- 20 Feb, 1996 9 commits
-
-
* sparc/sparc.md (cpu attr): Add all cpu variants. (negtf2,negdf2,abstf2,absdf2): Use isa attr, not arch attr in determining insn lengths. From-SVN: r11321
Doug Evans committed -
* sparc/sparc.h (TARGET_SWITCHES): Add -m{,no-}impure-text. (MASK_IMPURE_TEXT, TARGET_IMPURE_TEXT): Define. (LINK_SPEC): Only add `-assert pure-text' if -mimpure-text wasn't used. * sparc/sparc.h (TARGET_CPU_sparc{,let,lite,64}): Define. ({CPP,ASM}_DEFAULT_SPEC): Set from TARGET_CPU_foo. (SPARC_ARCH64 CPP_PREDEFINES): Define __arch64__. (CPP_SPEC): Add %(cpp_cpu). (CPP_CPU_SPEC): Define. (CC1_SPEC): Convert -m<cpu> to -mcpu=<cpu>. (ASM_SPEC): Add %(asm_cpu). (ASM_CPU_SPEC): Define. (EXTRA_SPECS,SUBTARGET_EXTRA_SPECS): Define. (OVERRIDE_OPTIONS): Call SUBTARGET_OVERRIDE_OPTIONS after sparc_override_options. ({MASK,TARGET}_SUPERSPARC): Delete. ({MASK,TARGET}_SPARCLET): Define. (MASK_ISA): Renamed from MASK_CPUS. ({MASK,TARGET}_FRW): Delete. (TARGET_SWITCHES): Delete no-{v8,sparclite}. (sparc_cpu,sparc_cpu_string): Declare. ({SUB,}TARGET_OPTIONS): Define. (FIXED_REGISTERS): Add definitions for sparc64 in 32 bit mode. (CONDITIONAL_REGISTER_USAGE): Don't set fixed_regs[234] if sparc64. Don't set call_used_regs[48..80] for sparc64 in 32 bit mode. Don't clobber fixed_regs[234] if -ffixed- was passed. (FRAME_POINTER_REQUIRED,INITIAL_FRAME_POINTER_OFFSET, BASE_{INCOMING_ARG,OUTGOING_VALUE}_REG,INCOMING_REGNO,OUTGOING_REGNO, FUNCTION_{PROLOGUE,EPILOGUE},DELAY_SLOTS_FOR_EPILOGUE): TARGET_FRW renamed to TARGET_FLAT. (ADJUST_COST): Change test for supersparc. From-SVN: r11320
Doug Evans committed -
From-SVN: r11319
Doug Evans committed -
* sparc/sparc.c (sparc_cpu_string,sparc_cpu): New globals. (sparc_override_options): Set ISA and CPU from sparc_cpu_string. Delete tests for v9 switches if not v9. From-SVN: r11318
Doug Evans committed -
* sparc/sp64-elf.h (TARGET_DEFAULT): Add MASK_APP_REGS. (CPP_PREDEFINES): Define __arch64__. From-SVN: r11317
Doug Evans committed -
* sparc/sp64-aout.h: #include aoutos.h. (TARGET_DEFAULT): Add MASK_APP_REGS. (JUMP_TABLES_IN_TEXT_SECTION,READONLY_DATA_SECTION): Delete. From-SVN: r11316
Doug Evans committed -
* sparc/sol2.h (CPP_SPEC): Use %(cpp_cpu). (ASM_SPEC): Likewise. (ASM_{DEFAULT,CPU}_SPEC): Use Solaris syntax for sparc64. From-SVN: r11315
Doug Evans committed -
* sparc/lite.h: #include aoutos.h. (TARGET_DEFAULT): Use MASK_FOO values. From-SVN: r11314
Doug Evans committed -
From-SVN: r11313
Richard Kenner committed
-
- 19 Feb, 1996 20 commits
-
-
From-SVN: r11312
Doug Evans committed -
* sparc/t-sparclite (MULTILIB_*): Delete msoft-float and mno-flat, they're the defaults. Add -mcpu=f934 as synonym for -mfpu. From-SVN: r11311
Doug Evans committed -
From-SVN: r11310
David Edelsohn committed -
support DImode const in TOC From-SVN: r11309
David Edelsohn committed -
From-SVN: r11308
David Edelsohn committed -
From-SVN: r11307
Doug Evans committed -
(SELECT_CC_MODE): return CC_Zmode if the operand is QImode. Allow LT and GE comparisons in CC_NOOVmode. (PREDICATE_CODES): add offsettable_memory_operand and alignable_memory_operand. From-SVN: r11306
Richard Earnshaw committed -
(*zeroextract[qs]i_compare0_scratch): Use const_int_operand for operands 1 and 2. (split patterns for aligned memory half-word operations): New patterns. (movhi): Handle memory accesses where the alignment is known in a more efficient manner. (*compareqi_eq0): Use CC_Zmode. From-SVN: r11305
Richard Earnshaw committed -
(alignable_memory_operand): New function. (gen_rotated_half_load): New function. (get_arm_condition_code): Extract the mode of the comparison and use it to generate the correct return value. From-SVN: r11304
Richard Earnshaw committed -
From-SVN: r11303
Michael Meissner committed -
From-SVN: r11302
Michael Meissner committed -
From-SVN: r11301
Richard Kenner committed -
the appropriate subword of extended precision CONST_DOUBLEs. From-SVN: r11300
Richard Earnshaw committed -
From-SVN: r11299
Jason Merrill committed -
From-SVN: r11298
Jason Merrill committed -
(yylex, case '0'..'9','.'): For cases '0' and '1', check for single digit constant before resorting to general number processing. From-SVN: r11297
Richard Kenner committed -
(ftruncdf2, ftruncsf2, muldf3, mulsidi3): Likewise. (smulsi3_highpart, umulsi3_highpart, umulsidi3): Likewise. From-SVN: r11296
Richard Kenner committed -
(TARGET_SWITCHES): Add -m68060. From-SVN: r11295
Richard Kenner committed -
From-SVN: r11294
Richard Kenner committed -
From-SVN: r11293
Richard Kenner committed
-
- 18 Feb, 1996 2 commits
-
-
From-SVN: r11292
Richard Kenner committed -
From-SVN: r11291
Richard Kenner committed
-
- 16 Feb, 1996 9 commits
-
-
* c-typeck.c (convert_for_assignment): Bring back conversion to union without a cast, undoing the Jan 16 change, but with the following differences: * The union must have the transparent_union attribute. * The conversion must be for a function argument. * Warn consistently about such conversions if pedantic. * Do not warn about an assignment incompatibility for one union member if another union member is compatible with no warning. From-SVN: r11290
Paul Eggert committed -
From-SVN: r11289
Stan Cox committed -
From-SVN: r11288
Stan Cox committed -
From-SVN: r11287
Richard Kenner committed -
(WCHAR_TYPE, WCHAR_TYPE_SIZE): Use short unsigned int. From-SVN: r11286
Richard Kenner committed -
From-SVN: r11285
Richard Kenner committed -
From-SVN: r11284
Richard Kenner committed -
From-SVN: r11283
Richard Kenner committed -
Don't include a29k.h or dbxcoff.h. From-SVN: r11282
Richard Kenner committed
-