1 2002-11-10 Hans-Peter Nilsson <hp@bitrange.com>
3 * config/mmix/mmix.h (FUNCTION_ARG_CALLEE_COPIES): Define the same
4 as FUNCTION_ARG_PASS_BY_REFERENCE.
6 2002-11-09 John David Anglin <dave@hiauly1.hia.nrc.ca>
8 * pa.h (STARTING_FRAME_OFFSET): Change offset for TARGET_64BIT to 16.
10 * config.gcc (hppa*64*-*-linux*): Shorten lines in tm_file define.
11 (hppa*64*-*-hpux11*): Likewise. Use elfos.h with gas.
12 * pa.c (output_millicode_call): Use symbol difference rather than
13 $PIC_pcrel$0 when using HP assembler.
14 * pa64-hpux.h (TARGET_GAS): Define to 1 or 0 depending on whether or
15 not elfos.h (i.e., gas) is being used.
16 (ASM_FILE_START, STRING_ASM_OP, TEXT_SECTION_ASM_OP,
17 DATA_SECTION_ASM_OP, BSS_SECTION_ASM_OP, ASM_OUTPUT_ALIGNED_COMMON,
18 ASM_OUTPUT_ALIGNED_LOCAL, GLOBAL_ASM_OP, ASM_DECLARE_FUNCTION_NAME,
19 ASM_OUTPUT_EXTERNAL, ASM_OUTPUT_EXTERNAL_LIBCALL,
20 ASM_OUTPUT_INTERNAL_LABEL, ASM_GENERATE_INTERNAL_LABEL): Define when
22 (TARGET_ASM_GLOBALIZE_LABEL): Undefine when using elfos.h.
23 (DWARF2_ASM_LINE_DEBUG_INFO): Delete.
24 (ASM_FILE_START): Add standard .SPACE and .SUBSPA defines when not
26 (TEXT_SECTION_ASM_OP, READONLY_DATA_SECTION_ASM_OP, DATA_SECTION_ASM_OP,
27 BSS_SECTION_ASM_OP): New HP style defines when not using elfos.h.
28 (TARGET_ASM_NAMED_SECTION, MAKE_DECL_ONE_ONLY, ASM_WEAKEN_LABEL):
29 Don't define when not using elfos.h.
30 (ASM_DECLARE_RESULT): Don't define.
31 * doc/install.texi (hppa*-hp-hpux*): Remove statement that HP assembler
32 doesn't work on hppa64-hp-hpux11.
33 (hppa*-hp-hpux11): Update.
35 2002-11-09 Jason Thorpe <thorpej@wasabisystems.com>
37 * config/mips/netbsd.h (SUBTARGET_ASM_SPEC): Don't pass -KPIC
38 to the assembler if -mno-abicalls was specified.
40 2002-11-09 John David Anglin <dave@hiauly1.hia.nrc.ca>
42 * pa-linux.h (PREFERRED_DEBUGGING_TYPE, DWARF2_ASM_LINE_DEBUG_INFO,
43 ASM_OUTPUT_DEF): Delete.
45 2002-11-09 Neil Booth <neil@daikokuya.co.uk>
47 * c-opts.c (COMMAND_LINE_OPTIONS): Fix -Wimplicit.
49 2002-11-08 Dale Johannesen <dalej@apple.com>
51 * dbxout.c (dbxout_type): Fix stabs info for vector types.
53 2002-11-08 Neil Booth <neil@daikokuya.co.uk>
57 * cpptrad.c (scan_out_logical_line): A '#' from a macro doesn't
58 start a directive. In assembler, #NUM is not a line directive.
60 2002-11-08 Neil Booth <neil@daikokuya.co.uk>
62 * cppmain.c (cpp_preprocess_file): Loop to pop any -included
65 2002-11-08 Kazu Hirata <kazu@cs.umass.edu>
67 * config/h8300/h8300.md (two anonymous test insns): New.
69 Fri Nov 8 11:20:19 CET 2002 Jan Hubicka <jh@suse.cz>
71 * jump.c (mark_jump_label): Handle subregs of label_refs.
73 2002-11-07 David Mosberger <davidm@hpl.hp.com>
75 * config/ia64/crtend.asm: Include "auto-host.h".
76 [HAVE_INITFINI_ARRAY]: Invoke __do_global_ctors_aux via .init_array.
77 * config/ia64/crtbegin.asm: Similarly.
78 * config/ia64/t-ia64 (crtbegin.o): Include from current directory.
79 (crtend.o, crtbeginS.o, crtendS.o): Likewise.
81 * aclocal.m4 (gcc_AC_INITFINI_ARRAY): New.
82 * configure.in: Use it if --enable-initfini-array not specified.
83 * doc/install.texi (Configuration): Document --enable-initfini-array.
84 * configure, config.in: Rebuild.
86 2002-11-07 Jason Thorpe <thorpej@wasabisystems.com>
88 * config/arm/arm-protos.h (arm_get_frame_size)
89 (thumb_get_frame_size): New prototypes.
90 * config/arm/arm.c (arm_get_frame_size)
91 (thumb_get_frame_size): New functions.
92 (use_return_insn, arm_output_epilogue, arm_output_function_epilogue)
93 (arm_compute_initial_elimination_offset, arm_expand_prologue): Use
95 (thumb_expand_prologue, thumb_expand_epilogue): Use
97 * config/arm/arm.h (PREFERRED_STACK_BOUNDARY): Define.
98 (machine_function): Add frame_size member.
99 (THUMB_INITIAL_ELIMINATION_OFFSET): Use thumb_get_frame_size.
101 2002-11-07 Richard Earnshaw <rearnsha@arm.com>
103 * arm.c (bit_count): Make argument unsigned long. Return unsigned.
104 Adjust code to use portable unsigned bit manipulation.
105 (insn_flags, tune_flags): Change type to unsigned.
106 (struct processors): Make flags unsigned long.
107 (arm_override_options): Change type of count and current_bit_count
110 2002-11-07 Richard Earnshaw <rearnsha@arm.com>
112 * arm/elf.h (TYPE_OPERAND_FMT): Prefix type with %.
114 Thu Nov 7 15:50:18 2002 J"orn Rennecke <joern.rennecke@superh.com>
116 * sh.h (DWARF_FRAME_RETURN_COLUMN): Use DWARF_FRAME_REGNUM.
118 Thu Nov 7 11:18:01 CET 2002 Jan Hubicka <jh@suse.cz>
120 * reg-stack.c (compensate_edge): Fix sanity check.
122 2002-11-05 Geoffrey Keating <geoffk@apple.com>
124 * config.gcc: Don't create crtbegin, crtend on Darwin; do create
125 crt2.o. Rearrange t-darwin makefiles.
126 * crtstuff.c [OBJECT_FORMAT_MACHO]: Delete.
127 * unwind-dw2-fde-darwin.c: New.
128 * unwind-dw2-fde-glibc.c: Correct comment.
129 * unwind-dw2-fde.c (__register_frame_info_bases)
130 [DWARF2_OBJECT_END_PTR_EXTENSION]: Clear fde_end.
131 (classify_object_over_fdes): Use last_fde.
132 (add_fdes): Likewise.
133 (linear_search_fdes): Likewise.
134 * unwind-dw2-fde.h (struct object)
135 [DWARF2_OBJECT_END_PTR_EXTENSION]: Add fde_end field.
137 * config/darwin.h (STARTFILE_SPEC): Include crt2.o not crtbegin.o.
138 (ENDFILE_SPEC): No crtend.o.
139 * config/t-darwin: New.
140 * config/i386/t-darwin: Delete.
141 * config/darwin-crt2.c: New.
142 * config/rs6000/t-darwin: Delete contents duplicated in t-rs6000
145 2002-11-06 David Edelsohn <edelsohn@gnu.org>
148 * config/rs6000/rs6000.md (movdi_internal64): Discourage
151 2002-11-06 Janis Johnson <janis187@us.ibm.com>
153 * doc/contrib.texi: Merge in the list from the Java web pages.
155 2002-11-06 David O'Brien <obrien@FreeBSD.org>
157 * config/sparc/freebsd: Fix typo.
159 2002-11-06 John David Anglin <dave@hiauly1.hia.nrc.ca>
161 * pa64-hpux.h (LDD_SUFFIX, PARSE_LDD_OUTPUT): Define.
163 2002-11-06 Alexandre Oliva <aoliva@redhat.com>
165 * config/mips/mips.md (call_value_multiple_internal2): Use dla for
166 non-SImode addresses.
168 2002-11-05 Bob Wilson <bob.wilson@acm.org>
170 * config/xtensa/elf.h (LIB_SPEC): Add "-lhal".
172 2002-11-05 John David Anglin <dave2hiauly1.hia.nrc.ca>
174 * pa64-hpux.h (LIB_SPEC): Fix p and pg options.
175 (STARTFILE_SPEC): Remove p and pg options.
177 2002-11-05 Andrew Haley <aph@redhat.com>
179 * fold-const.c (fold): Don't transform (a0 op compound(a1,a2))
180 to (compound(a1,a0 op a2)) if a0 or a1 have side effects.
182 2002-11-05 Richard Sandiford <rsandifo@redhat.com>
184 * config/mips/mips.h (CANNOT_CHANGE_MODE_CLASS): Move comment to...
185 * config/mips/mips.c (mips_cannot_change_mode_class): ...here.
187 2002-11-04 Dale Johannesen <dalej@apple.com>
189 * doloop.c (doloop_modify_runtime): Fix loop count computation
191 * loop.c (loop_invariant_p): Support calling from unroller.
193 2002-11-04 Ulrich Weigand <uweigand@de.ibm.com>
195 * config/s390/s390.c (s390_decompose_address): Use arg_pointer_rtx
198 2002-11-04 Aldy Hernandez <aldyh@redhat.com>
200 * hard-reg-set.h (REG_CANNOT_CHANGE_MODE_P): New.
202 * config/rs6000/rs6000.h (CLASS_CANNOT_CHANGE_MODE_P): Remove.
203 (CLASS_CANNOT_CHANGE_MODE): Remove.
204 (CANNOT_CHANGE_MODE_CLASS): New.
206 * config/alpha/alpha.h: Same.
208 * config/ia64/ia64.h: Same.
210 * config/mips/mips.h: Same.
212 * config/s390/s390.h: Same.
214 * config/sh/sh.h: Same.
216 * config/pa/pa64-regs.h: Same.
218 * config/sh/sh-protos.h (sh_cannot_change_mode_class): Add prototype.
220 * config/sh/sh.c (sh_cannot_change_mode_class): New.
222 * config/mips/mips-protos.h (mips_cannot_change_mode_class): Add
225 * config/mips/mips.c (mips_cannot_change_mode_class): New.
227 * doc/tm.texi (Register Classes): Remove
228 CLASS_CANNOT_CHANGE_MODE and CLASS_CANNOT_CHANGE_MODE_P.
229 Document CANNOT_CHANGE_MODE_CLASS.
231 * reload.c (push_reload): Use CANNOT_CHANGE_MODE_CLASS.
234 * simplify-rtx.c (simplify_subreg): Same.
236 * reload1.c (choose_reload_regs): Same.
238 * recog.c (register_operand): Same.
240 * regrename.c (mode_change_ok): Change to use new
241 CANNOT_CHANGE_MODE_CLASS infrastructure.
243 * regclass.c (cannot_change_mode_set_regs): New.
244 Declare subregs_of_mode.
245 (regclass): Use subregs_of_mode.
246 Remove references to reg_changes_mode.
247 (init_reg_sets_1): Remove class_can_change_mode and
248 reg_changes_mode code.
249 (invalid_mode_change_p): New.
250 (dump_regclass): Use invalid_mode_change_p instead of
251 class_can_change_mode.
253 (record_operand_costs): Do not set reg_changes_mode.
255 * local-alloc.c (struct qty): Remove changes_mode field.
256 (alloc_qty): Remove changes_mode initialization.
257 (update_qty_class): Remove set of changes_mode.
258 (find_free_reg): Use subregs_of_mode.
260 * global.c (find_reg): Use subregs_of_mode info.
262 * rtl.h (cannot_change_mode_set_regs): New prototype.
263 (invalid_mode_change_p): Same.
264 (REG_CANNOT_CHANGE_MODE_P): New macro.
266 * flow.c (mark_used_regs): Calculate subregs_of_mode. Remove
268 (life_analysis): Clear subregs_of_mode.
270 * combine.c (subst): Pass class to CLASS_CANNOT_CHANGE_MODE_P.
271 Remove use of CLASS_CANNOT_CHANGE_MODE.
272 (simplify_set): Same.
273 (gen_lowpart_for_combine): Calculate subregs_of_mode. Remove
276 * regs.h: Add extern for subregs_of_mode;
277 Include hard-reg-set and basic-block.
278 (REG_CHANGES_MODE): Delete.
280 2002-11-03 John David Anglin <dave@hiauly1.hia.nrc.ca>
282 * jump.c (never_reached_warning): Don't set contains_insn until the
283 first line note is seen.
285 2002-11-03 David Edelsohn <edelsohn@gnu.org>
287 * config/rs6000/rs6000.md (movti_string): Use string instructions.
289 2002-11-03 Roger Sayle <roger@eyesopen.com>
292 * c-typeck.c (c_expand_asm_operands): Defend against
293 error_mark_nodes in the output argument to avoid ICE.
295 2002-11-03 Eric Botcazou <ebotcazou@libertysurf.fr>
298 * genrecog.c (preds): Handle ADDRESSOF.
299 (validate_pattern): Mark it as an lvalue.
301 2002-11-02 David Edelsohn <edelsohn@gnu.org>
303 * config/rs6000/rs6000.c (rs6000_override_options): Use string
304 instructions when optimizing for size.
306 2002-11-02 Kazu Hirata <kazu@cs.umass.edu>
308 * config/h8300/h8300.h: Fix comment typos.
309 * config/h8300/h8300.md: Likewise.
310 * config/h8300/lib1funcs.asm: Likewise.
312 2002-11-02 Gerald Pfeifer <pfeifer@dbai.tuwien.ac.at>
316 *doc/install.texi (Installing GCC: Configuration): Clarify
317 the only supported ways to configure gcc.
319 2002-11-01 Kazu Hirata <kazu@cs.umass.edu>
321 * config/h8300/h8300.md (anonymous and:QI pattern): Use 'n'
322 instead of 'O' for the constraint for the second operand.
324 2002-11-01 Mark Mitchell <mark@codesourcery.com>
327 * toplev.c (rest_of_compilation): Do not refuse to output code for
328 an inline function in a local class.
330 2002-11-01 David O'Brien <obrien@FreeBSD.org>
332 * config/sparc/freebsd.h (CPP_CPU64_DEFAULT_SPEC): Define __arch64__.
333 (TRANSFER_FROM_TRAMPOLINE): Reformat.
336 2002-11-01 Kazu Hirata <kazu@cs.umass.edu>
338 * config/h8300/h8300.h (CAN_ELIMINATE): Simplify.
340 2002-11-01 Toshiyasu Morita <toshiyasu.morita@hsa.hitachi.com>
342 * config/h8300/h8300.h (OPTIMIZATION_OPTIONS): New.
344 2002-11-01 Steve Ellcey <sje@cup.hp.com>
346 * config/ia64/ia64.h (MASK_INLINE_DIV_LAT): Remove.
347 (MASK_INLINE_DIV_THR): Remove.
348 (TARGET_INLINE_DIV_LAT): Remove.
349 (TARGET_INLINE_DIV_THR): Remove.
350 (TARGET_INLINE_DIV): Remove.
351 (MASK_INLINE_FLOAT_DIV_LAT): New macro.
352 (MASK_INLINE_FLOAT_DIV_THR): New macro.
353 (MASK_INLINE_INT_DIV_LAT): New macro.
354 (MASK_INLINE_INT_DIV_THR): New macro.
355 (TARGET_INLINE_FLOAT_DIV_LAT): New macro.
356 (TARGET_INLINE_FLOAT_DIV_THR): New macro.
357 (TARGET_INLINE_INT_DIV_LAT): New macro.
358 (TARGET_INLINE_INT_DIV_THR): New macro.
359 (TARGET_INLINE_FLOAT_DIV): New macro.
360 (TARGET_INLINE_INT_DIV): New macro.
361 * config/ia64/ia64.md (divsi3): Change to use new macros.
365 (divsi3_internal): Ditto.
370 (divdi3_internal_lat): Ditto.
371 (divdi3_internal_thr): Ditto.
373 (divsf3_internal_lat): Ditto.
374 (divsf3_internal_thr): Ditto.
376 (divdf3_internal_lat): Ditto.
377 (divdf3_internal_thr): Ditto.
379 (divtf3_internal_lat): Ditto.
380 (divtf3_internal_thr): Ditto.
381 * config/ia64/ia64.c (ia64_override_options): Change
382 to check new macros for conflicts in settings.
383 * doc/invoke.texi (-minline-divide-min-latency): Remove.
384 (-minline-divide-max-throughput): Remove.
385 (-minline-float-divide-min-latency): New.
386 (-minline-float-divide-max-throughput): New.
387 (-minline-int-divide-min-latency): New.
388 (-minline-int-divide-max-throughput): New.
390 2002-11-01 Richard Earnshaw (rearnsha@arm.com)
393 * arm.c (use_return_insn): Don't use a return insn if there are
394 saved integer regs, but LR is not one of them.
396 Fri Nov 1 10:33:15 CET 2002 Jan Hubicka <jh@suse.cz>
398 * expr.c (emit_move_insn): Use SCALAR_FLOAT_MODE_P
399 * machmode.h (SCALAR_FLOAT_MODE_P): New macro.
401 2002-10-31 Nathanael Nerode <neroden@gcc.gnu.org>
404 * doc/md.texi: Document restriction on commutative operand
407 2002-10-31 Eric Christopher <echristo@redhat.com>
409 * explow.c (convert_memory_address): Use shallow_copy_rtx.
411 2002-10-31 Steve Ellcey <sje@cup.hp.com>
413 * expmed.c (store_bit_field): Check FUNCTION_ARG_REG_LITTLE_ENDIAN.
415 2002-10-31 Steve Ellcey <sje@cup.hp.com>
417 * config/ia64/hpux.h (MEMBER_TYPE_FORCES_BLK): Set for non-floats.
419 Thu Oct 31 Dale Johannesen <dalej@apple.com>
421 * config/rs6000/darwin.h: Correct formatting in previous.
423 Thu Oct 31 Dale Johannesen <dalej@apple.com>
425 * config/rs6000/darwin.h: Enable -falign-xxx options.
427 Thu Oct 31 18:08:00 CET 2002 Jan Hubicka <jh@suse.cz>
429 * i386.c (override_options): Set defaults for flag_omit_frame_pointer,
430 flag_asynchronous_unwind_tables, flag_pcc_struct_return.
431 * i386.c (optimization_options): Set flag_omit_frame_pointer,
432 flag_asynchronous_unwind_tables, flag_pcc_struct_return to 2.
433 Do not clear -momit-leaf-frame-pointer when profiling.
434 (ix86_frame_pointer_required): Frame pointer is allways required when
437 Thu Oct 31 16:09:44 CET 2002 Jan Hubicka <jh@suse.cz>
439 * i386.md (negdf2_ifs_rex64): Don't allow GPR operand.
441 Thu Oct 31 12:45:55 2002 J"orn Rennecke <joern.rennecke@superh.com>
443 * sh.h (binary_logical_operator): Declare.
444 * sh.c (binary_logical_operator): New function.
445 * sh.md (xordi3+1): New combiner splitter pattern.
447 2002-10-31 David O'Brien <obrien@FreeBSD.org>
449 * config/sparc/freebsd.h (TRANSFER_FROM_TRAMPOLINE): Define
450 __enable_execute_stack function.
452 2002-10-30 Aldy Hernandez <aldyh@redhat.com>
454 * c-common.c: Add GTY to vector_type_node_list.
456 2002-10-30 John David Anglin <dave@hiauly.hia.nrc.ca>
458 * pa-linux.h (ASM_OUTPUT_EXTERNAL_LIBCALL): Define.
459 * pa-protos.h (attr_length_millicode_call, attr_length_call,
460 pa_init_machine_status): Declare new global functions.
461 * pa.c (void copy_fp_args, length_fp_args, get_plabel): Declare and
462 implement new functions.
463 (attr_length_millicode_call, attr_length_call): Implement.
464 (total_code_bytes): Change type to long.
465 (pa_output_function_prologue): Compute total_code_bytes on TARGET_64BIT.
466 Reset counter if flag_function_sections.
467 (output_deferred_plabels): Set output alignment to 3 for TARGET_64BIT.
468 (output_cbranch): Move call to gen_label_rtx.
469 (output_millicode_call): Rewrite adding long TARGET_64BIT call, expose
470 delay slot in all variants, shorten pc-relative calls.
471 (output_call): Rewrite adding long TARGET_64BIT call, improved delay
472 slot usage and exposure, various new call variants, and shortened
473 sequences for some variants on TARGET_PA_20.
474 Miscellaneous format changes.
475 * pa.h (total_code_bytes): Change type to long.
476 (MASK_LONG_CALLS, TARGET_LONG_CALLS, TARGET_LONG_ABS_CALL,
477 TARGET_LONG_PIC_SDIFF_CALL, TARGET_LONG_PIC_PCREL_CALL): Define.
478 (TARGET_SWITCHES): Add "-mlong-calls" and "-mno-long-calls" options.
