPR_LINENO and EXPR_FILENAME
1266 * toplev.c (line_table): Changed type.
1267 (general_init): Update.
1268 (realloc_for_line_map): New function.
1269 (general_init): Allocate line_table using GC.
1270 * fix-header.c (line_table): Changed type.
1271 (read_scan_file): Update.
1272 (read_scan_file): Update.
1273 * c-ppoutput.c (maybe_print_line): Update.
1274 (print_line): Update.
1275 (cb_line_change): Update.
1276 (cb_define): Update.
1277 (pp_file_change): Update.
1278 * c-opts.c (c_common_init_options): Update.
1279 (finish_options): Update.
1280 (push_command_line_include): Update.
1281 * c-lex.c (cb_line_change): Update.
1282 (cb_def_pragma): Update.
1283 (cb_define): Update.
1285 (c_lex_with_flags): Use cpp_get_token_with_location.
1286 * input.h (line_table): Changed type.
1287 (location_from_locus): New macro.
1288 * tree.h (EXPR_FILENAME): No longer an lvalue.
1289 (EXPR_LINENO): Likewise.
1290 (expr_locus, set_expr_locus): Declare separately for
1291 USE_MAPPED_LOCATION.
1292 (expr_filename, expr_lineno): Changed return type.
1293 * gimplify.c (tree_to_gimple_tuple): Use SET_EXPR_LOCUS.
1294 * cfgexpand.c (expand_gimple_cond_expr): Use location_from_locus.
1295 (expand_gimple_basic_block): Likewise.
1296 * final.c (final_scan_insn): Use expanded_location.
1298 2007-09-06 Richard Guenther <rguenther@suse.de>
1300 PR tree-optimization/32586
1301 * tree-ssa-sccvn.c (simplify_binary_expression): Avoid
1302 folding if nothing changed.
1303 (simplify_unary_expression): New function. Do tree combining
1304 on conversion like codes.
1305 (try_to_simplify): Call it.
1306 * builtins.c (fold_builtin_cexp): Fold the built expressions.
1307 * fold-const.c (fold_unary): Test result of get_callee_fndecl().
1309 2007-09-06 Jan Hubicka <jh@suse.cz>
1312 * cse.c (fold_rtx): Avoid invalid sharing.
1314 2007-09-06 Richard Sandiford <richard@codesourcery.com>
1316 * config/mips/mips.md (fixuns_truncdfsi2, fixuns_truncdfdi2)
1317 (fixuns_truncsfsi2, fixuns_truncsfdi2): Update calls to real_2expN.
1319 2007-09-06 Revital Eres <eres@il.ibm.com>
1321 * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
1322 Fix offsettable memory reference for 750CL.
1324 2007-09-06 Pat Haugen <pthaugen@us.ibm.com>
1326 * reload.c (find_reloads_address_1): Try to preserve original
1327 base/index regclass of operands.
1328 * regrename.c (scan_rtx_address): Likewise.
1329 (replace_oldest_value_addr): Likewise.
1331 2007-09-06 Richard Guenther <rguenther@suse.de>
1333 PR tree-optimization/33302
1334 * domwalk.c (walk_dominator_tree): The exit block is
1335 interesting even if it is not reachable.
1337 2007-09-06 Richard Sandiford <richard@codesourcery.com>
1340 * optabs.c (avoid_expensive_constant): Do nothing if MODE is VOIDmode.
1342 2007-09-06 Basile Starynkevitch <basile@starynkevitch.net>
1344 * cfg.c (dump_bb_info, dump_edge_info): Added cfun test for
1347 2007-09-05 Ian Lance Taylor <iant@google.com>
1349 * tree-pretty-print.c (dump_decl_name): Cast LABEL_DECL_UID to int
1351 (dump_generic_node): Likewise.
1352 * print-rtl.c (print_decl_name): Likewise.
1353 * print-tree.c (print_node_brief): Likewise.
1354 (print_node): Likewise.
1355 * Makefile.in (RTL_BASE_H): Add alias.h.
1358 2007-09-06 Jesper Nilsson <jesper.nilsson@axis.com>
1360 * longlong.h [__CRIS_arch_version >= 8] (count_trailing_zeros):
1362 * config/cris/cris.md (ctzsi2, cris_swap_bits): Implemented.
1363 * config/cris/cris.h (CTZ_DEFINED_VALUE_AT_ZERO): Defined.
1365 2007-09-06 Jie Zhang <jie.zhang@analog.com>
1367 * config.gcc (tm_file): Add linux.h for bfin*-uclinux*.
1368 (tm_defines): Define UCLIBC_DEFAULT to 1.
1369 (extra_options): Add linux.opt.
1370 * config/bfin/linux.h (CPLUSPLUS_CPP_SPEC): Don't define.
1371 (CRT_CALL_STATIC_FUNCTION): Likewise.
1372 (NO_IMPLICIT_EXTERN_C): Likewise.
1373 (TARGET_OS_CPP_BUILTINS): Define as LINUX_TARGET_OS_CPP_BUILTINS.
1374 * config/bfin/elf.h (OBJECT_FORMAT_ELF): Don't define.
1375 * config/bfin/uclinux.h (CPLUSPLUS_CPP_SPEC): Don't define.
1376 (ENDFILE_SPEC): Don't define.
1377 (LIB_SPEC): Likewise.
1378 (CRT_CALL_STATIC_FUNCTION): Likewise.
1379 (NO_IMPLICIT_EXTERN_C): Likewise.
1380 (LINUX_TARGET_OS_CPP_BUILTINS): Likewise.
1381 (TARGET_OS_CPP_BUILTINS): Define as LINUX_TARGET_OS_CPP_BUILTINS.
1383 2007-09-06 Jan Hubicka <jh@suse.cz>
1385 * i386.c (ix86_expand_lround, ix86_expand_round): Update call of
1388 2007-09-06 Jan Hubicka <jh@suse.cz>
1390 * opts.c (common_handle_option): Enable inlining functions for
1393 2007-09-06 Jakub Jelinek <jakub@redhat.com>
1395 * builtin-attrs.def (ATTR_NONNULL_3): New.
1396 (DEF_FORMAT_ATTRIBUTE): Use just ATTR_NONNULL_##FA instead of
1397 ATTR_NOTHROW_NONNULL_##FA.
1398 (DEF_FORMAT_ATTRIBUTE_NOTHROW, DEF_FORMAT_ATTRIBUTE_BOTH): New macros.
1399 (ATTR_FORMAT_PRINTF_NOTHROW_2_0, ATTR_PRINTF_NOTHROW_2_3,
1400 ATTR_FORMAT_PRINTF_NOTHROW_3_0, ATTR_FORMAT_PRINTF_NOTHROW_3_4,
1401 ATTR_FORMAT_PRINTF_NOTHROW_4_0, ATTR_PRINTF_NOTHROW_4_5,
1402 ATTR_FORMAT_PRINTF_NOTHROW_5_0, ATTR_FORMAT_PRINTF_NOTHROW_5_6,
1403 ATTR_FORMAT_SCANF_NOTHROW_2_0, ATTR_FORMAT_SCANF_NOTHROW_2_3,
1404 ATTR_FORMAT_STRFTIME_NOTHROW_3_0, ATTR_FORMAT_STRFMON_NOTHROW_3_4):
1406 (ATTR_FORMAT_PRINTF_4_0, ATTR_PRINTF_4_5, ATTR_FORMAT_PRINTF_5_0,
1407 ATTR_FORMAT_PRINTF_5_6, ATTR_FORMAT_STRFTIME_3_0,
1408 ATTR_FORMAT_NOTHROW_3_4): Remove.
1409 * builtins.def (snprintf, sprintf, sscanf, vsnprintf, vsprintf,
1410 vsscanf, strfmon, strftime, __snprintf_chk, __sprintf_chk,
1411 __vsnprintf_chk, __vsprintf_chk): Use ATTR_FORMAT_*_NOTHROW_*
1412 instead of ATTR_FORMAT_*_*.
1414 2007-09-06 Hans-Peter Nilsson <hp@axis.com>
1416 * config/cris/cris.c (cris_emit_movem_store): Call copy_rtx on
1417 parts re-used for REG_FRAME_RELATED_EXPR.
1419 2007-09-05 Janis Johnson <janis187@us.ibm.com>
1421 * optabs.c (expand_float): Convert unsigned integer as signed only
1422 if it provides sufficient accuracy; add mode argument to real_2expN.
1423 (expand_fix): Fix comment typos; extend binary float into mode
1424 wider than destination for converion to unsigned integer; add mode
1425 argument to real_2expN.
1426 * real.c (real_2expN): Add mode argument to special-case decimal
1428 * real.h (real_2expN): Ditto.
1429 * fixed-value.c (check_real_for_fixed_mode): Add mode argument to
1431 (fixed_from_string): Ditto.
1432 (fixed_to_decimal): Ditto.
1433 (fixed_convert_from_real): Ditto.
1434 (real_convert_from_fixed): Ditto.
1435 * config/rs6000/rs6000.md (FP): Include DD and TD modes.
1436 * config/rs6000/dfp.md (extendddtd2, adddd3, addtd3, subdd3, subtd3,
1437 muldd3, multd3, divdd3, divtd3, cmpdd_internal1, cmptd_internal1,
1438 floatditd2, ftruncdd2, fixdddi2, ftrunctd2, fixddi2): New.
1440 2007-09-05 Ian Lance Taylor <iant@google.com>
1442 * init-regs.c (initialize_uninitialized_regs): Call
1443 emit_insn_before rather than add_insn_before.
1445 2007-09-05 Jakub Jelinek <jakub@redhat.com>
1447 * builtins.def (BUILT_IN_VA_ARG_PACK): New built-in.
1448 * tree.h (CALL_EXPR_VA_ARG_PACK): Define.
1449 * tree-inline.h (copy_body_data): Add call_expr field.
1450 * tree-inline.c (expand_call_inline): Initialize call_expr.
1451 (copy_bb): Append anonymous inline fn arguments to arguments
1452 when inlining a CALL_EXPR_VA_ARG_PACK call.
1453 * builtins.c (expand_builtin): Issue an error if
1454 BUILT_IN_VA_ARG_PACK is seen during expand.
1455 (fold_call_expr, fold_builtin_call_array): Don't fold
1456 CALL_EXPR_VA_ARG_PACK CALL_EXPRs or calls with
1457 __builtin_va_arg_pack () call as last argument.
1458 * gimplify.c (gimplify_call_expr): If last argument to a vararg
1459 function is __builtin_va_arg_pack (), decrease number of call
1460 arguments and instead set CALL_EXPR_VA_ARG_PACK on the CALL_EXPR.
1461 * expr.c (expand_expr_real_1): Issue an error if
1462 CALL_EXPR_VA_ARG_PACK CALL_EXPR is seen during expand.
1463 * tree-pretty-print.c (dump_generic_node): Handle printing
1464 CALL_EXPR_VA_ARG_PACK bit on CALL_EXPRs.
1465 * doc/extend.texi (__builtin_va_arg_pack): Document.
1467 2007-09-05 Adam Nemet <anemet@caviumnetworks.com>
1469 PR tree-optimization/21513
1470 * builtins.c (build_builtin_expect_predicate): New function.
1471 (fold_builtin_expect): Add argument for expected value.
1472 Distribute __builtin_expect over short-circuiting operations.
1473 Fold nested builtin_expects.
1474 (fold_builtin_2): Adjust call to fold_builtin_expect.
1476 2007-09-05 Rask Ingemann Lambertsen <rask@sygehus.dk>
1479 PR tree-optimization/13756
1480 * doc/invoke.texi (Options That Control Optimization): Document
1483 2007-09-05 Uros Bizjak <ubizjak@gmail.com>
1485 * gcc/config/i386/cpuid.h: New file.
1486 * gcc/config/i386/driver-i386.c: Include cpuid.h.
1487 (describe_cache): Shrink size and line strings to 100 bytes.
1488 (detect_caches_amd): Return "" for unsupported max_ext_level.
1489 Use __cpuid function.
1490 (detect_caches_intel): Return "" for unsupported max_level.
1491 Use __cpuid function.
1492 (host_detect_local_cpu): Change feature flag variables to
1493 unsigned int. Initialize only extended feature flag variables.
1494 Use __get_cpuid_max to determine max supported cpuid level.
1495 Use __cpuid function to determine supported features. Fix
1496 calculation of family id. Remove is_amd and check signature
1497 directly. Check for Geode signature. Handle family 4 id.
1498 [PROCESSOR_GENERIC32]: New default for unknown family id. Move
1499 cpu discovery code to lower part of the function.
1500 [PROCESSOR_PENTIUM, PROCESSOR_K6, PROCESSOR_ATHLON]: Do not tune
1501 for sub-architecture.
1502 [PROCESSOR_PENTIUMPRO]: Simplify cpu discovery code.
1503 [PROCESSOR_K8]: Add k8-sse3 architecture.
1504 [PROCESSOR_NOCONA]: Remove.
1505 [PROCESSOR_GENERIC64]: Ditto.
1506 * gcc/config/i386/x-i386 (driver-i386.o): Depend on cpuid.h.
1507 * gcc/config/i386/crtfastmath.c: Include cpuid.h. Use __get_cpuid
1508 to check for SSE and FXSAVE support.
1509 * gcc/config/i386/t-crtfm (crtfastmath.o): Depend on cpuid.h.
1510 Add -minline-all-stringops.
1511 * gcc/config.gcc (i[34567]86-*-*): Add cpuid.h to extra_headers.
1512 (x86_64-*-*): Ditto.
1514 2007-09-05 Jie Zhang <jie.zhang@analog.com>
1516 * config/bfin/linux-unwind.h: New file.
1517 * config/bfin/linux.h (MD_UNWIND_SUPPORT): Define.
1518 * config/bfin/uclinux.h (MD_UNWIND_SUPPORT): Define.
1520 2007-09-05 Laurynas Biveinis <laurynas.biveinis@gmail.com>
1522 * Makefile.in (stmt.o): Add alloc-pool.h dependency.
1523 * stmt.c: Include alloc-pool.h.
1524 (struct case_node): Remove GTY marker.
1525 (add_case_node): New parameter case_node_pool. Use it for
1527 (expand_case): New allocation pool case_node_pool. Initialize it,
1528 pass to add_case_node and free it.
1530 2007-09-05 Sandra Loosemore <sandra@codesourcery.com>
1531 David Ung <davidu@mips.com>
1532 Nigel Stephens <nigel@mips.com>
1534 Add mips16/nomips16 function attributes and -mflip-mips16 option
1535 for testing mixed-mode compilation.
1537 * config/mips/mips.opt (mflip-mips16): New.
1539 * config/mips/mips.h (SYMBOL_FLAG_MIPS16_FUNC): Define.
1540 (SYMBOL_FLAG_MIPS16_FUNC_P): Define.
1542 * config/mips/mips.c (mips_base_target_flags): New.
1543 (mips_base_mips16): New.
1544 (mips_base_schedule_insns): New.
1545 (mips_base_reorder_blocks_and_partition): New.
1546 (mips_base_align_loops): New.
1547 (mips_base_align_jumps): New.
1548 (mips_base_align_functions): New.
1549 (mips16_flipper): New.
1550 (mips_attribute_table): Add "mips16" and "nomips16" entries.
1551 (TARGET_SET_CURRENT_FUNCTION): Define.
1552 (mips_mips16_type_p, mips_nomips16_type_p): New.
1553 (mips_comp_type_attributes): Check mips16/nomips16 attributes.
1554 (mips_function_ok_for_sibcall): Make it deal with functions with
1556 (mips_init_split_addresses): New, split out from override_options.
1557 (mips_init_relocs): New, split out from override_options.
1558 (was_mips16_p): New.
1559 (mips_set_mips16_mode): New, split out from override_options.
1560 (mips_set_current_function): New.
1561 (override_options): Add sorry for unsupported mips16/pic
1562 combination. Remove error for mips16/dsp combination. Save
1563 base option settings.
1564 (mips_file_start): Move mips16 mode setting output from here....
1565 (mips_output_function_prologue): ....to here.
1566 (mips_output_mi_thunk): Check for mips16 function.
1567 (build_mips16_function_stub): Don't set .mips16 here.
1568 (build_mips16_call_stub): Likewise.
1569 (mips_expand_builtin): Error in mips16 mode.
1570 (mips_use_mips16_mode_p): New.
1571 (mips_encode_section_info): Check for mips16 function, and set
1572 SYMBOL_REF_FLAGS accordingly.
1574 * doc/extend.texi (Function Attributes): Document new
1575 mips16/nomips16 attributes.
1576 * doc/invoke.texi (Option Summary): Add -mflip-mips16.
1577 (MIPS Options): Document -mflip-mips16.
1579 2007-09-05 Sandra Loosemore <sandra@codesourcery.com>
1581 * config/mips/mips.c (mips_legitimize_tls_address): Call sorry
1582 if we encounter TLS address in MIPS16 mode.
1583 (mips_legitimize_const_move): Check cannot_force_const_mem for
1584 the (const (plus symbol offset)) case; this forces invalid TLS
1585 address in MIPS16 mode to be caught by the above call to sorry.
1586 (override_options): Don't reset targetm.have_tls in MIPS16 mode,
1587 because that now enables emutls, which is not ABI compatible
1588 with native TLS in non-MIPS16 mode.
1590 2007-09-05 Sandra Loosemore <sandra@codesourcery.com>
1592 Add target hook invoked when cfun changes.
1594 * doc/tm.texi (TARGET_SET_CURRENT_FUNCTION): Document.
1595 * target.h (struct gcc_target): Add set_current_function.
1596 * target-def.h (TARGET_SET_CURRENT_FUNCTION): Define.
1597 (TARGET_INITIALIZER): Add initializer for set_current_function.
1598 * tree.h (push_struct_function): New.
