1 2007-09-12 Dorit Nuzman <dorit@il.ibm.com>
3 PR tree-optimization/33373
4 * tree-vect-analyze (vect_determine_vectorization_factor): Call
5 TREE_INT_CST_LOW when comparing TYPE_SIZE_UNIT.
7 2007-09-12 Jan Hubicka <jh@suse.cz>
10 * i386.md (floatsisf2_mixed_memory, floatsisf2_sse_memory): Disable for
13 2007-09-12 Christian Bruel <christian.bruel@st.com>
15 * sh.h (SH_DBX_REGISTER_NUMBER): Added fpscr, fixed sr/gbr regs.
16 * linux-unwind.h (SH_DWARF_FRAME_GBR): fixed.
18 2007-09-12 Ira Rosen <irar@il.ibm.com>
20 * tree-vect-transform.c (vect_get_slp_defs): Don't build a vector
21 for oprnd1 if not required.
22 (vectorizable_operation): Use scalar operand in SLP in case of
23 shift with scalar argument.
25 2007-09-12 Ira Rosen <irar@il.ibm.com>
27 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change default and minimum
30 2007-09-11 James E. Wilson <wilson@specifix.com>
32 * defaults.h (DWARF2_UNWIND_INFO): Don't define if
33 TARGET_UNWIND_INFO is defined.
34 * config/ia64/ia64.h (INCOMING_RETURN_ADDR_RTX): Delete undef
37 2007-09-12 Kaz Kojima <kkojima@gcc.gnu.org>
39 * config/sh/sh.c (calc_live_regs): Use
40 current_function_saves_all_registers instead of
41 current_function_has_nonlocal_label.
42 (sh_allocate_initial_value): Likewise.
43 (sh_get_pr_initial_val): Likewise.
44 * config/sh/sh.h (SHMEDIA_REGS_STACK_ADJUST): Likewise.
45 * config/sh/sh.md (load_ra): Likewise.
47 2007-09-12 Hans-Peter Nilsson <hp@axis.com>
49 * config/cris/t-linux (LIMITS_H_TEST): Only define if not inhibit_libc.
52 * config/cris/cris.c (cris_expand_pic_call_address): Fix typo in
53 GET_CODE (x) == CONST_INT to CONST_INT_P (x) transformation.
55 2007-09-12 Sa Liu <saliu@de.ibm.com>
57 * config/spu/spu.c (spu_emit_branch_or_set): Handle NaN values as
58 operands to DFmode GE or LE compares.
60 2007-09-12 Bernd Schmidt <bernd.schmidt@analog.com>
62 * config/bfin/bfin.h (enum reg_class, REG_CLASS_CONTENTS,
63 REG_CLASS_NAMES): Add P0REGS.
64 (REGNO_REG_CLASS): Return it where appropriate.
65 (REG_CLASS_FROM_CONSTRAINT): Add 'qA'.
66 (CLASS_LIKELY_SPILLED_P): P0REGS is likely_spilled.
67 * doc/md.texi (Blackfin family): Document 'q' constraints.
69 2007-09-11 Steve Kenton <skenton@ou.edu>
71 * pa/linux-unwind.h: Guard with inhibit_libc.
72 * pa/hpux-unwind.h: Likewise.
74 2007-09-11 David Daney <ddaney@avtrex.com>
76 * doc/invoke.texi: Document new MIPS -mllsc and -mno-llsc options.
77 * doc/install.texi: Document new --with-llsc and --without-llsc
79 * config.gcc: Handle --with-llsc and --without-llsc configure options.
80 * config/mips/mips.md (sync, memory_barrier): Wrap sync instrunction
81 in %| and %- operand codes. Depend on GENERATE_SYNC instead of
83 (sync_compare_and_swap<mode>, sync_add<mode>, sync_sub<mode>,
84 sync_old_add<mode>, sync_old_sub<mode>, sync_new_add<mode>,
85 sync_new_sub<mode>, sync_<optab><mode>, sync_old_<optab><mode>,
86 sync_new_<optab><mode>, sync_nand<mode>, sync_old_nand<mode>,
87 sync_new_nand<mode>, sync_lock_test_and_set<mode>): Depend on
88 GENERATE_LL_SC instead of ISA_HAS_LL_SC.
89 * config/mips/mips.opt (mllsc): New option.
90 * config/mips/mips.c (mips_llsc): Define variable.
91 (mips_handle_option): Handle mllsc option.
92 (override_options): Set mips_print_operand_punct for '|' and '-'.
93 (print_operand): Add new %| and %- operand codes.
94 * config/mips/mips.h (mips_llsc_setting): New enum type.
96 (OPTION_DEFAULT_SPECS): Add llsc handling.
97 (GENERATE_SYNC): New macro.
98 (GENERATE_LL_SC): New macro.
99 (MIPS_COMPARE_AND_SWAP, MIPS_SYNC_OP, MIPS_SYNC_OLD_OP,
100 MIPS_SYNC_NEW_OP, MIPS_SYNC_NAND, MIPS_SYNC_OLD_NAND,
101 MIPS_SYNC_NEW_NAND, MIPS_SYNC_EXCHANGE): Wrap instructions
102 in %| and %- operand codes.
104 2007-09-11 Eric Botcazou <ebotcazou@adacore.com>
106 * tree-ssa-structalias.c (push_fields_onto_fieldstack): Deal with
107 TYPE_NONALIASED_COMPONENT like with DECL_NONADDRESSABLE_P.
109 2007-09-11 Jason Merrill <jason@redhat.com>
112 * stor-layout.c (layout_decl): Do pack variable size fields.
114 2007-09-11 Maxim Kuvyrkov <maxim@codesourcery.com>
116 * config/m68k/predicates.md (movsi_const0_operand,
117 non_symbolic_call_operand): New predicates.
119 * config/m68k/constraints.md: (Cs, Ci, C0, Cj, CQ, CW, CZ, CS, Ap, Ac):
121 * doc/md.texi (Constraints for Particular Machines: Motorola 680x0):
122 Document constraints N, O, P, R, S, T, Q, U, W, Cs, Ci, C0, Cj, CQ,
123 CW, CZ, CS, Ap and Ac.
125 * config/m68k/m68k.md (UNSPEC_IB): New constant.
126 (constraints.md): New include.
127 (cpu, type, type1, opx, opy, opx_type, opy_type, size, opx_access,
128 opx_mem, opy_mem, op_mem, guess, split): New attributes.
129 (movdf_internal): Name pattern. Fix to use alternatives. Add split.
132 (tstsi_internal): Name pattern. Fix to use alternatives. Specify
133 attributes. Split tstsi_internal_68020_cf from it.
134 (tstsi_internal_68020_cf): New pattern.
135 (tsthi_internal, tstqi_internal): Name pattern. Specify attributes.
136 (tst<mode>_cf): Specify attributea.
137 (cmpsi_cf): Name pattern. Specify attributes.
138 (cmp<mode>_68881, cmp<mode>_cf): Specify type attribute.
139 (pushexthisi_const): Fix to use alternatives. Specify
141 (movsi_const0): Split movsi_const0_68000_10 and movsi_const0_68040_60
142 from it. Fix to use alternatives. Specify attributes.
143 (movsi_const0_68040_10, movsi_const0_68040_60): New patterns.
144 (movsi_cf, movstrictqi_cf): Fix to use alternatives. Specify
146 (movsf_cf_soft): Specify attributes.
147 (movdf_cf_soft): Add split.
148 (pushasi, zero_extendhisi2_cf, zero_extendqisi2_cfv4,
149 cfv4_extendhisi2, 68k_extendhisi2, extendqihi2, cfv4_extendqisi2,
150 68k_extendqisi2, truncdfsf2_cf): Specify attributes.
151 (truncdfsf2_68881): Name pattern. Specify attributes.
152 (floatsi<mode>2_cf, floathi<mode>2_68881, floathi<mode>2_cf,
153 floatqi<mode>2_68881, floatqi<mode>2_cf, ftrunc<mode>2_cf,
154 fix<mode>qi2_cf, fix<mode>hi2_cf, fix<mode>si2_cf, adddi_dishl32):
156 (addsi3_5200): Fix to use alternatives. Specify attributes.
158 (add<mode>3_cf, subdi_dishl32): Specify attributes.
159 (subsi3): Add alternative for subq.l. Specify attributes.
160 (sub<mode>3_cf, mulhi3, mulhisi3): Specify attributes.
161 (mulhisisi3_s, mulsi3_68020, mulsi3_cf): Name pattern. Specify
163 (umulhisi3): Specify attributes.
164 (mulhisisi3_z): Name pattern. Specify attributes.
165 (fmul<mode>3_cf, div<mode>3_cf, negsi2_internal, negsi2_5200,
166 sqrt<mode>2_68881, clzsi2, one_cmplsi2_5200, subreghi1ashrdi_const32,
167 subregsi1ashrdi_const32, ashrsi3, subreg1lshrdi_const32, lshrsi3,
168 bsetmemqi): Specify attributes.
169 (bsetmemqi_ext): Name pattern. Specify attributes.
170 (bclrmemqi): Specify attributes.
171 (bclrmemqi_ext, scc, sls): Name pattern. Specify attributes.
172 (beq, bne, bgt, bgtu, blt, bltu, bge, bgeu, ble, bleu): Specify
174 (beq2, bne2, bgt2, bgtu2, blt2, bltu2, bge2, bgeu2, ble2, bleu2): Name
175 pattern. Specify attributes.
176 (jump): Specify attributes.
177 (tablejump_internal): Name pattern. Specify attributes.
178 (call_value): Split into non_symbolic_call_value,
179 symbolic_call_value_jsr, symbolic_call_value_bsr. Fix to use
180 alternatives. Specify attributes.
181 (non_symbolic_call_value, symbolic_call_value_jsr,
182 symbolic_call_value_bsr): New patterns.
183 (nop, return, unlink, indirect_jump): Specify attributes.
184 (trap): Fix condition. Specify attributes.
187 * config/m68k/m68k.c (m68k_symbolic_call_var): New variable.
188 (override_options): Initialize it. Initialize m68k_sched_cpu.
189 (CONST_METHOD): Rename to M68K_CONST_METHOD, move to m68k.h.
190 (const_method): Make global, rename to m68k_const_method.
191 (const_int_cost, output_move_const_into_data_reg): Update.
192 (output_move_double): Parametrize to emit rtl code, rename to
194 (output_reg_adjust, emit_reg_adjust, output_compadr, output_movsi,
195 emit_movsi): New static functions.
196 (output_move_double): New function with semantics of old
198 (m68k_emit_move_double): New function.
199 (m68k_sched_cpu): New variable.
200 (attr_op_type): New enum.
201 (sched_guess_p): New variable.
202 (sched_address_type, sched_operand_type, sched_attr_op_type):
203 New static functions.
204 (m68k_sched_attr_opx_type, m68k_sched_attr_opy_type,
205 m68k_sched_attr_size, m68k_sched_attr_op_mem): New functions.
206 (sched_branch_type): New static variable.
207 (m68k_sched_branch_type): New function.
208 * config/m68k/m68k.h (M68K_SYMBOLIC_CALL): New enum.
209 (m68k_symbolic_call_var): Declare.
210 (M68K_CONST_METHOD): Rename from CONST_METHOD. Move here from m68k.c.
211 (m68k_const_method, m68k_emit_move_double, m68k_sched_cpu,
212 m68k_sched_attr_opx_type, m68k_sched_attr_opy_type,
213 m68k_sched_attr_size, m68k_sched_attr_op_mem, m68k_sched_branch_type):
216 2007-09-11 Jakub Jelinek <jakub@redhat.com>
218 * builtins.def (BUILT_IN_VA_ARG_PACK_LEN): New builtin.
219 * builtins.c (expand_builtin) <case BUILT_IN_VA_ARG_PACK_LEN>: Issue
220 error if __builtin_va_arg_pack_len () wasn't optimized out during
222 * tree-inline.c (copy_bb): Replace __builtin_va_arg_pack_len ()
223 with the number of inline's anonymous arguments.
224 * doc/extend.texi: Document __builtin_va_arg_pack_len ().
226 2007-09-11 Zdenek Dvorak <ook@ucw.cz>
228 * fold-const.c (extract_muldiv_1): Do not simplify
231 2007-09-11 Jan Hubicka <jh@suse.cz>
233 * i386.h (ix86_tune_indices): Add X86_TUNE_INTER_UNIT_CONVERSIONS.
234 (TARGET_INTER_UNIT_CONVERSIONS): New.
235 * i386.md (floatsi expanders): Remove redundant check for SImode
236 source; offload to memory when asked for.
237 (floatsisf2_mixed, floatsisf2_sse, floatsidf2_mixed, floatsidf2_sse
238 floatdisf2_mixed, floatsisf2_sse, floatsidf2_mixed, floatsidf2_sse):
240 (floatsisf2_mixed_memory, floatsisf2_sse_memory,
241 floatsidf2_mixed_memory, floatsidf2_sse_memory
242 floatdisf2_mixed_memory, floatsisf2_sse_memory,
243 floatsidf2_mixed_memory, floatsidf2_sse_memory): New.
245 2007-09-11 Jan Hubicka <jh@suse.cz>
247 * toplev.c (process_options): all frontends now do unit-at-a-time.
248 * cgraphunit.c: update comments.
249 (cgraph_expand_function): call passmanager dirrectly; emit thunks.
250 * c-decl.c (finish_function): use cgraph_add_new_function.
251 * function.c (expand_function_end): We are always unit-at-a-time.
253 2007-09-11 Richard Sandiford <richard@codesourcery.com>
255 * config/mips/mips.c (mips_set_mips16_mode): Use separate anchor
257 (mips_use_anchors_for_symbol_p): Use default_use_anchors_for_symbol_p.
259 2007-09-11 Richard Sandiford <richard@codesourcery.com>
261 * config/mips/mips.c (mips_symbol_insns_1): Allow LEAs of
262 SYMBOL_FORCE_TO_MEM constants.
263 (mips_rtx_costs): Give a cost of 1 to force_to_mem_operands.
264 (mips16_rewrite_pool_refs_info): New structure.
265 (mips16_rewrite_pool_constant): New function, split out from...
266 (mips16_rewrite_pool_refs): ...here. Take a pointer to a
267 mips16_rewrite_pool_refs_info structure rather than a pointer
268 to a constant pool. Force force_to_mem_operands into memory.
269 (mips16_lay_out_constants): Update call to mips16_rewrite_pool_refs.
270 * config/mips/predicates.md (force_to_mem_operand): New predicate.
271 * config/mips/constraints.md (kf): New constraint.
272 * config/mips/mips.md (*movdi_64bit_mips16): Add a d <- kf alternative.
273 (*movsi_mips16): Likewise.
275 2007-09-11 Richard Sandiford <richard@codesourcery.com>
276 Nigel Stephens <nigel@mips.com>
277 David Ung <davidu@mips.com>
279 * config/mips/mips.h (CONSTANT_POOL_COST): Move to...
280 * config/mips/mips.c: ...here and set to 4 for TARGET_MIPS16.
281 (mips16_constant_cost, mips_immediate_operand_p, mips_binary_cost)
282 (mips_fp_mult_cost, mips_fp_div_cost, mips_sign_extend_cost)
283 (mips_zero_extend_cost): New functions.
284 (mips_rtx_costs): Treat COMPARE constants as having zero cost.
285 Use the new functions. Tweak many cost estimates, both here
286 and in the new subroutines. Return false when the cost of the
287 operands has not been calculated. Check for *clear_upper32.
288 Check for floating-point multiply-add, reciprocal and rsqrt
289 patterns. Handle comparison and rotation codes.
291 2007-09-11 Danny Smith <dannysmith@users.sourceforge.net>
293 * config/i386/cygming.h (TARGET_STRIP_NAME_ENCODING): Don't
295 * config/i386/i386.c (get_dllimport_decl): Don't strip
298 2007-09-10 Janis Johnson <janis187@us.ibm.com>
301 * config/dfp-bit.c: Don't skip TFmode conversions; move strto*
303 (DFP_TO_BFP): Use for either XFmode or TFmode.
304 (BFP_TO_DFP): Use for either XFmode or TFmode; always use cast
306 * config/dfp-bit.h: Include float.h.
307 (LONG_DOUBLE_HAS_XF_MODE, LONG_DOUBLE_HAS_TF_MODE): Define if long
308 double is one of these modes, rather than using LIBGCC_HAS_*F_MODE
309 which doesn't mean the same thing.
310 (BFP_KIND): Use 4 to mean TFmode.
311 (BFP_FMT): Specify the number of decimal digits based on the
312 number of mantissa digits.
313 (BFP_VIA_TYPE): Binary float type to use as cast for sprintf.
314 (BFP_TO_DFP, DFP_TO_BFP): Define names for TFmode variants.
315 (STR_TO_BFP): Use strtold for XFmode or TFmode.
316 (TFtype): Define if TFmode is supported.
317 * doc/libgcc.texi (Decimal float library routines): Document
318 TF conversion functions.
320 2007-09-10 Chao-ying Fu <fu@mips.com>
322 * config/mips/mips.c (mips_scalar_mode_supported_p): Declare.
323 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
324 (mips_emit_compare): Process fixed-point modes.
325 (mips_pad_arg_upward): Support fixed-point types.
326 (override_options): Allow fixed-point modes in accumulators.
327 (mips_pass_by_reference): Pass DQ, UDQ, DA, and UDA modes in registers.
328 (mips_vector_mode_supported_p): Support V2HQmode, V2UHQmode, V2HAmode,
329 V2UHAmode, V4QQmode, and V4UQQmode when TARGET_DSP.
330 (mips_scalar_mode_supported_p): New function to accept fixed-point
331 modes if the width is not greater than two BITS_PER_WORD.
332 * config/mips/mips.h (SHORT_FRACT_TYPE_SIZE, FRACT_TYPE_SIZE,
333 LONG_FRACT_TYPE_SIZE, LONG_LONG_FRACT_TYPE_SIZE,
334 SHORT_ACCUM_TYPE_SIZE, ACCUM_TYPE_SIZE, LONG_ACCUM_TYPE_SIZE,
335 LONG_LONG_ACCUM_TYPE_SIZE): Define.
336 * config/mips/mips.md ("d"): Update mode attribute for fixed-point
338 ("IMODE"): New mode attribute.
339 (mips-fixed.md): Include.
340 * config/mips/mips-modes.def: Create VECTOR_MODES for FRACT, UFRACT,
342 * config/mips/mips-fixed.md: New file.
344 2007-09-11 Ben Elliston <bje@au.ibm.com>
346 * config/spu/spu.md: Formatting fixes.
348 2007-09-10 Janis Johnson <janis187@us.ibm.com>
350 * config/dfp-bit.c (dfp_unary_func): Delete.
351 (dfp_unary_op): Delete.
352 (dfp_binary_op): Use decFloat functions instead of decNumber
353 functions for binary operations.
354 (d32_binary_op): Convert 32-bit operands to 64 bits for evaluation.
355 (dnn_binary_op): Call dfp_binary_op with decFloat rather than
357 (dfp_compare_op): Use decFloat functions insteadof decNumber
358 functions for comparisons.
359 (d32_compare_op): Convert 32-bit operands to 64 bits for evaluation.
360 (dnn_binary_op): Call dfp_compare_op with decFloat rather than
362 (DFP_ADD, DFP_SUB, DFP_MULTIPLE, DFP_DIVIDE): Use macros for
363 call to dxx_binary_op and decFloat function.
364 (DFP_EQ, DFP_NE, DFP_LT, DFP_GT, DFP_LE, DFP_GE): Use macros for
365 calls to dxx_binary_op and decFloat function.
366 * config/dfp-bit.h: Include decFloat header files.
367 (decFloat, DFP_BINARY_OP, DFP_COMPARE_OP, DEC_FLOAT_ADD,
368 DEC_FLOAT_SUBTRACT, DEC_FLOAT_MULTIPLY, DEC_FLOAT_DIVIDE,
369 DEC_FLOAT_COMPARE, DEC_FLOAT_IS_ZERO, DEC_FLOAT_IS_NAN,
370 DEC_FLOAT_IS_SIGNED: Define for each of 3 operand widths.
372 2007-09-10 Harsha Jagasia <harsha.jagasia@amd.com>
373 Jan Sjodin <jan.sjodin@amd.com>
375 * tree-vect-analyze.c (vect_analyze_operations): Change
376 comparison of loop iterations with threshold to less than
377 or equal to instead of less than. Reduce
378 min_scalar_loop_bound by one.
379 * tree-vect-transform.c (vect_estimate_min_profitable_iters):
380 Change prologue and epilogue iterations estimate to vf/2,
381 when unknown at compile-time. Change versioning guard
382 cost to taken_branch_cost. If peeling for alignment is
383 unknown at compile-time, change peel guard costs to one
384 taken branch and one not-taken branch per peeled loop.
385 If peeling for alignment is known but number of scalar loop
386 iterations is unknown at compile-time, change peel guard
387 costs to one taken branch per peeled loop. Change the cost
388 model equation to consider vector iterations as the loop
389 iterations less the prologue and epilogue iterations.
390 Change outside vector cost check to less than or equal to
391 zero instead of equal to zero.
392 (vect_do_peeling_for_loop_bound): Reduce
393 min_scalar_loop_bound by one.
394 * tree-vectorizer.h: Add TARG_COND_TAKEN_BRANCH_COST and
395 TARG_COND_NOT_TAKEN_BRANCH_COST.
396 * config/i386/i386.h (processor_costs): Add
397 scalar_stmt_cost, scalar_load_cost, scalar_store_cost,
398 vec_stmt_cost, vec_to_scalar_cost, scalar_to_vec_cost,
399 vec_align_load_cost, vect_unalign_load_cost,
400 vec_store_cost, cond_taken_branch_cost,
401 cond_not_taken_branch_cost.
402 Define macros for x86 costs.
403 * config/i386/i386.c:
404 (size_cost): Set scalar_stmt_cost, scalar_load_cost,
405 scalar_store_cost, vec_stmt_cost, vec_to_scalar_cost,
406 scalar_to_vec_cost, vec_align_load_cost,
407 vect_unalign_load_cost, vec_store_cost,
408 cond_taken_branch_cost, cond_not_taken_branch_cost to one.