479 (EXTRA_CONSTRAINT, GO_IF_LEGITIMATE_ADDRESS,
480 LEGITIMIZE_RELOAD_ADDRESS): Don't use long floating point loads and
481 stores on TARGET_ELF32.
482 *pa.md (define_delay): Allow insns in delay on TARGET_PORTABLE_RUNTIME.
483 (unnamed patterns for mulsi3, divsi3, udivsi3, modsi3, umodsi3 and
484 canonicalize_funcptr_for_compare expanders): Calculate attribute length
485 attr_length_millicode_call().
486 (call_internal_symref, call_value_internal_symref): Clobber register 1.
487 Calculate attribute length using attr_length_call().
488 (call_internal_reg_64bit, call_value_internal_reg_64bit): Move gp load
490 (sibcall, sibcall_value): Rewrite.
491 (sibcall_internal_symref, sibcall_value_internal_symref): Clobber
492 register 1. Use attr_length_call().
493 (sibcall_internal_symref_64bit, sibcall_value_internal_symref_64bit):
495 (unamed pattern for canonicalize_funcptr_for_compare): Rewrite.
496 * som.h (MEMBER_TYPE_FORCES_BLK): Define.
497 * t-pa64 (TARGET_LIBGCC2_CFLAGS): Add "-mlong-calls".
498 * doc/invoke.texi (mlong-calls): Document.
500 2002-10-30 Roger Sayle <roger@eyesopen.com>
502 * fold-const.c (fold_binary_op_with_conditional_arg): Improve
503 handling of cases where one or both branches of the conditional
504 have void type, i.e. throw an exception or don't return.
505 (fold): Only apply (and undo) type conversion to the non-void
506 branches of a COND_EXPR.
508 2002-10-30 Mark Mitchell <mark@codesourcery.com>
511 * varasm.c (asm_output_aligned_bss): Do not call
514 2002-10-30 David Edelsohn <edelsohn@gnu.org>
515 Torbjorn Granlund <tege@swox.com>
517 * config/rs6000/rs6000.md (load_toc_v4_PIC_1): Use preferred form
519 (load_toc_v4_PIC_1b): Same.
521 2002-10-30 Kazu Hirata <kazu@cs.umass.edu>
523 * config/h8300/h8300.c (h8300_eightbit_constant_address_p):
524 Truncate the addresses for H8/300 using HImode.
526 2002-10-29 Hans-Peter Nilsson <hp@bitrange.com>
528 * toplev.c (rest_of_type_compilation): Return early in case of
530 (check_global_declarations): Don't call debug_hooks->global_decl
533 2002-10-28 Andreas Bauer <baueran@in.tum.de>
535 * doc/c-tree.texi (Tree overview): Fix typos.
537 2002-10-29 Phil Edwards <pme@gcc.gnu.org>
539 * Makefile.in (gnucompare*): Only record bad comparisons
540 if there really was a bad comparison.
542 Tue Oct 29 19:32:16 CET 2002 Jan Hubicka <jh@suse.cz>
544 * i386.h (CONST_DOUBLE_OK_FOR_LETTER_P): Remove 'H'
545 * i386.md (movsf*, movdf*): Use 'C' instead of 'H'
546 * md.texi (machine dependent constraints): Document 'C'
548 * simplify-rtx.c (simplify_subreg): Fix const_int->vector subregging.
550 * i386.c (ix86_expand_vector_move): Fix.
552 * i386.c (ix86_expand_builtin): Use sse2_maskmovdqu_rex64.
553 * i386.md (sse2_maskmovdqu_rex64): New pattern
556 * xmmintrin.h (_mm_stream_pi, _mm_stream_pd): Fix cast.
557 (ix86_init_mmx_sse_builtins): Fix type.
559 2002-10-29 Jason Thorpe <thorpej@wasabisystems.com>
561 * gthr-posix.h: Include <unistd.h> for feature tests.
562 (sched_get_priority_max, sched_get_priority_min)
563 (pthread_getschedparam, pthread_setschedparam): Only use
564 if _POSIX_THREAD_PRIORITY_SCHEDULING is defined.
565 (__gthread_objc_thread_set_priority): Don't treat all non-zero
566 returns from sched_get_priority_max and sched_get_priority_min
569 2002-10-29 Kazu Hirata <kazu@cs.umass.edu>
571 * config/h8300/h8300.h (TARGET_DEFAULT): Make it
574 2002-10-29 Kazu Hirata <kazu@cs.umass.edu>
576 * config/h8300/h8300.c (h8300_eightbit_constant_address_p): New.
577 (h8300_tiny_constant_address_p): Likewise.
578 * config/h8300/h8300.h (EIGHTBIT_CONSTANT_ADDRESS_P): Use
579 h8300_eightbit_constant_address_p.
580 (TINY_CONSTANT_ADDRESS_P): Use h8300_tiny_constant_address_p.
581 * config/h8300/h8300-protos.h: Add the prototypes for the two
584 2002-10-29 Kazu Hirata <kazu@cs.umass.edu>
586 * reload1.c (update_eliminables): Unconditionally check if
587 frame_pointer_needed has changed.
589 Tue Oct 29 15:37:39 CET 2002 Jan Hubicka <jh@suse.cz>
591 * toplev.c (rest_of_compilation): Reorganize way reg_scan is called
594 2002-10-29 Eric Botcazou <ebotcazou@libertysurf.fr>
596 * expr.c (expand_expr) [PLUS]: Don't use simplify_binary_operation;
597 check for zero operands explicitly.
599 2002-10-29 Richard Sandiford <rsandifo@redhat.com>
601 * config/mips/mips.md (extv, extzv, insv): Set size of referenced
602 memory after adjusting to BLKmode.
604 2002-10-29 Kazu Hirata <kazu@cs.umass.edu>
606 * config/h8300/h8300.h (MASK_*): New.
607 (TARGET_*): Use MASK_*.
609 2002-10-28 Jason Thorpe <thorpej@wasabisystems.com>
611 * config.gcc (*-*-netbsd*): Add NETBSD_ENABLE_PTHREADS to
612 tm_defines if pthreads are enabled.
613 * config/netbsd.h (LIB_SPEC): Only support the -pthread option
614 if NETBSD_ENABLE_PTHREADS is defined.
616 2002-10-28 Kazu Hirata <kazu@cs.umass.edu>
618 * ChangeLog.1: Fix typos.
619 * cse.c: Fix a comment typo.
620 * reload1.c: Likewise.
622 2002-10-27 Hans-Peter Nilsson <hp@bitrange.com>
624 * fixinc/inclhack.def (libc1_G_va_list): Correct test_text.
625 * fixinc/tests/base/_G_config.h: New file.
627 2002-10-27 Kazu Hirata <kazu@cs.umass.edu>
629 * combine.c: Fix comment formatting.
632 * regclass.c: Likewise.
633 * regmove.c: Likewise.
634 * regrename.c: Likewise.
635 * reg-stack.c: Likewise.
636 * reload1.c: Likewise.
637 * reload.c: Likewise.
638 * reload.h: Likewise.
639 * unroll.c: Likewise.
641 2002-10-27 Kazu Hirata <kazu@cs.umass.edu>
643 * reload1.c (reload): Fix a comment typo.
645 Sun Oct 27 10:15:24 CET 2002 Jan Hubicka <jh@suse.cz>
647 * linux64.h (DEFAULT_PCC_STRUCT_RETURN): Define.
649 2002-10-27 Zdenek Dvorak <rakdver@atrey.karlin.mff.cuni.cz>
651 * Makefile.in (dwarf2out.o): Add dependendcy on hashtab.h.
652 * dwarf2out.c: Include hashtab.h.
653 (is_main_source): New static variable.
654 (attr_checksum, die_checksum): Modified to handle die references.
655 (same_loc_p, same_dw_val_p, same_attr_p, same_die_p, same_die_p_wrap,
656 unmark_all_dies, htab_cu_hash, htab_cu_eq, htab_cu_del, check_duplicate_cu,
657 record_comdat_symbol_number): New static functions.
658 (output_comp_unit, compute_section_prefix, is_type_die, break_out_includes,
659 mark_dies, unmark_dies, dwarf2out_start_source_file): Modified.
660 * toplev.c (rest_of_decl_compilation): Call of dwarf2out_decl for type
663 2002-10-26 Kazu Hirata <kazu@cs.umass.edu>
665 * config/h8300/h8300.c (initial_offset): Change to
666 h8300_initial_elimination_offset.
667 * config/h8300/h8300.h (INITIAL_ELIMINATION_OFFSET): Use
668 h8300_initial_elimination_offset.
669 * config/h8300/h8300-protos.h: Update the prototype.
671 2002-10-26 Hans-Peter Nilsson <hp@bitrange.com>
673 * config/mmix/mmix.h (LIBCALL_VALUE): Use
674 MMIX_RETURN_VALUE_REGNUM, not MMIX_OUTGOING_RETURN_VALUE_REGNUM.
675 (FUNCTION_VALUE_REGNO_P): Similar, but move code to...
676 * config/mmix/mmix.c (mmix_function_value_regno_p): New.
677 * config/mmix/mmix-protos.h: Remove needless ifdefs on TREE_CODE
679 (mmix_function_value_regno_p): Declare.
681 * config/mmix/mmix.md ("fixuns_truncdfdi2"): Replace unsigned_fix,
682 invalid for floating point mode result, with fix.
684 2002-10-25 Mike Stump <mrs@apple.com>
686 Fixes gcc.dg/warn-1.c.
687 * c-typeck.c (warn_for_assignment): Don't print argument number, if zero.
689 Sat Oct 26 01:44:46 CEST 2002 Jan Hubicka <jh@suse.cz>
691 * toplev.c (dump_file_index): Add DFI_ce3.
692 (dump_file_info): Likewise.
693 (rest_of_compilation): Run first ifcvt pass before tracer.
695 2002-10-25 Steve Ellcey <sje@cup.hp.com>
697 * config/ia64/hpux.h (BITS_BIG_ENDIAN): Remove.
699 2002-10-25 Richard Henderson <rth@redhat.com>
701 * real.c (real_to_decimal): If the >1 tens reduction loop results
702 in a negative exponent, fall into the <1 pten computation.
704 2002-10-25 Zack Weinberg <zack@codesourcery.com>
707 * c-objc-common.c (inline_forbidden_p): Can not inline
708 functions containing structures or unions containing VLAs.
709 * tree-inline.c (walk_tree): For all class 't' nodes, walk
710 TYPE_SIZE and TYPE_SIZE_UNIT.
711 (copy_tree_r): Copy types if they are variably modified.
713 2002-10-25 Ulrich Weigand <uweigand@de.ibm.com>
715 * config/s390/s390.md: Remove old-style peepholes.
717 2002-10-25 Ulrich Weigand <uweigand@de.ibm.com>
719 * config/s390/s390.c (s390_decompose_address): Do not range check the
720 displacement if base or index is the argument pointer register.
722 2002-10-24 Hans-Peter Nilsson <hp@bitrange.com>
727 * fixinc/inclhack.def (libc1_G_va_list): New fix.
728 * fixinc/fixincl.x: Regenerate.
729 * config/i386/linux.h: Move MD_FALLBACK_FRAME_STATE_FOR inside
730 ifndef IN_LIBGCC2. Wrap it together with signal.h and
731 sys/ucontext.h inclusion in ifndef USE_GNULIBC_1.
732 * configure.in (gcc_AC_CHECK_DECLS): Check vasprintf too.
733 * config.in, configure: Regenerate.
735 2002-10-24 Igor Shevlyakov <igor@microunity.com>
737 * varasm.c (struct rtx_const): Array size 16 for V16QImode.
739 2002-10-24 Richard Henderson <rth@redhat.com>
741 * config/i386/i386.c (x86_output_mi_thunk): Fix x86_64 pic jump.
743 2002-10-24 Kazu Hirata <kazu@cs.umass.edu>
745 * config/h8300/h8300.c (initial_offset): Simplify by using
748 2002-10-24 Marek Michalkiewicz <marekm@amelek.gda.pl>
750 * doc/install.texi (avr): Update required binutils version.
752 2002-10-24 Theodore A. Roth <troth@openavr.org>
754 * doc/install.texi: Point avr users at more up-to-date information.
756 2002-10-24 Ulrich Weigand <uweigand@de.ibm.com>
758 * config/s390/s390.md (movdi, movsi, movhi, movqi): Add peepholes2
759 to pull operands out of the literal pool where possible.
761 2002-10-24 Denis Chertykov <denisc@overta.ru>
763 * config/avr/avr.c (init_cumulative_args): Test fntype for zero.
765 2002-10-24 Steve Ellcey <sje@cup.hp.com>
767 * expr.c (convert_move): If unsignedp is less then zero there
768 is no equivalent code.
770 2002-10-24 Zack Weinberg <zack@codesourcery.com>
772 * tree.def: Delete mention of nonexistent ARRAY_TYPE fields.
774 2002-10-24 Ulrich Weigand <uweigand@de.ibm.com>
776 * config/s390/s390.h: Rework comments; re-sort target macro definitions
777 according to the sequence they are defined in the manual.
778 (POINTER_BOUNDARY): Remove.
780 2002-10-24 Kazu Hirata <kazu@cs.umass.edu>
782 * config/h8300/h8300.c (round_frame_size): Replace 8 with
785 2002-10-24 Kazu Hirata <kazu@cs.umass.edu>
787 * config/h8300/h8300.h (EIGHTBIT_CONSTANT_ADDRESS_P): Make it
789 (TINY_CONSTANT_ADDRESS_P): Likewise.
791 2002-10-24 Richard Henderson <rth@redhat.com>
793 * config/ia64/ia64.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): True.
794 (ia64_output_mi_thunk): Rewrite to use rtl, and to handle the
797 2002-10-24 Richard Henderson <rth@redhat.com>
800 * reload.c (find_reloads_toplev): Mode of X is not important
801 when simplifying subregs of constants.
803 2002-10-24 Richard Sandiford <rsandifo@redhat.com>
805 * config.gcc (mips64vr-*-elf*, mips64vrel-*-elf*): Add
806 MIPS_MARCH_CONTROLS_SOFT_FLOAT=1 to $tm_defines.
807 * config/mips/mips.c (MIPS_MARCH_CONTROLS_SOFT_FLOAT): Default to 0.
808 (override_options): Base default setting of MASK_SOFT_FLOAT on -march
809 if MIPS_MARCH_CONTROLS_SOFT_FLOAT.
811 2002-10-24 Richard Sandiford <rsandifo@redhat.com>
813 * optabs.c (expand_binop): Don't reuse the shift target in the
814 middle of shift sequences.
816 2002-10-23 Ziemowit Laski <zlaski@apple.com>
818 * objc/objc-act.c (get_static_reference): Remove unneeded
819 TYPE_BINFO initialization.
820 (get_object-reference): Likewise.
821 (build_constructor): Tighten precondition check.
822 (finish_message_expr): Likewise.
824 2002-10-23 Jakub Jelinek <jakub@redhat.com>
826 * config/i386/i386.c (local_symbolic_operand): Move LABEL_REF test
829 2002-10-23 Steve Ellcey <sje@cup.hp.com>
831 * config/ia64/ia64.c (hfa_element_mode): Don't allow 128 bit floats
834 2002-10-23 Richard Henderson <rth@redhat.com>
836 * config/alpha/alpha.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): True.
837 (alpha_output_mi_thunk_osf): Handle vcall_offset.
839 2002-10-23 Zack Weinberg <zack@codesourcery.com>
841 * langhooks.h (struct lang_hooks_for_tree_inlining): Add
843 * langhooks-def.h: Default for tree_inlining.var_mod_type_p is
844 hook_tree_bool_false.
846 * tree.c (variably_modified_type_p): Moved here from
847 cp/tree.c. Use lang_hooks.tree_inlining.var_mod_type_p for
848 language-specific cases. Due to this, must weaken some 'if
849 and only if' checks to merely 'if'.
850 * tree.h: Prototype variably_modified_type_p.
852 * tree-inline.c (walk_tree): #undef WALK_SUBTREE_TAIL at end.
854 2002-10-23 Ulrich Weigand <uweigand@de.ibm.com>
856 * config/s390/linux.h (CC1_SPEC, CC1PLUS_SPEC): Remove.
857 * config/s390/s390.c (optimization_options): Disable -fcaller-saves.
859 * config/s390/s390-protos.h (fp_operand): Remove.
860 * config/s390/s390.c (fp_operand): Remove.
861 * config/s390/s390.md ("movdi"): Replace fp_operand by FP_REG_P.
862 ("*movdi_lhi", "*movdi_lli", "*movdi_larl"): Likewise.
863 ("movsi", "*movsi_lhi", "*movsi_lli"): Likewise.
864 (movdi_31, movdf_31 splitters): Likewise.
866 * config/s390/s390.h (IEEE_FLOAT): Remove.
867 (TARGET_FLOAT_FORMAT): Define in terms of TARGET_IEEE_FLOAT.
868 (INT_REGNO_P): Rename to ...
869 (GENERAL_REGNO_P): ... this.
870 (FLOAT_REGNO_P): Rename to ...
871 (FP_REGNO_P): ... this.
872 (ADDR_REGNO_P): New macro.
873 (GENERAL_REG_P, ADDR_REG_P, FP_REG_P, CC_REG_P): New macros.
874 (REGNO_OK_FOR_DATA_P, REGNO_OK_FOR_FP_P): Remove.
875 (DATA_REG_P, FP_REG_P, ADDRESS_REG_P): Likewise.
876 (HARD_REGNO_NREGS): Adapt to macro renaming.
877 (HARD_REGNO_MODE_OK): Likewise.
879 2002-10-23 David Edelsohn <edelsohn@gnu.org>
880 Geoff Keating <geoffk@apple.com>
882 * config/rs6000/rs6000.c (rs6000_register_move_cost): New function.
883 (rs6000_memory_move_cost): New function.
884 * config/rs6000/rs6000-protos.h: Declare them.
885 * config/rs6000/rs6000.h: Use them.
887 2002-10-23 Ulrich Weigand <uweigand@de.ibm.com>
889 * libgcc2.c (__udiv_w_sdiv): Use attribute ((always_inline)) when
890 inlining it into other libgcc2 routines.
891 (__udivmoddi4): Likewise.
893 2002-10-22 Nathanael Nerode <neroden@gcc.gnu.org>
895 * doc/sourcebuild.texi (Test Suites): Improve.
897 2002-10-22 Stan Shebs <shebs@apple.com>
899 * config/rs6000/rs6000.c (rs6000_output_mi_thunk): Add missing
902 2002-10-22 Jim Wilson <wilson@redhat.com>
904 * config/i386/i386.md (subdi3_1): Add call to ix86_binary_operator_ok.
906 Wed Oct 23 01:52:36 CEST 2002 Jan Hubicka <jh@suse.cz>
909 * xmmintrin.h: Add const to the argument of loads.
911 * i386.md (pushv2di): New pattern.
913 * xmmintrin.h (_MM_TRANSPOSE4_PS): New.
915 2002-10-22 Richard Henderson <rth@redhat.com>
917 * target.h (gcc_target.asm_out): Merge output_mi_thunk and
918 output_mi_vcall_thunk into a single hook. Add can_output_mi_thunk.
919 * target-def.h (TARGET_ASM_OUTPUT_MI_THUNK): Don't conditionalize.
920 (TARGET_ASM_OUTPUT_MI_VCALL_THUNK): Remove.
921 (TARGET_ASM_CAN_OUTPUT_MI_THUNK): New.
922 (TARGET_ASM_OUT): Update.
923 * hooks.c (hook_bool_tree_hwi_hwi_tree_false): New.
924 (hook_bool_tree_hwi_hwi_tree_true): New.
925 (default_can_output_mi_thunk_no_vcall): New.
926 * hooks.h: Declare them.
927 * system.h (ASM_OUTPUT_MI_THUNK): Poison.
929 * config/alpha/alpha.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): New.
930 (alpha_output_mi_thunk_osf): Add VCALL_OFFSET parameter.
931 * config/arm/arm.c, config/cris/cris.c, config/frv/frv.c,
932 config/i960/i960.c, config/ia64/ia64.c, config/m68k/m68k.c,
933 config/mmix/mmix.c, config/pa/pa.c, config/sparc/sparc.c,
934 config/stormy16/stormy16.c: Similarly.
936 * config/i386/i386.c (x86_output_mi_thunk): Merge vcall_offset code.
937 Handle 64-bit properly. Streamline.
938 (x86_output_mi_vcall_thunk): Remove.
939 (x86_this_parameter): Rename from ia32_this_parameter; handle 64-bit.
940 (x86_can_output_mi_thunk): New.
941 (TARGET_ASM_OUTPUT_MI_VCALL_THUNK): Remove.
942 (TARGET_ASM_CAN_OUTPUT_MI_THUNK): New.
943 (override_options): Don't zap targetm.asm_out.output_mi_vcall_thunk.
945 * config/rs6000/rs6000.c (rs6000_output_mi_thunk): Rename from
946 output_mi_thunk; make static; always use function_section.
947 (TARGET_ASM_OUTPUT_MI_THUNK): New.