1599 * tree-inline.h (push_cfun, pop_cfun): Move declarations to...
1601 (set_cfun): Declare.
1602 * tree-inline.c (cfun_stack, push_cfun, pop_cfun): Moved to...
1604 (push_function_context_to): Use allocate_struct_function
1605 to create null context, not init_dummy_function_start. Use set_cfun.
1606 (pop_function_context_from): Use set_cfun.
1607 (in_dummy_function): New.
1608 (invoke_set_current_function_hook): New.
1610 (push_cfun, pop_cfun): Use set_cfun.
1611 (push_struct_function): New.
1612 (allocate_struct_function): Call invoke_set_current_function_hook
1614 (prepare_function_start): Don't set cfun here. Remove unused
1615 argument; fix all callers.
1616 (init_dummy_function_start): Fiddle with in_dummy_function. Call
1617 push_struct_function.
1618 (init_function_start): Set cfun here.
1619 (expand_dummy_function_end): Fiddle with in_dummy_function. Pop cfun.
1620 * omp-low.c (create_omp_child_function): Use push_struct_function
1621 and pop_cfun to save/restore state.
1622 (expand_omp_parallel): Remove unused saved_cfun variable.
1623 * cgraphunit.c (ipa_passes): Use set_cfun.
1624 * gimple-low.c (record_vars_into): Use push_cfun/pop_cfun here.
1625 * dwarf2out.c (dwarf2out_abstract_function): Likewise.
1626 * matrix-reorg.c (transform_allocation_sites): Likewise.
1627 (matrix_reorg): Use set_cfun.
1628 * gimplify.c (gimplify_function_tree): Use push_cfun/pop_cfun here.
1629 * tree-optimize.c (tree_rest_of_compilation): Remove one redundant
1630 assignment to cfun; use set_cfun for the other.
1631 * tree-cfg.c (move_sese_region_to_fn): Use set_cfun.
1632 (dump_function_to_file): Use push_cfun/pop_cfun here.
1633 * c-decl.c (finish_function): Use set_cfun.
1635 2007-09-05 Kenneth Zadeck <zadeck@naturalbridge.com>
1637 * regrename.c (rerename_optimize): Use deferred rescanning and
1638 insert explicit calls to rescan insns when changed.
1640 2007-09-05 Jan Hubicka <jh@suse.cz>
1642 * optabs.c (libfunc_entry): Change optab to integer.
1643 (hash_libfunc, convert_optab_libfunc, optab_libfunc,
1644 set_optab_libfunc, set_conv_libfunc): Use optable indexes instead
1645 of pointers for hashing.
1647 2007-09-05 Jan Hubicka <jh@suse.cz>
1649 * regrename.c (pass_regrename, pass_cprop_hardreg): Add RTL sharing
1651 * fwprop.c (pass_rtl_fwprop, pass_rtl_fwprop_add): Likewise.
1652 * see.c (pass_see): Likewise.
1653 * tracer.c (pass_tracer): Likewise.
1654 * postreload-gcse.c (pass_gcse2): Likewise.
1655 * postreload.c (pass_postreload_cse): Likewise.
1656 * mode-switching.c (pass_mode_switching): Likewise.
1657 * modulo-sched.c (pass_sms): Likewise.
1658 * cse.c (cse_insn): Likewise.
1659 * web.c (pass_web): Likweise.
1660 * combine-stack-adj.c (pass_stack_adjustments): Likewise.
1661 * dce.c (pass_ud_rtl_dce, pass_fast_rtl_dce): Likewise.
1662 * loop-init.c (pass_rtl_loop_init): Likewise.
1663 (pass_rtl_loop_done, pass_rtl_move_loop_, pass_rtl_unswitch,
1664 pass_rtl_unroll_and, pass_rtl_doloop): Likewise.
1665 * global.c (pass_global_alloc): Likewise.
1666 * ifcvt.c (pass_rtl_ifcvt, pass_if_after_combine,
1667 pass_if_after_reload): Likewise.
1668 * reload.c (pass_peephole2, pass_split_for_shoren_branches): Likewise.
1669 * dse.c (pass_rtl_dse1, pass_rtl_dse2): Likewise.
1670 * regmove.c (pass_regmove): Likewise.
1671 * function.c (pass_thread_prologugues_epilogues): Likewise.
1672 * gcse.c (pass_gcse): Likewise.
1673 * rtl-factoring.c (pass_rtl_seqabstr): Likewise.
1674 * lower-subreg.c (pass_lower_subreg2): Likewise.
1675 * bt-load.c (pass_branch_target_load): Likewise.
1676 * emit-rtl.c (pass_unshare_all_rtl): Likewise.
1677 * cfgcleanup.c (pass_jump, pass_jump2): Likewise.
1678 * combine.c (pass_combine): Likewise.
1679 * bb-reorder.c (pass_duplicate_comp, pass_reorder_blocks): Likewise.
1680 (pass_partition_blocks): Likewise.
1681 * var-tracking.c (pass_variable_track): Likewise.
1682 * reg-stack.c (pass_stack_regs_run): Likewise.
1683 * sched-rgn.c (pass_sched, pass_sched2): Likewise.
1684 * passes.c (pass_postreload): Likewise.
1685 (execute_function_todo): Add TODO_verify_rtl_sharing handling code.
1686 * tree-pass.h (TODO_verify_rtl_sharing): New.
1687 (TODO_update_ssa, TODO_update_ssa_no_phi, TODO_update_ssa_full_phi,
1688 TODO_update_ssa_only_virtuals, TODO_remove_unused_locals,
1689 TODO_set_props, TODO_df_finish, TODO_df_verify,
1690 TODO_mark_first_instance, TODO_rebuild_alias): Renumber.
1692 2007-09-05 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1694 * fold-const.c (all_ones_mask_p, sign_bit_p, simple_operand_p,
1695 fold_real_zero_addition_p, reorder_operands_p,
1696 div_if_zero_remainder, fold_undefer_overflow_warnings,
1697 int_binop_types_match_p, fold_convert_const_int_from_int,
1698 fold_convert_const_int_from_real,
1699 fold_convert_const_int_from_fixed,
1700 fold_convert_const_real_from_real,
1701 fold_convert_const_real_from_fixed,
1702 fold_convert_const_fixed_from_fixed,
1703 fold_convert_const_fixed_from_int,
1704 fold_convert_const_fixed_from_real, maybe_lvalue_p,
1705 fold_checksum_tree, fold_check_failed, print_fold_checksum,
1706 debug_fold_checksum, multiple_of_p): Constify.
1707 * tree-flow-inline.h (get_lineno): Likewise.
1708 * tree-flow.h (get_lineno): Likewise.
1709 * tree-object-size.c (compute_object_offset, addr_object_size,
1710 alloc_object_size, pass_through_call): Likewise.
1711 * tree-pretty-print.c (op_symbol, print_call_name,
1712 print_struct_decl, do_niy): Likewise.
1713 * tree.h (fold_undefer_overflow_warnings, multiple_of_p,
1714 debug_fold_checksum): Likewise.
1716 2007-09-04 Steven Bosscher <steven@gcc.gnu.org>
1719 * lower-subreg.c (resolve_clobber): If we remove a REG_LIBCALL
1720 note, remove the associated REG_RETVAL note.
1722 2007-09-04 Andrew Pinski <andrew_pinski@playstation.sony.com>
1724 * config.gcc (powerpc*-*-*): Install
1725 spu2vmx.h, vec_types.h, and si2vmx.h headers.
1726 * config/rs6000/spu2vmx.h: New header.
1727 * config/rs6000/si2vmx.h: New header.
1728 * config/rs6000/vec_types.h: New header.
1730 2007-09-05 Ben Elliston <bje@au.ibm.com>
1732 * varasm.c (initializer_constant_valid_p): Fix comment typo.
1734 2007-09-05 Ben Elliston <bje@au.ibm.com>
1736 * config/rs6000/ppu_intrinsics.h (__ldarx): Use `Z' constraint,
1738 (__stdcx): Make asm volatile.
1739 (__stwcx, __stdcx): Return only 0 or 1.
1741 2007-09-05 Jan Hubicka <jh@suse.cz>
1745 2007-09-04 Jan Hubicka <jh@suse.cz>
1746 * tree-tailcall.c (eliminate_tail_call): Expect unrenamed return value.
1748 2007-09-04 Richard Guenther <rguenther@suse.de>
1749 * tree-ssa-operands.c (add_virtual_operand): Only mark
1750 stores as has_volatile_ops if alias information is not available.
1752 2007-09-05 Jakub Jelinek <jakub@redhat.com>
1754 * config/rs6000/tramp.asm: Include config.h.
1755 Check __PIC__ or __pic__ macro instead of SHARED.
1757 PR rtl-optimization/32300
1758 * see.c (see_copy_insn): New function.
1759 (see_def_extension_not_merged, see_merge_one_use_extension,
1760 see_merge_one_def_extension): Use it. Avoid changing
1761 PREV_INSN/NEXT_INSN chains directly, insted emit insns
1762 into sequences. Call df_insn_delete on temporary insns
1763 that won't be emitted into the insn stream.
1764 (rest_of_handle_see): Turn off DF_DEFER_INSN_RESCAN
1765 and run df_process_deferred_rescans () before run_fast_dce.
1767 PR tree-optimization/33017
1768 * tree-data-ref.c (split_constant_offset) <case SSA_NAME>: Don't
1769 recurse for pure or const function calls.
1771 2007-09-04 Laurynas Biveinis <laurynas.biveinis@gmail.com>
1773 * c-format.c: Include alloc-pool.h.
1774 (check_format_info_main): New argument fwt_alloc. Use allocation
1775 pool instead of GC. Remove GC deallocation code.
1776 (check_format_arg): Create allocation pool, pass it to
1777 check_format_info_main and free it afterwards.
1778 * Makefile.in (c-format.o): Add alloc-pool.h dependency.
1780 2007-09-05 Ben Elliston <bje@au.ibm.com>
1782 * config/rs6000/ppu_intrinsics.h (__protected_stream_count):
1783 Rename count macro arugment for clarity.
1784 (__protected_unlimited_stream_set): Pass UG=1 to __dcbt_TH1000.
1785 (__protected_stream_set): Likewise, pass 0.
1787 2007-09-04 Jan Hubicka <jh@suse.cz>
1789 * tree-tailcall.c (eliminate_tail_call): Expect unrenamed return value.
1791 2007-09-04 Jan Hubicka <jh@suse.cz>
1793 * invoke.texi (-finline-small-functions): Document.
1794 * ipa-inline.c (cgraph_default_inline_p): Do not use DECL_INLINE
1795 when deciding what is inlinable.
1796 (cgraph_decide_recursive_inlining): Handle flag_inline_functions.
1797 (cgraph_decide_inlining_of_small_function): Handle new flags.
1798 (cgraph_decide_inlining_incrementally): Likewise.
1799 * opts.c (decode_options): Enable flag_inline_small_functions at -O2
1800 * common.opt (finline-small-functions): New.
1801 * Makefile.in (build/gengtype.o-warn): Work around PR29478
1803 2007-09-04 Richard Guenther <rguenther@suse.de>
1805 * tree-ssa-operands.c (add_virtual_operand): Only mark
1806 stores as has_volatile_ops if alias information is not available.
1808 2007-09-04 Andrew Haley <aph@redhat.com>
1810 * config/arm/libgcc-bpabi.ver: Add _Unwind_Backtrace as GCC_4.3.0.
1812 2007-09-04 Andrew Haley <aph@redhat.com>
1814 * config/arm/libunwind.S (UNWIND_WRAPPER _Unwind_Backtrace): New.
1815 * config/arm/unwind-arm.h (__gnu_Unwind_Backtrace): New.
1816 * config/arm/unwind-arm.c (__gnu_Unwind_Backtrace): New.
1818 2007-09-04 Emmanuel Thome <Emmanuel.Thome@inria.fr>
1821 * common.opt: Mention ABI version 2 in comment.
1823 2007-09-04 Jan Hubicka <jh@suse.cz>
1825 * optabs.c (debug_optab_libfunc): Update; make available to gdb.
1826 (libfunc_entry): New structure.
1827 (libfunc_hash): New hashtable.
1828 (hash_libfunc): New function.
1829 (eq_libfunc): New function.
1830 (convert_optab_libfunc): New function.
1831 (optab_libfunc): New function.
1832 (expand_binop, sign_expand_binop, expand_twoval_binop_libfunc,
1833 expand_unop, prepare_cmp_insn, prepare_float_insn, gen_add2_insn,
1834 expand_float, expand_fix, new_optab, new_convert_optab):
1835 Update for new libfunc API.
1836 (init_libfunc, init_integral_libfuncs,
1837 init_floating_libfuncs, init_interclass_conv_libfuncs
1838 init_intraclass_conv_libfuncs): Remove; reorganize all logic to:
1839 (gen_libfunc, gen_int_libfunc, gen_fp_libfunc, gen_int_fp_libfunc,
1840 gen_intv_fp_libfunc, gen_interclass_conv_libfunc,
1841 gen_int_to_fp_conv_libfunc, gen_ufloat_conv_libfunc,
1842 gen_int_to_fp_nondecimal_conv_libfunc, gen_fp_to_int_conv_libfunc,
1843 gen_intraclass_conv_libfunc, gen_trunc_conv_libfunc,
1844 gen_extend_conv_libfunc): New.
1845 (init_one_libfunc): Revamp for hashtables.
1846 (set_conv_libfunc): Likewise.
1847 (init_optabs): Initialize hashtable; use lazy initialization where
1849 * optabs.h (optab_handlers): Move out of GGC.
1850 (optab, convert_optab): Move out of GGC; add lazy gen info.
1851 (code_to_optab, convert_optab_table, optab_table): Move out of GGC.
1852 (optab_libfunc, convert_optab_libfunc): New.
1853 * builtins.c (expand_builtin_powi): Update for new API.
1854 * expr.c (convert_move): Likewise.
1855 * expmed.c (expand_divmod): Likewise.
1857 2007-09-04 Daniel Jacobowitz <dan@codesourcery.com>
1859 * config/rs6000/rs6000.c (rs6000_stack_info): Allocate space for the
1860 GOT pointer only if there is a constant pool. Use the allocated space
1863 2007-09-04 Daniel Jacobowitz <dan@codesourcery.com>
1865 * config/rs6000/linux-unwind.h (ppc_fallback_frame_state): Correct
1868 2007-09-04 Revital Eres <eres@il.ibm.com>
1869 Richard Guenther <rguenther@suse.de>
1870 R. Clint Whaley <whaley@cs.utsa.edu>
1872 * doc/invoke.texi (-fassociative-math, -freciprocal-math):
1874 * tree-tailcall.c (process_assignment): Use -fassociative-math
1875 when reodering operands of floating-point type.
1876 * fold-const.c (fold_comparison, fold_binary): Use
1877 -fassociative-math and -freciprocal-math instead of
1878 -funsafe-math-optimization flag.
1879 * toplev.h (set_unsafe_math_optimizations_flags): Declare function.
1880 * tree-ssa-math-opts.c (gate_cse_reciprocals): Use
1881 -freciprocal-math instead of -funsafe-math-optimizations.
1882 * opts.c (set_fast_math_flags): Set -freciprocal-math and
1883 -fassociative-math when -ffast-math is set.
1884 (set_unsafe_math_optimizations_flags): New Function
1885 to set -freciprocal-math and -fassociative-math when
1886 -funsafe-math-optimizations is set.
1887 (common_handle_option): Call it.
1888 * tree-vectorizer.c (vect_is_simple_reduction): Use
1889 -fassociative-math when doing reduction on floats.
1890 * loop-unroll.c (analyze_insn_to_expand_var): Use
1891 -fassociative-math when expanding an accumulator of type float.
1892 * simplify-rtx.c (simplify_binary_operation_1): Use
1893 -fassociative-math and -freciprocal-math when reordeing operands
1894 of floating-point type.
1895 * combine.c (combine_simplify_rtx): Likewise.
1896 * tree-ssa-reassoc.c (break_up_subtract_bb, reassociate_bb):
1898 * common.opt (-fassociative-math, -freciprocal-math): New flags.
1900 2007-09-04 Paolo Carlini <pcarlini@suse.de>
1903 * doc/install.texi (--enable-__cxa_atexit): Fix typo.
1905 2007-09-04 Jan Hubicka <jh@suse.cz>
1907 * regrename.c(copyprop_hardreg_forward_1): Unshare when doing copyprop.
1909 2007-09-04 Jan Hubicka <jh@suse.cz>
1911 * regrename.c (build_def_use): Don't share RTL between MATCH_OPERATOR
1912 and corresponding MATCH_DUP.
1914 2007-09-04 Uros Bizjak <ubizjak@gmail.com>
1917 * combine.c (subst): Do not try to simplify X if it represents load
1918 of FP constant from the constant pool via float extension.
1920 2007-09-04 Ben Elliston <bje@au.ibm.com>
1922 * c-opts.c: Include "tm_p.h".
1923 * Makefile.in (c-opts.o): Add $(TM_P_H) as a dependency.
1925 2007-09-04 Richard Guenther <rguenther@suse.de>
1927 PR tree-optimization/33291
1928 * tree-pretty-print.c (dump_generic_node): Dump all
1929 qualifiers for pointer types, not only first. Dump
1930 qualifiers for aggregate types as well.
1931 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Always
1932 use the canonical type for building ARRAY_REFs.
1933 * gimplify.c (canonicalize_addr_expr): Clean up. The
1934 correct validness check is compatibility of the pointer
1935 types. Always use the canonical type for building
1936 ARRAY_REFs and ADDR_EXPRs.
1937 * tree-ssa-forwprop.c (forward_propagate_addr_expr): Revert
1938 change that disabled propagation of ADDR_EXPRs into statements
1941 2007-09-03 Zack Weinberg <zack@codesourcery.com>
1943 * config/arm/arm.md (mulsidi3, umulsidi3, mulsi3_highpart)
1944 (umulsi3_highpart): Make into expanders; existing insns are
1945 now named *[insn]_nov6. Add stars to existing [insn]_v6
1946 counterparts' names.
