| 1 |
234 |
jeremybenn |
|
| 2 |
|
|
/*============================================================================
|
| 3 |
|
|
|
| 4 |
|
|
This C header file is part of the SoftFloat IEC/IEEE Floating-point Arithmetic
|
| 5 |
|
|
Package, Release 2b.
|
| 6 |
|
|
|
| 7 |
|
|
Written by John R. Hauser. This work was made possible in part by the
|
| 8 |
|
|
International Computer Science Institute, located at Suite 600, 1947 Center
|
| 9 |
|
|
Street, Berkeley, California 94704. Funding was partially provided by the
|
| 10 |
|
|
National Science Foundation under grant MIP-9311980. The original version
|
| 11 |
|
|
of this code was written as part of a project to build a fixed-point vector
|
| 12 |
|
|
processor in collaboration with the University of California at Berkeley,
|
| 13 |
|
|
overseen by Profs. Nelson Morgan and John Wawrzynek. More information
|
| 14 |
|
|
is available through the Web page `http://www.cs.berkeley.edu/~jhauser/
|
| 15 |
|
|
arithmetic/SoftFloat.html'.
|
| 16 |
|
|
|
| 17 |
|
|
THIS SOFTWARE IS DISTRIBUTED AS IS, FOR FREE. Although reasonable effort has
|
| 18 |
|
|
been made to avoid it, THIS SOFTWARE MAY CONTAIN FAULTS THAT WILL AT TIMES
|
| 19 |
|
|
RESULT IN INCORRECT BEHAVIOR. USE OF THIS SOFTWARE IS RESTRICTED TO PERSONS
|
| 20 |
|
|
AND ORGANIZATIONS WHO CAN AND WILL TAKE FULL RESPONSIBILITY FOR ALL LOSSES,
|
| 21 |
|
|
COSTS, OR OTHER PROBLEMS THEY INCUR DUE TO THE SOFTWARE, AND WHO FURTHERMORE
|
| 22 |
|
|
EFFECTIVELY INDEMNIFY JOHN HAUSER AND THE INTERNATIONAL COMPUTER SCIENCE
|
| 23 |
|
|
INSTITUTE (possibly via similar legal warning) AGAINST ALL LOSSES, COSTS, OR
|
| 24 |
|
|
OTHER PROBLEMS INCURRED BY THEIR CUSTOMERS AND CLIENTS DUE TO THE SOFTWARE.
|
| 25 |
|
|
|
| 26 |
|
|
Derivative works are acceptable, even for commercial purposes, so long as
|
| 27 |
|
|
(1) the source code for the derivative work includes prominent notice that
|
| 28 |
|
|
the work is derivative, and (2) the source code includes prominent notice with
|
| 29 |
|
|
these four paragraphs for those parts of this code that are retained.
|
| 30 |
|
|
|
| 31 |
|
|
=============================================================================*/
|
| 32 |
|
|
|
| 33 |
|
|
/*----------------------------------------------------------------------------
|
| 34 |
|
|
| The macro `FLOATX80' must be defined to enable the extended double-precision
|
| 35 |
|
|
| floating-point format `floatx80'. If this macro is not defined, the
|
| 36 |
|
|
| `floatx80' type will not be defined, and none of the functions that either
|
| 37 |
|
|
| input or output the `floatx80' type will be defined. The same applies to
|
| 38 |
|
|
| the `FLOAT128' macro and the quadruple-precision format `float128'.
|
| 39 |
|
|
*----------------------------------------------------------------------------*/
|
| 40 |
|
|
// OR1K's floating point currently only needs to simulate 32-bit floating point
|
| 41 |
|
|
// ops, so even the 64-bit stuff is overkill, but could be handy in future...
|
| 42 |
|
|
//#define FLOATX80
|
| 43 |
|
|
//#define FLOAT128
|
| 44 |
|
|
|
| 45 |
|
|
/*----------------------------------------------------------------------------
|
| 46 |
|
|
| Software IEC/IEEE floating-point types.
|
| 47 |
|
|
*----------------------------------------------------------------------------*/
|
| 48 |
|
|
typedef unsigned int float32;
|
| 49 |
|
|
typedef unsigned long long float64;
|
| 50 |
|
|
#ifdef FLOATX80
|
| 51 |
|
|
typedef struct {
|
| 52 |
|
|
unsigned long long low;
|
| 53 |
|
|
unsigned short high;
|
| 54 |
|
|
} floatx80;
|
| 55 |
|
|
#endif
|
| 56 |
|
|
#ifdef FLOAT128
|
| 57 |
|
|
typedef struct {
|
| 58 |
|
|
unsigned long long low, high;
|
| 59 |
|
|
} float128;
|
| 60 |
|
|
#endif
|
| 61 |
|
|
|
| 62 |
|
|
/*----------------------------------------------------------------------------
|
| 63 |
|
|
| Software IEC/IEEE floating-point underflow tininess-detection mode.