409 (i386_cost, i486_cost, pentium_cost, pentiumpro_cost,
410 geode_cost, k6_cost, athlon_cost, pentium4_cost, nocona_cost,
411 core2_cost, generic64_cost, generic32_cost): Set to default
413 (k8_cost, amdfam10_cost): Costs for vectorization tuned.
414 (x86_builtin_vectorization_cost): New.
416 2007-09-10 Janis Johnson <janis187@us.ibm.com>
417 Ben Elliston <bje@au.ibm.com>
419 * dfp.c: Include decimal128Local.h;
420 (dfp_byte_swap): Remove.
421 (encode_decimal32, decode_decimal32): Don't handle endianness.
422 (encode_decimal64, decode_decimal64): Ditto.
423 (encode_decimal128, decode_decimal128): Ditto.
424 * config/dfp-bit.c (host_to_ieee32, ieee_to_host_32): Ditto.
426 (host_to_ieee_64, ieee_to_host_64): Don't handle endianness.
428 (host_to_ieee_128, ieee_to_host_128): Don't handle endianness.
429 * Makefile.in (DECNUM_H): Add decimal128Local.h.
431 2007-09-10 David Daney <ddaney@avtrex.com>
433 * config/mips/mips.md (UNSPEC_MEMORY_BARRIER): New entry in
435 (memory_barrier): Rewrote as an insn that clobbers memory.
437 2007-09-10 Richard Sandiford <richard@codesourcery.com>
439 * config/mips/mips.c (mips_global_pointer): Check
440 call_really_used_regs instead of call_used_regs.
441 (mips_save_reg_p): Likewise. Save all call-saved registers
442 if current_function_saves_all_registers. Fix indentation.
443 No longer treat $18 as a special case.
444 (compute_frame_size): Guard FPR loop with TARGET_HARD_FLOAT.
446 2007-09-10 Richard Sandiford <richard@codesourcery.com>
448 * config/mips/mips.h (MIPS_ARCH_FLOAT_SPEC): New macro.
449 * config/mips/mips.c (mips_cpu_info_table): Mention it in the
450 the introductory comment.
451 (MIPS_MARCH_CONTROLS_SOFT_FLOAT): Delete.
452 (override_options): Don't test for it.
453 * config/mips/sde.h (MIPS_MARCH_CONTROLS_SOFT_FLOAT): Delete.
454 (DRIVER_SELF_SPECS): Add MIPS_ARCH_FLOAT_SPEC.
455 * config/mips/vr.h: As for config/mips/sde.h.
457 2007-09-10 Trevor Smigiel <trevor_smigiel@playstation.sony.com>
458 Revital Eres <eres@il.ibm.com>
460 * target.h (struct gcc_target.sched): New field: sms_res_mii.
461 (struct ddg): Define.
462 * target-def.h (TARGET_SCHED_SMS_RES_MII): Define.
463 (TARGET_SCHED): Add TARGET_SCHED_SMS_RES_MII.
464 * config/spu/spu.c: Include ddg.h.
465 (TARGET_SCHED_SMS_RES_MII): Define.
466 (spu_sms_res_mii): New function to calculate mii.
467 * modulo-sched (res_MII): Use it.
468 * doc/tm.texi: Document TARGET_SCHED_SMS_RES_MII.
470 2007-09-10 Andreas Krebbel <krebbel1@de.ibm.com>
472 * config/s390/s390.c (s390_dump_pool): Create copy of constant
473 pool entries since they might hold values that must not be shared.
475 2007-09-10 Uros Bizjak <ubizjak@gmail.com>
478 * gcc/config/i386/sse.md (ashr<mode>3): Change op2 mode to SImode.
479 Use 'N' operand constraint for op2.
480 (lshr<mode>3): Ditto.
481 (ashl<mode>3): Ditto.
482 (vec_shl_<mode>): Use const_0_to_255_mul_8_operand predicate for op2.
483 (vec_shr_<mode>): Use const_0_to_255_mul_8_operand predicate for op2.
485 * gcc/config/i386/i386.c (ix86_expand_builtin) [IX86_BUILTIN_PSLL?128,
486 IX86_BUILTIN_PSRA*?128, IX86_BUILTIN_PSRL?128]: Convert op1 to SImode.
488 2007-09-10 Andreas Krebbel <krebbel1@de.ibm.com>
490 * config/s390/s390.md ("fixuns_trunc<BFP:mode><GPR:mode>2"):
491 Change mode macro in the last real_2expN parameter to uppercase.
493 2007-09-10 Michael Matz <matz@suse.de>
495 * tree-pass.h (pass_cselim): Declare new pass.
496 * passes.c (init_optimization_passes): Link in pass_cselim.
497 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Renamed from
498 tree_ssa_phiopt; add do_store_elim parameter, handle it by calling
499 cond_store_replacement.
500 (condstoretemp): New static variable.
501 (cond_store_replacement): New function.
502 (tree_ssa_phiopt, tree_ssa_cs_elim): New wrappers around
503 tree_ssa_phiopt_worker.
504 (struct name_to_bb): New.
505 (get_non_trapping, name_to_bb_hash, name_to_bb_eq, add_or_mark_expr,
506 nt_init_block, nt_fini_block): New static functions.
507 (seen_ssa_names, nontrap_set): New static variables.
508 (gate_cselim, pass_cselim): Define new pass.
509 * common.opt (ftree-cselim): New flag.
510 * toplev.c (process_options): Set flag_tree_cselim if required.
512 2007-09-10 Hans-Peter Nilsson <hp@axis.com>
514 * simplify-rtx.c (simplify_relational_operation_1): For recent
515 canonicalization, don't recurse if op1 equals both PLUS arguments.
517 2007-09-09 David Daney <ddaney@avtrex.com>
519 * optabs.c (expand_sync_operation): Use plus insn if minus
521 (expand_sync_fetch_operation): Ditto.
523 2007-09-09 H.J. Lu <hongjiu.lu@intel.com>
525 * i386.md (*floatsisf2_mixed_vector): Use cvtdq2ps instead
527 (*floatsisf2_sse_vector): Likewise.
529 2007-09-09 Krister Walfridsson <cato@df.lth.se>
531 * config/netbsd.h (HANDLE_PRAGMA_PACK_PUSH_POP): Define to 1.
533 2007-09-09 H.J. Lu <hongjiu.lu@intel.com>
535 * config/i386/i386.h (ix86_tune_indices): Rename
536 X86_USE_VECTOR_CONVERTS to X86_TUNE_USE_VECTOR_CONVERTS.
537 (TARGET_USE_VECTOR_CONVERTS): Updated.
538 * config/i386/i386.c: Likewise.
540 2007-09-09 Sandra Loosemore <sandra@codesourcery.com>
541 Nigel Stephens <nigel@mips.com>
543 * doc/invoke.texi (Overall Options): Add .sx file extension
545 * cppspec.c (known_suffixes): Likewise.
546 * gcc.c (default_compilers): Likewise.
548 2007-09-09 Rask Ingemann Lambertsen <rask@sygehus.dk>
551 * config/i386/i386.h (CANONICALIZE_COMPARISON): Delete.
552 * simplify-rtx.c (simplify_relational_operation_1): Add the
553 canonicalization from i386.h.
554 * doc/md.texi (Canonicalization of Instructions): Document it.
556 2007-09-09 Jan Hubicka <jh@suse.cz>
557 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com>
559 * i386.h (ix86_tune_indices): Add X86_USE_VECTOR_CONVERTS.
560 (TARGET_USE_VECTOR_CONVERTS): New.
561 * i386.md: New post-reload splitters for converting SF to DF and DF to
563 (floatsi* expander): Special case vector conversions.
564 (floatsisf2_mixed_vector, floatsisf2_sse_vector_nointernunit,
565 floatsisf2_sse_vector_internunit, floatsisf2_sse_vector,
566 floatsidf2_mixed_vector, floatsidf2_sse_vector): New.
567 (floatsisf2_mixed, floatsisf2_sse, floatsidf2_mixed, floatsidf2_sse):
568 Disable when doing vector converts.
569 (floatsi<mode>2_i387): Disable when
570 * sse.md (vec_dupv2df): Export.
571 * i386.c (ix86_tune_features): Enable SSE conversions.
573 2007-09-09 Richard Guenther <rguenther@suse.de>
575 * tree-ssa-operands.c (add_virtual_operand): Only mark
576 stores as has_volatile_ops if alias information is not available.
578 2007-09-09 Revital Eres <eres@il.ibm.com>
580 * doc/contrib.texi: Add myself.
582 2007-09-09 Ira Rosen <irar@il.ibm.com>
584 * tree-vectorizer.h (stmt_vinfo_set_inside_of_loop_cost,
585 stmt_vinfo_set_outside_of_loop_cost): New functions.
586 * tree-vect-transform.c (vect_get_cost_fields): Remove.
587 (vect_model_simple_cost): Call
588 stmt_vinfo_set_inside/outside_of_loop_cost to set the relevant cost
589 field instead of calling vect_get_cost_fields.
590 (vect_model_store_cost, vect_model_load_cost): Likewise.
592 2007-09-09 Revital Eres <eres@il.ibm.com>
594 * config/rs6000/rs6000.c (paired_init_builtins): Add const
595 declaration to bdesc_paired_preds variable.
596 (paired_expand_builtin): Likewise.
598 2007-09-09 Revital Eres <eres@il.ibm.com>
600 * dbgcnt.def (sms_sched_loop): New counter.
601 * modulo-sched.c: Use sms_sched_loop instead of
602 MAX_SMS_LOOP_NUMBER to determine the maximum number of loops to
603 perform swing modulo scheduling on. Include dbgcnt.h.
604 * Makefile.in: Add DBGCNT_H to modulo-sched.o.
605 * params.def: Remove PARAM_MAX_SMS_LOOP_NUMBER.
607 2007-09-09 Uros Bizjak <ubizjak@gmail.com>
609 * config/i386/i386.md (X87MODEF12, SSEMODEF): Remove mode iterators.
610 Substitute all uses with ...
611 (MODEF): New mode iterator.
613 (fix_trunc<mode>_fisttp_i387_1): Remove operand constraints
614 from pre-regalloc define_insn_and_split splitter pattern.
615 (*fix_trunc<mode>_i387_1): Ditto.
617 (*fist<mode>2_1): Ditto.
618 (frndintxf2_floor): Ditto.
619 (*fist<mode>2_floor_1): Ditto.
620 (frndintxf2_ceil): Ditto.
621 (*fist<mode>2_ceil_1): Ditto.
622 (frndintxf2_trunc): Ditto.
623 (frndintxf2_mask_pm): Ditto.
625 (prologue): Use (const_int 0) as never generated filler insn.
627 (sibcall_epilogue): Ditto.
628 (eh_return_si): Ditto.
629 (eh_return_di): Ditto.
631 (add<mode>3): Rename from adddf3 and addsf3. Macroize expander
632 using MODEF mode iterator.
633 (sub<mode>3): Rename from subdf3 and subsf3. Macroize expander
634 using MODEF mode iterator.
635 (mul<mode>3): Rename from muldf3 and mulsf3. Macroize expander
636 using MODEF mode iterator.
637 (nearbyint<mode>2): Rename from nearbyintdf2 and nearbyintsf2.
638 Macroize expander using MODEF mode iterator.
640 (zero_extendsidi2): Remove operand constraints from expander.
641 (smuldi3_highpart): Ditto.
642 (indirect_jump): Ditto.
645 * config/i386/sse.md (storentv4sf): Ditto.
646 (storentv2df): Ditto.
647 (storentv2di): Ditto.
649 (sse2_cvtpd2ps): Ditto.
650 (vec_interleave_highv16qi): Ditto.
651 (vec_interleave_lowv16qi): Ditto.
652 (vec_interleave_highv8hi): Ditto.
653 (vec_interleave_lowv8hi): Ditto.
654 (vec_interleave_highv4si): Ditto.
655 (vec_interleave_lowv4si): Ditto.
656 (vec_interleave_highv2di): Ditto.
657 (vec_interleave_lowv2di): Ditto.
658 (sse2_maskmovdqu): Ditto.
659 * config/i386/mmx.md (mmx_maskmovq): Ditto.
661 2007-09-09 Ira Rosen <irar@il.ibm.com>
663 * tree-vectorizer.h (enum vect_def_type): Start enumeration from 1.
664 (struct _slp_tree, struct _slp_instance): Define new data structures
665 along macros for their access.
666 (struct _loop_vec_info): Define new fields: strided_stores,
667 slp_instances, and slp_unrolling_factor along macros for their access.
668 (enum slp_vect_type): New.
669 (struct _stmt_vec_info): Define new field, slp_type, and macros for its
671 (STMT_VINFO_STRIDED_ACCESS): New macro.
672 (vect_free_slp_tree): Declare.
673 (vectorizable_load): Add an argument of type slp_tree.
674 (vectorizable_store, vectorizable_operation, vectorizable_conversion,
675 vectorizable_assignment): Likewise.
676 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
677 Declare (make extern).
678 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
679 (new_loop_vec_info): Likewise.
680 (destroy_loop_vec_info): Free memory allocated for SLP structures.
681 * tree-vect-analyze.c: Include recog.h.
682 (vect_update_slp_costs_according_to_vf): New.
683 (vect_analyze_operations): Add argument for calls to vectorizable_ ()
684 functions. For not pure SLP stmts with strided access check that the
685 group size is power of 2. Update the vectorization factor according to
686 SLP. Call vect_update_slp_costs_according_to_vf.
687 (vect_analyze_group_access): New.
688 (vect_analyze_data_ref_access): Call vect_analyze_group_access.
689 (vect_free_slp_tree): New functions.
690 (vect_get_and_check_slp_defs, vect_build_slp_tree, vect_print_slp_tree,
691 vect_mark_slp_stmts, vect_analyze_slp_instance, vect_analyze_slp,
692 vect_make_slp_decision, vect_detect_hybrid_slp_stmts,
693 vect_detect_hybrid_slp): Likewise.
694 (vect_analyze_loop): Call vect_analyze_slp, vect_make_slp_decision
695 and vect_detect_hybrid_slp.
696 * tree-vect-transform.c (vect_estimate_min_profitable_iters): Take
697 SLP costs into account.
698 (vect_get_cost_fields): New function.
699 (vect_model_simple_cost): Make extern, add SLP parameter and handle
701 (vect_model_store_cost, vect_model_load_cost): Likewise.
702 (vect_get_constant_vectors): New function.
703 (vect_get_slp_vect_defs, vect_get_slp_defs,
704 vect_get_vec_defs_for_stmt_copy, vect_get_vec_defs_for_stmt_copy,
705 vect_get_vec_defs): Likewise.
706 (vectorizable_reduction): Don't handle SLP for now.
707 (vectorizable_call): Don't handle SLP for now. Add argument to
708 vect_model_simple_cost.
709 (vectorizable_conversion): Handle SLP (call vect_get_vec_defs to
710 get SLPed and vectorized defs). Fix indentation and spacing.
711 (vectorizable_assignment): Handle SLP.
712 (vectorizable_induction): Don't handle SLP for now.
713 (vectorizable_operation): Likewise.
714 (vectorizable_type_demotion): Add argument to
715 vect_model_simple_cost.
716 (vectorizable_type_promotion): Likewise.
717 (vectorizable_store, vectorizable_load): Handle SLP.
718 (vectorizable_condition): Don't handle SLP for now.
719 (vect_transform_stmt): Add a new argument for SLP. Check that there is
720 no SLP transformation required for unsupported cases. Add SLP
721 argument for supported cases.
722 (vect_remove_stores): New function.
723 (vect_schedule_slp_instance, vect_schedule_slp): Likewise.
724 (vect_transform_loop): Schedule SLP instances.
725 * Makefile.in: (tree-vect-analyze.o): Depend on recog.h.
727 2007-09-09 Andrew Haley <aph@redhat.com>
729 * optabs.c (sign_expand_binop): Set libcall_gen = NULL in the
732 2007-09-09 Hans-Peter Nilsson <hp@axis.com>
734 Divide REG_LABEL notes into REG_LABEL_OPERAND and REG_LABEL_TARGET.
735 * doc/rtl.texi (Insns): Specify when a label_ref makes a jump_insn.
736 Similar for what label_refs can go in the JUMP_TARGET field. Split
737 REG_LABEL documentation into REG_LABEL_TARGET and REG_LABEL_OPERAND.
738 * reload.c (find_reloads): Generate REG_LABEL_OPERAND, not
739 REG_LABEL when replacing an operand with a LABEL_REF for a
741 (subst_reloads): When replacing a LABEL_REG with a register,
742 instead of generating a REG_LABEL note, assert that there already
743 is one or that the label is a known target for the insn.
744 * rtlanal.c (computed_jump_p): Instead of looking for a REG_LABEL
745 note, check the JUMP_LABEL field. Remove "else" after return.
746 * reorg.c (emit_delay_sequence): Replace case for REG_LABEL with
747 cases for REG_LABEL_OPERAND and REG_LABEL_TARGET.
748 (fill_slots_from_thread): Handle both REG_LABEL_OPERAND and
749 REG_LABEL_TARGET notes, including the JUMP_TARGET field on JUMP_P
750 insns. Iterate over all notes; don't assume there's only one.
751 * cse.c (recorded_label_ref): Adjust comment to refer to
753 (cse_extended_basic_block): Do LABEL_REF check for all INSN_P
754 insns, not just NONJUMP_INSN_P.
755 (check_for_label_ref): For JUMP_P insns, check that the LABEL_REF
757 * jump.c (rebuild_jump_labels): Adjust head comment.
758 (init_label_info): Ditto. Remove REG_LABEL_OPERAND notes only;
759 don't reset REG_LABEL_TARGET notes, including the JUMP_LABEL field.
760 (mark_all_labels): For JUMP_P insns without a target, check if the
761 the target is noted on the previous nonjump insn.
762 (mark_jump_label_1): New function, guts from mark_jump_label.
763 <case IF_THEN_ELSE>: Handle first operand as a non-target when
764 marking jump target labels.
765 <case LABEL_REF>: Adjust for whether to generate a
766 REG_LABEL_TARGET or a REG_LABEL_OPERAND note.
767 For 'E' format rtl, iterate in descending element order.
768 (delete_related_insns): Handle both REG_LABEL_TARGET and
769 REG_LABEL_OPERAND notes. For JUMP_P insns with labels with zero
770 reference count, delete and fallthrough. Move finding-next-
771 non-deleted insn last in the function. Look at all INSN_P insns
772 for REG_LABEL_OPERAND notes.
773 (redirect_jump_2): Assert that OLABEL equals the old JUMP_LABEL of
775 * print-rtl.c (print_rtx): For JUMP_P insns and a non-NULL
776 JUMP_LABEL, output the INSN_UID of it.
777 * gcse.c: Adjust comments as appropriate to say REG_LABEL_OPERAND
778 and/or REG_LABEL_TARGET.
779 (add_label_notes): Only add REG_LABEL_OPERAND notes. Put in line
780 with jump.c copy by only adding notes for labels actually
781 referenced in the insn.
782 * emit-rtl.c (try_split): Don't assume only NONJUMP_INSN_P need
783 usage count increment; handle all INSN_P trial insns.
784 (emit_copy_of_insn_after): Change to not copy REG_LABEL_OPERAND
786 * rtl.h (struct rtx_def) <volatil>: Adjust to mention
787 REG_LABEL_TARGET and REG_LABEL_OPERAND.
788 (LABEL_REF_NONLOCAL_P): Allow REG_LABEL_TARGET and
790 * combine.c (distribute_notes): Adjust for REG_LABEL_TARGET on
791 JUMP_P insns and REG_LABEL_OPERAND everywhere.
792 * sched-rgn.c (is_cfg_nonregular): Check for REG_LABEL_OPERANDS
794 * reg-notes.def (LABEL_TARGET, LABEL_OPERAND): Split from LABEL.
795 * cfgrtl.c (delete_insn): Adjust to handle REG_LABEL_TARGET and
796 REG_LABEL_OPERAND notes.
797 * reload1.c (calculate_needs_all_insns): Adjust comments.
798 (set_label_offsets): Adjust to look for REG_LABEL_OPERAND notes.
799 * config/alpha/alpha.md (split for load of an address into a
800 four-insn sequence on Unicos/Mk): Adjust to use
802 * config/sh/sh.md (sh_reorg, final_prescan_insn): Ditto.
804 2007-09-09 Laurynas Biveinis <laurynas.biveinis@gmail.com>
807 2007-09-05 Laurynas Biveinis <laurynas.biveinis@gmail.com>
808 * regrename.c (copyprop_hardreg_forward_1): New variable next. Use
809 FOR_BB_INSNS_SAFE instead of for loop.
810 * cse.c (cse_extended_basic_block): Likewise.
811 * postreload.c (reload_cse_regs_1): New variable next. Make sure
812 that the for loop does not invoke NEXT_INSN on a deleted insn.
813 * function.c (instantiate_virtual_regs): Likewise.
814 * lower-subreg.c (remove_retval_note): Likewise.
815 (decompose_multiword_subregs): Use FOR_BB_INSNS_SAFE instead of
817 * emit-rtl.c (remove_insn): Set NEXT_INSN and PREV_INSN to NULL on
819 * cfgrtl.c (delete_insn): Set JUMP_LABEL to NULL on a deleted
820 insn, if it's a jump.
821 (try_redirect_by_replacing_jump): New variable jump_p. Call
822 tablejump_p before delete_insn_chain.
823 * reload1.c (reload): New variable next. Make sure that the for
824 loop does not invoke NEXT_INSN on a deleted insn.
825 (fixup_eh_region_note): Make the loop terminate if i becomes NULL.
826 (delete_output_reload): New variable prev. Make sure the the for
827 loops do not invoke PREV_INSN on a deleted insn.
829 2007-09-08 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
831 * pa/constraints.md: Remove 'X' from unused letters comment.
833 2007-09-08 Richard Guenther <rguenther@suse.de>
835 * tree-tailcall.c (find_tail_calls): If we don't have aliases
836 computed check stmt_ann->references_memory instead of counting
839 2007-09-08 Segher Boessenkool <segher@kernel.crashing.org>
841 * cse.c (fold_rtx): Use validate_unshare_change() instead of
842 validate_change() in one more case.