948 (TARGET_ASM_CAN_OUTPUT_MI_THUNK): New.
949 (rs6000_ra_ever_killed): Test no_new_pseudos not
950 targetm.asm_out.output_mi_thunk in conjunction with thunks.
951 * config/rs6000/rs6000-protos.h: Update.
952 * config/rs6000/sysv4.h (TARGET_ASM_OUTPUT_MI_THUNK): Remove.
953 * config/rs6000/xcoff.h (ASM_DECLARE_FUNCTION_NAME): Don't call
954 xcoffout_declare_function when using rs6000_output_mi_thunk.
956 * config/s390/s390.c (s390_output_mi_thunk): Rename from
957 s390_output_mi_vcall_thunk.
958 (TARGET_ASM_OUTPUT_MI_THUNK): Remove.
959 (TARGET_ASM_CAN_OUTPUT_MI_THUNK): New.
961 * config/vax/vax.c (vax_output_mi_thunk): Static; add vcall_offset.
962 (TARGET_ASM_OUTPUT_MI_THUNK, TARGET_ASM_CAN_OUTPUT_MI_THUNK): New.
963 * config/vax/vax-protos.h: Update.
964 * config/vax/vax.h (ASM_OUTPUT_MI_THUNK): Remove.
966 Wed Oct 23 00:33:11 CEST 2002 Jan Hubicka <jh@suse,cz>
968 * i386.c (standard_sse_constant_p): Accept vector and integer zeros too.
969 * i386.h (EXTRA_CONSTRAINT): Recognize 'C'
970 * i386.md (movti_internal): Use 'C'
972 * xmmintrin.h (_mm_cmplt_epi*): New.
974 2002-10-22 Ulrich Weigand <uweigand@de.ibm.com>
976 * config/s390/s390.md ("*movdi_64"): Fix op_type attribute.
977 ("*movdf_64"): Likewise.
978 ("*lshrdi3_64"): Likewise.
979 ("blockage"): Add length attribute.
982 Tue Oct 22 23:51:34 CEST 2002 Jan Hubicka <jh@suse.cz>
985 (sse2_cvtsi2sd, sse2_pslrdq): Fix template.
986 (sse2_umulv2siv2di3): Fix predicate.
987 (sse2_psadbw, ashrv8hi3, ashrv4si3, lshrv8hi3 lshrv4si3,
988 lshrv2di3, ashlv8hi3, ashlv4si3, ashlv2di3): Likewise.
989 * xmmintrin.h (_mm_mul_epu16): Rename to...
990 (_mm_mul_epu32): This one.
991 (_mm_cvtsi32_si128, _mm_cvtsi128_si32): New.
993 (contains_128bit_aligned_vector_p): Undo accidental checkin.
995 2002-10-22 Eric Christopher <echristo@redhat.com>
997 * config/sparc/sparc.h: Add #error.
999 2002-10-22 Ulrich Weigand <uweigand@de.ibm.com>
1001 * config.gcc [s390-*-linux]: Remove s390/t-linux from tmake_file.
1002 [s390x-*-linux*]: Likewise.
1003 * config/s390/t-linux: Remove.
1004 * config/s390/s390.h: Include fixdfdi.h when building libgcc2.
1006 Tue Oct 22 19:07:03 CEST 2002 Jan Hubicka <jh@suse.cz>
1008 * i386.c (builtin_description): Add IX86_BUILTIN_PUNPCKHQDQ128.
1009 (ix86_expand_builtin): Fix MASKMOVDQU expasion.
1010 * i386.h (ix86_builtins): Add IX86_BUILTIN_PUNPCKHQDQ128.
1011 * i386.md (mmx_punpck?dq): Simplify.
1012 (sse2_pubpcklqdq): Fix.
1013 (sse2_pubpckhqdq): New.
1014 * xmmintrin.h (_mm_unpackhi_epi32): New.
1016 * xmmintrin.h (_mm_cvt*, _mm_stream_pd): Fix prototypes.
1017 (_mm_shufflehi_epi16, _mm_shufflelo_epi16): Fix typo.
1019 2002-10-22 Nathan Sidwell <nathan@codesourcery.com>
1022 * fold_const.c (fold_binary_op_with_conditional_arg): Always
1023 build compound_expr if we used save_expr.
1025 2002-10-22 Alan Modra <amodra@bigpond.net.au>
1027 * output.h (SECTION_NOTYPE): Define.
1028 * varasm.c (default_section_type_flags_1): Set SECTION_NOTYPE for
1029 init array sections.
1030 (default_elf_asm_named_section): Mind SECTION_NOTYPE.
1031 * config/arm/arm.c (arm_elf_asm_named_section): Likewise. Also
1034 2002-10-21 Richard Henderson <rth@redhat.com>
1036 * real.c (sticky_rshift_significand): Return inexact, don't
1037 or it in immediately.
1038 (sub_significands): Accept incomming carry.
1039 (div_significands, rtd_divmod): Update for sub_significands change.
1040 (round_for_format): Update for sticky_rshift_significand change.
1041 (do_add): Don't involve the inexact bit in addition, do give the
1042 inexact bit as the subtraction carry-in.
1043 (encode_internal, decode_internal, real_internal_format): New.
1044 * real.h (real_internal_format): Declare.
1046 2002-10-21 Ulrich Weigand <uweigand@de.ibm.com>
1048 * libgcc2.c: Fix __udiv_w_sdiv breakage on platforms that
1049 don't define sdiv_qrnnd.
1051 2002-10-21 Kazu Hirata <kazu@cs.umass.edu>
1053 * config/h8300/h8300.h (EIGHTBIT_CONSTANT_ADDRESS_P): Simplify
1055 (TINY_CONSTANT_ADDRESS_P): Likewise.
1057 Tue Oct 22 00:04:20 CEST 2002 Jan Hubicka <jh@suse.cz>
1059 * i386.c (builtin_description): Add punpcklqdq and movdq2q
1060 (ix86_init_mmx_sse_builtins): Add v2di_ftype_void, di_ftype_v2di,
1061 v16qi_ftype_pchar, void_ftype_pchar_v16qi, v4si_ftype_pchar,
1062 void_ftype_pchar_v4si; Initialize __builtin_ia32_movdq2q,
1063 __builtin_ia32_loaddqa, __builtin_ia32_loaddqu, __builtin_ia32_loadd
1064 __builtin_ia32_storedqa, __builtin_ia32_storedqu, __builtin_ia32_stored
1065 __builtin_ia32_setzero128.
1066 (ix86_expand_builtin): Handle IX86_BUILTIN_CLRTI, IX86_BUILTIN_LOADDQA,
1067 IX86_BUILTIN_LOADDQU, IX86_BUILTIN_LOADD, IX86_BUILTIN_STOREDQA,
1068 IX86_BUILTIN_STOREDQU, IX86_BUILTIN_STORED, Ix86_BUILTIN_MOVQ.
1069 * i386.h (ix86_builtins): Add IX86_BUILTIN_LOADDQA, IX86_BUILTIN_LOADDQU,
1070 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_STOREDQU, IX86_BUILTIN_LOADD,
1071 IX86_BUILTIN_STORED, IX86_BUILTIN_CLRTI, IX86_BUILTIN_MOVDQ2Q,
1072 IX86_BUILTIN_PUNPCKLQDQ128, Ix86_BUILTIN_MOVQ.
1073 * i386.md (sse2_punpcklqdq, sse2_movqsse2_loadd, sse2_stored,
1074 sse2_movq): New patterns.
1075 (sse2_movdqa, sse2_movdqu, sse2_movdq2q): Fix.
1076 * xmmintrin.h (_mm_load_si128, _mm_loadu_si128, _mm_loadl_epi64,
1077 _mm_store_si128, _mm_storeu_si128, _mm_storel_epi64,
1078 _mm_setzero_si128, _mm_set_epi64, _mm_set_epi32, _mm_set_epi16,
1079 _mm_set_epi8, _mm_set1_epi64, _mm_set1_epi32, _mm_set1_epi16,
1080 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
1081 _mm_setr_epi8, _mm_unpacklo_epi64,_mm_set_moveq): New functions.
1082 (_mm_insert_epi16): Fix.
1084 2002-10-21 Dale Johannesen <dalej@apple.com>
1086 * config/rs6000/rs6000.c (rs6000_reverse_condition): Handle
1087 unsafe math reversals correctly for RTL generation.
1088 (output_cbranch): Replace rs6000_reverse_condition call
1089 by its former definition.
1091 2002-10-21 Jakub Jelinek <jakub@redhat.com>
1093 * config/i386/i386.c (x86_64_sign_extended_value): Add allow_rip
1094 argument. In CM_SMALL_PIC model consider SYMBOL_REFs binding locally or
1095 from constant pool or LABEL_REFs as sign extended if allow_rip.
1096 Change all +-1GB limits to +-16MB.
1097 (x86_64_general_operand, x86_64_szext_general_operand,
1098 x86_64_nonmemory_operand, x86_64_movabs_operand,
1099 x86_64_szext_nonmemory_operand, x86_64_immediate_operand,
1100 legitimate_address_p, ix86_expand_int_movcc): Update callers.
1101 (local_symbolic_operand): Don't allow offsets bigger than +-16MB
1102 in CM_SMALL_PIC model.
1103 (legitimate_pic_address_disp_p): Don't check offsets before
1104 calling local_symbolic_operand.
1105 (legitimize_pic_address): Force offsets bigger than +-16MB into
1107 * config/i386/i386.h (EXTRA_CONSTRAINT, CONST_COSTS): Likewise.
1108 * config/i386/i386-protos.h (x86_64_sign_extended_value): Update
1111 * configure.in: Test for @GOTNTPOFF and @INDNTPOFF on IA-32 too.
1112 Add x86-64 test. Set tls_first_minor to 14 on IA-32 and x86-64.
1113 * configure: Rebuilt.
1114 * config/i386/i386.c (x86_64_sign_extended_value): Don't allow TLS
1115 SYMBOL_REFs unless enclosed in UNSPEC. Handle UNSPEC_DTPOFF,
1116 UNSPEC_GOTNTPOFF and UNSPEC_NTPOFF.
1117 (legitimate_address_p): Allow foo@dtpoff(base) even on TARGET_64BIT
1119 (ix86_encode_section_info): Don't ever generate TLSGD or TLSLD for
1120 non-pic code if TARGET_64BIT.
1121 (legitimize_address): Generate 64-bit TLS sequences.
1122 (output_pic_addr_const): Support x86-64 TLS operators.
1123 (i386_output_dwarf_dtprel): Output 64-bit DTPOFF as .long f@DTPOFF, 0.
1124 (print_operand_address): Use %fs instead of %gs on TARGET_64BIT.
1125 Don't append (%rip) in 64-bit TLSGD and TLSLD sequences.
1126 (output_addr_const_extra): Support x86-64 TLS operators.
1127 (maybe_get_pool_constant): Handle TARGET_64BIT -fpic.
1128 (ix86_tls_get_addr): Use __tls_get_addr on TARGET_64BIT
1130 * config/i386/i386.md (*tls_global_dynamic_gnu): Renamed to...
1131 (*tls_global_dynamic_32_gnu): ..., add !TARGET_64BIT.
1132 (*tls_global_dynamic_sun): Renamed to...
1133 (*tls_global_dynamic_32_sun): ..., add !TARGET_64BIT.
1134 (tls_global_dynamic): Renamed to...
1135 (tls_global_dynamic_32): ... this.
1136 (tls_global_dynamic_64, *tls_global_dynamic_64): New.
1137 (*tls_local_dynamic_base_dynamic_gnu): Renamed to...
1138 (*tls_local_dynamic_base_dynamic_32_gnu): ..., add !TARGET_64BIT.
1139 (*tls_local_dynamic_base_dynamic_sun): Renamed to...
1140 (*tls_local_dynamic_base_dynamic_32_sun): ..., add !TARGET_64BIT.
1141 (tls_local_dynamic_base_dynamic): Renamed to...
1142 (tls_local_dynamic_base_dynamic_32): ... this.
1143 (tls_local_dynamic_base_dynamic_64,
1144 *tls_local_dynamic_base_dynamic_64): New.
1145 (*tls_local_dynamic_once): Renamed to...
1146 (*tls_local_dynamic_32_once): ... this.
1148 2002-10-21 Ulrich Weigand <uweigand@de.ibm.com>
1150 * libgcc2.c: Inline __udiv_w_sdiv when compiling __udivdi3,
1151 __divdi3, __umoddi3, or __moddi3.
1153 2002-10-21 Ulrich Weigand <uweigand@de.ibm.com>
1155 * c-opts.c (missing_arg): Use cl_options[opt_index].opt_code
1156 instead of just opt_index as switch expression.
1158 * calls.c (store_one_arg): Change type of 'excess_align'
1161 * profile.c (output_gcov_string): Change type of 'temp'
1164 2002-10-21 Ulrich Weigand <uweigand@de.ibm.com>
1166 * config/s390/fixdfdi.h (__fixunsdfdi, __fixdfdi): Add prototypes.
1167 (__fixunssfdi, __fixsfdi): Likewise.
1168 * config/s390/s390.c (s390_single_hi): Initialize 'value'.
1169 (s390_single_qi): Likewise.
1170 (s390_emit_epilogue): Initialize 'offset'. Remove signed vs.
1171 unsigned comparison warning.
1172 (s390_return_addr_rtx): New function.
1173 * config/s390/s390-protos.h (s390_return_addr_rtx): Declare it.
1174 * config/s390/s390.h (RETURN_ADDR_RTX): Use it.
1175 (HARD_REGNO_MODE_OK): Rewrite condition to silence warnings.
1177 2002-10-21 Ulrich Weigand <uweigand@de.ibm.com>
1179 * config/s390/s390.c (s390_output_mi_vcall_thunk): New function.
1180 (TARGET_ASM_OUTPUT_MI_VCALL_THUNK): Define target hook.
1181 (s390_output_mi_thunk): Remove.
1182 (TARGET_ASM_OUTPUT_MI_THUNK): Remove.
1184 2002-10-21 Kazu Hirata <kazu@cs.umass.edu>
1186 * config/h8300/h8300.h (N_REG_CLASSES): Parenthesize.
1188 2002-10-20 Richard Henderson <rth@redhat.com>
1190 * target.h (struct gcc_target): Line wrap.
1192 * config/alpha/alpha.c (alpha_output_mi_thunk_osf): Static.
1193 (TARGET_ASM_OUTPUT_MI_THUNK): Define here...
1194 * config/alpha/alpha.h: ... not here.
1195 * config/alpha/alpha-protos.h: Update.
1197 * config/arm/arm.c, config/arm/arm.h, config/arm/arm-protos.h
1198 config/cris/cris-protos.h, config/cris/cris.c, config/cris/cris.h,
1199 config/frv/frv-protos.h, config/frv/frv.c, config/frv/frv.h,
1200 config/i386/i386-protos.h, config/i386/i386.c, config/i386/openbsd.h,
1201 config/i386/unix.h, config/i960/i960-protos.h, config/i960/i960.c,
1202 config/i960/i960.h, config/ia64/ia64-protos.h, config/ia64/ia64.c,
1203 config/ia64/ia64.h, config/m68k/linux.h, config/m68k/m68k-protos.h,
1204 config/m68k/m68k.c, config/m68k/netbsd-elf.h, config/m68k/openbsd.h,
1205 config/mmix/mmix-protos.h, config/mmix/mmix.c, config/mmix/mmix.h,
1206 config/pa/pa-protos.h, config/pa/pa.c, config/pa/pa.h,
1207 config/s390/s390-protos.h, config/s390/s390.c, config/s390/s390.h,
1208 config/sparc/openbsd.h, config/sparc/sparc-protos.h,
1209 config/sparc/sparc.c, config/sparc/sparc.h,
1210 config/stormy16/stormy16-protos.h, config/stormy16/stormy16.c,
1211 config/stormy16/stormy16.h: Similarly.
1213 * config/m68k/m68k.c (m68k_output_mi_thunk): Replicate mnemonic
1214 selection logic from call patterns.
1216 2002-10-20 Mark Mitchell <mark@codesourcery.com>
1218 * config/m68k/m68k.c (m68k_output_mi_thunk): Fix typo.
1220 2002-10-20 Zdenek Dvorak <rakdver@atrey.karlin.mff.cuni.cz>
1223 * i386.c (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Define and
1224 expand __builtin_ia32_pslldqi128 and __builtin_ia32_psrldqi128.
1225 * i386.h (IX86_BUILTIN_PSLLDQI128, IX86_BUILTIN_PSRLDQI128): New.
1226 * xmmintrin.h (_mm_srli_si128, _mm_slli_si128): New.
1228 2002-10-20 Roger Sayle <roger@eyesopen.com>
1231 * toplev.c (flag_unsafe_profile_arcs): Remove.
1232 (flag_bounded_pointers): Remove.
1233 (flag_bounds_check): Correct comments.
1234 (lang_independent_options): Remove -funsafe-profile-arcs and
1235 -fbounded-pointers. Correct -fbounds-check comments.
1237 * flags.h: Correct flag_schedule_interblock comments.
1238 (flag_bounded_pointers): Remove prototype.
1239 (flag_bounds_check): Correct comments.
1241 * c-opts.c (c_common_init_options): No need to mark
1242 flag_bounds_check as unspecified.
1243 (c_common_post_options): And no need to set it from
1244 flag_bounded_pointers if its still unspecified.
1246 * doc/invoke.texi: Fix some overfull hboxes in "make dvi".
1247 Document --version, -feliminate-dwarf-2-dups, -fno-sched-interblock,
1248 -fno-sched-spec, -fsched-spec-load, -fsched-spec-load-dangerous,
1249 -fsched-verbose=n, -fno-branch-count-reg and -fbounds-check.
1251 2002-10-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1252 Mark Mitchell <mark@codesourcery.com>
1254 * alpha-protos.h (alpha_output_mi_thunk_osf): Update signature to
1256 * arm-protos.h, arm.c (arm_output_mi_thunk): Likewise.
1257 * cris-protos.h, cris.c (cris_asm_output_mi_thunk): Likewise.
1258 * frv-protos.h, frv.c (frv_asm_output_mi_thunk): Likewise.
1259 * i386-protos.h, i386.c (x86_output_mi_vcall_thunk,
1260 x86_output_mi_thunk): Likewise.
1261 * i960-protos.h, i960.c (i960_output_mi_thunk): Likewise.
1262 * ia64-protos.h, ia64.c (ia64_output_mi_thunk): Likewise.
1263 * m68k-protos.h, m68k.c (m68k_output_mi_thunk): Likewise.
1264 * mmix-protos.h, mmix.c (mmix_asm_output_mi_thunk): Likewise.
1265 * rs6000-protos.h, rs6000.c (output_mi_thunk): Likewise.
1266 * s390-protos.h, s390.c (s390_output_mi_thunk): Likewise.
1267 * stormy16-protos.h, stormy16.c (xstormy16_asm_output_mi_thunk):
1269 * vax-protos.h, vax.c (vax_output_mi_thunk): Likewise.
1271 * target.h (gcc_target): Update output_mi_thunk and
1272 output_mi_vcall_thunk to take a HOST_WIDE_INT delta and
1275 * config/alpha/alpha.c: Replace ASM_OUTPUT_MI_THUNK with
1276 TARGET_ASM_OUTPUT_MI_THUNK in comments.
1277 * config/alpha/vms.h (ASM_OUTPUT_MI_THUNK): Don't #undef it.
1278 (TARGET_ASM_OUTPUT_MI_THUNK): #undef it.
1279 * config/frv/frv.h (DEFAULT_VTABLE_THUNKS): Remove definition.
1280 * config/i386/i386-protos.h (x86_output_mi_vcall_thunk): Update
1282 * config/i386/i386.c (x86_output_mi_vcall_thunk): Likewise.
1283 * config/i386/openbsd.h: Replace ASM_OUTPUT_MI_THUNK with
1284 TARGET_ASM_OUTPUT_MI_THUNK in comments.
1285 * config/i960/i960.h (ASM_OUTPUT_MI_THUNK): Don't define.
1286 (TARGET_ASM_OUTPUT_MI_THUNK): Do define.
1287 * config/m68k/openbsd.h: Replace ASM_OUTPUT_MI_THUNK with
1288 TARGET_ASM_OUTPUT_MI_THUNK in comments.
1289 * config/rs6000/rs6000.c (rs6000_ra_ever_killed): Remove #ifdef
1290 ASM_OUTPUT_MI_THUNK and replace with check of targetm.
1292 * doc/tm.texi (TARGET_ASM_OUTPUT_MI_THUNK): Update signature.
1293 (TARGET_ASM_OUTPU_MI_VCALL_THUNK): Likewise.
1295 2002-10-19 Brad Lucier <lucier@math.purdue.edu>
1297 * real.c (do_add): Fix 0+0 sign corner case.
1298 (do_divide): Fix Inf/0 corner case.
1300 Sun Oct 20 00:31:31 CEST 2002 Jan Hubicka <jh@suse.cz>
1302 * i386.c (classify_argument): Pass MMX arguments in memory
1303 (ix86_expand_builtin): Expand proper address mode for cflush.
1304 * i386.md (movdqa): Fix typo.
1305 (sse2_cflush): Accept DImode addresses.