|
| 64 |
|
|
*----------------------------------------------------------------------------*/
|
| 65 |
|
|
extern signed char float_detect_tininess;
|
| 66 |
|
|
enum {
|
| 67 |
|
|
float_tininess_after_rounding = 0,
|
| 68 |
|
|
float_tininess_before_rounding = 1
|
| 69 |
|
|
};
|
| 70 |
|
|
|
| 71 |
|
|
/*----------------------------------------------------------------------------
|
| 72 |
|
|
| Software IEC/IEEE floating-point rounding mode.
|
| 73 |
|
|
*----------------------------------------------------------------------------*/
|
| 74 |
|
|
extern signed char float_rounding_mode;
|
| 75 |
|
|
enum {
|
| 76 |
|
|
float_round_nearest_even = 0,
|
| 77 |
|
|
float_round_down = 1,
|
| 78 |
|
|
float_round_up = 2,
|
| 79 |
|
|
float_round_to_zero = 3
|
| 80 |
|
|
};
|
| 81 |
|
|
|
| 82 |
|
|
/*----------------------------------------------------------------------------
|
| 83 |
|
|
| Software IEC/IEEE floating-point exception flags.
|
| 84 |
|
|
*----------------------------------------------------------------------------*/
|
| 85 |
|
|
extern signed char float_exception_flags;
|
| 86 |
|
|
enum {
|
| 87 |
|
|
float_flag_invalid = 1,
|
| 88 |
|
|
float_flag_divbyzero = 4,
|
| 89 |
|
|
float_flag_overflow = 8,
|
| 90 |
|
|
float_flag_underflow = 16,
|
| 91 |
|
|
float_flag_inexact = 32
|
| 92 |
|
|
};
|
| 93 |
|
|
|
| 94 |
|
|
/*----------------------------------------------------------------------------
|
| 95 |
|
|
| Routine to raise any or all of the software IEC/IEEE floating-point
|
| 96 |
|
|
| exception flags.
|
| 97 |
|
|
*----------------------------------------------------------------------------*/
|
| 98 |
|
|
void float_raise( signed char );
|
| 99 |
|
|
|
| 100 |
|
|
/*----------------------------------------------------------------------------
|
| 101 |
|
|
| Software IEC/IEEE integer-to-floating-point conversion routines.
|
| 102 |
|
|
*----------------------------------------------------------------------------*/
|
| 103 |
|
|
float32 int32_to_float32( int );
|
| 104 |
|
|
float64 int32_to_float64( int );
|
| 105 |
|
|
#ifdef FLOATX80
|
| 106 |
|
|
floatx80 int32_to_floatx80( int );
|
| 107 |
|
|
#endif
|
| 108 |
|
|
#ifdef FLOAT128
|
| 109 |
|
|
float128 int32_to_float128( int );
|
| 110 |
|
|
#endif
|
| 111 |
|
|
float32 int64_to_float32( long long );
|
| 112 |
|
|
float64 int64_to_float64( long long );
|
| 113 |
|
|
#ifdef FLOATX80
|
| 114 |
|
|
floatx80 int64_to_floatx80( long long );
|
| 115 |
|
|
#endif
|
| 116 |
|
|
#ifdef FLOAT128
|
| 117 |
|
|
float128 int64_to_float128( long long );
|
| 118 |
|
|
#endif
|
| 119 |
|
|
|
| 120 |
|
|
/*----------------------------------------------------------------------------
|
| 121 |
|
|
| Software IEC/IEEE single-precision conversion routines.
|
| 122 |
|
|
*----------------------------------------------------------------------------*/
|
| 123 |
|
|
int float32_to_int32( float32 );
|
| 124 |
|
|
int float32_to_int32_round_to_zero( float32 );
|
| 125 |
|
|
long long float32_to_int64( float32 );
|
| 126 |
|
|
long long float32_to_int64_round_to_zero( float32 );
|
| 127 |
|
|
float64 float32_to_float64( float32 );
|
| 128 |
|
|
#ifdef FLOATX80
|
| 129 |
|
|
floatx80 float32_to_floatx80( float32 );
|
| 130 |
|
|
#endif
|
| 131 |
|
|
#ifdef FLOAT128
|
| 132 |
|
|
float128 float32_to_float128( float32 );
|
| 133 |
|
|
#endif
|
| 134 |
|
|
|
| 135 |
|
|
/*----------------------------------------------------------------------------
|
| 136 |
|
|
| Software IEC/IEEE single-precision operations.