1948 2007-09-03 Daniel Jacobowitz <dan@codesourcery.com>
1950 * Makefile.in (DRIVER_DEFINES, gcc.o, gccspec.o, protoize.o)
1951 (unprotoize.o): Remove SHLIB_MULTILIB.
1953 2007-09-03 Eric Weddington <eweddington@cso.atmel.com>
1955 * doc/extend.texi: Document progmem attribute for AVR.
1957 2007-09-03 Anatoly Sokolov <aesok@post.ru>
1960 * config/avr/avr.h (TARGET_VTABLE_ENTRY_ALIGN): Define.
1962 2007-09-03 Zack Weinberg <zack@codesourcery.com>
1964 * optabs.c (expand_ffs): Initialize val and defined_at_zero
1967 * optabs.c: Remove unnecessary forward declarations.
1968 (expand_unop_direct): New, broken out of expand_unop.
1969 (expand_doubleword_clz): New.
1970 (expand_ctz): Move above expand_ffs. Use
1971 start_sequence, end_sequence, add_equal_note, and
1972 expand_unop_direct. Add more commentary.
1973 (expand_ffs): Try both ctz optab and expand_ctz.
1974 Generate a test and branch if the hardware doesn't give us
1975 a useful value for input zero. Style improvements similar to
1978 * config/arm/arm.md (ffssi2, ctzsi2): Delete.
1980 2007-09-03 Zack Weinberg <zack@codesourcery.com>
1982 * config.gcc: Delete stanza for arm-semi-aof and
1983 armel-semi-aof targets.
1984 * config/arm/arm-protos.h
1986 * config/arm/arm.h: Delete all #ifdef AOF_ASSEMBLER blocks;
1987 make all #ifndef AOF_ASSEMBLER blocks unconditional. Also
1988 delete aof_pic_label and remove mention of AOF in comments.
1989 * config/arm/arm.md: Delete patterns used only for AOF assembly.
1991 * config/arm/semiaof.h
1992 * config/arm/t-semi: Delete file.
1994 2007-09-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1996 * system.h (CONST_CAST2, CONST_CAST_TREE, CONST_CAST_RTX,
1997 CONST_CAST_BB): New macros for type-specific casts.
1998 (CONST_CAST): Add a TYPE parameter and define it in terms of
2001 * c-decl.c (c_make_fname_decl): Supply a TYPE for CONST_CAST.
2002 * c-lex.c (cb_ident, lex_string): Likewise,
2003 * c-typeck.c (free_all_tagged_tu_seen_up_to): Likewise,
2004 * config/i386/i386.c (ix86_function_regparm,
2005 ix86_function_sseregparm): Use CONST_CAST_TREE.
2006 * config/mmix/mmix.c (mmix_encode_section_info): Supply a TYPE for
2008 * gcc.c (set_spec, read_specs, for_each_path, execute, do_spec_1,
2009 give_switch, set_multilib_dir): Likewise,
2010 * gengtype-parse.c (string_seq, typedef_name): Likewise,
2011 * gimple-low.c (block_may_fallthru): Use CONST_CAST_TREE.
2012 * jump.c (reversed_comparison_code_parts): Use CONST_CAST_RTX.
2013 * passes.c (execute_one_pass): Supply a TYPE for CONST_CAST.
2014 * prefix.c (update_path): Likewise,
2015 * pretty-print.c (pp_base_destroy_prefix): Likewise,
2016 * rtlanal.c (keep_with_call_p): Use CONST_CAST_RTX.
2017 * tree-cfg.c (tree_can_merge_blocks_p,
2018 tree_block_ends_with_condjump_p): Use CONST_CAST_BB.
2019 * tree-eh.c (lookup_stmt_eh_region_fn): USe CONST_CAST_TREE.
2020 * tree.c (build_string): Supply a TYPE for CONST_CAST.
2021 (attribute_list_contained): Use CONST_CAST_TREE.
2023 2007-09-03 Krister Walfridsson <cato@df.lth.se>
2025 * config/netbsd-elf.h (USE_LD_AS_NEEDED): Define.
2027 2007-09-03 Richard Sandiford <richard@codesourcery.com>
2030 * optabs.c (avoid_expensive_constant): Canonicalize CONST_INTs
2031 before forcing them into a register.
2033 2007-09-03 Richard Sandiford <richard@codesourcery.com>
2035 * config/mips/mips.md (fetchop_bit): Use define_code_iterator
2036 rather than define_code_macro.
2038 2007-09-03 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2040 * gcc.c (xputenv): New function.
2041 (putenv_from_prefixes): Use xputenv instead of putenv.
2042 (process_command): Use xputenv instead of putenv.
2043 (set_collect_gcc_options): Use xputenv instead of putenv.
2044 (main): Use xputenv instead of putenv.
2046 2007-09-03 Jesper Nilsson <jesper.nilsson@axis.com>
2048 * config/cris/cris.h (TARGET_HAS_SWAP): Defined to describe
2049 availability of swap instruction.
2050 * config/cris/cris.md (bswapsi2): Implement using swap instruction.
2052 2007-09-03 Revital Eres <eres@il.ibm.com>
2054 * doc/invoke.texi (-fmodulo-sched-allow-regmoves): Update
2057 2007-09-03 Vladimir Yanovsky <yanov@il.ibm.com>
2058 Ayal Zaks <zaks@il.ibm.com>
2059 Revital Eres <eres@il.ibm.com>
2061 * modulo-sched.c (ps_insert_empty_row, verify_partial_schedule,
2062 compute_split_row): New functions.
2063 (ps_unschedule_node): Remove.
2064 (normalize_sched_times): Iterate over the already scheduled
2065 insns instead of the number of nodes.
2066 (MAX_SPLIT_NUM): New definition.
2067 (sms_schedule_by_order): Change the scheduling heuristic to
2068 avoid useless increases of initiation interval ii.
2069 (get_sched_window): Add dump printouts.
2071 2007-09-02 David Daney <ddaney@avtrex.com>
2073 * config/mips/mips.md (UNSPEC_COMPARE_AND_SWAP, UNSPEC_SYNC_OLD_OP,
2074 UNSPEC_SYNC_NEW_OP, UNSPEC_SYNC_EXCHANGE): New define_constants.
2075 (optab, insn): Add more attributes.
2076 (fetchop_bit): New code macro.
2077 (immediate_insn): New code macro attribute.
2078 (sync): Change condition to ISA_HAS_SYNC.
2079 (rdhwr): Change predicate for operand 0 to register_operand.
2080 (memory_barrier): New expand.
2081 (sync_compare_and_swap<mode>, sync_add<mode>, sync_sub<mode>,
2082 sync_old_add<mode>, sync_old_sub<mode>, sync_new_add<mode>,
2083 sync_new_sub<mode>, sync_<optab><mode>, sync_old_<optab><mode>,
2084 sync_new_<optab><mode>, sync_nand<mode>, sync_old_nand<mode>,
2085 sync_new_nand<mode>, sync_lock_test_and_set<mode>): New insns.
2086 * config/mips/mips.h (ISA_HAS_SYNC, ISA_HAS_LL_SC): New ISA predicates.
2087 (MIPS_COMPARE_AND_SWAP, MIPS_SYNC_OP, MIPS_SYNC_OLD_OP,
2088 MIPS_SYNC_NEW_OP, MIPS_SYNC_NAND, MIPS_SYNC_OLD_NAND,
2089 MIPS_SYNC_NEW_NAND, MIPS_SYNC_EXCHANGE): New Macros.
2091 2007-09-03 Jesper Nilsson <jesper.nilsson@axis.com>
2092 Hans-Peter Nilsson <hp@axis.com>
2094 * config/cris/cris.h (TARGET_HAS_LZ, CLZ_DEFINED_VALUE_AT_ZERO):
2095 Defined to describe availability and behavior of CLZ.
2096 * config/cris/cris.md (clzsi2): Implement using lz instruction.
2097 * config/cris/cris.opt: Tweak comment for "-metrax4".
2098 * config/cris/arit.c (LZ): When defined, define as __builtin_clz.
2099 * longlong.h [__CRIS__ && __CRIS_arch_version >= 3]
2100 (count_leading_zeros): Define.
2102 2007-09-01 Laurynas Biveinis <laurynas.biveinis@gmail.com>
2104 * ggc-page.c (ggc_pch_read): Call validate_free_objects.
2106 2007-09-02 Joseph Myers <joseph@codesourcery.com>
2109 * c-decl.c (grokdeclarator): Apply qualifiers to type of parameter
2112 2007-09-01 Kazu Hirata <kazu@codesourcery.com>
2114 * config/arm/arm.c, config/rs6000/ppu_intrinsics.h,
2115 config/spu/spu.c, df-scan.c, fixed-value.c, fold-const.c,
2116 ginclude/tgmath.h, haifa-sched.c, optabs.c, recog.c,
2117 sched-deps.c, sched-int.h, system.h, target.h,
2118 tree-ssa-live.c, tree-vect-transform.c, tree-vectorizer.c,
2119 tree.def: Fix comment typos.
2121 2007-09-01 Kazu Hirata <kazu@codesourcery.com>
2123 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove.
2124 (override_options): Turn on -mstrict-align on non-ColdFire
2126 * config/m68k/m68k.h (BIGGEST_ALIGNMENT): Update a comment.
2128 2007-09-01 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2130 * c-common.c (const_strip_array_types): Delete.
2131 * c-common.h (const_strip_array_types): Likewise.
2132 * emit-rtl.c (const_next_insn, const_previous_insn,
2133 const_next_nonnote_insn, const_prev_nonnote_insn,
2134 const_next_real_insn, const_prev_real_insn,
2135 const_next_active_insn, const_prev_active_insn, const_next_label,
2136 const_prev_label): Delete.
2137 * gimple-low.c (block_may_fallthru): Use expr_last.
2138 * jump.c (reversed_comparison_code_parts): Use prev_nonnote_insn.
2139 * rtl.h (const_previous_insn, const_next_insn,
2140 const_prev_nonnote_insn, const_next_nonnote_insn,
2141 const_prev_real_insn, const_next_real_insn,
2142 const_prev_active_insn, const_next_active_insn, const_prev_label,
2143 const_next_label): Delete.
2144 * rtlanal.c (keep_with_call_p): Use next_nonnote_insn.
2145 * system.h (CONST_CAST): Update comment.
2146 * tree-cfg.c (tree_can_merge_blocks_p): Use last_stmt.
2147 (const_first_stmt, const_last_stmt): Delete.
2148 (tree_block_ends_with_condjump_p): Use last_stmt.
2149 * tree-flow.h (const_first_stmt, const_last_stmt): Delete.
2150 * tree-iterator.c (const_expr_first, const_expr_last,
2151 const_expr_only): Delete.
2152 * tree.c (const_lookup_attribute): Likewise.
2153 (attribute_list_contained): Use lookup_attribute.
2154 * tree.h (const_lookup_attribute, const_expr_first,
2155 const_expr_last, const_expr_only): Delete.
2157 2007-09-01 Richard Guenther <rguenther@suse.de>
2159 * tree-ssa-alias.c (create_name_tags): Use TYPE_VOLATILE
2160 instead of TREE_THIS_VOLATILE.
2161 * tree-ssa-operands.c (add_virtual_operand): Do not set
2162 has_volatile_ops. Do not check s_ann for NULL.
2163 (get_indirect_ref_operands): Likewise.
2164 (get_expr_operands): Likewise. For all reference trees
2165 set has_volatile_ops based on TREE_THIS_VOLATILE. Also
2166 check base of reference for TREE_THIS_VOLATILE.
2168 2007-09-01 Richard Guenther <rguenther@suse.de>
2170 * tree-cfg.c (verify_gimple_expr): Fix check for conversions
2171 between integral types. Also allow conversions between
2174 2007-08-31 Andrew Pinski <andrew_pinski@playstation.sony.com>
2176 * optabs.c (prepare_float_lib_cmp): Use
2177 libgcc_cmp_return_mode instead of word_mode.
2178 * config/dfp-bit.h: Likewise.
2179 * config/fp-bit.h: Likewise.
2180 * config/spu/spu.c (spu_libgcc_cmp_return_mode):
2182 (spu_libgcc_shift_count_mode): New function.
2183 (TARGET_LIBGCC_CMP_RETURN_MODE): Define.
2184 (TARGET_LIBGCC_SHIFT_COUNT_MODE): Define.
2186 2007-08-31 Joseph Myers <joseph@codesourcery.com>
2188 * tree-pretty-print.c: Include output.h.
2189 (dump_decl_name, dump_generic_node): Use %wd with pp_printf, not
2190 HOST_WIDE_INT_PRINT_DEC.
2191 * Makefile.in (tree-pretty-print.o): Add dependency on output.h.
2193 2007-08-31 Zdenek Dvorak <ook@ucw.cz>
2195 PR rtl-optimization/33224
2196 * loop-iv.c (latch_dominating_def): Check that the definition belongs
2197 to the analysed region.
2199 2007-08-31 Richard Guenther <rguenther@suse.de>
2201 * function.c (allocate_struct_function): Do not set
2202 current_function_returns_pointer.
2203 * function.h (struct var_refs_queue): Remove.
2204 (struct simple_obstack_stack): Likewise.
2205 (struct function): Remove fixup_var_refs_queue member.
2206 Remove returns_pointer flag.
2207 (current_function_returns_pointer): Remove define.
2209 2007-08-31 Nick Clifton <nickc@redhat.com>
2211 * tree-ssa-coalesce.c (compare_pairs): Use the elements as
2212 secondary keys in order to obtain a stable sort.
2214 2007-08-31 Nick Clifton <nickc@redhat.com>
2217 * config/m32r/constraints.md: Add W constraint for integer zero.
2218 * config/m32r/m32r.md (get_pc): Use W and i constraints.
2219 Fix length calculation by using alternatives.
2221 2007-08-31 Richard Sandiford <richard@codesourcery.com>
2223 * optabs.c (shift_optab_p, commutative_optab_p): New functions,
2224 split out from expand_binop.
2225 (avoid_expensive_constant): New function.
2226 (expand_binop_directly): Remove commutative_op argument and
2227 (expand_binop_directly): Remove commutative_op argument and
2228 call cummutative_optab_p instead. Do not change op0 or op1
2229 when swapping xop0 and xop1. Apply avoid_expensive_constant
2230 to each argument after potential swapping. Enforce the
2231 canonical order of commutative operands.
2232 (expand_binop): Use shift_optab_p and commutative_optab_p.
2233 Update the calls to expand_binop_directly. Only force constants
2234 into registers when widening an operation. Only swap operands
2235 once a direct expansion has been rejected.
2236 (expand_twoval_binop): Only force constants into registers when
2237 using a direct expansion.
2239 2007-08-31 Maxim Kuvyrkov <maxim@codesourcery.com>
2241 * sched-deps.c (update_dep): Mark arguments with ATTRIBUTE_UNUSED.
2242 (sched_analyze_insn): Don't postprocess speculative dependencies when
2243 target has no scheduling.
2244 * tree-vect-transform.c (vect_estimate_min_profitable_iters): Fix
2245 'uninitialized' warning.
2247 2007-08-31 Richard Guenther <rguenther@suse.de>
2249 * tree.c (build_complex_type): Always set TYPE_NAME for
2252 2007-08-31 Olivier Hainque <hainque@adacore.com>
2254 * config/rs6000/xcoff.h (XCOFF_CSECT_DEFAULT_ALIGNMENT_STR): New
2255 internal macro. Default alignment factor for csect directives, chosen
2256 to match what BIGGEST_ALIGNMENT allows.
2257 (DATA_SECTION_ASM_OP): Use it.
2258 * config/rs6000/rs6000.c
2259 (rs6000_xcoff_output_readonly_section_asm_op):
2260 Use XCOFF_CSECT_DEFAULT_ALIGNMENT_STR.
2261 (rs6000_xcoff_output_readwrite_section_asm_op): Likewise.
2263 2007-08-31 Olivier Hainque <hainque@adacore.com>
2265 * dbxout.c (dbxout_block): Move declaration within the first
2266 #if defined (DBX_DEBUGGING_INFO) block. Rearrange another such
2267 block to include the definition, together with the associated
2270 2007-08-31 David Edelsohn <edelsohn@gnu.org>
2271 Revital Eres <eres@il.ibm.com>
2273 * doc/invoke.texi (-mpaired): Document flag.
2274 * config.gcc: Include paired.h in powerpc extra_headers and
2275 750cl.h in powerpc-*-linux*paired*.
2276 * config/rs6000/rs6000.opt (-mpaired): New flag.
2277 * config/rs6000/rs6000.c (paired_init_builtins,
2278 paired_expand_builtin, paired_expand_lv_builtin,
2279 paired_expand_stv_builtin, paired_expand_predicate_builtin):
2280 New functions to support the paired single builtin functions.
2281 (rs6000_hard_regno_mode_ok): Handle PAIRED_VECTOR_MODE.
2282 (def_builtin, bdesc_3arg, bdesc_2arg, bdesc_1arg): Add paired
2284 (bdesc_paired_preds): New structure for paired predicate
2286 (rs6000_expand_builtin): Expand paired single builtins.
2287 (rs6000_init_builtins): Init paired single builtins.
2288 (rs6000_common_init_builtins): Add v2sf_ftype_v2sf_v2sf_v2sf.
2289 Rename v2sf_ftype_v2sf_v2sf to v2sf_ftype_v2sf_v2sf_spe
2290 and v2sf_ftype_v2sf_spe to v2sf_ftype_v2sf. Add new types
2291 v2sf_ftype_v2sf_v2sf and v2sf_ftype_v2sf.
