- 31 Mar, 1993 40 commits
-
-
(truncxfdf2, truncxfsf2, floatsixf2, floathixf2, floatqixf2): Likewise. (ftruncxf2, fixxfqi2, fixxfhi2, fixxfsi2): Likewise. (addxf3, subxf3, mulxf3, divxf3, negxf2, absxf2, sqrtxf2): Likewise. From-SVN: r3973
Richard Stallman committed -
(ASM_OUTPUT_LONG_DOUBLE_OPERAND) New macro. (ASM_OUTPUT_FLOAT, ASM_OUTPUT_DOUBLE): Use REAL_VALUE_... macros. (ASM_OUTPUT_DOUBLE_OPERAND): Likewise. (ASM_OUTPUT_FLOAT_OPERAND): Likewise. Also, new argument CODE. (ASM_OUTPUT_FLOAT): Delete special def for CROSS_COMPILE. (ASM_OUTPUT_LONG_DOUBLE): New macro. (PRINT_OPERAND_EXTRACT_FLOAT): Definitions deleted. (PRINT_OPERAND_PRINT_FLOAT): Definitions deleted. From-SVN: r3972
Richard Stallman committed -
(standard_68881_constant_p): Initialize REAL_VALUE_TYPE binary constants from decimal values. Use REAL_VALUE_... macros. (standard_sun_fpa_constant_p): Likewise. (print_operand, floating_exact_log2): Use REAL_VALUE_... macros. (init_68881_table): New function. From-SVN: r3971
Richard Stallman committed -
used in call-by-reference situations. From-SVN: r3970
Richard Kenner committed -
(ASM_OUTPUT_DOUBLE, ASM_OUTPUT_FLOAT, PRINT_OPERAND_FLOAT) (PRINT_OPERAND): Use new REAL_VALUE... macros. (ASM_OUTPUT_LONG_DOUBLE): New macro. From-SVN: r3969
Richard Stallman committed -
Make new DECL_UID and TYPE_UID as comments say. From-SVN: r3968
Richard Kenner committed -
Use new REAL_VALUE... macros. From-SVN: r3967
Richard Stallman committed -
memcpy/bcopy is SIZETYPE, not Pmode. From-SVN: r3966
Richard Kenner committed -
Use new REAL_VALUE... macros. (ASM_OUTPUT_FLOAT_OPERAND, ASM_OUTPUT_DOUBLE_OPERAND): Likewise. From-SVN: r3965
Richard Stallman committed -
Use new REAL_VALUE... macros. (ASM_OUTPUT_LONG_DOUBLE): New definition. From-SVN: r3964
Richard Stallman committed -
(simplify_binary_operation, case MULT): When testing for floating-point equality, make sure we do so inside a region protected from traps. From-SVN: r3963
Richard Kenner committed -
make sure we do so inside a region protected from traps. From-SVN: r3962
Richard Kenner committed -
current_function_pretend_args_size from OFFSET for AP. From-SVN: r3961
Richard Kenner committed -
Use new REAL_VALUE... macros. (REAL_ARITHMETIC): Defined. From-SVN: r3960
Richard Stallman committed -
From-SVN: r3959
Richard Stallman committed -
From-SVN: r3958
Richard Stallman committed -
Use new REAL_VALUE... macros. (CHECK_FLOAT_VALUE): Call check_float_value. From-SVN: r3957
Richard Stallman committed -
From-SVN: r3956
Jeff Law committed -
From-SVN: r3955
Brendan Kehoe committed -
From-SVN: r3954
Brendan Kehoe committed -
From-SVN: r3953
James Van Artsdalen committed -
From-SVN: r3952
Richard Kenner committed -
want, don't do anything to the classes. From-SVN: r3951
Richard Kenner committed -
From-SVN: r3950
Jeff Law committed -
* pa.c (output_cbranch): New function. Given the operands of the branch, possible nullification, length of the branch, possible branch negation, and the branch insn itself, return the output template for the desired conditional branch. (output_bb): Likewise, but for branch on bit. From-SVN: r3949
Jeff Law committed -
From-SVN: r3948
Richard Stallman committed -
From-SVN: r3947
Richard Stallman committed -
From-SVN: r3946
Richard Stallman committed -
From-SVN: r3945
Richard Stallman committed -
From-SVN: r3944
Richard Stallman committed -
(simplify_binary_operation): real_value_truncate returns REAL_VALUE_TYPE. Convert rtx_code into tree_code for argument to REAL_ARITHMETIC. (simplify_unary_operation): Use new ..._RNDZINT macros instead of casting value of REAL_VALUE_FIX_TRUNCATE and REAL_VALUE_UNSIGNED_FIX_TRUNCATE. From-SVN: r3943
Richard Stallman committed -
From-SVN: r3942
Richard Stallman committed -
From-SVN: r3941
Richard Stallman committed -
From-SVN: r3940
Richard Stallman committed -
(real_value_truncate): If REAL_ARITHMETIC, replace by routine of same name in real.c. target_isinf, target_isnan, target_negative): Likewise. (const_binop): Delete `register' from decls using REAL_VALUE_TYPE. (fold_convert): Use REAL_VALUE_TO_INT. From-SVN: r3939
Richard Stallman committed -
If REAL_ARITHMETIC is defined, let real.c handle incompatible host and target floating point bit patterns. (init_emit_once): Add machine_mode argument to REAL_VALUE_ATOF. From-SVN: r3938
Richard Stallman committed -
from float mode CONST_DOUBLEs. From-SVN: r3937
Richard Stallman committed -
(yylex): Convert real decimal constants directly to the precision specified by the letter at the end of the number. (yylex): Convert real decimal constants directly to the precision specified by the letter at the end of the number. Pass mode arg to REAL_VALUE_ATOF to specify precision. Move the "out of range of double" error check. From-SVN: r3936
Richard Stallman committed -
(OBJS): Add real.o. From-SVN: r3935
Richard Stallman committed -
If REAL_ARITHMETIC is defined or if LONG_DOUBLE_TYPE_SIZE is 96, then define REAL_VALUE_ macros to invoke functions in real.c. (REAL_VALUE_RNDZINT, REAL_VALUE_UNSIGNED_RNDZINT): New macros truncate toward zero to integer value but return REAL_VALUE_TYPE. (REAL_VALUE_TO_DECIMAL): New macro defaults to fprintf if no REAL_ARITHMETIC, otherwise uses real.c for binary to decimal conversion. Used in ASM_OUTPUT_ macros. From-SVN: r3934
Richard Stallman committed
-