|
| 137 |
|
|
*----------------------------------------------------------------------------*/
|
| 138 |
|
|
float32 float32_round_to_int( float32 );
|
| 139 |
|
|
float32 float32_add( float32, float32 );
|
| 140 |
|
|
float32 float32_sub( float32, float32 );
|
| 141 |
|
|
float32 float32_mul( float32, float32 );
|
| 142 |
|
|
float32 float32_div( float32, float32 );
|
| 143 |
|
|
float32 float32_rem( float32, float32 );
|
| 144 |
|
|
float32 float32_sqrt( float32 );
|
| 145 |
|
|
char float32_eq( float32, float32 );
|
| 146 |
|
|
char float32_le( float32, float32 );
|
| 147 |
|
|
char float32_lt( float32, float32 );
|
| 148 |
|
|
char float32_eq_signaling( float32, float32 );
|
| 149 |
|
|
char float32_le_quiet( float32, float32 );
|
| 150 |
|
|
char float32_lt_quiet( float32, float32 );
|
| 151 |
|
|
char float32_is_signaling_nan( float32 );
|
| 152 |
|
|
char float32_is_nan( float32 );
|
| 153 |
|
|
|
| 154 |
|
|
/*----------------------------------------------------------------------------
|
| 155 |
|
|
| Software IEC/IEEE double-precision conversion routines.
|
| 156 |
|
|
*----------------------------------------------------------------------------*/
|
| 157 |
|
|
int float64_to_int32( float64 );
|
| 158 |
|
|
int float64_to_int32_round_to_zero( float64 );
|
| 159 |
|
|
long long float64_to_int64( float64 );
|
| 160 |
|
|
long long float64_to_int64_round_to_zero( float64 );
|
| 161 |
|
|
float32 float64_to_float32( float64 );
|
| 162 |
|
|
#ifdef FLOATX80
|
| 163 |
|
|
floatx80 float64_to_floatx80( float64 );
|
| 164 |
|
|
#endif
|
| 165 |
|
|
#ifdef FLOAT128
|
| 166 |
|
|
float128 float64_to_float128( float64 );
|
| 167 |
|
|
#endif
|
| 168 |
|
|
|
| 169 |
|
|
/*----------------------------------------------------------------------------
|
| 170 |
|
|
| Software IEC/IEEE double-precision operations.
|
| 171 |
|
|
*----------------------------------------------------------------------------*/
|
| 172 |
|
|
float64 float64_round_to_int( float64 );
|
| 173 |
|
|
float64 float64_add( float64, float64 );
|
| 174 |
|
|
float64 float64_sub( float64, float64 );
|
| 175 |
|
|
float64 float64_mul( float64, float64 );
|
| 176 |
|
|
float64 float64_div( float64, float64 );
|
| 177 |
|
|
float64 float64_rem( float64, float64 );
|
| 178 |
|
|
float64 float64_sqrt( float64 );
|
| 179 |
|
|
char float64_eq( float64, float64 );
|
| 180 |
|
|
char float64_le( float64, float64 );
|
| 181 |
|
|
char float64_lt( float64, float64 );
|
| 182 |
|
|
char float64_eq_signaling( float64, float64 );
|
| 183 |
|
|
char float64_le_quiet( float64, float64 );
|
| 184 |
|
|
char float64_lt_quiet( float64, float64 );
|
| 185 |
|
|
char float64_is_signaling_nan( float64 );
|
| 186 |
|
|
|
| 187 |
|
|
#ifdef FLOATX80
|
| 188 |
|
|
|
| 189 |
|
|
/*----------------------------------------------------------------------------
|
| 190 |
|
|
| Software IEC/IEEE extended double-precision conversion routines.
|
| 191 |
|
|
*----------------------------------------------------------------------------*/
|
| 192 |
|
|
int floatx80_to_int32( floatx80 );
|
| 193 |
|
|
int floatx80_to_int32_round_to_zero( floatx80 );
|
| 194 |
|
|
long long floatx80_to_int64( floatx80 );
|
| 195 |
|
|
long long floatx80_to_int64_round_to_zero( floatx80 );
|
| 196 |
|
|
float32 floatx80_to_float32( floatx80 );
|
| 197 |
|
|
float64 floatx80_to_float64( floatx80 );
|
| 198 |
|
|
#ifdef FLOAT128
|
| 199 |
|
|
float128 floatx80_to_float128( floatx80 );
|
| 200 |
|
|
#endif
|
| 201 |
|
|
|
| 202 |
|
|
/*----------------------------------------------------------------------------
|
| 203 |
|
|
| Software IEC/IEEE extended double-precision rounding precision. Valid
|
| 204 |
|
|
| values are 32, 64, and 80.