844 2007-09-08 Zdenek Dvorak <ook@ucw.cz>
846 PR tree-optimization/32283
847 * tree-ssa-loop-ivopts.c (may_eliminate_iv): Use
848 estimated_loop_iterations.
849 (determine_use_iv_cost_condition): Decrease cost of expressions
850 used in iv elimination.
852 2007-09-08 Richard Guenther <rguenther@suse.de>
854 * tree-cfg.c (verify_gimple_expr): Avoid building new
855 pointer types, use TYPE_POINTER_TO if available instead.
857 2007-09-08 Uros Bizjak <ubizjak@gmail.com>
860 PR rtl-optimization/26449
861 * config/i386/sse.md (mulv4si3): Do not expand sse2 sequence.
862 (*sse2_mulv4si3): New define_insn_and_split pattern. Split insn in
864 (mulv16qi3): Implement as define_insn_and_split pattern instead of
865 define_expand. Split insn in split1 pass.
868 2007-09-08 Dorit Nuzman <dorit@il.ibm.com>
870 PR tree-optimization/33301
871 * tree-vect-analyze (analyze_operations): Look at the type of the rhs
874 2007-09-07 Zdenek Dvorak <ook@ucw.cz>
876 PR tree-optimization/32183
877 * Makefile.in (tree-ssa-reassoc.o): Also depend on $(CFGLOOP_H).
879 * tree-ssa-reassoc.c: Include cfgloop.h.
880 (is_reassociable_op): Add a loop argument and return true only
882 (linearize_expr): Updated.
883 (should_break_up_subtract): Likewise.
884 (linearize_expr_tree): Likewise.
885 (init_reassoc): Call loop_optimizer_init with
886 AVOID_CFG_MODIFICATIONS. Remove calculate_dominance_info call
888 (fini_reassoc): Call loop_optimizer_finalize.
890 2007-09-07 Sterling Augustine <sterling@tensilica.com>
892 * config/xtensa/lib2funcs.S (__xtensa_sync_caches): Use an ISYNC even
893 if there is no i-cache.
895 2007-09-07 Richard Guenther <rguenther@suse.de>
898 2007-09-06 Richard Guenther <rguenther@suse.de>
900 PR tree-optimization/32586
901 * tree-ssa-sccvn.c (simplify_binary_expression): Avoid
902 folding if nothing changed.
903 (simplify_unary_expression): New function. Do tree combining
904 on conversion like codes.
905 (try_to_simplify): Call it.
907 2007-09-07 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
910 * gthr-posix.h (__gthread_active_p): Add implementation for hppa-hpux.
911 (__gthread_active,__gthread_start, __gthread_active_init): New.
912 * gthr-posix95.h: Likewise.
914 2007-09-07 Roman Zippel <zippel@linux-m68k.org>
916 * function.h (struct function): Rename calls_unwind_init
917 to saves_all_registers.
918 (current_function_saves_all_registers): Renamed from
919 current_function_calls_unwind_init.
920 * reload1.c (reload): Set current_function_saves_all_registers.
921 * except.c (expand_builtin_unwind_init): Likewise.
922 * config/m68k/m68k.c (m68k_save_reg): Use
923 current_function_saves_all_registers to save pic register.
925 2007-09-07 Janis Johnson <janis187@us.ibm.com>
927 config/m68k/m68k.c (floating_exact_log2): Update call to real_2expN.
928 config/s390/s390.md (fixuns_trunc<BFP:mode><GPR:mode>2): Ditto.
930 Reapply reverted changes:
932 2007-09-06 Jan Hubicka <jh@suse.cz>
934 * config/i386.c (ix86_expand_lround, ix86_expand_round): Update call of
937 2007-09-06 Richard Sandiford <richard@codesourcery.com>
939 * config/mips/mips.md (fixuns_truncdfsi2, fixuns_truncdfdi2)
940 (fixuns_truncsfsi2, fixuns_truncsfdi2): Update calls to real_2expN.
942 2007-09-05 Janis Johnson <janis187@us.ibm.com>
944 * optabs.c (expand_float): Convert unsigned integer as signed only
945 if it provides sufficient accuracy; add mode argument to real_2expN.
946 (expand_fix): Fix comment typos; extend binary float into mode
947 wider than destination for converion to unsigned integer; add mode
948 argument to real_2expN.
949 * real.c (real_2expN): Add mode argument to special-case decimal
951 * real.h (real_2expN): Ditto.
952 * fixed-value.c (check_real_for_fixed_mode): Add mode argument to
954 (fixed_from_string): Ditto.
955 (fixed_to_decimal): Ditto.
956 (fixed_convert_from_real): Ditto.
957 (real_convert_from_fixed): Ditto.
958 * config/rs6000/rs6000.md (FP): Include DD and TD modes.
959 * config/rs6000/dfp.md (extendddtd2, adddd3, addtd3, subdd3, subtd3,
960 muldd3, multd3, divdd3, divtd3, cmpdd_internal1, cmptd_internal1,
961 floatditd2, ftruncdd2, fixdddi2, ftrunctd2, fixddi2): New.
963 2007-09-07 Diego Novillo <dnovillo@google.com>
965 * tree-flow.h (const_block_stmt_iterator): Remove.
966 Update all users to use block_stmt_iterator.
967 * tree-iterator.h (const_tree_stmt_iterator): Remove.
968 Update all users to use tree_stmt_iterator.
970 2007-09-07 Sandra Loosemore <sandra@codesourcery.com>
972 * config/mips/mips.c (mips_set_current_function): Temporarily
973 make this a no-op to fix bootstrap errors, pending rewrite.
975 2007-09-07 Jan Hubicka <jh@suse.cz>
977 * reorg.c (dbr_schedule): Move code removing placeholder USEs later
980 2007-09-07 Dorit Nuzman <dorit@il.ibm.com>
982 PR tree-optimization/33299
983 * tree-vect-transform.c (vect_create_epilog_for_reduction): Update
984 uses for all relevant loop-exit phis, not just the first.
986 2007-09-07 Richard Guenther <rguenther@suse.de>
989 * tree-ssa-operands.c (access_can_touch_variable): An access
990 of the form (*p)[0] can touch a variable of same size.
992 2007-09-07 Jan Hubicka <jh@suse.cz>
994 * passes.c (init_optimization_passes): Add simple dce and addressable
996 * tree-ssa.c (execute_update_addresses_taken): New function.
997 (pass_update_address_taken): New.
998 * tree-ssa-dse.c (execute_simple_dse): New function.
999 (pass_simple_dse): New.
1000 * tree-pass.h (pass_simple_dse, pass_update_address_taken): Declare.
1002 2007-09-07 Tobias Burnus <burnus@net-b.de>
1005 * doc/invoke.texi: Fix -Wstrict-overflow= table.
1007 2007-09-07 Richard Guenther <rguenther@suse.de>
1010 2007-09-06 Richard Guenther <rguenther@suse.de>
1012 PR tree-optimization/32586
1013 * tree-ssa-sccvn.c (simplify_binary_expression): Avoid
1014 folding if nothing changed.
1015 (simplify_unary_expression): New function. Do tree combining
1016 on conversion like codes.
1017 (try_to_simplify): Call it.
1019 2007-09-07 Richard Guenther <rguenther@suse.de>
1020 Uros Bizjak <ubizjak@gmail.com>
1022 PR tree-optimization/32821
1023 * tree_if_conv.c (combine_blocks): Use alloc_stmt_list instead of
1024 NULL in the call to set_bb_stmt_list.
1026 2007-09-07 Richard Sandiford <richard@codesourcery.com>
1028 * config/mips/mips.c (build_mips16_call_stub): Emit all direct
1029 float calls here, rather than leaving some to the caller.
1030 Use call_internal_direct and call_value_internal_direct.
1031 * config/mips/mips.md (call_internal_direct): New pattern.
1032 (call_value_internal_direct): Likewise.
1034 2007-09-07 Richard Sandiford <richard@codesourcery.com>
1036 * config/mips/mips.c (mips_base_move_loop_invariants): New variable.
1037 (mips_set_mips16_mode): Restore flag_move_loop_invariants, then set
1039 (override_options): Set mips_base_move_loop_invariants.
1041 2007-09-07 Richard Sandiford <richard@codesourcery.com>
1043 * config/mips/mips.opt (mcode-readable=): Move to keep list
1046 2007-09-07 Richard Sandiford <richard@codesourcery.com>
1048 * doc/invoke.texi: Document the MIPS -mlocal-sdata, -mextern-sdata
1049 and -mgpopt options. Adjust the -G documentation to match.
1050 * config/mips/mips.opt (mextern-data, mgpopt, mlocal-data): New.
1051 * config/mips/mips.c (mips_rtx_constant_in_small_data_p): New
1052 function, split out from mips_classify_symbol. Return false for
1053 !TARGET_LOCAL_SDATA.
1054 (mips_classify_symbol): Call mips_rtx_constant_in_small_data_p.
1055 Only use GP-relative accesses for non-pool symbols if TARGET_GPOPT.
1056 (override_options): Check whether the -mgpopt setting is consistent
1057 with the other flags.
1058 (symbolic_expression_p): Delete.
1059 (mips_select_rtx_section): Use mips_rtx_constant_in_small_data_p
1060 and default_elf_select_rtx_section.
1061 (mips_in_small_data_p): Honor section attributes for MIPS16 too.
1062 Return false for local data unless TARGET_LOCAL_SDATA. Likewise
1063 external data and TARGET_EXTERN_SDATA.
1065 2007-09-07 Tobias Burnus <burnus@net-b.de>
1068 * doc/cpp.texi (Common Predefined Macros): Add __GFORTRAN__.
1070 2007-09-05 Laurynas Biveinis <laurynas.biveinis@gmail.com>
1072 * regrename.c (copyprop_hardreg_forward_1): New variable next. Use
1073 FOR_BB_INSNS_SAFE instead of for loop.
1074 * cse.c (cse_extended_basic_block): Likewise.
1075 * postreload.c (reload_cse_regs_1): New variable next. Make sure
1076 that the for loop does not invoke NEXT_INSN on a deleted insn.
1077 * function.c (instantiate_virtual_regs): Likewise.
1078 * lower-subreg.c (remove_retval_note): Likewise.
1079 (decompose_multiword_subregs): Use FOR_BB_INSNS_SAFE instead of
1081 * emit-rtl.c (remove_insn): Set NEXT_INSN and PREV_INSN to NULL on
1083 * cfgrtl.c (delete_insn): Set JUMP_LABEL to NULL on a deleted
1084 insn, if it's a jump.
1085 (try_redirect_by_replacing_jump): New variable jump_p. Call
1086 tablejump_p before delete_insn_chain.
1087 * reload1.c (reload): New variable next. Make sure that the for
1088 loop does not invoke NEXT_INSN on a deleted insn.
1089 (fixup_eh_region_note): Make the loop terminate if i becomes NULL.
1090 (delete_output_reload): New variable prev. Make sure the the for
1091 loops do not invoke PREV_INSN on a deleted insn.
1093 2007-09-06 Zdenek Dvorak <ook@ucw.cz>
1095 * cgraphbuild.c (rebuild_cgraph_edges): Export.
1096 * cgraph.h (rebuild_cgraph_edges): Declare.
1097 * tree-pass.h (pass_expand_omp_ssa): New.
1098 * omp-low.c (find_omp_clause): Export.
1099 (copy_var_decl): Split from omp_copy_decl_2.
1100 (build_omp_barrier): Return the call to emit instead of emitting
1102 (lower_rec_input_clauses, expand_omp_single): Gimplify the result of
1104 (extract_omp_for_data, expand_parallel_call, expand_omp_parallel,
1105 expand_omp_for_generic, expand_omp_for_static_nochunk,
1106 expand_omp_for_static_chunk, expand_omp_for, expand_omp_sections):
1107 Adapted to work on SSA form.
1108 (execute_expand_omp): Do not invalidate dominance information.
1109 (gate_expand_omp): Do not run with -fopenmp-ssa flag.
1110 (gate_expand_omp_ssa, pass_expand_omp_ssa): New.
1111 * gimplify.c (gimplify_omp_for): Ensure that the control variable is
1113 (force_gimple_operand): Allow gimplifying code expressions without
1115 * tree-predcom.c (mark_virtual_ops_for_renaming): Handle phi nodes.
1116 * common.opt (fopenmp-ssa): New.
1117 * tree-flow.h (find_omp_clause, copy_var_decl): Declare.
1118 * Makefile.in (tree-cfg.o): Add TREE_INLINE_H dependency.
1119 * tree-cfg.c: Include tree-inline.h.
1120 (struct move_stmt_d): Replace vars_to_remove by vars_map field.
1121 (replace_by_duplicate_decl, replace_ssa_name,
1122 mark_virtual_ops_in_region): New functions.
1123 (move_stmt_r, move_block_to_fn, move_sese_region_to_fn): Adapted
1124 to work on SSA form.
1125 * passes.c (init_optimization_passes): Add pass_expand_omp_ssa pass.
1126 * tree-ssa-operands.c (get_expr_operands): Handle operands of OMP
1129 2007-09-06 Laurynas Biveinis <laurynas.biveinis@gmail.com>
1131 * tree-loop-linear.c: Include obstack.h.
1132 (linear_transform_loops): New obstack lambda_obstack.
1133 Initialize it, pass it to gcc_loopnest_to_lambda_loopnest,
1134 lambda_loopnest_transform, lambda_loopnest_to_gcc_loopnest calls
1135 and free afterwards.
1136 * lambda.h (struct obstack): New forward declaration.
1137 (lambda_linear_expression_new): New parameter of type struct
1139 (lambda_loopnest_new): Likewise.
1140 (lambda_loopnest_transform): Likewise.
1141 (lambda_body_vector_new): Likewise.
1142 (lambda_body_vector_compute_new): Likewise.
1143 (gcc_loopnest_to_lambda_loopnest): Likewise.
1144 (lambda_loopnest_to_gcc_loopnest): Likewise.
1145 * lambda-code.c: Include obstack.h.
1146 (lambda_lattice_new): New parameter lambda_obstack. Use it for
1148 (lambda_body_vector_new): Likewise.
1149 (lambda_linear_expression_new): Likewise.
1150 (lambda_lattice_new): Likewise.
1151 (lambda_loopnest_new): Likewise. Additionally use obstack to
1152 allocate LN_LOOPS(ret).
1153 (lambda_lattice_compute_base): New parameter lambda_obstack. Pass
1154 it to lambda_lattice_new.
1155 (lambda_body_vector_compute_new): New parameter lambda_obstack.
1156 Pass it to lambda_body_vector_new.
1157 (lambda_lattice_compute_base): New paramater lambda_obstack. Pass
1158 it to lambda_lattice_new.
1159 (compute_nest_using_fourier_motzkin): New parameter lambda_obstack.
1160 Pass it to lambda_loopnest_new, lambda_linear_expression_new.
1161 (lambda_compute_target_space): Likewise.
1162 (lambda_compute_auxillary_space): New parameter lambda_obstack.
1163 Pass it to lambda_lattice_compute_base and
1164 compute_nest_using_fourieer_motzkin.
1165 (lambda_loopnest_transform): New parameter lambda_obstack. Pass
1166 it to lambda_lattice_compute_base, lambda_lattice_auxillary_space
1167 and lambda_lattice_compute_target_space.
1168 (gcc_tree_to_linear_expression): Nex parameter lambda_obstack.
1169 Pass it to lambda_linear_expression_new.
1170 (gcc_loop_to_lambda_loop): New parameter lambda_obstack. Pass it
1171 to gcc_tree_to_linear_expression.
1172 (gcc_loopnest_to_lambda_loopnest): New parameter lambda_obstack.
1173 Pass it to gcc_loop_to_lambda_loop and lambda_loopnest_new.
1174 (lambda_loopnest_to_gcc_loopnest): New parameter lambda_obstack.
1175 Pass it to lambda_body_vector_new and
1176 lambda_body_vector_compute_new.
1177 * Makefile.in (tree-loop-linear.o): Add $(OBSTACK_H) dependency.
1178 (lambda-code.o): Likewise.
1180 2007-09-06 Chao-ying Fu <fu@mips.com>
1182 * ginclude/stdfix.h: New file.
1183 * Makefile.in (USER_H): Add $(srcdir)/ginclude/stdfix.h.
1184 (convert.o): Add dependence on fixed-value.h.
1185 * c-convert.c (convert): Support FIXED_POINT_TYPE.
1186 * c-cppbuiltin.c (builtin_define_fixed_point_constants): New function
1187 to define fixed-point constants.
1188 (c_cpp_builtins): Define fixed-point constants.
1189 * convert.c (fixed-value.h): New include.
1190 (convert_to_real): Update comment to include fixed-point.
1191 Support FIXED_POINT_TYPE.
1192 (convert_to_integer): Update comment to include fixed-point.
1193 Support FIXED_POINT_TYPE.
1194 (convert_to_complex): Support FIXED_POINT_TYPE.
1195 (convert_to_fixed): New function.
1196 * convert.h (convert_to_fixed): Declare.
1197 * genopinit.c: Add comment about $Q for only fixed-point modes.
1198 (optabs): Add fract_optab, fractuns_optab, satfract_optab,
1199 satfractuns_optab, add_optab, ssadd_optab, usadd_optab, sub_optab,
1200 sssub_optab, ussub_optab, smul_optab, ssmul_optab, usmul_optab,
1201 ssmadd_widen_optab, usmadd_widen_optab, ssdiv_optab, udiv_optab,
1202 usdiv_optab, ssashl_optab, usashl_optab, neg_optab, ssneg_optab,
1203 usneg_optab for fixed-point modes.
1204 (gen_insn): Add force_fixed to track the $Q format for all fixed-point
1206 * optabs.c (optab_for_tree_code): For *DIV_EXPR, LSHIFT_EXPR,
1207 PLUS_EXPR, MINUS_EXPR, MULT_EXPR, NEGATE_EXPR, return signed or
1208 unsigned saturation optabs, when type is saturating.
1209 (shift_optab_p): Return true for SS_ASHIFT or US_ASHIFT.
1210 (expand_fixed_convert): New function.
1211 (gen_fixed_libfunc, gen_signed_fixed_libfunc,
1212 gen_unsigned_fixed_libfunc, gen_int_fp_fixed_libfunc,
1213 gen_int_fp_signed_fixed_libfunc, gen_int_fixed_libfunc,
1214 gen_int_signed_fixed_libfunc, gen_int_unsigned_fixed_libfunc,
1215 gen_fract_conv_libfunc, gen_fractuns_conv_libfunc,
1216 gen_satfract_conv_libfunc, gen_satfractuns_conv_libfunc): New
1218 (init_optabs): Initialize ssadd_optab, usadd_optab, sssub_optab,
1219 ussub_optab, ssmul_optab, usmul_optab, ssmadd_widen_optab,
1220 usmadd_widen_optab, ssmsub_widen_optab, usmsub_widen_optab,
1221 ssdiv_optab, usdiv_optab, ssashl_optab, usashl_optab, ssneg_optab,
1222 usneg_optab, fract_optab, fractuns_optab, satfract_optab,
1224 Initialize fixed-point libraries, including add, ssadd, usadd, sub,
1225 sssub, ussub, mul, ssmul, usmul, div, ssdiv, udiv, usdiv, ashl,
1226 ssashl, usashl, ashr, lshr, neg, ssneg, usneg, cmp, fract, satfract,
1227 fractuns, satfractuns.
1228 * optabs.h (enum optab_index): Add OTI_ssadd, OTI_usadd, OTI_sssub,
1229 OTI_ussub, OTI_ssmul, OTI_usmul, OTI_ssdiv, OTI_usdiv, OTI_ssneg,
1230 OTI_usneg, OTI_ssashl, OTI_usashl, OTI_ssmadd_widen, OTI_usmadd_widen,
1231 OTI_ssmsub_widen, OTI_usmsub_widen.
1232 (ssadd_optab, usadd_optab, sssub_optab, ussub_optab, ssmul_optab,
1233 usmul_optab, ssdiv_optab, usdiv_optab, ssneg_optab, usneg_optab,
1234 ssashl_optab, usashl_optab, ssmadd_widen_optab, usmadd_widen_optab,
1235 umsub_widen_optab, usmsub_widen_optab): Define.
1236 (enum convert_optab_index): Add COI_fract, COI_fractuns, COI_satfract,
1238 (fract_optab, fractuns_optab, satfract_optab, satfractuns_optab):
1240 (expand_fixed_convert): Declare.
1241 * expr.c (convert_move): Support the move of fixed-point modes.
1242 (emit_move_insn_1): Handle fixed-point mode to move via integer.
1243 (categorize_ctor_elements_1): Handle FIXED_CST.
1244 (count_type_elements): Handle FIXED_POINT_TYPE.
1245 (expand_expr_real_1): For VECTOR_CST, check MODE_VECTOR_FRACT,
1246 MODE_VECTOR_UFRACT, MODE_VECTOR_ACCUM, MODE_VECTOR_UACCUM.
1248 For PLUS_EXPR and MINUS_EXPR, support saturating and non-saturating
1249 multiply and add/subtract for fixed-point types.
1250 For MULT_EXPR, *DIV_EXPR, *SHIFT_EXPR, if the mode if a fixed-point
1251 mode, we jump to binop directly.
1252 Support FIXED_CONVERT_EXPR.
1253 (do_store_flag): Check FIXED_CST to put a constant second.
1254 (vector_mode_valid_p): Handle MODE_VECTOR_FRACT,
1255 MODE_VECTOR_UFRACT, MODE_VECTOR_ACCUM, MODE_VECTOR_UACCUM.
1256 (const_vector_from_tree): Support FIXED_CST.
1257 * doc/extend.texi (Fixed-Point): New node.
1258 * doc/md.texi (ssadd, usadd, sssub, ussub, ssmul, usmul, ssdiv, usdiv,
1259 ssmadd, usmadd, ssmsub, usmsub, ssashl, usashl, ssneg, usneg, fract,
1260 satfract, fractuns, satfractuns): Document them.
1262 2007-09-07 Bernd Schmidt <bernd.schmidt@analog.com>
1264 * config/bfin/bfin.h (PREFERRED_RELOAD_CLASS): Don't reload autoinc
1265 addresses into I registers.
