1 /*============================================================================
3 This C header file is part of the SoftFloat IEC/IEEE Floating-point Arithmetic
6 Written by John R. Hauser. This work was made possible in part by the
7 International Computer Science Institute, located at Suite 600, 1947 Center
8 Street, Berkeley, California 94704. Funding was partially provided by the
9 National Science Foundation under grant MIP-9311980. The original version
10 of this code was written as part of a project to build a fixed-point vector
11 processor in collaboration with the University of California at Berkeley,
12 overseen by Profs. Nelson Morgan and John Wawrzynek. More information
13 is available through the Web page `http://www.cs.berkeley.edu/~jhauser/
14 arithmetic/SoftFloat.html'.
16 THIS SOFTWARE IS DISTRIBUTED AS IS, FOR FREE. Although reasonable effort has
17 been made to avoid it, THIS SOFTWARE MAY CONTAIN FAULTS THAT WILL AT TIMES
18 RESULT IN INCORRECT BEHAVIOR. USE OF THIS SOFTWARE IS RESTRICTED TO PERSONS
19 AND ORGANIZATIONS WHO CAN AND WILL TAKE FULL RESPONSIBILITY FOR ALL LOSSES,
20 COSTS, OR OTHER PROBLEMS THEY INCUR DUE TO THE SOFTWARE, AND WHO FURTHERMORE
21 EFFECTIVELY INDEMNIFY JOHN HAUSER AND THE INTERNATIONAL COMPUTER SCIENCE
22 INSTITUTE (possibly via similar legal warning) AGAINST ALL LOSSES, COSTS, OR
23 OTHER PROBLEMS INCURRED BY THEIR CUSTOMERS AND CLIENTS DUE TO THE SOFTWARE.
25 Derivative works are acceptable, even for commercial purposes, so long as
26 (1) the source code for the derivative work includes prominent notice that
27 the work is derivative, and (2) the source code includes prominent notice with
28 these four paragraphs for those parts of this code that are retained.
30 =============================================================================*/
38 /*----------------------------------------------------------------------------
39 | Each of the following `typedef's defines the most convenient type that holds
40 | integers of at least as many bits as specified. For example, `uint8' should
41 | be the most convenient type that can hold unsigned integers of as many as
42 | 8 bits. The `flag' type must be able to hold either a 0 or 1. For most
43 | implementations of C, `flag', `uint8', and `int8' should all be `typedef'ed
44 | to the same as `int'.
45 *----------------------------------------------------------------------------*/
47 typedef uint8_t uint8;
51 typedef unsigned int uint32;
52 typedef signed int int32;
53 typedef uint64_t uint64;
54 typedef int64_t int64;
56 /*----------------------------------------------------------------------------
57 | Each of the following `typedef's defines a type that holds integers
58 | of _exactly_ the number of bits specified. For instance, for most
59 | implementation of C, `bits16' and `sbits16' should be `typedef'ed to
60 | `unsigned short int' and `signed short int' (or `short int'), respectively.
61 *----------------------------------------------------------------------------*/
62 typedef uint8_t bits8;
63 typedef int8_t sbits8;
64 typedef uint16_t bits16;
65 typedef int16_t sbits16;
66 typedef uint32_t bits32;
67 typedef int32_t sbits32;
68 typedef uint64_t bits64;
69 typedef int64_t sbits64;
71 #define LIT64( a ) a##LL
72 #define INLINE static inline
74 /*----------------------------------------------------------------------------
75 | The macro `FLOATX80' must be defined to enable the extended double-precision
76 | floating-point format `floatx80'. If this macro is not defined, the
77 | `floatx80' type will not be defined, and none of the functions that either
78 | input or output the `floatx80' type will be defined. The same applies to
79 | the `FLOAT128' macro and the quadruple-precision format `float128'.