1307 * xmmintrin.h (_mm_sqrt_sd): Accept two arguments.
1308 (_mm_max_sd): Fix pasto.
1309 (_mm_storeh_pd, _mm_storel_pd): Fix.
1311 * i386.c (bdesc_comi): Fix to match specification.
1312 (ix86_expand_sse_comi): Emit the comparison properly.
1313 * i386.md (sse_comi, sse2_comi, sse_ucomi, sse2_ucomi):
1314 Do not use comparison operator.
1315 (vnmaskcmp): Fix template.
1317 * xmmintrin.h (_mm_cvtps_pi16): Fix.
1319 2002-10-19 Sebastian Pop <s.pop@laposte.net>
1321 * dependence.c : Removed.
1322 * Makefile.in : Remove dependence.o.
1324 Sat Oct 19 10:46:52 CEST 2002 Jan Hubicka <jh@suse.cz>
1326 * mmintrin.h (__m64): typedef it to v2si.
1327 (_mm_cvtsi32_si64, _mm_cvtsi32_si64_mm_sll_pi16,
1328 _mm_sll_pi32, _mm_sll_pi64, _mm_slli_pi64, _mm_sra_pi16,
1329 _mm_sra_pi32, _mm_srl_pi16, _mm_srl_pi32, _mm_srl_pi64,
1330 _mm_srli_pi64, _mm_and_si64, _mm_andnot_si64,
1331 _mm_or_si64, _mm_xor_si64): Add neccesary casts.
1332 * xmmintrin.h (_mm_setzero_si64): Likewise.
1334 * i386.h (ALIGN_MODE_128): Update comment; add missing modes
1335 (SSE_REG_MODE_P, MMX_REG_MODE_P): New macros.
1338 Patch by Shawn Wagner
1339 * mmintrin.h: Replace pi64 by si64.
1341 2002-10-18 David Edelsohn <edelsohn@gnu.org>
1343 * rs6000.md (movdf_hardfloat32): Order alternatives consistently.
1344 Use length of 4 not *.
1345 (movdf_hardfloat64): Same. Support DFmode moves to/from CTR/LR.
1346 (movdf_softfloat64): Likewise.
1347 (movdi_internal32): Use length of 4 not *.
1348 (movti_power): Same.
1349 (ctrsi, ctrdi): Same.
1351 2002-10-18 Zack Weinberg <zack@codesourcery.com>
1353 * c-decl.c (start_decl): Point users of the old initialized-
1354 typedef extension at __typeof__.
1356 2002-10-18 Richard Henderson <rth@redhat.com>
1358 * real.c (cmp_significand_0, rtd_divmod, ten_to_mptwo): New.
1359 (real_to_decimal): Re-implement using the logic from the
1360 gcc 3.2 etoasc. Comment heavily.
1361 (div_significands): Simplify loop startup and comparison logic.
1363 2002-10-18 Mark Mitchell <mark@codesourcery.com>
1365 * target-def.h (TARGET_ASM_OUTPUT_MI_THUNK): Default to NULL.
1366 (TARGET_ASM_OUTPUT_MI_VCALL_THUNK): Likewise.
1367 (TARGET_ASM_OUT): Add them.
1368 * target.h (asm_out): Add output_mi_thunk and
1369 output_mi_vcall_thunk.
1370 * config/alpha/alpha.h (ASM_OUTPUT_MI_THUNK): Rename to ...
1371 (TARGET_ASM_OUTPUT_MI_THUNK): ... this.
1372 * config/arm/arm-protos.h (arm_output_mi_thunk): Declare.
1373 * config/arm/arm.c (arm_output_mi_thunk): Define.
1374 * config/arm/arm.h (ASM_OUTPUT_MI_THUNK): Rename to ...
1375 (TARGET_ASM_OUTPUT_MI_THUNK): ... this.
1376 * config/cris/cris.h (ASM_OUTPUT_MI_THUNK): Rename to ...
1377 (TARGET_ASM_OUTPUT_MI_THUNK): ... this.
1378 * config/frv/frv.h (ASM_OUTPUT_MI_THUNK): Rename to ...
1379 (TARGET_ASM_OUTPUT_MI_THUNK): ... this.
1380 * config/i386/i386-protos.h (x86_output_mi_thunk): Adjust
1382 (x86_output_mi_vcall_thunk): Declare.
1383 * config/i386/i386.c (override_options): Clear
1384 output_mi_vcall_thunk in 64-bit mode.
1385 (ix86_fntype_regparm): New function.
1386 (ix86_return_pops_args): Use it.
1387 (ia32_this_parameter): New function.
1388 (x86_output_mi_vcall_thunk): New function.
1389 (x86_output_mi_thunk): Use it
1390 * config/i386/unix.h (TARGET_ASM_OUTPUT_MI_THUNK): Adjust.
1391 (TARGET_ASM_OUTPUT_MI_VCALL_THUNK): Define.
1392 * config/i960/i960-protos.h (i960_output_mi_thunk): Declare.
1393 * config/i960/i960.c (i960_output_mi_thunk): New function.
1394 * config/i960/i960.h (ASM_OUTPUT_MI_THUNK): Adjust.
1395 * config/ia64/ia64-protos.h (ia64_output_mi_thunk): Declare.
1396 * config/ia64/ia64.c (ia64_output_mi_thunk): Define.
1397 * config/ia64/ia64.h (ASM_OUTPUT_MI_THUNK): Rename to ...
1398 (TARGET_ASM_OUTPUT_MI_THUNK): ... this.
1399 * config/m68k/m68k-protos.h (m68k_output_mi_thunk): New function.
1400 * config/m68k/linux.h (ASM_OUTPUT_MI_THUNK): Rename to ...
1401 (TARGET_ASM_OUTPUT_MI_THUNK): ... this.
1402 * config/m68k/netbsd-elf.h (ASM_OUTPUT_MI_THUNK): Rename to ...
1403 (TARGET_ASM_OUTPUT_MI_THUNK): ... this.
1404 * config/mmix/mmix.h (ASM_OUTPUT_MI_THUNK): Rename to ...
1405 (TARGET_ASM_OUTPUT_MI_THUNK): ... this.
1406 * config/pa/pa.h (ASM_OUTPUT_MI_THUNK): Rename to ...
1407 (TARGET_ASM_OUTPUT_MI_THUNK): ... this.
1408 * config/rs6000/sysv4.h (ASM_OUTPUT_MI_THUNK): Rename to ...
1409 (TARGET_ASM_OUTPUT_MI_THUNK): ... this.
1410 * config/s390/s390-protos.h (s390_output_mi_thunk): Declare.
1411 * config/s390/s390.c (s390_output_mi_thunk): Define.
1412 * config/s390/s390.h (ASM_OUTPUT_MI_THUNK): Rename to ...
1413 (TARGET_ASM_OUTPUT_MI_THUNK): ... this.
1414 * config/sparc/sparc.h (ASM_OUTPUT_MI_THUNK): Rename to ...
1415 (TARGET_ASM_OUTPUT_MI_THUNK): ... this.
1416 * config/stormy16/stormy16.h (ASM_OUTPUT_MI_THUNK): Rename to ...
1417 (TARGET_ASM_OUTPUT_MI_THUNK): ... this.
1418 * config/vax/vax-protos.h (vax_output_mi_thunk): Declare.
1419 * config/vax/vax.c (vax_output_mi_thunk): Define.
1420 * config/vax/vax.h (ASM_OUTPUT_MI_THUNK): Rename to ...
1421 (TARGET_ASM_OUTPUT_MI_THUNK): ... this.
1422 * doc/tm.texi: Adjust documentation.
1424 2002-10-18 Jason Thorpe <thorpej@wasabisystems.com>
1426 * config/netbsd.h (NETBSD_ENABLE_EXECUTE_STACK): Define
1427 __enable_execute_stack function.
1428 * config/alpha/netbsd.h (TRANSFER_FROM_TRAMPOLINE): Define
1429 as NETBSD_ENABLE_EXECUTE_STACK.
1430 * config/i386/netbsd-elf.h (TRANSFER_FROM_TRAMPOLINE): Ditto.
1431 * config/i386/netbsd.h (TRANSFER_FROM_TRAMPOLINE): Ditto.
1432 * config/i386/netbsd64.h (TRANSFER_FROM_TRAMPOLINE): Ditto.
1433 * config/sparc/netbsd-elf.h (TRANSFER_FROM_TRAMPOLINE): Ditto.
1434 * config/sparc/netbsd.h (TRANSFER_FROM_TRAMPOLINE): Ditto.
1436 2002-10-18 Jason Thorpe <thorpej@wasabisystems.com>
1438 * config/i386/i386.c (x86_initialize_trampoline): Emit a call
1439 to __enable_execute_stack with the address of the trampoline
1440 if TRANSFER_FROM_TRAMPOLINE is defined.
1441 * config/i386/i386.h (TARGET_64BIT): Expand to a compile-time
1442 constant if building libgcc2.
1444 2002-10-17 Roger Sayle <roger@eyesopen.com>
1446 * doc/c-tree.texi: Update description of COND_EXPR tree nodes.
1448 2002-10-17 Geoffrey Keating <geoffk@apple.com>
1450 * config/rs6000/rs6000.h (HARD_REGNO_MODE_OK): Allow arbitrary modes
1452 * config/rs6000/rs6000.md (movcc_internal1): Support CCmode moves
1454 (movsf_hardfloat): Support SFmode moves to/from CTR/LR/MQ.
1455 (movsf_softfloat): Likewise.
1457 2002-10-17 Janis Johnson <janis187@us.ibm.com>
1459 * Makefile.in (site.exp): Add ALT_CXX_UNDER_TEST and COMPAT_OPTIONS.
1461 2002-10-17 Jason Thorpe <thorpej@wasabisystems.com>
1463 * config/alpha/alpha.c (alpha_initialize_trampoline): Use
1464 tramp, not addr, to pass the trampoline address to
1465 __enable_execute_stack.
1467 Thu Oct 17 18:40:47 CEST 2002 Jan Hubicka <jh@suse.cz>
1469 * mmintrin.h: Guard by __MMX__
1470 * xmmintrin.h: Guard by __SSE__
1473 * xmmintrin.h (_MM_SHUFFLE2): New macro.
1474 (_mm_load*_?d): New functions.
1475 (_mm_set*_?d): New functions.
1476 (_mm_store*_?d): New functions.
1478 Wed Oct 16 15:01:29 CEST 2002 Jan Hubicka <jh@suse.cz>
1480 Really commit patch announced at Oct 14
1482 * predict.c (can_predict_insn_p): New function.
1483 (estimate_probability): Avoid unnecesary work.
1484 (process_note_prediction): Likewise.
1485 * toplev.c (rest_of_compilation): Account early branch prediction pass
1489 Found by Ian Ollmann
1490 * xmmintrin.h (_mm_shuffle_pd): Fix typo.
1491 (_mm_load?_pd): Likewise.
1492 (_mm_store?_pd): Likewise.
1495 * i386.c (builtin_description):Drop cmpg[te]s[sd].
1496 * xmmintrin.h (__mm_cmpg[te]_s[sd]): Rewrite using
1497 swapped alternative.
1500 * reload1.c (reload_inner_reg_of_subreg): New argument output;
1501 (push_reload): Update call.
1503 2002-10-17 Richard Sandiford <rsandifo@redhat.com>
1505 * config.gcc (mips*-*-*): Add OBJECT_FORMAT_ELF to $tm_defines
1506 if using mips/elf.h or mips/elf64.h.
1507 * config/mips/elf.h (OBJECT_FORMAT_ELF): Remove.
1508 * config/mips/elf64.h (OBJECT_FORMAT_ELF): Remove.
1510 2002-10-16 Aldy Hernandez <aldyh@redhat.com>
1512 * config/rs6000/rs6000.c (function_arg): Set inner mode of V1DI to
1515 2002-10-16 Ulrich Weigand <uweigand@de.ibm.com>
1517 * config/s390/linux.h (ASM_DOUBLE, _ASM_OUTPUT_LONG): Remove.
1518 (LPREFIX): Likewise.
1519 (ASM_COMMENT_START, LOCAL_LABEL_PREFIX, ASM_FORMAT_PRIVATE_NAME,
1520 ASM_OUTPUT_ADDR_VEC_ELT, ASM_OUTPUT_ADDR_DIFF_ELT,
1521 ASM_OUTPUT_ALIGN, ASM_OUTPUT_SKIP, ASM_OUTPUT_ALIGNED_BSS,
1522 TEXT_SECTION_ASM_OP, DATA_SECTION_ASM_OP, BSS_SECTION_ASM_OP,
1523 GLOBAL_ASM_OP, ASM_OUTPUT_MI_THUNK): Move to s390.h.
1525 * config/s390/s390.h (ASM_COMMENT_START, LOCAL_LABEL_PREFIX,
1526 ASM_FORMAT_PRIVATE_NAME, ASM_OUTPUT_ALIGN, ASM_OUTPUT_SKIP,
1527 ASM_OUTPUT_ALIGNED_BSS, TEXT_SECTION_ASM_OP, DATA_SECTION_ASM_OP,
1528 BSS_SECTION_ASM_OP): Move from linux.h.
1529 (ASM_OUTPUT_ADDR_VEC_ELT, ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1530 Also, use ASM_GENERATE_INTERNAL_LABEL instead of LPREFIX.
1532 * config/s390/s390.c (s390_function_profiler): Use
1533 ASM_GENERATE_INTERNAL_LABEL instead of LPREFIX.
1535 2002-10-15 Richard Henderson <rth@redhat.com>
1537 * real.c (real_to_decimal): Accept BUF_SIZE and CROP_TRAILING_ZEROS
1538 as arguments. Bound DIGITS by the available buffer size.
1539 (real_to_hexadecimal): Likewise.
1540 * real.h (real_to_decimal, real_to_hexadecimal): Update prototypes.
1541 (REAL_VALUE_TO_DECIMAL): Remove.
1542 * c-common.c, c-pretty-print.c, print-rtl.c, print-tree.c,
1543 sched-vis.c, config/arc/arc.c, config/c4x/c4x.c, config/fr30/fr30.c,
1544 config/i370/i370.h, config/i386/i386.c, config/i960/i960.c,
1545 config/ip2k/ip2k.c, config/m32r/m32r.c, config/m68hc11/m68hc11.c,
1546 config/m68k/hp320.h, config/m68k/m68k.h, config/m68k/sun2o4.h,
1547 config/m68k/sun3.h, config/mips/mips.c, config/ns32k/ns32k.c,
1548 config/pdp11/pdp11.h, config/vax/vax.h: Update all callers to
1549 use real_to_decimal directly, and with the proper arguments.
1550 * doc/tm.texi (REAL_VALUE_TO_DECIMAL): Remove.
1552 2002-10-15 Jim Wilson <wilson@redhat.com>
1554 * reload1.c (merge_assigned_reloads): After converting overlapping
1555 reloads to RELOAD_OTHER, abort if there are now conflicting reloads.
1557 * config/i386/i386.md (adddi3_1): Add call to ix86_binary_operator_ok.
1559 Tue Oct 15 22:08:35 CEST 2002 Jan Hubicka <jh@suse.cz>
1561 * expr.c (do_tablejump): Fix typo in my previous commit.
1563 2002-10-15 Richard Sandiford <rsandifo@redhat.com>
1565 * config/mips/vr.h (DRIVER_SELF_SPECS): Change %<mgp32 to %{<mgp32}.
1567 2002-10-15 Ulrich Weigand <uweigand@de.ibm.com>
1569 * config/s390/s390.c (s390_split_branches): Add return
1570 value. Add parameters TEMP_REG and TEMP_USED. Use unspec 104.
1572 (find_base_register_in_addr): New function.
1573 (find_base_register_ref): New function.
1574 (replace_base_register_ref): New function.
1576 (struct constant_pool): Add members pool_insn, insns, and anchor.
1577 Remove member last_insn.
1578 (s390_start_pool): Initialize them.
1579 (s390_end_pool): Emit pool placeholder insn.
1580 (s390_add_pool_insn): New function.
1581 (s390_find_pool): Use insns bitmap instead of addresses.
1582 (s390_dump_pool): Replace placeholder insn. Emit anchor.
1583 Replace unspec 104 by local-pool-relative references.
1584 (s390_output_constant_pool): Output anchor label if required.
1585 (s390_output_symbolic_const): Handle unspec 104 and 105.
1586 (s390_add_pool): Remove, replace by ...
1587 (s390_add_constant, s390_find_constant): ... these new functions.
1588 (s390_add_anchor): New function.
1590 (s390_chunkify_pool): Delete, replace by ...
1591 (s390_chunkify_start, s390_chunkify_finish,
1592 s390_chunkify_cancel): ... these new functions.
1593 (s390_optimize_prolog): Add parameter TEMP_REGNO.
1594 Recompute register live data for special registers.
1595 (s390_fixup_clobbered_return_reg): New function.
1596 (s390_machine_dependent_reorg): Rewrite to use new
1597 s390_chunkify_... routines.
1599 config/s390/s390.md ("reload_base"): Rename to ...
1600 ("reload_base_31"): ... this.
1601 ("reload_base_64"): New insn.
1602 ("reload_base2"): Remove.
1603 ("reload_anchor"): New insn.
1606 s390.c (s390_pool_overflow): Remove.
1607 s390.h (s390_pool_overflow): Likewise.
1608 s390.md ("cjump", "icjump", "doloop_si"): Remove s390_pool_overflow.
1610 Tue Oct 15 16:51:04 2002 J"orn Rennecke <joern.rennecke@superh.com>
1612 * sh.md (movv8qi_i+2): Don't split if source is -1.
1614 2002-10-15 Janis Johnson <janis187@us.ibm.com>
1616 * doc/install.texi: Formatting changes for conformance to HTML 4.01.
1618 2002-10-15 Ulrich Weigand <uweigand@de.ibm.com>
1621 * loop.c (loop_regs_scan): Mark registers used for function
1622 argument passing as MAY_NOT_OPTIMIZE.
1624 Mon Oct 14 20:33:12 CEST 2002 Jan Hubicka <jh@suse.cz>
1626 * i386.md (movv2di_internal): New pattern.
1627 (movv2df_internal, movv8hi_internal, movv16qi_internal): Fix predicate.
1628 (movv2di): New expander.
1629 * i386.c (ix86_preferred_reload_class): Return NO_REGS for vector operands.
1631 * i386.c (ix86_expand_timode_binop_builtin): Delete.
1632 (builtin_description): Add SSE1 logicals; rename SSE2 logicals.
1633 (ix86_init_mmx_sse_builtins): Kill SSE1 logicals.
1634 (ix86_expand_builtin): Likewise.
1635 * i386.h (sse_andti4_df_1, sse_andti3_df_2, sse_andti3_sf_1, sse_andti3_sf_2,
1637 sse_andnti4_df_1, sse_andti3_df_2, sse_andti3_sf_1, sse_andti3_sf_2,
1639 sse_orti4_df_1, sse_orti3_df_2, sse_orti3_sf_1, sse_orti3_sf_2,
1641 sse_xorti4_df_1, sse_xorti3_df_2, sse_xorti3_sf_1, sse_xorti3_sf_2,
1643 (sse_andv4sf3, sse_andnv4sf3, sse_orv2df3, sse_xorv2df3, sse_andv2df3,
1644 sse_andnv2df3, sse_orv2df3, sse_xorv2df3): New expanders.
1645 (*sse_andv4sf3, *sse_andnv2df3, *sse_orv4sf3, *sse_xorv4sf3, *sse_andv2df3,
1646 *sse_andnv2df3, *sse_orv2df3, *sse_xorv2df3): New patterns.
1647 (*sse_andsf3, *sse_andndf3, *sse_ordf3, *sse_xordf3, *sse_anddf3,
1648 *sse_andndf3, *sse_orv2df3, *sse_xorv2df3): New patterns.
1650 * xmmintrin.h (__m128i): Define as __v2di.
1653 (expand_expr): Use DECL_RTL_SET_P.
1655 2002-10-14 Roger Sayle <roger@eyesopen.com>
1657 * combine.c (simplify_set): Treat MODE_CC registers like cc0.
1659 2002-10-14 Roger Sayle <roger@eyesopen.com>
1660 Zack Weinberg <zack@codesourcery.com>
1662 * config/i386/i386.c (k6_cost): Correct typo.
1664 2002-10-14 Mark Mitchell <mark@codesourcery.com>
1666 PR optimization/6631
1667 * alias.c (objects_must_conflict_p): Check honor_readonly when
1668 examining TYPE_READONLY.
1669 * function.c (assign_stack_temp_for_type): Likewise.
1671 2002-10-14 Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
1673 * config/alpha/alpha.md (extendsidi2_nofix, extendsidi2_fix):
1674 Swap zero extension arguments.
1675 (umaxhi3): Fix instruction class.
1677 (prefetch): Fix prefetch instructions.
1679 (pkwb): Fix output constraint.
1681 2002-10-14 Alexandre Oliva <aoliva@redhat.com>
1683 * config/mips/mips.c (print_operand): Increase buffer size for
1686 2002-10-14 Richard Henderson <rth@redhat.com>
1689 * gcse.c (adjust_libcall_notes): Revert last change.
1690 * simplify-rtx.c (simplify_replace_rtx): Handle LO_SUM.