1267 2007-09-06 Jan Hubicka <jh@suse.cz>
1268 Andreas Tobler <a.tobler@schweiz.org>
1270 * config/darwin.c (machopic_indirect_data_reference): Avoid invalid
1272 (machopic_legitimize_pic_address): Likewise.
1274 2007-09-06 Andrew Pinski <andrew_pinski@playstation.sony.com>
1275 Jan Hubicka <jh@suse.cz>
1277 * config/spu/spu.md (floatsidf2): Use convert_optab_libfunc
1278 instead of ufloat_optab->handlers directly.
1279 (floatdidf2): Likewise.
1281 2007-09-06 Sandra Loosemore <sandra@codesourcery.com>
1283 * config/mips/mips.c: Include diagnostic.h.
1284 (mips_set_current_function): Check errorcount and sorrycount
1285 before generating RTL.
1287 2007-09-06 Richard Sandiford <richard@codesourcery.com>
1290 * config/mips/mips.c (mips_classify_symbolic_expression): New function.
1291 (mips_classify_address): Use it instead of mips_symbolic_constant_p.
1292 (print_operand_reloc): Likewise.
1294 2007-09-06 Janis Johnson <janis187@us.ibm.com>
1298 2007-09-06 Jan Hubicka <jh@suse.cz>
1300 * i386.c (ix86_expand_lround, ix86_expand_round): Update call of
1303 2007-09-06 Richard Sandiford <richard@codesourcery.com>
1305 * config/mips/mips.md (fixuns_truncdfsi2, fixuns_truncdfdi2)
1306 (fixuns_truncsfsi2, fixuns_truncsfdi2): Update calls to real_2expN.
1308 2007-09-05 Janis Johnson <janis187@us.ibm.com>
1310 * optabs.c (expand_float): Convert unsigned integer as signed only
1311 if it provides sufficient accuracy; add mode argument to real_2expN.
1312 (expand_fix): Fix comment typos; extend binary float into mode
1313 wider than destination for converion to unsigned integer; add mode
1314 argument to real_2expN.
1315 * real.c (real_2expN): Add mode argument to special-case decimal
1317 * real.h (real_2expN): Ditto.
1318 * fixed-value.c (check_real_for_fixed_mode): Add mode argument to
1320 (fixed_from_string): Ditto.
1321 (fixed_to_decimal): Ditto.
1322 (fixed_convert_from_real): Ditto.
1323 (real_convert_from_fixed): Ditto.
1324 * config/rs6000/rs6000.md (FP): Include DD and TD modes.
1325 * config/rs6000/dfp.md (extendddtd2, adddd3, addtd3, subdd3, subtd3,
1326 muldd3, multd3, divdd3, divtd3, cmpdd_internal1, cmptd_internal1,
1327 floatditd2, ftruncdd2, fixdddi2, ftrunctd2, fixddi2): New.
1329 2007-09-06 Tom Tromey <tromey@redhat.com>
1331 * tree-cfg.c (remove_bb): Only warn if line is non-zero.
1332 * c-pch.c (c_common_read_pch): Restore current location after
1334 * tree.c (expand_location): Update.
1335 (expr_filename): Changed return type. Unified the two cases.
1336 (expr_lineno): Likewise.
1337 (annotate_with_file_line): Don't use EXPR_LINENO and EXPR_FILENAME
1339 * toplev.c (line_table): Changed type.
1340 (general_init): Update.
1341 (realloc_for_line_map): New function.
1342 (general_init): Allocate line_table using GC.
1343 * fix-header.c (line_table): Changed type.
1344 (read_scan_file): Update.
1345 (read_scan_file): Update.
1346 * c-ppoutput.c (maybe_print_line): Update.
1347 (print_line): Update.
1348 (cb_line_change): Update.
1349 (cb_define): Update.
1350 (pp_file_change): Update.
1351 * c-opts.c (c_common_init_options): Update.
1352 (finish_options): Update.
1353 (push_command_line_include): Update.
1354 * c-lex.c (cb_line_change): Update.
1355 (cb_def_pragma): Update.
1356 (cb_define): Update.
1358 (c_lex_with_flags): Use cpp_get_token_with_location.
1359 * input.h (line_table): Changed type.
1360 (location_from_locus): New macro.
1361 * tree.h (EXPR_FILENAME): No longer an lvalue.
1362 (EXPR_LINENO): Likewise.
1363 (expr_locus, set_expr_locus): Declare separately for
1364 USE_MAPPED_LOCATION.
1365 (expr_filename, expr_lineno): Changed return type.
1366 * gimplify.c (tree_to_gimple_tuple): Use SET_EXPR_LOCUS.
1367 * cfgexpand.c (expand_gimple_cond_expr): Use location_from_locus.
1368 (expand_gimple_basic_block): Likewise.
1369 * final.c (final_scan_insn): Use expanded_location.
1371 2007-09-06 Richard Guenther <rguenther@suse.de>
1373 PR tree-optimization/32586
1374 * tree-ssa-sccvn.c (simplify_binary_expression): Avoid
1375 folding if nothing changed.
1376 (simplify_unary_expression): New function. Do tree combining
1377 on conversion like codes.
1378 (try_to_simplify): Call it.
1379 * builtins.c (fold_builtin_cexp): Fold the built expressions.
1380 * fold-const.c (fold_unary): Test result of get_callee_fndecl().
1382 2007-09-06 Jan Hubicka <jh@suse.cz>
1385 * cse.c (fold_rtx): Avoid invalid sharing.
1387 2007-09-06 Richard Sandiford <richard@codesourcery.com>
1389 * config/mips/mips.md (fixuns_truncdfsi2, fixuns_truncdfdi2)
1390 (fixuns_truncsfsi2, fixuns_truncsfdi2): Update calls to real_2expN.
1392 2007-09-06 Revital Eres <eres@il.ibm.com>
1394 * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
1395 Fix offsettable memory reference for 750CL.
1397 2007-09-06 Pat Haugen <pthaugen@us.ibm.com>
1399 * reload.c (find_reloads_address_1): Try to preserve original
1400 base/index regclass of operands.
1401 * regrename.c (scan_rtx_address): Likewise.
1402 (replace_oldest_value_addr): Likewise.
1404 2007-09-06 Richard Guenther <rguenther@suse.de>
1406 PR tree-optimization/33302
1407 * domwalk.c (walk_dominator_tree): The exit block is
1408 interesting even if it is not reachable.
1410 2007-09-06 Richard Sandiford <richard@codesourcery.com>
1413 * optabs.c (avoid_expensive_constant): Do nothing if MODE is VOIDmode.
1415 2007-09-06 Basile Starynkevitch <basile@starynkevitch.net>
1417 * cfg.c (dump_bb_info, dump_edge_info): Added cfun test for
1420 2007-09-05 Ian Lance Taylor <iant@google.com>
1422 * tree-pretty-print.c (dump_decl_name): Cast LABEL_DECL_UID to int
1424 (dump_generic_node): Likewise.
1425 * print-rtl.c (print_decl_name): Likewise.
1426 * print-tree.c (print_node_brief): Likewise.
1427 (print_node): Likewise.
1428 * Makefile.in (RTL_BASE_H): Add alias.h.
1431 2007-09-06 Jesper Nilsson <jesper.nilsson@axis.com>
1433 * longlong.h [__CRIS_arch_version >= 8] (count_trailing_zeros):
1435 * config/cris/cris.md (ctzsi2, cris_swap_bits): Implemented.
1436 * config/cris/cris.h (CTZ_DEFINED_VALUE_AT_ZERO): Defined.
1438 2007-09-06 Jie Zhang <jie.zhang@analog.com>
1440 * config.gcc (tm_file): Add linux.h for bfin*-uclinux*.
1441 (tm_defines): Define UCLIBC_DEFAULT to 1.
1442 (extra_options): Add linux.opt.
1443 * config/bfin/linux.h (CPLUSPLUS_CPP_SPEC): Don't define.
1444 (CRT_CALL_STATIC_FUNCTION): Likewise.
1445 (NO_IMPLICIT_EXTERN_C): Likewise.
1446 (TARGET_OS_CPP_BUILTINS): Define as LINUX_TARGET_OS_CPP_BUILTINS.
1447 * config/bfin/elf.h (OBJECT_FORMAT_ELF): Don't define.
1448 * config/bfin/uclinux.h (CPLUSPLUS_CPP_SPEC): Don't define.
1449 (ENDFILE_SPEC): Don't define.
1450 (LIB_SPEC): Likewise.
1451 (CRT_CALL_STATIC_FUNCTION): Likewise.
1452 (NO_IMPLICIT_EXTERN_C): Likewise.
1453 (LINUX_TARGET_OS_CPP_BUILTINS): Likewise.
1454 (TARGET_OS_CPP_BUILTINS): Define as LINUX_TARGET_OS_CPP_BUILTINS.
1456 2007-09-06 Jan Hubicka <jh@suse.cz>
1458 * i386.c (ix86_expand_lround, ix86_expand_round): Update call of
1461 2007-09-06 Jan Hubicka <jh@suse.cz>
1463 * opts.c (common_handle_option): Enable inlining functions for
1466 2007-09-06 Jakub Jelinek <jakub@redhat.com>
1468 * builtin-attrs.def (ATTR_NONNULL_3): New.
1469 (DEF_FORMAT_ATTRIBUTE): Use just ATTR_NONNULL_##FA instead of
1470 ATTR_NOTHROW_NONNULL_##FA.
1471 (DEF_FORMAT_ATTRIBUTE_NOTHROW, DEF_FORMAT_ATTRIBUTE_BOTH): New macros.
1472 (ATTR_FORMAT_PRINTF_NOTHROW_2_0, ATTR_PRINTF_NOTHROW_2_3,
1473 ATTR_FORMAT_PRINTF_NOTHROW_3_0, ATTR_FORMAT_PRINTF_NOTHROW_3_4,
1474 ATTR_FORMAT_PRINTF_NOTHROW_4_0, ATTR_PRINTF_NOTHROW_4_5,
1475 ATTR_FORMAT_PRINTF_NOTHROW_5_0, ATTR_FORMAT_PRINTF_NOTHROW_5_6,
1476 ATTR_FORMAT_SCANF_NOTHROW_2_0, ATTR_FORMAT_SCANF_NOTHROW_2_3,
1477 ATTR_FORMAT_STRFTIME_NOTHROW_3_0, ATTR_FORMAT_STRFMON_NOTHROW_3_4):
1479 (ATTR_FORMAT_PRINTF_4_0, ATTR_PRINTF_4_5, ATTR_FORMAT_PRINTF_5_0,
1480 ATTR_FORMAT_PRINTF_5_6, ATTR_FORMAT_STRFTIME_3_0,
1481 ATTR_FORMAT_NOTHROW_3_4): Remove.
1482 * builtins.def (snprintf, sprintf, sscanf, vsnprintf, vsprintf,
1483 vsscanf, strfmon, strftime, __snprintf_chk, __sprintf_chk,
1484 __vsnprintf_chk, __vsprintf_chk): Use ATTR_FORMAT_*_NOTHROW_*
1485 instead of ATTR_FORMAT_*_*.
1487 2007-09-06 Hans-Peter Nilsson <hp@axis.com>
1489 * config/cris/cris.c (cris_emit_movem_store): Call copy_rtx on
1490 parts re-used for REG_FRAME_RELATED_EXPR.
1492 2007-09-05 Janis Johnson <janis187@us.ibm.com>
1494 * optabs.c (expand_float): Convert unsigned integer as signed only
1495 if it provides sufficient accuracy; add mode argument to real_2expN.
1496 (expand_fix): Fix comment typos; extend binary float into mode
1497 wider than destination for converion to unsigned integer; add mode
1498 argument to real_2expN.
1499 * real.c (real_2expN): Add mode argument to special-case decimal
1501 * real.h (real_2expN): Ditto.
1502 * fixed-value.c (check_real_for_fixed_mode): Add mode argument to
1504 (fixed_from_string): Ditto.
1505 (fixed_to_decimal): Ditto.
1506 (fixed_convert_from_real): Ditto.
1507 (real_convert_from_fixed): Ditto.
1508 * config/rs6000/rs6000.md (FP): Include DD and TD modes.
1509 * config/rs6000/dfp.md (extendddtd2, adddd3, addtd3, subdd3, subtd3,
1510 muldd3, multd3, divdd3, divtd3, cmpdd_internal1, cmptd_internal1,
1511 floatditd2, ftruncdd2, fixdddi2, ftrunctd2, fixddi2): New.
1513 2007-09-05 Ian Lance Taylor <iant@google.com>
1515 * init-regs.c (initialize_uninitialized_regs): Call
1516 emit_insn_before rather than add_insn_before.
1518 2007-09-05 Jakub Jelinek <jakub@redhat.com>
1520 * builtins.def (BUILT_IN_VA_ARG_PACK): New built-in.
1521 * tree.h (CALL_EXPR_VA_ARG_PACK): Define.
1522 * tree-inline.h (copy_body_data): Add call_expr field.
1523 * tree-inline.c (expand_call_inline): Initialize call_expr.
1524 (copy_bb): Append anonymous inline fn arguments to arguments
1525 when inlining a CALL_EXPR_VA_ARG_PACK call.
1526 * builtins.c (expand_builtin): Issue an error if
1527 BUILT_IN_VA_ARG_PACK is seen during expand.
1528 (fold_call_expr, fold_builtin_call_array): Don't fold
1529 CALL_EXPR_VA_ARG_PACK CALL_EXPRs or calls with
1530 __builtin_va_arg_pack () call as last argument.
1531 * gimplify.c (gimplify_call_expr): If last argument to a vararg
1532 function is __builtin_va_arg_pack (), decrease number of call
1533 arguments and instead set CALL_EXPR_VA_ARG_PACK on the CALL_EXPR.
1534 * expr.c (expand_expr_real_1): Issue an error if
1535 CALL_EXPR_VA_ARG_PACK CALL_EXPR is seen during expand.
1536 * tree-pretty-print.c (dump_generic_node): Handle printing
1537 CALL_EXPR_VA_ARG_PACK bit on CALL_EXPRs.
1538 * doc/extend.texi (__builtin_va_arg_pack): Document.
1540 2007-09-05 Adam Nemet <anemet@caviumnetworks.com>
1542 PR tree-optimization/21513
1543 * builtins.c (build_builtin_expect_predicate): New function.
1544 (fold_builtin_expect): Add argument for expected value.
1545 Distribute __builtin_expect over short-circuiting operations.
1546 Fold nested builtin_expects.
1547 (fold_builtin_2): Adjust call to fold_builtin_expect.
1549 2007-09-05 Rask Ingemann Lambertsen <rask@sygehus.dk>
1552 PR tree-optimization/13756
1553 * doc/invoke.texi (Options That Control Optimization): Document
1556 2007-09-05 Uros Bizjak <ubizjak@gmail.com>
1558 * gcc/config/i386/cpuid.h: New file.
1559 * gcc/config/i386/driver-i386.c: Include cpuid.h.
1560 (describe_cache): Shrink size and line strings to 100 bytes.
1561 (detect_caches_amd): Return "" for unsupported max_ext_level.
1562 Use __cpuid function.
1563 (detect_caches_intel): Return "" for unsupported max_level.
1564 Use __cpuid function.
1565 (host_detect_local_cpu): Change feature flag variables to
1566 unsigned int. Initialize only extended feature flag variables.
1567 Use __get_cpuid_max to determine max supported cpuid level.
1568 Use __cpuid function to determine supported features. Fix
1569 calculation of family id. Remove is_amd and check signature
1570 directly. Check for Geode signature. Handle family 4 id.
1571 [PROCESSOR_GENERIC32]: New default for unknown family id. Move
1572 cpu discovery code to lower part of the function.
1573 [PROCESSOR_PENTIUM, PROCESSOR_K6, PROCESSOR_ATHLON]: Do not tune
1574 for sub-architecture.
1575 [PROCESSOR_PENTIUMPRO]: Simplify cpu discovery code.
1576 [PROCESSOR_K8]: Add k8-sse3 architecture.
1577 [PROCESSOR_NOCONA]: Remove.
1578 [PROCESSOR_GENERIC64]: Ditto.
1579 * gcc/config/i386/x-i386 (driver-i386.o): Depend on cpuid.h.
1580 * gcc/config/i386/crtfastmath.c: Include cpuid.h. Use __get_cpuid
1581 to check for SSE and FXSAVE support.
1582 * gcc/config/i386/t-crtfm (crtfastmath.o): Depend on cpuid.h.
1583 Add -minline-all-stringops.
1584 * gcc/config.gcc (i[34567]86-*-*): Add cpuid.h to extra_headers.
1585 (x86_64-*-*): Ditto.
1587 2007-09-05 Jie Zhang <jie.zhang@analog.com>
1589 * config/bfin/linux-unwind.h: New file.
1590 * config/bfin/linux.h (MD_UNWIND_SUPPORT): Define.
1591 * config/bfin/uclinux.h (MD_UNWIND_SUPPORT): Define.
1593 2007-09-05 Laurynas Biveinis <laurynas.biveinis@gmail.com>
1595 * Makefile.in (stmt.o): Add alloc-pool.h dependency.
1596 * stmt.c: Include alloc-pool.h.
1597 (struct case_node): Remove GTY marker.
1598 (add_case_node): New parameter case_node_pool. Use it for
1600 (expand_case): New allocation pool case_node_pool. Initialize it,
1601 pass to add_case_node and free it.
1603 2007-09-05 Sandra Loosemore <sandra@codesourcery.com>
1604 David Ung <davidu@mips.com>
1605 Nigel Stephens <nigel@mips.com>
1607 Add mips16/nomips16 function attributes and -mflip-mips16 option
1608 for testing mixed-mode compilation.
1610 * config/mips/mips.opt (mflip-mips16): New.
1612 * config/mips/mips.h (SYMBOL_FLAG_MIPS16_FUNC): Define.
1613 (SYMBOL_FLAG_MIPS16_FUNC_P): Define.
1615 * config/mips/mips.c (mips_base_target_flags): New.
1616 (mips_base_mips16): New.
1617 (mips_base_schedule_insns): New.
1618 (mips_base_reorder_blocks_and_partition): New.
1619 (mips_base_align_loops): New.
1620 (mips_base_align_jumps): New.
1621 (mips_base_align_functions): New.
1622 (mips16_flipper): New.
1623 (mips_attribute_table): Add "mips16" and "nomips16" entries.
1624 (TARGET_SET_CURRENT_FUNCTION): Define.
1625 (mips_mips16_type_p, mips_nomips16_type_p): New.
1626 (mips_comp_type_attributes): Check mips16/nomips16 attributes.
1627 (mips_function_ok_for_sibcall): Make it deal with functions with
1629 (mips_init_split_addresses): New, split out from override_options.
1630 (mips_init_relocs): New, split out from override_options.
1631 (was_mips16_p): New.
1632 (mips_set_mips16_mode): New, split out from override_options.
1633 (mips_set_current_function): New.
1634 (override_options): Add sorry for unsupported mips16/pic
1635 combination. Remove error for mips16/dsp combination. Save
1636 base option settings.
1637 (mips_file_start): Move mips16 mode setting output from here....
1638 (mips_output_function_prologue): ....to here.
1639 (mips_output_mi_thunk): Check for mips16 function.
1640 (build_mips16_function_stub): Don't set .mips16 here.
1641 (build_mips16_call_stub): Likewise.
1642 (mips_expand_builtin): Error in mips16 mode.
1643 (mips_use_mips16_mode_p): New.
1644 (mips_encode_section_info): Check for mips16 function, and set
1645 SYMBOL_REF_FLAGS accordingly.
1647 * doc/extend.texi (Function Attributes): Document new
1648 mips16/nomips16 attributes.
1649 * doc/invoke.texi (Option Summary): Add -mflip-mips16.
1650 (MIPS Options): Document -mflip-mips16.
1652 2007-09-05 Sandra Loosemore <sandra@codesourcery.com>
1654 * config/mips/mips.c (mips_legitimize_tls_address): Call sorry
1655 if we encounter TLS address in MIPS16 mode.
1656 (mips_legitimize_const_move): Check cannot_force_const_mem for
1657 the (const (plus symbol offset)) case; this forces invalid TLS
1658 address in MIPS16 mode to be caught by the above call to sorry.
1659 (override_options): Don't reset targetm.have_tls in MIPS16 mode,
1660 because that now enables emutls, which is not ABI compatible
1661 with native TLS in non-MIPS16 mode.
1663 2007-09-05 Sandra Loosemore <sandra@codesourcery.com>
1665 Add target hook invoked when cfun changes.
1667 * doc/tm.texi (TARGET_SET_CURRENT_FUNCTION): Document.
1668 * target.h (struct gcc_target): Add set_current_function.
1669 * target-def.h (TARGET_SET_CURRENT_FUNCTION): Define.
1670 (TARGET_INITIALIZER): Add initializer for set_current_function.
1671 * tree.h (push_struct_function): New.
1672 * tree-inline.h (push_cfun, pop_cfun): Move declarations to...
1674 (set_cfun): Declare.
1675 * tree-inline.c (cfun_stack, push_cfun, pop_cfun): Moved to...
1677 (push_function_context_to): Use allocate_struct_function
1678 to create null context, not init_dummy_function_start. Use set_cfun.
1679 (pop_function_context_from): Use set_cfun.
1680 (in_dummy_function): New.
1681 (invoke_set_current_function_hook): New.
1683 (push_cfun, pop_cfun): Use set_cfun.
1684 (push_struct_function): New.
1685 (allocate_struct_function): Call invoke_set_current_function_hook
1687 (prepare_function_start): Don't set cfun here. Remove unused
1688 argument; fix all callers.
1689 (init_dummy_function_start): Fiddle with in_dummy_function. Call
1690 push_struct_function.
1691 (init_function_start): Set cfun here.
1692 (expand_dummy_function_end): Fiddle with in_dummy_function. Pop cfun.
1693 * omp-low.c (create_omp_child_function): Use push_struct_function
1694 and pop_cfun to save/restore state.
1695 (expand_omp_parallel): Remove unused saved_cfun variable.
1696 * cgraphunit.c (ipa_passes): Use set_cfun.