2292 (rs6000_vector_mode_supported_p): Support paired vector mode.
2293 * config/rs6000/rs6000.h: (UNITS_PER_PAIRED_WORD,
2294 PAIRED_VECTOR_MODE, PAIRED_SIMD_REGNO_P, TARGET_PAIRED_FLOAT):
2296 (LOCAL_ALIGNMENT): Handle PAIRED_VECTOR_MODE.
2297 (DATA_ALIGNMENT): Likewise.
2298 (UNITS_PER_SIMD_WORD): Handle PAIRED_VECTOR_MODE.
2299 (rs6000_builtins): Add PAIRED builtins.
2300 * config/rs6000/rs6000.md: Include paired.md.
2301 * config/rs6000/paired.h: New.
2302 * config/rs6000/paired.md: New.
2303 * config/rs6000/750cl.h: New.
2304 * config/rs6000/spe.md: Recognize movv2sf instruction for 750cl.
2306 2007-08-30 Ollie Wild <aaw@google.com>
2308 * c-opts.c (c_common_handle_option): Support -fno-directives-only.
2309 * gcc.c (default_compilers): Add -fno-directives-only to
2310 @assembler-with-cpp.
2312 2007-08-30 Sandra Loosemore <sandra@codesourcery.com>
2316 * regclass.c (initial_fixed_regs): Revert previous change and make
2318 (initial_call_used_regs): Likewise.
2319 (initial_call_really_used_regs): Delete, reverting previous addition.
2320 (initial_reg_names): Likewise.
2321 (init_reg_sets): Revert previous change.
2322 (saved_fixed_regs): New.
2323 (saved_call_used_regs): New.
2324 (saved_call_really_used_regs): New.
2325 (saved_reg_names): New.
2326 (save_register_info): New.
2327 (restore_register_info): New.
2328 (init_reg_sets_1): Replace reset of register info with call to
2329 restore_register_info.
2330 * rtl.h (save_register_info): Declare.
2331 * toplev.c (backend_init): Call save_register_info.
2333 2007-08-30 Laurynas Biveinis <laurynas.biveinis@gmail.com>
2335 * tree-ssa-propagate.c (set_rhs): Remove the copied annotation
2336 from the original statement.
2338 2007-08-30 Chao-ying Fu <fu@mips.com>
2340 * c-lex.c (interpret_fixed): Declare.
2341 (interpret_float): Process _Fract and _Accum.
2342 (interpret_fixed): New function.
2343 * final.c (output_addr_const): Process CONST_FIXED.
2344 * simplify-rtx.c (simplify_const_unary_operation): Handle US_NEG.
2345 (simplify_binary_operation_1): Handle US_ASHIFT, SS_MULT, US_MULT,
2347 (simplify_const_binary_operation): Handle SS_MULT, US_MULT, SS_DIV,
2349 (simplify_immed_subreg): Support CONST_FIXED.
2350 Process MODE_FRACT, MODE_UFRACT, MODE_ACCUM, MODE_UACCUM.
2351 (simplify_subreg): Support CONST_FIXED.
2353 2007-08-30 Andrew Pinski <andrew_pinski@playstation.sony.com>
2355 * config/rs6000/ppu_intrinsics.h: New file.
2356 * config.gcc (powerpc*-*-* <extra_headers>): Install
2359 2007-08-30 Kenneth Zadeck <zadeck@naturalbridge.com>
2361 * cfg.c (dump_flow_info): Change to also print entry and exit
2363 * print-rtl.c (print_rtl_single): Allow to print rtl with
2366 2007-08-30 Kenneth Zadeck <zadeck@naturalbridge.com>
2368 * df-core.h (df_dump_region): New function.
2369 * df.h (df_dump_region): New function.
2370 * loop-invariant.c (find_defs): Add call to df_dump_region.
2371 * loop-iv.c (iv_analysis_loop_init): Changed call from df_dump to
2374 2007-08-30 Jakub Jelinek <jakub@redhat.com>
2377 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Return
2378 true if any of the compare_section_name calls returned true,
2379 rather than if any returned false.
2381 2007-08-30 Richard Guenther <rguenther@suse.de>
2383 PR tree-optimization/33199
2384 * tree-ssa-structalias.c (handle_lhs_call): New function.
2385 (find_func_aliases): In non-IPA mode make sure that for
2386 calls that return a pointer we add a constraint for the
2387 result to point to anything.
2389 2007-08-30 Richard Guenther <rguenther@suse.de>
2391 * doc/invoke.texi (-mveclibabi): Document new target option.
2392 * config/i386/i386.opt (-mveclibabi): New target option.
2393 * config/i386/i386.c (ix86_veclib_handler): Handler for
2394 vectorization library support.
2395 (override_options): Handle the -mveclibabi option, initialize
2396 the vectorization library handler.
2397 (ix86_builtin_vectorized_function): As fallback call the
2398 vectorization library handler, if set.
2399 (ix86_veclibabi_acml): New static function for ACML ABI style
2400 vectorization support.
2402 2007-08-30 Jakub Jelinek <jakub@redhat.com>
2404 * config/rs6000/rs6000.c (rs6000_emit_sync): For QI or HI mode
2405 used_m, even if it is 32-bit aligned, adjust used_m MEM to have
2406 SImode and update m. Don't run gen_lowpart_common on arbitrary
2407 memory address, force it to register first.
2410 * dce.c (dce_process_block): Don't delete setters of
2411 artificially used registers.
2413 2007-08-30 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
2415 * config/s390/s390.md ("*add<mode>3_alc_carry1_cc",
2416 "*add<mode>3_alc_carry1_cconly", "*add<mode>3_alc_carry2_cc",
2417 "*add<mode>3_alc_carry2_cconly"): New insn definitions.
2419 2007-08-30 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
2421 * config/s390/s390.c (s390_expand_addcc): Emit canonical alc
2423 * config/s390/s390.md ("*add<mode>3_alc_cc", "*add<mode>3_alc",
2424 "addti3", "*adddi3_31z", "*scond<mode>"): Make alc pattern canonical.
2426 2007-08-29 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2428 * system.h: Activate -Wcast-qual as warning-only.
2430 2007-08-29 Paolo Bonzini <bonzini@gnu.org>
2433 * config/rs6000/rs6000.c (compare_section_name): New function.
2434 (rs6000_elf_in_small_data_p): Compare section prefixes instead
2437 2007-08-29 Olivier Hainque <hainque@adacore.com>
2439 * xcoffout.c: #include debug.h.
2440 * Makefile.in (xcoffout.o): Add debug.h dependency.
2442 2007-08-29 Uros Bizjak <ubizjak@gmail.com>
2444 * simplify-rtx.c (simplify_binary_operation_1) [VEC_SELECT]:
2445 Change CONST_VECTOR_ELT to XVECEXP in elem calculation.
2447 2007-08-29 Richard Guenther <rguenther@suse.de>
2449 * tree-pretty-print.c (dump_generic_node): Print
2450 label DECL_UID as D.%u.
2451 * tree-dfa.c (dump_variable): Print DECL_UID as D.%u.
2453 2007-08-29 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2455 * gcc/builtin-types.def (BT_FN_PTR_PTR_SIZE): New type.
2456 * gcc/builtins.def (BUILT_IN_REALLOC): New builtin.
2458 2007-08-29 Douglas Gregor <doug.gregor@gmail.com>
2461 * tree.c (build_type_attribute_qual_variant): Set canonical types
2462 on the final, unqualified attribute variant before building the
2465 2007-08-29 Jie Zhang <jie.zhang@analog.com>
2467 * config/bfin/bfin.c (bfin_expand_builtin): Fix the argument
2468 order of __builtin_bfin_cmplx_mac and __builtin_bfin_cmplx_msu.
2470 2007-08-29 Jie Zhang <jie.zhang@analog.com>
2473 2007-08-29 Jie Zhang <jie.zhang@analog.com>
2474 * config/bfin/bfin.md (composev2hi): Put operands into vector
2477 2007-08-29 Jie Zhang <jie.zhang@analog.com>
2479 * config/bfin/bfin.md (composev2hi): Put operands into vector
2482 2007-08-29 Jie Zhang <jie.zhang@analog.com>
2484 * config/bfin/bfin.c (bfin_expand_call): Inline PLT with l1_text
2485 attribute when appropriate.
2486 (bfin_handle_l1_text_attribute): New.
2487 (bfin_handle_l1_data_attribute): New.
2488 (bfin_attribute_table): Add attributes: l1_text, l1_data,
2489 l1_data_A and l1_data_B.
2490 * doc/extend.texi (node Function Attributes): Document l1_text
2492 (Variable Attributes): Add Blackfin subsection. Document l1_data,
2493 l1_data_A and l1_data_B variable attributes.
2495 2007-08-28 Jie Zhang <jie.zhang@analog.com>
2497 * config/bfin/bfin.opt (minline-plt): Add.
2498 * config/bfin/bfin.c (bfin_expand_call): Inline PLT when emit
2499 call to global functions.
2500 * doc/invoke.texi (Option Summary): Mention -minline-plt.
2501 (Blackfin Options): Document -minline-plt.
2503 2007-08-28 Nathan Sidwell <nathan@codesourcery.com>
2505 * config/m68k/m68k.c (m68k_get_function_kind): Assert we're never
2506 given a non-function.
2507 (m68k_ok_for_sibcall_p): Only sibcall functions of the same kind.
2509 2007-08-28 DJ Delorie <dj@redhat.com>
2511 * config/sh/sh.c (sh_gimplify_va_arg_expr): Fix sh2a support.
2513 2007-08-28 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2515 * system.h (CONST_CAST): Avoid union for gcc-4.0.x.
2517 2007-08-28 Richard Guenther <rguenther@suse.de>
2519 * tree.h (struct tree_function_decl): Increase size of
2520 function_code bitfield.
2522 2007-08-28 Nathan Sidwell <nathan@codesourcery.com>
2523 Kazu Hirata <kazu@codesourcery.com>
2525 * gcc/config/m68k/linux.h
2526 (M68K_HONOR_TARGET_STRICT_ALIGNMENT): Redefine as 0.
2527 * config/m68k/m68k.c (TARGET_RETURN_IN_MEMORY): New.
2528 (m68k_return_in_memory): New.
2529 * gcc/config/m68k/m68k.h (M68K_HONOR_TARGET_STRICT_ALIGNMENT):
2532 2007-08-28 Uros Bizjak <ubizjak@gmail.com>
2535 * simplify-rtx.c (simplify_binary_operation_1) [VEC_SELECT]:
2536 Simplify nested VEC_SELECT (with optional VEC_CONCAT operator as
2537 operand) when top VEC_SELECT extracts scalar element.
2538 * config/i386/sse.md (*vec_extract_v4si_mem): New pattern.
2539 (*vec_extract_v4sf_mem): Ditto.
2541 2007-08-28 Jakub Jelinek <jakub@redhat.com>
2544 * passes.c (init_optimization_passes): Move pass_df_finish
2545 after pass_postreload sublist.
2547 PR rtl-optimization/33148
2548 * simplify-rtx.c (simplify_unary_operation_1): Only optimize
2549 (neg (lt X 0)) if X has scalar int mode.
2552 * dwarf2out.c (rtl_for_decl_init): If vector decl has CONSTRUCTOR
2553 initializer, use build_vector_from_ctor if possible to create
2554 VECTOR_CST out of it. If vector initializer is not VECTOR_CST
2555 even after this, return NULL.
2557 2007-08-28 Gabriel Dos Reis <gdr@integrable-solutions.net>
2559 * c-pretty-print.c (pp_c_constant): Handle COMPLEX_CST.
2561 2007-08-28 Richard Guenther <rguenther@suse.de>
2563 * tree.h (DECL_DISREGARD_INLINE_LIMITS): New.
2564 (struct tree_function_decl): Make function_code a bitfield.
2565 Add disregard_inline_limits flag.
2566 * cgraphunit.c (cgraph_process_new_functions): Check
2567 DECL_DISREGARD_INLINE_LIMITS instead of disregard_inline_limits_p.
2568 (cgraph_preserve_function_body_p): Likewise.
2569 * ipa-inline.c (compute_inline_parameters): Likewise.
2570 * c-decl.c (finish_function): Set DECL_DISREGARD_INLINE_LIMITS
2571 for GNU C extern inline functions.
2572 (merge_decls): Merge DECL_DISREGARD_INLINE_LIMITS.
2573 * tree-inline.c (disregard_inline_limits_p): Remove.
2574 * tree-inline.h (disregard_inline_limits_p): Likewise.
2575 * c-common.c (handle_always_inline_attribute): Set
2576 DECL_DISREGARD_INLINE_LIMITS.
2577 * langhooks.c (add_builtin_function): Verify the function code
2578 fits in the bitfield.
2580 2007-08-28 Mircea Namolaru <namolaru@il.ibm.com>
2581 Vladimir Yanovsky <yanov@il.ibm.com>
2582 Revital Eres <eres@il.ibm.com>
2583 Andrey Belevantsev <abel@ispras.ru>
2585 * config/spu/spu.md: Recognize doloop pattern when -fmodulo-sched
2587 * modulo-sched.c: Add documentation regarding do-loop.
2588 (doloop_register_get): Change number of arguments to support
2589 the new do-loop pattern and check whether COUNT_REG has no other
2590 occurences in the loop besides in the control part.
2591 (duplicate_insns_of_cycles): Do not duplicate the insn which
2592 changes count_reg as it is already adjusted.
2593 (generate_prolog_epilog): New argument to support the new
2594 do-loop pattern. Change the subtract instruction to use
2595 expand_simple_binop. Call duplicate_insns_of_cycles with new
2597 (sms_schedule): Call doloop_register_get and
2598 generate_prolog_epilog with new argument. Do not handle loops
2599 with single sets insns with subreg in their lhs.
2600 * loop-doloop.c (doloop_optimize): Support for another do-loop
2602 (doloop_condition_get): Gets an instruction instead of a pattern
2603 and change the return condition when the do-loop pattern is
2605 * ddg.c (create_ddg_dep_from_intra_loop_link): Handle only reg
2606 deps when considering to not create edges.
2608 2007-08-27 Alexandre Oliva <aoliva@redhat.com>
2610 * doc/extend.texi (gnu_inline funtion attribute): Document C++
2613 2007-08-27 Jason Merrill <jason@redhat.com>
2616 * gimplify.c (gimplify_modify_expr): Discard the assignment of
2617 zero-sized types after calling gimplify_modify_expr_rhs.
2619 2007-08-27 Sandra Loosemore <sandra@codesourcery.com>
2621 * regclass.c (init_reg_autoinc): Fix typo.
2623 2007-08-27 Daniel Berlin <dberlin@dberlin.org>
2625 Fix PR tree-optimization/33173
2626 * tree-ssa-alias.c (find_used_portions): Fix reversed test.
2628 2007-08-27 H.J. Lu <hongjiu.lu@intel.com>
2631 * config/i386/i386.h (VALID_DFP_MODE_P): New.
2632 * config/i386/i386.c (ix86_hard_regno_mode_ok): Allow DFP in
2635 2007-08-27 Sandra Loosemore <sandra@codesourcery.com>
2636 David Ung <davidu@mips.com>
2637 Nigel Stephens <nigel@mips.com>
2639 Separate target-specific initialization from general
2640 back-end initialization.
2642 * toplev.c (init_alignments): New, split out from...
2643 (process_options): ...here.
2644 (backend_init_target): New, split out from...
2645 (backend_init): ...here.
2646 (lang_dependent_init_target): New, split out from...
2647 (lang_dependent_init): ...here.
2648 (target_reinit): New.
2649 * toplev.h (target_reinit): Declare.
2650 * expr.c (init_expr_target): Renamed from init_expr_once, since it
2651 now can be called more than once. Update comments.
2652 * expr.h (init_expr_target): Likewise.
2653 * alias.c (init_alias_target): Renamed from init_alias_once, since it
2654 now can be called more than once. Explicitly zero
2655 static_reg_base_value.
2656 * emit-rtl.c (init_emit_regs): New, split out from...
2657 (init_emit_once): Here.
2658 * regclass.c (initial_fixed_regs, initial_call_used_regs): Make
2659 non-const, so that changes from command-line arguments can overwrite
2660 values provided by the static initializers.
2661 (initial_call_really_used_regs): New, used similarly to the above.
2662 (initial_reg_names): Likewise.
2663 (last_mode_for_init_move_cost): Promoted function-local static to
2664 file-scope static to make it accessible outside init_move_cost.
2665 (init_reg_sets): Do not initialize fixed_regs and call_used_regs yet.
2666 Do not initialize inv_reg_alloc_order yet, either. Do initialize
2667 reg_names since it is needed for parsing command-line options.
2668 (init_move_cost): Use last_mode_for_init_move_cost instead of
2669 function-local static.
2670 (init_reg_sets_1): Initialize fixed_regs, call_used_regs, and
2671 call_really_used_regs now. Reinitialize reg_names. Also
2672 initialize inv_reg_alloc_order. Zero reg_class_subunion and
2673 reg_class_superunion. Clear losing_caller_save_reg_set.
2674 Preserve global registers if called more than once. Reset
2675 move cost, may_move_in_cost, may_move_out_cost, and
2676 last_mode_for_init_move_cost.
2677 (init_reg_modes_target): Renamed from init_reg_modes_once, since it
2678 can now be invoked more than once. Update comments.
2679 (init_regs): Update comments.
2680 (fix_register): Update initial_fixed_regs, initial_call_used_regs,
2681 and initial_call_really_used_regs, instead of the non-initial
2682 variables. This allows us to save the command-line register settings
2683 after target reinitialization.
2684 (init_reg_autoinc): Zero forbidden_inc_dec_classes.
2685 * rtl.h (init_emit_regs): Declare.