1692 2002-10-14 Andrew Haley <aph@redhat.com>
1694 * tree-inline.c (remap_block): All local class initialization
1695 flags go in the outermost scope.
1696 (expand_call_inline): Call java_inlining_map_static_initializers.
1697 (expand_call_inline): Call java_inlining_merge_static_initializers.
1698 * java/lang.c (merge_init_test_initialization): New.
1699 (java_inlining_merge_static_initializers): New.
1700 (inline_init_test_initialization): New.
1701 (java_inlining_map_static_initializers): New.
1703 * tree-inline.c (expand_call_inline): Convert retvar to expected
1706 2002-10-14 Graham Stott <graham.stott@btinternet.com>
1708 * stmt.c (decl_conflicts_with_clobbers_p): Add REG_P check.
1710 2002-10-14 Aldy Hernandez <aldyh@redhat.com>
1712 * stmt.c: Fix typo in comment.
1714 Mon Oct 14 11:35:49 2002 J"orn Rennecke <joern.rennecke@superh.com>
1716 * c-common.c (c_common_type_for_mode): Add V2HImode case.
1717 * tree.c (build_common_tree_nodes_2): Initialize
1718 unsigned_V2HI_type_node and V2HI_type_node.
1719 * tree.h (enum tree_index): Add TI_UV2HI_TYPE and TI_V2HI_TYPE.
1720 (unsigned_V2HI_type_node, V2HI_type_node): Define.
1722 2002-10-14 Jakub Jelinek <jakub@redhat.com>
1724 * config/i386/i386.h (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP):
1725 Handle TARGET_64BIT.
1727 2002-10-14 Richard Sandiford <rsandifo@redhat.com>
1729 * config/mips/vr.h (DRIVER_SELF_SPECS): Define.
1730 * config/mips/t-vr (MULTILIB_OPTIONS): Remove mlong32.
1731 (MULTILIB_DIRNAMES): Remove long32.
1732 (MULTILIB_EXCEPTIONS): Don't build -mabi=32 -mgp32 multilibs.
1733 (MULTILIB_REDUNDANT_DIRS): Remove.
1735 2002-10-14 Richard Sandiford <rsandifo@redhat.com>
1737 * doc/tm.texi (DRIVER_SELF_SPECS): Document.
1738 * gcc.c (driver_self_specs): New variable.
1739 (do_self_spec): New function.
1740 (main): Use it to process driver_self_specs.
1742 2002-10-13 Richard Henderson <rth@redhat.com>
1744 * rtl.c (shallow_copy_rtx): Use memcpy for the entire node.
1746 2002-10-12 Neil Booth <neil@daikokuya.co.uk>
1748 PR preprocessor/7862
1749 PR preprocessor/8190
1750 * gcc.c (cpp_unique_options): Don't delete .d files.
1751 Remove stray whitespace.
1753 2002-10-12 Naohiko Shimizu <pshimizu@fa2.so-net.ne.jp>
1755 * pdp11.h (ASM_OUTPUT_SKIP): Add preceding 0 for octal constant.
1756 (ASM_OUTPUT_COMMON, ASM_OUTPUT_LOCAL): Likewise.
1757 * pdp11.c (pdp11_output_function_prologue): 0%o -> %#o.
1758 (pdp11_output_function_epilogue, output_ascii): Likewise.
1759 (output_addr_const_pdp11): Likewise.
1760 * pdp11.md (movdi): Use offsetable memory for floating store.
1761 (lshrsi3, negsi2): Delete irrelevant comment.
1763 2002-10-11 Geoffrey Keating <geoffk@apple.com>
1765 * cse.c (mention_regs): Set SUBREG_TICKED to the register number,
1766 not the address of the REG.
1767 (struct cse_reg_info): Make subreg_ticked unsigned.
1769 2002-10-11 Janis Johnson <janis187@us.ibm.com>
1771 * doc/compat.texi: Add info about C++ libraries.
1773 2002-10-11 Richard Henderson <rth@redhat.com>
1776 * gcse.c (adjust_libcall_notes): Also adjust notes for INSN.
1778 2002-10-11 John David Anglin <dave@hiauly1.hia.nrc.ca>
1780 * cfganal.c (dfs_enumerate_from): Use PARAMS.
1781 * genautomata.c (output_insn_code_cases): Likewise.
1782 * real.c (real_format): Likewise.
1783 * tree.c (tree_size): Revise expressions using TREE_CODE_LENGTH to
1784 ensure value is promoted before doing subtraction.
1786 Fri Oct 11 22:22:38 CEST 2002 Jan Hubicka <jh@suse.cz>
1788 * calls.c (expand_call): Simplify noreturn call.
1791 * cfgbuild.c (make_edges): Create edge cache when we do have
1793 * expr.c (do_tablejump): Note size of maximal jumptable.
1794 * function.c (prepare_function_start): Zero out size.
1795 * function.h (function): Add max_jumptable_ents.
1797 * cfgcleanup.c (insn_match_p): Verify sibcall flag for calls to.
1799 Fri Oct 11 12:34:33 2002 J"orn Rennecke <joern.rennecke@superh.com>
1801 * sh.md (movv8qi_i+2): For V8QI destinations, generate V4HI
1802 register for mperm_w operation.
1804 Fri Oct 11 10:56:17 2002 J"orn Rennecke <joern.rennecke@superh.com>
1806 * emit-rtl.c (gen_lowpart_common): When asked to make a vector from
1807 an integer, use simplify_gen_subreg.
1809 2002-10-10 Aldy Hernandez <aldyh@redhat.com>
1811 * extend.texi (Vector Extensions): Remove comment about single
1814 2002-10-10 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1816 * fold-const.c (size_htab_hash): Use htab_hash_pointer.
1817 * function.c (insns_for_mem_hash): Likewise.
1818 * varasm.c (STRHASH): Likewise.
1820 2002-10-10 Stuart Hastings <stuart@apple.com>
1822 * cse.c (struct cse_reg_info): Add subreg_ticked.
1823 (SUBREG_TICKED): New.
1824 (get_cse_reg_info): Initialize SUBREG_TICKED.
1825 (mention_regs): Use it.
1826 (invalidate): Set SUBREG_TICKED.
1827 (invalidate_for_call): Likewise.
1828 (addr_affects_sp_p): Likewise.
1830 2002-10-10 Jakub Jelinek <jakub@redhat.com>
1832 * config/i386/i386.md (tls_local_dynamic_base): Put pic reg
1833 into proper operand.
1835 2002-10-10 Denis Chertykov <denisc@overta.ru>
1837 * config/ip2k/ip2k.c (function_epilogue): Optimize stack
1839 * config/ip2k/libgcc.S: Combine routines used by function
1842 2002-10-10 Jim Wilson <wilson@redhat.com>
1844 * cse.c (fold_rtx): Don't perform associative optimization for DIV and
1847 2002-10-10 David Edelsohn <edelsohn@gnu.org>
1849 * config/rs6000/aix52.h: New file.
1850 * config/rs6000/t-aix52: New File.
1851 * config.gcc (rs6000-ibm-aix5.1.*): New entry.
1852 (rs6000-ibm-aix[56789].*): Default to AIX 5.2.
1854 Thu Oct 10 19:37:54 CEST 2002 Jan Hubicka <jh@suse.cz>
1857 * invoke.texi (-msse-math): Kill
1858 (-msse): Add note to mfpmath=sse.
1860 Thu Oct 10 17:08:30 CEST 2002 Jan Hubicka <jh@suse.cz>
1863 * i386.c (ix86_expand_vector_move): Do not generate const0->mem moves.
1865 2002-10-10 Neil Booth <neil@daikokuya.co.uk>
1867 PR preprocessor/8179
1868 * gcc.c (cpp_options): Add {ansi}, move %{m*} to same location
1870 (default_compilers): Pass debug options when preprocessing
1873 2002-10-06 Richard Henderson <rth@redhat.com>
1875 * toplev.c (rest_of_compilation): Revert opt/2960 change.
1877 2002-10-09 Zack Weinberg <zack@codesourcery.com>
1880 * c-decl.c (start_decl): Unconditionally issue error for
1881 'typedef foo = bar'.
1882 (finish_decl): Remove special case for TYPE_DECL with initializer.
1884 * doc/extend.texi: Delete "Naming Types" section. Change all
1885 cross-references to that section to refer to "Typeof" instead.
1886 Add the useful safe-max()-macro example from "Naming Types" to
1887 "Typeof", rewritten using that extension. Add some compatibility
1890 2002-10-09 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1892 * loop.c: Revert 2002-08-15 change.
1893 (LOOP_REGNO_NREGS): Ensure type is int.
1895 2002-10-09 David Edelsohn <edelsohn@gnu.org>
1897 * config/rs6000/rs6000.md (extenddftf2): Change to define_insn
1898 which copies first FPR and clears second.
1899 (extendsftf2): Same.
1900 (floatditf2): Fix typo.
1902 (fix_trunctfdi2): Same.
1903 (fix_trunctfsi2): Same.
1905 2002-10-09 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1907 * conflict.c (arc_hash): Change return type to hashval_t.
1908 * cselib.c (get_value_hash): Likewise.
1909 * genautomata.c (automaton_decl_hash, insn_decl_hash, decl_hash,
1910 state_hash, automata_list_hash): Likewise.
1911 * read-rtl.c (def_hash): Likewise.
1912 * tree.c (type_hash_hash): Likewise.
1914 2002-10-08 Aldy Hernandez <aldyh@redhat.com>
1916 * config/rs6000/rs6000.c (rs6000_ra_ever_killed): Call
1917 prologue_epilogue_contains instead of using REG_MAYBE_DEAD notes.
1919 Wed Oct 9 15:54:49 2002 J"orn Rennecke <joern.rennecke@superh.com>
1921 * sh.md (ffssi2): Fix emitted code.
1923 2002-10-09 Ulrich Weigand <uweigand@de.ibm.com>
1925 * cse.c (insn_live_p): Pass insn pattern, not full insn
1928 2002-10-09 Neil Booth <neil@daikokuya.co.uk>
1930 * cppmacro.c (paste_tokens): Only allow / to paste with =.
1932 2002-10-09 David Edelsohn <edelsohn@gnu.org>
1934 * config/rs6000/rs6000.md (movdf splitter): Use gen_int_mode on
1936 (movtf_internal): Reference correct displacement for second value
1938 (movtf splitter): Correct generation of constants in 64-bit mode.
1940 2002-10-09 Alan Modra <amodra@bigpond.net.au>
1942 * libgcc2.c (__floatdisf): Properly cure double rounding.
1944 2002-10-09 Gabriel Dos Reis <gdr@integrable-solutions.net>
1946 * c-common.c (cb_register_builtins): Define __WCHAR_MAX__.
1947 * doc/cpp.texi (Common Predefined Macros): Document.
1949 2002-10-09 Gabriel Dos Reis <gdr@integrable-solutions.net>
1952 * doc/invoke.texi (Option Summary): List
1953 -Wmissing-declarations as a C only option.
1955 2002-10-08 Jakub Jelinek <jakub@redhat.com>
1957 * config/sparc/t-linux64 (MULTILIB_OPTIONS): Remove
1958 mno-app-regs|mcmodel=medany.
1959 (MULTILIB_DIRNAMES, MULTILIB_OSDIRNAMES): Remove alt.
1960 (MULTILIB_EXCEPTIONS, MULTILIB_EXCLUSIONS, MULTILIB_MATCHES): Remove.
1961 (CRTSTUFF_T_CFLAGS): Define.
1963 2002-10-08 Roger Sayle <roger@eyesopen.com>
1966 * simplify-rtx.c (avoid_constant_pool_reference): Allow constant
1967 pool references that are constructed using LO_SUM.
1969 2002-10-08 Nathan Sidwell <nathan@codesourcery.com>
1971 * c-opts.c (c_common_decode_option): Add warn_strict_aliasing to
1973 * c-typeck.c (build_c_cast): Use warn_strict_aliasing, tweak
1975 * flags.h (warn_strict_aliasing): Declare.
1976 * toplev.c (warn_strict_aliasing): Define.
1977 (lang_independent_options): Add it.
1978 * doc/invoke.texi (-Wstrict-aliasing): Document it.
1980 2002-10-08 Zack Weinberg <zack@codesourcery.com>
1982 * system.h (GCCBUGURL): Delete.
1983 * version.c (bug_report_url): New. Add commentary about
1984 modifying both these strings in modified distributions.
1985 * version.h: Declare bug_report_url.
1987 * diagnostic.c, gcc.c, gcov.c: Globally replace GCCBUGURL with
1990 2002-10-08 Nick Clifton <nickc@redhat.com>
1992 * config/rs6000/spe.h (__ev_set_acc_u64): Use __ev_create_u64 to
1993 convert uint64_t into __ev64_opaque__.
1994 (__ev_set_acc_s64): Likewise, but using signed types.
1996 2002-10-08 Ulrich Weigand <uweigand@de.ibm.com>
1998 * config/s390/s390.md ("*doloop_si_long"): Add missing operand.
1999 ("*doloop_di_long"): Likewise.
2001 Tue Oct 8 16:50:10 CEST 2002 Jan Hubicka <jh@suse.cz>
2003 * print-rtl.c (print_rtx): Increase buffer size for real numbers.
2005 2002-10-08 Richard Sandiford <rsandifo@redhat.com>
2007 * config/mips/mips.md (define_attr cpu): Add r4111.
2009 2002-10-08 Anthony Green <green@redhat.com>
2011 * bitmap.c (bitmap_equal_p): Clear all bitmap_head fields.
2013 2002-10-08 Michael Hayes <m.hayes@elec.canterbury.ac.nz>
2015 * config/c4x/c4x.c (c4x_print_operand): Enlarge buffer
2016 for REAL_VALUE_TO_DECIMAL output.
2018 2002-10-07 David Edelsohn <edelsohn@gnu.org>
2020 * config/rs6000/rs6000.c (rs6000_override_options): Set
2021 real_format_for_mode for IBM extended format, if enabled.
2022 (easy_fp_constant): Add TFmode.
2023 (rs6000_legitimize_address): Add TFmode.
2024 (rs6000_legitimate_address): Same.
2025 (function_arg_advance): TFmode uses two FPRs.
2026 (rs6000_emit_prologue): Fix warning.
2027 (rs6000_output_function_epilogue): Add TFmode.
2028 (output_toc): Add TFmode.
2029 * rs6000.h (SLOW_UNALIGNED_ACCESS): Add TFmode.
2030 (LEGITIMATE_OFFSET_ADDRESS_P): Add TFmode.
2031 * rs6000.md (movtf splitter): Load TFmode constant.
2033 2002-10-07 Dale Johannesen <dalej@apple.com>
2035 * rtl.h: Add NOTE_PRECONDITIONED.
2037 * loop.c: Set loop_info->preconditioned from it.
2038 * doloop.c: Permit doloop treatment when loop_info->preconditoned.
2040 2002-10-07 Richard Henderson <rth@redhat.com>
2042 * config/i960/i960.c (i960_setup_incoming_varargs): Create a
2043 new rtx for comparing the argument pointer against zero.
2044 (i960_va_start): Similarly.
2046 2002-10-07 Richard Henderson <rth@redhat.com>
2048 * config/i960/i960.md (*): Use TFmode, not XFmode.
2049 * config/i960/i960.c (*): Likewise.
2050 (i960_arg_size_and_align): Remove XFmode alignment hack.
2051 (i960_round_align): Merge code from ROUND_TYPE_ALIGN.
2052 * config/i960/i960.h (LONG_DOUBLE_TYPE_SIZE): Use 128, not 96.
2053 (MAX_LONG_DOUBLE_TYPE_SIZE): Likewise.
2054 (DATA_ALIGNMENT, ROUND_TYPE_SIZE): Remove.
2056 2002-10-07 Richard Henderson <rth@redhat.com>
2058 * config/fp-bit.c (EXTENDED_FLOAT_STUBS): Flush out all XF/TFmode
2059 entry points; use void return value and argument list.
2061 2002-10-05 Naohiko Shimizu <nshimizu@keyaki.cc.u-tokai.ac.jp>
2063 * t-pdp11: Add MULTILIB support for msoft-float.
2064 * pdp11.h (LEGITIMATE_CONSTANT_P): Fix soft-float case.
2066 * t-pdp11: Add LIB2FUNCS_EXTRA.
2067 * pdp11.c (pdp11_output_function_prologue): Restrict offset to 16bit,
2068 add preceding 0 to the octal constant, rename 'fp' to 'r5', rename
2069 'fldd' to 'ldd', rename 'fstd' to 'std'.
2070 (pdp11_output_function_epilogue): Likewise.
2071 (output_move_quad): Make the comment gas compatible.
2072 (output_ascii): Add preceding 0 to the octal constant.
2073 (print_operand_address): Add pre_modify, post_modify.
2074 (output_addr_const_pdp11): Add preceding 0 to the octal constant.
2075 * pdp11.h (GO_IF_LEGITIMATE_ADDRESS) : Add 'movb' pre_modify case
2076 with the indication of Paul Koning.
2077 (PRINT_OPERAND): Fix floating constant.
2078 * pdp11.md (movdi): Restrict matching pattern.
2079 (movqi): Generalize the matching pattern.
2080 (movdf): Restrict matching pattern.
2081 (zero_extendqihi2): Change constant representation.
2082 (floatsidf2): Fix wrong operands.
2083 (addqi3): Fix wrong instruction name.
2084 (subqi3): Fix wrong instruction name.
2085 (andsi3, andhi3, andqi3): Simplify and fix to use 'bic'.
2086 (xorsi3): Fix wrong insn.
2087 (one_cmplqi2): Add two operand pattern.
2090 (call): Add register indirect case.
2091 (mod): Fix wrong subreg.
2093 2002-09-25 Eric Botcazou <ebotcazou@libertysurf.fr>
2094 Volker Reichelt <reichelt@igpm.rwth-aachen.de>
2097 * expr.c (expand_expr) [PLUS]: Simplify after the operands
2098 have been expanded in EXPAND_NORMAL mode.
2100 2002-10-06 Richard Henderson <rth@redhat.com>
2102 * config/rs6000/rs6000.md (load_toc_v4_PIC_2): Fix base constraint.
2104 2002-10-06 Richard Henderson <rth@redhat.com>
2106 PR optimization/2960
2107 * toplev.c (rest_of_compilation): Don't copy_loop_headers if
2110 2002-10-06 Alexandre Oliva <aoliva@redhat.com>
2112 * config/mips/mips.h (SIZE_TYPE, PTRDIFF_TYPE): Override
2113 previously definitions.
2115 2002-10-06 Frank Ch. Eigler <fche@redhat.com>
2117 * cppinit.c (init_standard_includes, parse_option): Use strncmp.
2118 * c-opts.c (find_opt): Similarly.
2120 2002-10-05 Jakub Jelinek <jakub@redhat.com>
2122 * gcc.c (set_multilib_dir): Don't access *end.
2123 Use memcpy instead of strncpy. Don't write beyond malloced buffer.
2124 (print_multilib_info): Don't show paths starting with ".:".
2125 * genmultilib: Add new option, "yes" if multilibs are enabled.
2126 Update comments. If multilibs not enabled, print .:${osdirout}
2127 for each directory. If multilibs are enabled, always print
2128 ${dirout}:${osdirout}, even if the two are the same.
2129 * Makefile.in (s-mlib): Pass @enable_multilib@ to genmultilib.
2130 Pass all MULTILIB_* variables to genmultilib even if
2131 --disable-multilib but MULTILIB_OSDIRNAMES is not empty.
2133 2002-10-04 Bruce Korb <bkorb@gnu.org>
2135 * fixinc/inclhack.def(hpux11_abs): use format fix
2136 * fixinc/fixincl.x: regenerate
2137 * fixinc/tests/base/stdlib.h: accommodate new fix test
2139 Sat Oct 5 19:42:45 CEST 2002 Jan Hubicka <jh@suse.cz>
2141 * c-common.c (cb_register_builtins): Use really_no_inline.
2143 2002-10-04 David Edelsohn <edelsohn@gnu.org>
2145 * unroll.c (copy_loop_body): Remove REG_EQUAL note attached to
2146 copied instruction if the note is not loop invariant.
2148 2002-10-04 Loren J. Rittle <ljrittle@acm.org>
2150 * gcc/ginclude/stddef.h: Support the FreeBSD 5 typedef system.
2152 2002-10-04 Steve Ellcey <sje@cup.hp.com>
2154 * doc/invoke.texi (HPPA): Add -mlinker-opt, -mgnu-ld,
2155 and -mhp-ld options to list of options. Add -mgnu-ld
2156 and -mhp-ld option descriptions.
2158 2002-10-04 Steve Ellcey <sje@cup.hp.com>
2160 * fixinc/inclhack.def (hpux11_abs): New.
2161 (stdio_va_list): change __va_list__ to __gnuc_va_list.
2162 * fixinc/fixincl.x: Rebuild.
2164 2002-10-04 Roger Sayle <roger@eyesopen.com>
2166 * config/i386/i386.h (processor_costs): Add new fields fadd,
2167 fmul, fdiv, fabs, fchs and fsqrt to costs structure.
2168 (RTX_COSTS): Use these fields to determine the RTX costs
2169 of floating point addition/subtraction, multiplication,
2170 division, fabs, negation and square root respectively.