80 *----------------------------------------------------------------------------*/
81 #ifdef CONFIG_SOFTFLOAT
82 /* bit exact soft float support */
86 /* native float support */
87 #if (defined(__i386__) || defined(__x86_64__)) && !defined(_BSD)
90 #endif /* !CONFIG_SOFTFLOAT */
92 #define STATUS_PARAM , float_status *status
93 #define STATUS(field) status->field
94 #define STATUS_VAR , status
96 /*----------------------------------------------------------------------------
97 | Software IEC/IEEE floating-point ordering relations
98 *----------------------------------------------------------------------------*/
100 float_relation_less = -1,
101 float_relation_equal = 0,
102 float_relation_greater = 1,
103 float_relation_unordered = 2
106 #ifdef CONFIG_SOFTFLOAT
107 /*----------------------------------------------------------------------------
108 | Software IEC/IEEE floating-point types.
109 *----------------------------------------------------------------------------*/
110 typedef uint32_t float32;
111 typedef uint64_t float64;
120 #ifdef WORDS_BIGENDIAN
128 /*----------------------------------------------------------------------------
129 | Software IEC/IEEE floating-point underflow tininess-detection mode.
130 *----------------------------------------------------------------------------*/
132 float_tininess_after_rounding = 0,
133 float_tininess_before_rounding = 1
136 /*----------------------------------------------------------------------------
137 | Software IEC/IEEE floating-point rounding mode.
138 *----------------------------------------------------------------------------*/
140 float_round_nearest_even = 0,
141 float_round_down = 1,
143 float_round_to_zero = 3
146 /*----------------------------------------------------------------------------
147 | Software IEC/IEEE floating-point exception flags.
148 *----------------------------------------------------------------------------*/
150 float_flag_invalid = 1,
151 float_flag_divbyzero = 4,
152 float_flag_overflow = 8,
153 float_flag_underflow = 16,
154 float_flag_inexact = 32
157 typedef struct float_status {
158 signed char float_detect_tininess;
159 signed char float_rounding_mode;
160 signed char float_exception_flags;
162 signed char floatx80_rounding_precision;
166 void set_float_rounding_mode(int val STATUS_PARAM);
167 void set_float_exception_flags(int val STATUS_PARAM);
168 INLINE int get_float_exception_flags(float_status *status)
170 return STATUS(float_exception_flags);
173 void set_floatx80_rounding_precision(int val STATUS_PARAM);
176 /*----------------------------------------------------------------------------
177 | Routine to raise any or all of the software IEC/IEEE floating-point
179 *----------------------------------------------------------------------------*/
180 void float_raise( int8 flags STATUS_PARAM);
182 /*----------------------------------------------------------------------------
183 | Software IEC/IEEE integer-to-floating-point conversion routines.
184 *----------------------------------------------------------------------------*/
185 float32 int32_to_float32( int STATUS_PARAM );
186 float64 int32_to_float64( int STATUS_PARAM );
187 float32 uint32_to_float32( unsigned int STATUS_PARAM );
188 float64 uint32_to_float64( unsigned int STATUS_PARAM );
190 floatx80 int32_to_floatx80( int STATUS_PARAM );
193 float128 int32_to_float128( int STATUS_PARAM );
195 float32 int64_to_float32( int64_t STATUS_PARAM );
196 float32 uint64_to_float32( uint64_t STATUS_PARAM );
197 float64 int64_to_float64( int64_t STATUS_PARAM );
198 float64 uint64_to_float64( uint64_t STATUS_PARAM );
200 floatx80 int64_to_floatx80( int64_t STATUS_PARAM );
203 float128 int64_to_float128( int64_t STATUS_PARAM );
206 /*----------------------------------------------------------------------------
207 | Software IEC/IEEE single-precision conversion routines.
208 *----------------------------------------------------------------------------*/
209 int float32_to_int32( float32 STATUS_PARAM );
210 int float32_to_int32_round_to_zero( float32 STATUS_PARAM );
211 unsigned int float32_to_uint32( float32 STATUS_PARAM );
212 unsigned int float32_to_uint32_round_to_zero( float32 STATUS_PARAM );
213 int64_t float32_to_int64( float32 STATUS_PARAM );
214 int64_t float32_to_int64_round_to_zero( float32 STATUS_PARAM );
215 float64 float32_to_float64( float32 STATUS_PARAM );
217 floatx80 float32_to_floatx80( float32 STATUS_PARAM );
220 float128 float32_to_float128( float32 STATUS_PARAM );
223 /*----------------------------------------------------------------------------
224 | Software IEC/IEEE single-precision operations.