|
| 205 |
|
|
*----------------------------------------------------------------------------*/
|
| 206 |
|
|
extern signed char floatx80_rounding_precision;
|
| 207 |
|
|
|
| 208 |
|
|
/*----------------------------------------------------------------------------
|
| 209 |
|
|
| Software IEC/IEEE extended double-precision operations.
|
| 210 |
|
|
*----------------------------------------------------------------------------*/
|
| 211 |
|
|
floatx80 floatx80_round_to_int( floatx80 );
|
| 212 |
|
|
floatx80 floatx80_add( floatx80, floatx80 );
|
| 213 |
|
|
floatx80 floatx80_sub( floatx80, floatx80 );
|
| 214 |
|
|
floatx80 floatx80_mul( floatx80, floatx80 );
|
| 215 |
|
|
floatx80 floatx80_div( floatx80, floatx80 );
|
| 216 |
|
|
floatx80 floatx80_rem( floatx80, floatx80 );
|
| 217 |
|
|
floatx80 floatx80_sqrt( floatx80 );
|
| 218 |
|
|
char floatx80_eq( floatx80, floatx80 );
|
| 219 |
|
|
char floatx80_le( floatx80, floatx80 );
|
| 220 |
|
|
char floatx80_lt( floatx80, floatx80 );
|
| 221 |
|
|
char floatx80_eq_signaling( floatx80, floatx80 );
|
| 222 |
|
|
char floatx80_le_quiet( floatx80, floatx80 );
|
| 223 |
|
|
char floatx80_lt_quiet( floatx80, floatx80 );
|
| 224 |
|
|
char floatx80_is_signaling_nan( floatx80 );
|
| 225 |
|
|
|
| 226 |
|
|
#endif
|
| 227 |
|
|
|
| 228 |
|
|
#ifdef FLOAT128
|
| 229 |
|
|
|
| 230 |
|
|
/*----------------------------------------------------------------------------
|
| 231 |
|
|
| Software IEC/IEEE quadruple-precision conversion routines.
|
| 232 |
|
|
*----------------------------------------------------------------------------*/
|
| 233 |
|
|
int float128_to_int32( float128 );
|
| 234 |
|
|
int float128_to_int32_round_to_zero( float128 );
|
| 235 |
|
|
long long float128_to_int64( float128 );
|
| 236 |
|
|
long long float128_to_int64_round_to_zero( float128 );
|
| 237 |
|
|
float32 float128_to_float32( float128 );
|
| 238 |
|
|
float64 float128_to_float64( float128 );
|
| 239 |
|
|
#ifdef FLOATX80
|
| 240 |
|
|
floatx80 float128_to_floatx80( float128 );
|
| 241 |
|
|
#endif
|
| 242 |
|
|
|
| 243 |
|
|
/*----------------------------------------------------------------------------
|
| 244 |
|
|
| Software IEC/IEEE quadruple-precision operations.
|
| 245 |
|
|
*----------------------------------------------------------------------------*/
|
| 246 |
|
|
float128 float128_round_to_int( float128 );
|
| 247 |
|
|
float128 float128_add( float128, float128 );
|
| 248 |
|
|
float128 float128_sub( float128, float128 );
|
| 249 |
|
|
float128 float128_mul( float128, float128 );
|
| 250 |
|
|
float128 float128_div( float128, float128 );
|
| 251 |
|
|
float128 float128_rem( float128, float128 );
|
| 252 |
|
|
float128 float128_sqrt( float128 );
|
| 253 |
|
|
char float128_eq( float128, float128 );
|
| 254 |
|
|
char float128_le( float128, float128 );
|
| 255 |
|
|
char float128_lt( float128, float128 );
|
| 256 |
|
|
char float128_eq_signaling( float128, float128 );
|
| 257 |
|
|
char float128_le_quiet( float128, float128 );
|
| 258 |
|
|
char float128_lt_quiet( float128, float128 );
|
| 259 |
|
|
char float128_is_signaling_nan( float128 );
|
| 260 |
|
|
|
| 261 |
|
|
#endif
|
| 262 |
|
|
|