2171 * config/i386/i386.c (size_cost): Provide instruction sizes
2172 for these new fields.
2173 (i386_cost, i486_cost, pentium_cost, pentiumpro_cost,
2174 k6_cost, athlon_cost, pentium4_cost): Provide typical cycle
2175 counts for these new fields for all x86 processor variants.
2177 2002-10-04 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2179 * mips.c (mips_const_double_ok): Delete unused variable.
2181 * gengtype.c (rtx_next): Change type to int.
2183 2002-10-04 Andreas Jaeger <aj@suse.de>
2185 * config/i386/t-linux64 (MULTILIB_OSDIRNAMES): Fix value.
2187 2002-10-04 Richard Henderson <rth@redhat.com>
2189 * real.h (SIGNIFICAND_BITS): Add one more word.
2190 (CONST_DOUBLE_FORMAT): Accomodate 6 words.
2191 * real.c (times_pten): New.
2192 (real_to_decimal, real_from_string): Use it.
2193 (sticky_rshift_significand): Use & to find modulus.
2194 (rshift_significand, lshift_significand): Likewise.
2195 (do_divide): Apply sticky bit after normalization.
2196 (real_to_decimal, real_to_hexadecimal): Fix sign of Inf and NaN.
2198 2002-10-03 Andreas Jaeger <aj@suse.de>
2200 * gengtype.c (adjust_field_rtx_def): Cast variables of type size_t
2201 to unsigned long, adjust printf format string.
2202 (output_mangled_typename): Likewise.
2204 2002-10-03 Jason Thorpe <thorpej@wasabisystems.com>
2206 * config/vax/vax.c (vax_output_function_prologue): Use asm_fprintf.
2207 * config/vax/vax.h (VAX_FUNCTION_PROFILER_NAME): New.
2208 (FUNCTION_PROFILER): Rewrite to use ASM_GENERATE_INTERNAL_LABEL,
2209 assemble_name, asm_fprintf, and VAX_FUNCTION_PROFILER_NAME.
2210 (ASM_OUTPUT_MI_THUNK): Use asm_fprintf instead of REGISTER_PREFIX.
2211 (PRINT_OPERAND_PUNCT_VALID_P): Fix comment.
2212 * config/vax/elf.h (FUNCTION_PROFILER): Remove.
2213 (VAX_FUNCTION_PROFILER_NAME): Redefine as "__mcount".
2215 2002-10-03 Mark Mitchell <mark@codesourcery.com>
2217 * doc/invoke.texi (-Wabi): Document mangling bug.
2219 2002-10-04 Alan Modra <amodra@bigpond.net.au>
2221 * config/rs6000/rs6000.c (rs6000_output_function_epilogue): Use a
2222 name for the tbtab label that depends on the function asm name.
2223 Don't output tbtab label unless optional_tbtab.
2224 (output_mi_thunk): Formatting.
2226 2002-10-03 Richard Henderson <rth@redhat.com>
2228 * config/m68k/m68k.h (OVERRIDE_OPTIONS): Move additional code ...
2229 * config/m68k/m68k.c (override_options): ... here.
2230 * config/m68k/m68kelf.h (OVERRIDE_OPTIONS): Remove.
2231 * config/m68k/m68kv4.h (OVERRIDE_OPTIONS): Remove.
2232 * config/m68k/linux.h (SUBTARGET_OVERRIDE_OPTIONS): Remove.
2233 * config/m68k/netbsd-elf.h (SUBTARGET_OVERRIDE_OPTIONS): Remove.
2235 2002-10-03 Richard Henderson <rth@redhat.com>
2237 * real.h (struct real_value): Use ENUM_BITFIELD.
2239 2002-10-03 Richard Henderson <rth@redhat.com>
2241 * config/i960/i960.md (call, call_value): Use emit_call_insn.
2243 2002-10-03 Steve Ellcey <sje@cup.hp.com>
2245 * config/pa/pa64-hpux.h (INIT_ENVIRONMENT): New.
2247 2002-10-03 Steve Ellcey <sje@cup.hp.com>
2249 * config.gcc (hppa*64*-*-hpux11*): Check gnu_ld.
2250 * config/pa/pa.h (MASK_GNU_LD): New.
2251 (TARGET_GNU_LD): New.
2252 * config/pa/pa64-hpux.h (LINK_SPEC): Set based
2253 on gnu-ld and MASK_GNU_LD.
2254 (SUBTARGET_SWITCHES): New gnu-ld & hp-ld flags.
2256 Thu Oct 3 23:35:51 CEST 2002 Jan Hubicka <jh@suse.cz>
2258 * i386.c (athlon_cost): Fix the move costs.
2260 Thu Oct 3 23:20:58 CEST 2002 Jan Hubicka <jh@suse.cz>
2262 * final.c (final): Use symbol name as function name for profiling.
2263 * profile.c (get_exec_counts): Likewise.
2264 (branch_prob): Likewise.
2266 2002-10-03 Jakub Jelinek <jakub@redhat.com>
2268 * longlong.h (__udiv_qrnnd): Remove PARAMS from prototype.
2270 2002-10-03 Jakub Jelinek <jakub@redhat.com>
2272 * gcc.c (print_multi_os_directory): New variable.
2273 (option_map): Support --print-multi-os-directory.
2274 (struct prefix_list): Add os_multilib field.
2275 (multilib_os_dir): New variable.
2276 (static_specs): Add multilib_options.
2277 (find_a_file): Add multilib argument. Search in GCC or OS multilib
2278 subdirs if non-zero.
2279 (read_specs, execute): Update callers.
2280 (find_file): Likewise. Don't prefix name with multilib_dir, instead
2281 pass 1 as multilib option.
2282 (display_help): Include --print-multi-os-directory.
2283 (add_prefix): Add os_multilib argument. Initialize pl->os_multilib.
2284 (process_command): Update callers. Handle --print-multi-os-directory.
2285 (do_spec_1) ['D']: Use multilib_os_directory if pl->os_multilib is
2287 (main): Update find_a_file and add_prefix callers.
2288 Handle print_multi_os_directory.
2289 (struct mdswitchstr): New.
2290 (mdswitches, n_mdswitches): New variables.
2291 (used_arg): Add MULTILIB_DEFAULT switches too if they are not
2292 present on the command line nor their mutually incompatible
2294 (default_arg): Optimize.
2295 (set_multilib_dir): Compute multilib_os_dir. Initialize mdswitches
2297 (print_multilib_info): Only print GCC multilib dir name, not OS
2299 * genmultilib: Add osdirnames parameter. Output multilib_options
2300 variable. If osdirnames is specified, output dirnames as
2302 * mklibgcc.in: Use MULTILIB_OSDIRNAMES, --print-multi-directory
2303 and --print-multi-os-directory instead of SHLIB_SLIBDIR_SUFFIXES
2304 to compute libgcc_s soname and install path.
2305 * Makefile.in (libgcc.mk): Pass MULTILIB_OSDIRNAMES instead of
2306 SHLIB_SLIBDIR_SUFFIXES to mklibgcc.
2307 (s_mlib): Pass MULTILIB_OSDIRNAMES or nothing as last genmultilib
2310 * config/sparc/t-linux64 (MULTILIB_OSDIRNAMES): Set.
2311 (SHLIB_SLIBDIR_SUFFIXES): Remove.
2312 * config/sparc/linux64.h (STARTFILE_SPEC32, STARTFILE_SPEC64,
2313 ENDFILE_SPEC32, ENDFILE_SPEC64, ENDFILE_COMMON): Remove.
2314 (STARTFILE_SPEC, ENDFILE_SPEC): Don't distinguish between -m32
2316 * config/sparc/t-sol2-64 (MULTILIB_OSDIRNAMES): Set.
2317 (SHLIB_SLIBDIR_SUFFIXES): Remove.
2318 * config/sparc/sol2-bi.h (STARTFILE_ARCH64_SPEC): Remove.
2319 (STARTFILE_ARCH_SPEC): Remove.
2320 * config/i386/t-linux64 (MULTILIB_OSDIRNAMES): Set.
2321 (SHLIB_SLIBDIR_SUFFIXES): Remove.
2322 * config/i386/linux64.h (STARTFILE_PREFIX_SPEC): Remove.
2323 * config/mips/t-iris6 (MULTILIB_OSDIRNAMES): Set.
2324 (SHLIB_SLIBDIR_SUFFIXES): Remove.
2326 Thu Oct 3 21:42:20 CEST 2002 Jan Hubicka <jh@suse.cz>
2328 * predict.c (choose_function_section): Avoid choice for linkonce functions.
2330 Thu Oct 3 15:15:00 CEST 2002 Jan Hubicka <jh@suse.cz>
2332 * i386.md (lea to mul peep2): Fix condition.
2334 2002-10-02 John David Anglin <dave@hiauly1.hia.nrc.ca>
2336 * pa-linux.h (FUNCTION_OK_FOR_SIBCALL): Delete macro.
2337 * pa32-linux.h (FUNCTION_OK_FOR_SIBCALL): Define.
2339 2002-10-02 Richard Henderson <rth@redhat.com>
2342 * config/i386/i386.c (ix86_register_move_cost): Increase cost
2343 for secondary_memory_needed pairs.
2345 2002-10-02 Nathanael Nerode <neroden@gcc.gnu.org>
2347 * doc/vms.texi: Blow away false include file section.
2349 2002-10-02 Roger Sayle <roger@eyesopen.com>
2351 PR optimization/6627
2352 * toplev.c (force_align_functions_log): New global variable.
2353 * flags.h (force_align_functions_log): Add extern prototype.
2354 * varasm.c (assemble_start_function): Use it to force minimum
2356 * config/i386/i386.h (FUNCTION_BOUNDARY): Set the correct
2357 minimum function alignment to one byte.
2358 (TARGET_PTRMEMFUNC_VBIT_LOCATION): Store the virtual bit in
2359 the least significant bit of vtable member function pointers.
2360 * tree.h (enum ptrmemfunc_vbit_where_t): Move definition to
2361 here from cp/cp-tree.h.
2363 Wed Oct 2 17:01:36 CEST 2002 Jan Hubicka <jh@suse.cz>
2365 * i386.c (print_operand_address): Use RIP addressing for offsetted
2368 2002-09-30 David S. Miller <davem@redhat.com>
2371 * config/sparc/sparc.md (movdi_insn_sp32_v9): Accept 'e' regs.
2372 (movdi reg/reg split): Match only on sparc32, and v9 when int regs.
2374 2002-10-01 Roger Sayle <roger@eyesopen.com>
2376 * unroll.c (loop_iterations): Revert 2002-09-08 change.
2378 2002-10-01 Richard Henderson <rth@redhat.com>
2380 * real.c (real_to_decimal): Crop trailing zeros for DIGITS < 0.
2381 (real_to_hexadecimal): Likewise.
2382 * print-rtl.c (print_rtx): If we are linked with real.c, don't
2383 dump the XWINT fields of a floating point CONST_DOUBLE.
2385 2002-10-01 Jason Thorpe <thorpej@wasabisystems.com>
2387 * config/vax/elf.h (FUNCTION_PROFILER): Fix __mcount call.
2389 2002-10-01 Richard Henderson <rth@redhat.com>
2391 * calls.c (precompute_register_parameters): Force non-legitimate
2392 constants into pseudos.
2394 2002-10-01 Nick Clifton <nickc@redhat.com>
2396 * config/rs6000/spe.md (spe_evrlwi): Add missing third operand
2397 to assembler template.
2399 2002-10-01 Richard Henderson <rth@redhat.com>
2401 * dwarf2out.c (loc_descriptor_from_tree): Relax requirement
2402 for TLS debug info to !DECL_EXTERNAL.
2404 2002-10-01 Matt Thomas <matt@3am-software.com>
2405 Jason Thorpe <thorpej@wasabisystems.com>
2407 * config.gcc (vax-*-netbsdelf*): Enable configuration.
2408 * config/elfos.h (PCC_BITFIELD_TYPE_MATTERS): Define only
2409 if not already defined.
2410 * config/vax/elf.h: New file.
2411 * config/vax/netbsd-elf.h: New file.
2412 * config/vax/vax.c: Include "debug.h".
2413 (vax_output_function_prologue): Add dwarf2 support. Use
2414 MAIN_NAME_P when checking for VMS_TARGET stack adjust.
2415 * config/vax/vax.h (CONST_OK_FOR_LETTER_P): Add cases for
2416 'J' [0..63], 'K' [-128..127], 'L' [-32768..32767],
2417 'M' [0..255], 'N' [0..65535], and, 'O' [-63..-1].
2418 (VAX_ISTREAM_SYNC): Remove.
2419 (INITIALIZE_TRAMPOLINE): Use gen_sync_istream.
2420 (JUMP_TABLES_IN_TEXT_SECTION): Define.
2421 (ASM_OUTPUT_REG_POP): Use reg_names for the stack pointer.
2422 (ASM_OUTPUT_ADDR_VEC_ELT): Use ASM_GENERATE_INTERNAL_LABEL
2424 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2425 (PRINT_OPERAND_PUNCT_VALID_P): Accept '|'.
2426 (PRINT_OPERAND): Output REGISTER_PREFIX for '|'.
2427 (INCOMING_RETURN_ADDR_RTX): Define.
2428 * config/vax/vax.md (VUNSPEC_BLOCKAGE)
2429 (VUNSPEC_SYNC_ISTREAM): Define.
2430 (blockage): Use VUNSPEC_BLOCKAGE.
2431 (sync_istream): New insn.
2433 2002-10-01 Richard Henderson <rth@redhat.com>
2435 * config/vax/vax.md (call_pop, *call_pop, call_value_pop)
2436 (*call_value_pop, call, call_value): Add dwarf2 EH support.
2439 2002-10-01 Nathan Sidwell <nathan@codesourcery.com>
2442 * c-typeck.c (build_c_cast): Warn about type punning which breaks
2443 type based aliasing.
2445 2002-10-01 Mark Mitchell <mark@codesourcery.com>
2447 * stor-layout.c (update_alignment_for_field): New function.
2448 (place_union_field): Use it.
2449 (place_field): Likewise.
2451 2002-10-01 Nathan Sidwell <nathan@codesourcery.com>
2454 * gcc.c (cc1_options): Add space on -auxbase-strip.
2456 2002-10-01 Jim Wilson <wilson@redhat.com>
2458 * config/v850/v850.h (EPILOGUE_USES): Define.
2460 2002-09-30 Andrew Haley <aph@redhat.com>
2462 * flow.c (insn_dead_p): When using non-call-exceptions, don't
2463 eliminate insns that may trap.
2464 * cse.c (insn_live_p): Likewise.
2466 2002-10-01 Richard Sandiford <rsandifo@redhat.com>
2468 * config/mips/mips.h (PROCESSOR_R4121): Rename to PROCESSOR_R4120.
2469 (TARGET_MIPS4121): Rename to TARGET_MIPS4120.
2470 * config/mips/mips.c (mips_cpu_info): Rename vr4121 to vr4120.
2471 * config/mips/mips.md: Apply same renaming here.
2473 2002-10-01 Richard Sandiford <rsandifo@redhat.com>
2475 * config/mips/mips.c (PROCESSOR_R4320, TARGET_MIPS4320): Remove.
2476 (GENERATE_MULT3_SI): Remove use of TARGET_MIPS4320.
2477 * config/mips/mips.c (mips_cpu_info): Remove vr4320 entry.
2478 * config/mips/mips.md (define_attr cpu): Remove r4320.
2479 Remove vr4320 scheduler and uses of TARGET_MIPS4320.
2481 2002-10-01 Richard Sandiford <rsandifo@redhat.com>
2483 * config/mips/mips.c (mips16_strings): New variable.
2484 (mips_output_function_epilogue): Clear the SYMBOL_REF_FLAG of every
2485 symbol in mips16_strings. Free the list.
2486 (mips_encode_section_info): Keep track of local strings.
2488 2002-10-01 Richard Sandiford <rsandifo@redhat.com>
2490 * config/mips/mips.md (bunge, bltgt, bungt): New define_expands.
2491 (sordered_df, sordered_sf): Remove.
2492 * config/mips/mips.c (get_float_compare_codes): New fn.
2493 (gen_int_relational, gen_conditional_move): Use it.
2495 2002-10-01 Richard Sandiford <rsandifo@redhat.com>
2497 * config/mips/mips-protos.h (mips_emit_fcc_reload): Declare.
2498 * config/mips/mips.h (PREDICATE_CODES): Add fcc_register_operand.
2499 * config/mips/mips.c (fcc_register_operand): New function.
2500 (mips_emit_fcc_reload): New function, extracted from reload_incc.
2501 (override_options): Allow TFmode values in float registers
2503 * cnfig/mips/mips.md (reload_incc): Change destination prediate
2504 to fcc_register_operand. Remove misleading source constraint.
2505 Use mips_emit_fcc_reload.
2506 (reload_outcc): Duplicate reload_incc.
2508 2002-09-30 Ulrich Weigand <uweigand@de.ibm.com>
2510 * longlong.h: Partially synchronize with GMP-4.1 version:
2511 Use i370 definitions also for s390.
2512 Add generic definition of umul_ppmm in terms of smul_ppmm.
2513 [s390] (umul_ppmm): Remove.
2514 [s390] (smul_ppmm): Fix incorrect assembler constraints.
2515 [s390] (smul_ppmm, sdiv_qrnnd): Rename __xx to __x.
2517 2002-09-30 Bob Wilson <bob.wilson@acm.org>
2519 * config/xtensa/xtensa.h (REG_CLASS_NAMES, REG_CLASS_CONTENTS):
2520 Add new RL_REGS register class.
2521 (PREFERRED_RELOAD_CLASS, PREFERRED_OUTPUT_RELOAD_CLASS):
2522 Call xtensa_preferred_reload_class for both input and output reloads.
2523 * config/xtensa/xtensa.c (xtensa_regno_to_class): Use new RL_REGS class.
2524 (xtensa_preferred_reload_class): Handle output reloads; use RL_REGS
2525 instead of either AR_REGS or GR_REGS classes.
2526 (xtensa_secondary_reload_class): Use new RL_REGS class.
2527 * config/xtensa/xtensa-protos.h (xtensa_preferred_reload_class): Update.
2529 2002-09-30 John David Anglin <dave@hiauly1.hia.nrc.ca>
2531 * pa.c (hppa_encode_label): Don't drop '*' from function labels.
2532 (pa_strip_name_encoding): Strip '@' and '*', in that order.
2533 * pa.h (ASM_OUTPUT_LABELREF): Output user_label_prefix except when
2534 there is a '*' prefix in NAME.
2536 Mon Sep 30 21:33:23 CEST 2002 Jan Hubicka <jh@suse.cz>
2538 * reload.c (push_reload): Handle subregs and secondary memory.
2539 * reload1.c (gen_reload): Likewise.
2541 * jump.c (reg_or_subregno): New function.
2542 * rtl.h (reg_or_subregno): Declare
2543 * unroll.c (find_splittable_givs): Handle subregs.
2545 2002-09-30 Mark Mitchell <mark@codesourcery.com>
2547 * store-layout.c (finish_record_layout): Add free_p parameter.
2548 (layout_type): Pass it.
2549 * tree.h (finish_record_layout): Update prototype.
2551 Mon Sep 30 14:57:18 CEST 2002 Jan Hubicka <jh@suse.cz>
2553 * i386.h (TARGET_CPP_CPU_BUILTINS): Define __SSE_MATH__.
2555 * gcse.c (cprop_jump): Check that the register has not
2557 (cprop_jump): Likewise.
2559 2002-09-30 Richard Earnshaw <rearnsha@arm.com>
2561 * arm.h (BASE_REG_CLASS): Always return LO_REGS for Thumb.
2562 (MODE_BASE_REG_CLASS, case Thumb): Only return BASE_REGS if we know
2563 that we have a SImode access, and only then if reload hasn't completed;
2564 for all other cases, use LO_REGS.
2566 2002-09-29 Richard Henderson <rth@redhat.com>
2568 * real.c (real_from_string): Apply sign last. Tidy exponent handling.
2570 2002-09-29 Richard Henderson <rth@redhat.com>
2573 * combine.c (force_to_mode): Handle FLOAT_MODE destinations
2576 2002-09-29 David Edelsohn <edelsohn@gnu.org>
2578 * real.h (ibm_extended_format): Declare.
2579 * real.c (encode_ibm_extended, decode_ibm_extended): New
2582 2002-09-29 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2584 * darwin-protos.h (darwin_asm_output_dwarf_delta): Prototype.
2586 * ia64.c (ia64_hpux_asm_file_end): Const-ify.
2588 2002-09-29 John David Anglin <dave@hiauly1.hia.nrc.ca>
2590 * expmed.c (extract_bit_field): Fix bit-field extraction from SUBREGs.
2592 2002-09-29 Kazu Hirata <kazu@cs.umass.edu>
2594 * builtins.def: Fix comment formatting.
2595 * c-common.def: Likewise.
2596 * cfgcleanup.c: Likewise.
2597 * combine.c: Likewise.
2598 * gengtype.c: Likewise.
2599 * params.def: Likewise.
2600 * predict.def: Likewise.
2601 * rtl.def: Likewise.