225 *----------------------------------------------------------------------------*/
226 float32 float32_round_to_int( float32 STATUS_PARAM );
227 float32 float32_add( float32, float32 STATUS_PARAM );
228 float32 float32_sub( float32, float32 STATUS_PARAM );
229 float32 float32_mul( float32, float32 STATUS_PARAM );
230 float32 float32_div( float32, float32 STATUS_PARAM );
231 float32 float32_rem( float32, float32 STATUS_PARAM );
232 float32 float32_sqrt( float32 STATUS_PARAM );
233 int float32_eq( float32, float32 STATUS_PARAM );
234 int float32_le( float32, float32 STATUS_PARAM );
235 int float32_lt( float32, float32 STATUS_PARAM );
236 int float32_eq_signaling( float32, float32 STATUS_PARAM );
237 int float32_le_quiet( float32, float32 STATUS_PARAM );
238 int float32_lt_quiet( float32, float32 STATUS_PARAM );
239 int float32_compare( float32, float32 STATUS_PARAM );
240 int float32_compare_quiet( float32, float32 STATUS_PARAM );
241 int float32_is_signaling_nan( float32 );
242 int float64_is_nan( float64 a );
244 INLINE float32 float32_abs(float32 a)
246 return a & 0x7fffffff;
249 INLINE float32 float32_chs(float32 a)
251 return a ^ 0x80000000;
254 /*----------------------------------------------------------------------------
255 | Software IEC/IEEE double-precision conversion routines.
256 *----------------------------------------------------------------------------*/
257 int float64_to_int32( float64 STATUS_PARAM );
258 int float64_to_int32_round_to_zero( float64 STATUS_PARAM );
259 unsigned int float64_to_uint32( float64 STATUS_PARAM );
260 unsigned int float64_to_uint32_round_to_zero( float64 STATUS_PARAM );
261 int64_t float64_to_int64( float64 STATUS_PARAM );
262 int64_t float64_to_int64_round_to_zero( float64 STATUS_PARAM );
263 uint64_t float64_to_uint64 (float64 a STATUS_PARAM);
264 uint64_t float64_to_uint64_round_to_zero (float64 a STATUS_PARAM);
265 float32 float64_to_float32( float64 STATUS_PARAM );
267 floatx80 float64_to_floatx80( float64 STATUS_PARAM );
270 float128 float64_to_float128( float64 STATUS_PARAM );
273 /*----------------------------------------------------------------------------
274 | Software IEC/IEEE double-precision operations.
275 *----------------------------------------------------------------------------*/
276 float64 float64_round_to_int( float64 STATUS_PARAM );
277 float64 float64_trunc_to_int( float64 STATUS_PARAM );
278 float64 float64_add( float64, float64 STATUS_PARAM );
279 float64 float64_sub( float64, float64 STATUS_PARAM );
280 float64 float64_mul( float64, float64 STATUS_PARAM );
281 float64 float64_div( float64, float64 STATUS_PARAM );
282 float64 float64_rem( float64, float64 STATUS_PARAM );
283 float64 float64_sqrt( float64 STATUS_PARAM );
284 int float64_eq( float64, float64 STATUS_PARAM );
285 int float64_le( float64, float64 STATUS_PARAM );
286 int float64_lt( float64, float64 STATUS_PARAM );
287 int float64_eq_signaling( float64, float64 STATUS_PARAM );
288 int float64_le_quiet( float64, float64 STATUS_PARAM );
289 int float64_lt_quiet( float64, float64 STATUS_PARAM );
290 int float64_compare( float64, float64 STATUS_PARAM );
291 int float64_compare_quiet( float64, float64 STATUS_PARAM );
292 int float64_is_signaling_nan( float64 );
294 INLINE float64 float64_abs(float64 a)
296 return a & 0x7fffffffffffffffLL;
299 INLINE float64 float64_chs(float64 a)
301 return a ^ 0x8000000000000000LL;
306 /*----------------------------------------------------------------------------
307 | Software IEC/IEEE extended double-precision conversion routines.