2686 (init_reg_modes_target, init_alias_target): Renamed as described
2688 * reload1.c (init_reload): Update comments.
2689 * optabs.c (init_optabs): Likewise.
2690 * cfgloopanal.c (init_set_costs): Explicitly zero target_avail_regs.
2692 2007-08-26 Chao-ying Fu <fu@mips.com>
2694 * rtl.h (XCNMPFV): Preserve const-ness of parameters through use of
2697 2007-08-26 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2699 * alias.c (memory_modified_1, memory_modified_in_insn_p):
2701 * basic-block.h (tree_predicted_by_p, rtl_predicted_by_p):
2703 * bb-reorder.c (copy_bb_p): Likewise.
2704 * cfghooks.c (can_remove_branch_p, can_merge_blocks_p,
2705 predicted_by_p, can_duplicate_block_p, block_ends_with_call_p,
2706 block_ends_with_condjump_p): Likewise.
2707 * cfghooks.h (can_remove_branch_p, can_merge_blocks_p,
2708 predicted_by_p, can_duplicate_block_p, block_ends_with_call_p,
2709 block_ends_with_condjump_p): Likewise.
2710 * cfglayout.c (cfg_layout_can_duplicate_bb_p): Likewise.
2711 * cfgrtl.c (can_delete_note_p, can_delete_label_p,
2712 rtl_can_merge_blocks, cfg_layout_can_merge_blocks_p,
2713 rtl_block_ends_with_call_p, rtl_block_ends_with_condjump_p,
2714 need_fake_edge_p, rtl_can_remove_branch_p): Likewise.
2715 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Use
2716 debug_true_const_tree instad of debug_true_tree.
2717 * debug.c (do_nothing_debug_hooks): Likewise.
2718 * debug.h (ignore_block, debug_true_const_tree): Constify.
2719 * dwarf2out.c (stack_adjust_offset, clobbers_queued_reg_save,
2720 dwarf2out_ignore_block, is_pseudo_reg, is_tagged_type,
2721 decl_ultimate_origin, block_ultimate_origin, lookup_decl_loc,
2722 same_dw_val_p, add_ranges, is_subrange_type, type_is_enum,
2723 dbx_reg_number, is_based_loc, field_type,
2724 simple_type_align_in_bits, simple_decl_align_in_bits,
2725 simple_type_size_in_bits, field_byte_offset, insert_float,
2726 type_tag, member_declared_type, is_redundant_typedef,
2727 secname_for_decl, is_inlined_entry_point): Likewise.
2728 * emit-rtl.c (const_next_insn, const_previous_insn,
2729 const_next_nonnote_insn, const_prev_nonnote_insn,
2730 const_next_real_insn, const_prev_real_insn,
2731 const_next_active_insn, const_prev_active_insn, const_next_label,
2732 const_prev_label): Likewise.
2733 * except.h (lookup_stmt_eh_region_fn, lookup_stmt_eh_region):
2735 * haifa-sched.c (may_trap_exp, haifa_classify_insn,
2736 find_set_reg_weight, no_real_insns_p, find_set_reg_weight,
2737 sched_insn_is_legitimate_for_speculation_p): Likewise.
2738 * jump.c (reversed_comparison_code_parts,
2739 reversed_comparison_code, reversed_comparison, condjump_label):
2741 * predict.c (rtl_predicted_by_p, tree_predicted_by_p): Likewise.
2742 * reg-stack.c (stack_regs_mentioned_p, stack_regs_mentioned):
2744 * rtl.h (const_previous_insn, const_next_insn,
2745 const_prev_nonnote_insn, const_next_nonnote_insn,
2746 const_prev_real_insn, const_next_real_insn,
2747 const_prev_active_insn, const_next_active_insn, const_prev_label,
2748 const_next_label, modified_between_p, modified_in_p,
2749 const_note_storeskeep_with_call_p, condjump_label,
2750 reversed_comparison, reversed_comparison_code,
2751 reversed_comparison_code_parts, memory_modified_in_insn_p,
2752 stack_regs_mentioned): Likewise.
2753 * rtlanal.c (modified_between_p, modified_in_p, const_note_stores,
2754 keep_with_call_p): Likewise.
2755 * sched-deps.c (deps_may_trap_p, sched_get_condition,
2756 conditions_mutex_p, sched_insns_conditions_mutex_psd_next_list,
2757 sd_lists_size, sd_lists_empty_p): Likewise.
2758 * sched-int.h (sched_insns_conditions_mutex_p,
2759 haifa_classify_insn, no_real_insns_p,
2760 sched_insn_is_legitimate_for_speculation_p, sd_next_list,
2761 sd_lists_size, sd_lists_empty_p): Likewise.
2762 * sdbout.c (sdb_debug_hooks): Likewise.
2763 * tree-cfg.c (tree_can_merge_blocks_p, is_ctrl_altering_stmt,
2764 stmt_ends_bb_p, tree_can_remove_branch_p, tree_can_duplicate_bb_p,
2765 -tree_block_ends_with_call_p, tree_block_ends_with_condjump_p):
2767 * tree-eh.c (lookup_stmt_eh_region_fn, lookup_stmt_eh_region,
2768 tree_can_throw_internal): Likewise.
2769 * tree-flow-inline.h (phi_nodes): Likewise.
2770 * tree-flow.h (phi_nodesstmt_ends_bb_p, is_ctrl_altering_stmt,
2771 tree_can_throw_internal, lookup_stmt_eh_region): Likewise.
2772 * tree-gimple.c (const_get_call_expr_in): Likewise.
2773 * tree-gimple.h (const_get_call_expr_in): Likewise.
2774 * tree.c (const_lookup_attribute, attribute_list_equal,
2775 attribute_list_contained): Likewise.
2776 * tree.h (attribute_list_equal, attribute_list_contained,
2777 const_lookup_attribute): Likewise.
2778 * vmsdbgout.c (vmsdbgout_ignore_block): Likewise.
2780 2007-08-26 Bernhard Fischer <aldot@gcc.gnu.org>
2782 * tree-ssa-sccvn.c, tree-ssa-sccvn.h, tree-vn.c,
2783 tree-ssa-pre.c: Remove unnecessary trailing whitespace.
2785 2007-08-26 H.J. Lu <hongjiu.lu@intel.com>
2788 * config/alpha/constraints.md: Delete "svn:mime-type" property
2789 with svn propdel "svn:mime-type".
2791 2007-08-26 H.J. Lu <hongjiu.lu@intel.com>
2794 * ifcvt.c (noce_get_alt_condition): Make sure that the previous
2795 non NOTE insn doesn't cross basic block.
2796 (noce_try_abs): Likewise.
2797 (noce_process_if_block): Likewise.
2799 2007-08-26 David Edelsohn <edelsohn@gnu.org>
2802 * config/rs6000/predicates.md (offsettable_mem_operand): Memory
2803 operand without auto-inc-dec.
2804 * config/rs6000/rs6000.md (floatsidf2_internal): Use
2805 offsettable_mem_operand.
2806 (floatunssidf2_internal): Same.
2807 (fix_truncdfsi2_internal): Same.
2808 (floatsidf_ppc64): Same.
2809 (floatunssidf_ppc64): Same.
2810 (fix_trunctfsi2_internal): Same.
2812 2007-08-26 Bernhard Fischer <aldot@gcc.gnu.org>
2815 * Makefile.in (libbackend.o): Add gcov-iov.h dependency and
2816 pass defines needed for version.c.
2818 2007-08-25 Joseph Myers <joseph@codesourcery.com>
2820 * configure.ac (--debug-prefix-map check): Change assembler
2821 version presumed to have feature to 2.18.
2822 * configure: Regenerate.
2824 2007-08-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2826 * c-common.c (vector_types_convertible_p,
2827 decl_with_nonnull_addr_p, c_promoting_integer_type_p,
2828 self_promoting_args_p): Constify.
2829 * c-common.h (has_c_linkage, decl_with_nonnull_addr_p,
2830 c_promoting_integer_type_p, self_promoting_args_p,
2831 anon_aggr_type_p, vector_types_convertible_p): Likewise.
2832 * c-decl.c (anon_aggr_type_p): Likewise.
2833 * * c-dump.c (dump_stmt): Likewise.
2834 * c-objc-common.c (has_c_linkage): Likewise.
2835 * c-tree.h (same_translation_unit_p): Likewise.
2836 * c-typeck.c (null_pointer_constant_p,
2837 tagged_types_tu_compatible_p, function_types_compatible_p,
2838 type_lists_compatible_p, lvalue_or_else, lvalue_p,
2839 comptypes_internal, struct tagged_tu_seen_cache,
2840 same_translation_unit_p, alloc_tagged_tu_seen_cache,
2841 c_size_in_bytes): Likewise.
2842 * ggc-common.c (compare_ptr_data, hash_descriptor, eq_descriptor,
2843 hash_ptr, eq_ptr): Likewise.
2844 * langhooks-def.h (lhd_decl_ok_for_sibcall,
2845 LANG_HOOKS_OMP_PRIVATIZE_BY_REFERENCE): Likewise.
2846 * langhooks.c (lhd_decl_ok_for_sibcall): Likewise.
2847 * langhooks.h (ok_for_sibcall, omp_privatize_by_reference):
2849 * optabs.c (optab_for_tree_code): Likewise.
2850 * optabs.h (optab_for_tree_code): Likewise.
2851 * rtl.h (simplify_rtx): Likewise.
2852 * simplify-rtx.c (simplify_rtx): Likewise.
2853 * tree-dump.c (queue, queue_and_dump_index, queue_and_dump_type,
2854 dump_flag, dump_node): Likewise.
2855 * tree-dump.h (struct dump_info, dump_stmt, queue_and_dump_index,
2856 queue_and_dump_type, dump_flag): Likewise.
2857 * tree-flow.h (vect_can_force_dr_alignment_p): Likewise.
2858 * tree-pass.h (dump_node): Likewise.
2859 * tree-vectorizer.c (vect_can_force_dr_alignment_p,
2860 supportable_narrowing_operation): Likewise.
2861 * tree-vectorizer.h (vect_can_force_dr_alignment_p,
2862 supportable_narrowing_operation): Likewise.
2863 * tree-vrp.c (needs_overflow_infinity, supports_overflow_infinity,
2864 is_negative_overflow_infinity, is_positive_overflow_infinity,
2865 is_overflow_infinity, vrp_val_is_max, vrp_val_is_min,
2866 nonnull_arg_p, get_value_range, vrp_operand_equal_p,
2867 update_value_range, add_equivalence, ssa_name_nonnegative_p,
2868 ssa_name_nonzero_p, fp_predicate): Likewise.
2869 * tree.c (auto_var_in_fn_p, empty_body_p): Likewise.
2870 * tree.h (empty_body_p, auto_var_in_fn_p, ssa_name_nonzero_p,
2871 ssa_name_nonnegative_p): Likewise.
2873 2007-08-25 Hans Kester <hans.kester@ellips.nl>
2875 * config.gcc : Add x86_64-elf target.
2877 2007-08-25 Jakub Jelinek <jakub@redhat.com>
2879 * expr.c (store_expr): Fix order of store_by_pieces arguments.
2881 2007-08-24 Sandra Loosemore <sandra@codesourcery.com>
2882 Nigel Stephens <nigel@mips.com>
2886 * doc/tm.texi (SET_RATIO, SET_BY_PIECES_P): Document new macros.
2887 (STORE_BY_PIECES_P): No longer applies to __builtin_memset.
2888 * expr.c (SET_BY_PIECES_P): Define.
2889 (can_store_by_pieces, store_by_pieces): Add MEMSETP argument; use
2890 it to decide whether to use SET_BY_PIECES_P or STORE_BY_PIECES_P.
2891 (store_expr): Pass MEMSETP argument to can_store_by_pieces and
2893 * expr.h (SET_RATIO): Define.
2894 (can_store_by_pieces, store_by_pieces): Update prototypes.
2895 * builtins.c (expand_builtin_memcpy): Pass MEMSETP argument to
2896 can_store_by_pieces/store_by_pieces.
2897 (expand_builtin_memcpy_args): Likewise.
2898 (expand_builtin_strncpy): Likewise.
2899 (expand_builtin_memset_args): Likewise. Also remove special case
2900 for optimize_size so that can_store_by_pieces/SET_BY_PIECES_P can
2901 decide what to do instead.
2902 * value-prof.c (tree_stringops_transform): Pass MEMSETP argument
2903 to can_store_by_pieces.
2905 * config/sh/sh.h (SET_BY_PIECES_P): Clone from STORE_BY_PIECES_P.
2906 * config/s390/s390.h (SET_BY_PIECES_P): Likewise.
2908 * config/mips/mips.opt (mmemcpy): Change from Var to Mask.
2909 * config/mips/mips.c (override_options): Make -Os default to -mmemcpy.
2910 * config/mips/mips.h (MIPS_CALL_RATIO): Define.
2911 (MOVE_RATIO, CLEAR_RATIO, SET_RATIO): Define.
2912 (STORE_BY_PIECES_P): Define.
2914 2007-08-24 Tom Tromey <tromey@redhat.com>
2916 * varpool.c (varpool_last_needed_node): Fix comment typo.
2917 * c-decl.c (duplicate_decls): Fix comment typo.
2918 (clone_underlying_type): Update comment.
2920 2007-08-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2922 * config/alpha/alpha.c (alpha_mangle_type, decl_has_samegp,
2923 alpha_in_small_data_p, alpha_split_complex_arg,
2924 alpha_stdarg_optimize_hook, TARGET_ASM_CAN_OUTPUT_MI_THUNK):
2926 * config/arm/arm-protos.h (arm_mangle_type): Likewise.
2927 * config/arm/arm.c (arm_comp_type_attributes, arm_mangle_type):
2929 * config/bfin/bfin.c (funkind, bfin_comp_type_attributes,
2930 TARGET_ASM_CAN_OUTPUT_MI_THUNK): Likewise.
2931 * config/darwin-protos.h (darwin_binds_local_p): Likewise.
2932 * config/darwin.c (darwin_binds_local_p): Likewise.
2933 * config/frv/frv.c (frv_string_begins_with, frv_in_small_data_p):
2935 * config/i386/i386-protos.h (i386_pe_binds_local_p,
2936 i386_pe_valid_dllimport_attribute_p): Likewise.
2937 * config/i386/i386.c (ix86_function_regparm,
2938 ix86_comp_type_attributes, ix86_ms_bitfield_layout_p,
2939 x86_can_output_mi_thunk, ix86_mangle_type,
2940 TARGET_USE_BLOCKS_FOR_CONSTANT_P): Likewise.
2941 * config/i386/winnt.c (i386_pe_valid_dllimport_attribute_p,
2942 i386_pe_binds_local_p): Likewise.
2943 * config/ia64/ia64.c
2944 (ia64_first_cycle_multipass_dfa_lookahead_guard_spec,
2945 ia64_needs_block_p, ia64_in_small_data_p, ia64_mangle_type,
2946 ia64_invalid_conversion, ia64_invalid_unary_op,
2947 ia64_invalid_binary_op, TARGET_ASM_CAN_OUTPUT_MI_THUNK):
2949 * config/m32c/m32c.c (m32c_comp_type_attributes): Likewise.
2950 * config/m32r/m32r.c (m32r_in_small_data_p): Likewise.
2951 * config/m68k/m68k.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Likewise.
2952 * config/mips/mips.c (mips_use_blocks_for_constant_p,
2953 mips_in_small_data_p, mips_use_anchors_for_symbol_p,
2954 mips_comp_type_attributes, TARGET_ASM_CAN_OUTPUT_MI_THUNK,
2955 mips_near_type_p, mips_far_type_p, mips_global_symbol_p,
2956 mips_symbol_binds_local_p, mips_classify_symbol): Likewise.
2957 * config/pa/pa.c (pa_commutative_p): Likewise.
2958 * config/rs6000/rs6000-protos.h (rs6000_elf_in_small_data_p):
2960 * config/rs6000/rs6000.c (rs6000_invalid_within_doloop,
2961 rs6000_ms_bitfield_layout_p, rs6000_mangle_type,
2962 rs6000_use_blocks_for_constant_p,
2963 rs6000_vector_alignment_reachable, rs6000_is_opaque_type,
2964 invalid_arg_for_unprototyped_fn, TARGET_ASM_CAN_OUTPUT_MI_THUNK,
2965 TARGET_SPLIT_COMPLEX_ARG, rs6000_elf_in_small_data_p): Likewise.
2966 * config/s390/s390.c (s390_mangle_type,
2967 TARGET_ASM_CAN_OUTPUT_MI_THUNK, TARGET_INVALID_WITHIN_DOLOOP):
2969 * config/score/score.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK,
2970 th_in_small_data_p): Likewise.
2971 * config/sh/sh.c (sh_ms_bitfield_layout_p,
2972 sh_dwarf_calling_convention, TARGET_ASM_CAN_OUTPUT_MI_THUNK):
2974 * config/sparc/sparc.c (sparc_can_output_mi_thunk,
2975 sparc_mangle_type): Likewise.
2976 * config/spu/spu.c (spu_vector_alignment_reachable): Likewise.
2977 * config/xtensa/xtensa.c (TARGET_SPLIT_COMPLEX_ARG): Likewise.
2978 * emit-rtl.c (const_fixed_htab_hash, const_fixed_htab_eq):
2980 * hooks.c (hook_bool_mode_const_rtx_false,
2981 hook_bool_mode_const_rtx_true,
2982 hook_bool_const_tree_hwi_hwi_const_tree_false,
2983 hook_bool_const_tree_hwi_hwi_const_tree_true,
2984 hook_int_const_tree_0, hook_int_const_tree_const_tree_1,
2985 hook_constcharptr_const_tree_null,
2986 hook_constcharptr_const_rtx_null,
2987 hook_constcharptr_const_tree_const_tree_null,
2988 hook_constcharptr_int_const_tree_null,
2989 hook_constcharptr_int_const_tree_const_tree_null): New.