2602 * stab.def: Likewise.
2603 * stor-layout.c: Likewise.
2604 * tree.def: Likewise.
2605 * config/darwin.c: Likewise.
2606 * config/darwin.h: Likewise.
2607 * config/dbxcoff.h: Likewise.
2608 * config/elfos.h: Likewise.
2609 * config/fp-bit.c: Likewise.
2610 * config/freebsd-spec.h: Likewise.
2611 * config/interix.h: Likewise.
2612 * config/libgloss.h: Likewise.
2613 * config/linux-aout.h: Likewise.
2614 * config/linux.h: Likewise.
2615 * config/lynx-ng.h: Likewise.
2616 * config/lynx.h: Likewise.
2617 * config/netbsd-aout.h: Likewise.
2618 * config/netbsd.h: Likewise.
2619 * config/netware.h: Likewise.
2620 * config/psos.h: Likewise.
2621 * config/ptx4.h: Likewise.
2623 2002-09-28 Kazu Hirata <kazu@cs.umass.edu>
2625 * ChangeLog.4: Fix typos.
2626 * ChangeLog.6: Likewise.
2627 * FSFChangeLog.10: Likewise.
2628 * genattrtab.c: Fix comment typos.
2629 * haifa-sched.c: Likewise.
2632 * config/arm/arm.c: Likewise.
2633 * config/arm/crti.asm: Likewise.
2634 * config/arm/crtn.asm: Likewise.
2635 * config/frv/frv.c: Likewise.
2636 * config/frv/frv.md: Likewise.
2637 * config/h8300/h8300.md: Likewise.
2638 * config/i386/rtemself.h: Likewise.
2639 * config/ia64/unwind-ia64.c: Likewise.
2640 * config/ip2k/ip2k.h: Likewise.
2641 * config/m88k/m88k.c: Likewise.
2642 * config/m88k/m88k.md: Likewise.
2643 * config/mips/sr71k.md: Likewise.
2644 * config/mmix/mmix.c: Likewise.
2645 * config/rs6000/rs6000.c: Likewise.
2646 * config/sh/sh.md: Likewise.
2648 2002-09-26 Theodore A. Roth <troth@verinet.com>
2650 * config/avr/avr.c: Eliminate use of _PC_ in pc relative insns.
2651 * config/avr/avr.md: Ditto.
2653 2002-09-27 Alexander N. Kabaev <ak03@gte.com>
2655 PR preprocessor/8055
2656 * cppmacro.c (stringify_arg): Do not overflow the buffer
2657 with the terminating NUL when the argument to be stringified
2660 2002-09-27 Richard Henderson <rth@redhat.com>
2662 * unroll.c (simplify_cmp_and_jump_insns): New.
2663 (unroll_loop): Use it. Use simplify_gen_foo+force_operand
2664 instead of expand_simple_foo.
2666 2002-09-27 Richard Henderson <rth@redhat.com>
2668 PR optimization/7520
2669 * cfganal.c (flow_active_insn_p): New.
2670 (forwarder_block_p): Use it.
2672 2002-09-27 Richard Henderson <rth@redhat.com>
2674 * emit-rtl.c (active_insn_p): Revert last change.
2676 2002-09-27 Jakub Jelinek <jakub@redhat.com>
2678 * doc/extend.texi (tls_model): Document.
2679 * varasm.c (decl_tls_model): New.
2680 * c-common.c (handle_tls_model_attribute): New.
2681 (c_common_attribute_table): Add tls_model.
2682 * config/alpha/alpha.c (alpha_encode_section_info): Use
2684 * flags.h (enum tls_model, flag_tls_default): Move...
2685 * tree.h (enum tls_model, flag_tls_default): ...here.
2686 (decl_tls_model): New prototype.
2687 * config/ia64/ia64.c (ia64_encode_section_info): Likewise.
2688 * config/i386/i386.c (ix86_encode_section_info): Likewise.
2689 * config/i386/i386.md (tls_global_dynamic, tls_local_dynamic_base):
2692 2002-09-27 Kazu Hirata <kazu@cs.umass.edu>
2694 * LANGUAGES: Follow spelling conventions.
2695 * rtl.def: Likewise.
2696 * sbitmap.c: Likewise.
2697 * sched-int.h: Likewise.
2698 * sched-rgn.c: Likewise.
2699 * sibcall.c: Likewise.
2700 * simplify-rtx.c: Likewise.
2702 * stab.def: Likewise.
2704 * stor-layout.c: Likewise.
2705 * target.h: Likewise.
2706 * timevar.c: Likewise.
2707 * toplev.c: Likewise.
2708 * tree-dump.c: Likewise.
2709 * tree-inline.c: Likewise.
2711 * tree.def: Likewise.
2713 * unroll.c: Likewise.
2714 * varasm.c: Likewise.
2715 * vmsdbgout.c: Likewise.
2716 * treelang/treelang.texi: Likewise.
2717 * treelang/treetree.c: Likewise.
2719 2002-09-27 Kazu Hirata <kazu@cs.umass.edu>
2721 * config/h8300/h8300.c (compute_saved_regs): Use a macro
2722 instead of a hard register number.
2723 (get_shift_alg): Use an enumerated type instead of numbers.
2724 (h8300_shift_needs_scratch_p): Likewise.
2726 2002-09-26 David S. Miller <davem@redhat.com>
2728 PR optimization/7335
2729 * calls.c (emit_library_call_value_1): Passing args by reference
2730 converts a CONST function into a PURE one.
2732 2002-09-26 David Edelsohn <edelsohn@gnu.org>
2734 * dbxout.c (FORCE_TEXT): Switch to current_function_decl, not
2736 * xcoffout.h (DBX_STATIC_BLOCK_START): Remove explicit change to
2738 * config/rs6000/rs6000.c (rs6000_override_options): Allow
2739 function-sections and data-sections functionality on AIX.
2741 2002-09-26 David Edelsohn <edelsohn@gnu.org>
2742 Dale Johannesen <dalej@apple.com>
2744 * config/rs6000/rs6000.c (rs6000_emit_move): Insert zero-extend
2745 in RTL for sub-word loads from memory.
2747 2002-09-26 Richard Henderson <rth@redhat.com>
2750 * sched-deps.c (sched_analyze_insn): Make clobber insns depend
2753 2002-09-26 Richard Henderson <rth@redhat.com>
2755 * emit-rtl.c (const_double_htab_eq): Remove unused variable.
2757 2002-09-26 Chris Lattner <sabre@nondot.org>
2759 * ssa.c (rename_insn_1): Handle RENAME_NO_RTX correctly when
2760 handling undefined values.
2762 2002-09-26 Richard Henderson <rth@redhat.com>
2765 * emit-rtl.c (active_insn_p): Consider a clobber of the
2766 function return value to be active even after reload.
2768 2002-09-27 Alan Modra <amodra@bigpond.net.au>
2770 * doloop.c (doloop_modify_runtime <biv skips initial incr>): Adjust
2771 by absolute loop increment, not loop increment.
2773 2002-09-26 Kazu Hirata <kazu@cs.umass.edu>
2775 * c-common.h: Follow spelling conventions.
2776 * cpplex.c: Likewise.
2777 * cpplib.h: Likewise.
2778 * gthr-dce.h: Likewise.
2779 * gthr-posix.h: Likewise.
2780 * optabs.c: Likewise.
2781 * output.h: Likewise.
2782 * profile.c: Likewise.
2783 * protoize.c: Likewise.
2784 * ra-rewrite.c: Likewise.
2786 * recog.c: Likewise.
2787 * reg-stack.c: Likewise.
2788 * regclass.c: Likewise.
2789 * regmove.c: Likewise.
2790 * reload.c: Likewise.
2791 * reload.h: Likewise.
2792 * reload1.c: Likewise.
2793 * reorg.c: Likewise.
2794 * resource.c: Likewise.
2796 * rtlanal.c: Likewise.
2798 2002-09-26 Steve Ellcey <sje@cup.hp.com>
2800 * config/ia64/ia64.c (ia64_expand_load_address): Ensure correct mode
2803 2002-09-24 Eric Christopher <echristo@redhat.com>
2805 * config/mips/elf.h: Add HANDLE_SYSV_PRAGMA.
2806 * config/mips/elf64.h: Ditto.
2808 2002-09-24 Eric Christopher <echristo@redhat.com>
2810 * except.c (expand_builtin_extract_return_address): Handle case
2811 where Pmode != ptr_mode.
2813 2002-09-26 Steve Ellcey <sje@cup.hp.com>
2815 * config/ia64/hpux.h (ASM_OUTPUT_EXTERNAL_LIBCALL): New
2817 2002-09-26 Steve Ellcey <sje@cup.hp.com>
2819 * config/ia64/hpux.h (TARGET_DEFAULT): Include TARGET_ILP32.
2821 2002-09-26 Igor Shevlyakov <igor@microunity.com>
2823 * combine.c (simplify_set): Don't call to force_to_mode if size
2824 of integer type is larger than HOST_BITS_PER_WIDE_INT.
2826 2002-09-26 Janis Johnson <janis187@us.ibm.com>
2828 * Makefile.in (qmtest-g++): Fix file path.
2830 2002-09-26 Ulrich Weigand <uweigand@de.ibm.com>
2832 * expr.c (expand_expr) [MINUS_EXPR]: Convert A - const to
2833 A + (-const) on RTX level, even for unsigned types.
2835 2002-09-26 Ulrich Weigand <uweigand@de.ibm.com>
2837 * reload.c (dup_replacements): New function.
2838 (find_reloads): Use it to duplicate replacements at the top level
2839 of match_dup operands.
2841 2002-09-26 Miles Bader <miles@gnu.org>
2843 * v850.md ("length"): Change default value to 4.
2845 2002-09-26 Kazu Hirata <kazu@cs.umass.edu>
2847 * ChangeLog.1: Follow spelling conventions.
2848 * ChangeLog.4: Likewise.
2849 * ChangeLog.6: Likewise.
2850 * FSFChangeLog.11: Likewise.
2851 * doc/cpp.texi: Likewise.
2852 * doc/invoke.texi: Likewise.
2853 * doc/tm.texi: Likewise.
2855 2002-09-26 Nick Clifton <nickc@redhat.com>
2857 * config.gcc: Add x prefix to v850e case for handling
2860 2002-09-25 David S. Miller <davem@redhat.com>
2863 * config/sparc/sparc.c (set_extends): SImode ASHIFT does not
2866 2002-09-25 Richard Henderson <rth@redhat.com>
2868 * emit-rtl.c (const_double_htab_eq): Distinguish integer and
2869 fp CONST_DOUBLE; use real_identical.
2871 2002-09-25 Mark Mitchell <mark@codesourcery.com>
2873 * doc/invoke.texi: Add more -Wabi examples.
2875 2002-09-25 Richard Sandiford <rsandifo@redhat.com>
2877 * config/mips/mips.h (TARGET_MIPS4100): Add missing bracket.
2879 2002-09-24 Denis Chertykov <denisc@overta.ru>
2881 * config/ip2k/ip2k.c (function_epilogue): Fix wrong numbers in
2882 cases of optimizing "add sp,w" to "inc sp".
2884 2002-09-24 Adam Nemet <anemet@lnxw.com>
2886 * config/arm/arm.c (thumb_unexpanded_epilogue): Don't generate
2887 epilogue for naked functions.
2889 2002-09-24 Adam Nemet <anemet@lnxw.com>
2890 Nick Clifton <nickc@redhat.com>
2892 * config/arm/arm.h (THUMB_FUNCTION_PROFILER): Remove.
2893 (FUNCTION_PROFILER): Only invoke THUMB_FUNCTION_PROFILER if it
2896 2002-09-24 Ulrich Weigand <uweigand@de.ibm.com>
2898 * config/s390/s390.c (preferred_la_operand_p): New function.
2899 * config/s390/s390-protos.h (preferred_la_operand_p): Declare it.
2900 * config/s390/s390.md ("addaddr_esame", "*la_ccclobber"): Replace by ...
2901 ("*la_64_cc", "*la_31_cc", splitters): ... these.
2902 ("*la_31"): Deactivate for TARGET_64BIT.
2903 ("*la_31_and", "*la_31_and_cc"): New.
2905 2002-09-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2907 * real.h (real_value): Make `exp' explicitly signed.
2909 2002-09-24 Kazu Hirata <kazu@cs.umass.edu>
2911 * config/elfos.h: Follow spelling conventions.
2912 * config/alpha/alpha.h: Likewise.
2913 * config/arc/arc.h: Likewise.
2914 * config/arm/arm.md: Likewise.
2915 * config/avr/avr.h: Likewise.
2916 * config/cris/cris.md: Likewise.
2917 * config/d30v/d30v.h: Likewise.
2918 * config/frv/frv.c: Likewise.
2919 * config/frv/frv.h: Likewise.
2920 * config/h8300/h8300.c: Likewise.
2921 * config/h8300/h8300.h: Likewise.
2922 * config/h8300/h8300.md: Likewise.
2923 * config/i386/cygwin.h: Likewise.
2924 * config/i386/i386.h: Likewise.
2925 * config/i386/sysv3.h: Likewise.
2926 * config/i960/i960.h: Likewise.
2927 * config/ia64/ia64.h: Likewise.
2928 * config/ia64/ia64.md: Likewise.
2929 * config/ip2k/ip2k.h: Likewise.
2930 * config/m32r/m32r.h: Likewise.
2931 * config/m68k/m68k.h: Likewise.
2932 * config/m88k/m88k.h: Likewise.
2933 * config/mcore/mcore.c: Likewise.
2934 * config/mcore/mcore.h: Likewise.
2935 * config/mcore/mcore.md: Likewise.
2936 * config/mips/mips.h: Likewise.
2937 * config/mmix/mmix.h: Likewise.
2938 * config/mmix/mmix.md: Likewise.
2939 * config/ns32k/netbsd.h: Likewise.
2940 * config/ns32k/ns32k.h: Likewise.
2941 * config/ns32k/ns32k.md: Likewise.
2942 * config/pa/pa.h: Likewise.
2943 * config/romp/romp.h: Likewise.
2944 * config/rs6000/rs6000.h: Likewise.
2945 * config/rs6000/rs6000.md: Likewise.
2946 * config/sparc/sparc.h: Likewise.
2947 * config/stormy16/stormy-abi: Likewise.
2948 * config/stormy16/stormy16.h: Likewise.
2949 * config/vax/vax.h: Likewise.
2951 2002-09-23 Zack Weinberg <zack@codesourcery.com>
2953 * version.c (version_string): Now const char[].
2954 * version.h: Update to match.
2956 2002-09-23 Richard Henderson <rth@redhat.com>
2958 * config/i386/i386.h (MASK_ACCUMULATE_OUTGOING_ARGS_SET, MASK_MMX_SET,
2959 MASK_SSE_SET, MASK_SSE2_SET, MASK_3DNOW_SET, MASK_3DNOW_A_SET): Kill.
2960 (TARGET_SWITCHES): Don't reference them.
2961 * config/i386/i386.c (override_options): Use target_flags_explicit
2962 to examine bits set by the user.
2964 2002-09-23 Dale Johannesen <dalej@apple.com>
2966 * dbxout.c (dbxout_parms): Set current_sym_code for params
2967 passed on stack by invisible reference.
2969 2002-09-23 Richard Earnshaw <rearnsha@arm.com>
2971 * arm/unknown-elf.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Always allocate
2972 at least one byte of space.
2974 2002-09-23 Mark Mitchell <mark@codesourcery.com>
2976 * c-common.h (flag_abi_version): Fix typo in comment.
2977 * doc/invoke.texi (flag_abi_version): Document default value.
2979 2002-09-23 Hans-Peter Nilsson <hp@axis.com>
2981 * doc/extend.texi (Extended Asm): Clarify that overlap between
2982 asm-declared register variables used in an asm and the asm clobber
2983 list is not allowed.
2984 * stmt.c (decl_conflicts_with_clobbers_p): New function.
2985 (expand_asm_operands): Keep track of clobbered registers. Call
2986 decl_conflicts_with_clobbers_p for each input and output operand.
2987 If no conflicts found before, also do conflict sanity check when
2990 2002-09-23 Richard Henderson <rth@redhat.com>
2992 * c-common.c (cpp_define_data_format): Remove.
2993 (cb_register_builtins): Don't define __WCHAR_BIT__, __SHRT_BIT__,
2994 __INT_BIT__, __LONG_BIT__, __LONG_LONG_BIT__, __FLOAT_BIT__,
2995 __DOUBLE_BIT__, __LONG_DOUBLE_BIT__.
2996 * doc/cpp.texi: Don't document them either.
2997 (__SCHAR_MAX__, __SHRT_MAX__, __INT_MAX__, __LONG_MAX__,
2998 __LONG_LONG_MAX__): Document.
2999 (__TARGET_FLOAT_FORMAT__): Remove.
3001 2002-09-23 Richard Henderson <rth@redhat.com>
3003 * real.c (do_multiply): Normalize U before addition.
3005 2002-09-23 Mark Mitchell <mark@codesourcery.com>
3007 * c-common.c (flag_abi_version): New variable.
3008 * c-common.h (flag_abi_version): Declare it.
3009 * c-opts.c (missing_arg): Add -fabi-version.
3010 (c_common_decode_option): Process -fabi-version.
3011 * doc/invoke.texi (-fabi-version): Document it.
3012 (-Wabi): Add information about bit-fields in unions.
3014 2002-09-22 Jason Thorpe <thorpej@wasabisystems.com>
3016 * config/mips/netbsd.h (SUBTARGET_ASM_SPEC): Always pass -KPIC
3017 unless -fno-pic or -fno-PIC is specified.
3019 2002-09-22 John David Anglin <dave@hiauly1.hia.nrc.ca>
3021 * c-common.c (preprocessing_trad_p): Define.
3022 * pa-hiux.h, pa-hpux.h, pa-hpux7.h (CPP_PREDEFINES): Delete.
3023 (TARGET_OS_CPP_BUILTINS, SUBTARGET_SWITCHES): Define.
3024 * pa-hpux10.h (TARGET_OS_CPP_BUILTINS, CPP_SPEC): Define.
3025 * pa-hpux11.h (TARGET_OS_CPP_BUILTINS): Define.
3026 * pa-linux.h (CPP_PREDEFINES): Delete.
3027 (TARGET_OS_CPP_BUILTINS, CPP_SPEC): Define.
3028 * pa32-linux.h, pa64-linux.h (CPP_SPEC): Delete.
3029 * pa-osf.h, pa-pro-end.h, rtems.h (CPP_PREDEFINES): Delete.
3030 (TARGET_OS_CPP_BUILTINS): Define.
3031 * pa.h (MASK_SIO, TARGET_SIO, TARGET_PA_10): Define.
3032 (TARGET_SWITCHES): Reformat. Use N_() macro. Add SUBTARGET_SWITCHES.
3033 (SUBTARGET_SWITCHES): Provide default definition.
3034 (TARGET_OPTIONS): Reformat. Use N_() macro.
3035 (CPP_PA10_SPEC, CPP_PA11_SPEC, CPP_PA20_SPEC, CPP_64BIT_SPEC,
3036 CPP_CPU_DEFAULT_SPEC, CPP_64BIT_DEFAULT_SPEC, SUBTARGET_EXTRA_SPECS,
3037 EXTRA_SPECS, CPP_SPEC, CPLUSPLUS_CPP_SPEC, CPP_PREDEFINES): Delete.
3038 (TARGET_CPU_CPP_BUILTINS): Define.
3039 (TARGET_OS_CPP_BUILTINS): Define for BSD-like systems.
3040 * doc/invoke.texi (msio, mwsio): Document new hppa options.
3041 * doc/tm.texi (TARGET_CPU_CPP_BUILTINS): Document macro
3042 preprocessing_trad_p().
3044 2002-09-22 Jason Thorpe <thorpej@wasabisystems.com>
3046 * doc/install.texi: Document behavior of --with-headers and
3047 --with-libs when arguments are omitted.
3049 2002-09-22 Kazu Hirata <kazu@cs.umass.edu>
3051 * dbxout.c: Follow spelling conventions.
3052 * defaults.h: Likewise.
3054 * diagnostic.h: Likewise.
3055 * doloop.c: Likewise.
3056 * dwarf2out.c: Likewise.
3057 * dwarfout.c: Likewise.
3058 * emit-rtl.c: Likewise.
3059 * except.c: Likewise.
3060 * explow.c: Likewise.
3061 * expmed.c: Likewise.
3064 * flags.h: Likewise.
3066 * fold-const.c: Likewise.
3067 * function.c: Likewise.
3068 * function.h: Likewise.
3070 * gcov-io.h: Likewise.
3073 * genattrtab.c: Likewise.
3074 * genconfig.c: Likewise.
3075 * genrecog.c: Likewise.
3076 * ggc-page.c: Likewise.
3078 * global.c: Likewise.
3079 * gthr-win32.h: Likewise.
3080 * integrate.c: Likewise.
3082 * langhooks.c: Likewise.
3083 * langhooks.h: Likewise.
3084 * line-map.h: Likewise.
3085 * local-alloc.c: Likewise.
3086 * longlong.h: Likewise.
3090 Tue Aug 27 22:26:35 CEST 2002 Jan Hubicka <jh@suse.cz>
3092 * i386.h (BIGGEST_FIELD_ALIGNMENT): Set proper default for x86_64.