308 *----------------------------------------------------------------------------*/
309 int floatx80_to_int32( floatx80 STATUS_PARAM );
310 int floatx80_to_int32_round_to_zero( floatx80 STATUS_PARAM );
311 int64_t floatx80_to_int64( floatx80 STATUS_PARAM );
312 int64_t floatx80_to_int64_round_to_zero( floatx80 STATUS_PARAM );
313 float32 floatx80_to_float32( floatx80 STATUS_PARAM );
314 float64 floatx80_to_float64( floatx80 STATUS_PARAM );
316 float128 floatx80_to_float128( floatx80 STATUS_PARAM );
319 /*----------------------------------------------------------------------------
320 | Software IEC/IEEE extended double-precision operations.
321 *----------------------------------------------------------------------------*/
322 floatx80 floatx80_round_to_int( floatx80 STATUS_PARAM );
323 floatx80 floatx80_add( floatx80, floatx80 STATUS_PARAM );
324 floatx80 floatx80_sub( floatx80, floatx80 STATUS_PARAM );
325 floatx80 floatx80_mul( floatx80, floatx80 STATUS_PARAM );
326 floatx80 floatx80_div( floatx80, floatx80 STATUS_PARAM );
327 floatx80 floatx80_rem( floatx80, floatx80 STATUS_PARAM );
328 floatx80 floatx80_sqrt( floatx80 STATUS_PARAM );
329 int floatx80_eq( floatx80, floatx80 STATUS_PARAM );
330 int floatx80_le( floatx80, floatx80 STATUS_PARAM );
331 int floatx80_lt( floatx80, floatx80 STATUS_PARAM );
332 int floatx80_eq_signaling( floatx80, floatx80 STATUS_PARAM );
333 int floatx80_le_quiet( floatx80, floatx80 STATUS_PARAM );
334 int floatx80_lt_quiet( floatx80, floatx80 STATUS_PARAM );
335 int floatx80_is_signaling_nan( floatx80 );
337 INLINE floatx80 floatx80_abs(floatx80 a)
343 INLINE floatx80 floatx80_chs(floatx80 a)
353 /*----------------------------------------------------------------------------
354 | Software IEC/IEEE quadruple-precision conversion routines.
355 *----------------------------------------------------------------------------*/
356 int float128_to_int32( float128 STATUS_PARAM );
357 int float128_to_int32_round_to_zero( float128 STATUS_PARAM );
358 int64_t float128_to_int64( float128 STATUS_PARAM );
359 int64_t float128_to_int64_round_to_zero( float128 STATUS_PARAM );
360 float32 float128_to_float32( float128 STATUS_PARAM );
361 float64 float128_to_float64( float128 STATUS_PARAM );
363 floatx80 float128_to_floatx80( float128 STATUS_PARAM );
366 /*----------------------------------------------------------------------------
367 | Software IEC/IEEE quadruple-precision operations.
368 *----------------------------------------------------------------------------*/
369 float128 float128_round_to_int( float128 STATUS_PARAM );
370 float128 float128_add( float128, float128 STATUS_PARAM );
371 float128 float128_sub( float128, float128 STATUS_PARAM );
372 float128 float128_mul( float128, float128 STATUS_PARAM );
373 float128 float128_div( float128, float128 STATUS_PARAM );
374 float128 float128_rem( float128, float128 STATUS_PARAM );
375 float128 float128_sqrt( float128 STATUS_PARAM );
376 int float128_eq( float128, float128 STATUS_PARAM );
377 int float128_le( float128, float128 STATUS_PARAM );
378 int float128_lt( float128, float128 STATUS_PARAM );
379 int float128_eq_signaling( float128, float128 STATUS_PARAM );
380 int float128_le_quiet( float128, float128 STATUS_PARAM );
381 int float128_lt_quiet( float128, float128 STATUS_PARAM );
382 int float128_is_signaling_nan( float128 );
384 INLINE float128 float128_abs(float128 a)
386 a.high &= 0x7fffffffffffffffLL;
390 INLINE float128 float128_chs(float128 a)
392 a.high ^= 0x8000000000000000LL;
398 #else /* CONFIG_SOFTFLOAT */
400 #include "softfloat-native.h"
402 #endif /* !CONFIG_SOFTFLOAT */
404 #endif /* !SOFTFLOAT_H */