1697 * gimple-low.c (record_vars_into): Use push_cfun/pop_cfun here.
1698 * dwarf2out.c (dwarf2out_abstract_function): Likewise.
1699 * matrix-reorg.c (transform_allocation_sites): Likewise.
1700 (matrix_reorg): Use set_cfun.
1701 * gimplify.c (gimplify_function_tree): Use push_cfun/pop_cfun here.
1702 * tree-optimize.c (tree_rest_of_compilation): Remove one redundant
1703 assignment to cfun; use set_cfun for the other.
1704 * tree-cfg.c (move_sese_region_to_fn): Use set_cfun.
1705 (dump_function_to_file): Use push_cfun/pop_cfun here.
1706 * c-decl.c (finish_function): Use set_cfun.
1708 2007-09-05 Kenneth Zadeck <zadeck@naturalbridge.com>
1710 * regrename.c (rerename_optimize): Use deferred rescanning and
1711 insert explicit calls to rescan insns when changed.
1713 2007-09-05 Jan Hubicka <jh@suse.cz>
1715 * optabs.c (libfunc_entry): Change optab to integer.
1716 (hash_libfunc, convert_optab_libfunc, optab_libfunc,
1717 set_optab_libfunc, set_conv_libfunc): Use optable indexes instead
1718 of pointers for hashing.
1720 2007-09-05 Jan Hubicka <jh@suse.cz>
1722 * regrename.c (pass_regrename, pass_cprop_hardreg): Add RTL sharing
1724 * fwprop.c (pass_rtl_fwprop, pass_rtl_fwprop_add): Likewise.
1725 * see.c (pass_see): Likewise.
1726 * tracer.c (pass_tracer): Likewise.
1727 * postreload-gcse.c (pass_gcse2): Likewise.
1728 * postreload.c (pass_postreload_cse): Likewise.
1729 * mode-switching.c (pass_mode_switching): Likewise.
1730 * modulo-sched.c (pass_sms): Likewise.
1731 * cse.c (cse_insn): Likewise.
1732 * web.c (pass_web): Likweise.
1733 * combine-stack-adj.c (pass_stack_adjustments): Likewise.
1734 * dce.c (pass_ud_rtl_dce, pass_fast_rtl_dce): Likewise.
1735 * loop-init.c (pass_rtl_loop_init): Likewise.
1736 (pass_rtl_loop_done, pass_rtl_move_loop_, pass_rtl_unswitch,
1737 pass_rtl_unroll_and, pass_rtl_doloop): Likewise.
1738 * global.c (pass_global_alloc): Likewise.
1739 * ifcvt.c (pass_rtl_ifcvt, pass_if_after_combine,
1740 pass_if_after_reload): Likewise.
1741 * reload.c (pass_peephole2, pass_split_for_shoren_branches): Likewise.
1742 * dse.c (pass_rtl_dse1, pass_rtl_dse2): Likewise.
1743 * regmove.c (pass_regmove): Likewise.
1744 * function.c (pass_thread_prologugues_epilogues): Likewise.
1745 * gcse.c (pass_gcse): Likewise.
1746 * rtl-factoring.c (pass_rtl_seqabstr): Likewise.
1747 * lower-subreg.c (pass_lower_subreg2): Likewise.
1748 * bt-load.c (pass_branch_target_load): Likewise.
1749 * emit-rtl.c (pass_unshare_all_rtl): Likewise.
1750 * cfgcleanup.c (pass_jump, pass_jump2): Likewise.
1751 * combine.c (pass_combine): Likewise.
1752 * bb-reorder.c (pass_duplicate_comp, pass_reorder_blocks): Likewise.
1753 (pass_partition_blocks): Likewise.
1754 * var-tracking.c (pass_variable_track): Likewise.
1755 * reg-stack.c (pass_stack_regs_run): Likewise.
1756 * sched-rgn.c (pass_sched, pass_sched2): Likewise.
1757 * passes.c (pass_postreload): Likewise.
1758 (execute_function_todo): Add TODO_verify_rtl_sharing handling code.
1759 * tree-pass.h (TODO_verify_rtl_sharing): New.
1760 (TODO_update_ssa, TODO_update_ssa_no_phi, TODO_update_ssa_full_phi,
1761 TODO_update_ssa_only_virtuals, TODO_remove_unused_locals,
1762 TODO_set_props, TODO_df_finish, TODO_df_verify,
1763 TODO_mark_first_instance, TODO_rebuild_alias): Renumber.
1765 2007-09-05 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1767 * fold-const.c (all_ones_mask_p, sign_bit_p, simple_operand_p,
1768 fold_real_zero_addition_p, reorder_operands_p,
1769 div_if_zero_remainder, fold_undefer_overflow_warnings,
1770 int_binop_types_match_p, fold_convert_const_int_from_int,
1771 fold_convert_const_int_from_real,
1772 fold_convert_const_int_from_fixed,
1773 fold_convert_const_real_from_real,
1774 fold_convert_const_real_from_fixed,
1775 fold_convert_const_fixed_from_fixed,
1776 fold_convert_const_fixed_from_int,
1777 fold_convert_const_fixed_from_real, maybe_lvalue_p,
1778 fold_checksum_tree, fold_check_failed, print_fold_checksum,
1779 debug_fold_checksum, multiple_of_p): Constify.
1780 * tree-flow-inline.h (get_lineno): Likewise.
1781 * tree-flow.h (get_lineno): Likewise.
1782 * tree-object-size.c (compute_object_offset, addr_object_size,
1783 alloc_object_size, pass_through_call): Likewise.
1784 * tree-pretty-print.c (op_symbol, print_call_name,
1785 print_struct_decl, do_niy): Likewise.
1786 * tree.h (fold_undefer_overflow_warnings, multiple_of_p,
1787 debug_fold_checksum): Likewise.
1789 2007-09-04 Steven Bosscher <steven@gcc.gnu.org>
1792 * lower-subreg.c (resolve_clobber): If we remove a REG_LIBCALL
1793 note, remove the associated REG_RETVAL note.
1795 2007-09-04 Andrew Pinski <andrew_pinski@playstation.sony.com>
1797 * config.gcc (powerpc*-*-*): Install
1798 spu2vmx.h, vec_types.h, and si2vmx.h headers.
1799 * config/rs6000/spu2vmx.h: New header.
1800 * config/rs6000/si2vmx.h: New header.
1801 * config/rs6000/vec_types.h: New header.
1803 2007-09-05 Ben Elliston <bje@au.ibm.com>
1805 * varasm.c (initializer_constant_valid_p): Fix comment typo.
1807 2007-09-05 Ben Elliston <bje@au.ibm.com>
1809 * config/rs6000/ppu_intrinsics.h (__ldarx): Use `Z' constraint,
1811 (__stdcx): Make asm volatile.
1812 (__stwcx, __stdcx): Return only 0 or 1.
1814 2007-09-05 Jan Hubicka <jh@suse.cz>
1818 2007-09-04 Jan Hubicka <jh@suse.cz>
1819 * tree-tailcall.c (eliminate_tail_call): Expect unrenamed return value.
1821 2007-09-04 Richard Guenther <rguenther@suse.de>
1822 * tree-ssa-operands.c (add_virtual_operand): Only mark
1823 stores as has_volatile_ops if alias information is not available.
1825 2007-09-05 Jakub Jelinek <jakub@redhat.com>
1827 * config/rs6000/tramp.asm: Include config.h.
1828 Check __PIC__ or __pic__ macro instead of SHARED.
1830 PR rtl-optimization/32300
1831 * see.c (see_copy_insn): New function.
1832 (see_def_extension_not_merged, see_merge_one_use_extension,
1833 see_merge_one_def_extension): Use it. Avoid changing
1834 PREV_INSN/NEXT_INSN chains directly, insted emit insns
1835 into sequences. Call df_insn_delete on temporary insns
1836 that won't be emitted into the insn stream.
1837 (rest_of_handle_see): Turn off DF_DEFER_INSN_RESCAN
1838 and run df_process_deferred_rescans () before run_fast_dce.
1840 PR tree-optimization/33017
1841 * tree-data-ref.c (split_constant_offset) <case SSA_NAME>: Don't
1842 recurse for pure or const function calls.
1844 2007-09-04 Laurynas Biveinis <laurynas.biveinis@gmail.com>
1846 * c-format.c: Include alloc-pool.h.
1847 (check_format_info_main): New argument fwt_alloc. Use allocation
1848 pool instead of GC. Remove GC deallocation code.
1849 (check_format_arg): Create allocation pool, pass it to
1850 check_format_info_main and free it afterwards.
1851 * Makefile.in (c-format.o): Add alloc-pool.h dependency.
1853 2007-09-05 Ben Elliston <bje@au.ibm.com>
1855 * config/rs6000/ppu_intrinsics.h (__protected_stream_count):
1856 Rename count macro arugment for clarity.
1857 (__protected_unlimited_stream_set): Pass UG=1 to __dcbt_TH1000.
1858 (__protected_stream_set): Likewise, pass 0.
1860 2007-09-04 Jan Hubicka <jh@suse.cz>
1862 * tree-tailcall.c (eliminate_tail_call): Expect unrenamed return value.
1864 2007-09-04 Jan Hubicka <jh@suse.cz>
1866 * invoke.texi (-finline-small-functions): Document.
1867 * ipa-inline.c (cgraph_default_inline_p): Do not use DECL_INLINE
1868 when deciding what is inlinable.
1869 (cgraph_decide_recursive_inlining): Handle flag_inline_functions.
1870 (cgraph_decide_inlining_of_small_function): Handle new flags.
1871 (cgraph_decide_inlining_incrementally): Likewise.
1872 * opts.c (decode_options): Enable flag_inline_small_functions at -O2
1873 * common.opt (finline-small-functions): New.
1874 * Makefile.in (build/gengtype.o-warn): Work around PR29478
1876 2007-09-04 Richard Guenther <rguenther@suse.de>
1878 * tree-ssa-operands.c (add_virtual_operand): Only mark
1879 stores as has_volatile_ops if alias information is not available.
1881 2007-09-04 Andrew Haley <aph@redhat.com>
1883 * config/arm/libgcc-bpabi.ver: Add _Unwind_Backtrace as GCC_4.3.0.
1885 2007-09-04 Andrew Haley <aph@redhat.com>
1887 * config/arm/libunwind.S (UNWIND_WRAPPER _Unwind_Backtrace): New.
1888 * config/arm/unwind-arm.h (__gnu_Unwind_Backtrace): New.
1889 * config/arm/unwind-arm.c (__gnu_Unwind_Backtrace): New.
1891 2007-09-04 Emmanuel Thome <Emmanuel.Thome@inria.fr>
1894 * common.opt: Mention ABI version 2 in comment.
1896 2007-09-04 Jan Hubicka <jh@suse.cz>
1898 * optabs.c (debug_optab_libfunc): Update; make available to gdb.
1899 (libfunc_entry): New structure.
1900 (libfunc_hash): New hashtable.
1901 (hash_libfunc): New function.
1902 (eq_libfunc): New function.
1903 (convert_optab_libfunc): New function.
1904 (optab_libfunc): New function.
1905 (expand_binop, sign_expand_binop, expand_twoval_binop_libfunc,
1906 expand_unop, prepare_cmp_insn, prepare_float_insn, gen_add2_insn,
1907 expand_float, expand_fix, new_optab, new_convert_optab):
1908 Update for new libfunc API.
1909 (init_libfunc, init_integral_libfuncs,
1910 init_floating_libfuncs, init_interclass_conv_libfuncs
1911 init_intraclass_conv_libfuncs): Remove; reorganize all logic to:
1912 (gen_libfunc, gen_int_libfunc, gen_fp_libfunc, gen_int_fp_libfunc,
1913 gen_intv_fp_libfunc, gen_interclass_conv_libfunc,
1914 gen_int_to_fp_conv_libfunc, gen_ufloat_conv_libfunc,
1915 gen_int_to_fp_nondecimal_conv_libfunc, gen_fp_to_int_conv_libfunc,
1916 gen_intraclass_conv_libfunc, gen_trunc_conv_libfunc,
1917 gen_extend_conv_libfunc): New.
1918 (init_one_libfunc): Revamp for hashtables.
1919 (set_conv_libfunc): Likewise.
1920 (init_optabs): Initialize hashtable; use lazy initialization where
1922 * optabs.h (optab_handlers): Move out of GGC.
1923 (optab, convert_optab): Move out of GGC; add lazy gen info.
1924 (code_to_optab, convert_optab_table, optab_table): Move out of GGC.
1925 (optab_libfunc, convert_optab_libfunc): New.
1926 * builtins.c (expand_builtin_powi): Update for new API.
1927 * expr.c (convert_move): Likewise.
1928 * expmed.c (expand_divmod): Likewise.
1930 2007-09-04 Daniel Jacobowitz <dan@codesourcery.com>
1932 * config/rs6000/rs6000.c (rs6000_stack_info): Allocate space for the
1933 GOT pointer only if there is a constant pool. Use the allocated space
1936 2007-09-04 Daniel Jacobowitz <dan@codesourcery.com>
1938 * config/rs6000/linux-unwind.h (ppc_fallback_frame_state): Correct
1941 2007-09-04 Revital Eres <eres@il.ibm.com>
1942 Richard Guenther <rguenther@suse.de>
1943 R. Clint Whaley <whaley@cs.utsa.edu>
1945 * doc/invoke.texi (-fassociative-math, -freciprocal-math):
1947 * tree-tailcall.c (process_assignment): Use -fassociative-math
1948 when reodering operands of floating-point type.
1949 * fold-const.c (fold_comparison, fold_binary): Use
1950 -fassociative-math and -freciprocal-math instead of
1951 -funsafe-math-optimization flag.
1952 * toplev.h (set_unsafe_math_optimizations_flags): Declare function.
1953 * tree-ssa-math-opts.c (gate_cse_reciprocals): Use
1954 -freciprocal-math instead of -funsafe-math-optimizations.
1955 * opts.c (set_fast_math_flags): Set -freciprocal-math and
1956 -fassociative-math when -ffast-math is set.
1957 (set_unsafe_math_optimizations_flags): New Function
1958 to set -freciprocal-math and -fassociative-math when
1959 -funsafe-math-optimizations is set.
1960 (common_handle_option): Call it.
1961 * tree-vectorizer.c (vect_is_simple_reduction): Use
1962 -fassociative-math when doing reduction on floats.
1963 * loop-unroll.c (analyze_insn_to_expand_var): Use
1964 -fassociative-math when expanding an accumulator of type float.
1965 * simplify-rtx.c (simplify_binary_operation_1): Use
1966 -fassociative-math and -freciprocal-math when reordeing operands
1967 of floating-point type.
1968 * combine.c (combine_simplify_rtx): Likewise.
1969 * tree-ssa-reassoc.c (break_up_subtract_bb, reassociate_bb):
1971 * common.opt (-fassociative-math, -freciprocal-math): New flags.
1973 2007-09-04 Paolo Carlini <pcarlini@suse.de>
1976 * doc/install.texi (--enable-__cxa_atexit): Fix typo.
1978 2007-09-04 Jan Hubicka <jh@suse.cz>
1980 * regrename.c(copyprop_hardreg_forward_1): Unshare when doing copyprop.
1982 2007-09-04 Jan Hubicka <jh@suse.cz>
1984 * regrename.c (build_def_use): Don't share RTL between MATCH_OPERATOR
1985 and corresponding MATCH_DUP.
1987 2007-09-04 Uros Bizjak <ubizjak@gmail.com>
1990 * combine.c (subst): Do not try to simplify X if it represents load
1991 of FP constant from the constant pool via float extension.
1993 2007-09-04 Ben Elliston <bje@au.ibm.com>
1995 * c-opts.c: Include "tm_p.h".
1996 * Makefile.in (c-opts.o): Add $(TM_P_H) as a dependency.
1998 2007-09-04 Richard Guenther <rguenther@suse.de>
2000 PR tree-optimization/33291
2001 * tree-pretty-print.c (dump_generic_node): Dump all
2002 qualifiers for pointer types, not only first. Dump
2003 qualifiers for aggregate types as well.
2004 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Always
2005 use the canonical type for building ARRAY_REFs.
2006 * gimplify.c (canonicalize_addr_expr): Clean up. The
2007 correct validness check is compatibility of the pointer
2008 types. Always use the canonical type for building
2009 ARRAY_REFs and ADDR_EXPRs.
2010 * tree-ssa-forwprop.c (forward_propagate_addr_expr): Revert
2011 change that disabled propagation of ADDR_EXPRs into statements
2014 2007-09-03 Zack Weinberg <zack@codesourcery.com>
2016 * config/arm/arm.md (mulsidi3, umulsidi3, mulsi3_highpart)
2017 (umulsi3_highpart): Make into expanders; existing insns are
2018 now named *[insn]_nov6. Add stars to existing [insn]_v6
2019 counterparts' names.
2021 2007-09-03 Daniel Jacobowitz <dan@codesourcery.com>
2023 * Makefile.in (DRIVER_DEFINES, gcc.o, gccspec.o, protoize.o)
2024 (unprotoize.o): Remove SHLIB_MULTILIB.
2026 2007-09-03 Eric Weddington <eweddington@cso.atmel.com>
2028 * doc/extend.texi: Document progmem attribute for AVR.
2030 2007-09-03 Anatoly Sokolov <aesok@post.ru>
2033 * config/avr/avr.h (TARGET_VTABLE_ENTRY_ALIGN): Define.
2035 2007-09-03 Zack Weinberg <zack@codesourcery.com>
2037 * optabs.c (expand_ffs): Initialize val and defined_at_zero
2040 * optabs.c: Remove unnecessary forward declarations.
2041 (expand_unop_direct): New, broken out of expand_unop.
2042 (expand_doubleword_clz): New.
2043 (expand_ctz): Move above expand_ffs. Use
2044 start_sequence, end_sequence, add_equal_note, and
2045 expand_unop_direct. Add more commentary.
2046 (expand_ffs): Try both ctz optab and expand_ctz.
2047 Generate a test and branch if the hardware doesn't give us
2048 a useful value for input zero. Style improvements similar to
2051 * config/arm/arm.md (ffssi2, ctzsi2): Delete.
2053 2007-09-03 Zack Weinberg <zack@codesourcery.com>
2055 * config.gcc: Delete stanza for arm-semi-aof and
2056 armel-semi-aof targets.
2057 * config/arm/arm-protos.h
2059 * config/arm/arm.h: Delete all #ifdef AOF_ASSEMBLER blocks;
2060 make all #ifndef AOF_ASSEMBLER blocks unconditional. Also
2061 delete aof_pic_label and remove mention of AOF in comments.
2062 * config/arm/arm.md: Delete patterns used only for AOF assembly.
2064 * config/arm/semiaof.h
2065 * config/arm/t-semi: Delete file.
2067 2007-09-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2069 * system.h (CONST_CAST2, CONST_CAST_TREE, CONST_CAST_RTX,
2070 CONST_CAST_BB): New macros for type-specific casts.
2071 (CONST_CAST): Add a TYPE parameter and define it in terms of
2074 * c-decl.c (c_make_fname_decl): Supply a TYPE for CONST_CAST.
2075 * c-lex.c (cb_ident, lex_string): Likewise,
2076 * c-typeck.c (free_all_tagged_tu_seen_up_to): Likewise,
2077 * config/i386/i386.c (ix86_function_regparm,
2078 ix86_function_sseregparm): Use CONST_CAST_TREE.
2079 * config/mmix/mmix.c (mmix_encode_section_info): Supply a TYPE for
2081 * gcc.c (set_spec, read_specs, for_each_path, execute, do_spec_1,
2082 give_switch, set_multilib_dir): Likewise,
2083 * gengtype-parse.c (string_seq, typedef_name): Likewise,
2084 * gimple-low.c (block_may_fallthru): Use CONST_CAST_TREE.
2085 * jump.c (reversed_comparison_code_parts): Use CONST_CAST_RTX.
2086 * passes.c (execute_one_pass): Supply a TYPE for CONST_CAST.
2087 * prefix.c (update_path): Likewise,
2088 * pretty-print.c (pp_base_destroy_prefix): Likewise,
2089 * rtlanal.c (keep_with_call_p): Use CONST_CAST_RTX.
2090 * tree-cfg.c (tree_can_merge_blocks_p,
2091 tree_block_ends_with_condjump_p): Use CONST_CAST_BB.
2092 * tree-eh.c (lookup_stmt_eh_region_fn): USe CONST_CAST_TREE.
2093 * tree.c (build_string): Supply a TYPE for CONST_CAST.
2094 (attribute_list_contained): Use CONST_CAST_TREE.
2096 2007-09-03 Krister Walfridsson <cato@df.lth.se>
2098 * config/netbsd-elf.h (USE_LD_AS_NEEDED): Define.
2100 2007-09-03 Richard Sandiford <richard@codesourcery.com>
2103 * optabs.c (avoid_expensive_constant): Canonicalize CONST_INTs
2104 before forcing them into a register.
2106 2007-09-03 Richard Sandiford <richard@codesourcery.com>
2108 * config/mips/mips.md (fetchop_bit): Use define_code_iterator
2109 rather than define_code_macro.
2111 2007-09-03 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2113 * gcc.c (xputenv): New function.
2114 (putenv_from_prefixes): Use xputenv instead of putenv.
2115 (process_command): Use xputenv instead of putenv.
2116 (set_collect_gcc_options): Use xputenv instead of putenv.
2117 (main): Use xputenv instead of putenv.
2119 2007-09-03 Jesper Nilsson <jesper.nilsson@axis.com>
2121 * config/cris/cris.h (TARGET_HAS_SWAP): Defined to describe
2122 availability of swap instruction.
2123 * config/cris/cris.md (bswapsi2): Implement using swap instruction.
2125 2007-09-03 Revital Eres <eres@il.ibm.com>
2127 * doc/invoke.texi (-fmodulo-sched-allow-regmoves): Update
2130 2007-09-03 Vladimir Yanovsky <yanov@il.ibm.com>
2131 Ayal Zaks <zaks@il.ibm.com>
2132 Revital Eres <eres@il.ibm.com>
2134 * modulo-sched.c (ps_insert_empty_row, verify_partial_schedule,
2135 compute_split_row): New functions.