3094 Tue Aug 27 20:07:01 CEST 2002 Jan Hubicka <jh@suse.cz>
3096 * i386.c (overwrite_options): Set -mpreferred-stack-boundary to 128
3097 for -Os/TARGET_64BIT too.
3099 2002-09-21 Kazu Hirata <kazu@cs.umass.edu>
3101 * ChangeLog: Follow spelling conventions.
3102 * ChangeLog.0: Likewise.
3103 * ChangeLog.1: Likewise.
3104 * ChangeLog.2: Likewise.
3105 * ChangeLog.3: Likewise.
3106 * ChangeLog.4: Likewise.
3107 * ChangeLog.5: Likewise.
3108 * ChangeLog.6: Likewise.
3109 * FSFChangeLog.10: Likewise.
3110 * FSFChangeLog.11: Likewise.
3111 * alias.c: Likewise.
3112 * basic-block.h: Likewise.
3113 * c-aux-info.c: Likewise.
3114 * c-common.c: Likewise.
3115 * c-common.h: Likewise.
3116 * c-decl.c: Likewise.
3117 * c-format.c: Likewise.
3118 * c-semantics.c: Likewise.
3119 * c-typeck.c: Likewise.
3120 * calls.c: Likewise.
3121 * cfganal.c: Likewise.
3122 * cfgloop.c: Likewise.
3123 * collect2.c: Likewise.
3124 * combine.c: Likewise.
3125 * conflict.c: Likewise.
3126 * cppexp.c: Likewise.
3127 * cppfiles.c: Likewise.
3128 * cpphash.h: Likewise.
3129 * cppinit.c: Likewise.
3130 * cpplex.c: Likewise.
3131 * cpplib.c: Likewise.
3132 * cpplib.h: Likewise.
3133 * cppmacro.c: Likewise.
3136 2002-09-21 Richard Earnshaw <rearnsha@arm.com>
3138 * netbsd-aout.h (NETBSD_LINK_SPEC_AOUT): New, takes old definition of
3140 (LINK_SPEC): Define to NETBSD_LINK_SPEC_AOUT.
3141 * arm/netbsd.h (SUBTARGET_EXTRA_SEPCS): Add NETBSD_LINK_SPEC_AOUT.
3142 (LINK_SPEC): Rework to use NETBSD_LINK_SPEC_AOUT).
3144 2002-09-21 Richard Earnshaw <rearnsha@arm.com>
3147 * cse.c (fold_rtx): Calculate old_cost before we fold each
3150 2002-09-21 Richard Henderson <rth@redhat.com>
3152 * c-common.c (cpp_define_data_format): Remove __GCC_LITTLE_ENDIAN__,
3153 __GCC_BIG_ENDIAN__, __TARGET_BITS_ORDER__, __TARGET_BYTES_ORDER__,
3154 __TARGET_INT_WORDS_ORDER__, __TARGET_FLOAT_WORDS_ORDER__,
3155 __TARGET_USES_VAX_F_FLOAT__, __TARGET_USES_VAX_D_FLOAT__,
3156 __TARGET_USES_VAX_G_FLOAT__, __TARGET_USES_VAX_H_FLOAT__.
3157 * doc/cpp.texi: Don't document them.
3159 2002-09-21 Richard Henderson <rth@redhat.com>
3161 * c-common.c (builtin_define_float_constants): Use real_format
3162 to get the floating-point parameters.
3164 2002-09-21 Richard Henderson <rth@redhat.com>
3166 * real.c (struct real_format): Move to real.h.
3167 (real_format_for_mode): Rename from fmt_for_mode; update all users;
3168 initialize with ieee defaults.
3169 (real_to_target_fmt, real_from_target_fmt): New.
3170 (ieee_single_format, ieee_double_format, ieee_extended_motorola_format,
3171 ieee_extended_intel_96_format, ieee_extended_intel_128_format,
3172 ieee_quad_format, i370_single_format, i370_double_format,
3173 c4x_single_format, c4x_extended_format): Rename from s/_format//.
3174 (ieee_quad_format): Fix emin.
3175 (format_for_size, init_real_once): Remove.
3176 * real.h (struct real_format): Move from real.c.
3177 (real_format_for_mode): Declare.
3178 (real_to_target_fmt, real_from_target_fmt): Declare.
3179 (ieee_single_format, ieee_double_format, ieee_extended_motorola_format,
3180 ieee_extended_intel_96_format, ieee_extended_intel_128_format,
3181 ieee_quad_format, vax_f_format, vax_d_format, vax_g_format,
3182 i370_single_format, i370_double_format, c4x_single_format,
3183 c4x_extended_format): Declare.
3184 * toplev.c (do_compile): Don't call init_real_once.
3186 * defaults.h (INTEL_EXTENDED_IEEE_FORMAT): Remove.
3187 * doc/tm.texi (INTEL_EXTENDED_IEEE_FORMAT): Remove.
3189 * config/alpha/alpha.h (TARGET_FLOAT_FORMAT): Define.
3190 * config/alpha/osf5.h (LONG_DOUBLE_TYPE_SIZE): 64, if vax mode.
3191 * config/alpha/alpha.c (override_options): Set real_format_for_mode
3192 for VAX, if enabled.
3194 * config/c4x/c4x.c (c4x_override_options): Set real_format_for_mode
3197 * config/i370/i370.h (OVERRIDE_OPTIONS): New.
3198 * config/i370/i370.c (override_options): New.
3199 * config/i370/i370-protos.h: Update.
3201 * config/i386/i386.c (override_options): Set real_format_for_mode
3202 for Intel 80-bit extended.
3203 * config/i386/i386.h (INTEL_EXTENDED_IEEE_FORMAT): Remove.
3205 * config/i960/i960.h (LONG_DOUBLE_TYPE_SIZE): Mind -mlong-double-64.
3206 (OVERRIDE_OPTIONS): Move code...
3207 * config/i960/i960.c (i960_initialize): ... here. Set
3208 real_format_for_mode for Intel 80-bit extended.
3210 * config/ia64/ia64.c (ia64_override_options): Set real_format_for_mode
3211 for Intel 80-bit extended, if enabled.
3213 * config/m68k/m68k.c (override_options): Set real_format_for_mode
3214 for Motorola 96-bit extended.
3216 * config/vax/vax.h (OVERRIDE_OPTIONS): New.
3217 * config/vax/vax.c (override_options): New.
3218 * config/vax/vax-protos.h: Update.
3220 2002-09-21 Alan Modra <amodra@bigpond.net.au>
3222 * config/rs6000/rs6000.md (builtin_setjmp_receiver): Add
3225 * config/rs6000/rs6000.md (floatdisf2_internal2): Combine
3226 insns. Supply missing clobber of scratch reg.
3228 2002-09-20 Kazu Hirata <kazu@cs.umass.edu>
3230 * config/m32r/m32r.c: Follow spelling conventions.
3231 * config/m32r/m32r.h: Likewise.
3232 * config/m32r/m32r.md: Likewise.
3233 * config/m68k/m68k.c: Likewise.
3234 * config/m88k/m88k.c: Likewise.
3235 * config/mcore/mcore.c: Likewise.
3236 * config/mips/mips.c: Likewise.
3237 * config/mips/mips.h: Likewise.
3238 * config/mmix/mmix.c: Likewise.
3239 * config/mn10200/mn10200.c: Likewise.
3240 * config/ns32k/ns32k.h: Likewise.
3241 * config/pa/pa.c: Likewise.
3242 * config/pa/pa64-linux.h: Likewise.
3243 * config/pdp11/pdp11.h: Likewise.
3244 * config/romp/romp.c: Likewise.
3245 * config/romp/romp.h: Likewise.
3246 * config/rs6000/eabi.asm: Likewise.
3247 * config/rs6000/linux64.h: Likewise.
3248 * config/rs6000/rs6000.c: Likewise.
3249 * config/rs6000/rs6000.h: Likewise.
3250 * config/rs6000/rs6000.md: Likewise.
3251 * config/rs6000/sysv4.h: Likewise.
3252 * config/rs6000/xcoff.h: Likewise.
3254 2002-09-20 Jim Wilson <wilson@redhat.com>
3256 * config/v850/v850/lib1funcs.asm (__muldi3): Change r5 to r28.
3258 2002-09-20 Jakub Jelinek <jakub@redhat.com>
3260 * config/i386/i386.md (UNSPEC_GOTNTPOFF, UNSPEC_INDNTPOFF): New.
3261 * config/i386/i386.c (legitimate_pic_address_disp_p): Handle
3262 UNSPEC_GOTNTPOFF and UNSPEC_INDNTPOFF like UNSPEC_GOTTPOFF.
3263 (legitimate_address_p): Likewise.
3264 (legitimize_address): Use @gotntpoff and @indntpoff.
3265 (output_pic_addr_const): Handle UNSPEC_GOTNTPOFF and UNSPEC_INDNTPOFF.
3266 (output_addr_const_extra): Likewise.
3268 2002-09-20 Jim Wilson <wilson@redhat.com>
3270 * combine.c (try_combine): When split an instruction pair, where the
3271 first has a sign_extend src, verify that the src and dest modes match.
3273 2002-09-20 Richard Henderson <rth@redhat.com>
3275 * config/mips/mips.c (dfhigh, dflow, sfhigh, sflow): Remove.
3276 (override_options): Do not initialize them.
3277 (mips_const_double_ok): Allow no fp constants except zero,
3278 and not even that for mips16.
3279 (const_float_1_operand): Use dconst1.
3280 * config/mips/mips.md (movsf, movsf_internal1, movsf_internal2,
3281 movdf, movdf_internal1, movdf_internal1a, movdf_internal2):
3282 Don't allow arbitrary constants; fix predicates and C constraint.
3284 2002-09-20 Neil Booth <neil@daikokuya.co.uk>
3286 * cppmacro.c: Don't warn about function-like macros without
3287 '(' during pre-expansion.
3289 2002-09-20 Jim Wilson <wilson@redhat.com>
3291 * config/v850/v850.c (current_function_anonymous_args): Delete.
3292 (expand_prologue): Use current_function_args_info.anonymous_args.
3293 (expand_epilogue): Delete use of current_function_anonymous_args.
3294 * config/v850/v850.h (struct cum_arg): Add anonymous_args field.
3295 (INIT_CUMULATIVE_ARGS): Clear anonymous_args field.
3296 (current_function_anonymous_args): Delete extern declaration.
3297 (SETUP_INCOMING_VARARGS): Set anonymous_args field.
3299 2002-09-20 Geoffrey Keating <geoffk@apple.com>
3301 * config/rs6000/rs6000.c (rs6000_emit_prologue): Update for change
3302 to load_macho_picbase.
3303 * config/rs6000/rs6000.md: Document Darwin-specific unspec IDs.
3304 (load_macho_picbase): Take the symbol to use as a parameter.
3305 (macho_correct_pic): New insn.
3306 (builtin_setjmp_reciever): On Darwin, restore the PIC register.
3308 * config/rs6000/rs6000.h (ELIMINABLE_REGS): Use
3309 RS6000_PIC_OFFSET_TABLE_REGNUM rather than hardcoding 30.
3310 (CAN_ELIMINATE): Likewise.
3311 (INITIAL_ELIMINATION_OFFSET): Likewise.
3312 (TOC_REGISTER): Likewise.
3314 2002-09-20 Richard Henderson <rth@redhat.com>
3316 * real.c (real_hash): New.
3317 * real.h: Declare it.
3318 * cse.c (canon_hash): Use it.
3319 * cselib.c (hash_rtx): Likewise.
3320 * emit-rtl.c (const_double_htab_hash): Likewise.
3321 * rtl.h (CONST_DOUBLE_REAL_VALUE): New.
3322 * varasm.c (struct rtx_const): Reduce vector size; separate
3323 integer and fp vectors.
3325 (const_hash_1): Rename from const_hash. Use real_hash. Do not
3326 take modulus MAX_HASH_TABLE.
3327 (const_hash): New. Do take modulus MAX_HASH_TABLE.
3328 (output_constant_def): Do not take modulus MAX_HASH_TABLE.
3329 (SYMHASH): Don't use HASHBITS.
3330 (decode_rtx_const): Copy only active bits from REAL_VALUE_TYPE.
3331 Fix CONST_VECTOR thinko wrt fp vectors. Fix kind comparison.
3332 (simplify_subtraction): Fix kind comparison.
3333 (const_hash_rtx): Return unsigned int. Don't use HASHBITS.
3334 Use a union to pun integer array.
3335 * config/rs6000/rs6000.c (rs6000_hash_constant): Use real_hash;
3336 only hash two words of integral CONST_DOUBLE.
3338 2002-09-20 Steve Ellcey <sje@cup.hp.com>
3340 * config/ia64/hpux.h (STARTFILE_SPEC): Modify.
3341 (STARTFILE_PREFIX_SPEC): New.
3342 (LINK_SPEC): Modify.
3346 2002-09-20 Jakub Jelinek <jakub@redhat.com>
3348 * config/i386/i386.c (legitimate_pic_address_disp_p): Allow
3349 UNSPEC_NTPOFF and UNSPEC_DTPOFF to be offsetted by constant.
3351 2002-09-20 Jeroen Dobbelaere <jeroen.dobbelaere@acunia.com>
3353 * config/arm/arm.md (sign_extract_onebit, not_signextract_onebit):
3354 Add clobber of the condition code register.
3356 2002-09-20 Richard Henderson <rth@redhat.com>
3358 * real.c (do_fix_trunc): Static.
3359 (encode_ieee_single, encode_ieee_double, encode_ieee_extended,
3360 encode_ieee_quad, encode_vax_f, encode_vax_d, encode_vax_g,
3361 encode_i370_single, encode_i370_double, encode_c4x_single,
3362 encode_c4x_extended): Add default abort case.
3364 2002-09-20 Richard Henderson <rth@redhat.com>
3366 * real.h (enum real_value_class, SIGNIFICAND_BITS, EXP_BITS,
3367 MAX_EXP, SIGSZ, SIG_MSB, struct real_value): Move from real.c.
3368 (struct realvaluetype): Remove.
3369 (REAL_VALUE_TYPE): Use struct real_value.
3370 (REAL_VALUE_TYPE_SIZE): Use SIGNIFICAND_BITS.
3371 (test_real_width): New.
3372 * real.c: Global replace struct real_value with REAL_VALUE_TYPE.
3373 (real_arithmetic): Avoid hoops for REAL_VALUE_TYPE parameters.
3374 (real_compare, real_exponent, real_ldexp, real_isinf, real_isnan,
3375 real_isneg, real_isnegzero, real_identical, exact_real_inverse,
3376 real_to_integer, real_to_integer2, real_to_decimal,
3377 real_to_hexadecimal, real_from_string, real_from_integer,
3378 real_inf, real_nan, real_2expN, real_convert, real_to_target,
3379 real_from_target): Likewise.
3380 * tree.h (struct tree_real_cst): Use real_value not realvaluetype.
3381 * gengtype-yacc.y (bitfieldopt): Accept an ID as well.
3383 2002-09-20 Richard Henderson <rth@redhat.com>
3385 * real.h (UNKNOWN_FLOAT_FORMAT, IEEE_FLOAT_FORMAT, VAX_FLOAT_FORMAT,
3386 IBM_FLOAT_FORMAT, C4X_FLOAT_FORMAT, TARGET_FLOAT_FORMAT): Move ...
3387 * defaults.h: ... here.
3388 * config/arm/arm.h, config/avr/avr.h, config/d30v/d30v.h,
3389 config/fr30/fr30.h, config/frv/frv.h, config/ia64/ia64.h,
3390 config/ip2k/ip2k.h, config/mips/mips.h, config/stormy16/stormy16.h,
3391 config/xtensa/xtensa.h (TARGET_FLOAT_FORMAT): Remove.
3393 2002-09-20 Hans-Peter Nilsson <hp@bitrange.com>
3395 * config/mmix/mmix.md ("negdf2"): Rewrite.
3396 ("*expanded_negdf2"): New.
3398 2002-09-19 Jim Wilson <wilson@redhat.com>
3400 * combine.c (simplify_set): When optimizing a subreg src with a
3401 cc0 dest, use GET_MODE (src) for mask instead of inner_mode.
3403 2002-09-19 Dale Johannesen <dalej@apple.com>
3404 * combine.c (make_extraction): Don't create
3407 2002-09-19 Ulrich Weigand <uweigand@de.ibm.com>
3409 * config/s390/s390.c (addr_generation_dependency_p): Handle SUBREG
3410 and STRICT_LOW_PART within SET_DEST.
3411 * config/s390/s390.md ("*extractqi", "*extracthi"): New insns with
3412 splitters, replacing pre-reload splitters.
3413 ("*zero_extendhisi2_31", "*zero_extendqisi2_31",
3414 "*zero_extendqihi2_31"): New insns.
3415 ("*zero_extendqihi2_64"): Do not clobber CC.
3417 2002-09-19 Steve Ellcey <sje@cup.hp.com>
3419 * config/ia64/hpux.h (CTORS_SECTION_ASM_OP): New.
3420 (DTORS_SECTION_ASM_OP): Ditto.
3421 (READONLY_DATA_SECTION_ASM_OP): Moved.
3422 (DATA_SECTION_ASM_OP): New.
3423 (SDATA_SECTION_ASM_OP): New.
3424 (BSS_SECTION_ASM_OP): New.
3425 (SBSS_SECTION_ASM_OP): New.
3426 (TEXT_SECTION_ASM_OP): New.
3428 2002-09-19 Kazu Hirata <kazu@cs.umass.edu>
3430 * config/fp-bit.c: Follow spelling conventions.
3431 * config/d30v/d30v.c: Likewise.
3432 * config/d30v/d30v.h: Likewise.
3433 * config/fr30/fr30.c: Likewise.
3434 * config/fr30/fr30.h: Likewise.
3435 * config/fr30/fr30.md: Likewise.
3436 * config/frv/frv.c: Likewise.
3437 * config/frv/frv.h: Likewise.
3438 * config/h8300/h8300.c: Likewise.
3439 * config/h8300/lib1funcs.asm: Likewise.
3440 * config/i370/i370.c: Likewise.
3441 * config/i386/i386.h: Likewise.
3442 * config/i386/i386.md: Likewise.
3443 * config/i386/pentium.md: Likewise.
3444 * config/i386/winnt.c: Likewise.
3445 * config/i960/i960.c: Likewise.
3446 * config/ia64/ia64.h: Likewise.
3447 * config/ip2k/ip2k.c: Likewise.
3448 * config/ip2k/ip2k.h: Likewise.
3449 * config/ip2k/ip2k.md: Likewise.
3450 * config/ip2k/libgcc.S: Likewise.
3452 2002-09-19 Stephen Clarke <stephen.clarke@superh.com>
3454 * config/sh/sh.h (UNSPEC_GOTOFF_P): Define.
3455 (GOTOFF_P): Extend to allow gotoff plus constant.
3457 2002-09-18 Richard Henderson <rth@redhat.com>
3459 * ifcvt.c (noce_process_if_block): Correctly detect X modified
3460 with INSN_B before COND_EARLIEST. Don't check A and B for
3461 modification in condition range. Reorder INSN_B for A==B properly.
3462 (if_convert): Iterate until no matches for a block.
3464 2002-09-18 Richard Henderson <rth@redhat.com>
3466 * calls.c (store_one_arg): Rename default_align to parm_align;
3467 always adjust parm_align for downward padding.
3469 2002-09-18 Richard Henderson <rth@redhat.com>
3471 * toplev.c (backend_init): Move init_real_once invocation ...
3472 (do_compile): ... here.
3474 2002-09-18 Richard Henderson <rth@redhat.com>
3476 * sibcall.c (optimize_sibling_and_tail_recursive_call): Also remove
3477 RTX_UNCHANGING_P markers for successful tail-recursive replacement.
3479 2002-09-18 Richard Henderson <rth@redhat.com>
3481 * real.c (round_for_format): Collect sticky as unsigned long, not bool.
3483 2002-09-19 Alan Modra <amodra@bigpond.net.au>
3485 * config/rs6000/rs6000.md: (floatdisf2): Rename to
3486 floatdisf2_internal1.
3487 (floatdisf2): New define_expand.
3488 (floatdisf2_internal2): Likewise.
3490 2002-09-18 Richard Henderson <rth@redhat.com>
3492 * real.c (sticky_rshift_significand): Collect sticky as
3493 unsigned long, not bool.
3495 2002-09-18 Ulrich Weigand <uweigand@de.ibm.com>
3497 * config/s390/s390.c (s390_address_cost): New function.
3498 config/s390/s390-protos.h (s390_address_cost): Add prototype.
3499 config/s390/s390.h (ADDRESS_COST): Call s390_address_cost.
3500 (RTX_COST): Use COSTS_N_INSNS.
3502 2002-09-18 Douglas Rupp <rupp@gnat.com>
3503 Donn Terry <donnte@microsoft.com>
3505 * stor-layout.c (place_field): Handle alignment of whole
3506 structures when MSVC compatible bitfields are involved.
3507 Change method of computing location of MS bitfields to
3508 be compatible with #pragma pack(n).
3510 * tree.h (record_layout_info): Add&nbs
OSDN Git Service