2990 (hook_bool_mode_rtx_false, hook_bool_mode_rtx_true,
2991 hook_bool_tree_hwi_hwi_tree_false,
2992 hook_bool_tree_hwi_hwi_tree_true, hook_int_tree_0,
2993 hook_int_tree_tree_1, hook_constcharptr_tree_null,
2994 hook_constcharptr_rtx_null, hook_constcharptr_tree_tree_null,
2995 hook_constcharptr_int_tree_null,
2996 hook_constcharptr_int_tree_tree_null): Delete.
2997 (default_can_output_mi_thunk_no_vcall): Constify.
2998 * hooks.h (hook_bool_mode_rtx_false, hook_bool_mode_rtx_true,
2999 hook_bool_tree_hwi_hwi_tree_false,
3000 hook_bool_tree_hwi_hwi_tree_true, hook_int_tree_0,
3001 hook_int_tree_tree_1, hook_constcharptr_tree_null,
3002 hook_constcharptr_rtx_null, hook_constcharptr_tree_tree_null,
3003 hook_constcharptr_int_tree_null,
3004 hook_constcharptr_int_tree_tree_null): Delete.
3005 (hook_bool_mode_const_rtx_false, hook_bool_mode_const_rtx_true,
3006 hook_bool_const_tree_hwi_hwi_const_tree_false,
3007 hook_bool_const_tree_hwi_hwi_const_tree_true,
3008 hook_int_const_tree_0, hook_int_const_tree_const_tree_1,
3009 hook_constcharptr_const_tree_null,
3010 hook_constcharptr_const_rtx_null,
3011 hook_constcharptr_const_tree_const_tree_null,
3012 hook_constcharptr_int_const_tree_null,
3013 hook_constcharptr_int_const_tree_const_tree_null): New.
3014 (default_can_output_mi_thunk_no_vcall): Constify.
3015 * integrate.c (function_attribute_inlinable_p): Likewise.
3016 * integrate.h (function_attribute_inlinable_p): Likewise.
3017 * jump.c (rtx_renumbered_equal_p): Likewise.
3018 * output.h (decl_readonly_section, categorize_decl_for_section,
3019 default_use_anchors_for_symbol_p, default_binds_local_p,
3020 default_binds_local_p_1): Likewise.
3021 * rtl.h (rtx_renumbered_equal_p, decl_default_tls_model):
3023 * target-def.h (TARGET_ASM_CAN_OUTPUT_MI_THUNK,
3024 TARGET_VALID_DLLIMPORT_ATTRIBUTE_P, TARGET_VECTOR_OPAQUE_P,
3025 TARGET_COMMUTATIVE_P, TARGET_USE_BLOCKS_FOR_CONSTANT_P,
3026 TARGET_COMP_TYPE_ATTRIBUTES,
3027 TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P,
3028 TARGET_MS_BITFIELD_LAYOUT_P, TARGET_MANGLE_TYPE,
3029 TARGET_IN_SMALL_DATA_P, TARGET_INVALID_CONVERSION,
3030 TARGET_INVALID_UNARY_OP, TARGET_INVALID_BINARY_OP,
3031 TARGET_DWARF_CALLING_CONVENTION): Likewise.
3032 * target.h (can_output_mi_thunk, needs_block_p,
3033 first_cycle_multipass_dfa_lookahead_guard_spec,
3034 vector_alignment_reachable, comp_type_attributes,
3035 function_attribute_inlinable_p, ms_bitfield_layout_p, mangle_type,
3036 commutative_p, use_blocks_for_constant_p,
3037 use_anchors_for_symbol_p, in_small_data_p, binds_local_p,
3038 vector_opaque_p, dwarf_calling_convention, stdarg_optimize_hook,
3039 invalid_within_doloop, valid_dllimport_attribute_p,
3040 split_complex_arg, invalid_arg_for_unprototyped_fn,
3041 invalid_conversion, invalid_unary_op, invalid_binary_op):
3043 * targhooks.c (default_invalid_within_doloop,
3044 hook_invalid_arg_for_unprototyped_fn,
3045 default_builtin_vector_alignment_reachable): Likewise.
3046 (hook_bool_rtx_commutative_p): Delete.
3047 (hook_bool_const_rtx_commutative_p): New.
3048 * targhooks.h (default_invalid_within_doloop,
3049 default_builtin_vector_alignment_reachable,
3050 hook_invalid_arg_for_unprototyped_fn): Constify.
3051 (hook_bool_rtx_commutative_p): Delete.
3052 (hook_bool_const_rtx_commutative_p): New.
3053 * varasm.c (bss_initializer_p, decl_default_tls_model,
3054 categorize_decl_for_section, decl_readonly_section,
3055 default_use_anchors_for_symbol_p, default_binds_local_p,
3056 default_binds_local_p_1): Constify.
3058 2007-08-24 Jie Zhang <jie.zhang@analog.com>
3060 * config/bfin/bfin.h (TARGET_CPU_CPP_BUILTINS): Define
3061 __NO_BUILTIN if -fno-builtin.
3063 2007-08-24 Jie Zhang <jie.zhang@analog.com>
3065 * config/bfin/bfin.c (print_operand): Report error instead of
3066 ICE for wrong operand.
3068 2007-08-24 Michael Matz <matz@suse.de>
3070 * Makefile.in (GTFILES_H): Use $(patsubst) instead of $(subst).
3072 2007-08-24 Richard Guenther <rguenther@suse.de>
3075 * tree-ssa.c (useless_type_conversion_p): Split into a
3076 recursive and a non-recursive part.
3077 (useless_type_conversion_p_1): New function.
3078 * tree-ssa-ccp.c (fold_stmt_r): Make sure that the result
3079 from maybe_fold_offset_to_reference is trivially convertible
3080 to the desired type before doing the simplification.
3082 2007-08-24 Jakub Jelinek <jakub@redhat.com>
3084 * expr.c (store_expr): Optimize initialization of an array
3086 * expr.h (builtin_strncpy_read_str): New prototype.
3087 * builtins.c (builtin_strncpy_read_str): Remove prototype.
3090 2007-08-24 Uros Bizjak <ubizjak@gmail.com>
3093 * ifcvt.c (noce_find_if_block): Do not clear if_info.cond_earliest
3094 field when initializing if_info structure.
3096 2007-08-24 Richard Guenther <rguenther@suse.de>
3098 * Makefile.in (tree-inline.o): Add $(TARGET_H) and $(INTEGRATE_H)
3100 * c-objc-common.c (c_cannot_inline_tree_fn): Remove.
3101 * langhooks.c (lhd_tree_inlining_cannot_inline_tree_fn): Likewise.
3102 * tree-inline.c (inlinable_function_p): Fold in common parts of
3103 the cannot_inline_tree_fn langhook.
3104 * langhooks-def.h (lhd_tree_inlining_cannot_inline_tree_fn): Remove.
3105 (LANG_HOOKS_TREE_INLINING_CANNOT_INLINE_TREE_FN): Likewise.
3106 (LANG_HOOKS_TREE_INLINING_INITIALIZER): Remove initializer for
3107 cannot_inline_tree_fn langhook.
3108 * langhooks.h (struct lang_hooks_for_tree_inlining): Remove
3109 cannot_inline_tree_fn member.
3111 2007-08-24 Richard Guenther <rguenther@suse.de>
3113 * expr.c (get_inner_reference): Do computation of bitoffset
3114 from offset in a way we can detect overflow reliably.
3116 2007-08-24 Jie Zhang <jie.zhang@analog.com>
3118 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define for uClibc.
3120 2007-08-24 Jie Zhang <jie.zhang@analog.com>
3122 * config/bfin/bfin.opt (mfast-fp): Add.
3123 * config/bfin/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Let
3124 libbffastfp override libgcc if -mfast-fp.
3125 * config/bfin/bfin.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
3126 * doc/invoke.texi (Option Summary): Mention -mfast-fp.
3127 (Blackfin Options): Document -mfast-fp.
3129 2007-08-24 Jakub Jelinek <jakub@redhat.com>
3131 PR tree-optimization/32573
3133 * tree-data-ref.c (initialize_data_dependence_relation): Clear
3134 DDR_SUBSCRIPTS, DDR_DIR_VECTS and DDR_DIST_VECTS at the beginning.
3135 (finalize_ddr_dependent): Clear DDR_SUBSCRIPTS after freeing it.
3136 (build_classic_dist_vector): Return false rather than true if
3137 DDR_ARE_DEPENDENT is non-NULL at the beginning. Return false
3138 if either subscript_dependence_tester_1 or build_classic_dist_vector_1
3139 returned false. Don't call save_dist_v before calling
3140 build_classic_dist_vector_1.
3141 (free_dependence_relation): Don't guard freeing DDR_SUBSCRIPTS
3142 with NULL DDR_ARE_DEPENDENT. Also free DDR_DIST_VECTS and/or
3143 DDR_DIR_VECTS vectors.
3145 2007-08-23 Brian Sidebotham <brian.sidebotham@gmail.com>
3147 * configure.ac (leb128): Modify sed statement to work with any
3148 binutils version string.
3149 * configure: Regenerate
3151 2007-08-23 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
3153 * config/arc/arc-protos.h arc_select_cc_mode, gen_compare_reg):
3154 Wrap in RTX_CODE macro guard.
3155 * config/arm/pe.c (arm_dllexport_p, arm_dllimport_p,
3156 arm_dllexport_name_p, arm_dllimport_name_p, arm_mark_dllexport,
3157 arm_mark_dllimport, arm_pe_encode_section_info,
3158 arm_pe_unique_section): Use ISO-C function declarations.
3159 * config/c4x/c4x-c.c (c4x_parse_pragma, c4x_pr_CODE_SECTION,
3160 c4x_pr_DATA_SECTION, c4x_pr_FUNC_IS_PURE,
3161 c4x_pr_FUNC_NEVER_RETURNS, c4x_pr_INTERRUPT, c4x_pr_ignored):
3163 * config/iq2000/iq2000.h (ASM_OUTPUT_SKIP): Fix format warning.
3164 * config/m68hc11/m68hc11.h (ASM_OUTPUT_EXTERNAL): Undef before
3166 * config/mips/mips.h (ASM_DECLARE_OBJECT_NAME): Fix format
3168 * config/mn10300/mn10300.h (OUTPUT_ADDR_CONST_EXTRA): Likewise.
3169 * config/pdp11/pdp11.c (pdp11_output_function_epilogue): Likewise.
3170 (register_move_cost): Use ISO-C function declarations.
3171 * config/pdp11/pdp11.h (PRINT_OPERAND): Fix format warnings.
3172 * config/score/score-protos.h (score_declare_object): Add
3174 * config/score/score.h (ASM_DECLARE_OBJECT_NAME): Fix format
3176 * final.c (profile_function): Avoid empty if-bodies.
3179 * calls.c (must_pass_in_stack_var_size,
3180 must_pass_in_stack_var_size_or_pad): Constify.
3181 * config/alpha/alpha-protos.h (function_value): Likewise.
3182 * config/alpha/alpha.c (alpha_return_in_memory,
3183 alpha_pass_by_reference, function_value,
3184 unicosmk_must_pass_in_stack, TARGET_PROMOTE_FUNCTION_ARGS,
3185 TARGET_PROMOTE_FUNCTION_RETURN, TARGET_PROMOTE_PROTOTYPES):
3187 * config/arc/arc.c (arc_return_in_memory, arc_pass_by_reference,
3188 TARGET_PROMOTE_FUNCTION_ARGS, TARGET_PROMOTE_FUNCTION_RETURN,
3189 TARGET_PROMOTE_PROTOTYPES): Likewise.
3190 * config/arm/arm-protos.h (arm_return_in_memory,
3191 arm_pad_arg_upward, arm_function_value): Likewise.
3192 * config/arm/arm.c (arm_pass_by_reference,
3193 arm_promote_prototypes, arm_return_in_msb, arm_must_pass_in_stack,
3194 TARGET_PROMOTE_FUNCTION_ARGS, TARGET_PROMOTE_FUNCTION_RETURN,
3195 arm_function_value, arm_return_in_memory, arm_pad_arg_upward):
3197 * config/arm/arm.h (ARM_DECLARE_FUNCTION_NAME): Likewise.
3198 * config/avr/avr-protos.h (avr_function_value): Likewise.
3199 * config/avr/avr.c (avr_return_in_memory,
3200 gas_output_limited_string, gas_output_ascii, avr_function_value,
3201 avr_return_in_memory): Likewise.
3202 * config/bfin/bfin-protos.h (bfin_return_in_memory): Likewise.
3203 * config/bfin/bfin.c (bfin_pass_by_reference,
3204 bfin_return_in_memory, TARGET_PROMOTE_PROTOTYPES,
3205 TARGET_PROMOTE_FUNCTION_ARGS, TARGET_PROMOTE_FUNCTION_RETURN):
3207 * config/cris/cris.c (cris_pass_by_reference,
3208 TARGET_PROMOTE_FUNCTION_ARGS): Likewise.
3209 * config/crx/crx.c (crx_return_in_memory): Likewise.
3210 * config/darwin.c (function_base, machopic_function_base_name):
3212 * config/fr30/fr30.c (fr30_must_pass_in_stack,
3213 TARGET_PROMOTE_PROTOTYPES): Likewise.
3214 * config/frv/frv.c (frv_must_pass_in_stack): Likewise.
3215 * config/h8300/h8300.c (h8300_return_in_memory): Likewise.
3216 * config/i386/i386-protos.h (ix86_return_in_memory,
3217 ix86_sol10_return_in_memory): Likewise.
3218 * config/i386/i386.c (ix86_function_value,
3219 ix86_function_sseregparm, ix86_must_pass_in_stack,
3220 type_natural_mode, classify_argument, examine_argument,
3221 construct_container, ix86_pass_by_reference, function_value_32,
3222 function_value_64, ix86_function_value_1, return_in_memory_32,
3223 return_in_memory_64, return_in_memory_ms_64,
3224 ix86_return_in_memory, ix86_sol10_return_in_memory,
3225 TARGET_PROMOTE_PROTOTYPES): Likewise.
3226 * config/ia64/ia64-protos.h (ia64_function_value,
3227 ia64_hpux_function_arg_padding): Likewise.
3228 * config/ia64/ia64.c (hfa_element_mode, ia64_return_in_memory,
3229 ia64_function_value, bundle_state_hash, bundle_state_eq_p,
3230 ia64_hpux_function_arg_padding): Likewise.
3231 * config/iq2000/iq2000-protos.h (function_arg,
3232 iq2000_function_value): Likewise.
3233 * config/iq2000/iq2000.c (iq2000_return_in_memory,
3234 iq2000_pass_by_reference, TARGET_PROMOTE_FUNCTION_ARGS,
3235 TARGET_PROMOTE_FUNCTION_RETURN, TARGET_PROMOTE_PROTOTYPES,
3236 function_arg, iq2000_function_value): Likewise.
3237 * config/m32c/m32c-protos.h (m32c_function_value,
3238 m32c_promote_function_return): Likewise.
3239 * config/m32c/m32c.c (m32c_pass_by_reference,
3240 m32c_promote_prototypes, m32c_promote_function_return,
3241 m32c_function_value): Likewise.
3242 * config/m32r/m32r.c (m32r_return_in_memory,
3243 m32r_pass_by_reference, TARGET_PROMOTE_PROTOTYPES,
3244 m32r_in_small_data_p): Likewise.
3245 * config/m68hc11/m68hc11-protos.h (m68hc11_function_arg_padding):
3247 * config/m68hc11/m68hc11.c (m68hc11_return_in_memory,
3248 m68hc11_function_arg_padding): Likewise.
3249 * config/m68k/m68k-protos.h (m68k_function_value): Likewise.
3250 * config/m68k/m68k.c (TARGET_PROMOTE_PROTOTYPES,
3251 m68k_function_value): Likewise.
3252 * config/mcore/mcore-protos.h (mcore_num_arg_regs,
3253 mcore_function_value): Likewise.
3254 * config/mcore/mcore.c (handle_structs_in_regs,
3255 mcore_return_in_memory, TARGET_PROMOTE_FUNCTION_ARGS,
3256 TARGET_PROMOTE_FUNCTION_RETURN, TARGET_PROMOTE_PROTOTYPES,
3257 mcore_num_arg_regs, mcore_function_value): Likewise.
3258 * config/mips/mips-protos.h (mips_pad_arg_upward,
3259 mips_function_value): Likewise.
3260 * config/mips/mips.c (mips_fpr_return_fields, mips_return_in_msb,
3261 mips_return_in_memory, mips_pass_by_reference, mips_callee_copies,
3262 TARGET_PROMOTE_FUNCTION_ARGS, TARGET_PROMOTE_FUNCTION_RETURN,
3263 TARGET_PROMOTE_PROTOTYPES, mips_pad_arg_upward,
3264 mips_function_value): Likewise.
3265 * config/mmix/mmix-protos.h (mmix_function_outgoing_value):
3267 * config/mmix/mmix.c (mmix_pass_by_reference,
3268 TARGET_PROMOTE_FUNCTION_ARGS, mmix_function_outgoing_value,
3269 mmix_encode_section_info): Likewise.
3270 * config/mn10300/mn10300-protos.h (mn10300_function_value):
3272 * config/mn10300/mn10300.c (mn10300_return_in_memory,
3273 mn10300_pass_by_reference, TARGET_PROMOTE_PROTOTYPES,
3274 mn10300_function_value): Likewise.
3275 * config/mt/mt-protos.h (mt_function_value): Likewise.
3276 * config/mt/mt.c (mt_pass_by_reference, mt_function_value,
3277 mt_pass_in_stack, TARGET_PROMOTE_PROTOTYPES): Likewise.
3278 * config/pa/pa-protos.h (function_arg_padding, function_value,
3279 pa_return_in_memory): Likewise.