2136 (ps_unschedule_node): Remove.
2137 (normalize_sched_times): Iterate over the already scheduled
2138 insns instead of the number of nodes.
2139 (MAX_SPLIT_NUM): New definition.
2140 (sms_schedule_by_order): Change the scheduling heuristic to
2141 avoid useless increases of initiation interval ii.
2142 (get_sched_window): Add dump printouts.
2144 2007-09-02 David Daney <ddaney@avtrex.com>
2146 * config/mips/mips.md (UNSPEC_COMPARE_AND_SWAP, UNSPEC_SYNC_OLD_OP,
2147 UNSPEC_SYNC_NEW_OP, UNSPEC_SYNC_EXCHANGE): New define_constants.
2148 (optab, insn): Add more attributes.
2149 (fetchop_bit): New code macro.
2150 (immediate_insn): New code macro attribute.
2151 (sync): Change condition to ISA_HAS_SYNC.
2152 (rdhwr): Change predicate for operand 0 to register_operand.
2153 (memory_barrier): New expand.
2154 (sync_compare_and_swap<mode>, sync_add<mode>, sync_sub<mode>,
2155 sync_old_add<mode>, sync_old_sub<mode>, sync_new_add<mode>,
2156 sync_new_sub<mode>, sync_<optab><mode>, sync_old_<optab><mode>,
2157 sync_new_<optab><mode>, sync_nand<mode>, sync_old_nand<mode>,
2158 sync_new_nand<mode>, sync_lock_test_and_set<mode>): New insns.
2159 * config/mips/mips.h (ISA_HAS_SYNC, ISA_HAS_LL_SC): New ISA predicates.
2160 (MIPS_COMPARE_AND_SWAP, MIPS_SYNC_OP, MIPS_SYNC_OLD_OP,
2161 MIPS_SYNC_NEW_OP, MIPS_SYNC_NAND, MIPS_SYNC_OLD_NAND,
2162 MIPS_SYNC_NEW_NAND, MIPS_SYNC_EXCHANGE): New Macros.
2164 2007-09-03 Jesper Nilsson <jesper.nilsson@axis.com>
2165 Hans-Peter Nilsson <hp@axis.com>
2167 * config/cris/cris.h (TARGET_HAS_LZ, CLZ_DEFINED_VALUE_AT_ZERO):
2168 Defined to describe availability and behavior of CLZ.
2169 * config/cris/cris.md (clzsi2): Implement using lz instruction.
2170 * config/cris/cris.opt: Tweak comment for "-metrax4".
2171 * config/cris/arit.c (LZ): When defined, define as __builtin_clz.
2172 * longlong.h [__CRIS__ && __CRIS_arch_version >= 3]
2173 (count_leading_zeros): Define.
2175 2007-09-01 Laurynas Biveinis <laurynas.biveinis@gmail.com>
2177 * ggc-page.c (ggc_pch_read): Call validate_free_objects.
2179 2007-09-02 Joseph Myers <joseph@codesourcery.com>
2182 * c-decl.c (grokdeclarator): Apply qualifiers to type of parameter
2185 2007-09-01 Kazu Hirata <kazu@codesourcery.com>
2187 * config/arm/arm.c, config/rs6000/ppu_intrinsics.h,
2188 config/spu/spu.c, df-scan.c, fixed-value.c, fold-const.c,
2189 ginclude/tgmath.h, haifa-sched.c, optabs.c, recog.c,
2190 sched-deps.c, sched-int.h, system.h, target.h,
2191 tree-ssa-live.c, tree-vect-transform.c, tree-vectorizer.c,
2192 tree.def: Fix comment typos.
2194 2007-09-01 Kazu Hirata <kazu@codesourcery.com>
2196 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove.
2197 (override_options): Turn on -mstrict-align on non-ColdFire
2199 * config/m68k/m68k.h (BIGGEST_ALIGNMENT): Update a comment.
2201 2007-09-01 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2203 * c-common.c (const_strip_array_types): Delete.
2204 * c-common.h (const_strip_array_types): Likewise.
2205 * emit-rtl.c (const_next_insn, const_previous_insn,
2206 const_next_nonnote_insn, const_prev_nonnote_insn,
2207 const_next_real_insn, const_prev_real_insn,
2208 const_next_active_insn, const_prev_active_insn, const_next_label,
2209 const_prev_label): Delete.
2210 * gimple-low.c (block_may_fallthru): Use expr_last.
2211 * jump.c (reversed_comparison_code_parts): Use prev_nonnote_insn.
2212 * rtl.h (const_previous_insn, const_next_insn,
2213 const_prev_nonnote_insn, const_next_nonnote_insn,
2214 const_prev_real_insn, const_next_real_insn,
2215 const_prev_active_insn, const_next_active_insn, const_prev_label,
2216 const_next_label): Delete.
2217 * rtlanal.c (keep_with_call_p): Use next_nonnote_insn.
2218 * system.h (CONST_CAST): Update comment.
2219 * tree-cfg.c (tree_can_merge_blocks_p): Use last_stmt.
2220 (const_first_stmt, const_last_stmt): Delete.
2221 (tree_block_ends_with_condjump_p): Use last_stmt.
2222 * tree-flow.h (const_first_stmt, const_last_stmt): Delete.
2223 * tree-iterator.c (const_expr_first, const_expr_last,
2224 const_expr_only): Delete.
2225 * tree.c (const_lookup_attribute): Likewise.
2226 (attribute_list_contained): Use lookup_attribute.
2227 * tree.h (const_lookup_attribute, const_expr_first,
2228 const_expr_last, const_expr_only): Delete.
2230 2007-09-01 Richard Guenther <rguenther@suse.de>
2232 * tree-ssa-alias.c (create_name_tags): Use TYPE_VOLATILE
2233 instead of TREE_THIS_VOLATILE.
2234 * tree-ssa-operands.c (add_virtual_operand): Do not set
2235 has_volatile_ops. Do not check s_ann for NULL.
2236 (get_indirect_ref_operands): Likewise.
2237 (get_expr_operands): Likewise. For all reference trees
2238 set has_volatile_ops based on TREE_THIS_VOLATILE. Also
2239 check base of reference for TREE_THIS_VOLATILE.
2241 2007-09-01 Richard Guenther <rguenther@suse.de>
2243 * tree-cfg.c (verify_gimple_expr): Fix check for conversions
2244 between integral types. Also allow conversions between
2247 2007-08-31 Andrew Pinski <andrew_pinski@playstation.sony.com>
2249 * optabs.c (prepare_float_lib_cmp): Use
2250 libgcc_cmp_return_mode instead of word_mode.
2251 * config/dfp-bit.h: Likewise.
2252 * config/fp-bit.h: Likewise.
2253 * config/spu/spu.c (spu_libgcc_cmp_return_mode):
2255 (spu_libgcc_shift_count_mode): New function.
2256 (TARGET_LIBGCC_CMP_RETURN_MODE): Define.
2257 (TARGET_LIBGCC_SHIFT_COUNT_MODE): Define.
2259 2007-08-31 Joseph Myers <joseph@codesourcery.com>
2261 * tree-pretty-print.c: Include output.h.
2262 (dump_decl_name, dump_generic_node): Use %wd with pp_printf, not
2263 HOST_WIDE_INT_PRINT_DEC.
2264 * Makefile.in (tree-pretty-print.o): Add dependency on output.h.
2266 2007-08-31 Zdenek Dvorak <ook@ucw.cz>
2268 PR rtl-optimization/33224
2269 * loop-iv.c (latch_dominating_def): Check that the definition belongs
2270 to the analysed region.
2272 2007-08-31 Richard Guenther <rguenther@suse.de>
2274 * function.c (allocate_struct_function): Do not set
2275 current_function_returns_pointer.
2276 * function.h (struct var_refs_queue): Remove.
2277 (struct simple_obstack_stack): Likewise.
2278 (struct function): Remove fixup_var_refs_queue member.
2279 Remove returns_pointer flag.
2280 (current_function_returns_pointer): Remove define.
2282 2007-08-31 Nick Clifton <nickc@redhat.com>
2284 * tree-ssa-coalesce.c (compare_pairs): Use the elements as
2285 secondary keys in order to obtain a stable sort.
2287 2007-08-31 Nick Clifton <nickc@redhat.com>
2290 * config/m32r/constraints.md: Add W constraint for integer zero.
2291 * config/m32r/m32r.md (get_pc): Use W and i constraints.
2292 Fix length calculation by using alternatives.
2294 2007-08-31 Richard Sandiford <richard@codesourcery.com>
2296 * optabs.c (shift_optab_p, commutative_optab_p): New functions,
2297 split out from expand_binop.
2298 (avoid_expensive_constant): New function.
2299 (expand_binop_directly): Remove commutative_op argument and
2300 (expand_binop_directly): Remove commutative_op argument and
2301 call cummutative_optab_p instead. Do not change op0 or op1
2302 when swapping xop0 and xop1. Apply avoid_expensive_constant
2303 to each argument after potential swapping. Enforce the
2304 canonical order of commutative operands.
2305 (expand_binop): Use shift_optab_p and commutative_optab_p.
2306 Update the calls to expand_binop_directly. Only force constants
2307 into registers when widening an operation. Only swap operands
2308 once a direct expansion has been rejected.
2309 (expand_twoval_binop): Only force constants into registers when
2310 using a direct expansion.
2312 2007-08-31 Maxim Kuvyrkov <maxim@codesourcery.com>
2314 * sched-deps.c (update_dep): Mark arguments with ATTRIBUTE_UNUSED.
2315 (sched_analyze_insn): Don't postprocess speculative dependencies when
2316 target has no scheduling.
2317 * tree-vect-transform.c (vect_estimate_min_profitable_iters): Fix
2318 'uninitialized' warning.
2320 2007-08-31 Richard Guenther <rguenther@suse.de>
2322 * tree.c (build_complex_type): Always set TYPE_NAME for
2325 2007-08-31 Olivier Hainque <hainque@adacore.com>
2327 * config/rs6000/xcoff.h (XCOFF_CSECT_DEFAULT_ALIGNMENT_STR): New
2328 internal macro. Default alignment factor for csect directives, chosen
2329 to match what BIGGEST_ALIGNMENT allows.
2330 (DATA_SECTION_ASM_OP): Use it.
2331 * config/rs6000/rs6000.c
2332 (rs6000_xcoff_output_readonly_section_asm_op):
2333 Use XCOFF_CSECT_DEFAULT_ALIGNMENT_STR.
2334 (rs6000_xcoff_output_readwrite_section_asm_op): Likewise.
2336 2007-08-31 Olivier Hainque <hainque@adacore.com>
2338 * dbxout.c (dbxout_block): Move declaration within the first
2339 #if defined (DBX_DEBUGGING_INFO) block. Rearrange another such
2340 block to include the definition, together with the associated
2343 2007-08-31 David Edelsohn <edelsohn@gnu.org>
2344 Revital Eres <eres@il.ibm.com>
2346 * doc/invoke.texi (-mpaired): Document flag.
2347 * config.gcc: Include paired.h in powerpc extra_headers and
2348 750cl.h in powerpc-*-linux*paired*.
2349 * config/rs6000/rs6000.opt (-mpaired): New flag.
2350 * config/rs6000/rs6000.c (paired_init_builtins,
2351 paired_expand_builtin, paired_expand_lv_builtin,
2352 paired_expand_stv_builtin, paired_expand_predicate_builtin):
2353 New functions to support the paired single builtin functions.
2354 (rs6000_hard_regno_mode_ok): Handle PAIRED_VECTOR_MODE.
2355 (def_builtin, bdesc_3arg, bdesc_2arg, bdesc_1arg): Add paired
2357 (bdesc_paired_preds): New structure for paired predicate
2359 (rs6000_expand_builtin): Expand paired single builtins.
2360 (rs6000_init_builtins): Init paired single builtins.
2361 (rs6000_common_init_builtins): Add v2sf_ftype_v2sf_v2sf_v2sf.
2362 Rename v2sf_ftype_v2sf_v2sf to v2sf_ftype_v2sf_v2sf_spe
2363 and v2sf_ftype_v2sf_spe to v2sf_ftype_v2sf. Add new types
2364 v2sf_ftype_v2sf_v2sf and v2sf_ftype_v2sf.
2365 (rs6000_vector_mode_supported_p): Support paired vector mode.
2366 * config/rs6000/rs6000.h: (UNITS_PER_PAIRED_WORD,
2367 PAIRED_VECTOR_MODE, PAIRED_SIMD_REGNO_P, TARGET_PAIRED_FLOAT):
2369 (LOCAL_ALIGNMENT): Handle PAIRED_VECTOR_MODE.
2370 (DATA_ALIGNMENT): Likewise.
2371 (UNITS_PER_SIMD_WORD): Handle PAIRED_VECTOR_MODE.
2372 (rs6000_builtins): Add PAIRED builtins.
2373 * config/rs6000/rs6000.md: Include paired.md.
2374 * config/rs6000/paired.h: New.
2375 * config/rs6000/paired.md: New.
2376 * config/rs6000/750cl.h: New.
2377 * config/rs6000/spe.md: Recognize movv2sf instruction for 750cl.
2379 2007-08-30 Ollie Wild <aaw@google.com>
2381 * c-opts.c (c_common_handle_option): Support -fno-directives-only.
2382 * gcc.c (default_compilers): Add -fno-directives-only to
2383 @assembler-with-cpp.
2385 2007-08-30 Sandra Loosemore <sandra@codesourcery.com>
2389 * regclass.c (initial_fixed_regs): Revert previous change and make
2391 (initial_call_used_regs): Likewise.
2392 (initial_call_really_used_regs): Delete, reverting previous addition.
2393 (initial_reg_names): Likewise.
2394 (init_reg_sets): Revert previous change.
2395 (saved_fixed_regs): New.
2396 (saved_call_used_regs): New.
2397 (saved_call_really_used_regs): New.
2398 (saved_reg_names): New.
2399 (save_register_info): New.
2400 (restore_register_info): New.
2401 (init_reg_sets_1): Replace reset of register info with call to
2402 restore_register_info.
2403 * rtl.h (save_register_info): Declare.
2404 * toplev.c (backend_init): Call save_register_info.
2406 2007-08-30 Laurynas Biveinis <laurynas.biveinis@gmail.com>
2408 * tree-ssa-propagate.c (set_rhs): Remove the copied annotation
2409 from the original statement.
2411 2007-08-30 Chao-ying Fu <fu@mips.com>
2413 * c-lex.c (interpret_fixed): Declare.
2414 (interpret_float): Process _Fract and _Accum.
2415 (interpret_fixed): New function.
2416 * final.c (output_addr_const): Process CONST_FIXED.
2417 * simplify-rtx.c (simplify_const_unary_operation): Handle US_NEG.
2418 (simplify_binary_operation_1): Handle US_ASHIFT, SS_MULT, US_MULT,
2420 (simplify_const_binary_operation): Handle SS_MULT, US_MULT, SS_DIV,
2422 (simplify_immed_subreg): Support CONST_FIXED.
2423 Process MODE_FRACT, MODE_UFRACT, MODE_ACCUM, MODE_UACCUM.
2424 (simplify_subreg): Support CONST_FIXED.
2426 2007-08-30 Andrew Pinski <andrew_pinski@playstation.sony.com>
2428 * config/rs6000/ppu_intrinsics.h: New file.
2429 * config.gcc (powerpc*-*-* <extra_headers>): Install
2432 2007-08-30 Kenneth Zadeck <zadeck@naturalbridge.com>
2434 * cfg.c (dump_flow_info): Change to also print entry and exit
2436 * print-rtl.c (print_rtl_single): Allow to print rtl with
2439 2007-08-30 Kenneth Zadeck <zadeck@naturalbridge.com>
2441 * df-core.h (df_dump_region): New function.
2442 * df.h (df_dump_region): New function.
2443 * loop-invariant.c (find_defs): Add call to df_dump_region.
2444 * loop-iv.c (iv_analysis_loop_init): Changed call from df_dump to
2447 2007-08-30 Jakub Jelinek <jakub@redhat.com>
2450 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Return
2451 true if any of the compare_section_name calls returned true,
2452 rather than if any returned false.
2454 2007-08-30 Richard Guenther <rguenther@suse.de>
2456 PR tree-optimization/33199
2457 * tree-ssa-structalias.c (handle_lhs_call): New function.
2458 (find_func_aliases): In non-IPA mode make sure that for
2459 calls that return a pointer we add a constraint for the
2460 result to point to anything.
2462 2007-08-30 Richard Guenther <rguenther@suse.de>
2464 * doc/invoke.texi (-mveclibabi): Document new target option.
2465 * config/i386/i386.opt (-mveclibabi): New target option.
2466 * config/i386/i386.c (ix86_veclib_handler): Handler for
2467 vectorization library support.
2468 (override_options): Handle the -mveclibabi option, initialize
2469 the vectorization library handler.
2470 (ix86_builtin_vectorized_function): As fallback call the
2471 vectorization library handler, if set.
2472 (ix86_veclibabi_acml): New static function for ACML ABI style
2473 vectorization support.
2475 2007-08-30 Jakub Jelinek <jakub@redhat.com>
2477 * config/rs6000/rs6000.c (rs6000_emit_sync): For QI or HI mode
2478 used_m, even if it is 32-bit aligned, adjust used_m MEM to have
2479 SImode and update m. Don't run gen_lowpart_common on arbitrary
2480 memory address, force it to register first.
2483 * dce.c (dce_process_block): Don't delete setters of
2484 artificially used registers.
2486 2007-08-30 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
2488 * config/s390/s390.md ("*add<mode>3_alc_carry1_cc",
2489 "*add<mode>3_alc_carry1_cconly", "*add<mode>3_alc_carry2_cc",
2490 "*add<mode>3_alc_carry2_cconly"): New insn definitions.
2492 2007-08-30 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
2494 * config/s390/s390.c (s390_expand_addcc): Emit canonical alc
2496 * config/s390/s390.md ("*add<mode>3_alc_cc", "*add<mode>3_alc",
2497 "addti3", "*adddi3_31z", "*scond<mode>"): Make alc pattern canonical.
2499 2007-08-29 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2501 * system.h: Activate -Wcast-qual as warning-only.
2503 2007-08-29 Paolo Bonzini <bonzini@gnu.org>
2506 * config/rs6000/rs6000.c (compare_section_name): New function.
2507 (rs6000_elf_in_small_data_p): Compare section prefixes instead
2510 2007-08-29 Olivier Hainque <hainque@adacore.com>
2512 * xcoffout.c: #include debug.h.
2513 * Makefile.in (xcoffout.o): Add debug.h dependency.
2515 2007-08-29 Uros Bizjak <ubizjak@gmail.com>
2517 * simplify-rtx.c (simplify_binary_operation_1) [VEC_SELECT]:
2518 Change CONST_VECTOR_ELT to XVECEXP in elem calculation.
2520 2007-08-29 Richard Guenther <rguenther@suse.de>
2522 * tree-pretty-print.c (dump_generic_node): Print
2523 label DECL_UID as D.%u.
2524 * tree-dfa.c (dump_variable): Print DECL_UID as D.%u.
2526 2007-08-29 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2528 * gcc/builtin-types.def (BT_FN_PTR_PTR_SIZE): New type.
2529 * gcc/builtins.def (BUILT_IN_REALLOC): New builtin.
2531 2007-08-29 Douglas Gregor <doug.gregor@gmail.com>
2534 * tree.c (build_type_attribute_qual_variant): Set canonical types
2535 on the final, unqualified attribute variant before building the
2538 2007-08-29 Jie Zhang <jie.zhang@analog.com>
2540 * config/bfin/bfin.c (bfin_expand_builtin): Fix the argument
2541 order of __builtin_bfin_cmplx_mac and __builtin_bfin_cmplx_msu.
2543 2007-08-29 Jie Zhang <jie.zhang@analog.com>
2546 2007-08-29 Jie Zhang <jie.zhang@analog.com>
2547 * config/bfin/bfin.md (composev2hi): Put operands into vector
2550 2007-08-29 Jie Zhang <jie.zhang@analog.com>
2552 * config/bfin/bfin.md (composev2hi): Put operands into vector
2555 2007-08-29 Jie Zhang <jie.zhang@analog.com>
2557 * config/bfin/bfin.c (bfin_expand_call): Inline PLT with l1_text
2558 attribute when appropriate.
2559 (bfin_handle_l1_text_attribute): New.
2560 (bfin_handle_l1_data_attribute): New.
2561 (bfin_attribute_table): Add attributes: l1_text, l1_data,
2562 l1_data_A and l1_data_B.
2563 * doc/extend.texi (node Function Attributes): Document l1_text
2565 (Variable Attributes): Add Blackfin subsection. Document l1_data,
2566 l1_data_A and l1_data_B variable attributes.
2568 2007-08-28 Jie Zhang <jie.zhang@analog.com>
2570 * config/bfin/bfin.opt (minline-plt): Add.
2571 * config/bfin/bfin.c (bfin_expand_call): Inline PLT when emit
2572 call to global functions.
2573 * doc/invoke.texi (Option Summary): Mention -minline-plt.
2574 (Blackfin Options): Document -minline-plt.
2576 2007-08-28 Nathan Sidwell <nathan@codesourcery.com>
2578 * config/m68k/m68k.c (m68k_get_function_kind): Assert we're never
2579 given a non-function.
2580 (m68k_ok_for_sibcall_p): Only sibcall functions of the same kind.
2582 2007-08-28 DJ Delorie <dj@redhat.com>
2584 * config/sh/sh.c (sh_gimplify_va_arg_expr): Fix sh2a support.
2586 2007-08-28 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2588 * system.h (CONST_CAST): Avoid union for gcc-4.0.x.
2590 2007-08-28 Richard Guenther <rguenther@suse.de>
2592 * tree.h (struct tree_function_decl): Increase size of
2593 function_code bitfield.
2595 2007-08-28 Nathan Sidwell <nathan@codesourcery.com>
2596 Kazu Hirata <kazu@codesourcery.com>
2598 * gcc/config/m68k/linux.h
2599 (M68K_HONOR_TARGET_STRICT_ALIGNMENT): Redefine as 0.
2600 * config/m68k/m68k.c (TARGET_RETURN_IN_MEMORY): New.