3280 * config/pa/pa.c (pa_pass_by_reference,
3281 TARGET_PROMOTE_FUNCTION_RETURN, TARGET_PROMOTE_PROTOTYPES,
3282 function_arg_padding, function_value, pa_return_in_memory):
3284 * config/pdp11/pdp11.c (pdp11_return_in_memory): Likewise.
3285 * config/rs6000/rs6000-protos.h (rs6000_function_value,
3286 function_arg_padding): Likewise.
3287 * config/rs6000/rs6000.c (rs6000_return_in_memory,
3288 rs6000_darwin64_record_arg_recurse, rs6000_darwin64_record_arg,
3289 rs6000_pass_by_reference, rs6000_must_pass_in_stack,
3290 TARGET_PROMOTE_FUNCTION_ARGS, TARGET_PROMOTE_FUNCTION_RETURN,
3291 function_arg_padding, altivec_expand_dst_builtin,
3292 altivec_expand_builtin, rs6000_expand_builtin, spe_init_builtins,
3293 altivec_init_builtins, rs6000_common_init_builtins,
3294 rs6000_function_value): Likewise.
3295 * s390/s390-protos.h (s390_function_value): Likewise.
3296 * config/s390/s390.c (s390_function_arg_size,
3297 s390_pass_by_reference, s390_return_in_memory,
3298 s390_function_value, TARGET_PROMOTE_FUNCTION_ARGS,
3299 TARGET_PROMOTE_FUNCTION_RETURN): Likewise.
3300 * config/score/score-protos.h (score_function_value): Likewise.
3301 * config/score/score.c (score_arg_partial_bytes,
3302 TARGET_PROMOTE_FUNCTION_ARGS, TARGET_PROMOTE_FUNCTION_RETURN,
3303 TARGET_PROMOTE_PROTOTYPES, score_return_in_memory,
3304 score_pass_by_reference, score_add_offset, score_function_value):
3306 * config/sh/sh-protos.h (sh_attr_renesas_p,
3307 sh_promote_prototypes): Likewise.
3308 * config/sh/sh.c (sh_return_in_memory, sh_pass_by_reference,
3309 sh_callee_copies, sh_promote_prototypes, shcompact_byref,
3310 sh_attr_renesas_p): Likewise.
3311 * config/sparc/sparc-protos.h (function_value,
3312 function_arg_padding): Likewise.
3313 * config/sparc/sparc.c (sparc_promote_prototypes,
3314 sparc_return_in_memory, sparc_pass_by_reference,
3315 TARGET_PROMOTE_FUNCTION_ARGS, TARGET_PROMOTE_FUNCTION_RETURN,
3316 function_arg_record_value_2, function_arg_record_value_1,
3317 function_arg_record_value, function_arg_record_value,
3318 function_arg_padding, function_value): Likewise.
3319 * config/spu/spu-protos.h (spu_function_value): Likewise.
3320 * config/spu/spu.c (spu_pass_by_reference, spu_return_in_memory,
3321 spu_function_value): Likewise.
3322 * config/stormy16/stormy16-protos.h (xstormy16_function_value):
3324 * config/stormy16/stormy16.c (xstormy16_return_in_memory,
3325 xstormy16_function_value, TARGET_PROMOTE_FUNCTION_ARGS,
3326 TARGET_PROMOTE_FUNCTION_RETURN, TARGET_PROMOTE_PROTOTYPES):
3328 * config/v850/v850.c (v850_return_in_memory,
3329 v850_pass_by_reference, TARGET_PROMOTE_PROTOTYPES): Likewise.
3330 * config/vax/vax.c (TARGET_PROMOTE_PROTOTYPES): Likewise.
3331 * config/xtensa/xtensa.c (xtensa_return_in_msb,
3332 xtensa_return_in_memory, TARGET_PROMOTE_FUNCTION_ARGS,
3333 TARGET_PROMOTE_FUNCTION_RETURN, TARGET_PROMOTE_PROTOTYPES):
3335 * explow.c (promote_mode, hard_function_value): Likewise.
3336 * expr.h (hard_function_value, promote_mode): Likewise.
3337 * function.c (aggregate_value_p): Likewise.
3338 * hooks.c (hook_bool_const_tree_true): New.
3339 * hooks.h (hook_bool_const_tree_true): New.
3340 * sdbout.c (SET_KNOWN_TYPE_TAG, plain_type_1): Constify.
3341 * target-def.h (TARGET_PROMOTE_FUNCTION_ARGS,
3342 TARGET_PROMOTE_FUNCTION_RETURN, TARGET_PROMOTE_PROTOTYPES,
3343 TARGET_RETURN_IN_MSB): Likewise.
3344 * target.h (promote_function_args, promote_function_return,
3345 promote_prototypes, return_in_memory, return_in_msb,
3346 pass_by_reference, must_pass_in_stack, callee_copies,
3347 function_value): Likewise.
3348 * targhooks.c (default_return_in_memory,
3349 hook_pass_by_reference_must_pass_in_stack,
3350 hook_callee_copies_named,
3351 hook_bool_CUMULATIVE_ARGS_mode_tree_bool_false,
3352 hook_bool_CUMULATIVE_ARGS_mode_tree_bool_true,
3353 default_function_value): Likewise.
3354 * targhooks.h (default_return_in_memory,
3355 hook_pass_by_reference_must_pass_in_stack,
3356 hook_callee_copies_named,
3357 hook_bool_CUMULATIVE_ARGS_mode_tree_bool_false,
3358 hook_bool_CUMULATIVE_ARGS_mode_tree_bool_true,
3359 default_function_value): Likewise.
3360 * tree-ssa-structalias.c (const_equiv_class_label_t): New.
3361 (equiv_class_label_hash, equiv_class_label_eq): Constify.
3362 * tree-vectorizer.c (bb_in_loop_p): Likewise.
3363 * tree.c (needs_to_live_in_memory): Likewise.
3364 * tree.h (struct tree_type, needs_to_live_in_memory,
3365 aggregate_value_p, must_pass_in_stack_var_size,
3366 must_pass_in_stack_var_size_or_pad): Likewise.
3367 * vmsdbgout.c (write_debug_addr, write_debug_delta4,
3368 write_debug_string, ASM_OUTPUT_DEBUG_STRING, write_rtnbeg,
3369 lookup_filename): Likewise.
3371 2007-08-23 Uros Bizjak <ubizjak@gmail.com>
3374 * config/i386/i386.c (ix86_expand_fp_compare): Expand fp comparison to
3375 fake fcomi i387 instruction for !TARGET_CMOVE.
3376 (ix86_expand_branch): Expand natural sequence with one jump for
3377 all targets, not only TARGET_CMOVE.
3378 * config/i386/i386.md (*cmpfp_0_cc): New define_insn_and_split
3379 pattern to implement fake fcomi sequence. Split instruction after
3380 reload to correct compare sequences.
3381 (*cmpfp_xf_cc): Ditto.
3382 (*cmpfp_<mode>_cc): Ditto.
3383 (*cmpfp_u_cc): Ditto.
3384 (*cmpfp_<mode>_cc): Ditto.
3386 2007-08-23 Richard Guenther <rguenther@suse.de>
3388 * tree-pretty-print.c (dump_generic_node): Annotate
3389 GIMPLE_MODIFY_STMTs with volatile ops with "{v}".
3391 2007-08-23 Richard Guenther <rguenther@suse.de>
3393 * builtins.c (expand_builtin_mathfn): Wrap argument in
3394 save_expr directly instead of re-building the call.
3395 (expand_builtin_mathfn_2): Likewise.
3396 (expand_builtin_mathfn_3): Likewise.
3397 (expand_builtin_interclass_mathfn): Likewise.
3398 (expand_builtin_int_roundingfn): Set arg.
3399 (expand_builtin_int_roundingfn_2): Likewise.
3401 2007-08-23 Paolo Bonzini <bonzini@gnu.org>
3403 * config/i386/sse.md (*sse_and<mode>3, *sse_ior<mode>3,
3404 *sse_nand<mode>3, *sse_xor<mode>3): New.
3406 2007-08-23 Uros Bizjak <ubizjak@gmail.com>
3408 * config/i386/i386.h (PRINT_OPERAND_PUNCT_VALID_P): Add ';' code.
3409 * config/i386/i386.c (print_operand): Handle ';' code. Output
3410 semicolon for TARGET_MACHO.
3411 * config/i386/sync.md (*sync_compare_and_swap<mode>): Use '%;' to
3412 emit semicolon after 'lock' prefix.
3413 (sync_double_compare_and_swap<mode>): Ditto.
3414 (*sync_double_compare_and_swapdi_pic): Ditto.
3415 (*sync_compare_and_swap_cc<mode>): Ditto.
3416 (sync_double_compare_and_swap_cc<mode>): Ditto.
3417 (*sync_double_compare_and_swap_ccdi_pic): Ditto.
3418 (sync_old_add<mode>): Ditto.
3419 (sync_add<mode>): Ditto.
3420 (sync_sub<mode>): Ditto.
3421 (sync_ior<mode>): Ditto.
3422 (sync_and<mode>): Ditto.
3423 (sync_xor<mode>): Ditto.
3425 2007-08-22 Chao-ying Fu <fu@mips.com>
3427 * rtl.c (rtx_code_size): Check CONST_FIXED to calcualte correct sizes
3429 (copy_rtx): Handle CONST_FIXED.
3430 (rtx_equal_p): Likewise.
3431 * rtl.h (fixed_value.h): New include.
3432 (rtx_def): Add a new field of fixed_value to u.
3433 (XCNMPFV): Define for accessing fixed_value.
3434 (CONST_FIXED_VALUE, CONST_FIXED_VALUE_HIGH, CONST_FIXED_VALUE_LOW):
3436 * rtl.def (CONST_FIXED): New constant.
3437 (SS_MULT, US_MULT, SS_DIV, US_DIV, FRACT_CONVERT,
3438 UNSIGNED_FRACT_CONVERT, SAT_FRACT, UNSIGNED_SAT_FRACT, US_NEG,
3439 US_ASHIFT): New codes.
3440 * doc/rtl.texi (Expressions): Document const_fixed, us_neg, ss_mult,
3441 us_mult, ss_div, us_div, us_ashift, fract_convert, sat_fract,
3442 unsigned_fract_convert, unsigned_sat_fract): Document them.
3443 * varasm.c (assemble_integer): Extend to support fixed-point constants
3444 by using different machine classes.
3445 (decode_addr_const): Handle FIXED_CST.
3446 (const_hash_1): Likewise.
3447 (compare_constant): Likewise.
3448 (copy_constant): Likewise.
3449 (const_rtx_hash_1): Handle CONST_FIXED.
3450 (output_constant_pool_2): Handle MODE_FRACT, MODE_UFRACT, MODE_ACCUM,
3451 MODE_UACCUM, MODE_VECTOR_FRACT, MODE_VECTOR_UFRACT, MODE_VECTOR_ACCUM,
3453 (initializer_constant_valid_p): Handle FIXED_CST.
3454 (output_constant): Support FIXED_POINT_TYPE.
3455 * gengenrtl.c (excluded_rtx): Check CONST_FIXED to exclude.
3456 * cse.c (hash_rtx): Support CONST_FIXED.
3457 (exp_equiv_p): Likewise.
3458 (cannon_reg): Likewise.
3459 (fold_rtx): Likewise.
3460 (equiv_constant): Likewise.
3461 (cse_process_notes_1): Likewise.
3462 (count_reg_usage): Likewise.
3463 * cselib.c (entry_and_rtx_equal_p): Check CONST_FIXED.
3464 (rtx_equal_for_cselib_p): Handle CONST_FIXED.
3465 (wrap_constant): Check CONST_FIXED.
3466 (cselib_hash_rtx): Support CONST_FIXED.
3467 (cselib_subst_to_values): Likewise.
3468 * df-scan.c (df_uses_record): Likewise.
3469 * gcse.c (want_to_gcse_p): Likewise.
3470 (oprs_unchanged_p): Likewise.
3471 (oprs_not_set_p): Likewise.
3472 (compute_transp): Likewise.
3473 (extract_mentioned_regs_helper): Likewise.
3474 * genemit.c (gen_exp): Likewise.
3475 * local-alloc.c (equiv_init_varies_p): Likewise.
3476 (contains_replace_regs): Likewise.
3477 (memref_referenced_p): Likewise.
3478 * loop-invariant.c (check_maybe_invariant): Likewise.
3479 (hash_invariant_expr_1): Likewise.
3480 (invariant_expr_equal_p): Likewise.
3481 * postreload-gcse.c (oprs_unchanged_p): Likewise.
3482 * regclass.c (reg_scan_mark_refs): Likewise.
3483 * regrename.c (scan_rtx): Likewise.
3484 * resource.c (mark_referenced_resources): Likewise.
3485 (mark_set_resources): Likewise.
3486 * rtlanal.c (rtx_unstable_p): Likewise.
3487 (rtx_varies_p): Likewise.
3488 (count_occurrences): Likewise.
3489 (reg_mentioned_p): Likewise.
3490 (modified_between_p): Likewise.
3491 (modified_in_p): Likewise.
3492 (volatile_insn_p): Likewise.
3493 (volatile_refs_p): Likewise.
3494 (side_effects_p): Likewise.
3495 (may_trap_p_1): Likewise.
3496 (inequality_comparisons_p): Likewise.
3497 (computed_jump_p_1): Likewise.
3498 (commutative_operand_precedence): Likewise.
3499 * sched-deps.c (sched_analyze_2): Likewise.
3500 * sched-vis.c (print_value): Likewise.
3501 * reload.c (operands_match_p): Likewise.
3502 (subst_reg_equivs): Likewise.
3503 * reload1.c (eliminate_regs_1): Likewise.
3504 (elimination_effects): Likewise.
3505 (scan_paradoxical_subregs): Likewise.
3506 * alias.c (rtx_equal_for_memref_p): Likewise.
3507 * Makefile.in (RTL_BASE_H): Add fixed-value.h.
3508 * emit-rtl.c (const_fixed_htab): New hash table.
3509 (const_fixed_htab_hash, const_fixed_htab_eq, lookup_const_fixed):
3511 (const_fixed_htab_hash, const_fixed_htab_eq, lookup_const_fixed,
3512 const_fixed_from_fixed_value): New functions.
3513 (verify_rtx_sharing): Handle CONST_FIXED.
3514 (copy_rtx_if_shared_1): Likewise.
3515 (reset_used_flags): Likewise.
3516 (set_used_flags): Likewise.
3517 (copy_insn_1): Likewise.
3518 (init_emit_once): Create const_fixed_htab.
3519 Store fixed-point scalar and vector zero and one to const_tiny_rtx.
3521 2007-08-22 Zdenek Dvorak <ook@ucw.cz>
3523 PR tree-optimization/32949
3524 * tree-ssa-loop-niter.c (scev_probably_wraps_p): Test nowrap_type_p
3525 before failing for ivs with non-constant step.
3527 2007-08-22 Hans-Peter Nilsson <hp@axis.com>
3529 * doc/md.texi (Iterators): Renamed from Macros. All contents
3530 changed to reflect rename of respectively define_code_macro and
3531 define_mode_macro to define_code_iterator and define_mode_iterator.
3532 (Mode Iterators, Code Iterators): Similar.
3533 * read-rtl.c (struct iterator_group, struct iterator_traverse_data)
3534 (uses_mode_iterator_p, apply_mode_iterator, uses_code_iterator_p)
3535 (apply_iterator_to_string, uses_iterator_p, apply_iterator_traverse)
3536 (initialize_iterators, find_iterator, check_code_iterator)
3537 (map_attr_string, apply_mode_maps, apply_iterator_to_rtx, add_mapping)
3538 (read_mapping, read_rtx_1): Similar.
3539 * config/alpha/sync.md, config/alpha/alpha.md, config/frv/frv.md,
3540 config/s390/s390.md, config/m32c/blkmov.md, config/m32c/m32c.md,
3541 config/spu/spu.md, config/sparc/sparc.md, config/sparc/sync.md,
3542 config/i386/i386.md, config/i386/mmx.md, config/i386/sse.md,
3543 config/i386/sync.md, config/crx/crx.md, config/xtensa/xtensa.md,
3544 config/cris/cris.c, config/cris/cris.md, config/ia64/sync.md,
3545 config/ia64/div.md, config/ia64/vect.md, config/ia64/ia64.md,
3546 config/m68k/m68k.md, config/rs6000/spe.md, config/rs6000/altivec.md,
3547 config/rs6000/sync.md, config/rs6000/rs6000.md,
3548 config/arm/vec-common.md, config/arm/neon.md, config/arm/iwmmxt.md,
3549 config/arm/arm.md, config/mips/mips-dsp.md, config/mips/mips.md,
3550 config/vax/vax.md, config/bfin/bfin.md: Similar.
3552 2007-08-22 David Daney <ddaney@avtrex.com>
3554 * doc/install.texi (Testing): Mention testing on a simulator.
3556 2007-08-22 Janis Johnson <janis187@us.ibm.com>
3558 * config/dfp-bit.c (DFP_TO_DFP): Check for overflow.
3560 * doc/libgcc.texi (Decimal float library routines): Fix formatting
3561 and rearrange floating point conversion functions into different
3564 2007-08-22 Maxim Kuvyrkov <maxim@codesourcery.com>
3566 * target.h (struct gcc_target.sched: dfa_pre_advance_cycle,
3567 dfa_post_advance_cycle): New scheduler hooks.
3568 * target-def.h (TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE,
3569 TARGET_SCHED_DFA_POST_ADVANCE_CYCLE): New macros to initialize
3571 (TARGET_SCHED): Use them.
3572 * doc/tm.texi (TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE,
3573 TARGET_SCHED_DFA_POST_ADVANCE_CYCLE): Document new hooks.
3574 * haifa-sched.c (advance_one_cycle): Invoke new hooks.
3576 * genautomata.c (insn_has_dfa_reservation_p): New DFA interface
3577 function to facilitate debugging.
3578 (INSN_HAS_DFA_RESERVATION_P_FUNC_NAME): New macro.
3579 (output_insn_has_dfa_reservation_p): New static function to output
3580 insn_has_dfa_reservation_p ().
3581 (write_automata): Use it.
3582 * genattr.c (main): Output declaration for
3583 insn_has_dfa_reservation_p ().
3585 2007-08-22 Christian Bruel <christian.bruel@st.com>
3586 Richard Guenther <rguenther@suse.de>
3588 * fold-const.c (fold_binary): Optimize A-A if -ffinite-math-only.
3589 * simplify_rtx (simplify_binary_operation_1): Likewise.
3591 2007-08-22 Rask Ingemann Lambertsen <rask@sygehus.dk>
3593 PR rtl-optimization/32557
3594 * df-problems.c (df_note_bb_compute): Use mws->start_regno instead
3595 of REGNO (mws->mw_reg).
3597 2007-08-22 Richard Guenther <rguenther@suse.de>
3600 * builtins.c (expand_builtin_int_roundingfn): Replace call
3601 argument wrapped with SAVE_EXPR directly.
3602 (expand_builtin_int_roundingfn_2): Likewise.
3604 2007-08-22 Richard Guenther <rguenther@suse.de>
3606 * tree-inline.c (inlinable_function_p): Restore disregarding
3607 inline limits for GNU extern inline functions.
3609 2007-08-22 Richard Guenther <rguenther@suse.de>
3612 * tree.c (host_integerp): Treat sizetype as signed as it is
3615 2007-08-21 Ian Lance Taylor <iant@google.com>
3617 PR tree-optimization/33134
3618 * tree-vrp.c (adjust_range_with_scev): Call
3619 set_value_range_to_value.
3621 2007-08-21 Aldy Hernandez <aldyh@redhat.com>
3623 * value-prof.h (gimple_remove_histogram_value): Remove duplicate
3626 2007-08-21 Aldy Hernandez <aldyh@redhat.com>
3628 * tree-flow-inline.h (next_imm_use_stmt): Remove unecessary
3631 2007-08-21 Nathan Froyd <froydnj@codesourcery.com>
3633 * config/rs6000/rs6000.c (expand_block_clear): Add TARGET_SPE
3634 cases to set eight bytes at a time.
3635 (expand_block_move): Likewise.
3637 2007-08-21 Jakub Jelinek <jakub@redhat.com>
3640 * dwarf2out.c (gen_decl_die): Don't call
3641 gen_tagged_type_instantiation_die if decl doesn't have tagged type.
3643 2007-08-21 Paul Brook <paul@codesourcery.com>
3644 Nathan Sidwell <nathan@codesourcery.com>
3645 Mark Mitchell <mark@codesourcery.com>
3646 Joseph Myers <joseph@codesourcery.com>
3648 * configure.ac: Add --with-pkgversion and --with-bugurl.
3649 * configure: Regenerate.
3650 * doc/install.texi: Document them.
3651 * version.c (version_string): Remove VERSUFFIX.
3652 (VERSUFFIX): Remove.
3653 (pkgversion_string): New.
3654 (bug_report_url): Do not hard-code initializer.
3655 * version.h (pkgversion_string): Declare.
3656 * Makefile.in (PKGVERSION_s, BUGURL_s, PKGVERSION, BUGURL_TEXI):
3658 (version.o): Define PKGVERSION and BUGURL.
3659 (gcc-vers.texi): Define VERSION_PACKAGE and BUGURL.
3660 (%.pod): Define BUGURL.
3661 * gcc.c (process_command, main): Use pkgversion_string.
3662 * toplev.c (compile_file, print_version): Likewise.
3663 * protoize.c (main): Likewise.
3664 * gcov.c (print_version): Likewise. Update copyright date.
3665 * gcov-dump.c (print_version): Likewise. Update copyright date.
3666 * mips-tdump.c (main): Likewise. Update copyright date.
3667 * mips-tfile.c (main): Likewise. Update copyright date.
3668 * doc/include/gcc-common.texi: Include VERSION_PACKAGE as subtitle.
3669 * doc/bugreport.texi: Use BUGURL for bug-reporting instructions;
3670 shorten description.
3671 * doc/gcc.texi: Include VERSION_PACKAGE in version description.
3672 * doc/gccint.texi: Likewise.
3673 * doc/invoke.texi: Use BUGURL for bug-reporting instructions.
3674 Update copyright date.
3676 2007-08-21 Uros Bizjak <ubizjak@gmail.com>
3678 * config/i386/i386.c (ix86_expand_clear): Use FLAGS_REG.
3679 (ix86_expand_strlensi_unroll_1): Ditto.
3680 (ix86_expand_branch): Use FLAGS_REG and FPSR_REG.
3681 (ix86_expand_carry_flag_compare): Update comment.
3683 2007-08-21 Jakub Jelinek <jakub@redhat.com>
3686 * fold-const.c (fold_unary): Optimize BIT_NOT_EXPR of VECTOR_CST.
3687 (fold_binary): Handle vectors in X | ~X and X ^ ~X optimizations.
3689 2007-08-21 Richard Guenther <rguenther@suse.de>
3691 * fold-const.c (fold_binary): Revert removing of index +p PTR folding.
3693 2007-08-21 Richard Guenther <rguenther@suse.de>
3696 * fold-const.c (fold_binary): Remove index +p PTR folding.
3697 Fix types of POINTER_PLUS_EXPR generated by folding of
3700 2007-08-20 Chao-ying Fu <fu@mips.com>
3702 * c-common.h (enum rid): Add new enumeration values of RID_SAT,
3703 RID_FRACT, and RID_ACCUM. RID_SAT needs to be inserted before
3704 RID_ONEWAY, so that it can be checked in declspecs_add_type.
3705 (c_common_fixed_point_type_for_size): Declare.
3706 * c-parser.c (reswords): Add _Fract, _Accum, and _Sat.
3707 (c_token_starts_typename): Handle RID_FRACT, RID_ACCUM, and RID_SAT.
3708 (c_token_starts_declspecs): Likewise.
3709 (c_parser_declspecs): Likewise.
3710 (c_parser_attributes): Likewise.
3711 * c-tree.h (enum c_typespec_keyword): Add cts_fract and cts_accum.
3712 (c_declspecs): Add saturating_p.
3713 * c-decl.c (build_null_declspecs): Initialize saturating_p.
3714 (declspecs_add_type): Avoid using complex with _Fract, _Accum, or
3715 _Sat. Handle RID_SAT.
3716 Avoid using void, bool, char, int, float, double, _Decimal32,
3717 _Decimal64, _Decimal128, and complex with _Sat.
3718 Handle RID_FRACT and RID_ACCUM.
3719 Make sure _Sat is used with _Fract or _Accum.
3720 (finish_declspecs): Handle cts_fract and cts_accum.
3721 * c-common.c (fixed-value.h): New include.
3722 (constant_expression_warning): Handle FIXED_CST.
3723 (overflow_warning): Likewise.
3724 (warnings_for_convert_and_check): Likewise.
3725 (c_common_fixed_point_type_for_size): New.
3726 (c_common_type_for_mode): Handle fixed-point modes to
3727 return various saturating/non-saturating, signed/unsigned types.
3728 (c_common_signed_or_unsigned_type): Support fixed-point types.
3729 (shorten_compare): Check fixed-point zero.
3730 Handle FIXED_POINT_TYPE.
3731 (c_common_truthvalue_conversion): Handle FIXED_CST.
3732 Handle FIXED_POINT_TYPE.
3733 (c_common_nodes_and_builtins): Record builtin types for fixed-point
3735 (handle_mode_attribute): Handle fixed-point modes. Need to check
3736 if the signness of base type and fixed-point modes are consistent.
3737 (handle_vector_size_attribute): Handle fixed-point modes.
3738 (same_scalar_type_ignoring_signedness): Handle FIXED_POINT_TYPE.
3739 (warn_for_div_by_zero): Check fixed-point zero.
3740 * c-typeck.c (c_common_type): Check FIXED_POINT_TYPE. Build
3741 a common fixed-point type based on fbit, ibit, sign, and saturation.
3742 (build_unary_op): Allow FIXED_POINT_TYPE for CONVERT_EXPR,
3743 NEGATE_EXPR, TRUTH_NOT_EXPR, PREINCREMENT_EXPR, POSTINCREMENT_EXPR,
3744 PREDECREMENT_EXPR, and POSTDECREMENT_EXPR.
3745 (convert_for_assignment): Support FIXED_POINT_TYPE.
3746 (digest_init): Handle FIXED_POINT_TYPE.
3747 (build_binary_op): Support FIXED_POINT_TYPE in *_DIV_EXPR,
3748 TRUTH_ANDIF_EXPR, TRUTH_ORIF_EXPR, TRUTH_AND_EXPR, TRUTH_OR_EXPR,
3749 TRUTH_XOR_EXPR, RSHIFT_EXPR, LSHIFT_EXPR, EQ_EXPR, NE_EXPR, LE_EXPR,
3750 GE_EXPR, LT_EXPR, GT_EXPR.
3751 * target-def.h (TARGET_FIXED_POINT_SUPPORTED_P): New.
3752 (TARGET_INITIALIZER): Add TARGET_FIXED_POINT_SUPPORTED_P.
3753 * target.h (gcc_target): Add fixed_point_supported_p.
3754 * targhooks.c (default_scalar_mode_supported_p): Handle MODE_FRACT,
3755 MODE_UFRACT, MODE_ACCUM, and MODE_UACCUM.
3756 (default_fixed_point_supported_p): Define.
3757 * targhooks.h (default_fixed_point_supported_p): Declare.
3758 * doc/tm.texi (TARGET_FIXED_POINT_SUPPORTED_P): Add.
3759 * doc/install.texi (Configuration): Add --enable-fixed-point.
3760 * configure.ac (--enable-fixed-point): New to enable fixed-point
3761 arithmetic extension to C. For mips targets, we enable it by default.
3762 * configure, config.in: Regenerate.
3764 2007-08-20 Pawel Sikora <pluto@pld-linux.org>
3766 * doc/invoke.texi (-Wnon-virtual-dtor): Update documentation.
3768 2007-08-20 David Edelsohn <edelsohn@gnu.org>
3770 * dwarf2out.c (text_section_used): Move declaration outside ifdef
3771 DWARF2_DEBUGGING_INFO.
3772 (cold_text_section_used): Same.
3773 (cold_text_section): Same.
3775 2007-08-20 Richard Guenther <rguenther@suse.de>
3777 * c-typeck.c (convert_for_assignment): Use the type of
3778 the member for the initialization.
3780 2007-08-20 Richard Guenther <rguenther@suse.de>
3782 * c-objc-common.c (c_disregard_inline_limits): Remove.
3783 * c-objc-common.h (c_disregard_inline_limits): Likewise.
3784 * cgraphunit.c (cgraph_process_new_functions): Call
3785 disregard_inline_limits_p.
3786 (cgraph_preserve_function_body_p): Likewise.
3787 * ipa-inline.c (compute_inline_parameters): Likewise.
3788 * langhooks-def.h (lhd_tree_inlining_disregard_inline_limits):
3790 (LANG_HOOKS_TREE_INLINING_DISREGARD_INLINE_LIMITS): Remove.
3791 (LANG_HOOKS_TREE_INLINING_INITIALIZER): Remove initializer for
3792 disregard_inline_limits langhook.
3793 * langhooks.c (lhd_tree_inlining_disregard_inline_limits):
3795 * langhooks.h (lang_hooks_for_tree_inlining): Remove
3796 disregard_inline_limits langhook.
3797 * tree-inline.c (disregard_inline_limits_p): New function.
3798 * tree-inline.h (disregard_inline_limits_p): Declare.
3800 2007-08-20 Richard Guenther <rguenther@suse.de>
3802 * langhooks-def.h (lhd_tree_inlining_auto_var_in_fn_p): Remove.
3803 (LANG_HOOKS_TREE_INLINING_AUTO_VAR_IN_FN_P): Likewise.
3804 (LANG_HOOKS_TREE_INLINING_INITIALIZER): Remove initializer for
3805 auto_var_in_fn_p langhook.
3806 * langhooks.c (lhd_tree_inlining_auto_var_in_fn_p): Rename and
3808 * tree.c (auto_var_in_fn_p): ... here.
3809 (find_var_from_fn): Call auto_var_in_fn_p directly.
3810 * langhooks.h (lang_hooks_for_tree_inlining): Remove
3811 auto_var_in_fn_p langhook.
3812 * tree-inline.c (remap_decls): Call auto_var_in_fn_p directly.
3813 (copy_body_r): Likewise.
3814 (self_inlining_addr_expr): Likewise.
3815 * tree.h (auto_var_in_fn_p): Declare.
3817 2007-08-20 Richard Guenther <rguenther@suse.de>
3819 * tree.c (WALK_SUBTREE): Call walk_tree_1.
3820 (walk_type_fields): Take lh parameter.
3821 (walk_tree): Rename to ...
3822 (walk_tree_1): ... this. Do not call the walk_subtrees
3823 langhook but the now passed callback. Pass lh on recursion.
3824 (walk_tree_without_duplicates): Rename to ...
3825 (walk_tree_without_duplicates_1): ... this. Take lh parameter
3826 and call walk_tree_1.
3827 * tree.h (walk_tree_lh): New typedef.
3828 (walk_tree_1): Declare.
3829 (walk_tree_without_duplicates_1): Likewise.
3830 (walk_tree): New define to walk_tree_1 with NULL lh parameter.
3831 (walk_tree_without_duplicates): New define to
3832 walk_tree_without_duplicates_1 with NULL lh parameter.
3833 * langhooks.c (lhd_tree_inlining_walk_subtrees): Remove.
3834 * langhooks.h (lang_hooks_for_tree_inlining): Remove walk_subtrees
3836 * langhooks-def.h (lhd_tree_inlining_walk_subtrees): Remove.
3837 (LANG_HOOKS_TREE_INLINING_WALK_SUBTREES): Likewise.
3838 (LANG_HOOKS_TREE_INLINING_INITIALIZER): Remove walk_subtrees
3841 2007-08-20 Andrew Pinski <andrew_pinski@playstation.sony.com>
3844 * tree-inline.c (optimize_inline_calls): Move the cgraph checking
3845 code in front of the compacting of basic blocks.
3846 Move the folding of statements inbetween the cgraph checking
3847 and compacting of basic blocks.
3849 2007-08-19 Andrew Pinski <andrew_pinski@playstation.sony.com>
3850 Serge Belyshev <belyshev@depni.sinp.msu.ru>
3853 * config/alpha/alpha.c (va_list_skip_additions): Check for
3854 POINTER_PLUS_EXPR in addition to PLUS_EXPR.
3855 (alpha_stdarg_optimize_hook): Look for POINTER_PLUS_EXPR instead of
3856 PLUS_EXPR when checking ap.__base.
3857 (alpha_va_start): Create POINTER_PLUS_EXPR instead of PLUS_EXPR
3858 when doing addition on pointer types. Use size_int instead of
3860 (alpha_gimplify_va_arg_1): Likewise, but use sizetype instead of
3861 ptr_type in the second operand.
3863 2007-08-19 Andrew Pinski <pinskia@gmail.com>
3866 * config/i386/darwin.h (CC1_SPEC): Add %(cc1_cpu) in front.
3868 2007-08-19 Andrew Pinski <andrew_pinski@playstation.sony.com>
3871 * cfgexpand.c (expand_one_register_var): Mark pointer
3872 DECL_ARTIFICIAL as REG_POINTER also.
3873 * stmt.c (expand_decl): Likewise.
3875 2007-08-19 Daniel Berlin <dberlin@dberlin.org>
3882 * tree-flow.h (struct stmt_ann_d): Remove makes_clobbering_call.
3883 * tree-ssa-alias.c (init_transitive_clobber_worklist): Add
3884 on_worklist argument and avoid adding things to worklist multiple
3886 (add_to_worklist): Ditto.
3887 (mark_aliases_call_clobbered): Mark entire structure clobbered if
3888 single SFT is clobbered.
3889 (set_initial_properties): Ditto.
3890 (compute_call_clobbered): Update for changes to function
3892 (create_overlap_variables_for): Always create SFT for offset 0.
3893 (create_structure_vars): Handle PHI's, since we are in SSA form at
3895 * tree-ssa-loop-ivopts.c (get_ref_tag): Don't return subvars.
3896 * tree-ssa-operands.c (access_can_touch_variable): Don't handle
3898 (add_vars_for_offset): Figure out aliases from access + points-to.
3899 (add_virtual_operand): Use add_vars-for_offset.
3900 (get_tmr_operands): Update for NMT changes, rewrite to be correct.
3901 (add_call_clobber_ops): Remove makes_clobbering_call set.
3902 (get_expr_operands): Always pass through the INDIRECT_REF
3904 * tree-ssa-structalias.c (struct constraint_graph): Remove
3906 Add pe, pe_rep, pointer_label, loc_label, pointed_by, points_to,
3907 address_taken, pt_used, number_incoming.
3908 (FIRST_ADDR_NODE): Removed.
3909 (merge_graph_nodes): Remove broken code for the moment.
3910 (init_graph): New function.
3911 (build_pred_graph): Remove code to init_graph.
3912 Add location equivalence support.
3913 (struct scc_info): Rename roots to deleted.
3915 (init_scc_info): Ditto
3916 (init_topo_info): Use graph->size.
3917 (compute_topo_order): Ditto.
3918 (do_da_constraint): Removed.
3919 (do_sd_constraint): Remove calls to find().
3920 set_union_with_increment should always get 0 as last arg here.
OSDN Git Service