2601 (m68k_return_in_memory): New.
2602 * gcc/config/m68k/m68k.h (M68K_HONOR_TARGET_STRICT_ALIGNMENT):
2605 2007-08-28 Uros Bizjak <ubizjak@gmail.com>
2608 * simplify-rtx.c (simplify_binary_operation_1) [VEC_SELECT]:
2609 Simplify nested VEC_SELECT (with optional VEC_CONCAT operator as
2610 operand) when top VEC_SELECT extracts scalar element.
2611 * config/i386/sse.md (*vec_extract_v4si_mem): New pattern.
2612 (*vec_extract_v4sf_mem): Ditto.
2614 2007-08-28 Jakub Jelinek <jakub@redhat.com>
2617 * passes.c (init_optimization_passes): Move pass_df_finish
2618 after pass_postreload sublist.
2620 PR rtl-optimization/33148
2621 * simplify-rtx.c (simplify_unary_operation_1): Only optimize
2622 (neg (lt X 0)) if X has scalar int mode.
2625 * dwarf2out.c (rtl_for_decl_init): If vector decl has CONSTRUCTOR
2626 initializer, use build_vector_from_ctor if possible to create
2627 VECTOR_CST out of it. If vector initializer is not VECTOR_CST
2628 even after this, return NULL.
2630 2007-08-28 Gabriel Dos Reis <gdr@integrable-solutions.net>
2632 * c-pretty-print.c (pp_c_constant): Handle COMPLEX_CST.
2634 2007-08-28 Richard Guenther <rguenther@suse.de>
2636 * tree.h (DECL_DISREGARD_INLINE_LIMITS): New.
2637 (struct tree_function_decl): Make function_code a bitfield.
2638 Add disregard_inline_limits flag.
2639 * cgraphunit.c (cgraph_process_new_functions): Check
2640 DECL_DISREGARD_INLINE_LIMITS instead of disregard_inline_limits_p.
2641 (cgraph_preserve_function_body_p): Likewise.
2642 * ipa-inline.c (compute_inline_parameters): Likewise.
2643 * c-decl.c (finish_function): Set DECL_DISREGARD_INLINE_LIMITS
2644 for GNU C extern inline functions.
2645 (merge_decls): Merge DECL_DISREGARD_INLINE_LIMITS.
2646 * tree-inline.c (disregard_inline_limits_p): Remove.
2647 * tree-inline.h (disregard_inline_limits_p): Likewise.
2648 * c-common.c (handle_always_inline_attribute): Set
2649 DECL_DISREGARD_INLINE_LIMITS.
2650 * langhooks.c (add_builtin_function): Verify the function code
2651 fits in the bitfield.
2653 2007-08-28 Mircea Namolaru <namolaru@il.ibm.com>
2654 Vladimir Yanovsky <yanov@il.ibm.com>
2655 Revital Eres <eres@il.ibm.com>
2656 Andrey Belevantsev <abel@ispras.ru>
2658 * config/spu/spu.md: Recognize doloop pattern when -fmodulo-sched
2660 * modulo-sched.c: Add documentation regarding do-loop.
2661 (doloop_register_get): Change number of arguments to support
2662 the new do-loop pattern and check whether COUNT_REG has no other
2663 occurences in the loop besides in the control part.
2664 (duplicate_insns_of_cycles): Do not duplicate the insn which
2665 changes count_reg as it is already adjusted.
2666 (generate_prolog_epilog): New argument to support the new
2667 do-loop pattern. Change the subtract instruction to use
2668 expand_simple_binop. Call duplicate_insns_of_cycles with new
2670 (sms_schedule): Call doloop_register_get and
2671 generate_prolog_epilog with new argument. Do not handle loops
2672 with single sets insns with subreg in their lhs.
2673 * loop-doloop.c (doloop_optimize): Support for another do-loop
2675 (doloop_condition_get): Gets an instruction instead of a pattern
2676 and change the return condition when the do-loop pattern is
2678 * ddg.c (create_ddg_dep_from_intra_loop_link): Handle only reg
2679 deps when considering to not create edges.
2681 2007-08-27 Alexandre Oliva <aoliva@redhat.com>
2683 * doc/extend.texi (gnu_inline funtion attribute): Document C++
2686 2007-08-27 Jason Merrill <jason@redhat.com>
2689 * gimplify.c (gimplify_modify_expr): Discard the assignment of
2690 zero-sized types after calling gimplify_modify_expr_rhs.
2692 2007-08-27 Sandra Loosemore <sandra@codesourcery.com>
2694 * regclass.c (init_reg_autoinc): Fix typo.
2696 2007-08-27 Daniel Berlin <dberlin@dberlin.org>
2698 Fix PR tree-optimization/33173
2699 * tree-ssa-alias.c (find_used_portions): Fix reversed test.
2701 2007-08-27 H.J. Lu <hongjiu.lu@intel.com>
2704 * config/i386/i386.h (VALID_DFP_MODE_P): New.
2705 * config/i386/i386.c (ix86_hard_regno_mode_ok): Allow DFP in
2708 2007-08-27 Sandra Loosemore <sandra@codesourcery.com>
2709 David Ung <davidu@mips.com>
2710 Nigel Stephens <nigel@mips.com>
2712 Separate target-specific initialization from general
2713 back-end initialization.
2715 * toplev.c (init_alignments): New, split out from...
2716 (process_options): ...here.
2717 (backend_init_target): New, split out from...
2718 (backend_init): ...here.
2719 (lang_dependent_init_target): New, split out from...
2720 (lang_dependent_init): ...here.
2721 (target_reinit): New.
2722 * toplev.h (target_reinit): Declare.
2723 * expr.c (init_expr_target): Renamed from init_expr_once, since it
2724 now can be called more than once. Update comments.
2725 * expr.h (init_expr_target): Likewise.
2726 * alias.c (init_alias_target): Renamed from init_alias_once, since it
2727 now can be called more than once. Explicitly zero
2728 static_reg_base_value.
2729 * emit-rtl.c (init_emit_regs): New, split out from...
2730 (init_emit_once): Here.
2731 * regclass.c (initial_fixed_regs, initial_call_used_regs): Make
2732 non-const, so that changes from command-line arguments can overwrite
2733 values provided by the static initializers.
2734 (initial_call_really_used_regs): New, used similarly to the above.
2735 (initial_reg_names): Likewise.
2736 (last_mode_for_init_move_cost): Promoted function-local static to
2737 file-scope static to make it accessible outside init_move_cost.
2738 (init_reg_sets): Do not initialize fixed_regs and call_used_regs yet.
2739 Do not initialize inv_reg_alloc_order yet, either. Do initialize
2740 reg_names since it is needed for parsing command-line options.
2741 (init_move_cost): Use last_mode_for_init_move_cost instead of
2742 function-local static.
2743 (init_reg_sets_1): Initialize fixed_regs, call_used_regs, and
2744 call_really_used_regs now. Reinitialize reg_names. Also
2745 initialize inv_reg_alloc_order. Zero reg_class_subunion and
2746 reg_class_superunion. Clear losing_caller_save_reg_set.
2747 Preserve global registers if called more than once. Reset
2748 move cost, may_move_in_cost, may_move_out_cost, and
2749 last_mode_for_init_move_cost.
2750 (init_reg_modes_target): Renamed from init_reg_modes_once, since it
2751 can now be invoked more than once. Update comments.
2752 (init_regs): Update comments.
2753 (fix_register): Update initial_fixed_regs, initial_call_used_regs,
2754 and initial_call_really_used_regs, instead of the non-initial
2755 variables. This allows us to save the command-line register settings
2756 after target reinitialization.
2757 (init_reg_autoinc): Zero forbidden_inc_dec_classes.
2758 * rtl.h (init_emit_regs): Declare.
2759 (init_reg_modes_target, init_alias_target): Renamed as described
2761 * reload1.c (init_reload): Update comments.
2762 * optabs.c (init_optabs): Likewise.
2763 * cfgloopanal.c (init_set_costs): Explicitly zero target_avail_regs.
2765 2007-08-26 Chao-ying Fu <fu@mips.com>
2767 * rtl.h (XCNMPFV): Preserve const-ness of parameters through use of
2770 2007-08-26 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2772 * alias.c (memory_modified_1, memory_modified_in_insn_p):
2774 * basic-block.h (tree_predicted_by_p, rtl_predicted_by_p):
2776 * bb-reorder.c (copy_bb_p): Likewise.
2777 * cfghooks.c (can_remove_branch_p, can_merge_blocks_p,
2778 predicted_by_p, can_duplicate_block_p, block_ends_with_call_p,
2779 block_ends_with_condjump_p): Likewise.
2780 * cfghooks.h (can_remove_branch_p, can_merge_blocks_p,
2781 predicted_by_p, can_duplicate_block_p, block_ends_with_call_p,
2782 block_ends_with_condjump_p): Likewise.
2783 * cfglayout.c (cfg_layout_can_duplicate_bb_p): Likewise.
2784 * cfgrtl.c (can_delete_note_p, can_delete_label_p,
2785 rtl_can_merge_blocks, cfg_layout_can_merge_blocks_p,
2786 rtl_block_ends_with_call_p, rtl_block_ends_with_condjump_p,
2787 need_fake_edge_p, rtl_can_remove_branch_p): Likewise.
2788 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Use
2789 debug_true_const_tree instad of debug_true_tree.
2790 * debug.c (do_nothing_debug_hooks): Likewise.
2791 * debug.h (ignore_block, debug_true_const_tree): Constify.
2792 * dwarf2out.c (stack_adjust_offset, clobbers_queued_reg_save,
2793 dwarf2out_ignore_block, is_pseudo_reg, is_tagged_type,
2794 decl_ultimate_origin, block_ultimate_origin, lookup_decl_loc,
2795 same_dw_val_p, add_ranges, is_subrange_type, type_is_enum,
2796 dbx_reg_number, is_based_loc, field_type,
2797 simple_type_align_in_bits, simple_decl_align_in_bits,
2798 simple_type_size_in_bits, field_byte_offset, insert_float,
2799 type_tag, member_declared_type, is_redundant_typedef,
2800 secname_for_decl, is_inlined_entry_point): Likewise.
2801 * emit-rtl.c (const_next_insn, const_previous_insn,
2802 const_next_nonnote_insn, const_prev_nonnote_insn,
2803 const_next_real_insn, const_prev_real_insn,
2804 const_next_active_insn, const_prev_active_insn, const_next_label,
2805 const_prev_label): Likewise.
2806 * except.h (lookup_stmt_eh_region_fn, lookup_stmt_eh_region):
2808 * haifa-sched.c (may_trap_exp, haifa_classify_insn,
2809 find_set_reg_weight, no_real_insns_p, find_set_reg_weight,
2810 sched_insn_is_legitimate_for_speculation_p): Likewise.
2811 * jump.c (reversed_comparison_code_parts,
2812 reversed_comparison_code, reversed_comparison, condjump_label):
2814 * predict.c (rtl_predicted_by_p, tree_predicted_by_p): Likewise.
2815 * reg-stack.c (stack_regs_mentioned_p, stack_regs_mentioned):
2817 * rtl.h (const_previous_insn, const_next_insn,
2818 const_prev_nonnote_insn, const_next_nonnote_insn,
2819 const_prev_real_insn, const_next_real_insn,
2820 const_prev_active_insn, const_next_active_insn, const_prev_label,
2821 const_next_label, modified_between_p, modified_in_p,
2822 const_note_storeskeep_with_call_p, condjump_label,
2823 reversed_comparison, reversed_comparison_code,
2824 reversed_comparison_code_parts, memory_modified_in_insn_p,
2825 stack_regs_mentioned): Likewise.
2826 * rtlanal.c (modified_between_p, modified_in_p, const_note_stores,
2827 keep_with_call_p): Likewise.
2828 * sched-deps.c (deps_may_trap_p, sched_get_condition,
2829 conditions_mutex_p, sched_insns_conditions_mutex_psd_next_list,
2830 sd_lists_size, sd_lists_empty_p): Likewise.
2831 * sched-int.h (sched_insns_conditions_mutex_p,
2832 haifa_classify_insn, no_real_insns_p,
2833 sched_insn_is_legitimate_for_speculation_p, sd_next_list,
2834 sd_lists_size, sd_lists_empty_p): Likewise.
2835 * sdbout.c (sdb_debug_hooks): Likewise.
2836 * tree-cfg.c (tree_can_merge_blocks_p, is_ctrl_altering_stmt,
2837 stmt_ends_bb_p, tree_can_remove_branch_p, tree_can_duplicate_bb_p,
2838 -tree_block_ends_with_call_p, tree_block_ends_with_condjump_p):
2840 * tree-eh.c (lookup_stmt_eh_region_fn, lookup_stmt_eh_region,
2841 tree_can_throw_internal): Likewise.
2842 * tree-flow-inline.h (phi_nodes): Likewise.
2843 * tree-flow.h (phi_nodesstmt_ends_bb_p, is_ctrl_altering_stmt,
2844 tree_can_throw_internal, lookup_stmt_eh_region): Likewise.
2845 * tree-gimple.c (const_get_call_expr_in): Likewise.
2846 * tree-gimple.h (const_get_call_expr_in): Likewise.
2847 * tree.c (const_lookup_attribute, attribute_list_equal,
2848 attribute_list_contained): Likewise.
2849 * tree.h (attribute_list_equal, attribute_list_contained,
2850 const_lookup_attribute): Likewise.
2851 * vmsdbgout.c (vmsdbgout_ignore_block): Likewise.
2853 2007-08-26 Bernhard Fischer <aldot@gcc.gnu.org>
2855 * tree-ssa-sccvn.c, tree-ssa-sccvn.h, tree-vn.c,
2856 tree-ssa-pre.c: Remove unnecessary trailing whitespace.
2858 2007-08-26 H.J. Lu <hongjiu.lu@intel.com>
2861 * config/alpha/constraints.md: Delete "svn:mime-type" property
2862 with svn propdel "svn:mime-type".
2864 2007-08-26 H.J. Lu <hongjiu.lu@intel.com>
2867 * ifcvt.c (noce_get_alt_condition): Make sure that the previous
2868 non NOTE insn doesn't cross basic block.
2869 (noce_try_abs): Likewise.
2870 (noce_process_if_block): Likewise.
2872 2007-08-26 David Edelsohn <edelsohn@gnu.org>
2875 * config/rs6000/predicates.md (offsettable_mem_operand): Memory
2876 operand without auto-inc-dec.
2877 * config/rs6000/rs6000.md (floatsidf2_internal): Use
2878 offsettable_mem_operand.
2879 (floatunssidf2_internal): Same.
2880 (fix_truncdfsi2_internal): Same.
2881 (floatsidf_ppc64): Same.
2882 (floatunssidf_ppc64): Same.
2883 (fix_trunctfsi2_internal): Same.
2885 2007-08-26 Bernhard Fischer <aldot@gcc.gnu.org>
2888 * Makefile.in (libbackend.o): Add gcov-iov.h dependency and
2889 pass defines needed for version.c.
2891 2007-08-25 Joseph Myers <joseph@codesourcery.com>
2893 * configure.ac (--debug-prefix-map check): Change assembler
2894 version presumed to have feature to 2.18.
2895 * configure: Regenerate.
2897 2007-08-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2899 * c-common.c (vector_types_convertible_p,
2900 decl_with_nonnull_addr_p, c_promoting_integer_type_p,
2901 self_promoting_args_p): Constify.
2902 * c-common.h (has_c_linkage, decl_with_nonnull_addr_p,
2903 c_promoting_integer_type_p, self_promoting_args_p,
2904 anon_aggr_type_p, vector_types_convertible_p): Likewise.
2905 * c-decl.c (anon_aggr_type_p): Likewise.
2906 * * c-dump.c (dump_stmt): Likewise.
2907 * c-objc-common.c (has_c_linkage): Likewise.
2908 * c-tree.h (same_translation_unit_p): Likewise.
2909 * c-typeck.c (null_pointer_constant_p,
2910 tagged_types_tu_compatible_p, function_types_compatible_p,
2911 type_lists_compatible_p, lvalue_or_else, lvalue_p,
2912 comptypes_internal, struct tagged_tu_seen_cache,
2913 same_translation_unit_p, alloc_tagged_tu_seen_cache,
2914 c_size_in_bytes): Likewise.
2915 * ggc-common.c (compare_ptr_data, hash_descriptor, eq_descriptor,
2916 hash_ptr, eq_ptr): Likewise.
2917 * langhooks-def.h (lhd_decl_ok_for_sibcall,
2918 LANG_HOOKS_OMP_PRIVATIZE_BY_REFERENCE): Likewise.
2919 * langhooks.c (lhd_decl_ok_for_sibcall): Likewise.
2920 * langhooks.h (ok_for_sibcall, omp_privatize_by_reference):
2922 * optabs.c (optab_for_tree_code): Likewise.
2923 * optabs.h (optab_for_tree_code): Likewise.
2924 * rtl.h (simplify_rtx): Likewise.
2925 * simplify-rtx.c (simplify_rtx): Likewise.
2926 * tree-dump.c (queue, queue_and_dump_index, queue_and_dump_type,
2927 dump_flag, dump_node): Likewise.
2928 * tree-dump.h (struct dump_info, dump_stmt, queue_and_dump_index,
2929 queue_and_dump_type, dump_flag): Likewise.
2930 * tree-flow.h (vect_can_force_dr_alignment_p): Likewise.
2931 * tree-pass.h (dump_node): Likewise.
2932 * tree-vectorizer.c (vect_can_force_dr_alignment_p,
2933 supportable_narrowing_operation): Likewise.
2934 * tree-vectorizer.h (vect_can_force_dr_alignment_p,
2935 supportable_narrowing_operation): Likewise.
2936 * tree-vrp.c (needs_overflow_infinity, supports_overflow_infinity,
2937 is_negative_overflow_infinity, is_positive_overflow_infinity,
2938 is_overflow_infinity, vrp_val_is_max, vrp_val_is_min,
2939 nonnull_arg_p, get_value_range, vrp_operand_equal_p,
2940 update_value_range, add_equivalence, ssa_name_nonnegative_p,
2941 ssa_name_nonzero_p, fp_predicate): Likewise.
2942 * tree.c (auto_var_in_fn_p, empty_body_p): Likewise.
2943 * tree.h (empty_body_p, auto_var_in_fn_p, ssa_name_nonzero_p,
2944 ssa_name_nonnegative_p): Likewise.
2946 2007-08-25 Hans Kester <hans.kester@ellips.nl>
2948 * config.gcc : Add x86_64-elf target.
2950 2007-08-25 Jakub Jelinek <jakub@redhat.com>
2952 * expr.c (store_expr): Fix order of store_by_pieces arguments.
2954 2007-08-24 Sandra Loosemore <sandra@codesourcery.com>
2955 Nigel Stephens <nigel@mips.com>
2959 * doc/tm.texi (SET_RATIO, SET_BY_PIECES_P): Document new macros.
2960 (STORE_BY_PIECES_P): No longer applies to __builtin_memset.
2961 * expr.c (SET_BY_PIECES_P): Define.
2962 (can_store_by_pieces, store_by_pieces): Add MEMSETP argument; use
2963 it to decide whether to use SET_BY_PIECES_P or STORE_BY_PIECES_P.
2964 (store_expr): Pass MEMSETP argument to can_store_by_pieces and
2966 * expr.h (SET_RATIO): Define.
2967 (can_store_by_pieces, store_by_pieces): Update prototypes.
2968 * builtins.c (expand_builtin_memcpy): Pass MEMSETP argument to
2969 can_store_by_pieces/store_by_pieces.
2970 (expand_builtin_memcpy_args): Likewise.
2971 (expand_builtin_strncpy): Likewise.
2972 (expand_builtin_memset_args): Likewise. Also remove special case
2973 for optimize_size so that can_store_by_pieces/SET_BY_PIECES_P can
2974 decide what to do instead.
2975 * value-prof.c (tree_stringops_transform): Pass MEMSETP argument
2976 to can_store_by_pieces.
2978 * config/sh/sh.h (SET_BY_PIECES_P): Clone from STORE_BY_PIECES_P.
2979 * config/s390/s390.h (SET_BY_PIECES_P): Likewise.
2981 * config/mips/mips.opt (mmemcpy): Change from Var to Mask.
2982 * config/mips/mips.c (override_options): Make -Os default to -mmemcpy.
2983 * config/mips/mips.h (MIPS_CALL_RATIO): Define.
2984 (MOVE_RATIO, CLEAR_RATIO, SET_RATIO): Define.
2985 (STORE_BY_PIECES_P): Define.
2987 2007-08-24 Tom Tromey <tromey@redhat.com>
2989 * varpool.c (varpool_last_needed_node): Fix comment typo.
2990 * c-decl.c (duplicate_decls): Fix comment typo.
2991 (clone_underlying_type): Update comment.
2993 2007-08-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2995 * config/alpha/alpha.c (alpha_mangle_type, decl_has_samegp,
2996 alpha_in_small_data_p, alpha_split_complex_arg,
2997 alpha_stdarg_optimize_hook, TARGET_ASM_CAN_OUTPUT_MI_THUNK):
2999 * config/arm/arm-protos.h (arm_mangle_type): Likewise.
3000 * config/arm/arm.c (arm_comp_type_attributes, arm_mangle_type):
3002 * config/bfin/bfin.c (funkind, bfin_comp_type_attributes,
3003 TARGET_ASM_CAN_OUTPUT_MI_THUNK): Likewise.
3004 * config/darwin-protos.h (darwin_binds_local_p): Likewise.
3005 * config/darwin.c (darwin_binds_local_p): Likewise.
3006 * config/frv/frv.c (frv_string_begins_with, frv_in_small_data_p):
3008 * config/i386/i386-protos.h (i386_pe_binds_local_p,
3009 i386_pe_valid_dllimport_attribute_p): Likewise.
3010 * config/i386/i386.c (ix86_function_regparm,
3011 ix86_comp_type_attributes, ix86_ms_bitfield_layout_p,
3012 x86_can_output_mi_thunk, ix86_mangle_type,
3013 TARGET_USE_BLOCKS_FOR_CONSTANT_P): Likewise.
3014 * config/i386/winnt.c (i386_pe_valid_dllimport_attribute_p,
3015 i386_pe_binds_local_p): Likewise.
3016 * config/ia64/ia64.c
3017 (ia64_first_cycle_multipass_dfa_lookahead_guard_spec,
3018 ia64_needs_block_p, ia64_in_small_data_p, ia64_mangle_type,
3019 ia64_invalid_conversion, ia64_invalid_unary_op,
3020 ia64_invalid_binary_op, TARGET_ASM_CAN_OUTPUT_MI_THUNK):
3022 * config/m32c/m32c.c (m32c_comp_type_attributes): Likewise.
3023 * config/m32r/m32r.c (m32r_in_small_data_p): Likewise.
3024 * config/m68k/m68k.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Likewise.
3025 * config/mips/mips.c (mips_use_blocks_for_constant_p,
3026 mips_in_small_data_p, mips_use_anchors_for_symbol_p,
3027 mips_comp_type_attributes, TARGET_ASM_CAN_OUTPUT_MI_THUNK,
3028 mips_near_type_p, mips_far_type_p, mips_global_symbol_p,
3029 mips_symbol_binds_local_p, mips_classify_symbol): Likewise.
3030 * config/pa/pa.c (pa_commutative_p): Likewise.
3031 * config/rs6000/rs6000-protos.h (rs6000_elf_in_small_data_p):
3033 * config/rs6000/rs6000.c (rs6000_invalid_within_doloop,
3034 rs6000_ms_bitfield_layout_p, rs6000_mangle_type,
3035 rs6000_use_blocks_for_constant_p,
3036 rs6000_vector_alignment_reachable, rs6000_is_opaque_type,
3037 invalid_arg_for_unprototyped_fn, TARGET_ASM_CAN_OUTPUT_MI_THUNK,
3038 TARGET_SPLIT_COMPLEX_ARG, rs6000_elf_in_small_data_p): Likewise.
3039 * config/s390/s390.c (s390_mangle_type,
3040 TARGET_ASM_CAN_OUTPUT_MI_THUNK, TARGET_INVALID_WITHIN_DOLOOP):
3042 * config/score/score.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK,
3043 th_in_small_data_p): Likewise.
3044 * config/sh/sh.c (sh_ms_bitfield_layout_p,
3045 sh_dwarf_calling_convention, TARGET_ASM_CAN_OUTPUT_MI_THUNK):
3047 * config/sparc/sparc.c (sparc_can_output_mi_thunk,
3048 sparc_mangle_type): Likewise.
3049 * config/spu/spu.c (spu_vector_alignment_reachable): Likewise.
3050 * config/xtensa/xtensa.c (TARGET_SPLIT_COMPLEX_ARG): Likewise.
3051 * emit-rtl.c (const_fixed_htab_hash, const_fixed_htab_eq):
3053 * hooks.c (hook_bool_mode_const_rtx_false,
3054 hook_bool_mode_const_rtx_true,
3055 hook_bool_const_tree_hwi_hwi_const_tree_false,
3056 hook_bool_const_tree_hwi_hwi_const_tree_true,
3057 hook_int_const_tree_0, hook_int_const_tree_const_tree_1,
3058 hook_constcharptr_const_tree_null,
3059 hook_constcharptr_const_rtx_null,
3060 hook_constcharptr_const_tree_const_tree_null,
3061 hook_constcharptr_int_const_tree_null,
3062 hook_constcharptr_int_const_tree_const_tree_null): New.
3063 (hook_bool_mode_rtx_false, hook_bool_mode_rtx_true,
3064 hook_bool_tree_hwi_hwi_tree_false,
3065 hook_bool_tree_hwi_hwi_tree_true, hook_int_tree_0,
3066 hook_int_tree_tree_1, hook_constcharptr_tree_null,
3067 hook_constcharptr_rtx_null, hook_constcharptr_tree_tree_null,
3068 hook_constcharptr_int_tree_null,
3069 hook_constcharptr_int_tree_tree_null): Delete.
3070 (default_can_output_mi_thunk_no_vcall): Constify.
3071 * hooks.h (hook_bool_mode_rtx_false, hook_bool_mode_rtx_true,
3072 hook_bool_tree_hwi_hwi_tree_false,
3073 hook_bool_tree_hwi_hwi_tree_true, hook_int_tree_0,
3074 hook_int_tree_tree_1, hook_constcharptr_tree_null,
3075 hook_constcharptr_rtx_null, hook_constcharptr_tree_tree_null,
3076 hook_constcharptr_int_tree_null,
3077 hook_constcharptr_int_tree_tree_null): Delete.
3078 (hook_bool_mode_const_rtx_false, hook_bool_mode_const_rtx_true,
3079 hook_bool_const_tree_hwi_hwi_const_tree_false,
3080 hook_bool_const_tree_hwi_hwi_const_tree_true,
3081 hook_int_const_tree_0, hook_int_const_tree_const_tree_1,
3082 hook_constcharptr_const_tree_null,
3083 hook_constcharptr_const_rtx_null,
3084 hook_constcharptr_const_tree_const_tree_null,
3085 hook_constcharptr_int_const_tree_null,
3086 hook_constcharptr_int_const_tree_const_tree_null): New.
3087 (default_can_output_mi_thunk_no_vcall): Constify.
3088 * integrate.c (function_attribute_inlinable_p): Likewise.
3089 * integrate.h (function_attribute_inlinable_p): Likewise.
3090 * jump.c (rtx_renumbered_equal_p): Likewise.
3091 * output.h (decl_readonly_section, categorize_decl_for_section,
3092 default_use_anchors_for_symbol_p, default_binds_local_p,
3093 default_binds_local_p_1): Likewise.
3094 * rtl.h (rtx_renumbered_equal_p, decl_default_tls_model):
3096 * target-def.h (TARGET_ASM_CAN_OUTPUT_MI_THUNK,
3097 TARGET_VALID_DLLIMPORT_ATTRIBUTE_P, TARGET_VECTOR_OPAQUE_P,
3098 TARGET_COMMUTATIVE_P, TARGET_USE_BLOCKS_FOR_CONSTANT_P,
3099 TARGET_COMP_TYPE_ATTRIBUTES,
3100 TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P,
3101 TARGET_MS_BITFIELD_LAYOUT_P, TARGET_MANGLE_TYPE,
3102 TARGET_IN_SMALL_DATA_P, TARGET_INVALID_CONVERSION,
3103 TARGET_INVALID_UNARY_OP, TARGET_INVALID_BINARY_OP,
3104 TARGET_DWARF_CALLING_CONVENTION): Likewise.
3105 * target.h (can_output_mi_thunk, needs_block_p,
3106 first_cycle_multipass_dfa_lookahead_guard_spec,
3107 vector_alignment_reachable, comp_type_attributes,
3108 function_attribute_inlinable_p, ms_bitfield_layout_p, mangle_type,
3109 commutative_p, use_blocks_for_constant_p,
3110 use_anchors_for_symbol_p, in_small_data_p, binds_local_p,
3111 vector_opaque_p, dwarf_calling_convention, stdarg_optimize_hook,
3112 invalid_within_doloop, valid_dllimport_attribute_p,
3113 split_complex_arg, invalid_arg_for_unprototyped_fn,
3114 invalid_conversion, invalid_unary_op, invalid_binary_op):
3116 * targhooks.c (default_invalid_within_doloop,
3117 hook_invalid_arg_for_unprototyped_fn,
3118 default_builtin_vector_alignment_reachable): Likewise.
3119 (hook_bool_rtx_commutative_p): Delete.
3120 (hook_bool_const_rtx_commutative_p): New.
3121 * targhooks.h (default_invalid_within_doloop,
3122 default_builtin_vector_alignment_reachable,
3123 hook_invalid_arg_for_unprototyped_fn): Constify.
3124 (hook_bool_rtx_commutative_p): Delete.
3125 (hook_bool_const_rtx_commutative_p): New.
3126 * varasm.c (bss_initializer_p, decl_default_tls_model,
3127 categorize_decl_for_section, decl_readonly_section,
3128 default_use_anchors_for_symbol_p, default_binds_local_p,
3129 default_binds_local_p_1): Constify.
3131 2007-08-24 Jie Zhang <jie.zhang@analog.com>
3133 * config/bfin/bfin.h (TARGET_CPU_CPP_BUILTINS): Define
3134 __NO_BUILTIN if -fno-builtin.
3136 2007-08-24 Jie Zhang <jie.zhang@analog.com>
3138 * config/bfin/bfin.c (print_operand): Report error instead of
3139 ICE for wrong operand.
3141 2007-08-24 Michael Matz <matz@suse.de>
3143 * Makefile.in (GTFILES_H): Use $(patsubst) instead of $(subst).
3145 2007-08-24 Richard Guenther <rguenther@suse.de>
3148 * tree-ssa.c (useless_type_conversion_p): Split into a
3149 recursive and a non-recursive part.
3150 (useless_type_conversion_p_1): New function.
3151 * tree-ssa-ccp.c (fold_stmt_r): Make sure that the result
3152 from maybe_fold_offset_to_reference is trivially convertible
3153 to the desired type before doing the simplification.
3155 2007-08-24 Jakub Jelinek <jakub@redhat.com>
3157 * expr.c (store_expr): Optimize initialization of an array
3159 * expr.h (builtin_strncpy_read_str): New prototype.
3160 * builtins.c (builtin_strncpy_read_str): Remove prototype.
3163 2007-08-24 Uros Bizjak <ubizjak@gmail.com>
3166 * ifcvt.c (noce_find_if_block): Do not clear if_info.cond_earliest
3167 field when initializing if_info structure.
3169 2007-08-24 Richard Guenther <rguenther@suse.de>
3171 * Makefile.in (tree-inline.o): Add $(TARGET_H) and $(INTEGRATE_H)
3173 * c-objc-common.c (c_cannot_inline_tree_fn): Remove.
3174 * langhooks.c (lhd_tree_inlining_cannot_inline_tree_fn): Likewise.
3175 * tree-inline.c (inlinable_function_p): Fold in common parts of
3176 the cannot_inline_tree_fn langhook.
3177 * langhooks-def.h (lhd_tree_inlining_cannot_inline_tree_fn): Remove.
3178 (LANG_HOOKS_TREE_INLINING_CANNOT_INLINE_TREE_FN): Likewise.
3179 (LANG_HOOKS_TREE_INLINING_INITIALIZER): Remove initializer for
3180 cannot_inline_tree_fn langhook.
3181 * langhooks.h (struct lang_hooks_for_tree_inlining): Remove
3182 cannot_inline_tree_fn member.
3184 2007-08-24 Richard Guenther <rguenther@suse.de>
3186 * expr.c (get_inner_reference): Do computation of bitoffset
3187 from offset in a way we can detect overflow reliably.
3189 2007-08-24 Jie Zhang <jie.zhang@analog.com>
3191 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define for uClibc.
3193 2007-08-24 Jie Zhang <jie.zhang@analog.com>
3195 * config/bfin/bfin.opt (mfast-fp): Add.
3196 * config/bfin/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Let
3197 libbffastfp override libgcc if -mfast-fp.
3198 * config/bfin/bfin.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
3199 * doc/invoke.texi (Option Summary): Mention -mfast-fp.
3200 (Blackfin Options): Document -mfast-fp.
3202 2007-08-24 Jakub Jelinek <jakub@redhat.com>
3204 PR tree-optimization/32573
3206 * tree-data-ref.c (initialize_data_dependence_relation): Clear
3207 DDR_SUBSCRIPTS, DDR_DIR_VECTS and DDR_DIST_VECTS at the beginning.
3208 (finalize_ddr_dependent): Clear DDR_SUBSCRIPTS after freeing it.
3209 (build_classic_dist_vector): Return false rather than true if
3210 DDR_ARE_DEPENDENT is non-NULL at the beginning. Return false
3211 if either subscript_dependence_tester_1 or build_classic_dist_vector_1
3212 returned false. Don't call save_dist_v before calling
3213 build_classic_dist_vector_1.
3214 (free_dependence_relation): Don't guard freeing DDR_SUBSCRIPTS
3215 with NULL DDR_ARE_DEPENDENT. Also free DDR_DIST_VECTS and/or
3216 DDR_DIR_VECTS vectors.
3218 2007-08-23 Brian Sidebotham <brian.sidebotham@gmail.com>
3220 * configure.ac (leb128): Modify sed statement to work with any
3221 binutils version string.
3222 * configure: Regenerate
3224 2007-08-23 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
3226 * config/arc/arc-protos.h arc_select_cc_mode, gen_compare_reg):
3227 Wrap in RTX_CODE macro guard.
3228 * config/arm/pe.c (arm_dllexport_p, arm_dllimport_p,
3229 arm_dllexport_name_p, arm_dllimport_name_p, arm_mark_dllexport,
3230 arm_mark_dllimport, arm_pe_encode_section_info,
3231 arm_pe_unique_section): Use ISO-C function declarations.
3232 * config/c4x/c4x-c.c (c4x_parse_pragma, c4x_pr_CODE_SECTION,
3233 c4x_pr_DATA_SECTION, c4x_pr_FUNC_IS_PURE,
3234 c4x_pr_FUNC_NEVER_RETURNS, c4x_pr_INTERRUPT, c4x_pr_ignored):
3236 * config/iq2000/iq2000.h (ASM_OUTPUT_SKIP): Fix format warning.
3237 * config/m68hc11/m68hc11.h (ASM_OUTPUT_EXTERNAL): Undef before
3239 * config/mips/mips.h (ASM_DECLARE_OBJECT_NAME): Fix format
3241 * config/mn10300/mn10300.h (OUTPUT_ADDR_CONST_EXTRA): Likewise.
3242 * config/pdp11/pdp11.c (pdp11_output_function_epilogue): Likewise.
3243 (register_move_cost): Use ISO-C function declarations.
3244 * config/pdp11/pdp11.h (PRINT_OPERAND): Fix format warnings.
3245 * config/score/score-protos.h (score_declare_object): Add
3247 * config/score/score.h (ASM_DECLARE_OBJECT_NAME): Fix format
3249 * final.c (profile_function): Avoid empty if-bodies.
3252 * calls.c (must_pass_in_stack_var_size,
3253 must_pass_in_stack_var_size_or_pad): Constify.
3254 * config/alpha/alpha-protos.h (function_value): Likewise.
3255 * config/alpha/alpha.c (alpha_return_in_memory,
3256 alpha_pass_by_reference, function_value,
3257 unicosmk_must_pass_in_stack, TARGET_PROMOTE_FUNCTION_ARGS,
3258 TARGET_PROMOTE_FUNCTION_RETURN, TARGET_PROMOTE_PROTOTYPES):
3260 * config/arc/arc.c (arc_return_in_memory, arc_pass_by_reference,
3261 TARGET_PROMOTE_FUNCTION_ARGS, TARGET_PROMOTE_FUNCTION_RETURN,
3262 TARGET_PROMOTE_PROTOTYPES): Likewise.
3263 * config/arm/arm-protos.h (arm_return_in_memory,
3264 arm_pad_arg_upward, arm_function_value): Likewise.
3265 * config/arm/arm.c (arm_pass_by_reference,
3266 arm_promote_prototypes, arm_return_in_msb, arm_must_pass_in_stack,
3267 TARGET_PROMOTE_FUNCTION_ARGS, TARGET_PROMOTE_FUNCTION_RETURN,
3268 arm_function_value, arm_return_in_memory, arm_pad_arg_upward):
3270 * config/arm/arm.h (ARM_DECLARE_FUNCTION_NAME): Likewise.
3271 * config/avr/avr-protos.h (avr_function_value): Likewise.
3272 * config/avr/avr.c (avr_return_in_memory,
3273 gas_output_limited_string, gas_output_ascii, avr_function_value,
3274 avr_return_in_memory): Likewise.
3275 * config/bfin/bfin-protos.h (bfin_return_in_memory): Likewise.
3276 * config/bfin/bfin.c (bfin_pass_by_reference,
3277 bfin_return_in_memory, TARGET_PROMOTE_PROTOTYPES,
3278 TARGET_PROMOTE_FUNCTION_ARGS, TARGET_PROMOTE_FUNCTION_RETURN):
3280 * config/cris/cris.c (cris_pass_by_reference,
3281 TARGET_PROMOTE_FUNCTION_ARGS): Likewise.
3282 * config/crx/crx.c (crx_return_in_memory): Likewise.
3283 * config/darwin.c (function_base, machopic_function_base_name):
3285 * config/fr30/fr30.c (fr30_must_pass_in_stack,
3286 TARGET_PROMOTE_PROTOTYPES): Likewise.
3287 * config/frv/frv.c (frv_must_pass_in_stack): Likewise.
3288 * config/h8300/h8300.c (h8300_return_in_memory): Likewise.
3289 * config/i386/i386-protos.h (ix86_return_in_memory,
3290 ix86_sol10_return_in_memory): Likewise.
3291 * config/i386/i386.c (ix86_function_value,
3292 ix86_function_sseregparm, ix86_must_pass_in_stack,
3293 type_natural_mode, classify_argument, examine_argument,
3294 construct_container, ix86_pass_by_reference, function_value_32,
3295 function_value_64, ix86_function_value_1, return_in_memory_32,
3296 return_in_memory_64, return_in_memory_ms_64,
3297 ix86_return_in_memory, ix86_sol10_return_in_memory,
3298 TARGET_PROMOTE_PROTOTYPES): Likewise.
3299 * config/ia64/ia64-protos.h (ia64_function_value,
3300 ia64_hpux_function_arg_padding): Likewise.
3301 * config/ia64/ia64.c (hfa_element_mode, ia64_return_in_memory,
3302 ia64_function_value, bundle_state_hash, bundle_state_eq_p,
3303 ia64_hpux_function_arg_padding): Likewise.
3304 * config/iq2000/iq2000-protos.h (function_arg,
3305 iq2000_function_value): Likewise.
3306 * config/iq2000/iq2000.c (iq2000_return_in_memory,
3307 iq2000_pass_by_reference, TARGET_PROMOTE_FUNCTION_ARGS,
3308 TARGET_PROMOTE_FUNCTION_RETURN, TARGET_PROMOTE_PROTOTYPES,
3309 function_arg, iq2000_function_value): Likewise.
3310 * config/m32c/m32c-protos.h (m32c_function_value,
3311 m32c_promote_function_return): Likewise.
3312 * config/m32c/m32c.c (m32c_pass_by_reference,
3313 m32c_promote_prototypes, m32c_promote_function_return,
3314 m32c_function_value): Likewise.
3315 * config/m32r/m32r.c (m32r_return_in_memory,
3316 m32r_pass_by_reference, TARGET_PROMOTE_PROTOTYPES,
3317 m32r_in_small_data_p): Likewise.
3318 * config/m68hc11/m68hc11-protos.h (m68hc11_function_arg_padding):
3320 * config/m68hc11/m68hc11.c (m68hc11_return_in_memory,
3321 m68hc11_function_arg_padding): Likewise.
3322 * config/m68k/m68k-protos.h (m68k_function_value): Likewise.
3323 * config/m68k/m68k.c (TARGET_PROMOTE_PROTOTYPES,
3324 m68k_function_value): Likewise.
3325 * config/mcore/mcore-protos.h (mcore_num_arg_regs,
3326 mcore_function_value): Likewise.
3327 * config/mcore/mcore.c (handle_structs_in_regs,
3328 mcore_return_in_memory, TARGET_PROMOTE_FUNCTION_ARGS,
3329 TARGET_PROMOTE_FUNCTION_RETURN, TARGET_PROMOTE_PROTOTYPES,
3330 mcore_num_arg_regs, mcore_function_value): Likewise.
3331 * config/mips/mips-protos.h (mips_pad_arg_upward,
3332 mips_function_value): Likewise.
3333 * config/mips/mips.c (mips_fpr_return_fields, mips_return_in_msb,
3334 mips_return_in_memory, mips_pass_by_reference, mips_callee_copies,
3335 TARGET_PROMOTE_FUNCTION_ARGS, TARGET_PROMOTE_FUNCTION_RETURN,
3336 TARGET_PROMOTE_PROTOTYPES, mips_pad_arg_upward,
3337 mips_function_value): Likewise.
3338 * config/mmix/mmix-protos.h (mmix_function_outgoing_value):
3340 * config/mmix/mmix.c (mmix_pass_by_reference,
3341 TARGET_PROMOTE_FUNCTION_ARGS, mmix_function_outgoing_value,
3342 mmix_encode_section_info): Likewise.
3343 * config/mn10300/mn10300-protos.h (mn10300_function_value):
3345 * config/mn10300/mn10300.c (mn10300_return_in_memory,
3346 mn10300_pass_by_reference, TARGET_PROMOTE_PROTOTYPES,
3347 mn10300_function_value): Likewise.
3348 * config/mt/mt-protos.h (mt_function_value): Likewise.
3349 * config/mt/mt.c (mt_pass_by_reference, mt_function_value,
3350 mt_pass_in_stack, TARGET_PROMOTE_PROTOTYPES): Likewise.
3351 * config/pa/pa-protos.h (function_arg_padding, function_value,
3352 pa_return_in_memory): Likewise.
3353 * config/pa/pa.c (pa_pass_by_reference,
3354 TARGET_PROMOTE_FUNCTION_RETURN, TARGET_PROMOTE_PROTOTYPES,
3355 function_arg_padding, function_value, pa_return_in_memory):
3357 * config/pdp11/pdp11.c (pdp11_return_in_memory): Likewise.
3358 * config/rs6000/rs6000-protos.h (rs6000_function_value,
3359 function_arg_padding): Likewise.
3360 * config/rs6000/rs6000.c (rs6000_return_in_memory,
3361 rs6000_darwin64_record_arg_recurse, rs6000_darwin64_record_arg,
3362 rs6000_pass_by_reference, rs6000_must_pass_in_stack,
3363 TARGET_PROMOTE_FUNCTION_ARGS, TARGET_PROMOTE_FUNCTION_RETURN,
3364 function_arg_padding, altivec_expand_dst_builtin,
3365 altivec_expand_builtin, rs6000_expand_builtin, spe_init_builtins,
3366 altivec_init_builtins, rs6000_common_init_builtins,
3367 rs6000_function_value): Likewise.
OSDN Git Service
